(Kit Fine (Auth.), Andrea Bottani, Massimiliano CA

INDIVIDUALS, ESSENCE AND IDENTITY
TOPOl LIBRARY
VOLUME 4
Managing Editor:
Ermanno Bencivenga, University of California, Irvine, U.S.A.
Editorial Board:
Daniel Berthold-Bond, Bard College, AnnantkUe-on-Hudson, U.S.A.
William James Earle, Baruch College, City University of New York, New York, U.S.A.
Ann Ferguson, University of Massachusetts, Amherst, U.S.A.
David Lloyd, Scripps College, Claremont, U.S.A.
Topoi Library is sponsored by the Department of Philosophy and the School of Humanities at
the University of California, Irvine
Scope:
Like the journal TOPOl, the TOPOl Library is based on the assumption that philosophy is a lively,
provocative, delightful activity, which constantly challenges our inherited habits, painstakingly
elaborates on how things could be different, in other stories, in counterfactual situations, in
alternative possible worlds. Whatever its ideology, whether with the intent of uncovering a truer
structure of reality or of shooting our anxiety, of exposing myths or of following them through, the
outcome of philosophical activity is always the destabilizing, unsettling generation of doubts, of
objections, of criticisms.
It follows that this activity is intrinsically a dialogue, that philosophy is first and foremost
philosophical discussion, that it requires bringing out conflicting points of view, paying careful,
sympathetic attention to their structure, and using this dialectic to articulate one's approach, to
make it richer, more thoughtful, more open to variation and play. And it follows that the spirit
which one brings to this activity must be one of tolerance, of always suspecting one's own
blindness and consequently looking with unbiased eye in every comer, without fearing to pass a
(fallible) judgment on what is there but also without failing to show interest and respect.
It is no rhetoric then to say that the TOPOl Library has no affiliation to any philosophical school
or jargon, that its only policy is to publish exciting, original, carefully reasoned works, and that its
main ambition is to generate serious and responsible exchanges among different traditions, to have
disparate intellectual tools encounter and cross-fertilize each other, to contribute not so much to
the notarization of yesterday's syntheses but rather to the blossoming of tomorrow's.
INDIVIDUALS,
ESSENCE AND IDENTITY
Themes of Analytic Metaphysics
edited by
ANDREA BaITANI
Universita di Bergamo, Italy
MASSIMILIANO CARRARA
Universitiz di Padova, Italy
and
PIERDANIELE GIARETIA
Universita di Padova, Italy
SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

A C.I.P. Catalogue record for this book is available from the Library of Congress.
ISBN 978-90-481-5988-8 ISBN 978-94-017-1866-0 (eBook)

DOI 10.100/978-94-017-1866-0
Printed on acid-free paper
Ali Rights Reserved

© 2002 Springer Science+Business Media Dordrecht
Originally published by Kluwer Academic Publishers in 2002
No part of the material protected by this copyright notice may be reproduced or
utilized in any form or by any means, electronic or mechanical,
including photocopying, recording or by any information storage and
retrieval system, without written permission from the copyright owner
Table of Contents
Introduction vii
Part One: Ontology and Analysis
KIT FINE / The Question of Realism 3
ACHILLE C. VARZI / Words and Objects 49
Part Two: Essence and Existence
ENRICO BERTI / Being and Essence in Contemporary
Interpretations of Aristotle 79
DAVID CHARLES / Some Comments on Prof. Enrico Berti's
Paper "Being and Essence in Contemporary Interpretations of
Aristotle" 109
ALEX ORENSTEIN / Existence, Identity, and an Aristotelian
Tradition 127
MAURO MARIANI/Orenstein on Existence and Identity 151
STEPHEN YABLO / Abstract Objects: A Case Study 163
E. JONATHAN LOWE / Kinds, Essence,
and Natural Necessity 189
KATHE TRETTIN / Kinds of Necessity: a Commentary
on EJ. Lowe's Paper 207
Part Three: Identity

MARIO MIGNUCCI / On the Notion of Identity in
Aristotle 217
vi Individuals, Essence, and Identity
PAOLO CRIVELLI / Sameness in Aristotle's Topics 239

DAVID WIGGINS / Identity and Supervenience 247
EDMUND RUNGGALDIER / Comments on Wiggins's Paper
"Identity and Supervienence" 267
TIMOTHY WILLIAMSON / Vagueness, Identity,
and Leibniz's Law 273
DOROTHY EDGINGTON / Williamson on Vagueness,
Identity, and Leibniz's Law 305
GRAEME FORBES / Origins and Identities 319
PENELOPE MACKIE / Forbes on Origins and Identities 341
Part Four: Time and Persistence

ANTHONY SAVILE / Leibniz, Composite Substances and the
Persistence of Organic Things 355
RICHARD GLAUSER / On Naturalising Leibniz (a Reply
to Anthony Savile) 369
PETER V AN INW AGEN / Temporal Parts and Identity Across
Time 387
ANDREA BOTTANI Van Inwagen on Temporal Parts
and Identity Across Time 413
UWE MEIXNER / Change and Change-Ersatz 427
CHRISTOPHER HUGHES / Starting Over 451
List of Contributors 477
Bibliography 479
Index of Names 497

Introduction
Andrea Bottani
Massimiliano Carrara
Pierdaniele Giaretta
What do we do when we do metaphysics? The aim of this introduction

is to give a provisional answer to this question, and then to explain the
subtitle of the volume. It is easy to observe that when we do meta-
physics we engage in a linguistic activity, mainly consisting of uttering
declarative sentences that are not very clear to most people. That is
true, but, of course, it is not very informative. What do we speak of
when we do metaphysics? A traditional answer could be: we speak of
what things really are, so suggesting that things can appear in a way
that is different from the way they really are. So understood, meta-
physics is about the sense, or the senses, of "real being". A question
that immediately arises is whether the sense of being is unique or is
different for different types of things. Another question is whether it is
possible that something could appear to be, but really not be.
Modem analytic metaphysicians usually answer that the sense of
being is unique, while acknowledging that there are different kinds of
things, and that to say that something could appear to be but really not
be is a plain contradiction, unless what is understood is that it could
appear to us that there is something having such and such features, but
viii Individuals, Essence, and Identity
really there is no such a thing. Everything exists in a unique general

sense of existence.
From this point of view most analytic metaphysicians are com-
mitted to theses which are not shared by ancient and other modem
metaphysicians. That is true of most, but not of all of them. In fact, for
example, Ryle does not agree with the thesis that the sense of being is
unique. This suggests that it is possible, and also preferable, to attach a
different meaning to the adjective "analytic". It could be vaguely char-
acterised as alluding to the adoption of a style of doing philosophy
mainly based on close conceptual analysis, attention to the ways in
which we ordinarily speak or to the ways in which thought may be best
expressed, in each case by working out rigorous explicit arguments.
From this different point of view it would be misleading to present
analytic metaphysics as the kind of metaphysics that most contempo-
rary analytic philosophers do, first because only a few of them engage
in metaphysics at all, and, second, because also some important past
metaphysicians adopted an analytic style. Aristotle, St. Thomas and
Leibniz are surely among them, even if their work was widely ne-
glected or criticised by the main founders of the analytic tradition, i.e.
Frege, Russell, Wittgenstein. Why did this happen? More interestingly,
could Aristotle, st. Thomas and Leibniz have been more fully taken
into account and appreciated by Frege, Russell and Wittgenstein? A
negative answer is quite plausible and does not depend only on the
possible dispute about the general question of the sense of being.
Even if it cannot be said that Frege, Russell and Wittgenstein
shared a common way of approaching the philosophical problems, they
converged to create the conditions for what was later called "the lin-
guistic tum", which was not generally favourable to metaphysics as a
discipline. The linguistic tum has been identified with a methodologi-
cal change of view, based on the idea that what is objective can be ex-
pressed through language and the very nature of language is informa-
tive about reality: whatever is objective can be shown or is presup-
posed by language. Typically Wittgenstein grounded the comprehen-
sion of reality on the notion of fact ("the world is totality of facts, not
of things"), i.e. on the very notion that is essential to the understanding
of the sense of a proposition and that could be seen as the projection
onto reality of the logical-linguistic notion of elementary proposition.
Introduction IX
In a relativistic and conventionalist mood, Carnap thought that every

relevant and comprehensible question of ontology was reducible to, or
identifiable with, a question of choice of a language, or reducible to
"internal" questions, where questions are "internal" when they are
raised within a linguistic framework, i.e. an interpreted formal lan-
guage. Internal questions are legitimate because they allow true or false
answers according to the rules constituting the linguistic framework
specified. Such rules specify methods of observation in the case of
empirical sciences such as physics and biology, methods of proof in
the case of the formal sciences such as logic and mathematics. From a
Carnapian point of view, the traditional general ontological or meta-
physical question concerning what kinds of things exist reduces to the
choice of a vocabulary and both the questions of the nature the kinds of
things and of the general principles ruling them are solvable by adopt-
ing the appropriate logical constraints, i.e. the appropriate definitions
or axioms. Even if very differently motivated, Wittgenstein's resort to
the notion of linguistic games ended by producing a similar reduction
of the traditional metaphysical questions.
It has been observed that Aristotle took the statement of ontologi-
cal concepts and their careful distinction as a way of clarifying the
meaning of words belonging to the natural language. As Chisholm no-
ted "Aristotle says that in discussing the categories, he is concerned in
part with our ordinary language. And he says this often enough to pro-
vide encouragement to those contemporary philosophers who believe
that the statements of metaphysicians, to the extent that they are not
completely empty, tells us something about our language"l It can even
be affirmed that in his ontological research Aristotle was guided by
some intuitions concerning the meaning of words in ordinary natural
language. Many of his ontological notions are defined or clarified with
reference to what can be said or not said in such a language. However
Aristotle took language into account only as a starting point and a
point of reference. He never presented the metaphysical questions as
questions concerning the general form a theory of meaning should
take, as for instance Dummett might say. Metaphysics is not taken into
account by Aristotle as far as it is required by the understanding of
language: it is not admitted only inside and for the needs of an analysis
of meaning. Metaphysics, book V is a good example of this Aristote-
x Individuals, Essence, and Identity
lian method. In this book there is an analysis of many philosophical

terms such as "principle", "cause", "nature", "one", "being", "sub-
stance", and so on. For example, the semantic analysis of the term
"being" starts from taking into account its use in common natural lan-
guage, but aims at establishing the base for an independent science of
being.
Aristotle's approach to metaphysics diverges also from some mod-
ern views such as, for example, Quine'S naturalised epistemology,
where metaphysics is placed in a larger context than a theory of
meaning and philosophical accounts are put on a par, and in continuity,
with scientific explanations. Such non-Aristotelian views share the
idea that ontology should not be pursued independently of the analysis
of language or the scientific view of the world or our conceptual
framework of thought.
However, more recently, some analytic philosophers such as
Strawson, Wiggins, Kripke, Putnam and others acknowledged, in very
different ways and degrees, that there are intuitions about what things
are that ground both our comprehension of words and our identifica-
tion of things. Most of them do not go beyond an appeal to ontological
intuitions. Only some of them engage in autonomously developing or
theorising more or less intuitive ontological concepts. They argue for
full-blown ontological theses, sometimes resorting to logical and for-
mal tools. Of course such philosophers cannot be said to belong to the
linguistic turn. Indeed the influence of their work has made some peo-
ple speak of "ontological turn" and it is also interesting to notice that
their work was not considered outside the field of analytic philosophy.
They can still be taken as analytic philosophers because of the atten-
tion to the language, and the application of rigorous, sometimes for-
mal, methodology. We are thinking of, for example, A. Prior, P.T.
Geach, or D.K. Lewis.
Modern analytic metaphysics is different from the traditional ap-
proach as concerns the general view of being or existence and the more
accurate styles and methods of arguing. Moreover, some absolutely
original methodological questions and ontological theses have been put
forward. One is the analysis of what part of a language has ontological
import. This problem was raised by Quine and answered in a famous
way: to find out what kinds of entities a given theory takes as existing
Introduction XI
one has to look at what kinds of entities are quantified over through
expressions like "there is" and "everything". It is the criterion of on-
tological commitment, also expressed as "to be is to be a value of a
bound variable" (Quine [1953], pp.14-5) The criterion is only a test for
detecting what entities one is committed to. Quine's criterion has been
widely accepted, but has generated also discussion concerning the re-
lation between existence as expressed by "there is" and other possible
senses of existence. As we have already said, not everyone has ac-
cepted that the sense of existence is unique and is expressed by the
existential quantifier.
The other principal innovations brought to the fore by modem
analytic research in ontology concern identity. In analytic metaphysics
it is taken into account from three new points of view. First, identity is
seen as an ontological relation presupposed by our basic practice of
identifying entities. This kind of topic was initiated by Strawson and
its ontological implications more fully pursued by Wiggins. The most
original aspect of this topic is given by the connection of identity with
our way of conceptual ising entities. Second, in this context, Wiggins
took identity as a relation that primitively applies also to entities
picked out at different times. So identity, sortally qualified, grounds
both identifying and reidentifying practices. Third, a methodological
question concerning identity was raised by Quine: the clarity with
which identity can be explicated, in short the definiteness of identity
criteria, confers ontological legitimacy on the entities for which the
identity criteria are stated. As we have already said, not all these theses
have been fully accepted, but each has constituted a new and original
topic for metaphysics.
These are some themes of "analytic metaphysics". But they are not
the only ones. Other themes can be mentioned which are as fully de-
bated even if not completely new. One of these is, for example, the
distinction between intrinsic and extrinsic properties.
However, rather than mention the specific topics dealt with by
analytic philosophers, it could be useful to give a glimpse of the way in
which metaphysics is taken and pursued by them. Let us try to formu-
late some ingredients of a provisional picture.
xii Individuals, Essence, and Identity
First of all, we should bear in mind that many analytic philoso-

phers use the term metaphysics as equivalent to "ontology". So, our
first question concerns what they mean by "ontology".
Usually, one way to explain the word "ontology" is to claim that it
is "the science of being as such" or, simply, "the science of being". If
ontology is "the science of being" one could observe that an ontologist
can be interested in giving an account either of the extension or of the
intension of the word "being". In the first sense, i.e. in the sense of an
explanation of the extension of the word "being", the goal of the on-
tologist or metaphysician is to give a list of everything there is without
excluding anything that does exist and without including anything that
does not exist. If we read "ontology" in this way we can claim - fol-
lowing van Inwagen - that it is a name for a study that is "productive
of ontologies". "Ontology" can be understood as the attempt to pro-
duce a complete catalogue of the furniture of the world (i.e. the exten-
sional sense of "being"). In this regard it is not a contingent list of
things that is to be understood, but rather an exposition of the catego-
ries of entities, a discussion of their properties and of the relations
among them.
On the other hand, in the second sense an ontologist or a metaphy-
sician is interested to explain the meaning of "being", and of germane
expressions like "there is" and "exists". An ontologist using
"ontology" in this second sense is interested in investigating the con-
cept or the concepts that these terms express and how being is related
to language and knowledge.
However some ontological research deals neither with an inven-
tory of the world, nor with an inquiry about being. Some ontologists
focus on some relations used to model the world. These relations are
usually inter-categorial and some of them can be taken as primitive.
They are, for example, "to be identical with", "to depend on" "to be
part of', "to be a cause of'.
Almost all the papers collected in the present volume concern on-
tology in the intensional sense, or deal with important ontological re-
lations. A few of them are explicitly related to some categories of enti-
ties. Such a classification could be the result of looking at what they
mainly focus on. From other points of view other classifications are
possible.
Introduction X11l
In order to avoid arbitrariness perhaps it is better to present the pa-

pers of this book by looking at the topics dealt with or not dealt with.
Little or nothing can be found here concerning a number of tradi-
tional metaphysical matters, such as, for example, the nature and exis-
tence of God, the nature and existence of immaterial substances, the
mind-body relation, and the nature of space and time. The volume does
not aim at giving an overall and systematic image, but rather, due
mainly to its collective nature, at weaving a web of philosophical dis-
cussions more or less closely connected to one another, where a few
historical contributions show the roots the metaphysical themes and
methods of today plunge into their Aristotelian and Leibnizian past.
Among the theoretical topics touched on in the volume, there are the
relationships between ontological and linguistic analysis, the questions
of realism and ontological commitment, the nature of abstract objects,
the existential meaning of particular quantification, the primitiveness
of identity, the question of epistemic versus ontological vagueness, the
necessity of origin, the nature of natural necessity, the possibility of
intermittent existence, the notion of a temporal part and its place in an
account of persistence, the question of identity and change across time
and possible worlds, and still others. Among the historical topics
touched on in the volume, there are the Aristotelian treatment of iden-
tity, the Aristotelian notions of being and essence and the Leibnizian
theory of the persistence of organic things.
Though not systematic in character, the book does not lack a
structure, the collected papers being grouped under four headings:
"Ontology and Analysis", "Essence and Existence", "Identity" and
"Time and Persistence". It is one of the aims of the book to show that
there is a very close semantic connection among the concepts of being,
essence, existence, identity and persistence (and indeed among all
these concepts and some other basic metaphysical notions). The first
section is essentially metatheoretical and methodological in character,
relating to the nature, method and meaning of metaphysical discussion.
What do we really do when we commit ourselves to the existence of
certain kinds of entities? And what do we do when we reduce some
kinds of entities to some others? What relations hold between onto-
logical analysis and linguistic analysis of logical form? Can the former
be decided by the latter or rather are they quite independent matters?
xiv Individuals, Essence, and Identity
All these questions directly relate to the notion of entity, which, since
Quine, has been naturally connected to (indeed, elucidated in terms of)
that of identity. So, the second section turns to the concept of identity.
Can identity be reduced to something else (indiscernibility, community
of parts or some other relation)? Can '=' be eliminated from our lan-
guage with no semantic loss? And what kind of relation is identity?
What connections are there - if any - between identity and sortal
predication? Can identity itself be vague? Can any objects be neither
identical nor different? The third section enlarges the field, bringing to
the fore modal themes and issues. What is essence? What is necessity?
What is a natural law? What do natural laws have to do with natural
kinds? Since the relation between essence and existence is one of the
central themes of ancient metaphysics, the questions of the nature of
existence and of the meaning of the existential quantifier find a place
in this section. In the last section, all the topics and issues discussed in
the previous sections find an application - though surely a very central
one - to the problem of identity, change and persistence across time.
The theme is taken up over a wide range of questions, concerning the
existence of temporal parts of ordinary objects like cats or trees, the
possibility of a genuine identity across time, the compatibility between
ordinary change and Leibniz's Law and some lessdiscussed issues such
as the possibility of intermittent existence.
According to some metaphysicians, the identity of a thing is given
by its origin. The origin of this book was a conference with the same
title, held in Bergamo (Italy) on June 22-24, 2000. Though the majority
of the papers published in the volume derive from the talks given at the
conference - and conversely the majority of the talks given at the con-
ference have become papers in this volume - this is not a volume of
proceedings, because a few of the contributors to the volume did not
take part in the conference.
We are indebted to many people for different reasons. First of all
we thank the authors who contributed to the conference or to the vol-
ume. We think to have been very lucky to gather so many distin-
guished philosophers together. Secondly, thanks to Alberto Castoldi,
rector of the University of Bergamo, for the firm support given to the
conference from which the book originated. Thanks also to Vittorio
Morato, Marzia Soavi, and Elisabetta Bonadeo: they were very helpful
Introduction xv
during the conference. Vittorio Morato did also almost all the work for
the camera-ready copy of the book. Carlo Nizzo checked most of the
editing work. Richard Davies made some corrections and gave some
advice about the English of some papers. Many thanks to them.
Thanks to the anonimous reader for the Kluwer Publisher for a wide
range of useful comments. And last, but not least, thanks to the series
editor of "Topoi Library" Ermanno Bencivenga for having supported
this long-standing enterprise.
The preparation of this book and the related conference have been
made possible by grants from the Department of Education and Com-
munication Sciences of the University of Bergamo and from a national
research project on "Knowledge and Cognition", coordinated by Paolo
Parrini (University of Firenze) - more specifically, from the local proj-
ects, on "Language, Thought, and Normativity", coordinated by Diego
Marconi (University of V ercelli) and on "Reference and Thought" co-
ordinated by Pierdaniele Giaretta (University of Padova). The projects
were sponsored by the Italian Ministry for University and Scientific
Research (MURST).
Notes
1 Chisholm [1996], p.8.
Part One
Ontology and Analysis

THE QUESTION OF REALISM
Kit Fine
My aim in this paper is to help lay the conceptual and methodological

foundations for the study of realism. I come to two main conclusions:
first, that there is a primitive metaphysical concept of reality, one that
cannot be understood in fundamentally different terms; and second,
that questions of what is real are to be settled upon the basis of consid-
erations of ground. The two conclusions are somewhat in tension with
one another, for the lack of a definition of the concept of reality would
appear to stand in the way of developing a sound methodology for de-
termining its application; and one of my main concerns has been to
show how the tension between the two might be resolved.
The paper is in two main parts. In the first, I point to the difficul-
ties in making out a metaphysical conception of reality. I begin by dis-
tinguishing this conception from the ordinary conception of reality
(§l) and then show how the two leading contenders for the metaphysi-
cal conception - the factual and the irreducible - both appear to resist
formulation in other terms. This leads to the quietist challenge, that
questions of realism are either meaningless or pointless (§4); and the
second part of the paper (§§5-10) is largely devoted to showing how
this challenge might be met. I begin by introducing the notion of
ground (§5) and then show how it can be used as a basis for resolving
questions both of factuality (§§6-7) and of irreducibility (§§8-9). I
3
A. Bottani et al. (eels. J. Individuals. Essence and Identity. 3-48.
© 2002 Kluwer Academic Publishers.
4 Individuals, Essence, and Identity
conclude with some remarks on the essential unity of these two ques-
tions and of the means by which they are to be answered (§1O).
1. REALITY
Among the most important issues in philosophy are those concerning

the reality of this or that feature of the world. Are there numbers or
other abstract objects? Is everything mental or everything physical?
Are there moral facts? It is through attempting to resolve such ques-
tions that philosophy holds out the promise of presenting us with a
world-view, a picture of how the world is and of our place within it.
However, as is so often true in philosophy, the difficulties begin
with the formulation of the question rather than with the attempt at an
answer. The antirealist about numbers maintains:
There are no numbers.
But most of us, in our non-philosophical moments, are inclined to
think:
There are prime numbers between 2 and 6.
And yet the second of these claims implies that there are numbers,
which is incompatible with the first of the claims. Similarly, the
antirealist about morality maintains:
There are no moral facts.
But he also thinks:
Killing babies for fun is wrong.
And yet the second claim implies that it is a fact that killing babies for
fun is wrong and, since this is a moral fact, its existence is incompati-
ble with the first claim.
How, in the light of such possible conflicts, should the realist and
antirealist claims be construed? Should we take the conflict between
antirealism and received non-philosophical opinion to be a genuine
Kit Fine, The Question of Realism 5
conflict or not? And if not, then how is the apparent conflict between
them to be dispelled?
If we take the conflict to be genuine, we obtain what has been
called an "eliminative" or "skeptical" conception of antirealism. The
antirealist will be taken to dispute what we ordinarily accept, the real-
ist to endorse it. Thus the antirealist about numbers will be taken to
deny, or to doubt, that there are prime numbers between 2 and 6; and
likewise, the moral antirealist will be taken to deny, or to doubt, that
killing babies for fun is wrong.
Of course, the mere rejection of what we ordinarily accept is per-
verse and so presumably the interest of antirealism, on this conception,
must derive from the assumption that philosophy is able to provide us
with some special reasons for dOUbting what we ordinarily accept.
Thus the antirealist may attempt to convince us that we have no good
reason to believe in a realm of abstract objects with which we can have
no causal contact or that, in moral matters, we can have no justification
for going beyond the mere expression of approval or disapproval. Our
world-view will therefore be the product of dealing with these doubts,
either by laying them to rest or by retreating into skepticism.
Anti-realism, as so understood, has a long and illustrious history;
and certainly its interest is not to be denied. However, in this age of
post-Moorean modesty, many of us are inclined to doubt that philoso-
phy is in possession of arguments that might genuinely serve to un-
dermine what we ordinarily believe. It may perhaps be conceded that
the arguments of the skeptic appear to be utterly compelling; but the
Mooreans among us will hold that the very plausibility of our ordinary
beliefs is reason enough for supposing that there must be something
wrong in the skeptic's arguments, even if we are unable to say what it
is. In so far, then, as the pretensions of philosophy to provide a world-
view rest upon its claim to be in possession of the epistemological high
ground, those pretensions had better be given up.
Is there room for another form of antirealism - and another ac-
count of philosophy's pretensions - that does not put them in conflict
with received opinion? If there is, then it requires that we be able con-
sistently to affirm that something is the case and yet deny that it is
really the case. I It requires, in other words, a metaphysical conception
of reality, one that enables us to distinguish, within the sphere of what
is the case, between what is really the case and what is only apparently
the case.
But what might this metaphysical conception of reality be? Two
main answers to this question have been proposed. According to the
first, metaphysical reality is to be identified with what is "objective" or
"factual". The antirealist, on this conception, denies that there are any
facts "out there" in virtue of which the propositions of a given domain
might be true. The propositions of the domain are not in the "business"
of stating such facts; they serve merely to indicate our engagement
with the world without stating, in objective fashion, how the world is.
As familiar examples of such a position, we have expressivism in eth-
ics, according to which ethical judgements are mere expressions of at-
titude; formalism in mathematics, according to which mathematical
statements are mere moves within a system of formal rules; and in-
strumentalism in science, according to which scientific theories are
mere devices for the prediction and control of our environment. Ac-
cording to the second conception, metaphysical reality is to be identi-
fied with what is "irreducible" or "fundamental". On this view, reality
is constituted by certain irreducible or fundamental facts; and in de-
nying reality to a given domain, the antirealist is claiming that its facts
are all reducible to facts of some other sort. Thus the ethical naturalist
will claim that every ethical fact is reducible to naturalistic facts, the
logicist that every mathematical fact is reducible to facts of logic, and
the phenomenalist that every fact about the external world is reducible
to facts about our sense-data.
We might see the anti factualist and reductionist as indicating two
different ways in which a proposition may fail to "correspond" to the
facts. For it may fail even to point in the direction of the facts, as it
were; or it may fail to indicate, at the most fundamental level, how the
facts are. In the one case, the propositions of a given domain will not
even represent the facts, while in the other, the propositions will not
perspicuously represent the facts - there will be some divergence be-
tween how the facts are "in themselves" and how they are represented
as being. If either of these metaphysical conceptions of reality is vi-
able, then it would appear to provide a way of upholding a non-
skeptical form of antirealism. For it will be perfectly compatible with
affirming any given proposition to deny that it is genuinely factual or
genuinely fundamental. The expressivist, for example, may affirm that

killing babies for fun is wrong and yet deny that, in so affirming, he is
making a factual claim; and the logicist may affirm that 5+7 = 12 and
yet deny that he is thereby stating something fundamental. Truth is one
thing, metaphysical status another.
But the problem now is not to defend the antirealist position but to
see how it could even be intelligible. Consider the antifactualist in
ethics. Since he is assumed to be non-skeptical, he will presumably be
willing to affirm that killing babies for fun is wrong. But then should
he not be prepared to admit that he is thereby making a claim about
how things are? And is not this a claim about how things are in the
world - the only world that we know, that includes all that is the case
and excludes whatever is not the case? So is he not then committed to
the proposition's being factual?
Of course, the antirealist will insist that he has been misunder-
stood. He will maintain that the proposition that killing babies for fun
is wrong does not make a claim about the real world as he conceives it
and that, even though it may be correct to affirm that killing babies for
fun is wrong, there still is no fact "out there" in the real world to which
it is answerable. But the difficulty then is in understanding the in-
tended contrast between his world - the real world "out there" - and
the world of common mundane fact. For what room is there, in our or-
dinary conception of reality, for any further distinction between what
is genuinely a fact and merely the semblance of a fact?
Similarly, the reductionist in ethics will claim that ethical facts are
reducible to facts of another sort and, on this ground, deny that they
are real. Now it may be conceded that there is a sense in which certain
facts are more fundamental than others; they may serve to explain the
other facts or perhaps, in some other way, be constitutive of them. But
how does this provide a ground for denying reality to the other facts?
Indeed, that they had an explanation or constitution in terms of the real
facts would appear to indicate that they themselves were real.
What then is this conception of reduction for which the reducible
will not be real?2 Just as there was a difficulty in understanding a
metaphysical conception of the facts, one that might serve to sustain a
metaphysical form of antirealism, so there is a difficulty in under-
standing a metaphysical conception of reduction. In either case, we
appear to avoid the absurdities of skepticism but only by buying in to

the obscurities of metaphysics, One kind of problematic high ground
has simply been exchanged for another,
2. FACTUALITY
Is there any way out of the previous difficulties? Can an intelligible

form of antifactualism or reductionism be sustained? Let us discuss
each question in tum.
In the case of anti factualism, it has commonly been supposed that
there is some feature of nondeclarative pronouncements - such as
'Ouch!' or 'Get out of here! , - that obviously renders them nonfactual
and is also possessed, though not so obviously, by the declarative
propositions of a given domain. 3 Thus despite these propositions being
declarative in form, they are to be classified with the nondeclarative
pronouncements as nonfactual. So for example, on the traditional ac-
count of this sort, a "noncognitive" or "nonfactual" proposition is
taken to be one that is not a candidate for being true or false, and the
antirealist is taken to deny that the propositions from a given disputed
class are candidates for being true or false.
But the problem with this approach is that what is regarded as a
non-obvious feature of the disputed propositions is in fact a feature
that they obviously lack. Thus, given that killing babies for fun is
wrong, it follows - in the ordinary, straightforward sense of 'true' -
that it is true that killing babies for fun is wrong; and so the proposi-
tion that killing babies for fun is wrong is a candidate for being true or
false after all. The traditional noncognitivist must therefore either be
using the terms 'true' and 'false' in some special metaphysically in-
flated sense that still needs to be explained, or he should reckon him-
self a complete skeptic who is unwilling to affirm any proposition
whatever from the disputed domain.
Nor does it help to appeal to other obvious factual characteristics
of propositions in place of candidacy for truth or falsehood. One might
suggest that a factual proposition is one capable of being believed or
asserted, or of figuring in inferences, or of being embedded in larger
linguistic contexts. But the same point applies. For in the ordinary
sense of 'believe', 'assert', etc., we do have moral beliefs and make
moral assertions, we do draw moral conclusions, and we do embed
moral propositions in larger linguistic contexts; and similarly for the
propositions of mathematics or of science or of other disputed areas.
Indeed, once given one of these characteristics, the rest seem to follow
- their possession is, for the most part, a "package deal". There there-
fore seems to be no reasonable hope of identifying a non-skeptical
form of factuality in terms of the possession of some of these charac-
teristics as opposed to the others.
Any reasonable, non-skeptical view should therefore grant that
propositions of the kind that figure in realist disputes will possess all
of the obvious trappings of factuality: they will be capable of being
true or false, believed or asserted, embedded in larger linguistic con-
texts, and so on. The antifactualist should therefore be a quasirealist
and attribute to the nonfactual all those features that were traditionally
thought to belong to the factual. But if the nonfactual is not to be dis-
tinguished from the factual in terms of the obvious trappings of factu-
ality, then how is it to be distinguished? What, in a word, is the differ-
ence between quasi realism and genuine realism?
Various more sophisticated criteria that appear to avoid these dif-
ficulties have been proposed. 4 I shall focus on one, Dummett's, ac-
cording to which realism for a given area of discourse is primarily a
matter of its conforming to the Principle of Bivalence, the principle
that every statement of the discourse should be either true or false. s I
shall argue that the criterion, even when supplemented, is unsatisfac-
tory, and then attempt to draw some broader conclusions.
We should note right away that the proposal, even if otherwise ac-
ceptable, does not answer to our needs. For we were after a non-
skeptical form of antirealism, one that was not at odds with received
opinion. But in regard to many areas of discourse, the received opinion
is that the statements are subject to Bivalence; and so any form of
Dummettian antirealism must to that extent be skeptical. Thus in so far
as philosophers have wished to espouse a completely non-skeptical
form of antirealism, the Dummettian criterion must be considered un-
satisfactory.
Another problem concerns the application of the criterion to par-

ticular statements. For surely we wish to be able to affirm or deny that
a particular statement - such as '7+5 = 12' - has realist import, that it
is or is not answerable to an "external" reality. But how is the criterion
to be applied in such a case? Presumably by associating the statement
with a particular area of discourse. But which one? The answer may
well depend upon what we say. In the case of '7+5 = 12', for example,
the area of discourse could be the language of equational arithmetic
(without variables or quantifiers), the first-order language for addition,
or the first-order language for addition and multiplication. But the fi-
nitist may well accept Bivalence for the first language though not the
others, while the constructivist may well accept Bivalence for the first
two languages though not the third. Thus no stable answer is assured.
(The problem here is analogous to the problem of determining a refer-
ence class for "single case" probabilities, and it besets several other
criteria for realism as well.)
Even in application to areas of discourse, Bivalence is not, on its
own, sufficient for realism. A simple counter-example runs as fol-
lows. 6 Suppose that an antirealist becomes completely opinionated
about the given discourse: he acquires a view (for reasons internal to
the discourse, though perhaps very bad ones) on the true/false answer
to every particular question that might arise. He would then be com-
mitted to each instance of Bivalence, and as long as he was aware of
having become completely opinionated, he would also be committed
to Bivalence holding of every statement of the discourse. But it seems
absurd to suppose that, on that fateful day in which the last question
falls under the sway of his opinion, he is destined to become a realist.
How can his being an antirealist prevent him from forming an opinion
on the matter?
Clearly, acceptance of Bivalence for a given area of discourse is
not enough to guarantee realism. It has to be acceptance for the right
reason. But is there any way to supplement Bivalence so as to ensure
that the acceptance will be for the right reason? Two proposals for
supplementing Bivalence have been considered (often in combination)
- one epistemic and the other semantic.
According to the epistemic proposal, the notion of truth that fig-
ures in Bivalence must be such that it is possible for a statement from
the given discourse to be true and yet unknowable - or even lacking in

any possible evidence in favor of its truth. Indeed, if a given statement
were true and yet unknowable, then that would appear to provide a
strong reason for taking the statement to have realist import quite apart
from any discourse in which it might be placed.
But I believe that even this plausible proposal is subject to coun-
terexample, though of a more sophisticated sort. Consider a semantics
in which the meaning of a sentence is given by the evidential situations
in which its assertion is warranted. The guiding principle of the se-
mantics is that one is warranted in asserting a sentence in a given evi-
dential situation iff the possibility of its vindication is never fore-
closed, i.e., iff for any improvement in one's evidential situations there
is a further improvement in which its assertion would be warranted. 7 It
should then be clear that the Law of Excluded Middle, and hence the
Principle of Bivalence, will be valid, i.e., be warranted in any eviden-
tial situation. s For take any improvement in that situation. Then either
it warrants -.A and hence has an improvement (viz. itself) that warrants
(A v -.A), 01: it has an improvement which warrants A and hence war-
rants (A v -.A).
Clearly, such a semantics might be adopted by an antirealist. And
so it remains to show how such an antirealist might be justified in
maintaining that there is no possible evidence for or against the truth
of a given statement A. It is not clear how this might be, for any given
evidential situation will either warrant -,A or will permit an improve-
ment that warrants A. Our antirealist, however, may be working with
an "objective" notion of warrant. It may be an objective matter - one
to which he does not necessarily have epistemic access - what the pos-
sible evidential situations are and hence what they warrant. The state-
ments in question might concern the external world, for example, and
the evidential situations might be given by the courses of experience
which someone might actually undergo. Our antirealist might then ar-
gue that even though I am objectively warranted in believing that
McCavity was not here, at the scene of the crime, since no counter-
evidence would ever present itself, still I can have no evidence for his
having not been here, since I can have no basis for excluding the pos-
sibility of counter-evidence. Of course, our antirealist is to some extent
a realist - for he is a realist about the objective possibilities of experi-
ence, but he is not a realist about the external world, which is what is
here in question.
The second strategy is to provide a semantic supplement to Biva-
lence: not only must Bivalence hold, but it must be a semantical matter
that it holds. But our previous counter-example still stands, since the
validity of Bivalence is a semantical matter under the antirealist se-
mantics that I proposed (given that there is a semantical underpinning
for applying the Law of the Excluded Middle to statements of war-
rant). It therefore appears necessary to go deeper into the mechanism
by which the language is to be interpreted. But what might that be?
One suggestion is that we require "acceptance of classical two-valued
semantics [... ] in its entirety" (Dummett [1993], p.468). This would
require not only Bivalence but the exclusion of empty terms, the stan-
dard clauses for the connectives, and so on. But the problem now is
that, given that our antirealist is willing to accept Bivalence, it is not
clear why he should be unwilling to accept the rest of the classical se-
mantics - though, of course, under his own understanding of what this
comes to. Another suggestion is that we require that our understanding
of the language should be truth-conditional, that our grasp of the
meaning of a statement should consist in knowledge of its truth-
conditions. But although philosophers use this phraseology as if they
knew what it meant, it is not at all clear that it can be explained in such
a way as to both imply realism and yet not presuppose that the truth-
conditions are already to be understood in realist fashion. One can, of
course, insist that the relevant notion of truth should conform to Biva-
lence or be evidence-transcendent. But this then leads us back to our
previous difficulties.
What, I believe, has made these various criteria so appealing is
that it is often hard to see how one could plausibly maintain that a
given criterion is satisfied (or not satisfied) and yet still be a realist (or
antirealist). Realism about mathematics, for example, is a reason - and
perhaps the only good reason - for holding that every mathematical
statement is true or false or that there might be mathematical truths
that are beyond our ken. But it should be recognized that, even though
the existence of an external reality may make it plausible that our lin-
guistic and epistemic contact with that reality is of a certain sort, this is
not in what the externality of the reality consists. In thinking about
these matters, we need to restore ourselves to a state of innocence in

which the metaphysical claims are seen to be about the subject-matter
in question - be it mathematics or morality or science - and not about
our relationship to that subject-matter.
Indeed, a broader conclusion may be justified. For we have seen
that, even as we piled on the conditions, we were unable to find a suf-
ficient condition for factualism (nor, I might add, for nonfactualism,
not that this is a case that I have considered). This therefore suggests
that there can be no sufficient condition at all for being factual (or non-
factual) - unless, of course, for the trivial reason that the condition
cannot be satisfied or for the question-begging reason that a problem-
atic conception of factuality has already been presupposed.
If this is right, then it means that it will not even be possible to
provide an adequate formulation of any particular factualist or anti-
factualist position, i.e., one that will imply that the position is indeed
committed to factualism or to antifactualism for the domain in ques-
tion; and examination of the actual formulations of such positions
bears this out. Consider expressivism, for example. The expressivist
wishes to maintain that moral affirmations are expressive in much the
same way as expressions such as 'Ouch'. But, of course, the mere
claim that moral affirmations are expressive does not serve to distin-
guish his position from that of the moral realist, since even he may be
willing to maintain that moral affirmations are used to express our
moral attitudes as well as to report the moral facts. The expressivist
must therefore be claiming that moral affirmations are merely expres-
sive, that they have no other feature that serves to make them factual.
Now in the case of 'Ouch' we can see why this should be so, since
'Ouch' is not used to say anything true or false. But it is on this exact
point that moral affirmations differ from expressions such as 'Ouch'.
Thus it remains completely opaque what exactly it is in the expressiv-
ist's position that obliges him to embrace the nonfactuality of moral
discourse; and similarly for the other particular forms of antifactualism
that have been proposed.
3. REDUCIBILITY
We tum to the second of the two metaphysical conceptions of reality,
the conception of reality as irreducible. This conception can be no
better off than the concept of reduction with which it is associated; and
so we may ask, "What is it for one proposition (or statement or sen-
tence) to be reducible to others?"
Three main lines of response to this question have been proposed.
According to the first, reduction is a matter of logical analysis. To say
that one sentence reduces to, or is analyzable in terms of, another is to
say that they express the same proposition but that the grammatical
form of the second is closer to the logical form of the proposition than
the grammatical form of the first. 9 Thus reduction reveals a discrep-
ancy between the "apparent" grammatical form of the sentence and the
"genuine" logical form of the proposition and serves to bring the two
in closer alignment. To take a paradigm example, the sentence 'The
average American is 5 feet tall' will reduce to 'The sum of heights of
all Americans divided by the number of Americans is 5 feet', since the
latter brings us closer to the logical form of the proposition that is ex-
pressed.
This approach suffers from at least two problems of detail. First, it
is unable to deal with one-many reductions. The philosopher who does
not believe in conjunctive facts will want to say that the truth of the
conjunction S /\ T reduces to the truth of its conjuncts S and T. But
here there is no question of a single proposition being expressed on left
and right. It is also unable to deal with one-one reductions in which the
reducing sentence is merely sufficient for the sentence to be reduced.
For example, when S is true and T false, we may wish to say that the
truth of the disjunction S v T reduces to the truth of the disjunct S. But
again, no single proposition is expressed.
A second difficulty concerns de re reductions. Just as there are de
re modal claims that are to be distinguished from their de dicto coun-
terparts, so there are de re reduction claims that are to be distinguished
from the corresponding de dicto claims. Thus we may wish to claim
not merely that the sentence 'The couple Jack and Jill is married' is
reducible to the sentence 'Jack is married to Jill', but also that the sat-
isfaction of the open sentence 'z is married' by the couple Jack and Jill
is reducible to the satisfaction of the open sentence 'x is married to y'
by Jack and Jill. But it is hard to see, on the proposed view, how this
could be a case of logical analysis. For the proposition expressed by
the open sentence 'z is married' under the assignment of the couple to
z is presumably the singular proposition that that couple is married and

the proposition expressed by the open sentence 'x is married to y' un-
der the assignment of Jack to x and Jill to y is presumably the singular
proposition that Jack is married to Jill; and yet these two propositions
should be taken to be distinct since, without a fine-grained notion of
propositional identity, we will be at a loss to explain how one gram-
matical form can be closer to the genuine logical form than another. 10
It might be thought odd that we express a reduction of couples to
their members by making reference to couples, since is not the point of
the reduction to show that couples are a "logical fiction" and hence not
really existent? But this line of thought represents a confusion between
the skeptical and non-skeptical forms of antirealism. Their being a
logical fiction, in the relevant sense, does not prevent us from making
non-philosophical claims about couples, such as that all the couples in
the room are married; and no more should it prevent us from making
philosophical claims about couples of the sort typified by reductions.
It is merely that consistency demands that these claims themselves
should, at some point, be reduced. 11
The most serious difficulty with the present approach is that it
rests upon a problematic conception oflogical form. To maintain that a
sentence concerning nations, say, expresses the same proposition, or
states the same fact, as one concerning individuals and yet is less close
in its form to that proposition or fact is already to buy into a meta-
physically loaded conception of logical form. There is nothing beyond
a metaphysical basis for making such a claim. One might attempt, of
course, to provide an account of logical form in metaphysically neutral
terms - perhaps in terms of what is required for a satisfactory explana-
tion of 'truth-conditions' or valid inference. But in so far as the ac-
count is successful in this respect, its metaphysical significance will be
unclear. Why should the most satisfactory account of valid inference
or of truth-conditions have any implications for how the world really is
unless this is something that is already required of such an explana-
tion?
According to the second approach, reduction is a semantical mat-
ter. It is taken to be a relation that holds in virtue of the meaning of the
sentences to which it applies; and what is most distinctive about this
relation is that, given that one sentence reduces to others, it should be
possible to acquire an understanding of the reduced sentence on the

basis of an understanding of the sentences to which it reduces.
Such an approach avoids the previous difficulties over one-many
and one-one reductions, since there is nothing, in general, to prevent a
sentence from simultaneously reducing to several sentences that jointly
provide a sufficient condition - though perhaps not a necessary condi-
tion - for the truth of the given sentence. 12 But it still flounders over
the problem of de re reductions. For where c is the couple Jack and
Jill, a is Jack and b is Jill, there is no semantic connection between c's
being a married couple and a's being married to b,l3 since there is
nothing semantic that might serve to indicate that a and b are the indi-
viduals that compose c. There is, of course, a semantical implication
from a's being married to b to a-and-b's being a married couple. But
this is not what we are after, since we want direct reference to the cou-
ple c on the right and not indirect reference, via the components a and
b. Or again, it may be an analytic truth that for any couple c there are
individuals a and b that compose c and are such that e's being married
reduces to a's being married to b. But this is a general claim and still
leaves unexplained what it is for the particular reductions to hold. We
might also note that the general formulation of the reduction in such
cases is not always an analytic truth. For we may want to say that, for
any quantity q of water, there are H20 molecules m l , m2, .•• such that
the existence of q at a given time reduces to the existence of m l , m2, .••
at that time; and this generalization is not a priori and hence presuma-
bly not analytic. Thus it is not even as if particular reductions can al-
ways be given a semantic backing.
Nor is it even clear that the existence of an analysis provides a
sufficient condition for reducibility in the metaphysical sense. For let
us grant that what it is for someone to be a bachelor is for him to be
unmarried and to be a man. Are we then obliged to say that the fact
that someone is a bachelor reduces to the fact that he is unmarried and
the fact that he is man and that there is therefore no real fact of his be-
ing a bachelor? Perhaps not. For we might believe that there really are
complex attributes in the world and that their attribution cannot, for
this reason, be reduced to the attribution of the simpler attributes of
which they are composed. Thus even though we may explain their
identity in terms of the simpler attributes, we do not reductively ac-

count for their attribution in those terms.
The lesson to be learnt from the foregoing criticisms, I believe, is
that reduction should be construed as a metaphysical rather than as a
linguistic or a semantical relation. In making claims of reduction, we
wish to talk, not about our representation of the facts, but about the
facts themselves. Thus in claiming that two nations' being at war re-
duces to such-and-such military activity on the part of their citizens,
we are not making a claim about our language for describing nations
and citizens, or even about our concepts of a nation or a citizen, but
about the nations and citizens themselves and the connection between
them. Again, we need to restore ourselves to a state of metaphysical
innocence in which reduction is seen to concern the subject-matter it-
self and not the means by which it might be represented or cognized.
According to the third, more recent approach, reduction is a modal
matter. One class of propositions will reduce to - or supervene upon -
another if, necessarily, any truth from the one is entailed by truths
from the other.14 This approach avoids the previous difficulties over
the possibility of de re reductions, since the propositions may them-
selves be de re; but it suffers from difficulties of its own. For one
thing, it faces the earlier problem of the "reference class", for whether
one proposition is reducible to others will depend upon the classes of
proposition with which they are associated. It will also not be properly
applicable to necessary domains, such as mathematics, since it is al-
ways a trivial matter that a necessary truth is entailed by any proposi-
tions whatever.
But even if we limit its application to contingent domains, there
are two other serious shortcomings to the approach. In the first place, it
is not able to capture the idea that the truth of a proposition must re-
duce to something more basic. Velocity at an instant, for example, su-
pervenes on velocity over an interval and vice versa, and yet we can-
not say, without circularity, that each reduces to the other. Nor does it
help to insist that the supervenience be one-way. For suppose that
there are three parameters and that the value of anyone parameter su-
pervenes on the values of the two others but not on the value of one of
them alone. Then the value of each parameter will one-way supervene
on the values of the others, and yet we cannot, without circularity, say
that the value of each parameter is reducible to the value of the others.
As a particular example, we might take the parameters to be the mass,
volume and density of a given body.15
Finally, the approach is no better able than the first to capture the
antirealist import of reductive claims. For as long as reduction is re-
garded as getting us closer to what is real, we will wish to deny the re-
ality of any fact that reduces to something else. But then how could the
mere existence of certain modal connections between one class of
propositions and another serve to establish the unreality (or the reality)
of the facts from either class?
Indeed, a broader conclusion may be justified, just as in the case
of factuality. For it is hard to see how there could be any sufficient
condition for one proposition to be reducible to others (trivial and cir-
cular cases aside). For whatever the sufficient condition might be, its
satisfaction would appear to be compatible with the adoption of a
thorough-going realist position, one which took every single fact to be
real, and hence compatible with the rejection of any given reductive
claim. 16
4. THE QUIETIST CHALLENGE
We see from the previous discussion that the prospects for defining the
notions of factuality and reducibility in fundamentally different terms,
or even for providing conceptually unproblematic sufficient conditions
for their application, do not look good. Although we arrived at this
negative conclusion as a result of detailed investigation, the conclusion
could perhaps have been anticipated from the start. For we were after a
form of antirealism that was not necessarily skeptical, at odds with re-
ceived opinion. Now, presumably there is nothing special about the
opinion's being received in this regard. In so far as an antirealist posi-
tion is independent of such opinion it should be independent of all
similar opinion, whether received or not. Indeed, central to our present
understanding of antirealism is a distinction between what one might
call "first-order" propositions, which merely say how things are with-
out regard to their metaphysical status, and the corresponding "second-
order" claims, which merely comment on the metaphysical status of

the first-order propositions. What would then appear to guarantee the
possibility of a non-skeptical form of antirealism is the general inde-
pendence of the second-order claims, in this sense, from the first-order
propositions with which they deal. So, as long as the criterion of factu-
ality or reducibility is stated in first-order terms, its inadequacy will
simply follow from this general form of independence. 17 Skepticism
has a very long arm; and if we are altogether to escape its grip, we
must embrace doctrines of nonfactualism or reducibility that are free
from any first-order encumbrance.
These results appear to be deeply disturbing, however. For if fac-
tuality or reducibility are not to be understood in ordinary first-order
terms, then how are they to be understood? How are we to make sense
of the idea that behind every putative fact there mayor may not be
something real in the world to which it corresponds? The results seem
especially disturbing in the case offactuality. For there will be no first-
order difference between factual and nonfactual propositions. The non-
factual propositions will be in the nature of imposters - with all of the
usual trappings of factuality but none of the substance. They will be
like "zombies" that display all of the outwards signs of consciousness
without themselves being conscious. But then how are we to distin-
guish between the two? At least in the case of the zombies, we can
perhaps tell from our own experience what it is like to be conscious.
But in the present case, there would appear to be no special vantage
point from which we could draw a distinction between what is real and
what is merely a "shadow" cast by our language or thought. There is
no stepping behind the putative facts to see what is really there.
It is considerations such as these that have led several present-day
philosophers - the "quietists" - to conclude that the metaphysical no-
tions of factuality and reducibility are devoid of content. IS And, of
course, once these notions go, then so does the metaphysical enterprise
associated with them. Philosophy, on this way of thinking, should
abandon its pretension of presenting us with a higher-order view of
how the world really is. Or rather, if there is a view, it is that there is
no such view to be had.
However, tempting as such a conclusion may be, it is not war-
ranted by the evidence. For the difficulty in defining the notions may
derive, not from their lack of content, but from their distinctive char-
acter. Indeed, it is hard to avoid the impression, once one surveys the
various attempts at definition, that we have, in the conception of real-
ity as objective or fundamental, a distinctively metaphysical idea.
From this point of view, the attempt to define these notions in other
terms would be akin to the naturalistic fallacy; and just as it would be a
mistake to infer the unintelligibility of normative notions from the dif-
ficulty of defining them in naturalistic terms, so it would be a mistake,
in the present case, to infer the unintelligibility of the notions of factu-
ality and reducibility from the difficulty of defining them in non-
metaphysical terms.
Of course, the quietist may have a general hostility to metaphysi-
cal concepts, but he, of all philosophers, is not in a good position to
justify such hostility in a principled way. For the usual basis for re-
jecting the intelligibility of a whole sphere of concepts is that they
cannot be rendered intelligible within some chosen world-view - one
that sees only the physical or only the psychological, for example, as
real. Now one might attempt to motivate the rejection of metaphysical
concepts by adopting a world-view that sees only the first-order facts-
as given in ethics or mathematics or science, etc. - as real. But the
adoption of such a view already presupposes the intelligibility of a
metaphysical concept of reality. Thus there is a real danger that the
quietist's position rules out as unintelligible the only ground that could
possibly make it plausible.
There is also strong intuitive evidence in favor of intelligibility;
for the fact that a notion appears to make sense is strong prima facie
evidence that it does make sense. Indeed, the indispensability of the
notions in formulating certain metaphysical issues would appear to
make their intelligibility almost impossible to deny. Consider the issue
dividing the "A-theorist" and the "B-theorist" as to whether temporal
reality is intrinsically tensed. This is an issue that cannot be rendered
intelligible without invoking the metaphysical conception of 'fact'. For
the A-theorist will want to affirm, and the B-theorist to deny, that there
are tensed facts in the world; and it is only the metaphysical rather
than the ordinary notion of 'fact' that can properly serve to represent
what is here at issue. I might also note that appeal to the metaphor of
an Archimedean standpoint is almost irresistible in this context. For
the B-theorist will want to adopt an Archimedean standpoint in which

temporal reality is described sub specie aeternitatis, while the A-
theorist will deny that there is any such standpoint to be had. Granted
the intelligibility of the issue, we should grant the intelligibility of the
notion and of the metaphor in application to this particular case. And if
in this particular case, then why not in general?
But even though the charge of unintelligibility cannot reasonably
be sustained, there is another more moderate objection that can be and
that is equally devastating in its implications for the pursuit of meta-
physics. 19 This quietism is methodological rather than conceptual in
orientation. The charge is not that there are no meaningful notions of
factuality or reducibility but that there is no way of ascertaining what
is or is not factual or what does or does not reduce to whaeo Given
that nonfactual propositions are in the nature of imposters, how are we
to tell them apart from the real thing? And given that reductions have
antirealist import, how are we to establish that any proposed connec-
tion between propositions will have such import? Judgments concern-
ing factuality and reducibility would appear to be metaphysical in the
pejorative sense of floating free from any considerations that might tell
for or against their truth.
The methodological quietist can perhaps concede that there is a
general presumption in favor of a proposition's being factual. He
might also concede that we appear to have a metaphysical bias against
certain kinds of propositions' being factual - e.g., those concerning
matters of taste; and he might be willing to grant the plausibility of
certain conditional judgements, such as 'If it is factual matter whether
P then it is factual matter whether not-P?' or 'If it is factual matter
whether snow is white then it is factual matter whether grass is green'.
But it will be agreed on all sides that such considerations will not take
us very far. What is needed are detailed considerations for or against a
given realist view; and what is not clear is what these considerations
might be.
Of course, the philosophical literature appears to be full of argu-
ments for or against this or that form of realism. It is often maintained,
for example, that it is easier for the antifactualist about mathematics to
account for the possibility of mathematical knowledge, since the factu-
alist faces the problem of explaining how we can be in appropriate
contact with an external realm of mathematical facts; and it is often

thought to be an advantage of the expressivist view in ethics that it can
account for the motivational role of moral belief. But all such argu-
ments, in so far as they are taken to bear upon the non-skeptical issue,
would appear to be subject to a devastating critique. For it is unclear
how they turn upon adopting a metaphysical, as opposed to an ordi-
nary, conception of reality and hence why they establish a non-
skeptical, as opposed to a skeptical, fonn of antirealism. Thus even if
we employ the thin, ordinary notion of a fact, there would still appear
to be a problem of explaining how we could have knowledge of
mathematical facts; and it remains unclear how this problem becomes
greater once we substitute the thick metaphysically inflated notion for
the thin ordinary notion. Or again, it is unclear what is it about belief
in the MORAL FACTS (the real thing) as opposed to belief in the
moral facts (the ordinary thing) that makes it any the less plausible to
suppose that moral beliefs are a kind of attitude.
This, I believe, is the truly serious problem raised by quietism. It is
not that the notion of factuality is senseless, but that it is useless; and
realist metaphysics should be abandoned, not because its questions
cannot be framed, but because their answers cannot be found. The real
world of the metaphysician is akin to Kant's noumenal world, a
something-we-know-not-what, and no progress is to be made by in-
quiring into its constitution. We might add that if the methodological
problem could be solved, the conceptual problem would then lose
much of its bite. For how can we seriously doubt the intelligibility of a
given discourse, when its employment in resolving disputes is not oth-
erwise in doubt?
What I would like to do in the remainder of the paper is to show
how these concerns can be met. I wish to make clear the role of the
concepts of factuality and reducibility in realist disputes and thereby
show how we might make progress in settling such disputes. Thus it is
not my aim to defend the coherence of these concepts. Indeed, for the
purpose of dispelling methodological doubts, it is better to throw con-
ceptual caution to the winds and adopt whatever models or metaphors
might help us understand how the concepts are to be employed. Nor it
is my aim to show how we might actually settle realist disputes. After
all, this is not something of which we nonnally consider ourselves ca-
pable in even the most unproblematic areas of philosophy. Rather I

wish to show how we might proceed. We need to know what it would
take to settle the disputes, even if we can have no assurance of settling
them in any given case.
Most anti-quietists have attempted to allay the quietist's concerns
by producing criteria for factuality or reducibility in terms of which
questions of realism might then be posed. One question has simply
been substituted for another. Our strategy for dealing with the quietist
is quite different. We attempt to see how questions of realism might
tum on other, more tractable questions, without presupposing that they
are to be rendered intelligible in terms of those other questions. We do
not thereby commit ourselves to the view that the key metaphysical
concepts cannot be defined in fundamentally different terms. But
clearly, in so far as we can remain neutral on this question, our defense
of realist metaphysics is likely to be far less contentious, and we lessen
the danger, to which all philosophy is prone, of making an issue clear
only by misrepresenting what it is.
5. GROUND
Metaphysical questions of realism are to tum on questions of a less

problematic nature. But what are these questions? I suggest that they
concern relationships of ground and so, before we proceed further, let
us attempt to explain what these are.
I recommend that a statement of ground be cast in the following
"canonical" form:
Its being the case that S consists in nothing more than its being the
case that T, U, ...
where S, T, U, ... are particular sentences. As particular examples of
such statements, we have:
Its being the case that the couple Jack and Jill is married consists
in nothing more than its being the case that Jack is married to
Jill.
Its being the case that Britain and Germany were at war in 1940
consists in nothing more than ... ,
where ' .. .' is a compendious description of the warring activity of
various individuals.
In such cases, we say that the propositions on the right (collec-
tively) ground the proposition on the left and that each of them partly
grounds that proposition. I shall normally assume that the grounded
proposition and its grounds are true, though one may also talk of
"ground" when the grounding propositions would ground the grounded
proposition were they true.
The notion of ground should be distinguished from the strict no-
tion of reduction. A statement of reduction implies the unreality of
what is reduced, but a statement of ground does not. Thus in saying
that the fact that P /\ Q reduces to the fact that P and the fact that Q,
we are implying that the conjunctive fact is unreal; but in saying that
the fact that P /\ Q is grounded in, or consists in, the fact that P and the
fact that Q, we are implying no such thing. We are adopting a meta-
physically neutral stand on whether there really are conjunctive facts
(or truths). Thus our view is that there is sense in which even a realist
about conjunctive facts may be willing to concede that the fact that P
/\ Q consists in the fact that P and the fact that Q; there is a position
here that may be adopted by realist and antirealist alike. 21
The notion of ground, like the notion of reduction, is also to be
distinguished from logical analysis. Indeed, the paradigm of logical
analysis ("the average American") is not for us a case of ground, since
the propositions expressed on both sides of the analysis are presuma-
bly the same and yet no proposition can properly be taken to ground
itself. For us, the potentially misleading surface appearance of gram-
mar is entirely irrelevant to questions of ground, since we are looking
to the propositions expressed by the sentences rather than to the sen-
tences themselves. Thus we distinguish between the essentially lin-
guistic matter of determining which proposition is expressed by a
given sentence (whether, for example, a term is a genuinely referring
expression) and the essentially metaphysical matter of determining
what grounds what.
We take ground to be an explanatory relation: if the truth that P is

grounded in other truths, then they account for its truth; P's being the
case holds in virtue of the other truths' being the case. There are, of
course, many other explanatory connections among truths. But the re-
lation of ground is distinguished from them by being the tightest such
connection. Thus when the truth of P causally explains the truth of Q,
we may still maintain that the truth of Q consists in something more
(or other) than the truth of P. Or again, the fact that someone broke a
promise may "normatively" account for his having done something
wrong, but that is still compatible with his wrongdoing's consisting in
something more than his having broken the promise. There is, how-
ever, no explanatory connection that stands to ground as ground stands
to these other forms of explanation. It is the ultimate form of explana-
tion; and it is perhaps for this reason that we are not inclined to think
of the truth of a grounded proposition as a further fact over and above
its grounds, even though it may be distinct from its grounds and even
though it may itself be a real fact. 22
Although we have talked of the truth of one proposition as being
grounded in the truth of others, this is not strictly necessary. For we
might express statements of ground in the form'S because T, U, ... ', as
long as the 'because' is taken in a suitably strong sense, and thereby
avoid all reference to propositions or facts or to the concept of truth.
The word 'ground' would, in effect, be a sentential operator, in the
same way as 'if-then' or 'unless'.23 This point is of some philosophical
interest, since it shows that there is no need to suppose that a ground is
some fact or entity in the world or that the notion of ground is inextri-
cably connected with the concept of truth. The questions of ground,
upon which realist questions turn, need not be seen as engaging either
with the concept of truth or with the ontology of facts.
6. SETTLING QUESTIONS OF FACTUALITY
In this and the next section we examine how questions of factualism

may be settled on the basis of considerations of ground. Our approach
is somewhat indirect. We begin by presenting an abstract argument to
the effect that any reasonable disagreement on the factual status of a

given proposition will lead to a disagreement on what grounds what;
and we then attempt to show how the ensuing questions of ground
might themselves be resolved. It will be seen that these latter questions
tum on whether it is the factualist or the antifactualist who is able to
provide the better account of our "practice".
We have suggested that there can be no conceptual guarantee of
factualism or nonfactualism on the basis of essentially different con-
siderations, and so in attempting to trace out the possible differences
between the factualist and the antifactualist, it will be necessary at
some point to appeal to what is plausible rather than to what is con-
ceptually required. Let us therefore suppose that two philosophers dif-
fer on the factual status of a given true proposition. We then wish to
show that their disagreement, as long as their respective positions are
themselves plausible, will lead to disagreement on some question of
ground.
We may illustrate the idea behind our argument with the proposi-
tion that abortion is wrong. Although the anti factualist will take this
proposition to be nonfactual, he will presumably agree with the factu-
alist on the factuality of the proposition that so-and-so said that abor-
tion is wrong. However, whereas the antifactualist will wish to say in
what the truth of this proposition consists without making any refer-
ence to wrongness, the factualist will hold that no such account can be
given. Thus they will differ on what may ground this further proposi-
tion.
Let us now attempt to state the argument in general form. It will
proceed in stages; and, at each stage, we shall make explicit the possi-
bly problematic assumptions that are employed. Articulating these as-
sumptions will help make clear the "dialectical space" or logic within
which questions of factuality are to be resolved; and once we have
completed the exposition of the argument, we shall attempt to show
that these assumptions are indeed defensible.
Stage 1. Let P be the true proposition upon whose factual status
the factualist and antifactualist disagree. 24 Say that a true proposition is
basic if it is not grounded in other propositions. We now ask the factu-
alist, "Is the given proposition P basic?" If he says "Yes", then we
proceed to the second stage. Ifhe says "No", we ask him, "Which ba-
sic propositions collectively ground the proposition P?" Granted:
(a) any true nonbasic factual proposition is grounded in basic

propositions
there will be some basic propositions that ground P; and so let us sup-
pose that they are Q, R, S, ....
If the antifactualist denies that these propositions ground P, then
we already have a disagreement on ground. So suppose he agrees that
they ground P. Then he must take one of them, say Q, to be nonfac-
tual, since
(b) no nonfactual proposition has a ground consisting entirely of

factual propositions.
The factualist, on the other hand, will take all of them to be factual,
since
(c) no factual proposition IS partly grounded III a nonfactual
proposition.
Thus the two philosophers will differ on the factual status of the
proposition Q; and we may substitute the present Q for the previous P
and proceed to the next stage.
Stage 2. Our factualist and antifactualist will differ on the factual
status of the proposition P, which the factualist takes to be basic.
Given that the antifactualist takes the proposition P to be nonfactual,
he must acknowledge that at least one of its constituents is nonfac-
tual/ s since
(d) any nonfactual proposition will contain a nonfactual constitu-

ent.
Thus in the case of the proposition that abortion is wrong, the nonfac-
tual constituent would presumably be the attribute wrong.
Let C, D, ... be the constituents of the proposition which our anti-
factualist takes to be nonfactual. Now it is conceivable that our factu-
alist might also take some of the constituents C, D, ... to be nonfactual
despite believing the given proposition to be factual. But any plausible
factualist and anti factualist position will surely agree on the question
of whether the given proposition would be nonfactual if the constitu-
ents C, D, ... were nonfactual. Thus even though they may disagree on
the factuality of the proposition that abortion is wrong, they will agree
that the proposition would be nonfactual if the attribute wrong were
nonfactual. But given that the factualist holds the given proposition to
be factual, and since he agrees with the antifactualist that the constitu-
ents C, D, ... , if nonfactual, would render the proposition nonfactual,
he must take one of those constituents to be factual.
Stage 3. Our factualist and antifactualist disagree on the factual
status of some constituent, say C, of the given proposition P. Say that a
proposition essentially contains a given constituent if its replacement
by some other constituent induces a shift in truth-value. Thus Socrates
is a essential constituent in the proposition that Socrates is a philoso-
pher though not in the proposition that Socrates is self-identical. Now
any plausible antifactualist view will presumably maintain
(e) any nonfactual constituent C is essentially contained in some
true factual proposition P+.
In the case of the ethical antifactualist, P+ might be the proposition
that so-and-so said that abortion is wrong (or attributed wrongness to
abortion) or the proposition that the word 'wrong' refers to wrongness.
The factualist, moreover, is plausibly taken to agree with the antifactu-
alist on this matter. There need be no disagreement about the proposi-
tion is essentially containing the given constituent or its truth; and the
factualist would appear to have even less reason than the antifactualist
for taking the proposition P+ to be nonfactual. Indeed, let us suppose
that the only other constituents in the proposition, besides C, are those
that both agree are factual. Then the proposition P+, for the factualist,
will only contain factual constituents and must therefore be factual.
Stage 4. Our factualist and antifactualist agree on the factuality of
the proposition P+ and yet disagree on the factual status of its con-
stituent C. Say that a proposition is imperfectly factual if it is factual
but contains a nonfactual constituent and that it is perfectly factual if it
is factual and contains only factual constituents. Thus the proposition
P+, for the antifactualist, is imperfectly factual. But then he will be-
lieve that it has a perfectly factual ground, i.e., one consisting entirely
of perfectly factual propositions, since
(t) any true imperfectly factual proposition has a perfectly factual
ground
The factualist, on the other hand, is plausibly taken to believe that the
proposition has no ground none of whose propositions involve the
constituent C, since
(g) whenever a constituent occurs in a true basic factual proposi-

tion and also occurs essentially in some true factual proposi-
tion, then any ground for the latter proposition must contain
the constituent.
Thus our antifactualist will maintain that there is some ground - R, S,
T, ... - for p+ that does not involve the constituent C, while our factu-
alist will deny that R, S, T, ... is a ground for P+. Disagreement on a
question of ground is thereby secured.
Let us now attempt to defend the assumptions (a)-(g) upon which
the argument depends. In this regard, it is important to bear in mind
that our argument is not fool-proof (and hence not philosopher-proof
either). We have not attempted to show that any factualist and anti-
factualist positions on a given proposition will lead, on conceptual
grounds alone, to a disagreement on some question of ground, but only
that any plausible factualist and antifactualist positions will lead to
such a disagreement. Indeed, if a factualist, let us say, were conceptu-
ally compelled to accept some statement S of ground which the anti-
factualist was not compelled to accept, then not-S would entail that the
given proposition was not factual without also entailing that it was
factual and so our stand on the "independence" of realist metaphysics
could no longer be maintained. Thus there are certain points in the ar-
gument where we must appeal to what it is plausible for the factualist
or antifactualist to accept in a given case rather than to what they are
compelled to accept. I should also point out that even if my specific
line of argument fails, there may be others like it that will succeed.
Assumption (a), that the nonbasic is grounded in the basic, is con-
troversial but also dispensable. For, as will become clear, the as sump-
tion will hold with 'fundamental' (or 'irreducible') in place of 'basic',

even in the presence of an infinite regress of grounds. 26
The truth of (b), that the factual can only ground the factual, seems
clear. For how can the truth of a nonfactual proposition consist entirely
in the truth of factual propositions? Would that not be enough to ren-
der the proposition factual? Of course, this is not to rule out senses of
'ground' or 'depend' in which the nonfactual might be grounded in, or
depend upon, the factual. Thus even an expressivist might agree that
the truth of any moral claim is "nonnatively" grounded in the truth of
certain naturalistic claims. But this is not the relevant sense of
'ground'; it is not being claimed that the truth of the moral claim con-
sists in no more than the truth of the naturalistic claims. Indeed, this
latter view would commit one to a fonn of naturalism and hence to a
denial that moral and naturalistic claims might differ in their factual
status.27
The truth of (c), that the nonfactual cannot partially ground the
factual, also seems clear. For how can the truth of something factual
partly consist in something nonfactual? Would that not be enough to
render the original proposition nonfactual? A possible counterexample
to the conjunction of (b) and (c) is the disjunction P of a factual truth
PI and a nonfactual truth P2 . For given that PI grounds P, it must be
factual by (b); and given that P 2 grounds P, it must be nonfactual by
(c). However, our view in such a case is that it is the two propositions
PI and P2 together that collectively ground P.2S
The truth of (d), that any nonfactual proposition contains a non-
factual constituent, is likewise apparent, since if a nonfactual proposi-
tion contained only factual constituents, its nonfactuality could have
no source; there would be nothing that could sensibly be said to render
it nonfactual. 29
Assumption (e), asserting the existence of a suitable imperfectly
factual proposition P+, is the main point of the argument at which con-
siderations of plausibility enter in, since there is nothing to force the
anti factualist into acknowledging the factuality of any given proposi-
tion, But it is surely very plausible that he will be willing to acknowl-
edge the factuality of some propositions of the required sort. Perhaps
he has doubts about belief-attributions. But can he not then consider
semantic attributions instead? Perhaps he has doubts about these. But
then can he not consider the question of when it is appropriate to

make such attributions? Indeed, I shall later suggest that there are fac-
tual considerations concerning any given domain that are almost
bound to arise, whatever one's view of the factuality of the domain
itself.
One might, of course, flatly declare all propositions to be nonfac-
tual. But this global form of antifactualism is not, in the present dia-
lectical setting, a viable option. For in attempting to argue for the non-
factuality of a given circumscribed range of propositions, the
antifactualist should not take for granted the nonfactuality of other
propositions, just as in arguing for the nonveridicality of a given range
of perceptual experiences, one should not take for granted the non-
veridicality of other perceptual experiences. Thus the antifactualist
should concede - if only temporarily, for the sake of argument - that
propositions outside of the given range are factual.
On this way of thinking, there is a general presumption in favor of
factuality and if global antifactualism is to be established at all, it is in
piecemeal fashion rather than by a general line of argument. One must
successively chip away at the apparent edifice of factuality; and it is
only then, when each part has been removed, that a global form of an-
tifactualism might emerge as a viable alternative.
According to assumption (f), any imperfectly factual truth must
have a perfectly factual ground, i.e., one that can be stated in factual
terms alone. For consider any truth containing a nonfactual constitu-
ent. If one asks the antifactualist why he takes it to be factual notwith-
standing the nonfactual constituent, then the only completely satisfac-
tory answer he can provide is that it has a perfectly factual ground. The
underlying metaphysical thought here is the inessentiality of the non-
factual in describing the factual. Even if the nonfactual were altogether
expunged from the ordinary world, we could still provide a complete
account of factual reality in terms of what remained; and this would
then provide a ground for all factual truths, whether formulated in
factual terms or not.
It is important, in this connection, not to be misled by our example
of someone's saying that abortion is wrong. For one might think that
an antifactualist in ethics could take this proposition to be ultimately
grounded in some facts relating the person to the concept wrong. But
in so far as this is plausible, the original proposition should be taken to

concern the concept wrong, which the anti factualist can legitimately
take to be a factual element, rather than wrongness itself. To clear up
any possible confusion on this score, take the proposition to be that the
person attributes wrongness to abortion. It would then be bizarre in the
extreme for the antifactualist to suppose that this proposition was ulti-
mately grounded in some facts relating the person to wrongness. For
how could there be real facts in the world relating the person to
wrongness and yet no real facts relating wrongness to things that were
wrong?
We may argue for assumption (g), concerning the ineliminability
of essential constituents, in the following way. If a given constituent C
occurs in a true basic factual proposition then it must be a fundamental
element of reality. But if some true factual proposition contains C es-
sentially, it must be true in virtue of some feature of C. But given that
C is a fundamental element of reality, this feature of C cannot be
grounded in something that did not itself involve C. 30
One possible kind of counterexample to this assumption is illus-
trated by the proposition that '5' refers to 5. For could not a realist
concerning numbers take its truth to consist simply in '5' being the
fifth counting-term, where this was something that did not involve the
number 5? However, a much more plausible view for the realist to take
is that its truth consists both in '5' being the fifth counting-term and in
5 being the fifth number. It is the connection between their both being
fifth in the respective series that then helps ground the fact that the one
refers to the other. Similarly, an ethical realist might suppose that the
fact that someone attributes wrongness to abortion is somehow
grounded in his behavior. But again, the ground is not plausibly taken
to be complete until the connection between his behavior and the at-
tribute wrongness is explicitly given. 3 !
It is worth remarking, in conclusion, on the critical role played by
the notion of ground in the above argument. If we had used a weaker
explanatory notion, then there would be no reason to suppose that the
various principles upon which the argument depends would hold.
There would be no objection, for example, to a nonfactual proposi-
tion's having an entirely factual, though normative, ground. This point
is important for understanding how the factualist and antifactualist
should be seen as facing different explanatory demands. For unless

these demands are understood in terms of the strict metaphysical no-
tion of ground, there is nothing to prevent the factualist and antifactu-
alist from meeting them in the very same way.32
7. SETTLING QUESTIONS OF GROUND
We have shown how to devise a "critical experiment" to test whether

to accept or to reject the hypothesis that a given proposition is factual.
For the claim that the proposition is factual will imply that a certain
related proposition has one kind of ground while the claim that it is not
factual will imply that the proposition lacks such a ground. By ascer-
taining the correct answer to the question of ground, we may thereby
ascertain the correct answer to the question of factuality.
But how are the questions of ground to be settled? The notion may
not be conceptually problematic in the same way as the notions of
factuality or reduction; for its application carries no realist or
antirealist import. But several of our previous arguments against a
definition or guarantee of reduction apply equally well to the notion of
ground; and in the absence of a definition or guarantee, we may have
similar methodological misgivings about how the notion is to be ap-
plied.
There are, I believe, two main sources of evidence for making
judgements of ground. The first is intuitive. We appear to be in posses-
sion of a wealth of intuitions concerning what does or does not ground
what. Some examples have already been given, but there are many
others. Thus what grounds the truth of a disjunction is the truth of
those of its disjuncts that are true, and what grounds the occurrence of
a compound event at a given time is the occurrence of its component
events. We also have intuitions about a wide range of negative cases
(quite apart from modal considerations). It is implausible, for example,
that what grounds facts about volume are facts about density and mass
or that what grounds the truth that a given object is red is the fact that
it is red or round and the fact that it is not round, even though the one
logically follows from the others.
The other main source of evidence is explanatory in character. As

we have mentioned, the relationship of ground is a form of explana-
tion; in providing the ground for a given proposition, one is explain-
ing, in the most metaphysically satisfying manner, what it is that
makes it true, Thus a system of grounds may be appraised, in much the
same way as any other explanatory scheme, on the basis of such con-
siderations as simplicity, breadth, coherence, or non-circularity.
Perhaps the most important virtue in this regard is explanatory
strength, the capacity to explain that which stands in need of explana-
tion and would otherwise be left unexplained. And here it is not simply
relevant that one grounds and hence accounts for certain truths but also
that, in so doing, one may account for the presence or absence of cer-
tain necessary connection between the propositions that are so
grounded.
Thus questions of ground are not simply to be settled on a case-
by-case basis but also on how well their answers fit into a general pat-
tern of explanation. Our critical experiment might have seemed to
have held out hope that a question of factuality could be decided on
the basis of a single question of ground. But in so far as we lack intui-
tions on such a question, any proposed answer must be placed within a
larger context of such answers and evaluated on the basis of largely
holistic considerations.
But what is this larger context and what are the considerations by
which its answers are to be assessed? In so far as the factuality of a
given proposition is in doubt, it will be because it is thought to contain
certain nonfactual constituents that occur in such a way as to render
the resulting proposition nonfactual. Let us call the class of proposi-
tions whose factuality is similarly in doubt the given domain and the
constituents which it is thought might be responsible for their nonfac-
tuality the contested constituents (or elements).
Associated with a given domain and a class of contested elements
will be another class, which we call the extended domain, consisting of
all those propositions which (essentially) contain the given elements
but which are agreed to be factual. We have already seen two kinds of
example of such propositions - the proposition that so-and-so said that
abortion is wrong and the proposition that 'wrong' refers to wrong-
ness. In so far as the given domain is taken to describe the "facts" of a
given area, the extended domain might be thought to describe our

"practice" of dealing with those facts. Thus where the one concerns
morality, science, or mathematics, let us say, the other will concern
our moral, scientific, or mathematical practice.
Two significant parts of our practice relate to our representation
and cognition of the given facts.33 Thus, in the case of morality, the
extended domain might include representational propositions to the
effect that we believe such-and-such a moral principle, or that we have
made such-and-such a moral claim, or that a moral term means what it
does, and it might also include cognitive propositions to the effect that
we know or are justified in holding a given moral belief, or that we are
morally sensitive, or that we are biased in our moral views. The ex-
tended domain may also include propositions peculiar to the area in
question. Thus it may include propositions to the effect that we have
been motivated by such-and-such a moral belief in the moral case or
propositions involving the application of mathematics in the mathe-
matical case.
For certain radical forms of antifactualism, many of these aspects
of our practice might themselves be taken to be nonfactual. Thus an
antifactualist about meaning might well take all propositions concern-
ing meaning, reference, and justification to be nonfactual. But I sus-
pect that, even in these cases, it will be possible to find aspects of our
practice upon whose factuality the parties to the dispute can agree. In
the first place, it seems to me that the antifactualist - in common with
the factualist - should be willing to acknowledge that there is a/actual
standard of correctness. Of course, the obvious standard of correctness
will be nonfactual; for the correctness of the judgement that abortion is
wrong, say, will simply amount to abortion's being wrong - which, for
the antifactualist, is a nonfactual matter. But this nonfactual standard
of correctness lives in the shadow, as it were, of a factual standard. For
the correctness of our judgements must somehow engage with the real
world; there must be something which we aim for in belief and whose
realization is a factual matter. So for the expressivist, for example, the
factual standard of correctness for a judgement might be that it faith-
fully reflects one's (possibly implicit) commitments; while for the
mathematical formalist, it could be that the judgement is in accordance
with the rules of the game. Thus, even in these cases, there is some.-
thing of a factual sort that counts as getting things right.
Similarly, there would appear to be a factual sense in which some-
one may be said to be responsive to the facts. Again, the most obvious
sense of being responsive to the facts for the antifactualist is one that is
itself nonfactual, since it will rest upon what one takes the facts to be.
But in the shadow of this sense, there is a another sense that is factual.
For our epistemic activity must somehow engage with the real world;
there must be something that we aim for - in aiming to be well-placed
epistemic agents - whose realization is a factual matter. So for the ex-
pressivist, it might consist in being appropriately sensitive to one's
(implicit) commitments in the formation of one's ethical beliefs; while,
for the formalist, it might consist in being appropriately responsive to
the rules of the game in executing a proof.
Granted that they can agree upon a common practice, the factualist
and anti factualist each owes us an account of what it is, of that in
which it consists. Or to put it more precisely, each should provide us
with an account of what might ground the propositions from the ex-
tended domain. But their accounts are subject to very different con-
straints. The antifactualist must provide an account of the practice
without making any reference to the contested constituents. The ex-
pressivist, for example, must be able to say what having a moral belief
might consist in without making any use of moral vocabulary, and the
formalist must be able to say what possessing a mathematical proof
might consist in without making any use of mathematical vocabulary.
It is this constraint that explains why a antifactualist must be able to
provide some alternative to a truth-conditional account of our under-
standing of language; for truth, in its application to the sentences of a
given nonfactual domain, is nonfactual and must therefore be elimi-
nab Ie. It also explains why the standard formulations of antifactualist
positions - expressivism, constructivism, formalism etc. - are com-
monly taken to be antifactualist, even though this is not strictly implied
by the formulations themselves. For they provide the general means by
which the constraint may be seen to be satisfied. It is immediately
clear from the expressivist's position, for example, how he would wish
to account for ethical belief without making use of ethical terms.
The factualist, by contrast, is obliged to make reference to the

contested elements (at least in so far as they are taken to be basic or
fundamental). The moral platonist, for example, cannot give an ac-
count of what it is sincerely to ascribe wrongness to a given act in
terms of having a con-attitude to that act, since the connection with the
attribute is thereby lost; and the arithmetical platonist cannot give an
account of what it is to refer to the natural numbers without making
appeal to the natural numbers. It is also this constraint that explains
why the factualist may find a truth-conditional account so congenial,
since it will connect our understanding of language in the required way
with the elements of reality with which it deals.
The factualist's account must in this sense be representational: it
must link up the practice with the underlying facts or subject-matter,
while the antifactualist's account will be nonrepresentational. In the
one case, the practice must be seen as engaging with the possible facts
and it must be understood - at least, in part - in terms of how it en-
gages with those facts. In the other case, the practice is taken to be dis-
engaged from the facts; and rather than understanding the practice in
terms of how it represents the possible facts, the facts themselves
should be understood in terms of how they are "projected" by the
practice. It is in this sense that they are subjective or not "out there".
For a nonfactual proposition is ultimately to be understood - not in
terms of its grounds, of what in the world makes it true - but in terms
of its role within a given practice. They are metaphysically incomplete
propositions, as it were, and should be understood, in much the same
manner as Russell's "incomplete symbols", by means of the context of
their use rather than by means of their isolated application to the
world.
The question of whether or not to be a factualist is therefore the
question of whether or not to adopt a representational account of what
grounds our practice. And this question, in its tum, is largely a matter
of determining which of the rival accounts is better able to meet the
explanatory demands that may be placed upon it. Can the moral factu-
alist account for the motivational role of moral belief or the moral anti-
factualist account for its inferential role? Can the mathematical factu-
alist account for the referential capacity of mathematical language, or
the constructivist for its application to science? Can the factualist
about meaning provide an adequate account of the grounds for the

meaning statements that we make, or the anti factualist an adequate ac-
count of the factual standards of correctness by which they appear to
be governed? It is on their answer to these and many other such ques-
tions that the correctness of a factualist or antifactualist position will
ultimately be settled. 34
Given that their accounts are subject to different constraints, the
factualist and the anti factualist will meet these demands in character-
istically different ways. But can we be confident that these differences
will enable us to adjudicate between their accounts? What is to rule out
the possibility of a stalemate in which the advantages and disadvan-
tages of the two accounts appear to be more or less equally matched?
Or even if one account seems preferable to another, perhaps it is not
itself so plausible as to be worthy of belief. Suppose, for example, that
the most plausible representational account of our mathematical prac-
tice is epiphenomenal: it consists of a nonrepresentational part and a
"parallel" mathematical part (e.g., the reference of '5' to 5 consisting
in '5' being the fifth counting-term and 5 the fifth number). It is then
hard to see how we are to choose between such an account and the cor-
responding nonrepresentational account.
It must be conceded that we have no a priori basis for excluding
such cases. But nor do we have any good reason to expect them. Some
philosophers, it is true, have been impressed by our repeated failure to
solve the problems of realism in the past and have become completely
pessimistic about our ability to make any progress on them in the fu-
ture. But I suspect that these philosophers have not fully appreciated
how much needs to be done before these problems can properly be ad-
dressed. For in providing an account of a given practice, we must
come up with what is in effect a complete epistemology, philosophical
psychology, and theory of language for the area in question; and in as-
sessing such an account, we must in effect solve all of the major philo-
sophical problems to which the area gives rise. Until we have settled
the question of whether moral beliefs necessarily have motivational
force, for example, we are in no position to say whether it is a point in
favor of a given account of our moral practice that it endows them
with such a force; and until we have decided whether mathematical
beliefs can be known a priori, we will be unable to say whether it is a
point in favor of an account of our mathematical practice that it allows

them to have such a status. A realist or antirealist conclusion therefore
represents the terminus of philosophical inquiry into a given area
rather than its starting point; and so it is hardly surprising that such
slight progress has been made within realist metaphysics, even by
comparison with other branches of philosophy.
8. REALITY AS FUNDAMENTAL
We have distinguished between two conceptions of reality - as factual

and as fundamental. We now tum to the second of these and, after
clarifying the concept in the present section, we attempt to show in the
next section how questions concerning its application might be re-
solved.
It is natural to understand the concept of fundamental reality in
terms of the relative concept of one thing being less fundamental than,
or reducible to, another - the fundamental being whatever does not
reduce to anything else (but to which other things will reduce). But we
appear thereby to play into the quietist's hands. For how can an ex-
planatory connection be determinative of what is and is not real? We
may grant that some things are explanatorily more basic than others.
But why should that make them more real?
What I would like to suggest, in the face of this difficulty, is that
we reject the idea that the absolute notion of fundamental reality is in
need of a relational underpinning. The conception of reality that we are
after is simply the conception of Reality as it is in itself. Thus even
though two nations may be at war, we may deny that this is how things
really or fundamentally are because the entities in question, the na-
tions, and the relationship between them, are no part of Reality as it in
itself. One might think of the world and of the propositions by which
the world is described as each having its own intrinsic structure; and a
proposition will then describe how things are in themselves when its
structure corresponds to the structure of the world. 3s Thus it is this
positive idea of the intrinsic structure of reality, rather than the com-
parative idea of reduction, that should be taken to inform the relevant

conception of what is fundamental or real.
It is also important to distinguish the notion of Reality in itself
from certain other notions of intrinsic reality. In talking of the intrinsic
nature of the physical world, for example, one might have in mind its
nonrelational or nondispositional features, but these features may be
no more a part of Reality in itself, on my understanding, than the rela-
tional and dispositional features. Or again, in talking of the intrinsic
nature of the physical world, one might be after a description that is
intrinsic to the world in the sense of being nonperspectivaP6 Thus
color terms might be excluded on the grounds that our understanding
of them is based upon a peculiar form of sensory awareness. But as
long as these terms pick out fundamental physical properties, I would
willing to countenance their use in the description of Reality in itself,
however they might have been understood.
Given the notion of reality as primitive, it is then possible to de-
fine the notion of reduction. Intuitively, one proposition will reduce to
others if they bring us closer to what is real. Now a necessary condi-
tion for the proposition P to reduce to the propositions Q, R, ... is for it
to be grounded in those other propositions; and a necessary and suffi-
cient condition for Q, R, ... to be closer to reality than P in such a case
is that P be unreal and each of Q, R, ... either be real or "en route" to
what is real. But the latter is presumably just a matter of the proposi-
tion's being grounded in what is real. Thus we arrive at the following
definition:
the true proposition P reduces to the propositions Q, R, ... iff (i) P
is not real; (ii) P is grounded in Q, R, ... ; and (iii) each of Q,
R, ... is either real or grounded in what is real. 37
On this approach, reduction is to be understood in terms of fundamen-

tal reality rather than the other way round, and there is no mystery as
to why reduction has antirealist import, since that import is built in to
the notion itself. It is also able to avoid the problems that afflicted the
previous accounts of reduction since it shares, with the underlying no-
tion of ground, the capability of enjoying one-many instances, of ap-
plying to the de re, and of having explanatory import.
9. SETTLING WHAT IS FUNDAMENTAL
Of course, the extreme quietist will not be happy with the concept of
Reality in itself. But it should be recalled that our target is the moder-
ate quietist. We have thrown conceptual caution to the winds and our
only question is: Given that the concept is intelligible, then how is its
application to be settled? How are we to determine, from all the possi-
ble truths, which are descriptive - or possibly descriptive - of such a
reality, and which are not?
In attempting to determine what is real in this way, we cannot
simply appeal to the fact that a given proposition is basic. For a basic
proposition may be nonfactual; and clearly no nonfactual truth is de-
scriptive of fundamental reality. However, any basic factual proposi-
tion will be real. For any true factual proposition is real or grounded in
what is real; and so the proposition, if basic, will be real.
Our previous methodology for determining what is factual can
therefore be of help in determining what is real. For once it is resolved
that a given domain is factual, then any basic propositions from that
domain can be taken to be real. But even without any help from that
methodology, there is perhaps a general presumption in favor of any
given proposition's being factual; and so the mere fact that a proposi-
tion is basic will give us reason to believe that it is real, in the absence
of any reason to the contrary.
Just as we cannot read off what is real from what is basic, so we
cannot read off what is unreal from what is nonbasic. Indeed, it is pos-
sible to imagine metaphysical scenarios in which the nonbasic, or
grounded, is plausibly taken to be real. Suppose, to take one kind of
case, that Aristotle is right about the nature of water and that it is both
indefinitely divisible and water through-and-through. Then it is plausi-
ble that any proposition about the location of a given body of water is
grounded in some propositions about the location of smaller bodies of
water (and in nothing else). The proposition that this body of water is
here, in front of me, for example, will be grounded in the proposition
that the one half is here, to the left, and the other half there, to the
right. But which of all these various propositions describing the loca-
tion of water is real? We cannot say some are real and some not, since
there is no basis upon which such a distinction might be made. Thus
we must say either that they are all real or that none are. But given that
the location of water is a factual matter, we should take all of them to
be real, notwithstanding the fact that each is grounded in propositions
of the very same sort. 38 Another kind of case involves "horizontal"
rather than "vertical" considerations. Imagine an ontology that takes
certain simple events and the causal relationships between them to be
real. Suppose now that one simple event causes a compound of simple
events. Then this presumably consists in its causing one component of
the compound and in its causing the other component. Now suppose
that a compound of simple events causes a simple basic event, even
though no component of the compound causes the event. Then it is not
clear what the ground might be. But if this causal relationship is taken
to be basic and hence real, then compound events should also be taken
to be real, and so causation of the compound - which is a real relation-
ship between real relata -should also be taken to be real, notwith-
standing its being grounded in other causal relationships.
So given that one proposition is grounded in others, how are we to
ascertain whether or not it is real? What I would like to suggest is that
there is a general presumption in favor of the grounded not being real.
In the absence of any reason to the contrary, such as those illustrated
by the cases above, we should assume that any given grounded propo-
sition is unreal.
The presumption may be justified by reference to the general aims
of realist metaphysics. For the distinction between what is and is not
real represents a general strategy for making metaphysical sense of the
factual world. For, of all of the structure that the world exhibits, some
may be taken to be real, to belong to the world itself, and some to be
only apparent and to be understood by reference to what is real. Let us
call a division of all propositions into those that are real and those that
are unreal a world-view. Thus a world-view will correspond to a par-
ticular attempt to see the world as intelligible in terms of the distinc-
tion between what is and is not real.
Let us now call a factual proposition moot if it is grounded and if
there is no special reason to think it real. There are then three possible
world-views one might adopt: the minimalist, which takes each moot
propOSItIon to be unreal; the maximalist, which takes each moot
proposition to be real; and the middling, which takes some moot
propositions to be unreal and some to be real. (We do not need to con-

sider nonfactual propositions since it is clear that they are unreal).
Now of these three alternatives, the third should be excluded on the
grounds that it draws an arbitrary distinction between moot proposi-
tions, taking some to be real and others not when there is no basis for
so doing. Thus it does not correspond to an intelligible metaphysical
conception of the world. The second should also be excluded. It is not
arbitrary in its treatment of moot propositions, but in treating them all
as real, it effectively abandons the explanatory strategy for which the
distinction between what is and is not real was intended. In terms of
that strategy, it is effectively equivalent to adopting a position that re-
fuses even to acknowledge the distinction between what is and is not
real. Thus the only reasonable alternative is the first; and it is this that
then justifies us in taking every moot proposition to be unreal.
10. THE UNITY OF REALIST METAPHYSICS
We see that questions of factuality and reality are to be answered by

essentially the same means. It is not merely that the determination of
what is factual is relevant to the determination of what is real but that,
in both cases, the questions are largely to be settled through considera-
tions of ground. In the one case, we must look to the propositions of
the extended domain to see whether an account of their grounds is best
given in representational or nonrepresentational terms; and in the other
case, we must look to the propositions of the given domain, assuming
them to be factual, and attempt to ascertain from the overall structure
of their grounds how the division into what is and is not real is best
effected. Thus once all questions of ground are decided, all questions
of what is real - either in the sense of what is factual or what is fun-
damental - can be resolved. Underlying this methodological unity is, I
believe, a significant conceptual unity. We have so far treated the fac-
tual and the fundamental as independent conceptions of metaphysical
reality. But they are intimately related. For it is clear that any (funda-
mentally) real proposition is factual and that any proposition grounded
in the factual, and hence any proposition grounded in the real, is fac-
tual. But also if a proposition is factual, then it must be rendered true

by the real world, and if it is not itself real, it must be grounded in the
real. We therefore arrive at the following definition:
a proposition is factual iff it is real or it is grounded in what is
real. 39
Realist metaphysics, on this view, has a single focus - the fundamen-
tally real - and our interest in other categories of reality will derive
from their connection with this more fundamental category. It is the
explanatory axis, as it were, upon which an account of the world will
tum. For a given proposition may either be identical to the real (the
real itself) or be reducible to the real (the unreal) or be neither identical
to nor reducible to the real (the nonfactualor irreal). And correspond-
ing to each type of proposition will be a characteristic account of the
proposition's metaphysical import - of how it relates to reality. It may
either be real, in which case there is nothing further to be said, since
the proposition bears its import "on its face"; or it may be unreal, in
which case its metaphysical import is given by its grounds; or it may
be irreal, in which case its metaphysical import is given by those fac-
tual propositions that reflect its use The aim of realist metaphysics is to
render the world intelligible in terms of the distinction between what is
and is not real; and its task is complete once it becomes clear how
what is apparent, or not real, is to be rendered intelligible in terms of
what is real. 40
Notes
1 I here ignore the possibility that reconciliation is to be achieved by modifying
our view of received opinion, either by not taking it to be matter of what we believe
or by supposing that its content is other than what we naturally take it to be. These
attempts at reconciliation, to my mind, merely shift the conflict with received opin-
ion to another place.
2 Many philosophers do not take reduction to have antirealist import. Their con-
cept of reduction seems to correspond more closely to what I later call 'ground'.
3 I here slide over the difficulty of whether the bearers of nonfactuality can
properly be said to be propositions.
4 A useful survey of such criteria is given in Wright [1992] and some general
critiques of them are to be found in Rosen [1994], Dworkin [1996] and Stroud
[2000], Chapters 1-2.
5 A similar proposal has been made by Gaifman [1975]. It should be noted that
Dummett [1993], p.467, is not inclined, as I am, to assimilate the nonfactualism of a
sophisticated form of expressivism with the nonfactualism of a constructivist posi-
tion. However, none of my criticisms will turn upon making this assimilation.
6 See Edgington [1980-81] and Winkler [1985] for further criticisms of Biva-
lence as a sufficient condition. I also believe that there are problems with Bivalence
as a necessary condition for realism even when the obvious sources of truth-value
gaps (such as vagueness or reference-failure) are removed, but this is not something I
shall discuss.
7 Let us use 't F A' for 't warrants A' and 't ~ s' for 't improves s' (I assume
that any situation is an improvement upon itself). Molecular formulas within the pro-
posed semantics may then be subject to the following clauses:
(i) t FB 1\ C ifftFB and t FC
(ii) t FBv Ciff(Vt* ~ t)(3u~ t"') (u FB or u FC)
(iii) t F~B iff (Vt* ~ t) (not-(t* FB)).
Given that atomic formulas p satisfy the non-foreclosure condition:
(*) t FP iff (Vt* ~ t)(3u ~ t*) (u FP)
then so will every formula A.
s Dummett [1978], pp.365-67, has reservations about making the transition from
the Law to the Principle, but 1 do not believe that they apply in the present case.
9 Sometimes the emphasis is on a correspondence with "facts" rather than
propositions.
10 The significance of de re reductive claims has not been properly appreciated.
They enable one to achieve a huge simplification in the formulation of many reduc-
tions and in certain cases - such as the bundle theory of particulars - they are essen-
tial to understanding the very point ofthe reduction.
11 Similarly, it is perfectly in order for the ethical antifactualist to express his
view that there are no genuine moral properties in the form "For any moral property
and any possible bearer of the property the bearer's possession of the property is a
nonfactual matter". We quantify over all moral properties in order to express the
view that none of them are "real".
12 Dummett propounds such a conception in [1993], pp.56-7.
13 Or to put it more linguistically, between the satisfaction of the condition 'z is mar-
ried' by c and the satisfaction of the condition 'x is married to y' by a and h.
14 Advocates of this approach include Armstrong [1997], p.12, Chalmers
[1996], p.48, and Jackson [1998], p.5. Many philosophers, I should note, do not take
supervenience to capture a metaphysically significant notion of reduction.
15 For further discussion of these points, see Kim [1993], pp.I44-46, and the
references contained therein.
16 We should note an asymmetry in the two cases, however. For even though
there may be no guarantee of nonfactuality, there will be a guarantee of nonreduci-
bility. For if R is reducible to S then S will necessarily imply R and so the possible
truth of Sand -,R (a purely modal fact) will guarantee that R is not reducible to S.
17 Ronald Dworkin has pointed out to me that first-order morality is naturally
taken to include the claim that if there are no objective moral facts then anything
goes (or some other such moral conclusion) and, given that this is so, the second-
order claim that there are no objective moral facts will not be independent of moral-
ity. However, for present purposes, I would not take the above conditional to be
strictly first-order.
18 Their number include Blackburn [1992], pp. 7, 34, 168, and [1998], p.319;
Dworkin [1996]; A. Fine [1984], pp.97-100; Putnam [1987], p.19. Other philoso-
phers, such as Rosen [1994], have flirted with quietism without actually embracing
it.
19 Rorty [1979], p.311, is someone who has espoused a more moderate form of
quietism.
20 It is an oddity in the logical positivist's critique of metaphysics that these two
charges were linked. For if there is no way of settling metaphysical questions, then
who cares whether or not they make sense?
21 Some philosophers have thought of supervenience as a metaphysically neutral
counterpart to the notion of reduction and to this extent, at least, what they have in
mind may correspond to our notion of ground.
22 I should note that I do not take all judgements of ground to be a priori. Thus
the philosophical investigation of reality should only be based upon those judge-
ments that are a priori or that can be given some kind of a priori backing.
23 I hold a similar view concerning the notions offactuality and irreducibility.
24 For simplicity, we assume that P (and its successor p+) are true, though all
that is strictly required for the argument is that they be possibly true.
25 Intuitively, a nonfactual constituent is one which can be a source ofnonfactu-
ality in a proposition to which it belongs. Perhaps the factuality of constituents can
be defined in terms of propositional factuality in the following way. With any con-
stituent c may be associated the class of propositions Pc in which c has its primary
employment. A constituent c is then nonfactual iff any proposition of Pc is nonfac-
tual.
I have presupposed that propositions are structured entities built up from their
constituents. Those who do not like this assumption may conduct a parallel argument
with sentences and their terms in place of propositions and their constituents.
26 It then becomes less clear whether the factualist need take the constituent C to
be "ineliminable" at stage 4. But as we shall see, the only plausible cases in which it
is eliminable are ones in which it gives way to constituents which the factualist and
anti factualist will recognize as equivalent, in their factuality status, to the given con-
stituent.
27 When P is the disjunction PI v P z of a true factual proposition PI and a false
nonfactual proposition P 2' it will be grounded by PI alone and we will then wish to
say, in conformity with (b), that P is factual. Thus factuality, on our current under-
standing, is a contingent matter. (On an alternative understanding, a proposition
might be taken to be nonfactual when it could be nonfactual in our current sense). I
might add that we do not think of vagueness as a source, per se, of nonfactuality,
since a vague proposition may still be aimed at the real world. Thus cases in which a
vague truth is grounded in a precise truth are also not counterexamples to (b).
28 The truth of P v -.P when P is a nonfactual truth is also not a counterexample
to (c) since P v "OP, in this case, should be taken to be nonfactual.
29 Consider the proposition that the last thing the Pope said is true and suppose
that the last thing that he said was that abortion is wrong. Then this proposition is
nonfactual (if ethics is nonfactual) and so we are obliged by this assumption to treat
some constituent of the proposition - presumably either 'thing' or 'true' - as non-
factual. This is awkward, since it means that the proposition is imperfectly factual
even when the last thing that the Pope said was factual. There is a related difficulty
for the thesis that no 'ought' can be derived from an 'is'. For from the assumptions
that the last thing the Pope said is true and that the last thing the Pope said is that
abortion is wrong, we can derive the conclusion that abortion is wrong. Thus the first
assumption should be taken to be an 'ought' even though it apparently contains no
moral terms. Perhaps there is some other way of dealing with such cases.
30 A related assumption (which he regards as a fallacy) has been adumbrated by
Horwich [1998], p.21. This is that "whenever a fact has a certain component, then
whatever constitutes this fact must contain either the same component or alterna-
tively something that constitutes it". Assumption (g) says that the constituting fact
must contain the same component when that component is itself fundamental, i.e.,
such as to occur in an unconstituted fact. But if the constituting fact does not contain
something that constitutes the component in his sense, there is no reason to suppose
that it is fundamental in my sense. Thus accepting (g) is perfectly compatible with
rejecting his fallacy.
31 On certain deflationary views, of the sort proposed by Field [2001], Chapters
4 and 5, these propositions would not even be taken to involve a relationship be-
tween a term or concept and an entity.
32 Thus Putnam's argument for scientific realism ([1978], p.lDO) and the argu-
ment that Harman [1977], Chapter 1, considers against moral realism both turn on
whether the best explanation of some phenomenon (the success of science, our moral
responses) does or does not involve reference to the facts that are in dispute. But in
so far as the relevant notion of explanation is not metaphysical, it is unclear why the
factualist or antifactualist should differ on this question. I also doubt, though this is a
separate matter, that these arguments can plausibly be brought to bear on the skepti-
cal form of antirealism.
33 Another aspect of our "practice" concern the metaphysical status of the
propositions from the given domain. Thus even if a antifactualist takes a proposition
to be nonfactual, he may still take it to be a factual matter that it is nonfactual or that
it is contingent or that it is grounded in certain other propositions. The question of
the metaphysical status of these metaphysical claims, though of enormous philo-
sophical interest, is not normally so relevant to adjudicating disputes over the factu-
ality of a given first-order domain.
34 Other philosophers (principally Dummett, passim, and Blackburn [1984],
p.169, and [1998], p.50) have also emphasized the role of practice in adjudicating
realist disputes. What is distinctive about my view is the precise way in which it ar-
ticulates what a practice is and what is involved in accounting for a practice.
35 This is merely a picture. It need not commit one to the view that there are
facts in the world whose structure might correspond to the structure of the proposi-
tions or sentences by which they are described, as in the standard exposition of logi-
cal atomism (Wisdom [1969]).
36 See Williams [1978], p.66, and [1985], p.241.
37 One might express the notion of reality by means of a connective 'it is con-
stitutive of reality that .. .', just as in the case of the notions of factuality and ground.
Considerable interest would then attach to developing the logic of such notions.
38 In making out this case, it is essential to distinguish between ground and re-
duction. If a ground is taken to bring one closer to what is real, then it is hard to see
how there could be an infinite regress of grounds (with nothing at the bottom). For
how can one get closer to what does not exist? But once grounds are taken to be
metaphysically neutral, there is no more difficulty than in the case of cause in con-
ceiving them to form an infinite regress. (Cf. the discussion of Leibniz's position on
this question in Adams [1994], pp.333-8).
39 It is possible to envisage a semi-quietist position that accepts the concept of
factuality but rejects the concept of fundamentality. The world would divide into an
objective and nonobjective part, on this view, but the objective part would be an un-
differentiated mass and could not be meaningfully be taken to possess any particular
intrinsic structure. The above definition would not then be available and the study of
what is factual would have to proceed independently of any consideration of what is
fundamental.
40 I have been very fortunate with the assistance I have received in writing this
paper. Participants of seminars at UCLA and Princeton and of talks at the Universi-
ties of Columbia, Michigan, Pennsylvania, and Rutgers made many excellent com-
ments; I have had valuable conversations or correspondence on the topics of the pa-
per with Rogers Albritton, Paul Boghossian, Ruth Chang, Hartry Field, Mark
Johnston, David Kaplan, David Lewis, Mark Moyer, Gideon Rosen, Stephen Schif-
fer and Bartosz Wieckowski; and, to top it all, Ronald Dworkin, Andre Gallois, Gil
Harman, Paul Horwich, Joshua Schechter and two referees from the journal Imprint
provided me with an extensive array of written comments which were of great help
in revising the paper. This paper first appeared in the electronic journal Philospher's
Imprint (voU, no. 1).
WORDS AND OBJECTS
Achille C. Varzi
1. INTRODUCTION
When we set ourselves to draw up an inventory of the world - a cata-

logue of all there is, was, and will or could be - we have to face two
tasks. First, we have to figure out what sorts of things there are, i.e., we
must identify and characterize the categories under which the items in
the inventory will fall. For example, we might want to draw a distinc-
tion between such things as chairs and tables, on the one hand, and
conferences, hurricanes, and stabbings, on the other. And we may
wonder what to do when it comes to such things (if such there be) as
numbers, jokes, haircuts, smiles, souls, shadows, and so on. The sec-
ond task is to figure out, for each category, how many different things
there are, i.e., how many individual items must be included in that
category. Is this chair the same as the chair that was here yesterday? Is
it something over and above the mereological fusion of the molecules
that constitute it? Am I the same as my body? Is Brutus's stabbing of
Caesar the same event as his killing of Caesar? Is it the same as the
assassination of Caesar? Is it the same as the violent assassination of
Caesar? And so on.
How do philosophers go about addressing these tasks? I think we
can say that two main tools have been developed to provide in each
case, if not a fool-proof algorithm for answering each question, at least
some help. The first is the method of linguistic analysis. Do we want to
49
A. Bottani et al. (eds.), Individuals, Essence and Identity, 49-75.
know what sorts of things there are or could be? Then let's see what
sorts of things there must be in order for what we truthfully say to be
true. This, in tum, requires that we pay due attention to what it is that
we say when we say something. For obviously we do not want to be
misled by the idiosyncrasies of the language we speak. Obviously we
do not want to say that there must be such things as age differences
(for example) just because the English sentence 'There is a difference
in age between John and Tom' is true. We do not want to say that there
exist such things as holes just because the sentence 'There is a hole in
that piece of cheese' is true. First we have to uncover the logical forms
of these sentences. We have to look at the deep, semantic structures
underlying their superficial, grammatical structures, and then we can
look at what sorts of things must exist in order for those semantic
structures to agree with the facts. For example, we can look at the ref-
erents of the logically proper names and at the values of the bound
variables. Do these include ages? Do they include holes? Does the do-
main of reference and quantification underlying our English assertions
include such things in addition to chairs, people, and chunks of
cheese? Does it include conferences, hurricanes, stabbings? Does it
also include numbers? Jokes? Haircuts? What else?
The second tool developed by philosophers is the one known as
Leibniz's law, broadly understood as a principle of substitutivity salva
veritate. Is this chair something over and above the mereological fu-
sion of the molecules that constitute it? Well, let us see - is there any
statement that is true of the chair but not of the mereological fusion, or
vice versa? If there is no such statement - if the chair and the
mereological fusion are indiscernible -then they are one and the same
thing, or so we may suppose. Otherwise they must be two things. Is
Brutus's stabbing of Caesar the same event as his killing of Caesar?
Well, if whatever is true of the stabbing is also true of the killing, and
vice versa, then the answer is: Yes, they are one and the same event;
otherwise the answer is: No. Of course it may actually be impossible to
check every statement, so in actual circumstances it may be impossible
to be sure that we have one thing or one event rather than two or more.
That's the obvious difficulty with one half of Leibniz's law, the one
that asserts the identity of the indiscernibles. (This is why it would be
good to have some identity criteria that are tailor-made for material
Achille Varzi. Words and Objects 51
objects, or for events, or for whatever categories of entities we counte-

nance, for such criteria would allow us to cut down the number of rele-
vant parameters.) However, the other half of the law - the indis-
cernibility of identicals - seems easy. If we do hit upon a suitable
statement that holds true of the chair but not of the mereological fusion
of its molecules, then we can be sure that we are dealing with two dis-
tinct things. If we hit upon a suitable statement that holds true of the
stabbing but not of the killing of Caesar, then we can be sure that we
are dealing with two distinct events. And so on.
All of this is very good, and I think one can hardly underestimate
the significance of these two tools - linguistic analysis in general, and
the principle of substitutivity in particular. But one thing is their sig-
nificance and another thing is their usefulness (ifI may put it this way),
and this is what I want to focus on in this paper. How much help do we
get from those tools when it comes to working out our inventory of the
world, or to convincing others that our inventory is a good one? How
useful are those tools for the working metaphysician? I am afraid the
answers that I can give are not very optimistic. Indeed it will be my
contention that those tools are pretty useless unless we already have a
good idea of what sort of inventory we want to draw up, and of how
we are going to count our items. If we do have such an idea then fine:
we can rely on linguistic analysis and on the principle of substitutivity
to double check our work and to clarify our views. Otherwise we are
stuck - which is to say that we have a lot of honest work to do before
we can find some use for those tools.
2. LOGICAL FORM AND ONTOLOGICAL COMMITMENT
Let me first focus on the business of logical form. The guiding idea,
here, is one that goes back to Frege and especially to Russell. We
know that we must make room for those things whose existence is im-
plied or presupposed by any statement that we can truthfully make. But
ordinary-language sentences may have a deceptive grammatical form
and therefore questions about their aboutness - questions concerning
their ontological commitment, as some like to say - only arise upon
suitable logical analysis. Before knowing what a sentence is about, or

even whether it is about anything at all, we must understand the logical
form of the sentence itself. For only the logical form is ontologically
transparent. The grammatical form is full of ontological traps.
I have already mentioned a couple of examples of the sort of traps
that we should beware of, but let me be more explicit. Take such fa-
miliar cases as Alice's answer to the White King:
(1) I can see nobody on the road.
Or think of Russell's paradigmatic example involving definite de-

scriptions:
(2) The king of France does not exist.
Sentences such as these - we are to understand - are not about the en-
tities they seem to be about. It's not that there is this guy, Mr. Nobody,
whom Alice is seeing ("And at that distance too!", as the King would
remark in a fretful tone.) It is not that there is this fellow, the present
king of France, who does not exist (as Meinong and others supposedly
held). To think so would be to be misled or deceived by superficial
grammatical features. It would be to treat as designators expressions
which are not, in fact, meant to designate anything at all. Deep down
such sentences have an entirely different form and that is the form that
matters when it comes to assessing their ontological import - or so the
story goes. The first sentence is the negation of an existential percep-
tual report:
(1') It is not the case that there is at least one person whom
I can see on the road.
The second amounts to the negation of a statement of existence and

uniqueness:
(2') It is not the case that there is one and only one king of
France.
Now, the traditional wisdom is that precisely this sort of consideration
is or should be at work when it comes to any sort of existential claim,
including assertions that involve an explicit existential quantifier of the
Achille Varzi, Words and Objects 53
fonn 'There is a P' or 'There are Ps'. The first example I mentioned
earlier was from Morton White:
(3) There is a difference in age between John and Tom.
Let us suppose that this sentence is true. Do we want to say that this
fact requires the existence of a suitable entity satisfying the existential
quantifier-the existence of an age difference? Of course not. That
would be falling in a trap. Here is what we are supposed to do instead:
We might begin by saying that we understand the relational predicate 'is
as old as' and that we test statements of the form 'x is as old asy' without
having to see that x has some queer thing called an age, thaty has one, and
that these ages are identical. In that event, the belief of the ordinary man
that there is a difference in age between John and Tom would be rendered
in language that is not misleading by saying instead, simply, 'It is not the
case that John is as old as Tom' [ ... ] We need not assert the existence of
age differences [ ... ] in communicating what we want to communicate.
(White [1956], pp.68-69)
In other words, when dealing with a sentence such as (3) we should

first of all recognize that the statement it makes can be rephrased more
perspicuously as
(3') It is not the case that John is as old as Tom.
And this statement carries no commitment whatsoever to a category of
individual ages. Likewise, if I say
(4) There is a hole in this piece of cheese
I need not be taken to be asserting something that commits me to the
existence of a truly immaterial entity - a hole - located in this piece of
cheese. The existential phrase 'There is a hole in' is misleading, or so
people have been telling me. Here is how things are explained by the
nominalist materialist featured in David and Stephanie Lewis's classic
dialogue:
I did say that there are holes in the cheese; but that is not to imply that
there are holes. [ ... ] When I say there are holes in something, I mean
nothing more nor less than that it is perforated. The synonymous shape-
predicates' ... is perforated' and 'there are holes in ... ' - just like any
other shape-predicate, say' ... is a dodecahedron' - may truly be predi-
cated of pieces of cheese, without any implication that perforation is due
to the presence of occult, immaterial entities. (Lewis and Lewis [1970],
p.4)
So we can truly assert sentence (4). But the underlying logical form is
not an existential statement but a simple subject-predicate statement:
(4') This piece of cheese is perforated.
Here are some more examples of logical analyses of this sort, taken
somewhat randomly from the literature:
(5) There is a strong chance that Professor Moriarty will

come.
(5') It's very likely that Professor Moriarty will come. 1
(6) Sue was dancing a waltz.

(6') Sue was dancing waltzly?
(7) The average star has 2.4 planets.

(7') There are 12 planets and 5 stars, or 24 planets and 10
stars, or ... 3
(8) This tomato and that fire engine have the same color.
(8') This tomato and that fire engine agree colorwise. 4
(9) There are many virtues which Tom lacks.

(9') Tom might conceivably be much more virtuous than
he is.s
Nor are these the only sort of cases that philosophers have been wor-
rying about. All of these are examples that illustrate an eliminativist
strategy: they show that we can analyze sentences which seem to in-
volve ontological commitment to certain entities as expressing propo-
sitions that are, in fact, ontologically neutral with respect to those enti-
ties. But there are also cases where the analysis goes in the opposite
direction, i.e., cases where the logical form discloses a hidden quanti-
fier, thereby introducing ontological commitments that do not appear

at the level of surface grammar. Davidson's account of the logical
form of action sentences is a good example of this introductionist
strategy. If I say
(10) Brutus stabbed Caesar with a knife
I am not just talking about Brutus, Caesar, and a knife (says Davidson).
I am not just saying that these three entities stand in a certain three-
place relation, x stabbed y with z, for otherwise I could not explain why
my statement logically implies
(11) Brutus stabbed Caesar
(a statement that would involve a different, two-place relation). Rather,
for Davidson (10) and (11) are to be understood as statements about a
certain event - a certain stabbing that took place a long time ago. Deep
down they are supposed to have the following forms:
(10') There was a stabbing by Brutus of Caesar, and it was
done with a knife.
(11 ') There was a stabbing by Brutus of Caesar.
And once we see that these are the statements corresponding to the
sentences in (10) and (11), the entailment is logically straightforward.
In Davidson's words:
There is, of course, no variable poliadicity. The problem is solved in the
natural way, by introducing events as entities about which an indefinite
number of things can be said. (Davidson [1967], pp.116-117)
3. IS THIS OF ANY HELP?
All of this is standard lore. Some of these analyses are so naturally ac-
cepted that they are now found in introductory textbooks in logic. This
is true of Russell's analysis of definite descriptions but also, for exam-
ple, of Davidson's analysis of action sentences. 6 Still, I think that there
are severe problems with this general picture. Some of these problems
are well known but some are less obvious and, I think, rather worri-
some, especially if we keep in mind the tasks of the working metaphy-
SICIan,
One obvious problem is that we can hardly eliminate or introduce
anything by mere armchair speculation. Philosophers do speak as
though we could banish entities from existence just by helping our-
selves with Occam's razor, or bring entities into existence just by
adding some quantifiers. But this is eerie and we should better take it
as a jat:;on de parler. All we can do is to show how the existence or
non-existence of certain entities would allow us to explain certain
facts, just as scientists sometimes posit the existence or non-existence
of certain entities in order to explain certain natural phenomena. In any
case, it is obvious that logical analysis per se can do very little. Para-
phrasability of sentences about holes does not per se eliminate holes
from the world just as the assertibility of such sentences does not
automatically introduce holes. Paraphrasability is a necessary condition
if we want to avoid commitment to such things, and assertibility is a
sufficient condition if we want to proclaim commitment, but neither is
necessary or sufficient to affect the ontology itself.?
A second obvious problem concerns the very idea that the logical
form of a sentence may allow us to withhold our commitments while
still communicating what we want to communicate. What test can we
apply to see whether a given English sentence can be understood as
having a certain logical form? How do we know, for example, whether
the logical form of a sentence such as (3) is correctly represented by
(3'), where there is no mention of age differences? Pretty clearly, if we
want to use (3') to communicate what we would be communicating
using (3), then (3') must express the same proposition as (3). It must be
the case that in uttering (3') we would be making the same assertion as
we would make if we uttered (3). Otherwise (3') would not represent a
legitimate analysis of (3). But then, in uttering (3') we would be talking
about the very same things we would be talking about if we uttered (3).
And why should one utterance be better than the other? Why should
(3') be ontologically more transparent than (3)? As William Alston
pointed out a while ago, this bears more than a passing resemblance to
the paradox of analysis:
In any context where questions of existence arise the problem is whether

or not we shaH assert that so-and-so exists, not whether we shaH choose
some particular way of making this assertion. (Alston [1958], p.50)
We can also put the problem as follows. Whether or not the truth of
our statement implies (or presupposes) the existence of age differences
does not depend on the words that we utter to make that statement - it
doesn't depend on the sentence that we use. So let us assume that sen-
tence (3') can be used to make the same statement as sentence (3),
though in a way that does not mention age differences explicitly. Then
we may as well say that (3) itself is a sentence that can be used to make
the same statement as (3 '), though in a way that does mention such
things explicitly. So from left to right (so to say) the analysis results in
an elimination; but from right to left it results in an introduction. How
do we choose?
Let me elaborate. 8 The idea behind the use of logical analysis is
that in order to assess the ontological commitment of ordinary-
language sentences one must first provide suitable logical paraphrases
that are "intrinsically non-misleading" (as Ryle put ie) and therefore
ontologically transparent. This amounts to a sort of linguistic recon-
structivism: the truth conditions of our sentences are determined by the
truth conditions of the corresponding paraphrases and do not therefore
require an independent ontology. Very well. The question is: Where do
these paraphrases come from? On what grounds should we look for the
logical forms that underlie our ordinary statements? Plainly, the very
issue of which sentences must be logically paraphrased - let alone how
they ought to be paraphrased - can only be addressed against the back-
ground of one's own philosophical inclinations. When Russell, for ex-
ample, says that (2) must be paraphrased as (2') it is because Russell
holds that the former, as it stands, is incompatible with our sense of
reality, with that robust "feeling for reality which ought to be preserved
even in the most abstract studies"lo (and which lies behind that
"aesthetic sense of us who have a taste for desert landscapes"ll, as
Quine later put it). The analysis yields no ontological discovery. It is
Russell's own ontological convictions that lead him through the quest
for an appropriate logical form for (2), not vice versa. So much so that
a philosopher such as Meinong might feel no need to take any action.
For him the grammatical fonn of (2) may well coincide with its logical
fonn because for him the present king of France does have ontological
dignity. It is just one among many characters that a complete inventory
of the world should include. 12
If things are so, however, then here is the problem: How shall we
go about detennining whether or not a given sentence can be taken at
face value? How do we know whether or not it needs to be analyzed or
rephrased before we can look at its ontological import? In the case of a
sentence such as (3) we may be inclined to look for a paraphrase that
avoids any reference to age differences because these would be "queer
things" (in White's own tenninology) to be included in our inventory.
But if we thought that such things are not queer, then we wouldn't
need any paraphrase. We could still accept the analysis but we might
be more inclined to read it from right to left and observe that whoever
says that John is not as old as Tom is actually asserting the existence of
an age difference between them.
For another example, let's go back to the holes in the cheese. Ifwe
are inclined to rewrite (4) as (4') it is because our strong "feeling for
reality" prevents us from taking this to be a statement about a hole: if
holes are not real then our statement can only be about the cheese. This
is understandable and may even justify the enonnous amount of work
that will be involved in analyzing every natural-language sentence that
seems to refer to or quantify over holes as expressing a proposition
which only involves reference to or quantification over perforated ob-
jects. Consider, for instance:
(12) There are seven holes in that piece of cheese.

(13) One hole in that piece of cheese is shaped like a
doughnut.
(14) There are as many trefoil-knotted, doughnut-shaped
interior holes in that piece of cheese as there are
cookies on your plate.
I am happy to assume that we could paraphrase these sentences as well

as every other.13 But what if! like holes (so to say)? What if! think that
a hole in the cheese is just a proper undetached part of the cheese's
complement, just as the crust of the cheese is a proper undetached part
of the cheese itself? Shall I still regard the paraphrases of these sen-
tences as expressing their logical form? Perhaps I should do exactly

the opposite. I should say that it is because there is a hole in it that the
piece of cheese is perforated, in which case it would be (4) that sup-
plies an "ontologically transparent" paraphrase of (4'), not vice versa.
Likewise in all other cases. If I like holes then there is no reason for me
to take the paraphrases into serious consideration - on the contrary. So
here is the impasse: On the one hand we have the hole-enemies, who
warn us against the dangers of ontological hallucination: We may have
the wrong impression of seeing holes where in fact there is nothing at
all. On the other hand we have the hole-realists, who warn us against
the danger of ontological myopia: Holes are ephemeral entities but
they are real nonetheless, and we should not pretend that they are not.
Does linguistic analysis help us in making up our mind?
Historically this tension has been particularly manifest in the case
of events. The standard, Davidsonian analysis is that a sentence such as
(10) should be analyzed as (10'), i.e., as involving quantification over
events. But there are also philosophers who view things the other way
around; for those philosophers it is (10) that provides the logical form
for (10') and there is no need whatsoever to posit the existence of an
event - a stabbing - which had Brutus as an agent and Caesar as a pa-
tient. 14 The reason for this different conception is that such philoso-
phers do not think that events are entities of a kind, so for them the
sentences in question cannot really be about stabbings. They are about
Brutus, about Caesar, and perhaps about a knife. And if this gives rise
to complications when it comes to explaining the logical validity of
certain inferential patterns (such as the adverb-dropping inference from
(10) to (11)), that simply means that we have a problem to solve. For
example, we need a logic of adverbs. But we would seem to have that
problem anyway, wouldn't we? For instance, we need a logic of ad-
verbs to explain the inference from (15) to (16):
(15) This mathematical series converges slowly.
(16) This mathematical series converges.
(Or do we want to say that these sentences are about an event of con-
vergence?IS) So here we are again. On the one hand, the logical form of
a simple atomic sentence is said to involve hidden existential quantifi-
cation over events. On the other hand, the situation is reversed and
what looks like a quantification is explained away by means of a sim-

ple atomic sentence, just as what looks like a quantification over holes
in (4) is explained away by means of a simple atomic sentence in (4').
How do we choose?
We can also get perverse results once we start playing this game.
For example, suppose you are a Davidsonian about action sentences. In
fact, suppose you think that stative sentences deserve a similar treat-
ment, as Terence Parsons has suggested, so that to explain the valid
inference from (17) to (18):
(17) John loves Mary passionately
(18) John loves Mary
we would have to understand these sentences as involving hidden
quantification over individual states: 16
(17') There is a loving state in which John is with respect to
Mary, and it is passionate
(18') There is a loving state in which John is with respect to
Mary.
Suppose, on the other hand, that your robust feeling for reality makes
you an eliminativist with regard to holes. Then you might want to say
that a sentence such as
(4) There is a hole in this piece of cheese
really has the form
(4') This piece of cheese is perforated
which really has the form
(4") There is a holey state in which this piece of cheese is.
Mirabile dictu, what seemed to be a quantification over holes in the
cheese turns into a quantification over states in which the cheese is.
This may well be fine. But the question of whether (4") really repre-
sents the logical form of (4), rather than vice versa, seems to me to be
entirely up for grabs. Let us just say that depending on what we think
there is we attach a meaning to what we say. Let us theorize explicitly
about what there is rather than attribute our views to the language that
we speak, and hence to the speakers who share our language. What
would entitle us to do that?
4. REVOLUTION AND INTERPRETATION
Here is another way of pressing this point. As far as ordinary practice

goes, the only way one could ultimately evaluate the success of a logi-
cal analysis is by testing it against our pre-analytical intuitions - by
comparing it with our understanding of the original sentence. How else
could we determine whether the analysis is acceptable? However, this
means that in order to analyze and eventually paraphrase a sentence it
is first necessary to understand it. We must attach a meaning to the
original sentence prior to the analysis. And how can we do that without
the background of a corresponding ontology?
In their recent book on nominalism in mathematics, John Burgess
and Gideon Rosen distinguish two ways in which the link between a
sentence A and its "transparent" paraphrase A' can be understood. 17 The
first is what they call the hermeneutic understanding. This is basically
what Russell and Davidson (and many others) have in mind when they
propose their logical analyses of certain types of natural-language sen-
tences. The analyses uncovers the deep structure of those sentences - it
reveals the truth conditions of the analysanda, those conditions which
are supposed to take us straight to the truth makers. The second way to
understand a paraphrase is what Burgess and Rosen call the revolu-
tionary way. On this understanding, the paraphrase or logical form is a
genuine revision of the given sentence. This is not what Russell and
Davidson have in mind but it is, for example, what Quine had in mind.
In Word and Object (section 33: "Aims and Claims of Regimentation")
Quine adamantly insists that a logical paraphrase does not reveal the
meaning of a sentence but changes it. Ultimately the purpose of a
paraphrase is to resolve ambiguity. And
if we paraphrase a sentence to resolve ambiguity, what we seek is not a
synonymous sentence, but one that is more informative by dint of resisting
some alternative interpretations. (Quine [1960], p.l59)18
I think this is also what motivates the linguistic analyses of philoso-

phers who find themselves on the business of massive paraphrasing,
such as the Lewisean nominalist mentioned earlier. Let me quote
agam:
When I say there are holes in something, I mean nothing more nor less
than that it is perforated. [ ... ] I am sorry my innocent predicate confuses
you by sounding like an idiom of existential quantification. [ ... ] But I have
my reasons. You, given a perforated piece of cheese [ ... ] employ an idiom
of existential quantification to say falsely 'There are holes in it.' Agree-
able fellow that I am, I wish to have a sentence that sounds like yours and
that is true exactly when you falsely suppose your existential quantifica-
tion over immaterial things to be true. That way we could talk about the
cheese without philosophizing. [ ... ] You and I would understand our sen-
tences differently, but the difference would not interfere with our conver-
sation until you start drawing conclusions which follow from your false
sentence but not from my homonymous true sentence. (Lewis and Lewis
[1970], p.4)
Thus, the revolutionary analysis is not meant as an ontologically trans-

parent paraphrase of what a given sentence really means. It is an analy-
sis of what the revolutionist means when she uses that sentence. The
sentence as such can be used to mean different things by different
speakers and the revolutionist is urging us to follow her practice. She
is not interested in understanding language. To the contrary her mani-
festo reads: Philosophers have hitherto tried to understand language;
now it's time to change it.
Now my point is that revolutionary paraphrases are perfectly all
right, but they don't play any role in our metaphysical investigations.
They play no role because they presuppose that we already have a
cause to fight for - that we already have a view about the way things
are. We just want to make sure that people don't draw the wrong infer-
ences from what we say, so we provide (only upon request, perhaps) all
the necessary linguistic amendments. On the other hand, the herme-
neutic paraphrases could be of great help, because they could be truly
revealing; yet it is very unclear where we can look for the relevant evi-
dence. In fact, it is not even clear whether there can be any evidence, or
whether the hermeneutic approach delivers a picture of natural lan-
guage that is intelligible at all. For the picture would be this: our daily
language - the language that we have learned and made ours since our
very first contacts with the surrounding world - that language would
consist of sentences whose real meaning often eschews us. It would
consist of sentences most of which are only acceptable as loose talk. It
would at best qualify as a sort of metalanguage with regard to the
regimented language of philosophy, the latter being the only genuine
object language and thus the only language that can express our genu-
ine ontological commitments. Is this an acceptable picture?
We thus come to what I regard as the main problem with the
whole idea of ontological transparency. As it turns out, both strategies
involve a duplication of languages. For neither is willing to give up
natural language altogether. Whether you are a revolutionist or a her-
meneuticist, you want to carry on speaking with the vulgar, hence you
are going to emphasize the pragmatic indispensability of ordinary lan-
guage against the philosophical value of the regimented language (on-
tologically impeccable but practically unspeakable). However, this du-
plication of languages only works fine for the revolutionist. For only
the revolutionist is always in a position to tell which language is being
spoken.
Take the Lewisean hole-e1iminativist once again. When speaking
with the vulgar she can give expression in English to the fact that some
cheese is perforated by asserting (4), but when speaking the regi-
mented language of philosophy she would assert the negation of (4).
More generally, she can assent to (19) when speaking loosely, and to
(20) when speaking strictly and literally:
(19) There are holes
(20) There are no holes.
This may be confusing to some people but the revolutionist will always
know when is when, and she will be happy to explain. Not so for the
hermeneuticist. If you are a hermeneuticist you do not have the same
leeway. To the extent that (19) is to be interpreted as (19'), (20) will
have to be interpreted as (20'):
(19') Something is perforated
(20') Nothing is perforated.
This is because (20) is just the negation of (19), so the paraphrase of

one must be the negation of the paraphrase of the other. But this is bi-
zarre. After all, if one thinks that holes do not exist, then (20) seems to
be a perfect way of expressing that view. Yet its paraphrase (20') is
plainly false. So the only way out for the hermeneuticist would be to
say that (20), unlike (19), is to be taken strictly and literally. Unlike
(19), (20) is not to be paraphrased. But this, too, is very bizarre. For
then the distinction between grammatical form and logical form be-
comes utterly arbitrary, and there appears to be no principled way of
discriminating the loosely true from the strictly false.
From this perspective, the situation is not different from what hap-
pens when linguistic revisions take place in science. To use an analogy
suggested by Peter van Inwagen,19 suppose we hear a Copernican as-
tronomer say something like
(21) It was cooler in the park after the sun had moved be-
hind the elms.
Would this be incoherent with the speaker's official view to the effect
that the sun does not move in the sky? Of course it would not. For our
astronomer would hasten to add that in uttering (21) she was speaking
loosely. If necessary, she could be more accurate and she would utter a
sentence that does not suggest that the sun has actually moved - for
example:
(21') It was cooler in the park when, as a result of the

earth's rotation, the elms ended up being in front of
the sun
(or something much more awkward than this). She can do that and she
knows how to do it because she is a revolutionist; when she uses cer-
tain sentences of English she actually means something that goes be-
yond the literal or customary meaning of her words. (She means to ex-
press propositions that have a different form, if you like.) Now sup-
pose I utter a sentence such as (21). Am I to be taken to imply that the
sun has moved? Well - no. I suppose in this case you are entitled to
reinterpret my statement too in a way that makes it consistent with the
heliocentric theory: you may be charitable. ("Achille Varzi couldn't
possibly mean to say that the sun had moved! He must have been
speaking loosely.") But that is only because you take me to be part of

the gang. You assume that I have myself subscribed to the revolution,
and that assumption justifies your hermeneutic attitude. I do have my
Copernican views on astronomy, and you know that, and you also
know that if the need arises I can be more precise on the basis of that
theory. You know this so well that you can take care of that on my be-
half. But metaphysics is not like physics, and when it comes to meta-
physics you can hardly base your interpretation of what I say (or what
anybody says) on the basis of the principle of charity. Ordinary speak-
ers do not need to be astronomy experts to know that the sun does not
move and the hermeneuticist may rely on this fact. But most people
who assert common-sense sentences about holes, or about events, or
about other "queer entities", are totally unaware of any metaphysical
theories about such things (if such there be). So how should one rein-
terpret those assertions? "The speaker couldn't possibly mean to say
that there is a hole in this piece of cheese! She was speaking loosely.
She meant to say that the cheese is perforated." Is this legitimate? Is it
not a biased interpretation? Or consider van Inwagen's own form of
linguistic revisionism. Strictly speaking, a sentence such as
(22) There is a table in the kitchen
should be understood as expressing the following proposition:
(22') There are xs in the kitchen, and such xs are arranged
tablewise.
This is because for van Inwagen tables do not exist; a thing would
have to possess certain properties in order to be properly called a 'ta-
ble', and nothing has those properties. I take it to be obvious that van
Inwagen is a revolutionist, and that is perfectly fine. But now suppose
that someone else utters sentence (22). "The speaker couldn't possibly
mean to say that there is a table in the kitchen! She must have been
speaking loosely. She meant to say that there is stuff arranged ta-
blewise in the kitchen." What sort of charitable reading is this? This is
plain misconstrual.
5. LEIBNIZ'S LAW AND THE COUNTING PROBLEM
So much for this part of the story - the tool of linguistic analysis. The
only way I can make sense of it is as an honest revolutionary tool. But
revolutions cannot be improvised and I would not engage in one unless
I had already sorted out my views in advance. Linguistic analysis can
be useful as a tool to clarify what I mean when I use certain sentences
- or what we all should mean - but not what those sentences must
mean. Hence it supplies no shortcut to metaphysical investigation. In
particular, linguistic analysis is of little guidance when it comes to the
first important task involved in the drawing up of an inventory of the
world - that of figuring out the ontological categories under which the
items in the inventory should fall. Let me now move on to the second
task - that of figuring out a way of counting the items in each category.
As I mentioned, here it is customary to rely on the general tool pro-
vided by Leibniz's law, broadly understood as a principle of substitu-
tivity salva veri/ate. If we are smart, for some categories we may have
concocted some kind of identity criterion, but generally speaking the
principle of substitutivity supplies at least a negative test for identity: If
we hit upon a statement that holds true of something x but not of
something y, then we can be sure that x and y are distinct.
Also in this case, I am afraid I have mostly negative things to say.
It seems to me that the basic intuition behind this strategy is seriously
flawed and that it only succeeds in raising a dust that obstructs the real
difficulty involved in our metaphysical task. To make my case, let me
briefly review some concrete examples of how this strategy is typically
implemented.
Consider again this chair in front of me and the mereological fu-
sion of the molecules that constitute it. Are they the same thing? Well,
it seems plausible to suppose that the chair would survive the annihi-
lation of a single molecule. But - the argument goes - surely the
mereological fusion would not: That fusion of molecules must consist
of those very molecules, it must include them by definition.20 Hence
the chair and the fusion of its molecules have different properties (dif-
ferent modal properties) and should be distinguished by Leibniz's law.
For a second, standard example, consider a statue and the lump of clay
that constitutes it: two things or just one? Well- the argument goes-
the artist made the statue this morning. But the clay was already there
yesterday. So the statue and the clay came into existence at different
times. Hence they have different properties (different temporal proper-
ties, in this case) and should be distinguished. The same line of argu-
ment is familiar also from the literature on events. Take Brutus's stab-
bing of Caesar. Is it the same event as Brutus's killing of Caesar?
Well, it seems reasonable to suppose that Caesar could have survived
the stabbing. But surely Caesar could not have survived his very kill-
ing. So, once again, an appeal to Leibniz's law would allow us to con-
clude that the stabbing and the killing are distinct events. And so on
and so forth. This line of argument is very popular and very pervasive
indeed?1 But is it legitimate?
Let us focus on one instance - say the chair and the mereological
fusion of the molecules that compose it, Xl ... X n . In that case the argu-
ment has the following structure: We come up with a statement which
is true of the chair but false of the fusion, and we conclude that the
statement must be about two things. Schematically:
(23) The chair in front of me could survive the annihilation
of molecule Xi.
(24) The fusion of molecules Xl ••. xn could not survive the
annihilation of molecule Xi.
Hence
(25) The chair in front of me is not the fusion of molecules
Xl .. · X n•
Of course, if this argument is accepted, then by the same pattern we

could also distinguish many other entities occupying the same region
of space in front of me: The mereological sum of molecule Xl plus the
rest of the chair, the mereological sum of molecule X2 plus the rest of
the chair, and so on. There would really be lots of entities in that re-
gion of space, not just one or two. But we need not go into this com-
plication now. Let us just ask: Is the argument above a good one?
Well, are the premises true? Obviously this depends on how we
read them. And there are two different ways of reading the premises,
depending on whether the terms occurring therein are understood de
dicto or de reo On a de dicto reading both premises are clearly true:
(23') There is a possible world w such that the thing which

is the chair in front of me in w lacks molecule Xi.
(24') There is no possible world w such that the thing which
is the fusion of molecules Xl .•. Xn in w lacks molecule
To deny the first would amount to making very strange assumptions

about what worlds are possible and what are not: there is nothing about
molecule Xi which makes it necessary for it to belong to whatever chair
is in front of me in every world w. As for the second premise, it simply
reflects the meaning of the term 'fusion of molecules XI ..• Xn', so it is
indeed true "by definition". On a de dicto reading, both premises are
therefore true. However, this is obviously beside the point. If we are
interested in the modal properties of the entity or entities that are in
front of me in the actual world, then we should not look at the possible
referents of our terms, 'the chair in front of me' and 'the fusion of
molecules XI . • . xn'. Plainly, if these terms have different senses (as
they do), then they may have different referents in different worlds.
But that is not the issue. The issue is not whether our terms could have
different referents. It is whether they do have different referents,
whether they have different referents in this world. And of course this
is not an issue that we can solve by looking at their senses. That would
be a well-known fallacy.
So it is the de re reading that matters if we want to apply Leibniz's
law. On that reading the argument is valid. But on that reading the
truth conditions of the two premises (23) and (24) are hardly obvious:
(23") The chair in front of me - that particular entity - is

such that there is a possible world w in which it lacks
molecule Xi.
(24") The fusion of molecules XI ... Xn - that particular en-
tity - is such that there is no possible world w m
which it lacks molecule Xi.
If the chair is not the same as the fusion of its molecules, then fine: We
are talking about two different entities and perhaps we can say that
(23") and (24") are both true. 22 Perhaps molecule Xi is an essential part
of the fusion but not of the chair. But this opposition would be prior to
our thought experiment - it cannot be inferred from it and calls for in-
dependent grounds. How could we have de re intuitions about the chair
and about the fusion if we didn't even know whether they are one thing
or two? How could we compare their properties if we didn't know
what they are - if we didn't even know whether they are distinct? Be-
sides, why should we be able to settle identity issues in this world by
looking at what goes on in other worlds? Don't we need to know how
many passengers we are bringing along before we can embark in other-
worldly philosophical excursions? On the other hand, if the chair is the
same thing as the fusion of its molecules - and to rule that out would
be to beg the question - then that particular entity is the same in both
cases, so (23") and (24") cannot be both true. And which one of them is
false is a genuine metaphysical question: maybe molecule Xi is an es-
sential part of that entity, in which case (23") would be false; maybe it
is not an essential part, in which case it is (24") that would be false. (Of
course, it would then be awkward to assert the true sentence of which
(24") is the negation, i.e., to say that the fusion of Xl ... Xn could survive
the annihilation of Xi. It would be better to make the same statement by
asserting (23"). But that awkwardness is a heritage of our inclination to
oscillate between de dicto and de re readings: It is not a falsehood in-
dicator and it is only relevant from a pragmatic perspective.) It does
not matter now which premise is false. As I said, that would be a
genuine metaphysical issue. As far as the argument goes the point is
that we cannot simply assume that both premises are true on pain of
begging the question. We can have a priori reasons to accept both
premises only if we already have reasons to distinguish between the
chair and the fusion of its molecules in the first place, and that is sup-
posed to be the conclusion of the argument. As it stands, on a de re
reading (the only one that makes the argument valid) the argument is
either unsound or viciously circular. Hence it is useless.
The same diagnosis applies to the other non-identity arguments
mentioned abov<:(, as well as to other statements along the same lines.
In each case, I submit, the reasoning is either invalid (if read de dicto)
or question begging (if read de re). Thus, surely the tenns 'the statue'
and 'the lump of clay' have different senses; but it does not follow that
they have different referents. And if their referent is the same, if they
name the same thing, then either that thing was already there yesterday
or it was not. (If it was, of course it would have been awkward to refer
to it as a statue prior to this morning, when the artist actually shaped it
as a statue; but that awkwardness would be purely linguistic or ideo-
logical and would have no bearing on the ontological level.) Or take
the events, Surely the predicates 'stabbing' and 'killing' have different
intensions, which entails that the event descriptions 'Brutus's stabbing
of Caesar' and 'Brutus's killing of Caesar' have different senses. It
does not, however, follow that the two descriptions have different ref-
erents. And if their referents are the same, if they describe the same
event, then either that event could have been survived by Caesar or it
could not. (If it could, it would of course be awkward to refer to it as a
killing when counterfactualizing about Caesar's survival. We should
rather refer to it as a stabbing. But once again that awkwardness would
be purely linguistic, or ideological, and would have no bearing on the
ontological level.) And so on and so forth. To be sure, there is a differ-
ence between this last case - the event descriptions - and the cases of
the chair and the statue. Perhaps Brutus's stabbing of Caesar was his
killing of Caesar, but there are lots of other stabbings that are not kill-
ings. On the other hand, if this chair in front of me is the same as the
fusion of its molecules, then that chair next to you is also the same as
the fusion of its own molecules. Every chair must be the same as the
fusion of its own molecules, or else no chair is. Likewise, either every
statue is identical with the lump of matter that constitutes it, or else no
statue is. I take these to be important metaphysical tenets. But having
said this, the trouble with the argument is the same in all cases.
Let me stress also that the trouble concerns the form of the argu-
ment, not the conclusion itself: whether the entities in question are one
or two remains open. Moreover, the analysis is not quite neutral with
respect to the issue of contingent identity. If you think that a chair can
be identical with the fusion of its parts in some worlds but not in others
(or at some times but not at others), then the objection would not quite
apply.23 In that case we could speculate about the modal or temporal
properties of the chair and of the fusion, and perhaps we could dis-
cover that these properties are distinct without begging the question of
whether the chair and the fusion are in fact distinct. However in that
case it would remain to be shown how we can use Leibniz's law to go
from the observation that there are worlds in which the chair and the
fusion have different properties to the observation that the actual chair
and the actual fusion have different properties - that is, different modal
properties. And unless we can do that the argument, though formally
non vicious, would still be pretty useless. Ditto in the other cases.
Now, I do not want to insist too much on this analysis here. 24 In its
general form it goes back to a point that Dummett made several years
ago in his first book on Frege, where he observed that whether or not
Leibniz's law can be used as a definition of identity, it cannot be made
to serve as a criterion for deciding the truth of identity statements.25
More recently, the same sort of consideration was put forward by Mi-
chael Della Rocca in the context of his discussion of essentialism and
by Stephen Neale in his analysis of event-referring descriptions. 26 It is
an obvious point, uncharitable as it may sound towards those philoso-
phers who do engage in non-identity arguments of this sort. In fact, I
do not even want to claim that Leibniz's law can never be used to pro-
vide grounds for a non-identity statement. I suppose that there are cir-
cumstances where one may find grounds for recognizing that a predi-
cate which is true of the bearer of one term is false of the bearer of an-
other term prior to any decision concerning whether the terms have the
same bearer. For example, I believe I have sufficient grounds for
making the statement
(26) Professor Bottani is sitting in this room.
Now we hear a scream coming from the room next door and I feel en-
titled to assert:
(27) The person who just screamed is not sitting in this
room.
From these two statements, together with the assumption that people
cannot be in two places at the same time, I can conclude that Professor
Bottani is not the person who screamed. Likewise, the truth of (26) to-
gether with that of
(28) Professor Bottani is not wearing a hat
is sufficient ground for me to conclude that, given the axioms of set
theory, the set of people sitting in this room is not the same as the set
of people wearing hats. All of this is fine. My point is, rather, that the
sort of evidence we need to rely on in cases such as these, where the
application of Leibniz's law is not question-begging, stems from a
complex web of factors. Among other things, the evidence builds on a
background theory about people and about sets. It certainly does not lie
in linguistic intuitions about the truth values of the relevant English
sentences, and perhaps that is why the non-identities that we can infer
come as no surprise. By contrast, it is typically linguistic intuitions that
seem to underlie ordinary non-identity arguments about such things as
chairs, mereological fusions, statues, and the like. And these intuitions
- I submit - are not warranted except on a de dicto reading. Hence they
are not warranted if we intend to apply Leibniz's law.
Some will reply that this is precisely the point to be stressed:
somehow there is a theory behind the claim that the premises of a non-
identity argument are true. It is not just linguistic intuitions. It is intui-
tions correlating the sortal terms that we use to talk about the entities at
issue - the sortals 'chair', 'statue', 'lump of clay', etc. - and the iden-
tity conditions that the theory associates with the entities falling under
those sortals, exactly as with people and sets. On this view, every sor-
tal comes with its identity and persistence conditions built in. And it is
such identity and persistence conditions that we rely on when it comes
to assessing the premises of a non-identity argument.
I do not intend to deny this. To the contrary I find this view per-
fectly reasonable. But it seems to me that the view is only acceptable
provided that the background theory is taken for what it is - a genuine
piece of metaphysics. If we want to say that a chair and a fusion of
molecules have different persistence conditions - that there are differ-
ent kinds of change that can and cannot be survived by the chair and by
the fusion, respectively - then we must do this properly. We must say
that it is a matter of what kind of thing a chair is, and what kind of
thing a mereological fusion is. We must do that before resorting to
Leibniz's law. And we can hardly do that simply by looking at our lin-
guistic practices and intuitions concerning the sortal terms 'chair' and
'fusion of molecules'. So, for example, David Wiggins and Jonathan
Lowe have famously argued that every individual is necessarily an in-
dividual of a kind, or sort, and that the kind or sort of thing that an in-
dividual is determines its identity and persistence conditions. Moreo-
ver, one fully grasps the nature of a given kind or sort of thing, and
hence the sense of a sortal tenn designating it (or even the sense of a
singular tenn that can be used to refer to a specimen of it), only when
one grasps the associated criterion of identity. In the cases under dis-
cussion, these philosophers would say for instance that the sortal tenns
'chair' and 'fusion of molecules'
have different criteria of identity associated with them, and [ ... ] no indi-
vidual of a sort <jl can intelligibly be said also to belong to a sort \jI if <jl and
\jI have different criteria of identity. (Lowe [1989], p.70)27
Very well. I have no objections against this. But that is precisely be-
cause I do not think that our linguistic practices and intuitions con-
cerning the sortal tenns 'chair' and 'fusion of molecules' provide any
evidence for or against the idea that these tenns have different criteria
of identity associated with them. Some philosophers think they do;
other philosophers do not.
In the tenninology introduced earlier, we can also say that the link
between a sortal tenn and its identity criterion - if we want to insist on
it - can only be viewed as part of a revolutionary analysis, not as a
piece of natural language henneneutics. Or do we really think that
there is something about our use of the word 'chair' that detennines
whether this word picks out entities that differ from the fusions of their
constitutive molecules? Do we really think that there is something
about our use of the word 'statue' that detennines whether this word
picks out entities that differ from the lumps of stuff that constitute
them (or whether it picks out three-dimensional rather than four-
dimensional entities, for that matter)? Answering in the affinnative
would, I think, amount to assigning the words of our language a meta-
physical strength that they do not and cannot have. Exactly as with the
case of logical fonn, rather than directly theorizing our ideas about
what there is and how things are we would surreptitiously attribute our
ideas to the language we share with others. We would surreptitiously
attribute them to all the speakers of our language even if our ontologi-
cal intuitions are very different from those of others. And this is the
step that I find illegitimate.
6. CONCLUSION
I realize that the picture that I have presented is mostly a negative one.
If we are interested in the tools of metaphysics - the tools that can help
us draw up an inventory of the world - then I have only offered skepti-
cal arguments against some common practices. Still, I don't intend this
to be exclusively a negative picture. To the contrary, all of this sug-
gests that we have to take metaphysics seriously, since we cannot hope
to derive it from our linguistic practices. And to me this is a positive
thing. This is the beginning of the book, so I thought I could take the
liberty of looking at the dark side first. By the end of the book this fog
of negativity will have dispelled and most of the problems - most of
the genuinely metaphysical problems - will (I am sure) have been
solved.
Notes
1 Burgess and Rosen [1997], pp.222-233.
2 Ducasse [1942], p.233.
3 Melia [1995], p.224.
4 Loux [1998], pp.66-67.
5 Alston [1958], p.47.

6 For example Forbes [1994].
7 One author who has emphasized this sort of skepticism towards the use of
logical analysis in ontology is Hacker [1982].

8 This paragraph elaborates on a point made in Carrara and Varzi [2000], which
in tum owes much to Marconi [1979].
9 Ry\e [1931-32].
10 The phrase is in Russell [1919], p.169.
11 Quine [1948], p.3.
12 I'm going along with the received doctrine here, but see Oliver [1999].
13 Though I am in fact skeptical: see Casati and Varzi [1994], Chapters 3 and
12.
14 See Horgan [1978] for a representative statement.
15 The point is made in Bennett [1988], p.176.
16 See Parsons [1987-88].
17 Burgess and Rosen [1997].
18 I am indebted to Chris Partridge on this point.

19 See van Inwagen [1990], p.lOI.
Achille Varzi. Words and Objects 75
20 The fusion of a bunch of xs is defined as something that overlaps those things

that overlap some of the xs. See Simons [1987], Chapter 1, and Casati and Varzi
[1999], Chapter 3.
21 For example, the first case (an object vs. the fusion of its parts) is illustrated
by Simons [1987] and Lowe [1989]; the second case (an object vs. its constitutive
matter) is illustrated by Johnston [1992], Baker [1997], and Thomson [1998]; and the
event case is illustrated in various forms by Goldman [1971], Thomson [1971], and
Brand [1977].
22 One can formulate these statements in terms of counterpart theory, if desired,
but here I will go along with the standard formulation in terms of cross-world iden-
tity.
23 This is the line taken by Gibbard [1975] and by Yablo [1987] and recently de-
fended by Gallois [1998].
24 For further elaborations I refer to Varzi [2000] and Pianesi and Varzi [2000],
§3.
25 Dummett [1973], pp.544-45.
26 See Della Rocca [1996] and Neale [1990], §4.6.

27 See also Wiggins [1980], passim.
Part Two
Essence and Existence

BEING AND ESSENCE IN CONTEMPORARY
INTERPRETATIONS OF ARISTOTLE*
Enrico Berti
1. BEING AND EXISTENCE IN CONTEMPORARY

"ANALYTICAL ONTOLOGY"
In the proceedings of the Conference on "Analytical Ontology", held

at University of Innsbruck in September 1997, Peter van Inwagen l
published an interesting paper, where, from a point of view which he
defines "broadly Quinean", he argued in favour of the following four
theses: (1) Being is not an activity; (2) Being is the same as existence;
(3) Being is univocal; (4) The single sense of being or existence is
adequately captured by the existential quantifier of formal logic. The
first thesis is clearly an anticipation of the third. It is supported by af-
firming that the differences of being, alleged by philosophers who
conceive it as an activity (e.g. Sartre and the existential phenomenol-
ogical tradition), are only differences in nature, which do not concern
being, once admitted the distinction between a thing's being and its
nature. The second thesis is defended by referring to Quine, and states
that there is no difference between what is expressed by 'there is' and
'exists'. The third thesis is defended by means of the observation that
existence is closely tied to number, because "to say that unicorns do
not exist is to say something very much like saying that the number of
unicorns is 0", while "to say that horses exist is to say that the number
of horses is 1 or more". On the basis of this observation van Inwagen
[1998] can conclude that "the univocacy of number and the intimate
79
connection between number and existence should convince us that

there is at least very good reason to think that existence is univocal".
The fourth thesis is the most developed, by means of arguments drawn
from formal logic, about which I am not able to judge. But they - this
is at least my impression - only explain and justify in a more sophisti-
cated way the main argument brought in defence of the third thesis.
What impressed me on reading this paper, as an old frequenter of
Aristotle's philosophy, was the third thesis, which van Inwagen for-
mulates in opposition to a modem philosopher who never concealed
his Aristotelian inspiration, namely, Gilbert Ryle. This philosopher, in
a famous page of The Concept ofMind, affirmed:
It is perfectly proper to say, in one logical tone of voice, that there exist
minds and to say, in another logical tone of voice, that there exist bodies.
But these expressions do not indicate two different species of existence
[ ... ] They indicate two different senses of 'exist', somewhat as 'rising'
has different senses in 'the tide is rising', 'hopes are rising' and 'the aver-
age age of death is rising'. A man would be thought to make a poor joke
who said that three things are now rising, namely the tide, hopes and the
average age of death. It would be just as good or bad a joke to say that
there exist prime numbers and Wednesdays and public opinion and na-
vies; or that there exist both minds and bodies. (Ryle [1949], p.23)
According to van Inwagen, "Ryle has made no case for the thesis that
existence is equivocal". And he adds - but it is not clear whether refer-
ring to Ryle or in general - "I know no argument for this thesis that is
even faintly plausible". This enables him to say: "We must therefore
conclude that existence is univocal".
In fact, Ryle was not the only philosopher who admitted different
senses of being. Before him John L. Austin, the first who introduced
Aristotle in the analytical Oxford philosophy, in his famous article en-
titled "The Meaning of a Word" (1940) claimed that" 'exist' is used
paronymously", i.e. with a "primary nuclear sense" and other senses
dependent on it, just as 'healthy' in Aristotle (Austin [1970], p.71). In
Sense and Sensibilia he wrote:
'real' (the translation of the Greek, 'on', i.e. 'being') is not a normal word
at all, but is highly exceptional; exceptional in this respect that, unlike
Enrico Berti, Being and Essence in Interpretations ofAristotle 81
'yellow' or 'horse' or 'walk', it does not have one single specifiable, al-
ways-the-same meaning. (Even Aristotle saw through this idea). Nor does
it have a large number of different meanings - it is not ambiguous, even
'systematically'. (Austin [1962], p.64, italics in the text)
As everybody knows, the doctrine that being, and perhaps also exis-
tence, is at least not univocal, if not equivocal, is distinctive of Aris-
totle. He refers to it several times. 2 Admittedly, Aristotle never brings
any argument in defence of this doctrine. By so doing, he gives the
impression of considering this doctrine perfectly evident, though he
was clearly persuaded that he was the first philosopher who discovered
this truth. He blames in fact not only Parmenides, Zeno and Melissus,
i.e. the Eleatics, for having conceived being as univocal, but also De-
mocritus, and even Plato. 3 In the whole Corpus Aristotelicum there is-
as far as I know - only one passage where Aristotle makes an attempt
to prove that being is not univocal, at Metaphysics B 3, 998b22-27.
This is an astonishing situation, but just for this reason the passage is
worthy of some attention, greater than that which is usually reserved to
it even by Aristotle's interpreters.
Even van Inwagen's commentator Gary Rosenkrantz fails to men-
tion this passage. On the one hand, he admits that, because of the inti-
mate connection between the 'is' of existence and the 'is' of predica-
tion, or the 'is' of existence and the 'is' of identity, "until van Inwagen
provides some reason to think that predication and identity are univo-
cal, his argument against the equivocacy of 'exist' is not clearly valid"
(Rosenkrantz [1998]). On the other hand, he argues, this time against
Ryle, that there are species (not genera or different modes) of exis-
tence, because the system of the ontological categories "has entity (or
entityhood) as its summum genus". But this is exactly what Aristotle
denies in the quoted passage. Before analysing this passage, however,
let us say something about the relation between being and existence in
Aristotle.
2. BEING AND EXISTENCE IN ARISTOTLE
In his article "On Aristotle's Notion of Existence", Jaakko Hintikka

claimed not only that Aristotle did not admit "the Frege-Russell thesis
about the ambiguity of verbs for being like the Greek einai", according
to which these verbs express either identity, or predication, or exis-
tence, or finally subsumption, but also that Aristotle consciously con-
sidered this thesis, and rejected it (Hintikka [1999], p.782). This would
be documented in Metaph. r 2, l003 b22-30, where Aristotle says that
"to be and to be one are the same [... ] since 'one man' and 'man' as
well as 'existent man' and 'man' are the same thing, in that the redu-
plication in the statement 'he is a man and an existent man' yields no
fresh meaning". Aristotle was operating with an unitarian concept of
being, says Hintikka, because "in the fullest sense, einai had to com-
prise all of the first three Frege-Russell senses of being, that is predi-
cation, existence, and identity".
This does not mean, Hintikka argues, that Aristotle did not intro-
duce distinctions between different uses, and even different senses, of
being other than the Frege-Russell ones. "The most prominent of them
are the distinction between being in different categories as well as the
contrast between potential being and actual being". In this way Hin-
tikka shows that it is impossible to isolate, in Aristotle, the existential
from the predicative sense of being, and that, since the latter sense is
multiple, inasmuch as there are many kinds of predicates, i.e. the
"categories", the existential sense is also multiple.
The same result had been reached five years before by Lesley
Brown. By focusing on the use of the verb 'to be' in Greek philoso-
phy, she had concluded that, while distinguishing 'to be something',
i.e. the predicative use of being, from 'to be haplos', i.e. its existential
use, Aristotle thinks that they are closely connected. So, when he
claims that 'is' is said in many ways, it is impossible to decide whether
he is analysing existential or predicative uses. Lesley Brown says
Aristotle insists on the inter-relations of the question 'Is Xl' and 'What is
Xl'. [ ... ] The distinctions he does consider philosophicalJy important -
chiefly that between essential and accidental being, and the different ways
in which, as he puts it, 'being is said' which correspond to the different
categories - cut across the syntactic distinction between complete and in-
complete, and do not correspond to the semantic distinction between 'ex-
ists' and the copula. (Brown [1994], pp.233-236).
The connection between the existential and the predicative sense of

being, which is not a confusion, is at the basis also of a famous paper
by G.E.L. Owen, published for the first time in 1965. According to
Owen, when Aristotle says that "being is said in many ways" (pol-
lachos legetai to on), for 'being' he means 'existence' or, to mark the
role of 'on' as a grammatical predicate, 'existent' (Owen [1965]). "At
various places - Owen observes - Aristotle says things which show
how the verb 'to be' in its existential role or roles can have many
senses". He says in De anima II, 4l5 b 13, that "for living things, to be
is to be alive", so that, when we say that a man 'is', we mean that he
'is living', because, if he is dead, we cannot say that he 'is'. Aristotle
generalizes the point when he speaks of the 'being' of a thing (its ou-
sia or einai) as what is explained by its definition, that is, by the ac-
count of the sort of thing it is. This means that the word 'is' is used in
a variety of ways corresponding to the conditions of the being of a
thing, as its material, its position, its time, etc: For instance, as Aris-
totle himself observes in Metaph. H 2, l042b 15-l043 a7), a threshold is,
in that it is situated thus and so: for it 'to be' means its being so situ-
ated; and 'the ice is' means that it is solidified in such and such way.
On the basis of Aristotle's theory of categories, Owen continues,
to be is always to be either a substance of a certain sort, or a quality of
a certain sort, or a quantity of a certain sort. The categories are the
most general headings under which other classifications are grouped.
No category is a species of another, and no category is a species of
being and what there is, for there is no such genus as being (cf. the
passage in Metaph. B). So the verb 'to be' in its existential role enjoys
a number of irreducibly different senses.
But Owen admits that in Aristotle there is also another sense of
existence. In the Analytica Posteriora he distinguishes the question ei
estin ("whether A exists") from the question Ii eslin ("what A is")
(89b34-90a l). In this case Aristotle does not say that 'exists' has a dif-
ferent sense for different kinds of SUbjects. Instead he says that we can
only be said to know of A's existence to the extent that we know what
it is to be A (93~1-33). According to Owen, this is the use of the verb
'to be' most commonly called 'existential' at present. It is the use
which is rendered by the French 'if Y a' or by the German 'es gibl',
and it is represented in predicate logic by the existential quantifier.
Owen proposed to indicate this sense of 'is' by two asterisks, i.e. is**,
while the first can be indicated by one asterisk, i.e. is*. "Then we can
say that while Arrowby is in existence there is** at least one man still
in existence, but if Arrowby dies there is** one man who is* no
more". This second sense of existence corresponds, as we can see, to
the existence that the Quinean tradition treats as univocal. According
to Owen, Aristotle nowhere distinguishes these two uses of the verb.
So it is impossible to say whether his analysis of the different predica-
tive senses of 'exist' applies to being* or to being**. We will see that
the argument against the univocity of being brought in Metaph. B ap-
plies not only to being*, but also to being**.
For the sake of completeness I would like to mention an article by
Paul Grice, who maintains that the passages quoted by Owen show
that the semantic multiplicity attributed by Aristotle to the verb 'to be'
concerns its existential use, but it is possible to find in his works other
passages, one of which is for instance Metaph. !1 7, 1017a23-31 (there
is no difference between 'man walks, or flourishes' and 'man is walk-
ing or flourishing'), from which it can be seen that this semantic mul-
tiplicity concerns the copulative use of being, i.e. the copula (Grice
[1988]t It is surprising that none of these interpreters, and of the
many others who treated the problem of semantic multiplicity of being
in Aristotle, engaged in a close analysis of the passage of Metaph. B,
where Aristotle tries to establish the reasons of this multiplicity. Only
recently, in a book entirely dedicated to homonymy, has Christopher
Shields provided a careful examination of that passage, and arrived at
the conclusion that it does not reach its aim, i.e. it does not demon-
strate that being has a multiplicity of meanings (Shields [1999]).
3. THE ARGUMENT OF METAP H. B 3
As is well known, the book B of Aristotle's Metaphysics is entirely

devoted to the discussion of the aporiai, i.e. of the main difficulties of
first philosophy. The seventh of these aporiai concerns the Platonic
doctrine that the most universal genera, i.e. Being (to on) and One (to
hen ), are the first principles of all things. 5 Aristotle criticises it by ob-
serving that "it is not possible for either One or Being to be a genus of
things" (992b22).6 We shall examine Aristotle's reasons for this thesis
later on. For the moment I would like to point out that not only Being,
which includes, as we will see, existence, but also One, i.e. the notion
to which existence is reduced in the article of van Inwagen, are not
univocal. This emerges from a passage of the Topics, where Aristotle
provides the following rule:
Look also at the genera of the predicates signified by the term, and see if
they are the same in all cases. For if they are not the same, then clearly the
term is homonymous. (I 15, 107a3-5)
The sentence "if they are not the same" genus means, I suppose, not
only that the predicate, i.e. the property, signified by the term in ques-
tion, is instantiated many times, i.e. belongs to different things, but
also that its many instantiations do not fall within one and the same
genus. This situation occurs when a property can be associated with
different categories, as we will see in the example brought by Aris-
totle. The categories, in fact, cannot be reduced to a single genus.
Resting on this, Aristotle may formulate the general rule, following
which every predicate, whose instantiations do not fall within the same
genus, is homonymous.
According to Cat. 1, the 'homonyma' are the things which "have
only a name in common and the definition of being which corresponds
to the name is different" (1 al_2). From the passage of the Topics it ap-
pears that Aristotle calls homonymous also the term which signifies
homonymous things. In particular there is no doubt, and all interpreters
agree on this, that a homonymous term has many meanings, i.e. is not
univocal.
The homonymous term Aristotle considers in the passage of the
Topics is "good" (agathon).
Good in the case offood is what is productive of pleasure, and in the case
of medicine what is productive of health, whereas as applied to the soul it
is to be a certain quality, e.g. temperate or courageous or just; and like-
wise also, as applied to a man. Sometimes it signifies what happens at a
certain time, as (e.g.) what happens at the right time; for what happens at
the right time is called good. Often it signifies what is of a certain quan-
tity, e.g. as applied to proper amount; for the proper amount too is called
good. (107"5-12)7
As it emerges from this example, the definitions of the properties sig-

nified by the term 'good', i.e. 'productive of pleasure', 'productive of
health', 'to be courageous', 'the right time', 'the proper amount', are
different or indicate different genera of things. They therefore satisfy
the definition of homonymy offered in Categories 1 and the rule in
Top. I 15. In the light of this we must conclude that also Being and
One must be homonymous. Neither of them, in fact, is a single genus,
because the property of being and of being one has many instantia-
tions, which do not fall within the same genus.
We might wonder why Being and One cannot be a single genus.
Aristotle provides us with a reason for this in Metaph. B 3, 998b23-24:
"it is necessary both for the differences of each genus to be and for
each of them to be one". It is evident that the verb 'to be' is used in
this sentence with an existential meaning. This is, in fact, the only
meaning of 'to be' which can be predicated of all things, and therefore
also of the differences between the species of a genus, as Aristotle
makes it clear a couple of lines before (998b21). When we say that
'man is a rational animal', we mean that man is an animal with a dif-
ference, i.e. with the property expressed by the word 'rational'. Now
we may say that this property is, in the sense that it exists, and that it is
one, in the sense that it has a unity and can be counted as one differ-
ence among others. Perhaps - but I am not sure of this - by taking to-
gether being and being one as predicated of the differences, Aristotle
supposes that they have the same existential meaning. If this is true,
we could say that he is using the verb 'to be', i.e. 'to exist', exactly in
the same sense indicated by the Quinean tradition, i.e. in the sense of
being at least an instance of a class. But this verb is not univocal for
Aristotle: not only can it be predicated of each genus, but also of the
differences of each genus of things.
Immediately after, Aristotle explains to us why a term predicated
of the differences of each genus cannot be a single genus.
It is impossible either for the species of the genus to be predicated of their
own differences or for the genus to be predicated apart from its species.
(998 b24-26).
The first point, as it was observed by W.D. Ross ([1953], I, p.235), is

irrelevant to what Aristotle is proving. Nobody claims that Being and
One are species, and probably it is made for the sake of completeness8 •
The second point must be understood in the sense that it is impossible
for the genus to be predicated of its own differences, as all interpreters
admit. 9 It is not clear what "apart from its species" means and no
commentator, as far as I know, explains it. Presumably it means that it
is impossible for the genus to be predicated of its differences "in ab-
sence of its species"lO, or "instead of being predicated of its species".
We know, in fact, that every genus can be predicated of its species.
But why is it impossible for the genus to be predicated of its own dif-
ferences?
As all interpreters admit, the answer to this question is given in
Top. VI 6, 144a32-b 3:
it seems that the genus is predicated, not of the differentia, but of the ob-
jects of which the differentia is predicated. Animal (e.g.) is predicated of
man and ox and other terrestrial animals, not of the differentia itself,
which we predicate of the species. I I
The differentia which we predicate, together with the genus 'animal',

of the species 'man', in order to give its definition, is 'rational'. Aris-
totle claims that the genus "animal" cannot be predicated of the differ-
entia 'rational', i.e. that it is impossible to say that 'rational is an ani-
mal'. This is impossible for two reasons:
(1) "For if animal is to be predicated of each of its differentiae,
then many animals (polla zoia) will be predicated of the
species; for the differentiae are predicated ofthe species";
(2) "moreover, the differentiae will be all either species or indi-
viduals, if they are animals; for every animal is either a spe-
cies or an individual".
The first argument is not immediately clear. In particular, it is not clear

what is meant by "many animals will be predicated of the species",
and why Aristotle considers this as impossible. Ross ([1953]) inter-
prets these words as follows: "If it were so predicated, the genus
would be predicated of the species many times over, since it would be
predicated of each of the successive differentiae which constitute the
species".12 But Aristotle is not saying that the same genus 'animal'
would be predicated 'many times' (pollachis) of the species, but rather
that 'many animals' (polla zoia) would be predicated of it. Aristotle
does not seem to be thinking that it is impossible to predicate the same
genus of many differentiae, but rather that it is impossible to predicate
the genus even of a single differentia, because in this way many gen-
era would be predicated of the same species.
Tricot ([1950]) explains: "If the genus were attributed to the
differentia, the genus would be many times over an attributes of the
species [ ... ]. Indeed, animal is said once of the human species, once of
the rational differentia, which would surely constitute a new species,
relative to which the same difficulty would arise anew". But in this
way too we should read 'many times' instead of 'many animals'. Be-
sides, the consequence of reducing the differentia to a species belongs,
as we will see, to the second argument and it cannot be attributed, at
least immediately, also to the first. Zadro ([1974], pA85), on his part,
interprets: "Given that the differentia is in its turn predicated of the
species, this would also means that zoion would be predicated of
'man', with the result tha zoion would be a property both of the e idos
and of the dia/ora, and the definition by way of the genus would be
redundant". I agree that, if the genus could be predicated of the differ-
entia, it would be predicated twice of the species, but the consequence
indicated by Zadro, that in this way the definition would become re-
dundant, does not seem to me impossible, as also Aristotle himself
thinks.
In his book entirely dedicated to homonymy in Aristotle, Shields
provides a close analysis of the passage in Metaph. B 3. He argues that
Aristotle is here concerned with being in its existential sense. Shields
focuses also on the passage in Top. VI 6, and criticises the other inter-
preters (e.g. Grice) for not having made use of it. But he follows Waitz
and Ross in interpreting polla zoia as pollachis to zoion, and argues
that Aristotle's argument, thus understood, "seems to appeal implicitly
to homonymy". According to Shields, if the difference, e.g. 'being
two-footed', counts as being an animal, the notion of 'animal' em-
ployed in saying 'being two-footed is an animal' must be significantly
unlike the notion employed in 'Callias is an animal'. Therefore, "if one
were to predicate 'animal' of its various differentiae, then one would
Enrico Berti, Being and Essence in Interpretations of Aristotle 89
render 'animal' homonymous by inventing non-standard senses of the

term [... ] Hence, predicating the genus of the differentiae renders the
genus term homonymous" (Shields [1999], pp.252-253).
Obviously, if Aristotle's argument is interpreted in this way, it is
easy to object - as indeed Shields does - that it is self-contradictory,
because it aims at demonstrating that being is not a genus, i.e. it is ho-
monymous, on the basis of the fact that, if being were a genus, and
were predicated of its differences, it would be homonymous, which
Aristotle does not admit, negating in this way his own thesis. For this
reason Shields concludes that Aristotle's argument is not valid, and
therefore it does not succeed in demonstrating that being is homony-
mous. For his own part, Shields thinks that, unlike good, being is not
homonymous, and in particular that existence is a univocal notion,
which can be predicated of everything, even of different categories,
and always with the same sense (Shields [1999], pp.265-266). Shields
finds himself in this way in agreement with van Inwagen, but (unlike
van Inwagen) by an examination and refutation of Aristotle's argu-
ment for the homonymy of being.
To all appearences, the objection Shields makes against the argu-
ment of Aristotle is valid. It has nevertheless a defect: it is not ad-
dressed to what Aristotle actually says, but rather to what Shields
conjectures that Aristotle says. Besides, Aristotle offers a second rea-
son why it is not possible to predicate the genus of its own differen-
tiae: "the differentiae will be all either species or individuals, if they
are animals; for every animal is either a species or an individ-
ual"(144 b l-3). If the differentiae were either species or individuals,
they would be ranged next to the species or individuals included in the
genus. In this way they could not be predicated of species, and there-
fore they would lose the function of distinguishing the species from
one another. As Aristotle says at the end of the passage in Metaph. B,
"if one or being is a genus, no difference will be either a being or a
one" (998 b26-27). The consequence of Plato's thesis that Being and
One are genera is therefore that there are no more differences, and that
all things are one, which is the monism of Parmenides.
I would like to propose an alternative reading of the first argument
of Aristotle. Yet this is not the interpretation of the argument proposed
by Arthur Madigan in his recent commentary on Metaphysics B. This

latter says:
To predicate a higher kind or genus of a difference is illegitimate for two
reasons. First, if this were legitimate, the higher kind or genus would be
predicable of each successive difference, and so would be predicated of a
multi-differenced species as many times as that species had differences;
this is taken to be absurd. Second, if it were legitimate to predicate the
higher kind or genus, such as animal, of its differences, then each of its
differences would be an animal, and since every animal is either an indi-
vidual animal or a species of animal, each difference would be either an
individual animal or a species animal; both possibilities are taken to be
absurd. (cf. Madigan (ed.) [1999], p.74)
In this interpretation, the first argument is still understood in the same

way as it was by Waitz and Ross.
Other considerations developed by Madigan are more convincing.
In particular, he defends Aristotle from a criticism advanced by Alex-
ander [206,12-207,4]. According to Alexander, Aristotle's arguments
are verbal rather than substantive. Madigan observes that they are
more than a mere verbalism, because "Aristotle supposes that there are
certain proper or canonical patterns of predication that mirror the
structures of being and certain improper patterns of predication that
fail to mirror the structures of being or even distort them". As an in-
stance of the latter he mentions predication of one accident of another
accident, as 'the white is musical', and predications which involve a
"doubling" or repeated mention of the same thing. Madigan continues:
Such a doubling would occur if one and being were all-encompassing
highest kinds or genera and predicated of all differences. Take, for exam-
ple, 'A human is a rational animal', and suppose that one and being are
predicated not only of animal, but also of rational. Results: 'A human is a
one one' (because animal is one, but rational is also one), and 'A human is
being being' (because animal is being, but rational is also a being). The
doubling of 'one' or 'being' does not simply violate a rule of dialectical
debate; it also gives a distorted view of a thing by suggestion that the
same item enters more than once into the constitution of the thing. (cf.
Madigan [1999], p.75)
Madigan is not particularly happy in his illustration of the "doubling".

First of all, he speaks of 'being' and 'one', whilst Aristotle speaks of
'many animals'. Secondly, there is no doubt that this "doubling" for
Aristotle is not only a redundancy in predication, i.e. a verbal incon-
venience, but also a sign of a distortion of the structures of being. On
the other hand, I completely agree with a more general consideration
Madigan offers, which goes as follows:
To speak of something as a kind presupposes some determinate content
and a contrast between that determinate content and the contents of other
things. Thus to speak of substance as a kind presupposes a contrast be-
tween it and the other kinds. Being and one fail to be kinds in this sense.
To speak of being as a kind implies no such contrast, for all things are
beings. To speak of one as a kind implies no such contrast, for all things
are ones. As terms of universal extension, one and being are too broad to
count as kinds. (cf. Madigan [1999], p.75)
This is completely true, but is not explicitly said in the passages of

Metaph. B and Top. VI.
The alternative interpretation of the first argument of Aristotle I
would like to propose goes as follows. If the genus 'animal' could be
predicated of its difference 'rational', then rational would be an ani-
mal, not because it would be identified with the animal, but because it
would be a particular instance of the genus 'animal'. In this case, the
genus 'animal' would enter in the definition of 'rational', so that the
'rational' would be defined as an animal with another particular differ-
ence.13 Now, as the genus and the difference must be both predicated
of the species 'man', two 'animals' would be predicated of this spe-
cies, i.e. the genus 'animal' and the difference 'animal', or - as Aris-
totle himself says - 'many animals' (polla zoia). The animals would be
as many as the differences of which the genus 'animal' can be predi-
cated. But in this way only genera, and no difference, would be predi-
cated of the species. There would remain nothing which could distin-
guish the species of the genus from one another. In other words, if the
genus could be predicated of the difference, the difference would be-
come itself a species and would lose its function of distinguishing one
species from the other species of the genus. If interpreted in this way,
the first argument of Aristotle would converge with the second, and
the interpretation of Tricot would be justified.
Even in the case in which 'animal' would be predicated of 'ra-
tional' without being its genus, but simply as an accident of it, it would
be predicated also of 'man', i.e. of the species of which the difference
is predicated, in virtue of the logical rule according to which the predi-
cate of a predicate is a predicate of the subject (nota notae, nota rei).
In any case, Aristotle's arguments ultimately rest on the doctrine of the
definition of a species by its genus and its differentia, where the genus
expresses what every species has in common with the other species
within the same genus, and the difference expresses what distinguishes
one species from the others species of the same genus. If the genus
could be predicated of the difference, either as its genus or as an acci-
dent, then the definition would only indicate the common aspects of
the species, losing what enables it to distinguish them from one an-
other. If Aristotle's arguments are valid, as I have tried to show, we
can conclude that his attempt to demonstrate the homonymy of being,
or at least that being is not univocal, is successful. Metaph. B makes it
clear that the non-univocity holds both for being conceived as exis-
tence as well as for being conceived as unity.
The 'univocacy of number', invoked by van Inwagen, does not
hold among objects belonging to different genera, as it is the case of
beings. We cannot, in fact, count the objects contained in a room, if
they belong to different genera. We can count, for instance, persons,
tables, chairs, books. In the Aristotelian language they are substances.
But we cannot count, together with them, the colours of the tables, the
weight of the books, the actions or the feelings of the persons, though
we must admit that each of these things does exist and is at least one
instantiation of its class. In particular, Aristotle said that "to be one
[ ... ] is specially to be the first measure of a kind"l\ and "the measure
is always homogeneous with the thing measured: the measure of spa-
tial magnitudes is a spatial magnitude, and in particular that of length
is a length, that of breadth a breadth, that of articulated sounds an ar-
ticulate sound, that of weight a weight, that of units a unity".15 This
doctrine was endorsed by modem analytical philosophers such as P.T.
Geach, M. Dummett and C. Wright and it was attributed by them to
one of the founders of this tradition, i.e. G. Frege. 16
4. BEING AND ESSENCE IN ARISTOTLE
But the risk of univocity is not yet completely eliminated. As we shall

see, even in admitting the multiplicity of the meanings of being, there
is still the possibility of conceiving one of these meanings as the es-
sence of being, which for Aristotle would be equivalent to admitting
the univocity of being. At the beginning of Metaph. r, after stating that
"there is a science that investigates being as being", Aristotle adds that
"there are many senses in which a thing may be said 'to be', but they
are related to one thing (pros hen), i.e. one definite kind of thing, and
are not homonymous". In order to illustrate this case, he brings two
examples:
Everything which is healthy is related to health, one thing in the sense that
it preserves health, another in the sense that it produces health, another in
the sense that it is a symptom of health, another because it is capable of it.
And that which is medical is relative to the medical art, one thing in the
sense that it possesses it, another in the sense that it is naturally adapted to
it, another in the sense that it is a function of the medical art. (Barnes (ed.)
[1985], 1053"24-27)
This happens also with being, where everything that it is said to be ei-
ther is a substance (ousia) or it is relative to substance:
some things are said to be because they are substances, others because
they are affections of substance, others because they are processes to-
wards substances, or destructions or privations or qualities of substance,
or productive or generative of substance, or of things that are relative to
substance, or negations of some of these things or of substance itself.
(1003"33-b l0)
This situation was considered by Austin [1970], as a case of 'parony-

mity'. Paronymity is described in Cat. 1 as different from both homo-
nymy, which implies an identical name with different connotations,
and synonymy, which implies an identical name with identical conno-
tations. "Paronymity - affirms Aristotle - belongs to things that have
different names, but derived from one of them, e.g. when the gram-
marian gets his name from grammar, or the brave gets his name from
bravery" (Cat. 1, 1al2-15). This is not exactly the case of 'being', or
'healthy', or 'medical'. Indeed, these names remain always the same.

However, Austin considered the multiplicity of meanings of a word,
which are all relative to one of them, a particular form of paronymity.
He described it as the situation in which "a word may possess conno-
tations which are partly identical and partly different". He explained
this fact as follows:
when we speak of 'healthy exercise', the world 'healthy' has a connota-
tion which is only partly the same as that which it has in the phrase 'a
healthy body': a healthy exercise is a exercise which produces or pre-
serves healthiness in bodies. Hence healthiness a, when predicated of an
exercise, means 'productive or preservative of healthinessb, i.e. of
healthiness in the sense which is predicated of bodies. Thus 'healthinessb'
and 'healthiness a' have connotations which are partly identical and partly
different. (Austin [1970], p.27)
Speaking of the word agathon, i.e. 'good', Austin argued that

"sometimes it means 'x', sometimes 'productive, etc., of x', etc.; and
clearly it is only the 'nuclear' meaning of 'x' which is common to
both, with which they are concerned".!7 In his article 'The meaning of
a word', speaking of the word 'healthy', Austin wrote:
In this case there is what we may call a primary nuclear sense of
'healthy': the sense in which 'healthy' is used of a healthy body. I call this
nuclear because it is 'contained as a part' in the other two senses which
may be set out as 'productive of healthy bodies' and 'resulting from a
healthy body'. (Austin [1970], p.71)
The same situation holds - said Austin - also for the word 'exist', i.e.
being.!S Now, it seems to me that, if the Aristotelian doctrine of the
multiplicity of meanings relative to one were interpreted in this way,
i.e. if substance (ousia) was taken as the nuclear meaning, "contained
as a part" in all the other meanings, and "common" to all of them, as
suggested by Austin, substance would become the genus of being and
the other meanings would be only specifications of this genus. They
would specify the genus without modifying it. The genus is in fact the
common part of the definition, and it is specified by the suitable dif-
ferences. If therefore the definitions of different meanings have a part
in common, and this part is always the same, this part would be neces-
sarily the genus. Moreover, given that the genus is part of the essence,
substance, conceived as a genus, would be nothing but an essence, the
essence of being.
But the examples given by Aristotle do not suggest this idea. The
different meanings of 'healthy' are in different relations with health.
Apparently, the relations Aristotle is thinking of are those of produc-
tion, preservation and manifestation of health. They are proper to
things such as a healthy medicine, a healthy climate, a healthy com-
plexion, which do not belong to the same genus and do not have the
same essence. We cannot say that health is the essence of the healthy
medicine, or the healthy climate, or the healthy complexion. We
should say that health is the product of the healthy medicine, or the
thing preserved by the healthy climate, or finally the thing manifested
by the healthy complexion. In all these cases there is surely a relation
between the single healthy thing and health. This relation is neverthe-
less different in each case. The different healthy things cannot be
therefore considered as mere modifications, or qualifications, of
health. This is also the case of being, at least in my opinion. Each
meaning of being, i.e. each category of being, substance, quality,
quantity, relation, time, place, etc., is in a certain relation with the sub-
stance and it is said to be in virtue of this relation. But substance is not
the essence of quality, quantity, relation, time, place, etc.. Nor are
these latter mere modifications, or qualifications, of substance in the
sense that they should be substances with some accidental qualifica-
tions: they are other kinds of being, different from the substance,
though dependent on it.
In the Eudemian Ethics, illustrating the different kinds of friend-
ship, Aristotle himself explains that they are all relative to one, which
is the primary, just as in the case of the word 'medical', and adds:
"Everywere, then, we seek for the primary. But because the universal
is primary, they also take the primary to be universal, and this is an
error" (Eth. Eud. VII 2, 1236a22-25). The target of his criticism are the
Platonists, who took the primary as a universal, i.e. a common aspect,
a common predicate, like the genus. According to Aristotle, this is a
mistake: the primary meaning of friendship is not the genus of which
the other meanings are the species. "The primary is - Aristotle says -
that of which the notion is present in us" (en hemin), not 'in the defi-
nition of all' (en pasin), as many interpreters believe. 19 This means that
the primary is only a term of reference, i.e. that to which the other are
in relation (pros), and it is common to all just for this reason, and not
because it is a universal, in conformity with which (kata) the other are
said.
Interpreters do not completely agree with one another on this
point. In a famous article entitled 'Aristotle and the Ambiguity of Am-
biguity', Hintikka distinguishes three cases: the case in which the
multiple applications of a term have totally different definitions
(homonymy), the case in which the multiple applications of a term
have totally identical definitions (synonymy), and finally the case in
which the multiple applications of a term have definitions which are
partially identical and partially different. According to Hintikka, this
last case is that of being. This is described by Aristotle sometimes by
the relation of its applications to one central point (pros hen), and
sometimes by the affirmation that their definitions are derivable from
each other "by adding to and taking away".20
There is a passage of Metaph. Z 4 which illustrates the latter
situation, and which Hintikka quotes in support of his interpretation. It
sounds:
Essence (to ti en einai) will belong, just as the 'what' (to ti estin) does,
primarily and in the simple sense to substance, and in a secondary way to
the other categories also, - not essence simply, but the essence of a quality
or of a quantity. For it must be either homonymously that we say that they
are, or by making qualifications or abstractions (prostithentas /wi aphai-
rountas) (in the way in which what is not known may be said to be
known), - the truth being (to ge orthon) that we use the word neither
homonymously nor in the same sense, but just as we apply the word
'medical' when there is a reference to one and the same thing, not mean-
ing one and the same thing, nor yet speaking homonymously; for a pa-
tient and an operation and an instrument are called medical neither homo-
nymously nor in virtue of one thing, but with reference to one thing (pros
hen). (1030"29-b 3)
In this passage it is not clear whether the use of the verb to be "by
making qualifications or abstractions" coincides or not with the use of
this verb with reference to one and the same thing. In the first case, the
meaning of being in the other categories (quality, quantity, etc.) would

be the result of a qualification of the meaning that being has when it is
used for substance. Substance would be the purest instance of being,
i.e. its essence. In this way what is associated with the other categories
would be a substance with or without some further qualification. In the
second case, the substance would be only the tenn of reference for the
other categories, i.e. the thing with which they are in relation, without
being their essence. It seems to me that Aristotle does not identify the
two uses of being. The example Aristotle brings to illustrate "by mak-
ing qualifications or abstractions", i.e. "the way in which what is not
known may be said to be known", coincides with what is mentioned
some lines before, i.e. with the case of the people who say that that
which is not is (l030 a25-26). Now, Aristotle himself qualifies these
people as speaking "in a mere verbal way" (logikos) , and contrasts
them - I suppose - with "the truth (to ge orthon)", which consists in
using the word neither homonymously nor in the same sense, but just
as we apply the word 'medical"'.21
A further clarification of Aristotle's doctrine of being was offered
by G.E.L. Owen in his paper entitled "Logic and Metaphysics in Some
Earlier Works of Aristotle". He introduced the expression 'focal
meaning' to describe the first of the meanings of a word said in many
senses, but all relative to one of them (Owen [1960]). This is much
better than 'nuclear meaning': it does not presuppose that the first
meaning is a part of the others, but indicates that it is only the focus,
i.e. the tenn of reference, of the others. Nevertheless, the way Owen
describes focal meaning turns out to be close to what was already sug-
gested by Austin. According to Owen, in fact, a sense of a word is
primary "in that its definition reappears as a component in each of the
other definitions". He speaks also of a "reductive translation" in con-
nection with the relation between the definition of the substance and
the definitions of the other categories, explaining that "all the senses of
on must be defined in tenns of ousia, 'substance"'.22 Finally, he makes
it clear that this is the relation of "logical priority", i.e. priority in logos
or definition. All these expressions could be applied also to the genus,
which is also contained in the definition of the species, and therefore is
logically prior to it.
Logical priority as well as natural priority, or priority in being, is

ascribed to substance in Metaph. Z 1. Here Aristotle says that sub-
stance is prior "in notion" (toi logo i) to the other categories, because
"in the definition of each term the notion of the substance must be pre-
sent"(1028 a35-36), obviously not as their genus but rather as their term
of reference. Given that, as we know that substance is prior to the
other categories not as the genus in conformity with which (kata) they
are said, but as the principle with which they are in relation (pros)23,
the logical priority of the substance must be interpreted in a particular
way, which does not emerge from Owen's interpretation. 24 Even the
natural, or ontological, priority of the substance with respect to the
other categories must different from that of the genus respect to the
species. This latter is presumably the ontological priority admitted by
Plato (cf. Metaph.!:1 11, lOI9 a l-6), but it is explicitly distinguished
by Aristotle from the priority of the substance respect to the other
categories (Ibidem).
The claim of Metaph. Z 1 that substance is not only that which is
"primarily" (to protos on), but also that which is "simply" (haplos)
rather than being something (ou ti on) (1028 a30), might appear to pose
some problems for this interpretation. The expression 'which is sim-
ply', opposed to 'being something', might suggest that substance is
pure being, without qualifications, perhaps pure existence, i.e. the es-
sence of being, or the essence of existence. This interpretation was
advanced by the German translation of Frede and Patzig: "das, was
primar Seiendes und nicht nur in bestimmter Hinsicht, sondern
uneingeschrankt Seiendes, die ousia ist". On the basis of this transla-
tion, in fact, the substance turns out to be the 'unlimited being', which
seems to be the being for essence, the essence itself of being.
5. ESSENCE AND PRIMARY SUBSTANCE
This interpretation is suggested also by another contribution of Patzig

and Frede, concerning the relation between the substance in general
and the primary kind of substance, i.e. the unmovable substance. In an
article concerning the relationship between ontology and theology in
Aristotle's Metaphysics, Gunther Patzig, apparently unaware of the
contributions of Austin and Owen, employed paronymity not only to
explain the dependence of the other categories on substance, i.e. the

unity of ontology, but also to explain the dependence of the various
kinds of substance on the unmovable substance (Patzig [1960-1961]).
For Patzig, not only is being used in many senses, all referred to one of
them, i.e. to substance, but also substance is used in many senses. Sub-
stance may mean three kinds of things, the movable and corruptible
substance (the sublunar bodies), the movable and incorruptible sub-
stance (the heavenly bodies) and finally the unmovable substance (the
movers of the spheres). The last substance is the cause of the others,
and it is 'first' and 'principle' with regard to them. Therefore, on the
basis of the rule exposed by Aristotle in Metaph. r 2, 1003b 16-17
("everywhere science deals chiefly with that which is primary, and on
which the other things depend, and in virtue of which they get their
names"), the science of substance is primarily the science of this sub-
stance. This is the reason why Aristotle calls it theology. But, as the
science of substance, in virtue of the paronyrnity of being qua being, is
also the science of being, theology would coincide with ontology, and
it would be "universal in this way, because it is first" (Metaph. E 1,
1026a30-31).
Patzig founded this interpretation on many passages, where Aris-
totle actually says that the unmovable substance is "principle"
(Metaph. A 7, lOnb ll-14: arche) and is "the first" (Metaph. A 8,
1073 a30: prote). The first mover - says Patzig - is the paronyrnic prin-
ciple of all substances, "the substance of substances". First philosophy
is first not because it comes before the others (on the contrary, it
comes last). Nor is it first because of the dignity of its object. First
philosophy is first because it is the science of the first substance. In
this way we can explain the double characterisation of first philosophy
as ontology and as theology. This consists of a unique and consequen-
tial process of thinking, which goes from books A - E to book A of the
Metaphysics, even if in books Z-8, which for Patzig are posterior,
paronyrny is substituted by analogy.25
The same interpretation has been recently endorsed, and further
developed, by Michael Frede [1987]. In the light of the article of Owen
[1960], Frede no longer speaks ofparonyrnity but of 'focal meaning'.
He shows no hesitation in affirming that substance is the focal mean-
ing of being, and that the unmovable substance is the focal meaning of
the substance, and therefore of the whole being. In his interpretation

the unmovable substance, i.e. the divine substance, would be the par-
ticular way of being, "in terms of which all other ways of being have
to be explained". The unmovable substance would be nothing but the
focal way (or sense) of being. Apparently, Frede extends the use of the
notion of focal meaning, as it was defined by Owen, that is to say as
implying logical as well as ontological (or natural) priority. He makes
use of this notion to describe both the relation between substance and
the other categories and the relation between the unmovable substance
and the other kinds of substance. The expression 'in terms of which'
seems in fact to be used to describe the situation in which a certain
notion is contained in another, i.e. precisely a situation of logical pri-
ority.
Frede is aware that in Aristotle's texts this thesis is not explicitly
stated. Nevertheless he thinks that this thesis is implied in what Aris-
totle affirms about sensible substance in book Z of Metaphysics. Here,
as it is well known, Aristotle identifies the "substantial form" - the ex-
pression is used by Frede - with the "first substance", i.e. with the sub-
stance "in terms of which the substantiality of the sensible substances
has to be explained". Now, the unmovable substances are nothing but
substantial forms separated from matter, they have therefore the same
type of substantiality as belongs to the substantial forms. They are
prior, as substances, to the sensible ones, and "we shall achieve a full
understanding of the substantiality of sensible substances only when
we have understood the substantiality of non-sensible substances". The
conclusion that Frede proposes, on the basis of a passage where Aris-
totle says that the simple and actual substance, i.e. God, is the first in-
telligible object (Metaph. A 7, IOna26 ss.), is that "ultimately nothing
is intelligible unless it is understood in its dependence on God". In the
light of this it is clear that Frede is ready to accept that God is not only
ontologically but also logically prior to the other beings.
This is confirmed by his presentation of the entire Aristotelian
ontology in terms of a 'scale of perfection', in which the lower forms
of being somehow imitate higher forms of being: on the one hand,
animals procreate to imitate the eternity of the heavens, and by so do-
ing they assure eternity to their own species; on the other, the heavens
eternally rotate to imitate, as they can, the unchanging nature of the
unmovable mover. The unmovable movers - continues Frede - are

"beings in a paradigmatic way, in that they are pet;fectly real", and the
separate substances are "paradigmatic as substances", because they
have the necessary qualifications for substantiality, which are to be the
last subject of predication and also to be separate. In this way general
metaphysics has as its core the study of the way of being of the divine
substances, and ultimately coincides with theology.
It seems to me that this is a Platonic rather than Aristotelian con-
ception of being, or in any case it is the result of an interpretation of
Aristotle in a Platonizing, or even in a Neoplatonizing vein. 26 The re-
lation of imitation that it establishes among the various ways of being
is in fact the same as the relation which Plato admits between Ideas
and sensible things. This can be supported only by some Aristotelian
texts, being in contrast with others. It is true that Aristotle considers
procreation as an imitation of the eternity of the heavens (De Gen.
Corr. II 10, 336b32-337 a7; De an. II 4, 415 a26-b2). But Aristotle never
says that the circular motion of the heavens is an imitation of the im-
mobility of the unmovable mover. This is only an interpretation of his
thought, whose Platonizing character was already denounced by
Theophrastus, who attributed the conception of heaven's desire as
imitation of the unmovable mover to "people who admit the One and
the numbers", i. e. the Platonists. 27 Aristotle, on the contrary, sharply
criticizes the exemplaristic, i.e. "paradigmatic", causality of those
separate substances that are Platonic Ideas, and therefore also their
utility for understanding the sensible substances, affirming for instance
that the cause of Achilles is not the universal man, who does not exist,
but Peleus, i.e. his efficient cause, and "of you, your father" (Metaph.
A 5, 107I a21-22).
Frede's interpretation, for its tendency to consider the causality of
unmovable substances in terms not only of ontological, but also of
logical priority, ultimately depends on analytical philosophy, that is to
say on the analysis of ontological relations only in terms of logico-
linguistical relations, whose model is Owen's analysis of 'focal
meaning'. At least in this case it singularly converges with Platonism.
The results of Frede's interpretation are exactly the same as those ob-
tained, 50 years ago, by a well-known Thomist interpreter of Aris-
totle's Metaphysics, Father J. Owens, a pupil of E. Gilson in the Pon-
tifical Institute of Mediaeval Studies in Toronto and belonging to that

stream which emphasizes the Platonic and Neoplatonic elements of
Thomism. In his book of 1951, on The Doctrine of Being in the Aris-
totelian Metaphysics, Owens claimed that.between the sensible and the
separate substance there is the same relation of pros hen legesthai
which exists between the other categories and the substance, and that
this latter (called by him 'entity'), in its primary instance, is the form,
which can be the form of a compound or a pure, i.e. separate, form.28
Owens too, like most of the Platonizing interpreters, considered the
causality of the unmovable mover as an exemplaristic causality, af-
firming that the heavens move circularly in order to imitate the immo-
bility of the unmovable mover. 29
What is doubtful in the interpretation proposed by Patzig and
Frede, at least to my mind, is that the unmovable substance is logically
prior to the other kinds of substance. I do not see how the definition of
the movable substance can contain the notion of the unmovable sub-
stance. There is only one passage which could go in this direction. In
Metaph. A 7, lOna27-32, Aristotle affirms that "the primary (ta
prota) object of desire and thought are the same", and that they are
"the substance which is simple and exists actually". This substance -
Aristotle adds - like all the terms of the positive series, is intelligible
by itself (noete kath 'hauten). This surely means that the notion of the
unmovable substance does not contain other notions, i.e. that it be-
longs to the things better known by nature, not for us, because it is the
farthest from perception. Does this imply that the notion of the un-
movable substance is contained in the definition of the other sub-
stances? The notion of form, or the notion of actuality, is certainly
contained in the definition of all the other substances. From this we are
nevertheless not entitled to conclude that the definition of separate
form, or pure actuality, is contained in the definition of the other sub-
stances. It does not seem to me that these notions are without qualifi-
cation, as Frede claims. Separateness and purity are important qualifi-
cations, which we discover only at the end of the philosophical re-
search, that is to say after having demonstrated the existence of the
unmovable mover.
I can agree with Frede's claim that "nothing is intelligible unless it
is understood in its dependence on God", but only if by dependence
we mean ontological, not logical, dependence, and ontological in the

sense of causal dependence, not in the sense of the Platonic depend-
ence of the species on the genus. I can also agree with the affIrmation
that God is the being "in terms of which all other ways of being have
to be explained", but only if'in terms of which' means 'in dependence
on which', and the dependence in question is ontological in the speci-
fied sense. I cannot agree with the claim that God is being 'in a para-
digmatic way'. Indeed, I do not believe that Aristotle admits a relation
of exemplarity, i.e. of imitation, between God and the other sub-
stances. This seems to me to belong to the Platonic tradition rather
than to Aristotle. What I want to emphasize is that, for Aristotle, to be
the 'first' does not mean necessarily to be a model, a perfect exemplar,
the highest degree, the purest instance, but it can also mean principle,
or cause, or moving cause.
6. ARISTOTLE'S REFUSAL OF AN ESSENCE AND A

SUBSTANCE OF BEING: METAPH B 4
If the primary substance were the purest instance of being, it would be

the essence of being. In other words, there should be a substance,
whose essence would be being itself. This is the concept of God as
Esse ipsum subsistens, which is present in all the religious interpreta-
tions of Greek philosophy, i.e. in the Jewish theology of Philo of Al-
exandria, in the Muslim theology of Avicenna and in the Christian
theology of Thomas Aquinas. 30 In general, the supporters of this con-
ception do not pay enough attention to the fact that Aristotle not only
knew this conception but ascribed it to Plato and criticized it by argu-
ments which are closely connected to his doctrine of the multiplicity of
the meanings of being. His criticism against this conception is once
again contained in the book B of the Metaphysics, which is considered
by all the interpreters as an aporetic, or dialectical book, and presuma-
bly for this reason it is not taken seriously.3)
In the eleventh aporia of book B, which is said "the most difficult
of all even to study and the most necessary for knowledge of the
truth", Aristotle asks "whether being and one are really the substances
of beings (ousiai ton onton), and whether each of them, without being
something else, is being or one respectively, or whether it is necessary
to inquire what being and one really are, supposing that another nature
underlies them as a subject" (1001 a4_8). He attributes the first opinion
to some philosophers and the second to others. As supporters of this
opinion he gives the names of Plato and the Pythagoreans, whose doc-
trines are described as follows: "Plato and the Pythagoreans think that
neither being nor one is something different, but that this is what their
nature is, supposing that its ousia is to be one and being" (1001 a 9-
12). In Aristotle's language ousia means not only 'substance' but also
'essence', and this last meaning should be preferred when the term is
followed by the genitive (ousia of something). But if so, we must con-
clude that, for Aristotle, Plato and the Pythagoreans conceived of be-
ing and one as substances, whose essence was respectively to be being
and to be one.
Aristotle's criticism of the second opinion, which he ascribes to
Empedocles, ultimately rests on arguments which have sense only
from an Academic, i.e. Platonic, point of view. After this criticism,
Aristotle goes back to the first opinion, which he reformulates in the
following way: "If there exists some one itself (auto hen) and being,
then one and being are necessarily their substance (ousia). For nothing
different is predicated of them universally, but rather they themselves"
(l001a27-29). Aristotle normally uses the pronoun "itself' (auto) to
designate Platonic Forms.32 In the following line he applies it also to
being (auto on). Also in the light of this it seems that he is thinking of
Plato rather than the Pythagoreans, and that he is ascribing to Plato a
doctrine which considers Being itself, i.e. the Form of being, and One
itself, i.e. the Form of one (it is not clear whether they coincide or not),
as the substances, i.e. the formal causes, of all the beings. This corre-
sponds partially to the doctrine of the One and the indefinite Dyad as
principles of all things, ascribed to Plato in book A of Metaphysics.
What is most interesting is the criticism which Aristotle addresses
to this doctrine:
But on the other hand, if there is to be some being itself (auto on) and one
itself (auto hen), there is much aporia about how anything different will
exist alongside them: I mean, how beings will be more than one. For that
which is different from being is not. So, in line with the argument of Par-
menides, the necessary consequence is that alI beings are one and that this
is being (to on). But either way it is difficult. (l 00 1"29_b1)
Notice that the sentence "that which is different from being is not" is
justified only "in line with the argument of Parmenides", i.e. only if
we admit that being has only one meaning, the doctrine that Aristotle
usually ascribes to Parmenides. 33 Ross thinks that here Aristotle is not
justified in ascribing to Plato the Parmenidean notion of being, be-
cause - he says - "for Parmenides it (to on) means 'what is', i.e. the
universe; for the Platonists it means 'being', i.e. the attribute of exis-
tence. It is this abstraction that they make a substance, and there is
nothing in this to prevent their recognizing of other substances".34
But Aristotle's interpretation of Plato's notion of being is not
without reasons. Elsewhere he says that the Platonists admitted two
principles, i.e. the One and the indefinite Dyad, because
they frame the difficulty in an old-fashioned way (aporesai archaikos),
for they thought that all things that are would be one - viz. Being itself
(auto to on), if one did not join issue with and refuted the saying of Par-
menides: 'For never will this be proved, that things that are not are'. They
thought it necessary to prove that that which is not is; for thus - of that
which is and something else - could the things that are be composed, if
they are many. (Metaph. N 2, 1089"1-6)
Apparently, the Platonists thought that, if we do not admit two princi-

ples, i.e. the One and the Dyad, which are equivalent to Being and
Not-being, we are not able to account for the multiplicity of things.
The reason for this is that Parmenides as well as the Platonists con-
ceived being as having only one meaning. The objection that Aristotle
immediately addresses to them is that "being has many senses"
(1089 a7). But if being has many senses, there is no need of another
principle, opposed to it, i.e. not-being, to account for the multiplicity
of things.
Plato conceived being as having only one meaning because he
conceived it as a genus, that is to say as an universal predicate ex-
pressing only what is common to all things, i.e. only a single aspect of
things. This was, indeed, the condition for conceiving it as a separate
FOTIll,i.e. an Idea. And this was also the condition for conceiving be-
ing and one as the essence of a substance, i.e. being itself (ipsum esse
subsistens) and one itself (ipsum unum subsistens). In conclusion, if
the primary substance is the essence of being, being must be under-
stood univocally. If being has an essence, it is this essence. It cannot
be many essences. But this is impossible; because we see many things,
and their differences are existing and each of them is one. This is the
core of Aristotle's criticism of Plato as it is exposed in Metaph. B 4.
This criticism ultimately rests on the argument offered in Metaph. B 3,
and the view that Being and One cannot be genera.
Notes
Parts of this paper first appeared in the Proceedings of Aristotelian Society,
vol. CI part 2, 185-207; they are here reprinted by courtesy of the Editor of the Aris-
totelian Society © 2001.
I See van Inwagen [1998].
2 Cf. Aristotle Metaph. r 2, 1003'33; E 4, 1028'5; Z 1, 1028'10; N 2, 1089'7.

3 Aristotle Soph. El. 33, 182b25-27; Aristotle Metaph.A 4, 985 b7, N 2,1089'1-3.
4 lowe my knowledge of this article to the kindness of David Charles.
5 The Platonic origin of this doctrine is rightly emphasized by Madigan (ed.)
[1999], p.73).
6 Translation by A. Madigan (ed.) [1999]), with some modifications.
7 Translation by 1. Barnes (ed.) [1985]).
8 Madigan (cf. Madigan (ed.) [1999]) agrees with Ross about the irrelevance of
this point.
9 This can be seen in the Revised Oxford Translation (Barnes (ed.) [1985]: "it is
not possible for the genus to be predicated of the differentiae") and from the com-
mentary by Madigan ([1999], p.73).
!O This is the translation by Shields (ed.) [1999], pp.247-248.
II Revised Oxford Translation (cf. Barnes (ed.) [1985]).
12 Ross is clearly following Waitz (cf. Waitz (ed.)[1844-1846], II, p.500), who
interprets polla zoia as pollachis to zoion .
13 I am grateful to Stephen Menn for sending me, as yet unpubblished, his book
on The Aim and the Argument of Aristotle's Metaphysics, in which he shows how
this case would produce a sort of absurd infinite regress.
14 Aristot. Metaph. I I, 1052 b I6-18.
15 Ibid., 1053'24-27.
16 Geach [1980], p.63; Dummett [1981], p.547; Wright [1983], p.3.

17 Austin [1970], pp.27-28.
18 Austin [1970], p.7!.

19 Cf. Walzer and Mingay (eds.) [1991]. This is the text of all the manuscripts,
while 'en pasin' is an emendation made by Bonitz.
20 Hintikka [1959], p.142.
21 This interpretation was in part suggested to me by Paolo Fait in a seminar on
Aristotle held in Padua on April 2000. It is present also in Bumyeat [1979].The
commentary of Frede and Patzig ((eds.) [1988] II, pp.70-71) is clearly oriented in
favour of the identification of the two uses.
22 Owen [1960], p.184 and n.16. The "reductionist" tendency of these expres-
sions was strongly criticised by Leszl [1970]. It was also corrected by Owen himself
in his later articles.
23 The substance is called 'principle' (arche) in Metaph. r 2, 1003 b 6, and the
things said "in conformity with one" (kath 'hen) are distinguished from the things
said "in relation (pros) to one nature" in 1003 b I2-14.
24 This was noted by Leszl. On the interpretation of Owen see also Code [1996].
25 In the English translation of his article, published in Barnes et al.(eds.)
[1975], pp.33-49, Patzig corrects his interpretation, distinguishing the paronymity

from the 'focal meaning' and founding the unity of metaphysics only on the latter.
26 This has been noted also by an Italian scolar who shares the interpretation of
Frede, i.e. Donini [1995], p.1 O!.
27 Theophrastus, Metaph. Z, 5 a25-27(cf. Laks and Most (eds.) [1993]).
28 Owens [1978a], p.395, 457.
29 Cf. also Owens, [1978b].
30 I cannot document here this affirmation, for which I refer to Beierwaltes
[1972].
31 More than twenty years ago I wrote an article on this subject, which, though
republished in an important collection, has remained without replies: cf. Berti

[1978].
32 Cf. Madigan (ed.) [1999], p.ll!.
33 Cf. Madigan (ed.) [1999], p.113. Apparently, the Aristotelian criticism of the
existence of a being itself is shared by Owens [1973], p.21-35, esp. 33: "If the exis-
tence known in observable things is viewed as having a content within itself, it is left
open to Parmenides' relentless reasoning [ ... ] It leaves no room for anything else".
34 Cf. Ross (ed.) [1953], I, p.245.
SOME COMMENTS ON PROF. ENRICO
BERTI'S "BEING AND ESSENCE IN
CONTEMPORARY INTERPRETATIONS OF
ARISTOTLE"
David Charles
1. INTRODUCTION
In his interesting and wide-ranging paper Enrico Berti makes a number

of important exegetical claims. I shall mention a few of them:
(a) Aristotle did not separate the existential from the predica-
tive sense of 'is'. One possibility is that Aristotle analysed
'A exists' (in the case of substances) as 'A is one existent
substance'. Alternatively, he failed to distinguish clearly
between the existential and the predicative sense of 'is'.
(b) Since, according to Aristotle, the predicative sense of 'is' is
multiple, it follows (given (a» that ' ... exists' will also have
multiple senses.
(c) For Aristotle, being is not a genus. (Prof. Berti proposes a
new way of understanding one of Aristotle's argument in
favour of this claim.)
(d) According to Aristotle, all senses of 'is' are to be defined in
terms of the central, or focal, case: the being of substances.
(e) However, it is not the case that the being of all substances
is to be defined in terms of the being of a focal case of sub-
stance: that of pure actuality.
109
A. Bottani et al. (eds.). Individuals. Essence and Identity. 109-126.
(t) The relation between pure actuality and other substances is

that of efficient causation.
As a good Aristotelian, Prof. Berti accepts reputable scholarly opinion

in cases where he sees no objection to it. Thus, in discussing (a), (b)
and (d), he cites with approval such recent authorities as G .E.L. Owen
[1965], Jaakko Hintikka [1999], and Lesley Brown [1994]. However,
in discussing (e) and (t), he finds problems with certain suggestions
once made by Michael Frede [1987] concerning the unity of Aris-
totle's metaphysics.
Nor is Prof. Berti's interest solely exegetical. In discussing (b) and
(d), he expresses some sympathy for Aristotle's claim that 'exists'
(while not a chance homonym) is not univocal, and contrasts this with
the currently fashionable view which takes it to be univocal. As he
rightly emphasizes, Aristotle's views in this area are of continuing
philosophical interest. Some philosophers, notably Gilbert Ryle, John
Austin and David Wiggins, have sought to find an alternative to the
following options: either a term (e.g. 'exists', 'is good', 'is true') has a
univocal sense or it is a chance homonym. Aristotle's discussion of
focal meaning and analogy appears to offer an outline of precisely the
route required to escape from this limited range of options.
Take, as an example, one well-known modem discussion of 'ex-
ists'. In the sentences 'AI Gore exists' and 'unicorns exist', 'exists'
appears not to be a chance homonym. It is generally assumed that if a
term is not a chance homonym, it must be univocal and as such have
one semantic role. The difficulty is that in the first sentence cited 'ex-
ists' appears to be a first-order predicate true of individuals, while in
the second it appears to be a second-order predicate which says
(roughly) that a certain feature is instantiated (or, perhaps, that the
number of unicorns is not equal to zero). Many writers have assumed
that these appearances are misleading, and that 'exists' must play the
same semantic role in these two sentences. However, this assumption
has generated results which appear counterintuitive. Thus, some have
sought to assimilate the second sentence to the first by analysing 'Uni-
corns exist' as 'Some unicorns exist', taking the quantifier to range
over both mythical and real objects and interpreting 'exists' as a first-
order predicate. Others have sought to assimilate the first sentence to
David Charles, Some Comments on Prof Berti 111
the second by analyzing 'AI Gore' as shorthand for a description (e.g.:

'the man who stood as Democratic candidate in the American election
in 2000') and taking the sentence as a whole to say that the property
picked out by this description is instantiated. Neither move would be
required if 'exists' can fail to be univocal (by having more than one
semantic role) without being a chance homonym. Aristotle's semanti-
cal framework, with its interest in focal and analogical meaning, seems
to focus on just the right option, one which avoids both the Scylla of
univocality and the Charybdis of chance homonymity. The central is-
sue is: does his suggestion stand up to critical scrutiny?l
A full discussion of this topic alone, let alone of all the issues
raised by Prof Berti, would require a lengthy paper. In this short essay,
I shall focus exclusively on his first two claims: (a) and (b). If these
are correct, Aristotle's suggestion that 'exists' is a case of focal
meaning rests crucially on his failure (or unwillingness) to distinguish
existential and predicative senses of the verb 'to be'. If so, his conten-
tion will only appeal to those who reject the commonly accepted mod-
em assumption that these two senses are to be distinguished. My
method will be indirect: I shall outline a position, drawing on some
Aristotelian texts, which maintains the multiplicity of existential
senses of 'exists' while separating the existential from the predicative
sense of 'is'. This position I shall call that of Cautious Aristotle. I shall
then argue that the evidence cited by Prof Berti, drawing on the work
of other writers, does not show that Aristotle's views on this topic
were non-identical with those of Cautious Aristotle. (There may, of
course, be further pieces of evidence which shows that the views of
Aristotle and Cautious Aristotle are non-identical. But investigation of
these lies outside the scope of the present essay.) I shall conclude with
a few remarks on the more general philosophical issue raised by Aris-
totle's view.
2. CAUTIOUS ARISTOTLE INTRODUCED
The position of Cautious Aristotle can be defined by the following

claims.
(1) In Post. An. Bl and 2 (89 a33, 90b 12ff), there is talk of sub-
stances such as the moon, the sun, and the earth as 'being without
qualification (haplos)' which is contrasted with their 'being some-
thing', such as white, dark etc. This discussion appears, at first sight,
to mark the distinction as between 'A exists' and 'A is F', the existen-
tial and the predicative uses of the word 'to be'. At least, this is the
distinction that Cautious Aristotle is committed to maintaining. Else-
where, Aristotle seems to draw precisely this distinction when he dis-
cusses under what conditions 'a is F' entails that 'a is'. (Soph. El.
1.167a l-2, 180a36-8: see de Int. 21 a25-33). For the inference in ques-
tion seems to be one from the predicative to the existential sense of 'to
be'.
(2) According to Cautious Aristotle, there will be a variety of dif-

ferent ways in which sentences of the form
' ... exists'
may be true depending on whether the blank is filled by a substance-

term or a quality-term. The term 'exists' will be used with a different
sense depending on which of the ways is invoked. Differences in the
sense of 'exists' correspond to differences in the ways in which sub-
stances and qualities exist. Indeed, there are as many senses of exis-
tence as there are different categories. In the case of terms denoting
substances, 'A exists' will be true if and only if A is a substance. In
other cases 'B will exist' if and only if B is a quality. In the latter
cases, B will exist if and only if B inheres in some substance in the
way that it is appropriate for qualities to inhere in substances. So un-
derstood, what it is for a quality to exist will be different from what it
is for a substance to exist. (This is what Cautious Aristotle means by
the condensed argument in Metaph. 11, l017a22-27. For he holds that
there are as many existential senses of the verb 'to be' as there are dif-
ferent types categorial predication of the form 'A is a substance', 'A is
a quality' etc.)
(3) Notwithstanding (2), for Cautious Aristotle, the verb 'to exist'
is used in its primary sense in the case of substances. If so, in his view,
the following will be true:
If 'A exists without qualification' is true, and 'exists' is used in its

primary sense, A is a substance.
Similarly, 'exists' will be used in a derivative sense in the case of
qualities. Thus, if 'B exists without qualification' is true, and 'exists' is
used in its relevant derivative sense, B is a quality. The case of sub-
stance is primary to that of quality because one cannot state what it is
for any quality to exist without mentioning that some substance exists,
but one can state what it is for a substance to exist without mentioning
the existence of any quality. In this way, 'exists' exhibits focal mean-
ing. (This is how Cautious Aristotle intends his discussion in Metaph.
r, lO03 a33-blO and Metaph. Z 1, 1028a30, cited by Prof. Berti, to be
understood. )
(4) Cautious Aristotle does not infer from (1) and (2) that 'A ex-
ists' is always incomplete and needs to be understood as (e.g.) 'A is
one substance'. It will, of course, be true that if 'A exists' is true, and
'A' is used in its primary sense, then 'A is one substance' will be true.
But the latter does not amount to a semantic analysis of 'A exists'. It is
rather something which will necessarily be true if 'A exists' is true
and 'exists' is used in its primary sense. While A can only exist in the
primary sense if A is a substance, the latter (metaphysical) claim is not
part of the meaning of 'A exists'. In this way, there can be several
senses of 'exists' corresponding to the different categories of existents,
without requiring any conflation of existential and predicative senses
of the verb 'to be.'
(5) In the light of (4), Cautious Aristotle infers from (1) and (2)
that one should separate the question 'Does A exist?' from the ques-
tions 'Is A a substance?' or 'Is A one substance?' For, according to
Cautious Aristotle, the second type of question would most naturally
be taken to mean (e.g.):
(i) 'Given that A exists, is it a substance?'
However, since (i) is not an existential question, it must be a different
question from 'Does A exist?' If so, 'A exists' cannot be analyzed se-
mantically as 'A is a substance'. If, however, the question 'Is A a sub-
stance' is taken to mean
(ii) 'Is there a substance which is identical with A?'

this question embeds an existential question in the phrase 'is there a
.. .', and so does not constitute an alternative to the existential sense of
'exists'. (ii), in effect, asks 'Does there exist one substance which is
identical with AT. However, so understood, (ii), so far from doing
away with the existential sense of 'to be', distinguishes this from the
predicative sense of 'is', and presupposes the existential sense in rais-
ing a further predicative question: is the substance which exists identi-
cal with A?
Cautious Aristotle is rightly so-called for one crucial reason. He is not
attempting to give a reductive semantic analysis of the verb 'to be'.
Perhaps he despairs of being able to carry through this type of reduc-
tive operation. Perhaps he is merely non-committal. But either way he
is not engaged in providing an analysis of this basic notion. He treats it
rather as a primitive notion which can be elucidated in terms of (but
not reduced to) other basic notions. He is not engaged in the ambitious
task of analyzing 'A exists' in terms of 'A is (e.g.) a substance'.
Rather, he is seeking to defend the claim that 'exists' has different but
related senses in the sentences 'AI Gore exists', 'Unicorns exist' and
'Folly exists', while separating existential and predicative senses of the
verb 'to be'. In the latter case, 'Folly exists' will be true if and only if
the relevant quality belongs to some individuals. (This latter claim, it
should be noted, is not an attempt to reduce the existence of Folly to
the existence of foolish individuals. For Cautious Aristotle requires not
merely that such individuals exist, but also that Folly belongs to
them).2
3. IS ARISTOTLE IDENTICAL WITH CAUTIOUS

ARISTOTLE? THE CASE AGAINST
Many will say that, despite the evidence so far noted, Aristotle's views
cannot be those expressed by Cautious Aristotle. Indeed, it has fre-
quently been suggested, at least since G.E.L. Owen's influential paper
[1965], that for Aristotle to be is, in the final analysis, to be a sub-
stance (or a substance of a given type) or a quality (or a quality of a

given type) or a quantity (or a quantity of a given type), etc.. Thus, his
claim that 'to be' is said in many ways rests on his favoured analysis
of 'to be' as 'to be (e.g.) a substance. I shall consider four reasons that
have been taken to show that this is the case. None is, in my view,
compelling.
Reason 1. In de Anima II 1, 41S b 13, Aristotle writes that 'for ani-
mals to be is to be alive'. Surely, it will be said, in writing this he
commits himself to the claim that 'to be' when predicated of an animal
means 'to be alive.'
Reply. The conclusion does not follow. Even Cautious Aristotle

can agree that the sentence
'Animal A exists' is true
if and only if
'Animal A is alive.'
Indeed, he might go further and say that the first sentence is true in
virtue of the second being true. This might well be a profound meta-
physical truth expressible by the claim:
'For animals, existence consists in their being alive.'
But this latter claim need not be part of the meaning of the term 'ex-
ists'. Indeed, one might come to know that animal A exists, and then
go on to discover the existence of an animal is grounded in its being
alive. The latter claim would only be grasped when one came to know
what it is to be an animal: the essence of animality. For only then will
one come to know the relevant metaphysical truth.
Aristotle could incautiously conclude that the sentence:
'Animal A exists'
is to be analysed as
'Animal A is alive,'
only if he believed that whatever would have to be true if A existed

should be included in the semantic analysis of 'A exists'. But this can-
not be his view. For he is clear that in many cases we can grasp the
significance of a term, discover that it signifies something that exists
and then go on to find out the essence of the kind in question (Post.
An.. B 10, 93 b29-33). While it may well be that, in Aristotle's view,
difference in significance between different types of object or property
(such as pride: Post. An. B 13, 97 b 13ff) rests on relevant metaphysical
differences in their respective essences, it does not follow that the sig-
nificance of the term itself incorporates any reference to their essences.
Reason 2. In Metaph. r 2, 1003 b22-30, Aristotle appears to make
the following claims:
1. To be and to be one are the same (i.e. they accompany one an-
other but are not expressed by the same definition).
2. For one man and existent man and man are the same thing and
the expressions 'one man', 'man' and 'existent man' do not 'show'
anything different. (For man, existent man, and one man come to be at
the same time and pass away at the same time).
Ifhe is suggesting that 'one man', 'man' and 'existent man' 'show' the
same thing, is he not thereby identifying the existential sense, the
predicative and identificatory sense of being? For if 'being' and 'one'
had differed in sense, surely this difference would have shown up in
the complex expression of which they are parts. If so, these expres-
sions would have had to mean something different from each other.
Hence, if these two expressions do not differ in meaning (according to
Aristotle), this must be because 'being' and 'one' (in his view) have
the same meaning. Similar arguments apply in the case of the predicate
'being a man', if this is taken to mean the same as 'existent man.' If
so, Aristotle must have collapsed the distinction between the existen-
tial and predicative senses of being.
Reply. Care needs to be taken with Aristotle's terminology here.
He seems to be suggesting that (at the ontological level) one man,
man, and existent man are 'the same, but different in definition.' Cor-
respondingly, at the linguistic level, the expressions refer to the same
thing but do so in a different way. In similar fashion, the expression

'the road from A to B' refers to the same thing as 'the road from B to
A', but does so in a different way (Physics r 3, 202b 13ff). Does it fol-
low from this fact alone that the expressions 'one man', 'man' and
'existent man' have the same sense? Clearly not, unless one also ac-
cepts (as Aristotle does not) that the expression 'the road from A to B'
has the same sense as 'the road from B to A'. For while all might agree
that these latter expressions have the same referent, few would agree
that they have the same sense. Similarly while all might agree that the
expressions 'one man', 'man' and 'existent man' specify the same ob-
ject in a given context (e.g. Socrates), few would infer from this that
the expressions have the same meaning. Rather, as Aristotle implies, it
will be true that if one expression is correctly applied, so is the other,
not that the expressions have the same meaning. It merely follows that
if it is true to claim that 'Socrates is a man' it is also true to claim that
'Socrates is an existent man'. Indeed, if one claim is accepted, the sec-
ond is redundant ("epanadiploumenon").
Aristotle would only be committed to the claim that these expres-
sions have the same meaning if he held the following semantical
claim:
If any expressions 'a' and 'b' always refer to the same things, they
have the same meaning.
But why should be hold this? Why cannot he hold that the complex
expressions 'the road from A to B' and 'the road from B to A' have dif-
ferent meanings, even if they are necessarily co-referential? Similarly
with 'two' and 'the product of one and one'. It appears that Aristotle,
in general, takes two expressions to have the same significance if the
accounts corresponding to the expressions are the same (see Cat. 1a6_
10) But the accounts corresponding to 'one man', 'man' and 'existent
man' are said in Metaph. r 2, 1003 b24 to be different, as are those cor-
responding to 'the road from A to B' and 'the road from B to A'. If so,
as one would expect, the relevant expressions will differ in significa-
tion (meaning) although they have the same extension.
This point can be made in a different way. Let us stipulate, for the
sake of argument, that the views of Aristotle are identical with those of
Cautious Aristotle. The latter holds that there are distinct senses of
being, corresponding to the existential, predicative and identificatory

senses. But, he could still hold that the expressions 'man', 'one man',
and 'existent man' show the same thing, or, as he might have said,
point to the same objects. He need see no inconsistency between this
claim and his attachment to the view that there are distinct senses of
being. Indeed, Cautious Aristotle might accept (plausibly enough) that
it is redundant to add 'existing' (or 'actual') to the sentence 'He is one
man', or 'one' to the sentence 'He is a man'. But this in no way shows
that the sense of 'A man exists' is the same as 'A man is one thing' or
'A man is an existing thing'. Indeed, one who distinguished existential
and predicative senses of the verb 'to be' could make precisely Aris-
totle's point and find the expressions 'one thing' and 'an existing
thing' redundant in these contexts.
Reason 3. While in Post. An. B 2, Aristotle does indeed appear to
speak of existence (being without qualification/hap/os), appearances
must be deceptive since he goes on to request that statements con-
cerning being hap/os are proved by finding a middle term. But how
can this be? Surely, the existence of substances cannot be established
by finding a middle term? Substances exist in virtue of what they are
in themselves, not because something else can be predicated of them.
Some have concluded from this that Aristotle is not asking the ques-
tion
'Does man exist?'
but is focusing rather on some question of predicative form, such as
'Is man an Fl'
where 'F' might be (e.g.) 'substance' or 'animal'. If so, it is suggested,
there could be a middle term, which explains why being F belongs to
man. So, if search for a middle term is to be possible in this case, the
statement of being hap/os cannot be a statement of existence.
Rep/y. There is an alternative interpretation of these passages,
more in line with Aristotle's procedure later in Post. An. in B.8 and
elsewhere in Metaph. Z 17. When confronted by an existence question
of the form
'Does A exist?'
he frequently reformulates question using the account of what the

name 'A' signifies. Thus, in the case of questions concerning the exis-
tence of thunder, he would begin his search into the existence of the
phenomenon by focusing on the account of what 'thunder' signifies
(e.g. 'a kind of noise in the clouds').3 At this stage, he will ask the
question
'Does noise belong in the clouds?'
and answer it (in the optimal case) by finding a basic (or immediate)
middle term (e.g.: 'fire being quenched') which specifies the essence
of thunder. If he finds such a middle term, he will know that there is
one unified phenomenon, thunder, which exists. In this way one will
have established the existence of thunder as a kind.
The question
'Does noise belong on the clouds?'
is a question about the connection between two entities in the world:

noise and the clouds. In it, the expressions introduced in the account of
what the term 'thunder' signifies are used to refer to objects in the
world. Thus, the question is distinct from the questions:
Is the expression 'noise in the clouds' true of any objects?
or
Are there any objects which satisfy the linguistic expression 'noise
in the clouds' .
The latter may be answered by saying

There are (es gibtlil y a) objects which satisfy the linguistic ex-
pression/fall under a given concept
where this answer requires us to invoke (e.g.) concepts and a general
sense of existence (corresponding to that captured by the existential
quantifier). But Aristotle's question requires none of this machinery to
answer it. It can be answered (as in Post. An. B 8, 93 a29-31) while
staying at the level of objects (and properties) and noting their inter-
connections.
Aristotle makes several moves to articulate the question:
'Does man exist?'
In Post. Anal. B 8, the question can be phrased using (e.g.) 'a specific
type of animal' or perhaps as elsewhere 'biped animal' (see Metaph. r
4, l006 a3lff), as the account of what 'man' signifies. In this case, to
establish that man exists is to establish that (e.g.) bipedality belongs to
animal by finding as a middle term one which specifies (in the optimal
case) the essence of man: e.g. being rational. In Metaph. Z.17, the ar-
ticulation of the question appears to be rather
'Why is matter of this type some definite thing?'
or more precisely
'Why is matter of this type arranged thus?'
to which the answer would be given in terms of what it is to be a man
(i.e. man's essence). For this explains why matter of this type is ar-
ranged as it is in this case.
Aristotle's aim is to regiment existence questions in such a way
that he can answer them using the resources of demonstration, and in
this way prove that (e.g.) man is indeed one properly unified kind. But
none of this requires him to say that the statement
'Man is without qualification/haplos'
must be elliptical for 'Man is an F'. It merely shows that the way in
which Aristotle seeks to prove (in his favoured demonstrative form)
that man exists rests on his complex accounts of what the term signi-
fies. For the latter introduce claims of the right form to be proved by
the presence of a middle term. In this way, scientific proof of the ex-
istence of a kind can depend on finding a middle term even though the
statement of existence is of the straightforward existential form:
'Man exists.'
Reason 4. In Meta H 2, I042 b I5-1043 a7, Aristotle states that for a

threshold to be is for it to be situated thus, and suggests (if the cases
are parallel) that for ice to be is for it to be solidified in a certain way.
If so, it might seem that to ask the question
'Does ice exist?'
is to ask the question
'Does being solidified in a certain way belong to ice?'
Here, too, it appears that an apparently existential question is to be re-
placed by some form of predicative one. Surely, here, too, the expla-
nation is that Aristotle lacks the purely existential form of being, and
thus has to rephrase apparently existential questions in predicative
form.
Reply. This passage also needs to be handled with considerable
care. For Aristotle goes on to give a fuller specification of 'ice' as 'so-
lidified water' (1043 a IOf). He then can ask the existence question
'Does ice exist?'
in a form closer to that suggested for 'thunder' in the Analytics:
'Does being solidified belong to water?'
If he can answer this question in the affirmative, he establish that ice
exists.
The point is a general one. If Aristotle wishes to establish that
(e.g.) ice exists in a demonstrative way, he seeks to establish that a
certain predicative sentence is true. Thus, if the account of what ice
signifies mentions 'frozen water', he will (in demonstrating that ice
exists) demonstrate that
'Being frozen belongs to water.'
For in demonstration one requires a (two-term) proposition which can
be proved and a further term ('the middle term ') from which the dem-
onstration proceeds. It will be true that
'Ice exists'
if and only if it is true that

'Being frozen belongs to water.'
But it does not follow from this that the latter sentence is a (semantic)
analysis of the sentence
'Ice exists.'
Indeed, it is not such an analysis since it presupposes (while 'Ice ex-
ists' does not) that water exists.
It is important to give Aristotle's account of what the term signi-
fies in the Analytics-style just indicated. For, if he had followed an al-
ternative route, his method would have been in danger of collapse.
Thus, if he had taken 'ice' to signify the same as 'solidified water' and
then asked the question:
'Does being solidified belong to solidified water?'
the affirmative answer
'Being solidified belongs to solidified water'
would have been tautologous, and the negative answer
'Being solidified does not belong to solidified water'
would have been self-contradictory. However, in raising such ques-
tions in the Posterior Analytics Aristotle does not include the term
'thunder' in the account of what the term 'thunder' signifies. Rather,
the latter is standardly given by such phrases as 'noise in the clouds'. It
therefore seems plausible to find a similar regimentation at work in
Metaphysics H 2. Nor is there any need to see a conflation of existen-
tial and predicative senses in either case. Rather in both a paraphrase is
suggested, apparently based on an account of what the term in question
signifies. Using this paraphrase, Aristotle proceeds to establish the ex-
istence of the required phenomenon by proving that (e.g.) noise be-
longs in the clouds or that being solidified belongs to water. On this
basis, Aristotle can claim that what it is for thunder to exist is for noise
to occur in the clouds, and (in parallel fashion) what it is for ice to ex-
ist is for water to be solidified. (Sometimes, it is true, Aristotle talks as
if he is proving that thunder exists by establishing that thunder be-

longs in the clouds (Post. An. B 8, 93 b lOff) or as if he is establishing
that ice exists by establishing that ice is solidified (Metaph. H 2,
l042 b28. But these formulations are, it appears, shorthand for the
longer and more accurate paraphrases, involving respectively noise
and water).
Owen [1965] suggested an alternative way of understanding Ar-
istotle's account of such claims as 'thunder exists' or 'ice exists.' Ac-
cording to Owen, these claims are sometimes to be understood (in Ar-
istotle's account) as claiming that the concepts 'thunder' or 'ice' are
presently instantiated. If so, an investigator will proceed by first find-
ing an account of what the term 'thunder' signifies, and then asking
whether there are currently any objects of which that account is true.
But, if I am correct, Aristotle did not need to make any such moves.
Indeed, as I have suggested above, is there no indication that he made
them in his central discussion of these issues in Post. An. B 8, 93az8ff.
Owen was forced into his alternative analysis because he did not pay
sufficient attention to the way in which, in the Anaiytics, Aristotle used
accounts of what terms signify to articulate questions about existence
so as to make them susceptible to proof in his favoured demonstrative
form.
Interim conclusion. The four pieces of evidence, so far examined,
have persuaded some that Aristotle failed to separate the existential
and predicative senses of the verb 'to be'. But in no case is their con-
clusion secured. Indeed, if Aristotle followed his Analytics-style ac-
count (as I have interpreted it) in these contexts, he would have been
able (without difficulty) to distinguish in each case these senses of
being. For all that has been said so far, Aristotle's views on this issue
can be same as those of Cautious Aristotle. In different contexts, Ar-
istotle may have focused on claims which will be true if (e.g.) man
exists (e.g.: man is a substance, man is a biped animal, being a biped
belongs to animal etc.). But none of these amount to an analysis of the
term 'exists.' They are rather a variety of claims, useful for different
purposes, which will be true if man exists.
If this line of argument can be sustained, Aristotle's discussion of
these issues will constitute an advance over Plato's. For interlocutors
in Plato's dialogues frequently ask what might be taken as existential

questions in predicative form:
'Do you call something A?' (kale is ti to A): see, for example,
Cratylus 38S b2, 421 as, Alcibiades 128b S, Gorgias 464a l.
Similarly, in the Protagoras, Socrates asks the apparently predicative
question
'Do you say that a certain thing is justice?' (330 d) (phate pragma
ti einai ten dikaiosunen?)
when he appears to wish to raise the existential question:
'Do you think that there is such a thing as justice?'
In these passages, Plato appears to conflate existential and predicative
questions in a manner which suggests that he has not separated exis-
tential and predicative senses of the verb 'to be.' However, if I am cor-
rect, Aristotle in the Analytics successfully distinguished questions
about the significance of the term 'A' from ones about the existence
about A's existence in just the way required to ask such existential
questions as:
'Does justice exist?' 'Does A exist?'
Perhaps he saw himself (here as elsewhere) as freeing himself from
Plato's mistakes about being.
4. WAS CAUTIOUS ARISTOTLE CORRECT TO DETECT

DIFFERENT SENSES OF THE VERB 'TO EXIST'?
Cautious Aristotle suggested that there are different, but focally re-
lated, senses of the existential verb 'to be' corresponding to different
types of categorial claims such as 'A is a substance', 'B is a quality,
etc .. In his view, there have to be different senses of 'exists' since
there are major differences in the ways in sentences such as 'A exist'
and 'B exist' are true. While he did not analyze 'A exists' as 'A is a
substance', he took differences in sense of the verb 'to be' to reflect

differences in the ways in which entities in different categories exist.
Some will say that Cautious Aristotle is not cautious enough. They
will object that he should have rested content with the claim that there
are different ways in which substances and qualities exist, and not
concluded from this that there are different senses of the verb 'to ex-
ist'. An analogy may serve to bring out the objectors' point. All will
agree that there are many ways in which one can travel to Rome from
Milan. One may travel by plane, by train, by donkey, on foot. Perhaps
one could, given enough time, make the journey by boat. But the pres-
ence of these different ways of traveling to Rome do not show that
there are different senses of the verb 'to travel'. Similarly, it will be
said, there are many different ways in which different types of entity
may enter into the realms of existence (some by the substance route,
some by the quality route, etc.), but all will in the end arrive in the
same place. They will all be existents. In each case the number of the
relevant existents will be greater than zero. Why should 'exists' differ
in sense while 'traveling' does not?
This objection, as it stands, is not decisive. For the analogy rests
on the assumption that, just as Rome is the endpoint of all the journeys
mentioned, so there is one thing, existence, which is achieved by all
existents. But this is precisely what Cautious Aristotle is concerned to
deny. In his view, what it is for a substance to exist will be different in
kind from what it is for a quality to exist. While the existence of the
quality depends on its belonging to some substance, the existence of a
substance does not depend on its belonging to anything. This differ-
ence in metaphysical understanding of what it is for qualities and sub-
stances to exist is reflected (according to Cautious Aristotle) in the dif-
fering senses of the verb 'to exist' in these cases.
A full defense of Cautious Aristotle's view lies outside the scope
of this paper. It should be noted, however, that his approach will ap-
peal to (at least) two groups of philosophers. The first group will con-
sist of those who share the dissatisfaction noted at the beginning of this
essay with the suggested univocal analyses of 'exists'. The second will
be made up of those who think that a full account of what we under-
stand in grasping terms such as 'exist' should reflect our understanding
of the metaphysical grounding of existence claims. For the latter
group, it will not be enough to establish univocality by pointing to

similar patterns of inference involving 'exists' (such as the claim that
from 'Fa' we can infer that both a and the property F exists). For they
will seek some understanding of why these inferences are valid, and
for this they will turn to a metaphysical account of what it is for sub-
stances or properties to exist. The numbers of philosophers in these
two (non-exclusive) groups is, and should be, considerably greater
than zero. 4
Notes
For discussion of this issue, see Williams (1996]
I
This point is well made by Barnes [1995], p.81 ff.
2
3 For further discussion of these issues, see Charles [2000].
4 I am indebted to Adam Beresford, Paolo Crivelli, Kit Fine and Peter van In-
wagen for their comments on an earlier version ofthis paper.
EXISTENCE, IDENTITY AND AN
ARISTOTELIAN TRADITION
Alex Orenstein
Shouldn't we expect a proper account of existence sentences to do jus-

tice to claims involving empty names, especially those denying the
existence of objects such as Pegasus and Vulcan? Pegasus is a product
of ancient mythology and Vulcan an unnecessary and discarded posit
of nineteenth century astronomy. A test of such accounts is how they
deal with the problem Quine named 'Plato's Beard'. I begin by posing
the Plato's Beard problem as it appears in Quine and then as involving
identity.! David Wiggins extended the scope of this problem. Quine's
version concerned objects that do not exist. Wiggins extended it to
objects that exist but might not have done so. I discuss Wiggins' solu-
tion and one of two conceptions of "free logic". I then suggest a solu-
tion along the lines of a second conception of free logic that has its
roots in an older Aristotelian tradition.
1. QUINE: PLATO'S BEARD AND IDENTITY
Quine opens "On What There Is" by endorsing the claim that every-
thing exists. The question:
What is there? [ ... ] can be answered [ ... ] in a word - 'Everything' - and
everyone wiJI accept this answer as true.
127
A. Bottani et aL (eds.), Individuals, Essence and Identity, 127-149.
After three paragraphs on non-existence, he tells us

This is the old Platonic riddle of nonbeing. Nonbeing must in some sense
be, otherwise what is it that there is not? This tangled doctrine might be
nicknamed Plato's Beard [... ] (Quine, [1948], pp.I-2)
The puzzle bears on vacuous/empty names and inferences involving

them such as:
(1) PegasusNulcan does not exist
(2) Something does not exist. (Quine [1960], p.1 76)
Using identity to explicate singular existentials, e.g., 'Julius Caesar
exists' is explicated as (3y)(y = J.C.), the problem can be put as fol-
lows:
(I') -,(3y) (y = Pegasus/Vulcan)
(2') (3x)-,(3y) (y = x)
Interpreting the English quantifier, 'something' and the first order

quantifiers '(3x)', '(3y)' as expressing existence, the conclusion ap-
pears as what Quine elsewhere calls "a contradiction in terms":
There exists something that does not exist.
Moreover, (2) contradicts the opening claim that everything exists.
When we represent (2) by using identity as (2'), we reveal another as-
pect of the riddle. The conclusion (2')
(3x)-,(3y) (y = x)
is inconsistent with what Quine and most of his readers take as an ax-
iom of identity theory, its total reflexivity:
(\ix) (x = x).
From this axiom we can derive '(\ix)(3y) (y = x)' which is an identity
rendering of the English 'Everything exists'. Doing this gives further
support to Quine's opening endorsement. Furthermore, (\ix)(3y) (y =
x) is equivalent to
-,(3x)-,(3y) (y = x)
Alex Orenstein, Existence, Identity, and an Aristotelian Tradition 129
and the latter is the denial of the conclusion (2'). The identity explica-
tion of the issues intensifies the problem.
In a word, the conclusion is a logical falsehood given the total re-
flexivity of identity and the standard rules of predicate logic. In using
identity we back up Quine's English language claim that everything
exists, and that the conclusion is a "contradiction in terms".
Much the same problem was presented earlier in Mathematical
Logic (p .150). (Identity renderings of the relevant claims inserted by
me appear in square brackets and I have used bold italics for a clause
that we will return to in Part 3.)
To say that something does not exist [(3x) ....{3y) (y = x)], or that there is
something which is not, is clearly a contradiction in terms; hence '(Vx) (x
exists), [(Vx)(3y) (y = x)] must be true. Moreover we should expect leave
to put any primitive name of our language for the'x' of any matrix' .. .x ... ',
and to infer the resulting singular statement from '(Vx)(x exists)'; it is dif-
ficult to contemplate any alternative logical rule for reasoning with
names. But this rule of inference leads from the truth '(Vx) (x exists), [... ]
to the controversial conclusion 'God exists' [(3y) (y = God)] and the false
conclusion 'Pegasus exists' [(3y)(y = Pegasus)] [... ]. The atheist seems
called upon to repudiate the very name 'God', thus depriving himself of
vocabulary in which to affirm his atheism, and those of us who disbelieve
in Pegasus would seem to be in a similar position.
2. WIGGINS: EXTENDING THE PLATO'S BEARD PROBLEM
For Quine, the Plato's Beard problem arises when we allow for
empty/vacuous names in predicate logic with identity and acknowledge
that some denials of singular existentials express contingent truths. In
his paper "The Kant-Frege-Russell View of Existence: towards the
Rehabilitation of the Second-Level View"2, David Wiggins shows how
the problem arises even when one does not allow empty names or true
denials of singular existentials.
Wiggins presents a revisionist Fregean account of singular exis-
tentials. He disagrees with Frege and Russell, who went to some pains
to argue that singular existentials, such as 'Julius Caesar exists', are
meaningless. Wiggins does not say why he disagrees with Frege and
Russell. I assume that unlike the climate of opinion surrounding the
acceptance of Frege and Russell's views, today Wiggins and most of
us prefer accounts of logical form which are more accommodating of
natural language locutions. It is simply too high handed to dismiss sin-
gular existentials as being meaningless. 3 Aside from this revision,
Wiggins wishes to remain true to Frege's central insights. While
holding that 'J.e. exists' expresses a contingent truth, central Fregean
themes are adhered to.
He treats singular existentials in terms of Frege's view of exis-
tence and of quantification as second level property/concept instantia-
tion. Wiggins quasi-fuses the identity sign '=' with the singular term
'J.C.', and treats '[= J.e.]' as representing a Haecceitian property or
individual concept. The English 'J.e. exists' is then accommodated as
the Fregean higher level property instantiation (3x) x[= J.e.]. Saying
that Julius Caesar exists is to be thought of as saying that the property
of being identical with Julius Caesar is instantiated. Wiggins is uni-
formly employing Frege's views of existence and existential quantifi-
cation as property instantiation. For Frege, to state that horses exist is
to hold that the property or concept of being a horse is instantiated.
Wiggins is merely extending this property instantiation view of exis-
tence and quantification to the Haecceity or quasi-Quinian like prop-
erty of being identical with J.C. 4
So far so good. Wiggins has made his case for accommodating
singular existentials as second level Fregean property instantiation. But
questions remain concerning those denials of singular existentials
Quine and others consider. Wiggins examines the contingently false 'It
is not the case that lC. exists'. But he does not in the same vein con-
sider prima facie contingently true denials of singular existentials such
as: 'Vulcan/Pegasus does not exist'.
He takes an object dependent approach to empty names. Object
dependent accounts are not so much a solution to Quine'S original
puzzle for empty names as a way of avoiding it by not allowing any
significant role to empty names. On these views, empty names do not
play any role in logic and argumentation. All the customary principles
of logic stay in place. Empty names simply do not enter the picture.
They and the sentences containing them do not occasion revising stan-
dard logic. The sentence 'Vulcan does not exist' does not furnish a
truth vehicle let alone a truth. s Bivalence is not in question. There is
nothing to be true, false, or have some other value.
As I see it, object dependence comes in several forms. For Evans
and Salmon (though for different reasons), sentences with empty
names are meaningless and do not provide truth vehicles. For Straw-
son, they are meaningful but do not provide us with a truth vehicle (are
not used to make statements). Wiggins holds that the sentence 'Vulcan
does not exist' fails to express a truth. 6
By restricting himself to something like statements as truth vehi-
cles, the strings 'Vulcan does not exist' and its identity predicate logic
correlate '-{3x) (x = Vulcan)' never have to be taken account of.
There is no problem of empty names since the principles of logic do
not apply to sentences containing them. The situation is reminiscent of
the net with spaces between the netting large enough to have no effect
on fish of a smaller size.
While Wiggins does not have to worry about reasoning from 'Eve-
rything exists' to 'Vulcan exists', he does deal with a related problem.
As we have seen, if one accepts the total reflexivity of identity, '(Vx)
(x = x)" as a logical truth, then (given standard predicate logic and
representing 'a exists' as '(3y) (y = a)') the following are theorems and
in that sense would also be considered logical truths:
(1) (Vx)(3y) (y =x), i.e., Everything exists.
(2) (3y) (y = Julius Caesar), i.e., Julius Caesar exists.
The principle of necessitation sanctions a transition from non-modal to
modal logic. It states that
If A is logically true, then it is necessary that A.
Given Necessitation, the following, (3), is a theorem of modal logic

and thereby considered a truth of modal logic:
(3) Nec(3y) (y = Julius Caesar), i.e., It is necessary that Julius Cae-
sar exists.
Since Wiggins' object dependent view does not recognize 'Pega-
susNulcan does not exist' as a truth vehicle, the original Plato's Beard
problem does not arise. Wiggins, though, allows that 'Julius Caesar
exists' expresses a contingent truth. He presents its contingent status
modally:
Julius Caesar might not have existed: Pos--.(3y) (y = Julius Cae-
sar),
and this clashes with (3) Nec(3y) (y = Julius Caesar).

Wiggins suggests a solution to this problem. He would have us
abandon the principle of necessitation for existential/particular gener-
alization: (3) would not follow from total reflexivity, or from (1) or
(2). There would appear to be no case for 3 being true.
However, Wiggins' solution is unacceptable. He is treating a
symptom and not the cause. The problem does not arise in connection
with a transition from non-modal to modal logic and at that point with
existential generalization. It can be stated solely in terms of non-
modal predicate logic with identity. That J.e. exists follows from the
assumed theorem/logical truth that everything exists, (Vx)(3y) (x = y),
and so would be considered a theorem of logic. As such it should be a
logical truth. Wiggins' intuition that 'J.C. exists' is "contingent" can
be captured without relying on modal notions. We can say that it is not
a logical truth or should not be derivable from logic alone. Other ways
of putting the point consist of saying that singular existentials are nei-
ther analytic in the sense of being logical truths nor a priori in the
sense of being provable by logic alone.
Moreover, limiting necessitation in this way seems ad hoc. Why
limit only existential/particular generalization? If existential/particular
generalization is analogous to disjunctive addition, why not limit dis-
junctive addition as well? What of versions of predicate logic where
existential/particular generalization is merely a derivative principle?
Which basic principle or principles would be revised as per necessita-
tion?
Before going on to the next section, let us comment on an aspect
of Wiggins' object dependent account of 'Vulcan does not exist'.
While he denies that such sentences provide us with truths, he does not
completely ignore them. They appear to occasion changing the subject
and focusing instead on the Evans inspired assertion that Vulcan does
not really exist. Instead of a straightforward denial of existence we are
served up a denial of a different sort; of a claim involving a 'really'

operator. When a waiter brings you the wrong dish, you tell him so. Is
there any reason for reacting differently to Wiggins' and Evans' sug-
gestion?
Their suggestion quite simply changes the subject. When the athe-
ist says that God does not exist, he is contradicting/denying the theist's
claim that God exists. The atheist is not merely making some sort of
incompatible claim about God not really existing. Even if we grant that
such 'really' claims can playa role in talk about existence, substituting
them for denials of existence is beside the point. 7
It is to the credit of current generations of philosophical logicians,
including Wiggins, to acknowledge (contra Frege and Russell) the
meaningfulness and contingent truth of singular existentials. This rec-
ognition of natural language claims should accord the same status to
true simple denials/negations of singular existentials. In section 4, it
will be argued that Frege's views can be adapted to this end. I will also
argue that the Fregean treatment of existence as property instantiation
Wiggins is trying to rehabilitate can also be employed in accounting
for the truth of 'Vulcan does not exist'.
Let us be candid. Why do Evans, Wiggins and others go to such
pains to make a case for adopting 'Vulcan does not really exist' in
place of 'Vulcan does not exist'? 'Vulcan does not really exist' really
is an ersatz creation being marketed by and for those who cannot get
over their need for the real thing: 'Vulcan does not exist'. They are
driven by philosophical theories to downgrade these simple denials of
existence. In Evans, his object dependence leads him to view sentences
with vacuous nouns as not expressing propositions. Similar results ac-
crue on holding, as do Salmon, Almog and others, the ordered n-tuple
view of Russellian singular propositions. Here the singular term's ref-
erent is actually part of the proposition and when the singular term is
vacuous and there is no referent, there is no proposition. Wiggins is
driven in part by his views on identifYing the referents of names.
Sainsbury has challenged Wiggins on this point and allows that 'Vul-
can does not exist' expresses a truth. Sainsbury would then have to
face Quine's as well as Wiggins' version of Plato's Beard. Those in
the grip of combinations of ideas of direct reference, rigid designation,
and causal theories of names sacrifice the varieties of singular sen-
tences without fully exploring the possibility of reconciling such theo-

ries with taking empty names seriously.
Wiggins and Evans should have simply discarded vacuous singular
existentials, and left the matter at that. They did not. They tried via
related sentences with a 'really' operator, to get some of the effects of
the simple denial of existence. Why not acknowledge as data that such
denials of existence yield contingent truths and then revise ones theo-
retical assumptions? They seem to feel the tug of the data that the
claim that Vulcan does not exist is really and truly a contingent truth.
Ironically the quote from Nietzsche which appears at the outset of
Wiggins' paper uses the claim that God does not exist in just those
terms. "The best atheist joke: God's only excuse is that he does not
exist." Stating true negative singular existentials is not at all rare. A
favorite of mine occurs in the Marx brother's film "A Night at the Op-
era". Chico "denies" the existence of sanity clauses by pointing out
that Santa Claus does not exist. He expresses this by saying: there ain't
no sanity clause.
3. TWO TRADITIONS OF FREE LOGIC
Some address problems concerning empty names in terms of what is

known as "free logic". The phrase was originally used to refer to logics
free of existence assumptions. As such it would allow for the presence
of empty/vacuous nouns. s Most people still use it in this way. I will
follow this practice in section 4. However, the phrase is used in a nar-
rower way by K. Lambert and E. Bencivenga, and they are two of the
foremost figures in this area. According to them, free logics are logics
that are free of existence assumptions in that they allow for empty
names, etc. and in addition interpret the '(3x)' quantifier as having ex-
istential force.
These logicians revise the ordinary quantifier rules. One is prohib-
ited from deriving
(3x) (Fx)
from
Fa
alone (as on the customary statement of the rule). An additional prem-
ise is required to the effect that a exists. They also reject the ordinary
rule of universal instantiation. In much of what follows what I say ap-
plies to these free logicians who state the additional premise in the
form of the identity: (3x) (x = a). Their revised universal instantiation
and existential/particular generalization rules appear as:
(Vx) (-:X-)
(3x) (x = a) i.e., a exists
therefore ~
and
~
(3x) (x = a)
therefore, (3x) (-:X-)

On these new predicate logic rules, 'Vulcan exists' and 'Julius CaesaL
exists' are not theorems oflogic and do not follow from '(Vx)(x = x)'.
Given these new rules, if one accepted that everything exists, and/or
that (Vx)(3y) (y = x), neither of the previous sentences would follow.
Wiggins does not adopt this free logic line, even though it would block
deriving 'Nec(3x) (x = J.C)'. He does not give his reasons.
Three considerations lead me to reject this narrower conception of
free logic: it appears to be driven by a philosophical thesis and not by
matters of logic, it is not conservative, and it goes too far in dealing
with the problem at hand.
I share Lambert's and Bencivenga's laudable goal of allowing for
empty names and true denials of singular existentials. I have misgiv-
ings on their, as I see it, narrow notion of free logic. Their conception
requires adopting the Frege-Russell-Quine philosophical thesis that
"existence is what existential quantification expresses." I have argued
elsewhere that this philosophical interpretation conflicts with logical
constraints on the quantifiers ("Is Existence What Existential Quantifi-
cation Expresses?"). Revising logical principles is not to be taken
lightly and should not be opted for on non-logical grounds, if more

conservative alternatives are available within logic. I take a broader
view of free logic requiring no more than that it be free of existence
assumptions. It is not necessary that the quantifiers have existential
force and consequently that the ordinary rules of quantification be re-
vised to deal with the Plato's Beard problem. In fact, there is a slogan
from traditional logic that existential import is not connected with
quantifiers but with the quality of propositions: "Quality, not quantity,
determines existential import" (see Orenstein, "Reconciling Aristotle
and Frege"). In section 4, I will say more about this alternative.
The second problem is that "narrow" free logic is not sufficiently
conservative. We have to pay a high price for such a free logic. When
the customary rules for quantifiers are revised, straightforward infer-
ences that have been accepted as correct are no longer valid as they
stand. The following otherwise valid arguments are judged to be inva-
lid:
Browny is a dog.
Something is a dog.
Everything is in space.
That chair is in space.

The universal instantiation rule applied in the previous example, of
which Quine said "it is difficult to contemplate any alternative logical
rule," is incorrect on this view. The standard expansions of universal
and particular generalizations as conjunctions and disjunctions would
be wrong. We would have to revise Behman's completeness proof for
monadic predicate logic. I surmise that ordinary duality relations
among quantifiers no longer would hold. These revisions are not dic-
tated by purely logical considerations but by the philosophical one of
linking existence and quantification.
The last reason for not following this brand of free logic is that in
dealing with Plato's Beard problems it goes too far in revising quanti-
fier rules. The formal versions of these problems depend on identity.
Identity is the source of the problem and not the background logic of
quantifiers.
4. A TRADITIONAL ARISTOTELIAN APPROACH
In this last section a traditional Aristotelian approach will be sketched.

Though it has its roots in traditional logic, it will be implemented in
terms of predicate logic. I will be ecumenical and, like Wiggins, where
possible adapt themes from Frege. The material is arranged in three
parts.
(1) Along revisionist Fregean lines, I indicate how one can make
room for empty names and allow that 'Vulcan does not exist' is a con-
tingent truth.
(2) I adopt one of Frege's main themes (existence is property in-
stantiation) and deny another (existence is linked with quantification).
Wiggins' title speaks of "The Kant-Frege-Russell View of Existence".
He and many others fail to see an important difference between Kant
and Frege. Frege links existence as property instantiation with exis-
tence as quantification. Kant holds a copula version of the existence as
property instantiation approach. However, there is good reason to think
that Kant (and those in the Aristotelian tradition of logic he worked in)
would deny that existence is a matter of generalization/quantification.
Kant's views are derived from an older treatment of existence. It con-
ceives of existence in terms of quality, an affirmative/negative distinc-
tion (linked with property instantiation), and not quantity, a univer-
sal/particular distinction. It has roots in the heritage of Terminist logic,
and is also a feature of a Lesniewskian approach.
(3) I call attention to another difference between the older tradition
and Quine, Wiggins, and "narrow" Free Logicians. The older traditon
denies the total reflexivity of identity. We shall replace it with the par-
tial reflexivity of identity. Revising total reflexivity goes to the heart of
the matter in Plato's Beard problems.
(1) How a Revisionist Fregean can concede that Vulcan Does Not
Exist
Let us adapt to our own purposes a view that Frege put forward in "On
Sense and Reference." He proposed there that a sentence containing
the vacuous name 'Odysseus' expresses a proposition. Frege assumed
that it did not have a truth value and that it should not be construed as
part of our serious truth telling use of language, but only as figuring in
stories, novels, etc .. The proposition/thought expressed presumably is
composed of an individual property/concept for 'Odysseus' and of a
property for the predicate 'while deeply asleep was disembarked at
Ithaca'. The individual property/concept for 'Odysseus' can be taken in
the spirit of Wiggins' suggestion. Just as Wiggins formed the individ-
ual concept for 'Julius Caesar': [= J.C.], one can form an individual
concept for the empty name 'Odysseus': [= Odysseus]. Empty individ-
ual concepts/properties were not a problem for Frege. He already al-
lowed for empty general properties, such as the property of being a
unicorn, or of being phlogiston.
We follow Frege and recognize a role for such propositions in sto-
ries and novels (where they might be governed by an implicit "in the
story" or "pretense" operator, or playa part in a game of make be-
lieve). But, unlike Frege, I propose that these propositions also playa
role in serious discourse. Such propositions/sentences can serve as
truth vehicles.
As a further step, I take it that in these serious non-fictional con-
texts, these sentences/propositions are false. This last step, treating
vacuous singular sentences/propositions as false, has precedents. It is
the view taken in the Lesniewskian tradition. It is also found in Kant,
and further back in Terminist logicians such as Ockham and Buridan.
Moreover, the English sentences, 'Vulcan is a planet', 'Vulcan does
exist' will receive the same truth value of falsity as they would on Rus-
sell's theory of descriptions. For Russell and those following him, such
sentences having the surface form of a singular sentence are translated
using his theory of descriptions into "existential" generalizations that
are false when the existence component is not satisfied.
Following Lambert's and Bencivenga's terminology, call this "the
negative solution." Such sentences-propositions are special cases of
the falsity of singular sentences/propositions. On an Aristotelian in-

spired correspondence theory of truth a sentence/proposition with a
vacuous/empty name is false. There is no reality for it to correspond to.
From the standpoint of standard model theoretic truth conditions, sin-
gular sentences are not true (and assuming bivalence are false), when
the object which is the semantic value of the singular term is not a
member of the set which is the semantic value of the predicate. The
usual way this applies is when there are semantic values for the singu-
lar term and for the predicate, and the first semantic value is not a
member of the second. This model theoretic account can readily be
adapted to consider singular sentences with empty names as false. If
one prefers propositions rather than sentences as truth vehicles, we can
achieve the same result. The singular proposition that the sentence
'Vulcan is a planet' expresses is composed of the property of being
Vulcan, and the property of being a planet. Singular propositions are
true when the individual that instantiates the singular term property is
one ofthe individuals instantiating the predicate property. Since 'being
Vulcan' is not instantiated by any individual, the singular proposition
is false.
A number of figures on the current scene have also considered and
endorsed the negative solution. Saul Kripke in lectures (though, so far
as I know, not in print) has argued that some of Strawson's line of ar-
guing for a neutral solution are equally compelling for the negative
solution. Steven Schiffer has recently talked of adopting such a view in
a forthcoming paper entitled "Meaning". Tyler Burge has given the
matter greater consideration than these others. However, Kripke, Schif-
fer and Burge are not interested (as I am) in maintaining a connection
with themes from classical and traditional Aristotelian accounts.
Burge's views can be found in a 1974 paper "Truth and Singular
Terms". In it he presents an axiom system and truth conditions for first
order logic with identity. It is designed to provide a negative solution.
While I share his motivation of arriving at an account of non-denoting
singular terms along the lines of a negative solution, there are a num-
ber of differences between his approach and the one taken in this pa-
per. To begin with there is the Aristotelian orientation that I am argu-
ing for. Another difference consists in his having an axiom system and
my use of trees (which can readily be restated in the form of a system
of natural deduction). His truth conditions differ from my Mates in-

spired treatment. A major difference of a somewhat more philosophi-
cal sort is that I am following an Aristotelian tradition that does not
connect existentials and the so-called existential quantifier but does
connect existential import with something like the copula. Burge
seems to adhere to the Frege-Russell-Quine view that "existence is
what existential quantification expresses". I have also replaced (Vx) (x
= x) (which is Burge's axiom 3) with (Vx)«3y) x = y ~ x = x). As ax-
iom 7 he asserts (Vx)(3y) (x = y) which he takes to be valid. I remind
the reader that (on the Frege-Russell-Quine reading of '(3x)') this is
the formal statement of 'Everything exists' and Burge endorses it. On
my Aristotelian oriented account this sentence is a contingent false-
hood and as such neither an axiom, a theorem nor a valid sentence.
(2) Quality/Property Instantiation and not Quantity/Quantification
Determines Existential Import
On both an Aristotelian and a Fregean view, singular sentences and
quantifications based on them require existing objects in order for
them to be true. However, the Fregean ties existence to quantification,
even quantification that is not based on property instantiation (such as
when the quantification is based on a negation of a singular sentence -
a denial of property instantiation). This difference can be clarified by
making explicit an ambiguity in the notions of 'instance' and 'instanti-
ate'.9 In one sense, 'instance' and 'instantiate' apply to instantiating a
property. So the truth that Socrates is human involves the fact that the
property of being human is instantiated in Socrates, in other words,
Socrates is an instance of the property of being human. In this sense of
instantiation, in the typical case an instance is an individual object and
instantiation is not much different than having a property. This is the
'having a property' sense of 'instance'. It is this sense of instantiation
that explicates existence and that is common to the Fregean and Aris-
totelian views.
A denial of existence is connected with a denial of instantiation in
this first sense. So true external negations of the form -,Fa, such as
'-,Socrates is a reindeer', '-,Socrates is a unicorn', record the fact that
the properties of being a reindeer and of being a unicorn are not in-
stantiated in Socrates. A true denial of existence as in '-,Unicorns ex-
ist' records the fact that the property of being a unicorn is not instanti-
ated in anything. The property of being Vulcan is like the property of
being a unicorn in not being instantiated, and '-,vulcan exists' is like
'-.,Unicorns exist' in being true.
A major difference between Frege and the older tradition concerns
a second use of 'instance'. A second sense of 'instance' and 'instanti-
ate' occurs in saying that the sentence/proposition stating that
Socrates is in space
is an instance of the sentence/proposition stating that
Everything is in space.
We speak of universal "instantiation." Similarly, the sentence or
proposition stating that
Socrates is human
is said to be an instance that is the basis for so called "existentially"
generalizing to the sentence or proposition stating that
Something is human.
In this second sense a sentence, a proposition, or some other truth ve-
hicle is an instance and not an individual object as in the first "property
instantiation" sense. This second sense is the "generalization to or
from an instance" notion of an instance.
The sentence
-.,Vulcan exists
records a fact concerning the absence of property instantiation (in-
stance in sense one) and this sentence, the denial, can serve as the ba-
sis, an instance in the second sense, for generalizing to the generaliza-
tion
Something does not exist.
When we generalize from the denial of property instantiation (sense
one), to the generalization that something is not a unicorn, we are not
committed to the existence of anything. From not instantiating (sense
one) and so not existing, nothing follows about what exists; e.g., from
being a unicorn is not instantiated, nothing follows about what exists
(what is instantiated), even though 'Something is not a unicorn' does
follow. Since the second - the generalization sense of an instance -
allows for generalizations on not having a property (not instantiating),
the generalizing (quantity/quantifier) sense of instantiation does not
have the force of existence. The "generalization to or from an in-
stance" sense of 'instance' is in a way broader than the having a prop-
erty sense of 'instance'. Generalizations represent non-existence as
well as existence.
On the Fregean view, existence is tied to both senses of instantia-
tion. I regard this as a mistake. On the older view, negations of singu-
lar sentences do not require existents in order to be true. Neither such
negations nor any generalizations (or truth functional constructions)
derived from them require existents. So quantification, so called
"existential" quantification, has existential import only when it is
grounded in singular/atomic sentences - genuine property instantia-
tion. Quantification does not have existential force when it is grounded
in negations of singular sentences (a property not being instantiated).
The same view is found in traditions that make serious use of the
copula such as in Kant and in Lesniewski. Kant thinks of existence in
terms of the copula. It is the copula tie between singular noun/term and
the traditional predicate noun/term that is linked with existential im-
port and not the quantifier 'some'. "Logically, it [being] is merely the
copula of a judgment." (Kant, Critique of Pure Reason, Smith (ed.)
[1953] p.504) This copula account comes to the same as the property
instantiation one, if we are allowed to put the copula account of the
truth of singular judgements along the following lines. A singular
judgement is true when the individual the singular term/concept refers
to instantiateslhas the property that is represented by the predi-
cate/noun. Kant uses language employing notions of positing. A sin-
gular judgement is true when the object the subject concept posits is
posited by the predicate concept. When no object is posited the judge-
ment is false (the negative solution), (cf. Orenstein [1978], Ch. 4).
The above points are encapsulated in the maxim of traditional
logic that quality, not quantity, determines existential import (see
Thompson). Instantiating a property (existence) is recorded in atomic
sentences. Such sentences are affirmative, and the property being in-
stantiated is recorded in the predicate being applied (syntactically) to
the singular term. The sentence is true when there is an individual the
singular term represents and it has/instantiates that property. The sen-
tence is false and its negation true, when the individual does not in-
stantiate the property. This can be when there is an individual and it
does not have the property, or when there is no such individual (e.g.
Vulcan). Non-existence, as failure of a property to be instantiated, is a
sufficient condition for the falsity of an atomic sentence and the truth
of its negation. Quality determines existential import, in the sense that
any sentence whose truth follows from the negation of an atomic sen-
tence inherits that lack of existential import. Only sentences whose
truth requires the truth of an atomic sentence (property instantiation)
have existential import.
(3) Non-existents are not self-identical
But what of the purported theorem/logical truth:

(\fx)(3y) (x = y)
which clashes with the contingent truth that Vulcan does not exist:
-,(3x) (x = Vulcan).
It is said to be a theorem because it follows from the total reflexivity of

identity
(\fx) (x = x),
which is taken to be an axiom of the logic of identity. When this pur-

ported axiom is examined in the light of the negative solution for
atomic sentences with empty names, it loses its plausibility. If 'Vulcan
= Vulcan' is false, then '(\fx) (x = x)' is false as well. While it seems
undeniable that every existing thing is self identical, (\fx)(3y) (yexists
~ x = x) is true, the same cannot and need not be said for everything
is self identical, (\fx) (x =x).
In the history of philosophy, we find various ways of thinking
about the negative solution. There is the dictum: Nihili nullae pro-
prietates sunt (Mates [1971], p.345). (Nothing [what does not exist]
has no properties.) This serves as an ontological underpinning for the

position that atomic/singular sentences with empty subject terms are
not true, and so are false. In Kant we find a slightly more semantic ver-
sion of it. Non entis nulla sunt predicata (all that is asserted of the [
non-existent] object, whether affirmatively or negatively, is erroneous)
(Kant, Critique of Pure Reason, Smith (ed.) [1953], p.627). From the
ontological version, it follows that non-existents such as Vulcan do not
have any properties - not even that of being self-identical. With se-
mantic ascent the semantical version dictates that no predicate truly
applies to Vulcan, even the predicate '= Vulcan'.
We abandon total reflexivity and in its place adopt the partial re-
flexivity of identity, i.e., ('v'x)«3y) y = x ~ x = x). There is no longer
any clash with true denials of singular existentials and there are no
problematic modal claims such as that J.C. exists necessarily. Revis-
ing total reflexivity provides a more conservative solution than that of
the narrow free logician. It does not modify/mutilate the underlying
predicate logic that precedes identity theory. Whatever was blocked by
the modified rules is blocked here as well. By questioning total re-
flexivity, we go to the heart of the problem.
Denying total reflexivity has quite a history. The Lesniewskian
tradition (Lejewski, Waragai) rejects it. We find Kant telling us that
"- in an identical proposition - if its [the subject term's purported
referent] existence is rejected, we reject the thing with all its predi-
cates; and no question of contradiction can then arise" (Critique of
Pure Reason, Smith (ed.) [1929], p.502). As I read this material from
Kant, it seems clear to me that when he rejects an identity claim be-
cause of non-existence, he is applying the negative solution as embod-
ied in his remark Non entis nulla sunt predicata (all that is asserted of
the [non-existent] object, whether affirmatively or negatively, is erro-
neous). I would change the translation, substituting 'predicated' for
'asserted', to read: all that is predicated of the [non-existent] object,
whether affirmatively of negatively is erroneous (Ibid., p.627). When
Kant in the passage on identity speaks of rejecting "the thing with all
its predicates," he is applying the Non entis dictum.
Even earlier Ockham, Buridan, and others "in effect" rejected the
total reflexivity of identity. They did this when considering and re-
jecting Boethius's claim "no predication is more true than that in
Alex Orenstein, Existence, identity, and an Aristotelian Tradition 145
which the same thing is predicated of itself'. I say "in effect" because
the issues surrounding total reflexivity were discussed under the topic
of self-predication. For the Terminists this was the crucial test case for
existential import. They dealt with it in connection with the sophism
A chimera is a chimera.
Their discussion assumes that generalizations are explained in terms of
a doctrine of descent to the singulars. Universal and particular gener-
alizations are true or false when conjunctive and disjunctive expan-
sions of suitable singular sentences are true. So
All/Some men are mammals
requires the truth of
This man is this mammal and/or that man is that mammal and/or
etc.
A singular sentence figuring as a conjunct/disjunct underlying the
sophism would look like
This chimera is this chimera.
Such a singular sentence is false when the subject term is vacuous.
Anachronistically speaking Gumping from Ockham to identity in
predicate logic) such false singular sentences are counterexamples to
the total reflexivity of identity.
The views put forward in section 4 can be summarized for first order
predicate logic with identity:
1. Allow atomic/singular sentences (propositions, or your favorite
truth vehicle) of the form 'Fa' where the 'a' position can be
taken by empty names. Allow for truth functional compositions
and generalizations.
2. Inspired by the correspondence conception of truth such atomic
sentences are true provided the subject denotes one of the ob-
jects the predicate applies to. When the subject term is empty
the sentence is not true. Assuming bivalence, it is false. Its ne-
gation is true.
3. Assume the usual truth conditions for connectives and a variant

of Mates' truth conditions for generalizations (see the appendix
and Orenstein [2000a])
a. A corollary of 1-3 is that (\Ix) (x = x) is false when a
sentence of the form 'a = a' is false.
b. Adopt the partial reflexivity of identity.
1-3 deal with the logical side of things. The philosophical side per-
taining to the notion of existence has a positive and a negative compo-
nent. They caution us that quality and not quantity determines existen-
tial import.
4. Think of existence in terms of property instantiation, i.e., what
is required for the truth of an atomic sentence. Atomic sen-
tences are the most basic form of an affirmative sentence. They
are used to define distinctions of quality (an affirma-
tive/negative distinction). Only sentences requiring the truth of
atomic sentences, i.e., property instantiation, have existential
import.
5. Don't interpret/read '(3x)' as 'There exists'. Existence is not
what so called "existential" quantification expresses.
Appendix: Semantics - Truth Conditions
The truth condition for basic canonical instances such as 'Fa', in non-
vacuous cases is similar to the ones found in most logic texts for non-
vacuous atomic sentences. 'Fa' will be true when there is an a and it is
an F (a is a member of the set 'F' has as its semantic value). While dif-
ferent accounts can be given of vacuous singular sentences, in this pa-
per I adopt the negative solution and treat them as false. Like Terminist
logicians and Lesniewskians, 'Fa' is false when either 'a' is vacuous
or there is an object that is not part of the semantic value of 'F'. Such
singular sentences are false and their negations true when they contain
a vacuous term. I assume the usual conditions for negation, conjunc-
tion and the other connectives.
The truth conditions for generalizations are an adaptation of Ben-
son Mates method of beta-variants. Mates provided an account of un-
restricted quantifiers of first order logic which relies on taking a given
interpretation of a first order fonnula and systematically reinterpreting

it (such reinterpretations are called "beta-variants") so that a given sin-
gular tenn is assigned different objects in different interpretations
(beta-variants). A universal generalization is true when it is true for an
instance on a given interpretation and remains true for every reinter-
pretation (every new assignment to that singular tenn). '(\Ix) -x-' is
true iff an instance of it, e.g., '-a--' is true under every interpretation.
The singular tenn, individual constant, 'a' is assigned different objects.
Tom Baldwin pointed out that the substituend, the constant, func-
tions like (is analogous to) the demonstratives 'this' and 'that' of Eng-
lish in being used to refer to different objects. It is as if the instance
'This is an F' turns out true for any object you use the same singular
expression 'This' to refer to. Note that it is the semantic (in the sense
of truth conditions and the theory of reference) and not the pragmatic
i.e., contextual - indexical character, of demonstratives that is being
appealed to, 'This' and 'That' can be assigned different individuals
and nonetheless be of the category of singular tenns. Baldwin has in
mind the aspect of demonstratives that allows the same expression to
be assigned different referents and not the contextual and user relative
notion of someone actually using a demonstrative while identifying
and pointing at some object that will serve as its referent. Note further
that: (1) Baldwin's point is construed as a helpful analogy and not as a
substitute for the full truth condition. And, to repeat: (2) It is the se-
mantic (in the sense of truth conditions and the theory of reference)
and not the pragmatic or contextual-indexical character of demonstra-
tives that is being appealed to, i.e., that 'this', the same expression, can
be assigned different individuals and nonetheless is of the category of a
singular tenn.
A basic canonical sentence Fa is true iff a exists and it is a mem-
ber of the val F.
[As an aid to understanding this condition, consider the functional
equivalence of Fa with either (3x) (x = a /\ Fx) or (a = a /\ Fa)]
The usual conditions for negation, conjunction and the other con-
nectives.
val (\Ix) - x - is true iff val -a-- is true under every beta-
variant.
val (3x) -:X'- is true iff val-a-- is true under at least one beta-
variant.
Consider the following examples of how these truth conditions are ap-
plied. A sentence such as
Something exists, i.e., (3x)(3y) (x = y)
is true, since at least one instance is true on at least one beta-variant,
e.g., (3y) (a = y) on at least one interpretation 'a' is assigned an object
e.g. Kant.
Something does not exist, i.e., (3x)---,(3y) (x = y)
is true. Consider the case where the substitution instance for 'x' is
'Vulcan'. '---,(3y) (Vulcan = y)' is true given the falsity of the sentence
it negates, i.e. '(3y) (Vulcan = y)'. There is no atomic sentence 'Vulcan
= a' which is true; no matter what object is assigned to 'a'. Since there
is no beta-variant on which 'Vulcan = a' is true, i.e., its negation, i.e.,
'---,(3y) (Vulcan = y)' is true. As true it serves as the interpretation that
accounts for the truth of '(3x)---,(3y) (x = y)'. Recall that '(3x) -:X'-' is
true if there is an instance of it that is true on a given interpretation.
Notes
I There are several ways of dealing with these problems. In Orenstein [2000] I
examine the problem without appealing to identity. In Orenstein [1990] I take up
Quine'S statement of the problem in Mathematical Logic (see Quine [1940]).
2 Cf. Wiggins [1994]. There are many varieties of the slogan that exists is not a
real predicate. It is a mistake to cite Kant as in the Frege-Russell-Quine tradition of
construing existence in terms of quantification. Kant in his general/formal logic is in
the Aristotelian Terminist-Lesniewskian tradition of construing existentials in terms
ofthe copula. See the last section of this paper and Orenstein [1978].
3 Wiggins on p.94 cites two strings which are translations into English of Ger-
man sentences judged meaningless by Frege 'There is Julius Caesar' and 'Julius Cae-
sar exists'. I believe most who disagree with Frege would say that the second makes
sense and is meaningful, but the first does not appear to be well formed and to con-
stitute a sentence by itself. The first appears more well formed when we insert an
article, i.e., 'There is a Julius Caesar'.
4 Wiggins suggestion differs from Quine's in that for Quine '= a' is a fused
predicate and has no logical structure so that one cannot existentially generalize with
respect to 'a' (Quine [1960], pp.176-81). Wiggins allows such generalizations.

5 Not to be confused with views such as van Fraassen's supervalutations and
many-valued approaches where there is a truth vehicle but it may not have a truth
value or may have more values than truth and falsity.
6 I think it fails to do so for him because he believes that it does not meet certain
conditions for identifying the referent of a name (see Sainsbury [1990] and Wiggins
[1994]).
7 Numerous questions remain about the 'really' operator:
(a) Wiggins quite rightly extends Evans thoughts on 'really' from story telling to
hypothesizing. While this is all to the good, what it indicates is the ubiquity of
'really' claims. We can add 'really' to just about any claim. Instead of saying the
battIe of Hastings took place on such and such a date we can say that it really took
place on that date. As for fiction and elsewhere, we are not limited in using 'really'
in connection with exists. Instead of saying that Sherlock Holmes did not live on
Baker street we can say that Holmes did not really live on Baker street. See Wiggins
own example of actors - it is not an existential.
(b) 'Really' brings background assumptions such as beliefs or imaginings into
the foreground, Within the background assumptions we are calling attention to their
status as background assumptions.
(c) The Spice Girls iteration is derived from a common form of speech: What do
you want? What do you really want? What do you really really want? 'Really' has a
force not mentioned by Evans and Wiggins: that of emphasis of some sort or of
forcing one to look at the reasons for making such a claim.
(d) In their deployment of the 'really' operator in connection with existentials,
Wiggins and Evans are being ad hoc. They are arbitrarily singling out one sort of
case where 'really' may be used without saying how it has a special use here. They
are leaving out the use of 'really' to express surprise: I really never thought he would
- , to cast doubt: Did you really think -?, to simply reiteratc, as used for contrasting
in general not just for fiction/pretense or hypothesizing e.g., to ask for speaker's justi-
fication in making the non-really assertion or by the speaker of elliptically offering or
claiming assurance as to the justification.
(e) Besides, if 'really' has scope, then, given compositionality, we need an ex-
planation of the sub-sentence, i.e., 'Vulcan does not exist'.
8 It has also been used for issues concerning the empty domain though this as-
pect will not concern us here. (see Orenstein [1995]).

9 This ambiguity is a case in point of one of the distinctions Kevin Mulligan
made in his paper at the Bergamo meetings.

ORENSTEIN ON EXISTENCE AND
IDENTITY
Mauro Mariani
1. A COUNTER-INTUITNE ASPECT OF ORENSTEIN'S

PROPOSAL
The problem Orenstein is faced with can be set out in a form which has
a venerable past in the history oflogic - that of an inconsistent triad. In
other words, among the following three statements
(I) The language of first-order logic contains empty individual
terms.
(II) Negative sentences whose subject is an empty individual term
are true, while the affirmative ones are false.
(III) The quantifiers are interpreted referentially (using Tarski's
truth conditions).
at least one must be discarded. Otherwise, the logical law
(V'X) A(x) ~ A(t)
where t is an empty name, would be invalid!: in the Tarskian interpre-

tation of quantifiers, in fact, in every model that assigns the entire do-
main to the predicative letter A, the antecedent is true, but on the basis
of (II) the consequent is false. The assumption that the domain is not
empty - it is worth noticing - is irrelevant, since in this case the ante-
cedent is vacuously true.
151
As often happens with inconsistent triads, the possible solutions

are very many, each raising its own difficulties, since (I)-(III) appear at
first sight to be all intuitively plausible assumptions. Frege, for in-
stance, finds himself forced to deny (II): if the denotation of a complex
expression is defined compositionally, and if the relation name/bearer
is the paradigm of denoting, apophantic sentences in which empty
names occur tum out to lack a truth value. For Frege, however, the
presence of empty names is a defect in natural languages, to be elimi-
nated in the scientific ones: hence for the latter (I) does not hold - and
(II) becomes irrelevant, while for the former it is (II) that is false.
In these comments I do not intend to discuss the virtues or defects
of the various solutions that have been put forward. Orenstein's criti-
cisms, in any case, seem to me to be on the whole acceptable: as I have
said already, each proposed solution presents counter-intuitive features
or technical complications (especially in the case of the varieties of
free logic), that are not difficult to pick up. The point is that Oren-
stein's proposed solution also presents at least one feature that is
strongly counter-intuitive. Let us see what this is.
To put it as synthetically as possible, Orenstein proposes to discard
(III) and to replace the Tarskian truth conditions for quantified sen-
tences with Mates', modified by allowing for empty names. In short (to
avoid pointlessly repetition of what has already been clearly indicated
in Orenstein's text) ('I7'x) A(x) is true under the interpretation I when
A(f3lx) is true under every f3-variant of I. So, if t is an empty name and
therefore A(t) is false on the basis of (II), then (\Ix) A(x) will be false
(even if the entire domain is - in Tarskian style - assigned to A), and
therefore (\Ix) A(x) ~ A(t) will be true. In this way one obtains the ad-
vantages of not arbitrarily regimenting the language (eliminating the
empty names), of avoiding gaps in the truth values, and finally of not
modifying the standard first-order logic (as in the case of the varieties
of free logic).
In contrast to all these virtues, however, as I have already said,
there is in my opinion one strongly counter-intuitive feature. Let us as-
sume in fact that a strange gas has turned all objects existing at the pre-
sent time red: in this case each one of us would be willing to claim
'Everything is red'. Let us suppose for the moment that someone
wishes to protest against this statement by arguing 'Something is not
Mauro Mariani, Orenstein on Existence and Identity 153
red' is true, on the basis of the fact that 'The white lady' (a character
in a Walter Scott novel) is obviously not red. I believe that the standard
reaction would be that this person is joking, or else that her use of
quantifiers deviates from its normal use in English. In other words the
difficulty lies in derivation of a sentence, that we expect will tell us
something about the real world, simply from the fact that a certain
name, let us suppose 't', is not denoting (when 't' is without denota-
tion the only condition of truth of 't is not B' is indeed the lack of de-
notation of 't'). From Frege's perspective this truth condition would
not be totally uninformative, because it would mean that to a given
sense no denotation corresponds (for example that 'the most divergent
series' is a non-denoting name is a truth of the analysis); but from a
Mill-Kripke perspective, in which the proper nouns do not have sense,
that a proper noun does not denote is a purely linguistic fact.
I am not of course arguing here that this interpretation of quantifi-
ers is inconsistent, or that it is not possible to construct first-order logic
in this way. I am simply saying that this interpretation does not con-
stitute an adequate formalization of the standard use of quantifiers in
natural languages. What Orenstein seems to suggest is that in natural
language quantifiers do not have existential import in themselves, and
that they receive this only from a faulty interpretation of the first-order
logic: we could go further and say that according to Orenstein, it would
be perfectly natural for us to say that there are things that do not exist,
if it were not that a Tarskian interpretation had been given to quantifi-
ers.
Orenstein rightly claims that his interpretation of quantifiers is not
substitutional, and he levels against such an interpretation the criticism
of not having enough names. This is by no means the only problem it
runs into: for example, as Dunn and Belnap have pointed ouf, the
strong completeness theorem would be no longer valid. If we employ,
however, the Henkin "trick" of adding to our language a denumerable
set of new names, we are able to meet the above objections. So, to my
mind, the substitutional interpretation of quantifiers, if emended along
these lines, is equivalent to Orenstein's (of course if we stipulate, un-
like R. Barcan Marcus 3, that A(t) is always false when A is a predica-
tive letter and t is an empty name).
In general I am rather skeptical about the idea that substitutional

quantification may, in itself, be a solution to puzzles that are unsolv-
able otherwise. In some cases, however, it may be useful, to eliminate
irrelevant ontological problems. For example in Protothetic, a referen-
tial interpretation of propositional quantifiers leads to thorny questions
(for example, as to the denotation of apophantic sentences and as to the
ontological status of propositions) that the substitutional interpretation
allows us to avoid. However, as far as the truth value (and probably
also the conditions of truth) of quantified sentences themselves is con-
cerned, the two interpretations do not differ. In other cases the solution
obtained through the substitutional quantification is more apparent
than real. For example D. Kaplan's "Quantifying in"4 only apparently
solves the well-known puzzles arising from quantifying into opaque
contexts. In the last analysis, the problem becomes one of specifying
what it is necessary to understand by 'standard name' or 'vivid name',
and I am unable to escape from the impression that it may be really dif-
ficult to arrive at an understanding of what a standard name is without
falling back, more or less covertly, much into a deprecated essential-
ism (adopting which, as Quine himself admitted, to give a sense to
modal contexts is no longer a problem).
2. PRECEDENTS FOR ORENSTEIN'S PROPOSAL
2.1. The Aristotelian tradition
Orenstein observes that his way of understanding negative sentences

and quantifiers is rooted in an ancient Aristotelian tradition, that found
its most complete expression in terministic logic of the late Middle
Ages. Orenstein is undoubtedly right in saying that in the Aristotelian
tradition, and in Aristotle himself, truth value gaps do not exist, and
that negative singular sentences with non-denoting subject are, on the
whole, true, whereas affirmative ones are false (cf. the locus classicus
Cat. 10, 13 b27-35). It is nevertheless necessary to make some distinc-
tions, which will tum out to be important also in relation to Oren-
stein's observations on Kant. In De Int. 11, in fact, Aristotle seems to
deny that from 'Homer is a poet' should follow 'Homer is', on the ba-
sis of the fact that in the first of these sentences 'to be' is predicate by
accident. According to some modem interpreters the reason for this is
that in 'Homer is a poet' the predicate expresses an accidental feature
of the subject, whereas 'to be' for Homer means 'to be a man' (or
something of that kind), in other words to enjoy a certain essential
property: thus from the simple fact that Homer is a poet 'Homer is'
does not follow, not because this latter sentence is false, but because
from the predication of an accident the predication of an essence does
not follow. However, medieval tradition is nearly unanimous in giving
another, in my opinion more correct, interpretation to this passage. In
brief, in the context of De Int. 11 predication by accident means predi-
cation depending on another, and the fallacy discussed by Aristotle is
analogous to that of deriving 'Socrates is good' from 'Socrates is a
good cobbler': just as Socrates is good only insofar as he is a cobbler,
but not in an absolute sense, so Homer 'is' only insofar as something is
predicated of him, but not in an absolute sense. That this is the correct
interpretation is confirmed by the fact that immediately after this ex-
ample, Aristotle observes that in an analogous way, from 'the not be-
ing is thinkable' it does not follow that 'the not being is', where it is
not a question of predicating essences or accidents. If this is correct,
then in De Int. 11 Aristotle is questioning the fact that from the truth of
an affirmative sentence the existence of what is denoted by the subject
term (in brief, of its subject) always follows: i.e. there can be true af-
firmative sentences with a non-existent subject, and therefore, since
the Principle of Excluded Middle is not called into question in this
case, false negative sentences with a non-existent subject. In addition,
in De Int. 10, 19b 14-19 Aristotle introduces, among the primary forms
of affirmation and negation, "strange" sentences of the kind 'Every
man is' and 'Every man is not'. In my opinion the best way to make
sense of these sentences is to suppose that a universal can be predi-
cated truly also of subjects that do not at present exist. If also 'Homer
is a man' is true on the same grounds as 'Homer is a poet' was true,
while 'Homer is' is false, then 'Every man is' is a meaningful false
sentence. I shall return to this point in a short while.
In one medieval tradition there were at least two solutions (in
practice equivalent) to this type of exception to the rule that from 'A is
B' it follows that 'A is'. The first was based on the theory of the am-
plialio and the restrictio of the term subject, or on the fact that the
suppositio personalis5 of the term A could be amplified until it in-
cluded both the denotation of A in the past and (perhaps) in the future 6 ,
and the objects of thought, whereas 'is' (in its use de secundo adi-
acenle) continued to co-signifY the time and therefore to refer only to
time in the present. The second was based on the distinction between
internal predication and external predication. The sentence 'Homer is a
poet' constitutes an example of internal predication, because being a
poet connotes Homer (in Frege's terms it is part of the sense that al-
most everyone connects to the proper name 'Homer'), and therefore it
is true even though 'Homer' supponit for something that is not present;
whereas 'Homer is sick' represents an example of external predication
(the 'normal' predication), and can be true only if 'Homer' supponil
for something that is present. In other words the rule that from 'A is B'
follows 'A is' admits of exceptions only if the premise 'A is B' consti-
tutes a case of internal predication.
Now if we apply the theory of ampliatio also to 'Aristotelian'
quantifiers we can suppose that they range or do not range over the
amplified supposilio of the term 'man'. In the second case 'Every man
is' is logically true (at least if we admit that men exist), while in the
first it can be, or rather certainly is, false. If we translate this distinc-
tion in terms of "Fregean" quantifiers, and read 'U(x)' as un instance
of internal predication, a sentence of the type '(::Ix) (U(x) 1\ -'E!(x))' -
where 'E!' corresponds to 'is' de secundo adiacente7 - can be true.
This apparently contradicts what I previously said about the meaning
of quantifiers in natural language, but only apparently. The domain of
quantifiers can include or not include the amplified suppositiones of
the terms, but the interpretation of the quantifiers remains in any case
referential. If the domain does not include them we have the normal
use of the quantifiers, and if it includes them we have a special use that
remains anyway referential.
This Aristotelian tradition (represented among others by Peter of
Spain), notwithstanding the fact that it too admits the possibility of
quantifYing over the non-existent, is deeply different from the one
Orenstein is referring to. First of all the truth of a sentence with non-
denoting subject does not depend only on the quality of the sentence
itself, but also on the type of predication. If it is a case of internal

predication, also the affirmative sentences can be true. In addition the
ampliatio of the suppositio of a term to objects not at present existing,
or also to objects of thought, leads more in the direction of Meinong
(with his distinction between what exists and what subsists) than in the
direction of Russell's thinking, to which also Orenstein belongs de-
spite his different way of interpreting quantifiers. After all, what Oren-
stein is trying to do is combine the realism of Russell with the idea that
it is possible to formulate true sentences with a non-denoting subject.
There is however another Aristotelian tradition (if we wish to label
it, we may call it Ockhamist) in which, as a matter of fact, the sen-
tences with a non-denoting subject are true if and only if they are
negative, and this is the one which Orenstein is thinking of. In this tra-
dition quantification is explained in terms of the different types of sup-
positio personalis and descensus, but it seems this explanation cannot
be applied to sentences in which empty (i.e. without suppositio) uni-
versal terms occur. Let us consider the sentence' Some A is not E'. To
every universal term C is connected a sequence of individual terms q,
C2, •.• Cn, ... so that 'a is C' if and only if 'a is cn' for some Cn belonging
to the above-mentioned sequence. 8 In a negative particular, for exam-
ple in 'Some A is not B', A has suppositio determinata, whereas B has
suppositio confusa distributiva, i.e. (leaving apart futile technicalities)
'Some A is not B' is equivalent to or, according to other interpreta-
tions, implies
(i) (al is neither b l nor b2 nor... nor bn nor... ) or (a2 is neither b l
nor b2 nor...nor bn nor... ) or. .. or (an is neither b l nor b2
nor... nor bn nor ... ) or. ..
where ah a2, ... an, ... and b h b2,... bn,... are the sequences of terms con-
nected respectively to A and B.
This descensus, though it may be considered a distant ancestor of
the substitutional interpretation of quantifiers, presupposes that the
terms have suppositio. But in Ockhamist semantics, this is not true of
terms like 'unicorn', and therefore 'Some unicorn is not white' cannot
be interpreted on the same lines as (i). In effect Ockham says that a
negative particular is true when:
(ii) the subject term supponit for something for which the predi-
cate does not supponit (when i.e. (i) holds);
or
(iii) the subject term does not have suppositio. 9
Orenstein on the other hand considers only condition (i), which he lays
out, however, in a slighty different form:
(iv) (This A is neither b l nor b2 nOL .. nor bn nor... ) or (that A is

neither b l nor b2 nOL .. nor bn nor ... ) or...
The difference between (iv) and (i) is that, since in (iv) 'this A' and
similar expressions are proper names, it is a sufficient condition of (iv)
that all these names are empty, i.e. that A supponit pro nihilo; whereas,
if this latter circumstance occurs, (i) cannot even be formulated and it
is necessary to use the clause (iii). As a consequence the interpretation
of quantifiers in the Ockhamist tradition is on the whole nearer to that
of a referential type, even if the negative sentences with non-denoting
terms are considered true.
2.2. Kant
According to a consolidated interpretative tradition (not without its ex-
ceptions, for example Hintikka lO), Frege shares with Kant the thesis
that existence, properly understood, is a second order predicate. It is
therefore a bit surprising that Orenstein attributes to Kant the thesis
that so-called existential quantification actually has existential signifi-
cance only when it is based on an affirmative sentence, but not "when
it is grounded in negations of singular sentences".
In actual fact Orenstein justifies his interpretation through a certain
number of quotations, that, taken out of context, end up by meaning
the opposite of what Kant really wanted to say. For example
"Logically, being is merely the copula of a judgment" is understood as
if it meant that the existential significance of an affirmative sentence
lay, from a logical point of view, in the copula. On the contrary, in the
context from which this quotation was taken (Critique of Pure Reason,
Transcendental Dialectic, Book II, Chapter ill, Section IV) Kant op-
poses (a) the copulative use of being, in which 'is' is limited to putting
the predicate in relation with the subject, without (thereby) anything

being 'posited', to (b) the existential use (i.e. de secundo adiacente),
where the subject is posited with all the predicates that belong to it
analytically. In other words, for Kant the copulative use of 'is' has no
existential significance, at least when we have to do with analytical
judgments.
That the above paragraph accurately depicts the situation is con-
firmed by the fact that the supreme principle of all analytical judg-
ments is the principle of contradiction. To deny an analytical judgment
means indeed to affirm and deny the same thing of the same subject,
since the predicate is contained in the subject itself. As Kant writes:
If the judgment is analytic, whether negative or affirmative, its truth can
always be adequately known in accordance with the principle of contra-
diction. (Critique of Pure Reason, Analytic of Principles, Book II, Chap-
ter II, Section I)
The same point is stressed in this passage:

If, in an identical proposition, I reject the predicate while retaining the
subject, contradiction results; and I therefore say that the former belongs
necessarily to the latter. But if we reject subject and predicate alike, there
is no contradiction, for nothing is then left that can be contradicted. (Cri-
tique of Pure Reason, Transcendental Dialectic, Book II, Chapter III,
Section IV)
In other words, an analytical judgment (this is in fact the meaning of

'identical proposition' - cf. Kant, Logic, §37) whose subject is non de-
noting is true because its truth is, as we have seen, given by the simple
lack of contradiction. II
Rather ironically, Orenstein quotes this same passage as proof that
for Kant sentences of the form 'A = A' are false if A does not exist.
Apart from the fact that this means taking 'identical proposition' in a
much more limited sense than the Kantian one of analytical judgment,
Orenstein interprets 'if I suppress the subject together with the predi-
cate', as if it meant the rejection of the identical proposition when the
subject does not exist. But such an interpretation cannot be sustained
because, as I have said, the consequence of this suppression is the ab-
sence of contradiction, and therefore the truth of the analytical judg-

ment.
Orenstein attributes to Kant the unconditional acceptance of the
principle Non entis nulla sunt predicanda and stresses that the (sup-
posed) Kantian rejection of 'A = A' (if A does not exist) is grounded on
it. It is not clear what meaning Orenstein attributes to this principle,
which could mean either (a) that of not being nothing is predicated
with truth in an affinnative proposition, or else (b) that of not being
nothing is predicated with truth, neither affinnatively nor negatively,
Indeed the second interpretation of the principle leads to truth value
gaps (something that Orenstein in general does not accept); but it
seems to correspond better to the Kantian fonnulation "all that is as-
serted of the object, whether affinnatively or negatively, is erroneous,
and consequently we cannot arrive apagogically at knowledge of the
truth through refutation of the opposite." (Critique of Pure Reason,
Transcendental Doctrine of Method, Ch. I, Section N). However, at
least in this case, "what is asserted of the object negatively" corre-
sponds to infinite predicates (for example 'not-A'), and therefore also
in the Kantian fonnulation, the principle Non entis nulla sunt predi-
canda concerns sentences with an affinnative fonn. This turns out to
be evident from the following text, in which Kant argues that, if the
subject does not exist, the negative sentences are true, whereas the in-
finite ones are false (since their fonn is affirmative):
If therefore we say that the world is either infinite in extension or is not in-
finite (non est infinitus) and if the former proposition is false, its contra-
dictory opposite, that the world is not infinite, must be true. And I should
thus deny the existence of an infinite world, without affirming in its place
a finite world. But if we said that the world is either infinite or finite (non-
infinite), both statements might be false. (Critique of Pure Reason, Tran-
scendental Dialectic, Book II, Chapter II, Section VII)
All this seems in contradiction with what I said before, that affirmative
sentences with non-existent subject can be true, but the contradiction is
only apparent. Indeed all the Kantian applications of the principle Non
entis nulla sunt predicanda concern synthetic sentences 12, whereas for
the analytical ones this principle seems not to apply: so in Kant too we
find, at least implicitly, the distinction between internal and external
predication. 13 The former concerns analytical sentences and does not

have in itself existential import; the latter concerns synthetic sentences
(a priori or a posteriori), and always has, when the form is affirmative,
existential import.
To conclude, Orenstein is right when he affirms that for Kant
negative sentences are true when the subject does not exist, but is
wrong to argue that for Kant this holds only for negative sentences.
Affirmative sentences with non-existent subject can be true too, if they
are identical.
Notes
1 To avoid complications 1 assume that A is a predicative letter, to be able to
limit (II) to atomic sentences.
2 Cf. Dunn and Belnap [1968], pp.I77-85.
3 Cf. Marcus [1962], pp.252-59.
4 Cf. Kaplan [1969], pp.178-14.
S Actually there are many varieties of suppositio, but in this paper 1 am con-
cerned only with suppositio personalis: so by suppositio 1 will mean henceforth sup-
positio personalis.
6 1 say "perhaps" because in many cases, especially as far as individuals are con-
cerned, the future existence is uncertain. It is not by chance that medieval people of-
ten used as an example Antichrist, whose future existence was guaranteed by the
Book of Revelation.
7 In this case 'Every A is' would always be true, since its standard translation in
the logic of predicates would be (Vx) (U(x) ~ E!(x)).
8 For example if x is man then x is Peter or x is John, etc. (the 'is' of identity is
not distinct from 'is' of predication)
9 Cf. Matthews [1973], pp.13-24.
10 Cf. Hintikka [1986], pp.249-67.
11 The same thing goes for Leibniz of course, and it is probable that Kant him-
self depends on Leibniz.
12 In particular the cosmological antinomies. Kant takes great care to demon-
strate that this is not a dialectical game nor a questioning of the Principle of the Ex-
cluded Middle, and observes that the antinomies are made up of pairs of sentences of
the form 'A is B' and 'A is not B', whose joint falseness demonstrates only that A
does not exist (I do not intend here to go into the question of what Kant means by the
worlds not existing).
13 Of which 1 have spoken at some length concerning one ofthe Aristotelian tra-
ditions.
ABSTRACT OBJECTS: A CASE STUDY
Stephen Yablo
1. NECESSITY
Not a whole lot is essential to me: my identity, my kind, my origins,

consequences of these, and that is pretty much it. Of my intrinsic prop-
erties, it seems arguable that none are essential, or at least none spe-
cific enough to distinguish me from other humans. And, without get-
ting into the question of whether existence is a property, it is certainly
not part of my essence to exist.
I have by contrast huge numbers of accidental properties, both in-
trinsic and extrinsic.
So, if you are looking for an example of a thing whose "essence" is
dwarfed by its "accense", you could not do much better than me. I
should add, in fairness to the other candidates, that you could not eas-
ily do much worse than me, either. Accense dwarfs essence for just
about any old object you care to mention: table, mountain, horse, or
what have you.
Any old concrete object, I mean. Abstract objects, especially pure
abstracta like 11 and the empty set, are a different story. I do not know
what the intrinsic properties of the empty set are, but odds are that they
are mostly essential. Pure sets are not the kind of thing we expect to go
through intrinsic change between one world and another. Likewise in-
tegers, reals, functions on the reals, and so on. I
The pattern repeats itself when we turn to relational properties.
163
A. Bottani et at. (eds.), Individuals, Essence and Identity, 163-188.
My relations to other concrete objects are almost all accidental. But the
number II's relations to other abstract objects (especially other num-
bers) would seem to be essential.
The most striking differences, though, have to do with existence.
Concrete objects (with the possible exception of "the world" on one
construal of that phrase) are one and all contingent. But the null set and
the number 11 are thought to exist in every possible world. To add to
the mystery, one normally supposes that existence is inversely related
to essence: the bigger x's essence, the "harder" it is for x to exist, and
so the fewer worlds it inhabits. And yet here is a class of objects ex-
tremely well endowed in the essence department, and yet incapable of
not existing.
You would have to be in a coma not to wonder what is going on
here. Why is it that so much about abstract objects is essential to them?
What is it about numbers et al. that makes it so hard for them not to
exist? And should not objects that turn up under all possible conditions
have impoverished essences as a result?
It may be that I have overstated the phenomenon. Not everyone
agrees that numbers even exist, so it is certainly not agreed that they
exist necessarily. There would be more agreement if we changed the
hypothesis to: numbers exist necessarily provided they can exist, that
is, unless they are impossible2 . And still more if we made it: numbers
exist necessarily provided they do exist. But these are nuances and de-
tails. I think that it is fair to say that everyone, even those who opt in
the end for a different view, has trouble with the idea that 11 could go
missing.
So I ask: Why should a numberless world seem impossible (al-
lowing that the appearance may be only prima facie)? Why should it
seem impossible for numbers to have had different intrinsic properties,
or different relational properties vis a vis other abstract objects? Why
should numbers seem to be absolutely modally inflexible?
2. APRIORITY
A second prima facie difference between the concrete and abstract

realms is epistemological. Our knowledge of concreta is a posteriori.
But our knowledge of numbers at least has often been considered a
Stephen Yablo, Abstract Objects: a Case Study 165
priori. That 3+5 = 8 is a fact that could be known on the basis of expe-
rience - experience of counting, say, or of being told that 3+5 = 8. But
the same is true of most things we know a priori. It is enough for a
priority that experience does not have to figure in our justification.
And this seems true of many arithmetical claims. One can determine
that 3+5 = 8 just by thinking about the matter.
Like the felt necessity of arithmetic, its felt apriority is puzzling
and in need of explanation. It is a thesis of arithmetic after all that
there are these things called numbers. And it is hard to see how one
could be in a position to know a priori that things like that really ex-
isted.
It helps here to remember the two main existence-proofs philoso-
phers have attempted. The ontological argument tries to deduce God's
existence from God's definition, or the concept of God. The knock
against this has been the same ever since Kant; from the conditions a
thing would have to satisfy to be X, nothing existential follows unless
you have reason to think that the conditions are in fact satisfied. Then
there is Descartes's cogito. This could hardly be expected to give us
much guidance about how to argue a priori for numbers! In any case,
the argument does not seem to be a priori. You need to know that you
think, and that knowledge would seem to be based on your experience
of self.
I said that the ontological argument and the cogito were the two
best-known existence-proofs in philosophy. Running close behind is
Frege's attempted derivation of numbers themselves. If the Fregean
line is right, then numbers are guaranteed a priori by logic together
with definitions. Shouldn't that be enough to make it a priori that
numbers exist? Perhaps, if the logic involved were ontology-free. But
Frege's logic affirms the existence of all kinds of higher-type objects.
(Frege would not have wanted to call them objects because they are
not saturated; but there is little comfort in that.) The Fregean argument
cannot defeat doubts about a priori existence, because it presupposes
they have been defeated in presupposing the apriority of Fregean logic.
A different strategy for obtaining a priori knowledge of numbers
goes via the "consistency-truth principle": in mathematics, a consistent
theory is a true theory. If we can know a priori that theory T is consis-
tent, and that the consistency-truth principle holds, we have a priori

warrant for thinking T is true, its existential claims included.
There are a lot of things one could question in this strategy. How
do we gain knowledge of the consistency-truth principle? You may say
that it follows from the fact that consistent theories have (intended)
models, and that truth is judged relative to those models. But this ar-
gument assumes the truth of model theory. And a priori knowledge of
model theory does not seem easier to get than a priori knowledge of
arithmetic.
Even if we do somehow know the consistency-truth principle a
priori, a problem remains. Not all consistent theories are on a par. Pe-
ano Arithmetic, one feels, is true, and other theories of the numbers
(PA#) are true only to the extent that they agree with PA. It does not
help to say that PAis true of its portion of mathematical reality, while
PA# is true of its. That if anything only reinforces the problem, be-
cause it makes PA# just as true in its own way as PA. It begins to look
as though arithmetical truth can be a priori only if we downgrade the
kind of truth involved. A statement is not true/false absolutely but only
relative to a certain type of theory or model.
3. ABSOLUTENESS
I take it as a given that mathematical truth does not feel relative in this
way. It feels as though 3+5 is just plain 8. It feels as though the power
set of a set is just plain bigger than the set itself.
One could argue that the notion of truth at work here is still at
bottom a relativistic one: it is truth according to standard math, that is,
truth according to whatever theories are accepted in the mathematical
community.
But this is a far cry from what we want, and act like we have. For
now the question becomes, why is this math standard and not that? It
would be nice if the answer adverted at some point to the theory's be-
ing true or correct in a way that logically coherent alternatives are not
correct. To run the point the other way around: If all PA can claim over
the competition is greater utility or naturalness given our projects and
cognitive dispositions, then that gives us absolute truth in name only.

Why is it that there are infinitely many primes? Because we wound up
passing on the coherent alternative theory according to which the
number of primes is finite - and for reasons having nothing to do with
truth. Neither theory is truer than the other.
Another problem is sociological. 3 plus 5 was seen to be 8 long be-
fore anyone had formulated a theory of arithmetic. How many people
even today know that arithmetic is something that mathematicians
have a theory of? Saul and Gloria (my non-academic parents) are not
thinking that 3+5 = 8 is true-relative-to-the-standard-theory, because
they have no idea that such a theory exists, and because they would
most likely think that the theory was standard because it was true. Are
they just confused? If so, someone should pull the scales from their
eyes. Someone should make them realize that the truth about numbers
and sets is (like the truth about what is polite or stylish) relative to an
unacknowledged standard, a standard that is in relevant respects quite
arbitrary.
I would not want to attempt it, and not only because I do not like
my parents angry at me. If they would balk at the notion that there is
no more to be said for standard mathematics than for a successful code
of etiquette, I suspect they are probably right.
Admittedly, there are parts of mathematics, especially parts of set
theory, where a relative notion of truth seems not out of place. Perhaps
the most we can say about the continuum hypothesis is that in some
nice-looking models it is true, while in others it is false. I admit then
that the intuition of absolute truth may not extend to all cases. But
even in set theory it extends pretty far. A set theory denying, say, In-
finity, or Power Set, strikes us as wrong, even if we have yet to put our
finger on where the wrongness is coming from.
Could the explanation be as simple as this? If a model does not
satisfy Power Set, or Infinity, then we do not see it as modeling 'the
sets'. That Infinity holds in all models of 'the sets' is a trivial conse-
quence ofthat linguistic determination. It is not as if there is a shortage
of models which include only finite set-like objects. It is just that these
objects are at best the pseudo-sets, and that makes them irrelevant to
the correctness of Infinity taken as a description of the sets. Infinity is
'true' because models that threaten to falsify it are shown the door;
they are not part of the theory's intended subject matter.
Call this the debunking explanation of why it seems wrong to deny
the standard axioms. I do not say that the debunking explanation is out
of the question; I do not want to rule it out that ZF serves in effect as a
reference-fixer for 'set' But, that certainly is not how it/eels. If some-
one wants to argue that Infinity is wrong - that the hereditarily finite
sets are the only ones there are - our response is not "save your breath!
deny Infinity and you're changing the subject". Our response is: "that
sounds unlikely, but let's hear the argument". No doubt we will end up
thinking that the Infinity-denier is wrong. The point is that what he is
wrong about is the sets. It has to be, for if he is not talking about the
sets, then we are not really in disagreement.
Suppose though the debunkers are right that ZF is true because it
sets the standard for what counts as a set. This still does not quite ex-
plain our sense that ZF is correct. Why should we be so obsessed with
the sets as opposed to the pseudo-sets? To the extent that ZF and 'sets'
are a pair, curiosity about why ZF seems so right is a lot like curiosity
about why the sets seem so right. It does not matter how the questions
individuate, as long as they are both in order. And so far nothing has
been said to cast doubt on this. So I ask again, why do ZF and the sets
seem so right?
4. ABSTRACTNESS AND NECESSITY
Three puzzles, then: one about necessity, one about apriority, one
about absoluteness. Let us start with necessity; the other two puzzles
will be brought in shortly.
What accounts for 11 's tenacious grip on reality? One natural
thought is that there is something about abstractness that prevents a
thing from popping in and out of existence as we travel from world to
world. 3 Crispin Wright and Bob Hale observe that it is hard to think
what conditions favorable for the emergence of numbers would be, and
it is hard to think of conditions unfavorable for their emergence. It is
however easy to think of conditions favorable for the emergence of Mt.
McKinley. The reason, one imagines, is that numbers are abstract and
Mt. McKinley is not.
But, granted that numbers do not show up only when conditions
are right, how does that bear on their necessity? Explanations come to
an end somewhere, and when they are gone we are left with the brute
facts. Why shouldn't the existence/nonexistence of numbers be a brute
fact? Traditionally existence has been the paradigm of a phenomenon
not always admitting of further explanation. So, granted that numbers
aren't contingent on anything, one still wants to know why they should
not be contingent full stop.
A second possible explanation is that it is part of the concept of an
abstract object (a "pure" abstract object, anyway) to exist necessarily if
at all. An object that made an appearance in this world, but was miss-
ing from others, would by that alone not be abstract.
Suppose that's right; an otherwise qualified object that does not
persist into all worlds does not make the cut. Let us think about these
contingent would-be abstracta some more. What sort of object are we
talking about here? The obvious thought is that they are exactly like
real abstracta except in the matter of necessary existence. But the ob-
vious thought is strange, and so let us ask explicitly: Could there be
"shmabstract" objects that are just like their abstract cousins except in
failing to persist into every world?
Fiddling with an object's persistence conditions is generally con-
sidered harmless. If I want to introduce, or call attention to, a kind of
entity that is just like a person except in its persistence-conditions - it
is missing (e.g.) from worlds where the corresponding person was born
in Latvia - then there would seem to be nothing to stop me. If you can
have shmersons alongside persons, why not shmumbers along with
numbers?
You may think that there is a principled answer to this: a princi-
pled reason why abstracta cannot be "refined" so as to exist in some
worlds but not all. If so, though, then you hold the view that we started
with: there's something about abstractness that precludes contingency.
What is it? Earlier we looked at the idea that where abstracta are con-
cerned, one world is as good as another. But although it is true that a
contingently existing number has no apparent rationale for turning up
here rather than there, why should that bother us? Why should the
choice of worlds not be arbitrary? This is only one suggestion, of

course, but as far as I am aware, the route from abstractness to neces-
sity has never been convincingly sketched.
Back then to the view that abstract objects can be "refined". There
is nothing wrong with shmabstract objects, on this view, it is just that
they should not be confused with abstract objects.
Once again, though, this seems to deepen the mystery. Why do we
attach so much importance to a concept - abstractness - that rules out
contingent existence, as opposed to another - shmabstractness - that
differs from the first only in being open to contingent existence? Does
the salience of numbers as against shmumbers reflect no more than a
random preference for one concept over another? One would like to
think that more was involved.
5. CONSERVATIVENESS AND NECESSITY
So much for "straight" explanations, meaning explanations that accept

the phenomenon as genuine and try to say why it arises. Now for a
"subversive" explanation. Hartry Field does not think there are any
numbers. So he is certainly not going to try to validate our intuition of
necessary existence. He might however be able to explain the intuition
away, in other words reinterpret it as really being an intuition not of
necessity but something related. He does make a suggestion along
these lines. A theory is conservative if
it is consistent with every internally consistent theory that is 'purely about
the physical world'. (Field [1989], p.240)
Conservative theories are theories compatible with any story that might
be told about how things go physically, as long as that story is consis-
tent in itself. (I am going to skate lightly over the controversy over how
best to understand 'consistent' and 'compatible' here. The details are
not important for what follows.)
Now, one obvious way for a mathematical theory to be conserva-

tive is for it to be necessary. A theory that cannot help but be true is
automatically compatible with every internally consistent physical the-
ory.
But, although necessity guarantees conservativeness, you can have
conservativeness without it. A necessary theory demands nothing;
every world has what it takes to make the theory true. A conservative
theory makes no demands on the physical world; if the theory is false,
it is false not for physical reasons but because the world fails to com-
ply in some other way. To put this a little more explicitly, T is conser-
vative iff for each world in which T is false, there is another, physi-
cally just like the first, in which T is true. The theory is false then only
due to the absence of non-physical objects like numbers.
You might think of the foregoing as a kind of necessity. A conser-
vative theory T is "quasi-necessary" in the sense that necessarily, Tis
satisfiable in the obtaining physical circumstances. Here again is
Field:
mathematical realists have held that good mathematical theories are not
only true but necessarily true; and a clear part of the content of this (the
only clear part, I think) is that mathematics is conservative [ ... ]. Conser-
vativeness might loosely be thought of as 'necessary truth without the
truth'. [ ... ] I think that the only clear difference between a conservative
theory and a necessarily true one is that the conservative theory need not
be true. [ ... ] Perhaps many realists would be content to say that all they
meant when they called mathematical claims necessarily true was that they
were true and that the totality of them constituted a conservative theory.
(Field [1989], p.242)
From this it seems a small step to the suggestion that the only distinc-
tively modal intuition we have about mathematical objects is that the
theory of those objects is conservative. So construed, the intuition is
quite correct. And it is correct in a way that sits well with our feeling
that existence is never "automatic" - that nothing has such a strong
grip on reality that it is incapable of not showing up.
So: is our intuition of the necessity of '3+5 = 8' just a (confused)
intuition of quasi-necessity, that is, conservativeness?
I think this is very unlikely. Yes, every world has a physical dupli-
cate with numbers. But one could equally go in the opposite direction:
every world has a physical duplicate without them. If the permanent
possibility of adding the numbers in makes for an intuition of neces-
sity, then the permanent possibility of taking them out should make us
want to call numbers impossible. And the second intuition is largely
lacking. A premise that is symmetrical as regards mathematical exis-
tence cannot explain why numbers seem necessary as opposed to im-
possible.
A second reason why necessity is not well-modeled by conserva-
tiveness is this. Arithmetical statements strike us as individually neces-
sary. We say, "this has got to be true", not "this considered in the con-
text of such and such a larger theory has got to be true". But the latter
is what we should say if our intuition is really of conservativeness. For
conservativeness is a property of particular statements only seen as ex-
emplars of a surrounding theory. A statement that is conservative in
the context of one theory might fail to be conservative in the context of
another. (It might be inconsistent with the other.) Nothing like that
happens with necessity.
A third problem grows out of the discussion above of consistency
as sufficient for truth. Suppose that two theories contradict each other.
Then intuitively, they cannot both be necessary; indeed if one is neces-
sary then the other is impossible. But theories that contradict each
other can both be conservative.
Someone might reply that if contradictory means syntactically
contradictory, then contradictory theories can so be necessary. All we
have to do is think of them as describing different domains, e.g., dif-
ferent subsets of the universe of sets.
That is true in a technical sense. But the phenomenon to be ex-
plained - our intuition of necessity - occurs in a context where contra-
dictory theories are (the technical point notwithstanding) experienced
as incompatible. If I affirm Infinity and you deny it, we take ourselves
to be disagreeing. But both of us are saying something conservative
over physics.
When two statements contradict each other, they cannot both be
necessarily true. Or rather, they cannot unless the truth is relativized
somehow: to the background theory, a certain type of model, a certain
portion of mathematical reality. Once we relativize, though, standard

mathematics ceases to be right (full stop). And a lot of it feels right
(full stop). Once again, then, our problems about apriority and neces-
sity are pushing us toward a no less problematic relativism.
6. FIGURALISM
The conservativeness gambit has many virtues, not the least of them
being its short way with abstract ontology. At the same time there are
grounds for complaint. One would have liked a theory that made
arithmetic "necessary" without making it in a correlative sense
"impossible". And one would have liked a theory less friendly to rela-
tivism. The best thing of course would be if we could hold onto the
advantages of the Field proposal without giving up on "real" necessity,
and without giving up on the intuition of absolute truth or correctness.
Is this possible? I think it just may be - or at least I have an idea to
this effect that is not immediately obviously hopeless. I can indicate
the intended direction by hazarding (what may strike you as) some ex-
tremely weird analogies:
(A) '7 is less than 11'

'the frying pan is not as hot as the fire'
'a molehill is smaller than a mountain'
'pinpricks of conscience register less than pangs of conscience'
(B) '7 is prime'

'the back burner is where things are left to simmer'
'the average star has a rational number of planets'
'the real estate bug doesn't sting, it bites'
(C) 'primes over two are not even but odd'

'butterflies in the stomach do not sit quietly but flutter about'
'pounds of flesh are not given but taken'
'the chips on people's shoulders never migrate to the knee'
(D) 'the number ofFs is large iff there are many Fs'
'your marital status changes iff you get married or ... '
'your identity is secret iff no one knows who you are'

'your prospects improve iff it becomes likelier that you'll succeed'
(E) 'the Fs outnumber the Gs iff# {xlFx} > # {xIGx}'

'you are more resolute ... iff you have greater resolve'
'these are more available ... iff their market penetration is greater'
'he is more audacious ... iffhe has more gall'
(F) 'the # ofFs = the # ofGs iff there are as many Fs as Gs'
'your whereabouts = our whereabouts iff you are where we are'
'our greatest regret = yours iffwe most regret that. .. and so do
you'
'our level of material well-being = yours iffwe are equally well
off
Now let me list some of the ways in which these statements are analo-
gous. (I'll focus for the time being on necessity).
All of the statements seem (I hope) true. But their truth does not
depend on what may be going on in the realm of concrete objects and
their contingent properties and relations. There is no way, we feel, that
7 could fail to be less than 11. Someone who disagrees is not under-
standing the sentence as we do. There is no way that molehills could
fail to be smaller than mountains, even if we discover a race of mutant
giant moles. Someone who thinks molehills could be bigger is con-
fused about how these expressions work.
Second, all of the statements employ a distinctive vocabulary -
'number', 'butterflies', '{xIFx}', 'market penetration' - a vocabulary
that can also be used to talk about concrete objects and their contingent
properties. One says that 'the number of English Kings is large', 'her
marital status is constantly changing', and so on.
Third, its suitability for making contingent claims about concrete
reality is the vocabulary's reason for being. Our interest in stomach-
butterflies does not stem from curiosity about the aerodynamics of
fluttering. All that matters to us is whether people have butterflies in
the stomach on particular occasions. Our interest in 11 has less to do
with its relations to 7 than with whether, say, the eggs in a carton have
11 as their number, and what that means about the carton's relation to
other cartons whose eggs have a different number.
Fourth, the vocabulary's utility for this purpose does not depend
on conceiving of its referential-looking elements as genuinely standing
for anything. It does not depend on conceiving its referential-looking
elements any other way, either. Those if any who take stomach-
butterflies, party affiliations, and numbers dead seriously derive the
exact same expressive benefit from them as those who think the first
group insane. And both groups derive the exact same expressive bene-
fit as the silent majority who have never given the matter the slightest
thought.
7. NECESSITY AS BACK-PROPAGATED
I said that all of the statements strike us as necessary, but I did not of-
fer an explanation of why. When it comes to the non-mathematical
statements, an explanation is quickly forthcoming.
Stomach-butterflies and the rest are representational aids. They
are "things" that we advert to not (not at first, anyway) out of any in-
terest in what they are like in themselves, but because of the help they
give us in describing other things. Their importance lies in the way
they boost the language's expressive power.
By making as if to assert that I have butterflies in my stomach, I
really assert something about how I feel - something that it is difficult
or inconvenient or perhaps just boring to put literally. The real content
of my utterance is the real-world condition that makes it sayable that S.
The real content of my utterance is that reality has feature BLAH: the
feature by which it fulfills its part of the S bargain.
The reason it seems contingent that Pat's marital status has
changed is that, at the level of real content, it is contingent: she could
have called the whole thing off. The reason it seems necessary that
Pat's prospects have improved iff it has become likelier that she will
be successful is that, at the level of real content, it is necessary, for the
two sides say the very same thing.
How does the world have to be to hold up its end of the 'the num-
ber of apostles is even' bargain? How does the world have to be to
make it sayable that the number of apostles is even, if we are going to
say there are numbers? There have to be evenly many apostles. So, the
real content of 'the number of apostles is even' is that there are evenly
many apostles.
Now, that there are evenly many apostles is something that need
not have been the case, and that it takes experience to confirm. At the
level of real content, then, 'the number of apostles is even' is epistemi-
cally and metaphysically contingent. But there might be other number-
involving sentences whose real contents are necessary. To the extent
that it is the real content of these sentences that we hear them as ex-
pressing, it would be natural for us to put those sentences on the list of
necessary truths.
This is why (some) number-involving sentences, e.g., 'the number
of Fs = the number of Gs iff the Fs and Gs are equinumerous' feel
necessary at the same time as we have trouble seeing how they could
be necessary. Our two reactions are to two different contents. The
sentence feels necessary because at the level of real content it is a tau-
tology: the Fs and Gs are equinumerous iff they are equinumerous.
And tautologies really are necessary.
The reason we have trouble crediting our first response is that the
sentence's literal content - that there is this object, a number, that be-
haves like so - is to the effect that something exists. And it is baffling
how anything could cling to existence so tightly as to be incapable of
not turning up.
Why do the two contents get mooshed together in this way? One
reason is this. A sentence's conventional content - what it is generally
understood to say - can be hard to tell apart from its literal content. It
takes work to remember that the literal meaning of 'he is not the
brightest guy in town' leaves it open that he is the second brightest. It
takes work to remember that (literally) pouring your heart out to your
beloved would involve considerable mess and a lengthy hospital stay,
not to mention the effect on your beloved. Since there is no reason for
us to do this work, it is not generally realized what the literal content in
fact is.
Consider now '7<11 '. To most (!) people it means that seven
things are fewer than eleven things. But the literal content is quite dif-
ferent. The literal content makes play with entities 7 and 11 that meas-
ure pluralities size-wise, and encode by their internal relations facts
about their outnumbering relations. Of course, these plurality-measures
7 and 11 are no more on the speaker's mind than blood is on the mind
of someone pouring their heart out to their beloved. '7<11' is rarely
used to describe numbers as such, and so one forgets that the literal
content is about nothing else.
Someone might say that the literal contents of pure-mathematical
statements are quickly recovered, once we set our minds to it. It is the
real contents that need further explanation. I do not actually think that
the real contents are always the same, so there is a considerable
amount of exaggeration in what follows. But that having been said, the
claim will be that arithmetic is, at the level of real content, a body of
logical truths - specifically, logical truths about cardinality, while set
theory is, at the level of real content, a body of logical truths of a com-
binatorial nature.
8. ARITHMETIC
Numbers enable us to make claims which have as their real contents

things we really believe, and would otherwise have trouble putting into
words.
One can imagine introducing number-talk for this purpose in vari-
ous ways, but the simplest is probably this. Imagine that we start out
speaking a first-order language with variables ranging over concreta.
Numerical quantifiers '3 v x Fx' are defined in the usual recursive way.4
Now we adopt the following rule (*S* means that it is to be assumed
or imagined that S):
(N) if 3 n x Fx, then *there is a thing n = the number of Fs*

Since (N)'s antecedent states the real-world condition under which we
are to make as if the Fs have a number, F should be a predicate of con-
crete objects. But the reasons for assigning numbers to concrete plu-
ralities apply just as much to pluralities of numbers (and pluralities of
both together). So (N) is strengthened to
(N) if *3 n x Fx* then *there is a thing n = the number of Fs*

Now F is a predicate of concreta and/or numbers. Because the rule
works recursively in the manner of Frege, we get "all" the numbers
even if there are only finitely many concreta. 0 is the number of non-
self-identical things, and k+ 1 is the number the numbers <k.
Making as if there are numbers is a bit of a chore; why bother?
Numbers are there to expedite cardinality-talk. Saying '#Fs = 5' in-
stead of '3sx Fx' puts the numeral in a quantifiable position. And we
know the expressive advantages that quantification brings. Suppose
you want to get it across to your neighbor that there are more sheep in
the field than cows. Pre-eN) this takes (or would take) an infinite dis-
junction: there are no cows and one sheep or there are no cows and two
sheep or there is one cow and there are two sheep, and etc. post-eN) we
can say simply that the number of sheep, whatever it may be, exceeds
the number of cows. The real content of '#sheep > #cows' is the infi-
nite disjunction, expressed now in finite compass. s
But although this may clarify what the real contents of applied ar-
ithmetical statements are, statements of pure arithmetic are another
matter.
Take first quantifierless addition statements. What does the con-
crete world have to be like for it to be the case that, assuming numbers,
3+5 = 8? Assuming numbers is assuming that there is a number k
numbering the Fs if there are k Fs. But that is not all. One assumes too
that if no Fs are Gs, then the number of Fs and the number of Gs have
a sum = the number of things that are either F or G. All of that behind
us, the real-world condition that makes it OK to suppose that 3+5 = 8
is that
33X Fx /\ 3 sy Gy /\ Vx -'(Fx /\ Gx) ~ 3gZ (Fz v Gz)
This is a logical truth. Consider next quantifierless multiplication

statements. What does the concrete world have to be like for it to be
the case that, assuming numbers, 3x5 = 15? Well, it is part of the
number story that if n = the number of Fls = the number of F2s =
.... the number of Fms, and there is no overlap between the Fjs, then m
and n have a product mxn = the number of things that are FI or F2 or
.. . .Fm But then the real-world condition that makes it OK (on the as-
sumption of numbers) to suppose that 3x5 = 15 is that
(3)X FIX /\ ... /\ 3 3x Fsx /\ -'3x (FIX /\ F2X) /\ ... )
~ 3lsX(Flx v ... V Fsx)
Once again, this is a logical truth. Negated addition and multiplication

statements are treated similarly; the real content of 3+5 :f::- 9, for exam-
ple, is that
=hx Fx /\ 3 sy Gy /\ "ix-' (Fx /\ Gx) ~ -,3 9z (Fz v Gz)
Of course most arithmetical statements, and all of the "interesting"

ones, have quantifiers. Can logically true real contents be found for
them?
They can, if we make a few assumptions. First, the real content of
a universal (existential) generalization over numbers is given by the
countable conjunction (disjunction) of the real contents of its in-
stances. Second, conjunctions all of whose conjuncts are logically true
are logically true. Third, disjunctions any of whose disjuncts are logi-
cally true are logically true. Given these we can show that
The real content of any arithmetical truth is a logical truth.
This has already been seen for atomic and negated-atomic truths 6 •
These give us all the arithmetical truths (up to logical equivalence)
when closed under four operations: (1) conjunctions of truths are true;
(2) disjunctions with truths are true; (3) universal generalizations with
only true instances are true; (4) existential generalizations with any
true instances are true. It is easy to check that these operations preserve
the property of being logically true at the level of real content. We can
illustrate with case (4). Suppose that 3x cf>(x) has a true instance cf>(n).
By hypothesis of induction, cf>(n) is logically true at the level of real
content. But the real content of 3x cf>(x) is a disjunction with cf>(n) as a
disjunct. So the real content of3x cfJ(x) is logically true as well.
9. SET THEORY
Sets are nice for the same reason as numbers. They make possible
sentences whose real contents we really believe, but would otherwise
have trouble putting into words. One can imagine introducing set-talk
for this purpose in various ways, but the simplest is probably this. "In
the beginning" we speak a first-order language with quantifiers ranging
over concreta. These quantifiers can be singular or plural; one can say
'there is a rock such that it ... ' and also 'there are some rocks such
that they ... '. Now we adopt the following rule:
(S) if there are some things a, b, c ... ,
then *there is a set {a, b, c, ... }*.
Since the antecedent here states the real-world condition under which
we are to make as if a, b, c, ... form a set, a, b, c, ... are limited to con-
crete objects. But the reasons for collecting concreta into sets apply
just as much to the abstract objects introduced via (S). So (S) is
strengthened to
(8) if *there are some things a, b, c ... ,*
then *there is a set {a, b, c, ... }*.
This rule, like (N) in the last section, works recursively. On the first
go-round we get sets of concreta. On the second go-round we get sets
containing concreta and/or sets of concreta. On the third we get sets
containing concreta, sets of them, and sets of them. And so on through
all the finite ranks. Assuming that there are only finitely many con-
creta, our output so far is the hereditarily finite sets: the sets that in
addition to being themselves finite have finite sets as their members,
and so on until we reach the concrete objects that started us off.
What now? If we think of (8) as being applied at regular intervals,
say once a minute, then it will take all of eternity to obtain the heredi-
tarily finite sets. No time will be left to obtain anything else, for exam-
ple, the first infinite number 0).
The answer to this is that we are not supposed to think of (8) as
applied at regular intervals; we are not supposed to think of it as ap-
plied at all. (8) does not say that when we establish the pretense-
worthiness of 'there are these things', it becomes pretense-worthy that
'they form a set'. (8) says that if as a matter of fact (established or not)
*there are these things*, then *there is the set of them*. If in fact
*there are the hereditarily finite sets*, then *there are the von Neu-
mann integers (0 = 4>, n+l = {0,1, ... n} )*. And now (8) tells us that
*there is the set {O, 1,2,3, ... }*, in other words, *there is 0)*. I cannot
argue this here, but similar reasoning shows that we get all sets of rank
<X for each ordinal <x. (S) yields in other words the full tower of sets:
the full cumulative hierarchy.
Now, to say that (S) yields the full cumulative hierarchy might
seem to suggest that (S) yields a certain fIXed bunch of sets, viz. all of
them. That is not the intention. There would be trouble if it were the
intention, for (S) leaves no room for a totality of all sets. To see why,
suppose for contradiction that *a, b, c, ... are all the sets*. (S) now tells
us that *all the sets form a set V*. This set V must for familiar reasons
be different from a, b, c, ... So the proposed totality is not all-
encompassing. (I will continue to say that (S) yields the full cumulative
hierarchy, on the understanding that the hierarchy is not a fixed bunch
of sets, since any fixed bunch you might mention leaves something
out. This does not prevent a truth-definition, and it does not prevent us
from holding that some sentences are true of the hierarchy and the rest
false?).
Conjuring up all these sets is a chore; why bother? The reason for
bothering with numbers had to do with cardinality-type logical truths.
Some of these truths are infinitely complicated, but with numbers you
can sum them up in a single finite sentence. Something like that is the
rationale for sets as well. The difference is that sets help us to deal
with combinatorial logical truths - truths about what you get when you
combine objects in various ways.
An example will give the flavor. It is a theorem of set theory that if
x = y, then {x, u} = {y, v} iff u = v. What combinatorial fact if any does
this theorem encode? Start with '{x, u} = {y, v}'. The real content of
that is that theyxu are themyv - or, to dispense with the plurals, that (x =
y v x = v) /\ (u = Y v u = v) /\ (y = X V Y = v) /\ (v = X V V = u). So what
our theorem is really saying is that
Ifx = y, then
([(x = y v x = v) /\ (u = Y v u = v) /\ (y = X V Y = v) /\
(v = X V V = u)] iff u = v).
This is pretty simple as logical truths go. Even so it is not really com-
prehensible; I at least would have trouble explaining to someone else
what it says. If truths as simple as this induce combinatorial boggle-
ment, it should not be surprising that the set-theoretic formulations are
found useful and eventually indispensable.
A second example is Cantor's Theorem. What is the logical truth

here? One can express parts of it using the plural quantifier 3X ('There
are some things such that ... ') Numerical plural quantifiers are defined
using the standard recursive trick:
30X t/J(X) iff VX -,t/J(X)
3 n+IX t/J(X) iff3Y(t/J(Y) A 3nX (t/J(X) A -,.¥= Y)
Consider now 3¢Vy (Xy ~ Fy). I cannot give this a very natural
paraphrase, because English does not quantify over pluralities of plu-
ralities. But roughly the claim is that there are four ways of making a
selection from the Fs.s This lets us express part of what Cantor's Theo-
rem is "really saying", viz. that ifthere are n Fs, then there are 2n ways
of selecting just some of the Fs, as follows:
This is a second-order logical truth, albeit a different such truth for

each value of n. But we are still a long way from capturing the Theo-
rem's real content, because it applies to infinite pluralities as well.
There is (as far as I know) no way with these resources to handle the
infinite case. 9 It all becomes rather easy, though, if we are allowed to
encode the content with sets. Then the claim is that every set, finite or
infinite, has more subsets than it has members. (IP(x) I = 21X1 >JX1).
Now let me try to give a general recipe for finding real contents. It
will be simplest if we limit ourselves to talk of hereditarily finite sets;
the procedure I think generalizes but that remains to be checked. Take
first atomic sentences, that is, sentences of the form x = y and x E Z. A
reduction function r is defined:
(AI) r(x E z) is
(a) 3y «VUEZY = u) A X = y) if z has members

(b) 3y (y 7= Y A X = y) if z is the empty set
(c) X E Z if z is not a set. to
Note that the first line simplifies to v YEZ x = y; that is in practice what I
will take the translation to be. (The reason for the quantified version is
that it extends better to the case where z is the empty set.) The third
line marks the one place where E is not eliminated. If z is not a set,
then it is (literally) false to say that x belongs to it, which is the result
we want. The rule for identity-statements is
(A2) rex = y) is
(a) Vu (u EX B u E y) if x andy are sets

(b) x = y if either is not a set
In the "usual" case, x and y have members, and Vu (u EX B U E y)

simplifies to (I\UEx VVEY U = v) 1\ (I\VEY V UEX V = u). If x has members and
y is the null set, it simplifies to VU(VZEX U = z B U *- u). If both x andy
are the null set, we get Vu(u *- u B U *- u). Otherwise r leaves un-
touched x = y. Non-atomic statements reduce to truth-functional com-
binations of atomic ones by the following rules:
(R 1) r(-'cf» is -'r( cf»

(R2) r(l\; cf>;) is I\i r( c/>i)
(R3) rev; cf>;) is Vi r(c/>i)
(~) r(Vx cf>(x» is I\z=z r( cf>(z».
(Rs) r(3x cf>(x» is Vz=z r( cf>(z».
The real content of cf> is found by repeatedly applying r until you reach
a fixed point, that is, a statement cf>* such that r( cf>*) = cf>*. This fixed
point is a truth-functional combination of "ordinary" statements true or
false for concrete (non-mathematical) reasons. These ordinary state-
ments are to the effect that x = y, where x and yare concrete, or x = y,
where one is concrete and the other is not, or x E z, where z is con-
crete. 1I
How do we know that a fixed point will be reached? If cf> is a gen-
eralization, the (R;)s tum it into a truth-functional combination of at-
oms 'If. If 'If is an atom talking about sets, then the (A;)s tum it into a
generalization about sets of a lower rank, and/or non-sets. Now we ap-
ply the (R;)s again. Given that cf> contains only finitely many quantifi-
ers, and all the sets are of finite rank, the process must eventually bot-
tom out. 12 The question is how it bottoms out, that is, the character of
the sentence cf>* that gives cf>'s real content.
I claim that if cf> is a set-theoretic truth, then cf>* is, not quite a logi-
cal truth, but a logical consequence of basic facts about concreta:
identity and distinctness-facts, and facts to the effect that concreta

have no members. To have a word for these logical consequences, let
us call them logically true over concrete combinatorics, or for short
logically true cc . Three assumptions will be needed, analogous to the
ones made above for arithmetic. First, the real content of a universal
(existential) generalization is given by the countable conjunction (dis-
junction) of its instances. Second, conjunctions all of whose conjuncts
are logically truecc are themselves logically true cc . Third, disjunctions
any of whose disjuncts are logically true cc are logically truecc •
Every set-theoretic truth has a logically true cc real content.
The set-theoretic truths (recall that we are limiting ourselves to

hereditarily finite sets) are the closure of the atomic and negated-
atomic truths under four rules: (1) conjunctions of truths are true; (2)
disjunctions with truths are true; (3) universal generalizations with
only true instances are true; (4) existential generalizations with any
true instances are true. The hard part is to show that atomic and ne-
gated-atomic truths are logically true cc at the level of real content. The
proof is by induction on the ranks of x and y.
Basis Step
(a) If x and yare concrete, then the real content of x = y is that

x = y. This is logically true cc if true, because it is a conse-
quence of itself. Its negation is logically truecc if true for the
same reason.
(b) If x is concrete and y is a set, then x -:f:. y is true. Its real
content x -:f:. y is logically truecc , because a consequence of
the fact that x -:f:. y.
(c) If x and yare the null set, then x = y is true. Its real content
Vu (u -:f:. u B U -:f:. u) is a logical truth, hence logically truecc .
(d) If x is a non-empty set and y is the null set, then X-:f:.y is true.
Its real content -'Vu (VZEX u = z B U -:f:. u) is logically true cc ,
hence logically true cc .
(e) If y is a non-set then x~y is true. Its real content x~y is
logically truecc because a consequence of itself.
(f) Ify is the null set then x~y is true. Its real content -'3z (z -:f:. z
1\ X = z) is logically true cc because logically true.
Recursion Step
(a) Ifx andy are nonempty sets, then r(x:= y) is (I\UEX VVEY u:=
v) 1\ (I\VEY V UEXV:= u). (al) If it is true that x:= y, then r(x:=
y) is a conjunction of disjunctions, each of which has a true
disjunct u := v. By hypothesis of induction, these true dis-
juncts have logically truecc real contents. So rex := y) has a
logically truecc real content. And the real content of rex = y)
is also that of x := y. (a2) If it is true that x :t:. y, then rex :t:. y)
is a disjunction of conjunctions, each of which is built out
of true conjuncts. By hypothesis of induction, these true
conjuncts are logically truecc at the level of real content. So
rex :t:. y) has a logically true cc real content. And the real
content of rex :t:. y)is also that of x :t:. y.
(b) If z is a nonempty set, then rex E z) is vYEZ X = Y (bl) If it is
true that x E z, this has a true disjunct x = y. By hypothesis
of induction, x = y has a logically truecc real content. But
then rex E z) is logically true cc at the level of real content,
whence so is x E z. (b2) If the truth is rather that x (i!: z, then
rex (i!: z) is a conjunction of true conjuncts. By hypothesis of
induction, these conjuncts are logically truecc at the level of
real content. So rex (i!: y) has a logically truecc real content,
whence so also does x (i!: y.
10. SUMMING UP
The view that is emerging takes something from Frege and something
from Kant. One might call it "Kantian logicism". The view is Kantian
because it sees mathematics as arising out of our representations.
Numbers and sets are "there" because they are inscribed on the specta-
cles through which we see other things. It is logicist because the facts
that we see through our numerical spectacles are facts of first-order
logic.
And yet the view is in another way the opposite of Kantian. Kant
thinks necessity is imposed by our representations, and I am saying that
necessity is imposed on our representations by the logical truths they
encode. Another possible name then is "anti-Kantian logicism"! I will

stick with the original name. (I take comfort in the fact that "Kantian"
contains "anti"; we can think of the "anti" as inert or ert, as the situa-
tion demands.)
Back now to our three questions. Why does mathematics seem
(metaphysically) necessary, and a priori, and absolute? The first and
second of these we know how to answer. It seems necessary because
the real contents of mathematical statements are logical truths. And
logical truths really are necessary. It seems a priori because the real
contents of mathematical statements are logical truths. And logical
truths really are a priori.
That leaves absoluteness. It might seem that we can once again
cite the absoluteness of logical truth; real contents are not logically true
relative to this system or that, they are logically true period.
But there is an aspect of the absoluteness question that this fails to
address. It does perhaps tell us why arithmetic and set theory seem in a
non-relative sense correct. It does not tell us why Peano Arithmetic
strike us as superior to arithmetical theories that contradict it. It does
not tell us why the Zermelo-Fraenkel theory of sets strikes us as supe-
rior to set theories that contradict it. It could be, for all we have said,
that PAis not the only arithmetical theory - ZF is not the only set the-
ory - with the property that its real content is logically true. PA# and
ZF# could be (at the level of real content) just as logically true as PA
andZF.
If ZF# has a logically true real content, this is not the content in-
duced by the game sketched above: the game based on principle (S).
(ZF# has as a theorem some sentence A such that it is a theorem of ZF
that -.A. Unless something has gone very wrong, A and -.A will not
come out assertible in the same game). ZF# can be "correct" only if
real contents are judged relative to a different principle than
if it is to be imagined there are some things x, y, z, ... , then it's to
be imagined that there's a set of those things.
This gives us a way out of our difficulties. I said early on that you can-
not accuse someone of changing the subject just because they deny
some principle of ZF. But principle (S) is a great deal more basic than
anything found in ZF. If someone has trouble with the idea behind (S)
- the idea that when you have got a detenninate bunch of things, you
are entitled to the set of those things - then that person arguably does
not mean the same thing by 'set' as we do. l3 '
Suppose we call a theory 'ZF-like' if it represents the sets as

fonning a cumulative hierarchy. Then here is an argument that only
ZF-like theories get the sets right. If ZF# is not ZF-like, then by defini-
tion it does not represent sets as fonning a cumulative hierarchy. But
the cumulative hierarchy comes straight out of (S), the rule that says
that if you have got the objects, you have got the set of them as well.
So, whatever it is that ZF# describes, it is not a system of entities
emerging (S)-style out of their members. Emerging (S)-style out of
your members is definitive, though, of the sets as we understand them.
ZF# may well get something right, but that something is not the sets. 14
Notes
I Although on the Frege-Russell definition of number, there is, arguably, intrin-
sic change. The empty set can change too, if as Lewis suggests it is definable as the
sum of all concreta. But I am talking about what we intuitively expect, and no one
would call these definitions intuitive.
2 Hale and Wright suggest in "Nominalism and the Contingency of Abstract
Objects" that Field might not accept even that much. Field does say that numbers are
conceptually contingent. But it would be hard to pin a metaphysical contingency the-
sis on him, for two reasons. (1) He is on record as having not much use for the notion
of metaphysical necessity. (2) To the extent that he tolerates it, he understands it as
conceptual entailment by contextually salient metaphysical truths. If salient truths
include the fact that everything is concrete, then (assuming they are not concrete)
numbers will come out metaphysically impossible
3 Impure abstracta like singleton-Socrates are not thought to be necessarily ex-
istent. So really I should be talking about pure-abstractness. I will stick to "abstract"
and leave the qualification to be understood. (Thanks here to Marian David.)
4 30X Fx =df 'ix (Fx ~ x"* x) and 3n+1X Fx =df 3y (Fy /\ 3"x (Fx /\ x"* y))
5 There is an analogy here with Hartry Field's views on "the reason" for having
a truth-predicate, in the absence of any corresponding property.
6 Mario Gomez-Torrente pointed out that some atomic truths have not been pro-
vided out with real contents, a fortiori not with logically true real contents. An exam-
ple is (3+2)+1 = 6. This had me worried, until he pointed that these overlooked
atomic truths were logically equivalent to non-atomic truths that had not been over-
looked. For instance, (3+2)+ 1 = 6 is equivalent 3y «3+2 = y) /\ (y+ 1 = 6)). A quick
and dirty fix is to think of overlooked sentences as inheriting real content from their
not overlooked logical equivalents. A cleaner fIx would be desirable, but Mario has
not provided one yet.
7 See the last few pages of Putnam [1967] and "Putnam Sematics" in Hellman
[1989].
8 Alternatively, there are some things all of which are Fs, and some things not
the same as the fIrst things all of which are Fs, and etc. (Say there are two Fs. You
can pick both of them, either taken alone, or neither of them. Note that 'all the Fs'
and 'none of them' are treated here as limiting cases of 'some of the Fs').
9 You could do it with plural quantifIcation over ordered pairs
10 The idea is that XEZ describes x as one of the things satisfying the condition of
membership in z. The condition for membership {a, b, c, ... } is x = a v x = b v x = c

*
v ... The condition for membership in the null set is x X.
11 Statements of the fIrst type are necessarily true or necessarily false, depending
on whether x is indeed identical to y. Statements of the second and third types are
necessarily false, since concreta cannot be sets or have members.
12 The same argument would seem to work with sets of infInite rank; there are
no infInite descending chains starting from infInite ordinals either.

13 This might sound funny, given the widespread view that there are some things
(the sets) that are too many to form a set. This widespread view is at odds with (S)
only if it is supposed that there is some defInite bunch of things including all and only
the sets. If the sets are a defInite bunch of things, it is very hard to understand what
could be wrong with gathering them together into a further set. I agree with Putnam
when he says that "no concrete model [of Zermelo set theory] could be maximal -
nor any non-concrete model either, as far as that goes. Even God could not make a
model for Zermelo set theory that it would be mathematically impossible to extend,
and no matter what 'stuff He might use. [ ... ] it is not necessary to think of sets as
one system of objects [... ] in order to follow assertions about all sets" (Putnam
[1967], p.21).
14 I am grateful to a number of people for criticism and advice; thanks in par-
ticular to Gideon Rosen, Kit Fine, Gil Harman, Mario Gomez-Torrente, Marian
David, Ted Sider, Paul Horwich, and Stephen Schiffer.
KINDS, ESSENCE, AND NATURAL
NECESSITY
E. Jonathan Lowe
It is commonly assumed that there is a kind of necessity, natural or

physical necessity, that is in some sense "weaker" than logical neces-
sity and is somehow related to natural laws, or, more specifically, to
causal laws - whence this supposed kind of necessity is often also
called 'nomic' or 'causal' necessity. Of course, many empiricist phi-
losophers have followed David Hume in being sceptical either about
the existence or, less radically, at least about our epistemic access to
any such species of necessity.l But in recent years the foregoing con-
cept of natural necessity has come under attack from another quarter,
namely, from certain philosophers - some of whom describe them-
selves as 'scientific essentialists' - who hold that natural laws are in
fact necessary in the strongest possible sense: that is, who hold that the
necessity of such laws is no weaker than, and just as "absolute" as, the
necessity of logical truths. 2
Apparently, it is not possible to object to this essentialist doctrine
simply on the grounds that natural laws are only discoverable empiri-
cally, because Saul Kripke has taught us that there not only may be but
certainly are a posteriori truths that are necessary in the strongest pos-
sible sense, that is, in the sense of being true in every logically possible
world. 3 The paradigm examples of such truths are truths of identity,
189
A. Bottoni et al. (eds.), Individuals, Essence and Identity, 189-206.
such as the proposition that Hesperus is Phosphorus. According to

Kripke, we can establish a priori the truth of the conditional statement
'If Hesperus is identical with Phosphorus, then it is necessary that
Hesperus is identical with Phosphorus' - where the necessity in ques-
tion is of the strongest possible kind - even though we can only estab-
lish empirically that Hesperus is identical with Phosphorus. Given,
however, that we do know, on empirical grounds, that Hesperus is
identical with Phosphorus, we can then use the a priori truth just men-
tioned as the conditional premise in a modus ponens inference to the
conclusion that, indeed, it is necessary that Hesperus is identical with
Phosphorus. This, then, provides us with a model for the possible ac-
quisition of a posteriori knowledge of truths which are necessary in
the strongest possible sense. And, it is suggested, this model is appli-
cable to our knowledge of natural laws, where these are construed as
being necessary in the strongest possible sense. Let us, following the
now customary usage, call this strongest possible kind of necessity
metaphysical or broadly logical necessity.4 Then, it seems, there can be
no particular difficulty in maintaining that natural laws are at once
metaphysically necessary and knowable a posteriori.
One seeming advantage in holding this view of laws is that it pro-
vides us with a perfectly straightforward account of the nature of natu-
ral necessity, for according to this view there is really only one kind of
necessity - metaphysical or broadly logical necessity - which may be
explicated, in the language of possible worlds, in terms of truth in all
possible worlds. We need not, then, try to find any place for natural
necessity as a kind of necessity which is somehow weaker than logical
necessity but which, nonetheless, does not collapse into mere contin-
gency. On the other hand, many philosophers have strong intuitions
that natural laws are not necessary in the strongest possible sense - that
a natural law which obtains in this, the actual world, need not obtain in
every possible world. So we would need to be given good reasons for
thinking that these intuitions are mistaken, as well as some explanation
for our possession of those intuitions despite their being mistaken
ones.
But why should we want to embrace any notion of natural neces-
sity at all, whether or not this is construed as being weaker than logical
necessity? One reason commonly given is that we need to embrace this
E. Jonathan Lowe, Kinds, Essence, and Natural Necessity 191
notion in order to distinguish between genuine laws of nature and mere

accidental regularities. (Incidentally, I should explain that here and in
what follows, I distinguish between laws and statements of law, or
lawlike statements - a law being the kind of state of affairs which
makes a statement of law true, if indeed it is true. S) To use a well-
known example, it is a merely accidental regularity that every lump of
gold is less than a mile in diameter (assuming that this is indeed the
case), whereas it is no merely accidental regularity that every lump of
uranium is less than a mile in diameter. We want to say that no lump
of uranium could be a mile in diameter, of natural or physical neces-
sity, because the critical mass of uranium sufficient to produce a nu-
clear explosion occupies a spherical volume much less than one mile
in diameter. By contrast, nothing in nature prevents us from amassing
enough gold to make a lump of it one mile in diameter, even ifno such
lump ever has been or ever will be formed in the entire history of the
universe.
At this point, we need to consider what logical form a natural law
should be construed as having. (Strictly, it is statements of natural law
that have "logical form", but in an extended sense we can speak of the
lawlike state of affairs which makes a statement of law true as having a
"logical form".) According to the "Humean" or "regularity" account of
laws, a law is simply a universal generalisation which quantifies over
particulars - in the simplest case, something of the form 'For all x, if
Fx, then Gx'. Against this proposal, then, we find the objection raised
that it fails to distinguish between lawlike and accidental generalisa-
tions, according both the same logical form. As part of this objection,
it is observed that laws support counterfactual conditionals, whereas
mere universal generalisations of the sort just described do not. Letting
'F' be 'is a lump of gold/uranium' and 'G' be 'is less than one mile in
diameter', we find that 'For all x, if Fx, then Gx' does not support the
counterfactual conditional 'If the universe had contained another lump
of gold/uranium, it would have been less than one mile in diameter' -
and yet we want to say that, in the case of uranium, the counterfactual
conditional is true and should be seen to follow from the law. One
proposed solution, then, is to say that laws are not mere universal gen-
eralisations but are, rather, necessitations of these: that is, that a law, in
the simplest case, has the logical form 'Necessarily, for all x, if Fx,
then Gx', where the kind of necessity in question is "natural" necessity,

whatever that is. 6 As I shall explain later, I do not consider that this
sort of proposal really provides a genuine solution to the problem in
hand.
However, another sort of proposal, advanced by David Armstrong
and others, is that laws do not concern particulars at all, but, rather,
universals. According to this view, the logical form of a law, in the
simplest case, is 'N(F,G)', or 'F-ness necessitates G-ness'.7 Such a law
is then supposed to entail the corresponding universal generalisation
concerning particulars, namely, 'For all x, if Fx, then Gx', although the
latter does not, of course, entail the former. One supposed advantage of
this account is that it offers an explanation for the non-accidental
regularity obtaining amongst particulars in terms of a special "second-
order" relation of necessitation amongst the universals exemplified by
the particulars in question. Armstrong himself regards laws as being,
nonetheless, contingent, in the sense that although it may be the case
that F-ness necessitates G-ness in this, the actual world, there may be
other possible worlds in which F-ness and G-ness exist but in which
the former does not necessitate the latter. s Here it is important to note
that Armstrong is an "Aristotelian" or "immanent" realist concerning
universals - that is, he holds that a universal can only exist in a world
in which it is exemplified, at some time and some place, by some par-
ticular. 9 It should also be remarked that, in his more recent writings,
Armstrong has tended to play down the idea that what relates the uni-
versals involved in a law is some species of "necessitation", preferring
to use causal idioms of 'bringing about' or 'making happen' to de-
scribe this alleged relation. 1o The idea, then, seems to be that particular
instances of a law relating F-ness to G-ness will be particular cases of
something's being Fbringing about something's being G.
Finally, I should mention my own preferred view concerning the
logical form of natural laws. 11 This is that, in the simplest case, a natu-
rallaw has the form 'K is F' (where 'K' is a mass noun) or 'Ks are F'
(where 'K' is a count noun) - as, for example, 'Gold is electrically
conductive', 'Electrons have unit negative charge' or 'Planets move in
elliptical orbits', where these so-called generic propositions are by no
means logically equivalent to the corresponding universal generalisa-
tions quantifying over particulars, such as 'For all x, if x is an electron,
then x has unit negative charge' or 'For all x, if x is a planet, then x

moves in an elliptical orbit'. Thus I too, like Armstrong, conceive of
laws as involving universals rather than particulars, but distinguish, as
Armstrong does not, between substantial universals, or kinds, and non-
substantial universals, or properties and relations. 12 In the simplest
case, a law consists in a kind's possessing, or being characterised by, a
property. More complex cases involve two or more kinds standing in a
relation, as with the law that water dissolves common salt or the law
that positive and negative charges attract one another.
An advantage of this view over Armstrong's, I consider, is that it
requires no appeal to any "second-order" relation between universals
in saying what constitutes a law. Sometimes, indeed, laws are rela-
tional in form, as in the case of the law that water dissolves common
salt, but in such cases the relation involved is not a "second-order" re-
lation, since it is a relation in which particular objects can stand to one
another - and in which they do so stand in particular instances of the
law. A particular instance or exemplification of this law, thus, is any
particular case of some water's dissolving some common salt. Another
but closely related difference between Armstrong's proposal and mine
is that he seems to regard all particular instances of laws as involving
the exemplification of some general relation of 'bringing about' or
'making happen', whereas I favour the view that there is no such gen-
eral relation, only a broad family of specific causal relations, such as
dissolving, breaking, pushing, tearing, attracting, and so forth. On my
view, talk of 'bringing about', 'making happen' or 'producing' is just
an abstraction from more specific ways of talking about causal inter-
action. Moreover, my view provides a unified account both of causal
laws, such as the law that water dissolves common salt, and of non-
causal laws, such as the law that electrons have unit negative charge. 13
It may be objected to my view that it incorporates no species of
necessity at all into the constitution of natural laws and thus falls prey
to the complaint that it fails to distinguish between lawlike and acci-
dental generalisations. But this charge is unfair, for I do indeed distin-
guish between these two kinds of generalisation, albeit not in terms of
any notion of necessity. According to my account, a law has, in the
simplest case, the form 'Ks are F' (as in 'Electrons have unit negative
charge' or 'Lumps of uranium are less than one mile in diameter'),
whereas accidental generalisations are mere universal quantifications

over particulars, of the form 'For all x, if x is a K, then x is F', which
does not entail 'Ks are F'. Indeed, on my view, 'For all x, if x is a K,
then x is F', neither entails nor is entailed by 'Ks are F', because I hold
that laws - at least, non-fundamental laws - admit of possible excep-
tions. 14 The law that electrons have unit negative charge is fundamental
and has no possible exceptions, but a non-fundamental law, such as the
law that planets move in elliptical orbits (Kepler's First Law of Plane-
tary Motion), most certainly does have exceptions, because interfering
factors (such as the gravitational influence of other planets) can pre-
vent the instantiation of the law in particular cases.
Against me, it may perhaps be urged that, for example, it is in fact
true, but not a law, that lumps of gold are less than one mile in diame-
ter. But I would simply deny that this is true, even if it should happen
to be true that every particular lump of gold that ever has existed or
ever will exist is less than one mile in diameter. It is not true because
to assert 'Lumps of gold are less than one mile in diameter', as that
sentence would standardly be understood, is to assert something false
concerning the nature of the kind of stuff, gold, not something true
about some or all particular instances of that kind that have existed or
will exist. Anyone who asserted 'Lumps of gold are less than one mile
in diameter', meaning thereby to say either that some particular lumps
of gold are less than one mile in diameter or that all particular lumps
of gold are, would be using this sentence in a non-standard and mis-
leading way. This is clear from the fact that such a thing is not what is
standardly meant, mutatis mutandis, by someone who asserts, truly,
'Lumps of uranium are less than one mile in diameter', for such a
speaker evidently means thereby to say something about the nature of
the kind of stuff, uranium. 15
Here I should ~mphasise that I am by no means contending that all
so-called generic propositions, as linguists would identify this class of
propositions, are expressive of putative natural laws. For instance,
'Elephants are numerous in Africa' is not. This is because the predicate
involved here does not express a property that individual elephants can
possess, so that the proposition must be interpreted as expressing
something about the world's elephant popUlation - a collection or plu-
rality of individual animals - not something about the animal kind,
elephant, of which all of these individuals are instances. As for such

propositions as 'Electrons are smaller than dogs' or 'Cats are a nui-
sance', which may perhaps be doubted to possess lawlike status, I
would say merely that the former expresses a derivative law - one that
is derivable from laws concerning the sizes of electrons and dogs re-
spectively - and that the latter is an elliptical expression of a far-from-
fundamental sociological law concerning a relationship between cats
and some kinds of people. As I implied earlier, the test of whether a
generic proposition is expressive of a putative law is whether it is in-
terpretable as affirming that some kind is characterised by a certain
property or, in more complex cases, that two or more kinds stand in a
certain relation, and the last two propositions just mentioned certainly
seem to pass this test. Not all putative laws are putatively fundamental
laws and many of them may be of little interest to serious science.
All this being granted, it may now be objected against me that,
precisely because I do not distinguish between lawlike and accidental
generalisations in terms of necessity, I cannot explain why laws sup-
port counterfactual conditionals. Indeed, it may be charged against me
now that I have left no room for the notion of natural necessity at all
and have gone over entirely to the "Humean" camp. Not so, on either
count. First of all, we need to take into account how the semantics and
pragmatics of counterfactual conditionals work in these contexts,
which is as follows. 16 When entertaining a counterfactual possibility,
for the purposes of evaluating whether or not a given counterfactual
conditional is true, we have to hold certain facts concerning actuality
"fixed", while "unfixing" others. When evaluating a counterfactual
conditional such as 'If the universe had contained another lump ofura-
nium, it would have been less than one mile in diameter', we hold
fixed, amongst other things, the laws of nature, including the law that
lumps of uranium are less than one mile in diameter. This is why we
can and should evaluate that counterfactual conditional as being true.
(Obviously, we do not always hold fixed the laws of nature when
evaluating counterfactual conditionals: for instance, we patently do not
when evaluating one which begins 'If the laws of nature had been dif-
ferent ... '. However, I am not presently concerned with such condition-
als, only with conditionals such as 'If the universe had contained an-
other lump of uranium, it would have been less than one mile in di-
ameter', and with respect to these my remarks are, it seems clear,

highly plausible.)
Here it may be objected that an adherent of the "regularity" ac-
count of laws can make exactly the same claim and thus urge that on
that account, too, the counterfactual conditional should be evaluated as
being true. But in order to make this move the regularity theorist must
trade upon an ambiguity. According to the regularity theorist, the law
which obtains in the actual world, and which is consequently to be
held fixed, consists in the fact that all actually existing lumps of ura-
nium are less than one mile in diameter: but supposing that all of these
lumps of uranium are still less than one mile in diameter in the coun-
terfactual situation is perfectly consistent with there being an addi-
tionallump of uranium in that situation which is not less than one mile
in diameter. Merely to hold fixed the existence of a true proposition of
the form 'For all x, if x is a lump of uranium, then x is less than one
mile in diameter' is not to hold fixed the law, as this is conceived by
the regularity theorist. By contrast, if what is held fixed in the counter-
factual situation is not merely a fact concerning all actual instances of
the kind uranium but rather a fact concerning the nature of the kind
uranium itself, as on my own account of laws, then indeed we can see
why the counterfactual conditional should be evaluated as being true.
Exactly the same may be said on behalf on Armstrong's view of laws,
of course, so that this is not a respect in which my view has any ad-
vantage over his, only a respect in which both his and my views have
an advantage over the regularity theorist's view. 17
From this it will be obvious how I reply to the charge that I have
left no room for the notion of natural necessity at all. My view is that
there is such a thing as natural necessity, if that is what one wants to
call it, but that it is only a species of relative necessity. To say that a
state of affairs is 'naturally necessary' is merely to say that it is a state
of affairs which must be the case given the laws of nature. In possible-
worlds idiom, it is something that is the case in every possible world in
which the (actual) laws of nature obtain - or, at least, something that is
the case in every possible world in which all such (actual) laws obtain
as are relevant to the state of affairs in question. 18 (Thus we need not
assume, perhaps implausibly, that any two laws which obtain in one
world also co-obtain in any other world in which either of them ob-
tains.) Analogously, something is "legally" necessary, or obligatory, in

the forensic sense, if it must be the case given the laws of the land - as,
for example, it is legally necessary, or obligatory, that a citizen does
not steal or commit murder. Against this view, the scientific essential-
ist holds, as we saw earlier, that natural necessity is in fact nothing
other than metaphysical necessity, and so is "absolute" necessity. The
scientific essentialist may agree that a state of affairs is naturally nec-
essary if and only if it is the case in every possible world in which the
(actual) laws of nature obtain - but since he also contends that these
laws are themselves metaphysically necessary and so obtain in every
possible world whatever, this is consistent with his view that a natu-
rally necessary state of affairs is one which obtains in every possible
world whatever. We have yet to see in detail what might be said for or
against this essentialist doctrine. But all I want to claim at this stage is
that there is an alternative and ostensibly coherent account of laws and
natural necessity which does not treat laws as mere regularities
amongst particulars, which treats such laws as metaphysically contin-
gent, and which nonetheless accommodates both an explicable notion
of natural necessity and our conviction that laws of nature support
counterfactual conditionals.
Indeed, there is more than one such account of laws, since both
Armstrong's account and my own meet these criteria. But as for the
other "non-Humean" account of laws mentioned above, according to
which laws are necessitations of universal quantifications over par-
ticulars, my opinion is that this does not meet the criteria in question,
because it leaves unexplained the nature of the necessity at issue.
Plainly, one cannot non-circularly explain this notion of necessity in
terms of truth in all worlds in which the actual laws of nature obtain, if
one appeals to the very notion of necessity in question in explaining
what constitutes a law. Note here that there is no such circularity in
Armstrong's (original) view, according to which a law has, in the sim-
plest case, the form 'F-ness necessitates G-ness'. For it is clear that the
necessitation relation between universals which, on this view, helps to
constitute a law of nature is not to be confused with the "natural neces-
sity" which may be said to characterise states of affairs, where this is
explicated in terms of the obtaining of those states of affairs in every
possible world in which the (actual) laws of nature obtain. 19 On the
other hand, this very fact suggests that it is at least misleading to speak
of that alleged relation as one of "necessitation" and this may help to
explain Armstrong's tendency to downplay such talk in his more re-
cent writings.
Let us assume, in what follows, that laws of nature do indeed con-
cern universals rather than particulars. Scientific essentialists may be
expected to agree, because one of the primary claims often made in
support of their view is that natural kinds and properties depend for
their very identity upon the laws into which they enter.20 Thus, one
contention is that a property owes its identity to the contribution it
makes to the (perhaps conditional) causal powers of physical objects
possessing that property, as determined by the causal laws governing
interactions involving such objects. For example, if we ask what prop-
erty mass is, we may be told that mass is that property in a body which,
amongst other things, makes it necessary for a force to be applied to
the body in order to accelerate it, in accordance with Newton's Second
Law of Motion, F = Ma. Again, we may be told that sphericity is that
property which, amongst other things, makes a body which possesses it
liable to roll down an inclined plane (provided that the body is also
rigid and subject to a gravitational force), in accordance with various
natural laws. In support of this view, it may be pointed out that we can
only detect the properties of physical objects by interacting with them
or with other objects affected by their activities. So it seems reasonable
to adopt something like the following criterion of identity for (physi-
cal) properties: the property of being F is identical with the property of
being G if and only if F and G make the same contribution to the
causal powers of physical objects possessing them. Then, given that
the causal powers of objects are determined by the natural laws gov-
erning the properties (universals) which those objects possess, the
foregoing criterion seems to reduce to this: the property of being F is
identical with the property of being G if and only if F and G enter into
the same laws in the same ways.
Thus, consider all the laws into which the property of mass, M,
enters, such as the law that force equals the product of mass and accel-
eration (F = Ma) and the law (Newton's Law of Gravitation) that the
gravitational force between two bodies is proportional to the product of
their masses and inversely proportional to the square of the distance
between them (F = GMI M2!-?). (Purely for the purposes of argument, I

am assuming the truth of classical Newtonian physics rather than of
Einsteinian relativistic physics.) Then, any property P, which entered
into exactly the same lawful relations with other properties as does M,
the property of mass, would have to be regarded as being identical
with the property of mass, since there would be no possible way of
distinguishing a body's possession of P from its possession of M in
terms of any effect it could have on us or other bodies. But then, it may
seem, we are committed to the essentialist doctrine that the laws of
nature are metaphysically necessary - that any law which obtains in
this, the actual world, must obtain in every possible world. For if the
law involves certain properties (universals), PI to Pn , and these prop-
erties, like all properties, owe their very identity to the laws in which
they are involved, then any world in which those very properties exist
will be a world in which the law in question obtains, and any world in
which they do not exist will be a world in which the law obtains vacu-
ously.21 Hence the law will obtain in any world whatever and so be
metaphysically necessary.
Unfortunately for the scientific essentialist, this reasoning is quite
fallacious, because it illicitly assumes that a criterion of identity for
properties in this, the actual, world must also serve as a principle of so-
called transworld identity. Consider, by way of analogy, Donald Da-
vidson's famous (though now abandoned) criterion of identity for
events, whereby event el is identical with event e2 if and only if el and
e2 have exactly the same causes and effects?2 However reasonable this
criterion may be - and, certainly, it is not easy to dispute its truth, even
if it may be objected that it is implicitly circular as a criterion of iden-
titj3 - it cannot at all plausibly be made the basis of a principle of
transworld identity for events, because we have a strong intuition that
any particular event could have had at least some causes and effects
other than those it actually has. Indeed, if we do tum this criterion into
a principle of transworld identity, we saddle ourselves with the doc-
trine that all singular causal relations are metaphysically necessary:
that if an event el in fact caused an event e2, then el causes e2 in every
possible world in which either of these events exists.
Now, perhaps the scientific essentialist would be happy to accept
this too - indeed, it may be that, to be consistent, he must accept it. But
this does not detract from my point that, so far, we have been given no
good reason to adopt this position, nor the corresponding position re-
garding the transworld identity of properties. Quite generally, one can-
not advance directly from an intraworld criterion of identity to an in-
terworld criterion of identity. As a reductio ad absurdum of the suppo-
sition that one can, consider the claim that because two material ob-
jects of the same kind cannot occupy the same place at the same time -
so that spatiotemporallocation provides a criterion of identity for ma-
terial objects of any given kind - it therefore follows that no material
object of a given kind could have occupied a spacetime location differ-
ent from its actual spacetime location - that this very chair, say, could
not have been located where that chair is now. Plainly, no such conclu-
sion does follow.
At this point, the scientific essentialist may reply that properties
are relevantly different from particular objects and events where these
issues of identity are concerned. 24 Objects and events are non-
repeatable entities with unique spatiotemporallocations, so that we can
identify them in this world in ways which differ from the ways in
which we identify them "across" worlds. He may try to press home the
point by urging that it simply does not make sense to suppose that the
very same property - mass, let us say - might, in another possible
world, make the same contribution to the laws of that world that, let us
say, electrical charge does in this, the actual world. Or, to take another
example, he may urge that it simply doesn't make sense to suppose
that there might be a world in which the kind electron exists, but in
which the law is that electrons have unit positive charge instead of unit
negative charge. These otherworldly "electrons" would surely just be
positrons.
However, one can oppose scientific essentialism without being
driven to accept these extreme consequences. If we consider less ex-
treme variations in the laws, the scientific essentialist's case seems
much less plausible. Must we say, for instance, that a world in which
the law of 'gravitation' is an inverse cube law, or a law making the
force inversely proportional to a non-integral power of the distance
very slightly different from 2, is a world in which the property of mass
does not exist? What about a world in which the universal constant of
gravitation, G, has a slightly different value from its value in this world
- is that likewise a world in which the actual world's property of mass

does not exist? Again, must we say that a world in which the ratio
between the value of G and the value of the charge on the electron dif-
fers somewhat from the ratio obtaining in the actual world is a world in
which neither mass as we know it nor electrons as we know them ex-
ist? It would be highly speculative to respond to any of these questions
by urging that a future physics will one day discover mathematically
necessary connections between all the constants of nature which render
it metaphysically impossible that these constants should have values at
all different from their actual values. Certainly, nothing in current
physics can give us any confidence in such a prospect. Indeed, we can-
not even treat as sacrosanct the idea that the so-called "constants" of
nature, such as G, really are constant over time. 25 But ifG were discov-
ered to be gradually increasing over time, would it not be absurd to
conclude that the property of mass which exists now is not the same
property of 'mass' which existed at some time in the past?
Let it be clear that I do not want to challenge scientific essential-
ism in its entirety. I am prepared to accept, or at least to countenance,
the Kripke/Putnam view that water is essentially H2 0 and that common
salt is essentially sodium chloride, NaCp6 What I dispute is that it in
any way follows from this that the natural law that water dissolves
common salt is metaphysically necessary. In every possible world in
which both water and common salt exist, I am prepared to accept,
those substances are composed of molecules consisting, respectively,
of two hydrogen ions and one oxygen ion and of one sodium ion and
one chlorine ion. Likewise, I am prepared to accept, for instance, that
an oxygen atom, in any possible world in which it exists, consists of a
nucleus containing eight protons and a closely comparable number of
neutrons, surrounded by eight orbital electrons. Now, it is true that the
chemical interactions of water and common salt are determined by
their atomic structure and the laws governing their atomic constituents,
especially their orbital electrons. But, provided that we can accept that
atomic constituents of those very kinds - electrons, protons and so
forth - can exist in worlds in which the constants of nature have
somewhat different values, or in which power laws differ somewhat in
the values of the exponents involved, then we must accept that water
and common salt may exist in worlds in which the characteristic
chemical interactions of those substances differ significantly from

those which they exhibit in this, the actual world - perhaps even to the
extent of there being worlds in which, although water and common salt
both exist, the former lacks the power to dissolve the latter, so that it is
not a law in that world that water dissolves common salt.
At this point I should like to challenge one of the mainstays of sci-
entific essentialism, namely, its presumption that the Kripkean expla-
nation of how a posteriori knowledge of metaphysically necessary
truths is possible is extensible to our knowledge of natural laws, con-
ceived as constituting metaphysically necessary truths. We saw earlier
how this explanation works with a simple truth of identity concerning
particular material objects, such as the fact that Hesperus is identical
with Phosphorus. We know a priori that if Hesperus is identical with
Phosphorus, then it is metaphysically necessary that Hesperus is iden-
tical with Phosphorus (accepting, for current purposes, the validity of
the Barcan-Kripke proof of the necessity of identity).27 And we can
establish empirically that Hesperus is identical with Phosphorus, be-
cause we can determine by astronomical observation that their orbital
positions coincide. Hence we can infer, by modus ponens, that it is
metaphysically necessary that Hesperus is identical with Phosphorus.
But, of course, we can only establish the empirical truth that Hesperus
is identical with Phosphorus because we can appeal in this case to an-
other a priori principle, namely, a criterion of identity for planets, con-
ceived as a kind of material object. This criterion tells us that no two
planets can exist in the same place at the same time, whence we can
conclude, from the astronomical observation that Hesperus and Phos-
phorus coincide in their orbital positions, that Hesperus is identical
with Phosphorus.
However, as will be obvious, this criterion of identity for planets is
an intraworld criterion, not an interworld or "transworld" criterion.
And that fact is vital to our ability to settle the identity question con-
cerning Hesperus and Phosphorus empirically. If the only admissible
principle for determining the identity or diversity of Hesperus and
Phosphorus were a transworld principle, then we would not, after all,
be able to determine their identity empirically, for it would not then
suffice to adduce facts concerning those planets which obtain in the
actual world and which are therefore empirically accessible to us, such
as facts concerning their actual orbital positions - we should need also

to adduce facts concerning them which obtain in every other possible
world in which they exist, and such facts are not empirically accessible
to us.
Now we are in a position to construct a dilemma for the scientific
essentialist, as follows. Either the essentialist allows, in the case of
properties and kinds, that a distinction is to be drawn between an in-
traworld and an interworld criterion of identity for such entities, or else
he does not allow this and holds that the same criterion necessarily
governs both intraworld and interworld identity in such cases. If the
essentialist adopts the first option, then his claim that laws of nature
are metaphysically necessary is undermined. For according to this op-
tion, while it may be accepted, say, that the property of being F is
identical with the property of being G if and only if F and G enter into
the same (actual) laws in the same ways, it must at the same time be
agreed that, because this is only an intraworld criterion of identity, it is
also metaphysically possible for F or G to have entered into different
laws. 28 On the other hand, if the essentialist adopts the second option,
as he really seems bound to do, he must relinquish the claim to be able
to apply the Kripkean model of how our a posteriori knowledge of
natural laws is possible and hence render altogether mysterious how
empirical knowledge of such laws, conceived as metaphysically neces-
sary, is possible.
Consider, for instance, the law that electrons have unit negative
charge. How do we know that this law is true? The obvious thing to
say is that we know that it is true, or at least have very good empirical
grounds for believing that it is true, because scientists have conducted
many experiments (such as Millikan's oil drop experiment) to measure
the charge on particular electrons and have found that it consistently
takes a certain negative value, which has consequently been defined as
the unit negative charge. But to conduct such an experiment we must
in principle be able to identify some particular entity empirically as
being of the kind electron and be able to determine by empirical means
that this entity possesses the property unit negative charge. Yet how
can we determine empirically, in any given case, that we are indeed
confronted with exemplifications of the universals electron and unit
negative charge, if the very identity of these universals turns on what
laws they enter into in every possible world (including, of course, the
actual world)? If it is part of the very essence of electronhood and of
unit negative charge that electrons have unit negative charge, because
this law involving them obtains in every possible world, how do we
know that our world is one in which these very universals do indeed
exist and are exemplified by particular entities, as opposed to being a
world in which the law in question is merely vacuously true and the
entities that we experiment upon exemplify quite different universals?
The scientific essentialist may be tempted at this point to appeal to
something like Hilary Putnam's notion of a 'stereotype'?9 According
to Putnam, a natural kind term, such as 'water' or 'lemon', has associ-
ated with it a bundle of stereotypical properties - in the case of
'lemon', the properties, say, of being yellow, juicy and acidic - which
do not necessarily belong to every individual exemplar of the natural
kind in question but which do belong to "typical" exemplars and which
can therefore be used, albeit only defeasibly, to identify something as
being an exemplar of that kind. However, in the first place, if these
stereotypical properties are indeed genuine properties - universals -
possessed by individual objects, then the scientific essentialist ought,
in all consistency, to say that these properties too owe their very iden-
tity to the laws into which they enter, in which case it becomes equally
mysterious how we can ever know empirically that anything exempli-
fies them. Secondly, when we are concerned with fundamental kinds
such as the kind electron, it appears that we can no longer suppose that
there are stereotypical properties associated with the kind but which
not every individual exemplar of the kind necessarily possesses: all
electrons, it seems, are necessarily exactly alike in respect of their in-
trinsic properties. Hence, the proposed model for our ability to identify
exemplars of natural kinds empirically, even if it served the scientific
essentialist's purposes in some cases - which I do not think it does -
would not serve them in the case of fundamental natural kinds. I con-
clude that the scientific essentialist has an undischarged burden of ex-
plaining how, in general, our empirical knowledge of natural laws can
be possible if, as he maintains, the very essence of a property resides in
its lawlike connections with other properties. 30
Notes
1 There is controversy amongst commentators as to whether Hume was an
antirealist or a sceptical realist concerning causal necessity: see Strawson [1989]. I
adopt no particular position in this paper on this historical issue.
2 See, especially, Shoemaker [1998], Fales [1993], and Ellis [1999]. See also
Shoemaker [1980] and Martin [1993].

3 The locus classicus is Kripke [1980].
4 The term 'broadly logical necessity' is Alvin Plantinga's: see Plantinga [1974]
p.2. See further Lowe [1998] p.13ff.

5 Here I follow the lead ofD. M. Armstrong: see Armstrong [1983], p.8.
6 For a partial defence of the view that laws have this form, see Foster [1982].
7 See Armstrong [1983], p.85
8 See Armstrong, [1983], p.158ff

9 See Armstrong, [1989b], pp.75ff.
10 See Armstrong's replies to Smart and to Menzies in Bacon et al. (eds.),
[1993], p.l72 and p.229, and also Armstrong, [1997], p.223ff (he endorses the
'making happen' idiom at pp.210-11).
11 See further Lowe [1989], Chapter 8.
12 For Armstrong's dismissive view of kinds, see Armstrong [1997], p.65ff.
13 Armstrong maintains that non-causal laws are supervenient: see Armstrong,
[1997], p.231 ff.

14 See further Lowe [1989], Chapter 8, and Lowe [1987].
15 Compare Drewery [2000].
16 For a fuller account of my views about counterfactual conditionals, see Lowe
[ 1995].
17For Armstrong's deployment of a similar argument on behalf of his own view
oflaws, see his reply to Fales in Bacon et al. (eds.), [1993], p.144ff.
18 But see further Lowe [1987], for reasons to modify this proposal in certain
ways which, however, do not undermine the argument of the present paper. For more
on the distinction between "relative" and "absolute" necessity, see Lowe [1998],
p.18ff.
19 Of course, Armstrong himself is not a realist concerning possible worlds: see
Armstrong [1989] and Armstrong [1997], p.l72ff. For my own view of the ontologi-
cal status of possible worlds, see Lowe [1998], p.256ff. Nothing I say in the present
paper in the language of possible worlds is intended to imply a commitment to their
reality.
20 See the papers cited in note 2 above.
21 See further Ellis [1999], p.30.
22 See Davidson [1980]. He abandons the proposal in Davidson [1985].
23 For more on the circularity objection to Davidson's criterion, see Lowe
[1998], p.43.
24 One essentialist who does at least discuss the matter - though not, to my
mind, entirely satisfactorily- is Sydney Shoemaker: see Shoemaker [1998].

25 In fact, E.A. Milne's cosmological theory of kinematic relativity, which was
perfectly respectable from a scientific point of view, proposed a secular increase in
the value of the "constant" of gravitation: see Bondi [1961], Chapter 11.
26 See Kripke [1980], and Putnam [1975].
27 In point of fact, I have doubts about the Barcan-Kripke proof: see Lowe
[ 1982].
28 I do not mean to suggest that I endorse this intraworId criterion of identity for
properties myself: in fact, I reject it on the grounds that it is circular. For my own
account ofthe identity-conditions of properties (universals), see Lowe [1999].
29 See Putnam [1975].
30 I am very grateful for comments received when this paper was presented at
the Bergamo Conference on 'Individuals, Essence, and Identity: Themes of Analytic
Metaphysics', and especially grateful to Dr. Kathe Trettin, who replied to the paper. I
am also grateful to an anonymous referee for some criticisms which prompted me to
clarify certain aspects of the paper.
KINDS OF NECESSITY: A COMMENTARY
ON E.l. LOWE'S PAPER "KINDS, ESSENCE,
AND NATURAL NECESSITY"
Kathe Trettin
Jonathan Lowe offers a subtle but nonetheless determined attack on

the view, recently espoused by some philosophers, that a Theory of
Causality and Natural Laws requires nothing less than "metaphysical
necessity", if natural laws are to be something more than mere regu-
larities. On this view, natural or nomological necessity is supposed to
be as strong as logical necessity in virtue of combining causal neces-
sitation with the logic of possible worlds. For Jonathan Lowe, how-
ever, laws of nature are not metaphysically necessary. Instead, he de-
fends a Contingency Theory of Natural Laws, be they causal or not,
although he rejects a Regularity Theory. I shall briefly reconstruct
Lowe's argument and then point to some questions which arise about
it.
One way to reconstruct an argument, is to employ the scholastic
method of introducing a proponent and an opponent. As the paper of
Jonathan Lowe in a way invites this method, I shall use it in order to
reconstruct the argumentation. So, in this scenario, the Proponent is
played by Sydney Shoemaker, Evan Fales, and C.B. Martin. They are
the "bad guys", the so-called 'Scientific Essentialists'. But, as it turns
out, they are not all that bad. This is because of a mediating authority,
207
A. Bottani et aI. (eds.J, Individuals, Essence and Identity, 207-213.
commonly called 'Saul Kripke'. The Opponent is, of course, played by

Jonathan Lowe himself, allied by a force called David Armstrong, with
the caveat that between Lowe and Armstrong a subsidiary argument
comes into play. Finally, in the background, we find a prominent
authority, namely David Hume, whose Regularity Theory, however, is
considered both by Proponent and Opponent as a spent force. So much
for the setting.
Now, what are the claims of the Proponent to be critically assessed
by our Opponent?
1. Causal features of properties are essential to them - are fea-
tures they have in all possible worlds. (Shoemaker [1998],
p.66)
2. Properties having the same causal features are identical.
(Shoemaker [1998], p.64)
3. Causal (and nomological) connections are necessary. (Fales
[1993], p.l22)
4. Giving an account of law-supported counterfactuals requires
objective necessities. (Fales [1993], p.l26)
5. There can be no objective truth conditions for a counterfactual
unless there is something fixed across worlds. (Fales [1993],
p.127)
6. There is inter-world and trans-temporal identity of causal fea-
tures.(Shoemaker [1998], p.66)
7. Causal necessity is a special case of metaphysical necessity,
and is necessity in the strongest sense of the term. (Shoemaker
[1998], p.61)
8. Necessary truths can be known a posteriori, i.e., empiri-
cally.(Shoemaker [1998], p.60)
9. The necessity of causal relations is primitive and sui generis.
(Martin [1993], p.140)
Kathe Trettin, Kinds a/Necessity: a Comment on E.J Lowe 209
10. Laws are relations between properties taken either as universals

(Shoemaker, Fales) or as particulars (Martin).
In other words, the Proponent not only rejects the Regularity Theory of
Causality, but advocates a Strong Necessity Theory of Causal Laws,
combining logical and natural necessities into something called 'meta-
physical necessity', thereby stressing the epistemological point that
metaphysically necessary truths are in principle empirically accessible.
Against this rich proposition, the Opponent offers an equally rich
contraposition.
1. Natural laws, and therefore, causal laws, are contingent.
2. There is natural necessity, but it is relative to the conditions
obtaining in the actual world.
3. One cannot extend intra-world identity to inter-world identity.
4. Genuine laws and mere regularities are to be distinguished.
They are grounded in different entities and their statements
have different logical forms.
5. Natural laws (causal or not) are ontologically grounded in
kinds and their possessing a property which have in the sim-
plest case the form 'K is F' (for mass nouns) or 'Ks are F' (for
count nouns). They support "context-sensitive" counterfactu-
also
6. Kinds are "substantial universals", properties and relations are
"non-substantial universals".
7. Regularities, by contrast, are ontologically grounded in par-
ticulars and properties, usually having the form of universal
quantification over particulars: 'For all x, if Fx , then Gx'. They
do not support counterfactuals.
8. Essentialism does not entail metaphysical necessity: if water is
essentially H20, and common salt is essentially sodium chlo-
ride, NaCl, it does not follow, that the natural law that water
dissolves common salt is metaphysically necessary, i.e., a law
in all possible worlds.
In other words, the Opponent just like the Proponent is an adherent of

something stronger than the Regularity Theory of Causal Laws and
thus maintains that laws are to be ontologically reconstructed in a re-
alistic spirit. And here the kinds come in. What Jonathan Lowe, at
least on my interpretation, tries to disentangle, is a confusion of essen-
tial entities that constitute a law with the strongest possible necessity
of a law itself. While a law is contingent or relative to the actual and
therefore empirically accessible world, the entities which constitute the
law cannot be accidental but have to be essential for there to be a law
at all, and not a mere regularity. To put it even more simply: endorsing
essentialism in metaphysics is not the same as endorsing the meta-
physical necessity of natural laws.
So much for summing up the controversial proposals. But as is
usually the case in philosophy, proposals and standpoints are far less
interesting than the arguments leading to them. In order to do justice to
the arguments, at least partially, I shall pick out two lines of reasoning
which seem, at least to my mind, to be at the core of the controversy
pointed out by Jonathan Lowe. The first one concerns the question:
what are the identity conditions for entities in order to constitute a law
in its strictest and most general sense? The second concerns the fol-
lowing question: why are kinds, or "substantial universals", privileged
to meet these conditions?
Let us start with the identity question. Proponent as well as Oppo-
nent agree on two presuppositions: (1) Natural laws must support at
least some sort of counterfactual conditionals. (2) Natural laws must
be in principle empirically accessible. The controversy arises over the
question how counterfactuals can be supported, i.e. whether and how
the identity of entities which enter into counterfactuals is in principle
knowable by actual epistemic agents. Whereas the Proponent claims
that the identity of entities is fixed by their causal or nomological rela-
tions across all possible worlds, the actual world being one of them,
there is no problem of accessibility, our Opponent rejects the view that
the identity of entities is determined by the laws they enter into. "How
can we determine empirically, in any given case", he asks, "that we are
indeed confronted with exemplifications of the universals electron and
unit negative charge, if the very identity of these universals turns on
what laws they enter into in every possible world?" One cannot extend
Kathe Trettin, Kinds of Necessity: a Comment on E.J Lowe 211
intra-world identity conditions to inter-world conditions, because one

has no empirical clue to those conditions.
This is, prima jacie, a very plausible objection: if the Necessitari-
ans. want to know their natural laws in the actual world, at least in
principle, (and where else could they know anything at all), they
should not derive the identity conditions from universes which nobody
can know determinately. Notice, however, that this objection is a
purely epistemological one, whereas identity conditions are ontologi-
cal conditions, at least, if identity is not construed as a conventional
term, or taken as purely dependent on our capabilities of identifying
things. Jonathan Lowe is, of course, quite aware of the fact that a
purely epistemic objection will not suffice, especially when Proponent
as well as Opponent agree on a realistic spirit in ontology. What is
needed, then, is an argument which supplies good ontological candi-
dates to substitute for the rejected trans-world necessity.
The ingenious move of Jonathan Lowe is, now, actually to build
up a classical category as the ontological supporter of laws in general,
a category which is not otherworldly, but belongs to the actual philo-
sophical tradition from at least Aristotle, namely, the species, or kind.
Kinds, e.g., gold or electrons, are supposed to be the entities to ensure
a "natural necessity" which is, however, not dependent on possible
world conditions, but relative to conditions of the actual world. This
move is ingenious, because kinds, being the universals they are, save
one, on the one hand, from relapsing into the Regularity Theory which
is built upon particulars. Moreover, the kinds-view seems to offer a
good alternative to the Necessitarian view, because at least fundamen-
tal or natural kinds are purported to have essential features. Hence, a
lot depends on what kinds really are. And here not only a second
question, but a whole slew of questions arise, and they more or less
involve the subsidiary argument which our Proponent has with David
Armstrong: why are kinds privileged entities to support laws? What
exactly is the difference between "substantial" and "non-substantial"
universals? And, are kinds really irreducible, if they have, on Lowe's
account, to be instantiated or be at least instantiable, just like Arm-
strong's universals?
The shortest answer, at least to the third question, that I have
found in Jonathan Lowe's work, is the one he states in the introductory
chapter of his book Kinds of Being (1989): "[ ... ] the notion of an 'indi-
vidual' and of a 'sort' or 'kind' are opposite sides of a single concep-
tual coin: each is only understandable in tenns of the other. Individuals
are necessarily individuals ofa kind, and kinds are necessarily kinds of
individuals". Therefore, "a realism with regard to particulars or indi-
viduals [... ] implies realism with regard to sorts or kinds" (Lowe
[1989], p.4f.). Although this quasi-Hegelian fonnulation sounds a fa-
miliar note to my Gennan ear and therefore I have no trouble under-
standing it, I would argue that this interesting mutual dependence
should be spelled out not only in conceptual but in ontological tenns.
What is ontological dependence and in what respect does it affect the
dependence between individuals and kinds? Although Jonathan Lowe
points to that very deep and interesting question throughout this book,
I cannot see an entirely satisfactory answer. Are kinds primary enti-
ties? On Lowe's view, they are at least indispensable, whenever one
accepts individuals in one's ontology. But if kinds have to be instanti-
ated or exemplified by individuals in order to count as real entities,
why should individual qualities not be able to build up those kinds, for
example, by constituting resemblance classes? In what sense, then, are
kinds ontologically stronger than resemblance classes of tropes? On
the other hand, if one holds, as our Opponent does, that individuals
and kinds necessarily depend on each other and that kinds are the very
tenns of lawlike statements, why should one not characterise this de-
pendence as some sort of metaphysical or ontological necessity? Al-
though this is implied by Lowe's own proposal, one might have
wished a more explicit assertion in order to make sure that metaphysi-
cal necessity is not to be identified with a brand of causal-cum-
possible-worlds-realism.
There is, of course, a good reason for Jonathan Lowe not to reduce
kinds. For ifhe did, he would eventually either fall back to the camp of
David Hume, or he would eventually find himself slapped on the back
by one of the Necessitarians with a cordial welcome to their club. In
order to avoid unacceptable company, Lowe is quite right to fortify his
favourite entity, the Kind, and make it as strong as possible. The new
label is: 'substantial universal'. This is, at least to my understanding, a
clear announcement to the effect that not any old universal can serve
as a supporter of laws. Moreover, the distinction between substantial
Kathe Trettin, Kinds of Necessity: a Comment on E.J Lowe 213
and non-substantial universals is supposed to dispense with David

Armstrong's higher-order universals. Of course, I cannot discuss this
issue here, but what I can do, is to point to the immediate questions
which arise from this strategy.
1. What are the identity conditions for kinds or substantial univer-
sals, if they are not reducible to resemblance classes of indi-
vidual qualities or tropes?
2. How do we know kinds ifnot by their instantiations?
I am sure that Jonathan Lowe has good answers to these questions,
even if I presumably will not agree with him on some of them. In any
case, I am very grateful that he has stressed three decisive points: (1)
that natural laws have to be accounted for in a realist way; (2) that
qualities construed in one way or another enter essentially into laws
without assuming a special transworld relation of necessitation, and (3)
that dependence of entities in ontology is not a deplorable state of af-
fairs.
Part Three
Identity
ON THE NOTION OF IDENTITY IN
ARISTOTLE
Mario Mignucci
1.
As is well known, a chapter in the first book of the Topics is dedicated

to exploring the various senses or uses of the word 'same' (tauton),
and its style is to a certain extent reminiscent of the famous lexicon of
the philosophical terms in Metaphysics ~. The reason for its insertion
after the definition ofpredicables probably lies in the fact that a refer-
ence to identity is made not only in relation to definition, where it is
said that checking a definition in most cases consists in testing identity
and difference,! but also in relation to genus, where the question
whether two things fall under the same genus is relevant in many
ways.2 If we stick to the received view that the Topics is one of Aris-
totle's earliest works,3 the analysis of identity we find in I 7 is proba-
bly his first attempt to make a map of the senses of sameness. 4 He
starts by introducing a tripartite division of the use of 'tauton'. Let us
read the passage:
(A) We may regard the same as being divided, in outline, in three parts,
for we are accustomed to describe what is the same as in number or in
species or in genus. Those are the same in number which have several
names but the thing is one, for example cloak and coat. Those are the
same in species which, though many, are indistinguishable with respect to
species, for instance a man and another man and a horse and another horse
217
A. Bottoni et aI. (eds.J, Individuals, Essence and Identity, 217-238.
(for those things are said to be the same in species which fall under the
same species). Similarly, those are the same in genus which fall under the
same genus, as horse and man. (Top. I 7, 103'7-14, translation by Robin
Smith slightly modified, see Smith (ed.) [1997])
This is not the only place where an examination of the word 'tauton' is
offered,s or where this tripartition of the uses of 'sameness' is men-
tioned. 6
Numerical identity is probably the only kind of identity that one
would expect to find under the heading of sameness, so much so that,
below, when referring to numerical identity, we shall simply call it
'identity' without any further qualification, unless this is required by
the obvious necessity offollowing Aristotle's terminology.
Aristotle claims that 1..1 and ~ are (numerically) the same if they
have different names denoting one thing. To formalise his characteri-
sation of identity and the ways in which, according to him, 'tauton' is
used, we must make use of a language slightly richer than the usual
one adopted in a standard first order calculus with identity. We have to
introduce not only 'a', 'b', 'c', ... , as arbitrary names for individuals
and 'x', 'y', 'z', ... , as variables ranging over a domain of individuals,
but also '1..1', 'v', '~', ... , as arbitrary names for general or individual
entities, and 'p', '0', 'T', ... , as variables ranging over a domain of
general or individual entities. Then, if we take Aristotle's words as a
sort of definition of identity, we can say that 1..1 and ~ are the same, i.e.
(1)1J=~ if (i) '1..1' is a name for 1..1

(ii) '~' is a name for ~
(iii) '1..1' "* '~'
(iv)* '1..1' and '~' denote one thing
We can express condition (iv)* more elegantly by stating
(iv) '1..1' and '~' are co-referential
This characterisation of identity deserves some comment. First of all,
one should be aware that in Aristotle's view identity is a relation be-
tween things and not between names, despite the linguistic approach
chosen here. It is not that the names 'cloak' and 'coat' are the same. In
Mario Mignucci, On the Notion of Identity in Aristotle 219
fact, they are not the same: 'cloak' is a different name from 'coat'. It is
what is denoted by 'cloak' that is the same as what is denoted by
'coat'. Therefore, identity holds between things, pragmata, and not
linguistic entities. As we shall see, this observation is crucial for the
development of our analysis. 7
There are, however, at least two disturbing aspects of this way of
putting things. The first depends on the example of numerical identity
offered by Aristotle in the text. We would expect him to quote the case
of an individual with two names, as for instance 'Tullius' and 'Cicero'
for Cicero. But he mentions 'cloak' and 'coat', which are general
terms. For those of us raised in the protective shadow of Frege it may
be shocking to accept an identity relation between the denotata of non-
individual terms. The impression that Aristotle is not interested in dis-
tinguishing the case of sameness among individuals and sameness
among general or abstract entities is confirmed by the fact that else-
where he states a proposition expressing sameness of a particular with
itself as an example of (numerical) identity. 8 I will not discuss here this
apparent anomaly of Aristotle's approach to identity. Let me only ob-
serve that his view does not seem to be conditioned by the context in
which the analysis of identity is carried out in the Topics. In the Meta-
physics, where he seems very concerned to avoid assigning an onto-
logical import to the denotata of general terms, we find the same free
attitude towards the bearers of the identity relation: they can be not
only individuals but also what is referred to by universal terms. 9
Let us concentrate on the second awkward feature of Aristotle's
way of characterising identity in text (A). The linguistic aspect of it
should not by any means go unnoticed. Identity is said to occur when
the names of the entities involved by the relation refer to one and the
same object. As Paolo Crivelli in his comments has acutely observed,
this approach to identity implies that things to which identity applies
must have a name, and this assumption is not at all obvious. Worse
than that, identity seems to apply only to things which have at least
two possible names, as condition (iii) in (1) suggests. Apart from the
fact that it is not at all clear that we can give a name to a grain of sand
in the sea, although we can claim that it is self-identical, Aristotle is
well aware that names cannot match things, since the former are finite
and the latter infinite in number. 10 Even if we include definite de scrip-
tions in the notion of an Aristotelian name we do not solve the prob-

lem. All possible combinations of a finite number of names are them-
selves finite and they cannot equate the number of infinite things to
which sameness applies. Of course, identity has a linguistic aspect. We
cannot express identity without referring to a language and a way of
picking out things since, as we have seen, identity holds between
things. But this obvious remark does not entail that we are compelled
to assign a name (or a definite description) to any object of which we
say, for instance, that it is identical to itself. In a natural language we
have quantifiers and expressions such as 'thing', 'object', 'entity' to
help cope with the limited number of names at our disposal, and some
refined versions of these devices are used in formal languages.
Needless to say, (1) cannot be counted as a definition of sameness.
Condition (iv) (or (iv)*) contains a clear reference to identity and the
same probably holds for condition (iii). Therefore, conditions (i)-(iv)
cannot be taken as a proper definiens of sameness, since they include
what must be defined. As one should expect, (1) can be considered
only as an elucidation of the notion of identity, and this elucidation
naturally depends on the context in which sameness is supposed to op-
erate. The context of Aristotle's analysis is dialectic, i.e. the technique
of discussion between two opponents. It is from this point of view that
definitions are examined and, as we have seen, discussion about defi-
nitions and genera is one of the main reasons for appealing to identity.
For instance, a definition can be disproved if one is able to show that
its alleged definiens is not the same as the definiendum. Similarly, we
can reject the claim that 1.1 is the genus of v and ; if it can be main-
tained that TT is the genus of;, 1.1 is the genus of v, and 1.1 is different
from TT.
If this is the context of Aristotle's use of identity it should not sur-
prise us too much that he has recourse to a linguistic approach in his
elucidation of this notion. In some sense definientia and definienda can
be taken as names of entities, and genera can be considered in the
same way. II This view is confirmed by the fact that when Aristotle
treats identity in the context of his ontology, his approach changes.
Consider for instance the following passage:
(8) The same has several meanings: we sometimes mean the same nu-
merically; again we call a thing the same if it is one both in definition and
in number, e.g. you are one with yourself both in form and in matter; and
again if the definition of its primary substance is one, e.g. equal straight
lines are the same, and so are equal and equal-angled quadrilaterals -
there are many such, but in these equality constitutes unity. (Metaph. I 3,
1054a32-b3, Oxford revised translation modified, see Barnes (ed.) [1985])
For our purposes, it is sufficient to underline the different perspective

according to which identity is characterised. It is no longer explained
in terms of sameness of names' denotata, but in terms of sameness of
the ontological constituents of the bearers of the relation, matter and
form. The details of the passage are far from clear. It is not obvious in
what sense numerical identity differs from unity of definition and
number. Maybe, as the pseudo-Alexander thinks,12 Aristotle is here
hinting at the distinction between accidental and essential identity. \3
Nor need we find perspicuous the example of the equal straight lines
as a special case of sameness in definition, i.e. specific identity. How-
ever, what really matters for us is the ontological shift in the charac-
terisation of sameness with respect to the approach in the Topics. We
find the same ontological commitment in characterising identity in
other passages of the Metaphysics in which no mention of the co-
referentiality of names is made. 14
There is however an aspect in which the ontological characterisa-
tion of identity and the linguistic coincide, and this is the attempt that
Aristotle makes to reduce sameness to oneness. In the Topics it is said
that J.1 and ~ are the same if their names denote one thing and in text
(B) it is oneness in number that establishes sameness. More explicitly,
in another passage of the Metaphysics Aristotle clearly says that iden-
tity is a kind of oneness. 15 It is difficult to evaluate the meaning of this
reduction. One might be tempted to take it seriously and spell it out in
terms of indistinguishability: J.1 and ~ are the same if they are one, i.e.
if they cannot be distinguished, and, of course, J.1 and ~ cannot be dis-
tinguished if they share all properties and attributes. '6 We could ex-
press this by positing
(1 *) J.1 =~ if \/F (F(J.1) B F(~»

222 Individuals. Essence. and Identity
But the textual evidence for such a move is not very strong,17 and one
might wonder what is to be gained by such a characterisation of iden-
tity. Surely (1 *) is not a real definition of identity since the notion of
sharing the same attributes involves the notion of identity.
(Numerical) sameness is contrasted with specific and generic
identity. We can say that two individuals a and b are specifically iden-
tical, namely Spec (a, b), if the species to which a belongs is the same
as the species to which b belongs, i.e.
(2) Spec (a, b) if S (a) = S (b)
It should be clear that' Sex), in (2) stands for 'the species of x'. Here
we can safely use, as Aristotle does, identity to define specific identity
because we can draw a clear-cut line between the two notions. To
speak in the Aristotelian way, if a and b are specifically (but not nu-
merically) identical they still count as two, which does not happen
with (numerical) identity. However, specific identity is based upon
(numerical) identity, because specific identity occurs only when the
species of a is (numerically) the same as the species of b. Moreover,
numerical identity implies specific identity, in the sense that
(3) a = b ~ Spec (a, b)
That is easy to see. Suppose that a = b. Then, the individual denoted

by both 'a' and 'b', is one and the same, its species being e. Thus,S
(a) is e, and the same holds for S(b)'. Therefore, numerical identity
implies the corresponding specific one. The converse of (3) does not
hold, because it may be that (2) is satisfied also in the case in which a
f:. b, as the example of Coriscus and Socrates shows, being different
individuals in the same species.
Generic identity does not add very much to specific identity, in the
sense that a similar pattern works in this case, with the only difference
that here not only individuals but also species can enjoy generic iden-
tity. Therefore, we can write
(4) Gen (I.I,~) if g(~) = g(~)

where' g(p)' stands for 'the genus of p' and 'Gen (p, 0)' for 'p and 0
are generically the same'. As before, generic identity is implied by
(numerical) identity and is based upon the latter. We can also add that
things which enjoy specific identity enjoy also generic identity, be-
cause individuals which fall under the same species fall also under the
same genus. 18
The problem that these definitions raise concerns the role that we
have to assign to them. Are they introducing distinctions in the notion
of identity in such a way that, for instance, we are allowed to equate
specific and generic sameness to something like relative identity en-
dorsed by Peter Geach and criticised by David Wiggins?19 Or are they
specifying a generic notion of identity in three different kinds of
sameness? I believe that we must strongly resist these views. The no-
tion of identity which is implied in the tripartition of the meanings of
'tauton' is always the same and it is simple identity. In the case of the
so-called 'numerical sameness', it is identity of what is denoted by two
names for objects, while in the case of specific and generic identity it
is identity concerning the species or the genus of two objects. I am in-
clined to think that Aristotle is not classifying kinds of sameness but
taking into account linguistic uses of 'same' in connection with propo-
sitions such as 'Coriscus and Dion are the same in species' or 'man
and horse are the same in genus'. Consider for instance the proposition
'Coriscus and Dion are the same in species'. Aristotle's analysis is
probably intended to warn anyone who is dealing with this proposition
that it must be taken as asserting not that Coriscus and Dion are the
same thing, but that they share the same species. No new notion of
identity is introduced, and Aristotle's distinctions are simply meant to
make it clear what identity refers to.
2.
In the same chapter from which we quoted text (A) there is a puzzling
passage concerning (numerical) identity, which is worth considering.
Aristotle says:
(C) What is one in number is most uncontroversially called the same in

everyone's judgement. But even this is customarily indicated in several
ways. The strictest and primary case is when that which is the same is in-
dicated by means of a word or a definition, e.g. coat for cloak or
two-footed terrestrial animal for man. The second case is when it is indi-
cated by means of a proprium, e.g. capable of knowledge for man or car-
ried upwards by nature for fire. The third case is when it is indicated with
an accident (apo tou sumbebekotos), e.g. the one sitting or the musical one
for Socrates. All these case are intended to signify what is one in number.
(Top. I 7, 103'23-31, Robin Smith's translation modified)
This text is meant not only to present cases in which (numerical) iden-
tity or oneness is expressed but also offer a sort of hierarchy of them.
Aiming to have a better understanding of what is behind this strange
passage, let us reverse the order of cases and start from the last, i.e. the
case of identity concerning individuals. If we put Aristotle's examples
in a proper sentential fonn we get propositions asserting a relation of
identity in which one and the same particular is picked out by a proper
name, 'Socrates', and by what looks like a definite description, 'the
one sitting' or 'the musical one'. If this view is correct we can fonnu-
late one of these examples in a semi-fonnal way by stating:
(5) Socrates = the one sitting
Aristotle says that in such a case the definite description is taken "from
an accident", surely because 'being sitting' is an accident of Socrates.
Since in (5) identity is expressed by means of an accident one can
confidently infer that (5) and all proposition of this kind are contingent
statements. By making a further step, one might be tempted to con-
clude that (5) and the other similar propositions express contingent
identity. We say that a = b expresses identity contingently if its truth
does not rule out the possibility of a f. b. The argument in favour of
this further conclusion might be as follows. We can generalise (5) by
introducing the 'I' operator used by Russell to represent definite de-
scriptions, so that we can express the fonn of a definite description by
'lxF(x)', which means 'the unique x which F_S'.20 In this way, the logi-
cal fonn of Aristotle's example is
(6) a = IxF(x)
Moreover, the accident which plays a role in (5) is 'being sitting' and
it is obviously an accident of Socrates. Therefore, the condition is
added that F(a) is an accidental predication. What is characteristic of
accidental predicates is that they need not hold of the subjects to which
they are attributed. Therefore, if F(a) is constituted by an accidental
predication we can state F(a) and O-F(a). Now, we can easily derive
F(a) from a = IxF(x), so that the following implication
(7) a = IxF(x) ~ F(a)
is the case. If we put ourselves in a first order predicate calculus with

identity and we add a modal basis as strong as T to ie t we can use the
necessitation rule to deduce
(8) D(a = IxF(x) ~ F(a»
from (7). By distributing the necessity operator across implication (8)

we get
(9) Da = IxF(x) ~ DF(a)
and by contraposition we obtain
(10) O-F(a) ~ Oa"* IxF(x)

Therefore, in every case in which a definite description is made out of
an accidental attribute we are allowed to conclude that a = IxF(x) is a
contingent proposition, and one might be led to infer that identity in-
volved is contingent as well. We can reconstruct the argument in a dif-
ferent way and present it as a proof ad absurdum. In any case, how-
ever, in order to get the conclusion, what we need is only the assump-
tion that the attribute involved by the description of the particular at
issue is an accidental predicate of that particular. It is sufficient to
postulate this condition and use few obvious logical principles to de-
rive that a = IxF(x) is a contingent statement. 22
If we put this result within the framework of Aristotle's ontology
and his theory of definition we might reinforce it by saying that in
every case in which identity refers to an individual and implies a defi-
nite description, we have to do with a contingent statement and, there-
fore, with a contingent statement of identity. To prove this assertion, it
is sufficient to show that every definite description of a particular en-
tails a reference to accidental properties of that particular, so that the
argument we used for (5) can apply to every case of identity. Of
course, to put this claim in a proper logical form, we should accom-

modate our language to include in a definite description not only a
simple accidental attribute, but also a complex condition that is
uniquely satisfied by an individual. Then, the claim would be that,
however complicated the condition might be, it holds contingently of
the individual to which it is supposed to apply. But for the sake of
simplicity, we can omit this complication and stick with the plain case
of a simple accident involved in a description.
The reason that leads one to ascribe to Aristotle the view that
every definite description of a particular is accidental in the sense de-
scribed is that he underlines in several texts that individuals cannot be
defined. 23 In his view, a definition is meant to answer the question ti
esti, 'what is it?', with reference to something by picking out a com-
plex of attributes which can apply separately to many things but alto-
gether apply only to the thing defined. 24 To count as a definition, the
complex of attributes must identify the definiendum in a permanent
and stable way, by separating it from all other things. If we are going
to define 'man', a definiendum, we have to look for a definiens which,
by applying only to 'man', divides it from all other animals and, in
general, from all sorts of things which are not a man. This cannot hap-
pen in the case of an individual identified by a definite description apo
tou sumbebekotos, because such a description is able to isolate the in-
dividual at a certain moment and in a certain context, but not perma-
nently and from all other individuals. Were a complex of attributes
such that it identified a particular in a stable way when applied to it,
we could conclude that there are definitions for individuals. Suppose
that there were such a permanent definite description of a and call it
, IxK(x),. By hypothesis, IxK(x) refers to a property (or a group of
properties) which is not accidental to a. Being non-accidental, such a
property belongs to a necessarily. Therefore, description IxK(x) (i) ap-
plies uniquely to a and (ii) applies necessarily to it. Thus, description
IxK(x) is able to discriminate a from any other individual different
from a under any circumstance. Insofar as IxK(x) is supposed to satisfy
these conditions, why should we not take it as the correct answer to the
question 'what is a'? Therefore, description IxK(x) counts as a defini-
tion of a. But there is no definition of a; consequently, there is no such
description.
It should be clear that Aristotle's refusal of the possibility of de-

fining particulars does not imply a denial of the view that, given two
particulars a and b, it is always possible to find a definite description
of one of them which is able to discriminate it from the other. His po-
sition is consistent with the view that if a =f:. b then there is at least one
attribute of a which does not belong to b or an attribute of b which
does not belong to a. This means simply that we are able to discrimi-
nate by description a from b but it does not entail that we can get a de-
finitory description of a, since a definitory description should isolate a
not only from b, but also from whatever else is not a in a permanent
and stable way.
3.
Shall we conclude from this discussion that propositions expressing

identity concerning individuals, where definite descriptions are in-
volved, are all contingent and Aristotle is a defender of contingent
identity? I think that we can answer this question only if we reformu-
late it in a clearer way.
To achieve this goal, let us first make a further move along the
lines of our previous analysis. We might ask whether the claim about
contingency could be extended from statements of identity involving
definite descriptions of individuals to any identity statement concern-
ing individuals, in the sense that also simple propositions such as a=b
must be treated as contingent. Taken in this extreme way, the question
must be answered negatively. For instance, a = a cannot be conceived
of as a sentence expressing contingent identity. It is impossible for an
individual to be different from itself: a =f:. a implies a contradiction.
On the other hand, suppose that one and the same individual is
picked out by two different names. Can we claim that, e.g., 'Tullius =
Cicero' is a statement of contingent identity? Well, if we stick to the
previous argument, the answer depends on the way we interpret a
proper name and its relation to the particular which it refers to. If we
take, for instance, 'Tullius' and 'Cicero' as definite descriptions such
as 'the one called 'Tullius", and 'the one called 'Cicero", then in
'Tullius = Cicero' identity involving two definite descriptions is stated

and the proposition must be taken as contingent for the same reason
we have concluded that (5) is contingent. On the other hand, if we
consider 'Tullius' and 'Cicero', insofar as they are proper names, as
simple ways to address or to point to one and the same individual, so
that 'Tullius' and 'Cicero' do not function as predicates but denote di-
rectly by simply indicating their denotatum, we should conclude that
'Tullius = Cicero' expresses non-contingent, i.e. necessary, identity no
more and no less than a = a.
I am inclined to think that it is in this second way that we must
interpret our 'Tullius = Cicero', and admit that proper names, the real
ones, behave differently from definite descriptions as far as their refer-
ence is concerned. In other words, while a proper name, qua proper
name, points to an individual, being a sort of index finger, and its main
function consists in this act, a definite description implies the use of an
attribute which is supposed to be predicated uniquely and contingently
(in the Aristotelian perspective) of the individual which the description
applies to. In this sense a definite description, properly speaking, has
no denoting power and it can identify an object only because it applies
uniquely to it. 25
There is maybe some evidence for this claim in Aristotle's way of
describing another type of numerical identity in our text (C). As we
have seen, he maintains that a case of numerical identity is given
when it is stated by means of two synonymous words. Take for in-
stance 'cloak' and 'coat'. If we say 'cloak = coat' then we have an
identity statement comparable to the statement we get when we relate
a dejiniens, call it dC(, to a dejiniendum a and we obtain dcx = a. If the
type of identity which is appropriate for the proposition 'cloak = coat'
is of the same sort as identity which occurs in the attribution of a de-
jiniens to a dejiniendum, it is difficult to believe that 'cloak = coat' is a
contingent statement of identity, since a definition, in the Aristotelian
perspective, cannot be classified as contingent. Therefore, 'cloak =
coat' is a necessary statement of identity. Now, not only in the case of
the attribution of a dejiniens to a definiendum but also in a proposition
such as 'cloak = coat' a relation is established between entities which
are supposed to be universal or general, whatever this may in fact
mean. If we were thinking in modern semantic terms, we could maybe
say that here 'cloak' stands for the class of the individuals which are
cloaks, or one might imagine that 'cloak' and 'coat' are names of con-
cepts. Whatever the ontology of these things may be, it is hard to reject
the idea that 'cloak' and 'coat' refer to special entities, being names, or
rather, proper names for them.
Suppose now that these names are not directly referring, but must
be expanded into descriptions, 'the object called 'cloak", or 'the ob-
ject called 'coat", in the same way we imagined that 'Tullius' and
'Cicero', taken as descriptions, had to be expanded into 'the one called
'Tullius", 'the one called 'Cicero". It would be plausible to maintain
that in 'the object called 'cloak' (or 'coat')' the property of being
called 'cloak' ('coat') is accidental with respect to the object referred
to in the description. After all, according to Aristotle, names are as-
signed by convention and not by nature. 26 Therefore, if 'Tullius ==
Cicero' is a contingent statement of identity under the assumption that
'Tullius' and 'Cicero' stand for definite descriptions, since these de-
scriptions involve an accidental property of the thing designated, the
same should be true for 'cloak = coat' because also in this case de-
scriptions referring to accidental properties are used. But we have seen
that 'cloak == coat' cannot be taken as a contingent statement of iden-
tity since it is put on the same footing as definitional identity.
To avoid this conclusion, we must give up either the assumption
that (i) 'cloak' and 'coat' are proper names, or that (ii) "being called
'cloak' ('coat')" is an accidental attribute for the object referred to, or
finally that (iii) 'cloak' and 'coat' function as definite descriptions. It
is hard to dismiss the first assumption, because 'cloak' and 'coat' are
called by Aristotle onomata, "names", in text (A) and they are said to
be names for one and the same pragma, "object". Therefore, they do
not play the role of kategoroumena, "predicates", of a subject, their
main function being to point to something.
One might challenge the second assumption along the following
lines. It is true that in some sense "being called 'cloak' ('coat')" is an
accidental attribute of the object so described, since one might imagine
a different name or group of names for this object. But still a differ-
ence remains with the case of 'Tullius' and 'Cicero'. Synonymous and
homonymous terms are given in a language. Therefore, we must pre-
liminarily fix a language in order to decide about identity involving
synonymous words. In this language we could easily think of Cicero as

an individual with a different name. Call Cicero 'Petrus' and we are
not compelled to abandon the language in which it makes sense to call
Cicero 'Tullius'. But we cannot assign a different name to the object
called 'cloak' or 'coat' without putting ourselves outside the language
we have chosen as our reference point. This observation may be true
(even if there are modifications in a language which are internal to it)
but I do not think that it is sufficient to make the point its possible
supporters intend to. It would be strange to maintain that 'being called
'cloak' (or 'coat')' denotes a necessary property of the object to which
it attaches even if we transform the description into "being called
'cloak' ('coat') in (e.g.) English". Is it really necessary that the object
named 'cloak' or 'coat' be so designated in English? I do not even see
that it is necessary that it be named in some way in English. It might
have been that in English there was no word for cloak.
If we want to be faithful to Aristotle, we must perforce give up our
third assumption, namely the idea that the names 'cloak' and 'coat'
function as definite descriptions. But if so, we do not see why 'Tullius'
and 'Cicero' should be conceived of as definite descriptions. Thus, my
conclusion is that every proper name, of general objects or individuals,
has a direct referential power, as we have explained.
The consequence of this analysis seems unavoidable. Not only
propositions such as "cloak = coat", i.e. statements involving proper
names for general entities, but also propositions such as "Tullius =
Cicero", where proper names for individuals are used, should be taken
as necessary. If 'cloak' refers to a certain object 11 and the same holds
for 'coat', then 'cloak = coat' simply means that the object denoted by
'cloak', namely 11, is the same as the object indicated by 'coat', i.e. that
11 is the same as 11, a necessary statement. Exactly the same can be said
for 'Tullius = Cicero'. Take a as the object denoted by 'Tullius' and
'Cicero'. Therefore what 'Tullius = Cicero' states is that a is a, again a
necessary proposition.
An objection could be raised to this interpretation since it is easy to
observe that it presupposes that a proposition such as a = a (or 11 = 11)
is a necessary proposition. But it is a widespread view that Aristotle
did not formulate a Principle of Identity in his philosophy;27 and, a
fortiori, he did not represent it as a necessary statement. This view is
to a certain extent true, since we cannot find for the Principle of Iden-
tity assertions as explicit as those concerning the related Principle of
Non-Contradiction, which is said to be one of the most reliable axioms
of philosophf8 and a necessary truth. 29 However, first of all, there is
evidence to maintain that for Aristotle a = a (or 1-1 = 1-1) is a sound
statement. In Metaph. !!. 9, 10 lSa S-9 that a thing is the same as itself is
quoted as an obvious case of numerical identity and in I 3, 1054a35
"you are the same as you" is offered as an evident example of numeri-
cal oneness. Moreover, in a passage of the Sophistical Refutations the
Principle of Identity is used to build up a sophistical argument con-
cluding that one and the same thing is the same as itself and different
from itself,30 and the explanation of the paradox does not involve any
restriction on the Principle itself. Finally, the idea that a proposition
such as 1-1 = 1-1 constitutes a philosophical principle is suggested by a
passage in the Metaphysics where Aristotle says:
CD) To inquire why a thing is itself is not a real inquiry, since for it the
fact or that the thing is must already be evident (e.g. that the moon is
eclipsed), but the fact that a thing is itself is the only reason and the only
cause to be given for all such questions as why man is man or musical is
musical; unless one were to say that each thing is indistinguishable from
itself and its being one just meant this; but this is common to all things
and is a short and easy way with the question. (Metaph. Z 17, 1041 "14-20
Oxford revised translation modified)
There is a standard and, I believe, correct interpretation of this texe t A

scientific inquiry trying to establish the reason in accordance with
which man is man or musical is musical is pointless. In general, when
we ask why a predicate belongs to a subject the fact that the predicate
belongs to that subject must be already known. But in the case of 1-1 =
1-1, stating 1-1 = 1-1 is self-explanatory and we do not need to look for a
reason for it. A possible justification of 1-1 = 1-1 does not depend on the
nature of 1-1 but, at most, is based on the fact that 1-1 cannot be distin-
guished from itself and on the assumption that indistinguishability is
what characterises identity. It is therefore clear that for Aristotle the
fact that one thing is the same as itself is something which applies to
everything for reasons which are not tied to the particular nature of one
thing or another. We are entitled to infer from this that 1-1 = 1-1 is a gen-
eral principle of philosophy and, as such, a necessary statement. Our

interpretation is confirmed.
4.
We are in big trouble. On the one hand, we have claimed that identity
can be contingent in the sense that a = IxF(x) is contingent and, on the
other, we have concluded that a = b is a statement of necessary iden-
tity. A contradiction arises. Since a = b is a necessary statement we are
allowed to write
(l1)a=b~Da=b
Substitute'lxF(x), for 'b' and we get
(12) a = IxF(x) ~ Da = IxF(x)
which is the negation of the claim that statements of identity involving

definite descriptions express contingent sameness.
How can we get out of this difficulty? I think that we cannot argue
against the claim that a = b is a necessary identity, since it is based on
two very simple assumptions, namely that' a' and 'b' are not disguised
descriptions and that a = a is a necessary statement. Fortunately, mod-
em logic can help us to revise the argument that led us to conclude that
a = IxF(x) expresses contingent identity. As logicians have repeatedly
pointed out, the use of the 'I' descriptor is inadequate when we have to
deal with modalities, since it does not allow us to show the exact scope
of the modal operators. 32 Let us use the normal way to get rid of defi-
nite descriptions introduced by means of the' I' operator. As every-
body knows after Russell, a sentence such as (6) can be rephrased in
the following way
(13) F(a) /\ \/x (F(x) ~ x = a)
Assume now that the property on which a description of an individual

is based belongs accidentally to the individual in question, i.e. take
<>-'F(a). Starting from the obvious thesis that
(14) (F(a) /\ \/x (F(x) ~ x = a» ~ F(a)

by the same argument we produced before we can easily derive
(15) 0-, (F(a) /\ Vx (F(x) ~ x = a»
Statement (15) is true and it is actually implied by (5) if we stick to the

view that (5) is a contingent proposition.
But admitting (15) does not entail that we must reject the view that
identity is conceived by Aristotle as necessary. When we claim that
(15) holds, we are giving up the idea that (5) should be considered as a
necessary proposition, since (15) rules out
(16) O(F(a) /\ Vx (F(x) ~ x = a»
But why should one take an identity statement like (5) as implying a
necessary proposition in the sense of (16)1 When we say that Socrates
is the one sitting, we are not thereby committed to the claim that nec-
essarily Socrates is the only sitting object. It might have been that Soc-
rates is not sitting or that he is not the only one to sit. We can more
plausibly take (5) to mean that one and the same person, Socrates, is
described as the only sitter and addressed as 'Socrates'. It may be that
the description is based on an accident of Socrates but Socrates, how-
ever described, is always the same Socrates as the one called 'Socra-
tes', and this is a necessary fact. In this perspective if we would like to
make explicit the kind of necessity involved by the use of identity in a
statement such as (5) when (5) is interpreted as (13) and identity is
supposed to satisfy (11), we should state something like
(17) F(a) /\ Vx (F(x) ~ Ox = a)
which corresponds exactly to our intuition about (5): the individual

called 'Socrates' is necessarily the same individual contingently de-
scribed as the one sitting. If we adopt the view that identity is neces-
sary (17) is an immediate consequence of (13). But (17) is by no way
inconsistent with (15). Therefore, (5) can be considered as a contingent
statement, which is what Aristotle does, but this claim does not com-
mit us to attribute him the view that identity has to be taken as contin-
gent. On the contrary, there is evidence for the opposite. If our inter-
pretation of proper names is correct, we should conclude that he fa-
voured a system where identity is necessary in the sense described by
(11).33
5,
Let us come back to text (C). Various cases of identity are mentioned
and a hierarchy among them is established. We have already consid-
ered two of these cases. One occurs when identity is expressed by at-
tributing a definite description to an individual; the other is the case of
two synonyms for the same general entity. It seems legitimate to
equate to the latter the case of two synonyms for the same individual,
i.e. the case of identity expressed by propositions such as 'Tullius =
Cicero'. After all, in a passage parallel to text (C) Aristotle mentions
the proposition 'you are you' as an example of the strongest case of
identity, and this fact shows that he does not distinguish between the
situation of general entities and individuals. 34
We must briefly explore now the two remaining cases, i.e. the case
of the attribution of a proprium to its subject and the case of definition.
Here, once again, Aristotle is applying identity to entities denoted by
general terms, taking the latter as proper names for the former. In this
perspective, we must consider the dejiniendum which appears as a
member of definitional identity as expressing the name of a class or
concept, no more and no less than 'Tullius' is the name for an individ-
ual and a way to indicate it. If we take this point of view seriously, we
can think of a dejiniens as a special kind of definite description. As we
have seen, 'the one sitting' is a definite description of Socrates apo tou
sumbebekotos. In the same way, we can consider 'two-footed terres-
trial animal' as a definite description of what is designated by the
name 'man'. The only difference with respect to Socrates and 'the one
sitting' is that 'two-footed terrestrial animal' is not apo tou sumbebe-
kotos with respect to man, being not based on an accidental attribute of
man, but it expresses what answers the question 'ti esti anthropos;',
'what is a man?', i.e. it picks out the essence of man.
We can interpret the case of proprium along the same lines of ap-
proach. Basically, a proprium is an attribute which applies to a subject
and no other. In this sense, 'capable of knowledge' is a proprium of
man because only men are capable of knowledge. Consider for in-
stance the statement 'man = capable of knowledge' and, as before,
take 'man' as a name of the entity designated by 'man'. Then 'capable
of knowledge' can be considered as a definite description of this en-
tity. A definite description construed from a proprium differs, how-

ever, from a definite description based on a definition, because a pro-
prium, according to Aristotle, does not reveal the essence of the object
to which it is applied. On the other hand, a definite description based
on a proprium differs from a definite description apo tou sumbebeko-
tos, because a proprium is not an accidental attribute of the thing to
which it applies.
It should be clear that Aristotle has a different approach to propria
when he gives the official definition of this notion. 3s 'Man is capable
of knowledge', which is viewed as an identity statement in text (C), is
considered here as a universal equivalence, being treated as equivalent
to 'everything is human if and only if it is capable of knowledge',
whose formal structure is
(20) \Ix (F(x) B G(x»
What is interesting is that for Aristotle the two different ways of con-
sidering propria are equivalent and interchangeable, even if the onto-
logical commitment of the two approaches is quite different. But I
would not like to be trapped by this difficult question.
It is more appropriate to point out that we are now in a position to
explain what the cases of identity considered by Aristotle are. It should
be clear that it is not a difference in identity: a and b cannot be more
(or less) identical than c and d are. If they are identical they are simply
the same. Nor can it be a question of "strong" versus "weak" identity,
i.e. necessary and contingent identity. As we have seen, identity in
Aristotle's view is always "strong" in the sense that it is necessary. We
must take not only 'man = two-footed terrestrial animal' as a necessary
proposition but also 'Socrates = the one sitting' as involving necessity
in the sense we have explained.
I think that the cases described by Aristotle in text (C) refer to dif-
ferent ways in which identity can be expressed. We can express iden-
tity by picking two names for the same object, or we can use a definite
description. As we have seen, definitions, propria and definite de-
scriptions for individuals all fall in the latter case, although they mani-
fest differences. In the case of a definition a description is offered
which reveals the essence of the definiendum. In the case of a pro-
prium we have a description based on a necessary but not essential at-
tribute, and in the case of individuals the description is construed by

means of an accident, which is neither essential nor necessary to the
particular referred to.
We can now face our last question, namely the question of the or-
der in which Aristotle puts his various types of identity statement. The
order constitutes a hierarchy of them. At the top are those identity
statements which are expressed by definitions and synonyms; we find
at the bottom definite descriptions for individuals and, at the interme-
diate level, there are identity statements expressed by means of pro-
pria. Why this order? My hypothesis is that Aristotle was driven by
the idea that any identity statement exhibits a slightly embarrassing
feature, because it expresses sameness by means of different terms. 36
When we state, for instance, 'Tullius = Tullius' we use two different
occurrences of the same name for one object while in 'Tullius =
Cicero' two different names are employed to designate one and the
same particular. Thus, one could think that the various types of iden-
tity statements are ordered according to the grade of difference of the
terms by which the identity is expressed. Aristotle does not mention
here the strongest case, i.e. a = a, or more generally 1.1 = 1.1. He does so
elsewhere. 37 In text (C), his starting point is the case of synonyms and
definitions, where there is a difference between the terms through
which identity is formulated, but the difference is less than the one in-
volved by propria or accidents. Synonyms do not present any differ-
ence in content and a definiens manifests what its definiendum is. A
greater distance occurs between a proprium and its subject because
they have different definitions. The last position is occupied by acci-
dents, which attach in a loose way to their subjects, being not even
necessary to them.
Aristotle's theory of identity, at least as it appears in the Topics, is
quite different from the modem approach especially from the point of
view of the ontology implied. The ancient philosopher looks less par-
simonious than contemporary authors do about allowing entities in the
universe. This probably depends on the fact that he had a different no-
tion of and different criteria for existence. To explore this aspect of his
thought would be, I think, rather fascinating but it cannot be pursued-
here.
Notes
I Top. I 5, 102"7-9.
2 Top. I 5, 1028 36 ff.
3 See e.g. Maier [1969-1970], II b, p.78; Hambruch, [1904], p.3 ff.; Solmsen
[1929], p.194; Ross [1939], pp.251-272; Bochenski [1968], pp.22-24; During [1966],
p.69 ff.; Barnes [1981], pp.17-59.
4 I use 'sameness' and 'identity' as synonyms. Therefore, by attributing a theory
of sameness to Aristotle we intend to ascribe to him a theory of identity, pace White
[1971], pp.I77-197. See also Miller [1973], pp.483-490.
5 E.g. Metaph. 119; 13, 10548 32 ff.
6 Texts in Bonitz [1955 2], 125"57 ff.
7 I am grateful to Professor Enrico Martino for having made me aware of the
importance of this point for the interpretation of Aristotle's position.

8 Metaph. I 3, 1054"32-5.
9 See e.g. Metaph. I 3, 10548 34-b3 where the unity, i.e. identity, of a definition is
in question.
10 SE 1,165"10-13.
II I am once more grateful to Paolo Crivelli for his useful comments, which
have greatly helped me to correct my previous view and better understand Aristotle's
position.
12 Alex., In Metaph. 615.20 ff.
13 The same view in Ross [1953], II, p.287).
14 See e.g. Metaph. !l. 9, 1018"4-9.
15 Metaph. !l. 9, 1018 8 7-9.
16 It should be kept in mind that according to Aristotle relations are treated as

properties, so that in his language there are no n-place predicates where Ii> 1.
17 Aristotle seems to consider unity in terms of indivisibility, e.g., at Metaph. !l.
6, 1016"32 ff.; Z 17, 1041 "14-20; I 1, 1052"29 ff. But it is not clear to me whether we
must confer a deep and philosophically important sense to this way of speaking.
18 Aristotle is well aware of the fact that (numerical) identity implies the corre-
sponding specific and generic one, while specific identity implies generic identity
but not vice versa (see Metaph. !l. 6, 1016b35 ff.).
19 P.T. Geach [1972], p.239 ff; Wiggins [1980], p.15 ff.
20 Russell and Whitehead [19272 ], I , p.173 ff.
21 For the system T see Hughes and Cresswell [1996], p.23 ff.
22 It should be pointed out that in the derivation of (7) we need to use some ver-
sion of the law of substitutivity, which can be formalised as follows:
(LS) a = b ~ (.A ~ IJ) (8differing from .Aonly in having b in 0 or more oc-
currences of a in -A).
(LS) is notoriously at odds with systems of contingent identity. However, also these
systems normally adopt a non-modal version of it, in the sense that (LS) is supposed
to hold if -Aand IJdo not contain modal operators. We have respected this condition
in using (LS) since (6) contains no modal operator (see Hughes and Cresswell [1996]
pp.332-334).
23 E.g. Metaph. Z 15, 1039b 27 ff.
24 A.Po. II 13, 96"24 ff.
25 The fact that proper names are supposed to have a primary referential func-
tion does not necessarily imply that they do not possess a sense. On this question see
D. Wiggins [1976], pp. 221-255
26 Int. 2, 16"19-20
27 See e.g. Berti [1987], pp.212-213.
28 See e.g. Metaph. r 3, 1005 b 5 ff.
29 E.g. Metaph. r 3,1005 b I9-22.
30 SE 30, 181 b 13 ff.
31 See e.g. Ross [1953], II, pp. 206-207, Reale [1993], III, pp.396-397, Frede
and Patzig [1988], II, pp.310-312.
32 Russell was aware of the ambiguity involved by the use of the ',' operator
and to avoid it he introduced the notion of scope of a description (Principia Mathe-
matica, I, p. 173). The ambiguity of ',' increases in a modal context (see Hughes and
Cresswell [1996], pp.324-325).
33 I am aware of the fact that using formulas with arbitrary names in a modal
context can raise difficulties (on them see e.g. Hughes and Cresswell [1996], p.274
ff.). However, I do not think that I need take care of them at the level of analysis on
which I am moving.
34 Metaph. I 3, 1054"34-35.
35 For the definition of proprium see Top. I 5, 102"18-24.
36 This idea is implied by Metaph. 11 9, 1018"7-9. See also L. Wittgenstein
[1922], 5.5303: "Beiliiufig gesprochen: von zwei Dingen zu sagen, sie seien
identisch, ist ein Unsinn, und von Einem zu sagen, es sei identisch mit sich selbst,
sagt gar nichts. "
37 Metaph. 119, 1018"7-9; Metaph. I 3, 1054"33-35.
SAMENESS IN ARISTOTLE'S TOPICS
Paolo Crivelli
Aristotle's discussion of sameness in the Topics. In the Topics Aris-

totle discusses sameness to an extent and with a care that strike us as
rather unusual: elsewhere he tends to concentrate on unity rather than
sameness, and his remarks about sameness are introduced almost as
asides in his reflections on unity. As several commentators have no-
ticed, there is probably a twofold reason why in the Topics Aristotle is
so interested in sameness: on the one hand, he thinks that sameness is
intimately linked to definition; on the other hand, definition is one of
his major concerns in the Topics.!
In the Topics Aristotle discusses two issues concerning sameness:
first, he distinguishes and defines the main kinds of sameness or uses
of the expression 'the same,2 (this he does in chapter I 7); secondly, he
presents some 'commonplace rules' which involve sameness (and, in
fact, depend on some form of Leibniz's Law) (this he does in chapters
VII 1-2 and in lines 133 a24-134a4 of chapter V 4). I am going to con-
centrate on what Aristotle says in chapter 17, in particular on his char-
acterization of numerical sameness.
The structure of chapter I 7. Topics I 7 divides into three parts:
The first part (103 a6-14) distinguishes and defines three kinds of
sameness: numerical sameness, sameness in species, and sameness in
genus. 3
239
The second part (103 a 14-23) discusses a puzzle concerning the

second kind of sameness, i.e., sameness in species: should portions of
water coming from the same spring be regarded as the same in spe-
cies?
The third part (103 a23-39) distinguishes various cases in which
the first kind of sameness, i.e., numerical sameness, may hold. These
various cases depend on the types of linguistic expression which can
be used to refer to the items between which numerical sameness holds.
A merely apparent circularity. Here are the main points of Aristotle's
distinction and definition of three kinds of sameness in the first part of
the chapter:
(Tl) Those things are numerically the same of which the names are many
but the thing is one, as cloak and coat. [ ... J Those things are said to be the
same in species which fall under the same species. Similarly, those are the
same in genus which fall under the same genus. (Top. 17. 103'9-14)
This definition induces one immediately to raise the objection of cir-

cularity: the dejinientia of the definitions of sameness in species and
sameness in genus use the expression 'the same'. There is an obvious
answer to this objection: the expression 'the same', as used in the de-
jinientia in question, has a different sense from that in which it is being
defined. In particular, it is natural to assume that 'the same', as used in
the dejinientia of sameness in species and sameness in genus, ex-
presses (neither sameness in species nor sameness in genus, but) nu-
merical sameness (which has already been defined). If this is correct, it
has two noteworthy consequences.
The first consequence is that numerical sameness holds (not only
of individuals, but also) of species and genera (universals):4 for it is to
species and genera that 'the same' is applied in the dejinientia of the
definitions of sameness in species and sameness in genus. This conse-
quence does not constitute a problem: Aristotle's example illustrating
numerical sameness is most naturally interpreted as involving univer-
sals (the universals cloak and coat).
The second consequence is that numerical sameness turns out to
be prior ('prior in definition') to the other two kinds of sameness. s It
follows that numerical sameness is more fundamental than the other
Paolo Crivelli, Sameness in Aristotle's Topics 241
two kinds of sameness,6 and we have some justification for concen-

trating on numerical sameness and leaving on one side sameness in
species and sameness in genus.
The linguistic character of Aristotle's definition of numerical same-
ness: a puzzle. Aristotle's definition of numerical sameness has a 'lin-
guistic character' in that it speaks of the names of things: 'Those things
are numerically the same of which the names are many but the thing is
one'. Despite this linguistic character, it is important to be clear about
what this definition is not: it is not a formulation of the truth condi-
tions of a statement of identity. Aristotle is not giving the truth condi-
tions of a statement of identity' a is b' by saying that' a is b' is true iff
the thing named by 'a' and the thing named by 'b' are one. If this were
what Aristotle is doing, it would be easy to understand why Aristotle
speaks of names: the names in question would be those which occur at
the two ends of the statement of identity whose truth conditions are
being spelled out. However, Aristotle is not giving the truth conditions
of a statement of identity: in fact, his enterprise is to explain what it is
for something to be numerically the same as something.
Multiplicity of names cannot be an essential trait of numerical same-
ness. Given that this is Aristotle's enterprise, the linguistic character of
his definition of numerical sameness is more puzzling than ever. In-
deed, on one possible interpretation, the linguistic character of Aris-
totle's definition has paradoxical consequences. For, on one possible
interpretation, in his definition of numerical sameness Aristotle is
claiming that the multiplicity of names is an essential trait of numeri-
cal sameness. Aristotle's definition of numerical sameness would thus
amount to something like the following:
(1) a is numerically the same as b iff a and b are one, and some
name of a is distinct from some name of b.
(1) entails that a thing is numerically the same as itself only if it has at
least two names. A consequence of this is that nameless objects (in-
deed, objects with at most one name) should not be numerically the
same as themselves (e.g., grains of sand at the bottom of the sea would
not be numerically the same as themselves). This consequence is so
paradoxical that it induces one to abandon the suggestion that the mul-
tiplicity of names is an essential trait of numerical sameness. It might

be replied that every object has a name: that every grain of sand at the
bottom of the sea has a name in that there is a definite description that
picks it out. But I doubt that this reply is available to Aristotle: his
characterization of a name in chapter 2 of de Interpretatione (16 a 19-
26) requires that no name should be a phrase, so that most definite de-
scriptions (in particular, those rather complex definite descriptions
which would be needed to pick out grains of sand at the bottom of the
sea) would not count as names for Aristotle.
Multiplicity of names must be a warning. Since the linguistic character
of Aristotle's definition of numerical sameness cannot be interpreted
by assuming that the multiplicity of names is an essential trait of nu-
merical sameness, an alternative interpretation of this linguistic char-
acter must be found. The most plausible alternative interpretation is
one that somewhat plays down the importance of the linguistic char-
acter of Aristotle's definition: the reason why Aristotle says that things
which are numerically the same have distinct names is not that this is
an essential trait of numerical sameness, but merely to warn that nu-
merical sameness is compatible with multiplicity of names. More pre-
cisely, Aristotle is warning that an object a can be numerically the
same as an object b even if one of the expressions which can be used
to refer to a is distinct from one of the expressions which can be used
to refer to b.
But why should Aristotle be giving such a warning? At least two
answers to this question come to mind.
(i) Sameness, unity, and duality. The first answer is very speculative.
One of the most remarkable aspects of the concept of numerical same-
ness, at a very rough and intuitive level, is the fact that it brings in both
duality and unity: on the one hand, numerical sameness is a binary re-
lation, and as such it has two 'empty slots to fill'; on the other hand,
numerical sameness can by its very nature concern only one object
(because numerical sameness is that relation that everything bears to
exactly one thing, i.e., itself). This combination of duality and unity
might appear somewhat awkward or even paradoxical, and Aristotle
was probably sensitive to this issue, as is shown by the following pas-
sage:
(T2) Sameness is a kind of unity either of the being of several things or of

a thing when it is treated as several, as when one says that a thing is the
same as itself (for one treats it as two). (Metaph. t:. 9. 1018"7-9)
I cannot address here the interpretation of this passage in the context of

Metaphysics ~ 9: let me just say that when he says that 'when one says
that a thing is the same as itself [ ... ] one treats it as two', Aristotle is
probably alerting the reader to the awkward feature of numerical
sameness that I pointed out above, i.e., that numerical sameness brings
in both duality and unity. 7
Now, it is well known that Aristotle does not have a clear concept
of a binary relation. Maybe the linguistic character of Aristotle's defi-
nition of numerical sameness in Topics 17, i.e., the emphasis he puts
on the multiplicity of names, is an inchoate attempt at expressing the
idea that numerical sameness is a binary relation, so that at least part of
what Aristotle is doing is alerting the reader to the awkward feature of
numerical sameness that I pointed out above, i.e., its bringing in both
duality and unity. (Note that even now, when a logic teacher tries to
introduce students to the concept of a binary relation, the most natural
starting-point is linguistic: consider a complete sentence like 'Paris
loves Helen', subtract the two names, and draw attention to what is
expressed by the remainder of this linguistic subtraction.)
(ii) Sameness and definition. The second answer to the question of
why Aristotle is warning the reader that an object a can be numerically
the same as an object b even if one of the expressions which can be
used to refer to a is distinct from one of the expressions which can be
used to refer to b, has to do with the reason why the Topics contain a
comparatively extensive discussion of sameness: the relevance of
sameness to definitions.
The relevance of sameness to definitions is emphasized by Aris-
totle already in chapter I 5:
(T3) One may call 'definitory' the question 'Are perception and knowl-
edge the same or different?,8 - for arguments about definitions are mostly
concerned with questions of sameness and difference. [ ... ] For if we are
able to argue that two things are the same or are different, then by the
same turn of argument we shall be well supplied with lines of attack upon
their definitions; for when we have shown that they are not the same we
shall have demolished the definition. But the converse of this last state-
ment does not hold: for to show that they are the same is not enough to
establish a definition. To show, however, that they are not the same is
enough of itself to overthrow it. (Top. I 5. 102"6--17)9
Given that definitions are sentences (linguistic expressions) in which

'is' links two distinct noun phrases (the definiendum-expression and
the definiens-expression) which signify the same thing, it is natural
that in his discussion of numerical sameness Aristotle should empha-
size that numerical sameness can link items which are referred to by
means of distinct linguistic expressions.
That the connection of numerical sameness with definitions should
be one of Aristotle's reasons for the linguistic character of his defini-
tion of numerical sameness in I 7 is confirmed by the third part of I 7
(103"23-39), a passage which seems to be closely connected to T3. Let
me spell this connection out in some detail. In T3 Aristotle is pointing
out that the commonplace rules involving sameness can be used to
demolish, but not to establish, a suggested definition. What he has in
mind is roughly the following. Suppose that in a dialectical debate one
is successful at establishing or 'proving' the sentence 'a is b'. Can this
amount to establishing or proving that' a is b' is the definition of a? In
T3 Aristotle answers 'No'. Suppose, alternatively, that in the dialecti-
cal debate one is successful at demolishing or 'disproving' the sen-
tence 'a is b'. Can this amount to establishing or proving that 'a is b'
is not the definition of a? In T3 Aristotle answers 'Yes'.
Now, T3's second claim is uncontroversial: clearly, if 'a is b' is
demolished or disproved, then 'a is b' is not the definition of a. How-
ever, one might have some doubts about T 3's first claim: one might
wonder why establishing or proving 'a is b' does not suffice to estab-
lish or prove that' a is b' is the definition of a. The third part of Topics
I 7 goes some way towards answering this doubt. In this passage, Ar-
istotle distinguishes four cases of numerical sameness. The distinction
is based on the types of linguistic expression which might be used to
refer to the items between which numerical sameness holds: if a is
numerically the same as b, the expression used to refer to b can be ei-
ther (i) a name of it, or (ii) the definiens-expression of its definition, or
(iii) an expression of a characteristic proper to it, or finally (iv) an ex-

pression of a characteristic incidental to it. In other words, there are
four kinds of case in which the sentence 'a is b' can be true: either
(i) 'b' is a name of a, or (ii) 'b' is the definiens-expression of a's defi-
nition, or (iii) 'b' is an expression of a characteristic proper to a, or
finally (iv) 'b' is an expression of a characteristic incidental to a. But
then, clearly, establishing or proving' a is b' does not suffice to estab-
lish or prove that' a is b' is the definition of a: e.g., the true 'a is b'
might simply be expressing 'property-sameness' (i.e., it might fall un-
der case (iii)). Thus, the third part of Topics 17 provides some theo-
retical underpinning for the commonplace rules involving numerical
sameness which are relevant to definitions. This confirms that the link
with definitions is at least one of the reasons for the linguistic character
of Aristotle's definition of numerical sameness.
Notes
J Cf. Alex. Aphr. in Top. 57, 23-4; 60, 11-2; Pacius [1597], p.355; Rolfes
[1919], p.205; Brunschwig (ed.) [1967], p.124; Smith (ed.) [1997], p.68.
2 I shall freely (and somewhat sloppily) use "kinds of sameness" as equivalent to
"uses of the expression 'the same'''.
3 Cf. VII 1. 152b 30-3; Ph. VII 1. 242a32-~; de An. 15. 411 b20-1; HA II 1.
497b9-13. The classifications of the uses of 'the same' offered by Aristotle in the
Metaphysics (89. 1017b27-1018 a9; 13. 1054 a32-b3) are different from the one we
find in Topics I 7 (the main difference is that they involve the concept of matter,
which at the time of the Topics was not yet a philosophical tool of Aristotle's).
4 Cf. Alex. Aphr. in Top. 60, 17-61,5; Pacius [1597], p.355).
5 For 'priority in definition', see Metaph. Z 1. 1028 a31-6; M 2. 1077b2-4 (cf. Ph.
VIII 9. 265"22-4; Metaph. 8 11. 1018b30-6; Z 13. 1038b27-8; e 8. 1049b I2-4).
6 Aristotle explicitly recognises that numerical sameness is the most fundamental
kind of sameness: see 17.103"23-4; VII 1. 151 b28-30.

7 The connection between lines 1018 a7-9 of Metaphysics 8 9 and lines 103"7-9
of Topics I 7 is already claimed to hold by Miller [1973], p.484. For a different inter-
pretation ofthe passage from Metaphysics 8 9, see White [1971], pp.187-8.
8 We have here, of course, an implicit reference to Plato's Theaetetus, where
the second attempt at a definition of knowledge is the thesis that 'knowledge is per-
ception' (see 151 e2-3; e6; 152<5-6).
9 Cf. VII 2. 152b 36-153 a5.
IDENTITY AND SUPERVENIENCE
David Wiggins
1.
Identity is an equivalence relation, a congruence relation, and a fully

determinate relation. Moreover, where identity is concerned, there can
be no almosts and no near misses. Identity seems unique. I But that is
not enough to show that identity and its peculiarities cannot be ac-
counted for somehow in other terms. It does not show there can be no
reduction of sameness to other properties and relations.
Philosophers rarely avow the explicit belief that identity must be
reducible to other properties and relations. But some such belief must
surely have been at work when, a long time ago, Jaakko Hintikka was
a
moved to making ruling in which indefinitely many philosophers
have at one time or another concurred:
Each possible world contains a number of individuals with certain proper-
ties and with certain relations to each other. We have to use these proper-
ties and relations to decide which member (if any) of a given possible
world is identical with a given member of another possible world. 2 (Hin-
tikka [1970], p.4lO)
Along with a true insight about constraints upon identity, Hintikka

here conveys something more questionable, namely his rejection of the
idea that a possible world may be specified by saying inter alia that
there, in that world, Jack Adams and Jill Jones in particular walk up
247
A. Bottani et al. (eds.), Individuals, &sence and Identity, 247-265.
Bredon Hill and then another possible world may be simply specified
by saying that there, in that second world, the very same boy and girl,
namely Jack Adams and Jill Jones, walk up Muswell Hill. Why is that
rejected? Why can things not be as simple as that?
Is the thought that, strictly, the identities of the objects in a world
are not to be stipulated in this way but must be founded, however indi-
rectly, in their other properties and relations? Is the thought that, in the
two would-be world-specifications just indicated, the names, 'Jack
Adams' and 'Jill Jones' are really shorthand in the constructor's vo-
cabulary for descriptions on the basis of which more elaborate identifi-
cations of the sort that Hintikka envisages might have been effected;
that otherwise - is this the idea? - these vocables are only the names
that two persons who are yet to be identified bear in the world under
construction?
If nobody has any inclination to say something analogous to this
about the ordinary properties and relations, then why is the claim so
perennially attractive where identity is concerned? Why is identity seen
as a poor relation of other properties and relations? Why is it problem-
atic to find Jill Jones in world one and in world two but unproblematic
to find the colour blue in world one and in world two? Why are we so
tempted to elide the difference between the reasonable claim that we
are not free to stipulate in a possible world whatever we like and the
questionable claim that the identities in a possible world need a foun-
dation? That is the question I begin from.
2.
Perhaps the relevant thought is that the identity of objects does not an-
nounce itself; that identity itself is unobservable, and unobservable
precisely because it is somehow consequential on other things. But if
so, then the linguistic philosopher within me wants to protest: "Do you
really mean that? Can one not see straight off and entirely directly
(however fallibly) that it is one's brother, not one's uncle, who has
come to call on one?" But, at best, that is a straw in the wind.
David Wiggins. Identity and Supervenience 249
Perhaps the relevant thought is that, even allowing for the point
just registered in my quietist cum language-centred protest, identity is
inscrutable except as resemblance; that identity is the vanishing point
of resemblance, and resemblance can only be a matter of properties
and relations.
If that were the thought, the quietist response (and my response)
would be that it hardly takes deep philosophy to reflect that resem-
blance itself is neither sufficient for identity, nor yet necessary. It is
only necessary to this extent: at any given time, x must exactly resem-
ble x. "Identity is all right", the quietist says, "leave it alone. Don't
spoil it. Try not to paint the lily."
3.
For some philosophers, philosophical exchanges of this sort may suf-

fice to restore the commonsensical view that it is simply wrong to say
that identity is unobservable. For others, however, they will only
deepen the apparent mystery about identity and strengthen the empiri-
cist resolve to account for our practical mastery of it in terms of what-
ever may be philosophically and cognitively less taxing. Since we
work so happily with identity (they whisper), how can it help but ride
on the back of something that is more manifest and manifestly man-
ageable?
Within technical philosophy, it may be that some think there is
hope of such a quasi-reduction in the "elimination" of identity for
which Hilbert and Bernays first gave the recipe and which Quine de-
scribes as follows:
Still nothing has been said [in Word and Object] as to the make-up of the
admissible vocabulary of unanalysed general terms. But of this particular
we may be sure: [the predicate] '=' will in effect be present, whether as an
unanalysed general term or in complex paraphrase, at least provided that
the vocabulary ofunanalysed general terms is finite. For, write 'if Fx then
Fy' and vice-versa with each ofthe absolute general terms of the vocabu-
lary in place of 'F'; also write '(z) (if Fxz then Fyz)' and '(z) (if Fzx then
Fzy)' and vice versa, with each of the dyadic relative terms in place of F;
and so on to '(z)(w) (if Fxzw then Fyzw)" etc. The conjunction of all these
formulae is coextensive with 'x = y' if any formula constructible from the
given vocabulary is; and otherwise we can without conflict adopt that
conjunction as our version of identity. In so doing we impose a certain
identification of indiscemibles, but only in a mild way. (Quine [1960],
p.230)
Wherever such eliminations are possible (the thought may run), there
is a clear sense in which the predicate 'is the same as' is indeed re-
ducible to other predicates and relation words. For 'x is the same as y'
is replaceable by another and longer open sentence with the same free
variables x and y, an open sentence that will have to have the same
extension. How better or more satisfyingly could the whole mystery be
dissipated of identity and our grasp of identity? The identity of x and y
comes down to the reflexivity and congruence of some relation in
which x stands to y. That is all there is to identity, or so it may be said.
4.
The first thing to remark is that the recipe Quine has given for the con-
fection of this first level surrogate of identity is indeed a sound recipe.
Its technical correctness is beyond doubt. The only question is what the
recipe shows about the relation of identity. Nor is there any doubt that
Quine furnishes a single and unitary recipe. One of the questions that
remains open, however, is whether the results of applying this recipe
do justice to the evident univocity of the sentence form 'x is the same
asy'.
Suppose that, for each of several languages unlike English in hav-
ing a clearly delimited set of predicates, each different from all the
others in respect of its basic vocabulary, we frame by exhaustion its
surrogate identity predicate, and then we replace the extant '=' predi-
cate of the language by the surrogate predicate. What do the surrogates
framed for these various languages have in common? Each expresses a
certain congruence with respect to the predicates of the particular lan-
guage into which it is introduced. But it may trouble us that, so long as
we restrict ourselves carefully to that which can be expressed in first-
David Wiggins, Identity and Supervenience 251
level terms and we are confined to that which can register within
Quine's various first-level versions of 'x = y', we cannot express what
it is that constitutes each surrogate an identity or congruence predicate.
For Quine's method works by surreptitious allusion to and silent de-
pendence upon an idea which is essentially second level. The second
level idea is that identity is the relation whose holding between a and b
ensures that every property of a is a property of b and vice versa. The
trouble with such a second-level account of identity, however, is that it
involves quantifying over all first-level properties, including identity
(unreconstructed identity) itself and the countless other properties that
latently involve identity.
It might be replied that none of this matters for the philosophical
purposes we are pursuing so long as there is for each language some
surrogate identity predicate.
Such a defence would raise difficult questions about how it is that
speakers with different active vocabularies that determine different
surrogates understand one another - unless on the basis of a second
level conception of identity. But let us not dwell on this here because
there is ground for a more specific reservation. It relates to ontology
and it can be put in Quine's words:
If the universe is taken as that of persons, and the predicates are inter-
preted in ways depending on nothing but people's incomes then the pro-
posed manner of defining 'x = y' will equate any persons who have equal
incomes; so here indeed is an unfavourable case where 'x y' [defined in
=
the manner indicated here, as in 'Reply to Professor Marcus'] does not
come out with the sense of genuine identity. (Quine [1963], § 1, p.15)
Quine'S doctrine about this difficulty - if it is a difficulty - has been as

follows: straighten the matter out by reconstruing the members of the
universe as whole income groups. (Cf. Quine [1953], p.70.) But Quine
also says that, whether or not we reconstrue ontology in this way, "no
discrepancies between [the surrogate for identity] and genuine identity
can be registered in terms of the theory itself'.
5.
No doubt there are purposes for which Quine's remedy is fully ade-
quate. But would it give everything he needed to the kind of empiricist
philosopher I have undertaken to answer, the one (I mean) who worries
about the empirical credentials of umeconstructed identity and thinks
he sees what he is looking for in the idea that one can dismantle the
identity relation into reflexiveness and congruence?
Suppose one believes that the proper interpretation of a living lan-
guage depends on the interpreter's taking the language in its full prac-
tical context and explaining the speakers' use of the predicates of the
language by reference to its speakers' constant commerce with one an-
other and with various objects that are manifest in the portion of the
world to which they are seen by the interpreter as responding reasona-
bly in perception, in thought and feeling, and in action. If that is the
picture one has of interpretation (and this might as well be the empiri-
cist's picture too), then one will want to question whether absolutely
everything that grounds interpretation for a given language must be
such as to register in the explicit predicative repertoire of the language
itself. Surely it can be practically manifest to the interpreter that the
speaker is concerned with people not groups, for instance, and practi-
cally manifest in ways that outrun explicit speech. If that is right, one
concludes, then Quine'S reconstrual of ontology cannot fill the bill.
All right, some defender of the surrogate may say. Let me allow
that an interpreter needs to interact with the objects his subjects inter-
act with. Let me allow that one who interprets has to deploy practical
understanding in tracking his subjects' practical understanding. But I
still side with Quine. There must be some process (the defender may
say) by which an interpreter's implicit or practical understanding of the
role of language in the life and conduct of its speakers can be made
more and more explicit. Let the interpreter draw on his own predica-
tive repertoire. And then, where necessary, let him (however notion-
ally) fill all the gaps in the language he interprets. That will sort out the
people/income groups problem. It will sort out any other ontological
difference that is a real difference.
Yes, I reply, but now one must ask whether this reply will seem
satisfactory so soon as one reflects that the process of explication need
David Wiggins. Identity and Supervenience 253
not terminate. In so far as the process does not terminate, it seems that,
even if the predicative resources and repertoire of the language are
constantly extended towards semantic saturation, the intended defini-
tion by means of exhaustion of predicates may never be available. Al-
ways the explicit predicative repertoire may be expected to lag behind
the demonstrative cum practical capacities of subjects and interpreters.
6.
There are questions I do not want to try to resolve here, not least about
Quinean doctrines of indeterminacy or unscrutability of reference. It is
more important for present purposes to gain a fuller appreciation of the
scale of the difficulty we are concerned with. A further consideration is
this. John Wallace points out that Quine's method of reducing '='
yields rather stranger results than any that Quine himself has explicitly
countenanced. 3 Applied to a quantificational language with just three
unanalysed predicates 'x is a forest', 'x is a tree' and 'x grows in y',
Quine'S proposal forces the truth-value false upon the sentence 'In
every forest there grow many trees' .
The defender of surrogate identity can claim that such troubles as
this will disappear so soon as we consider first-order theories with an
expressive power more closely approximating to that of English.
Wallace suggests that the claim of eliminability would need to be re-
phrased as follows: if a theory has a finite number of unanalysed predi-
cates, then a finite number of predicates can be added to the theory, so
that, in the resulting expanded theory, what we usually intend by iden-
tity is eliminable by the prescribed method. But now - at least for the
purposes of the evaluation of the idea that identity can be reduced or
explained in this way - everything is seen to depend on what predi-
cates are added in order to force the desired truth-value upon sentences
involving the constructed predicate that is to be introduced in lieu of
'='. Suppose that the appearances can only be saved if monadic and
polyadic predicates presupposing identity or place-, time- or thing-
individuation are supplied. If so, it is upon the presence of these that
the success of the elimination recipe will depend. But in that case, no
real elimination has been achieved of the kind that that empiricist I be-
gan with ought to have an interest in. The whole case remains to be
made for the philosophical claim that the presence in a language of the
identity predicate boils down to some predicable's introducing a re-
flexive relation that is a congruence relation with respect to the prop-
erties imported by the predicates of the language.
7.
If Wallace's suggestion were to be fully explored on Quine's behalf,

then it would matter which predicates were really and truly identity-
free. But at this point the empiricist whose outlook I set out to charac-
terize may want to try another tack. He may say this: "Maybe it was a
mistake for me to get involved in questions about reduction and quasi-
reduction. The thing that I want to hold out for may amount to much
less than that. My thought is that, if identity is to be empirically tracta-
ble, then it must be that in any given situation and context, all the
identities of the things involved are fixed or fastened down from some
arbitrarily large sufficiency of information about all the various other
predicates and relations that are instantiated there. Is not the fixation of
identity by other properties and relations the condition of significance
and of workability for the idea of identity? Whatever questions may
remain open about the reducibility of identity or the elimination of
Quine'S elimination recipe, identity must supervene on other properties
and relations."
8.
How is this idea to be evaluated? The empiricist speaks of other prop-

erties and relations fixing identities, fastening them down, or being the
basis for ascertaining identities. But these may not be exactly the
same. So, before we enter into that, let us have a case before us where
an identity is determined in a striking and unexpected way from some
narrative about a whole group of entities.
In Conan Doyle's story Silver Blaze we have the following char-

acters:
(1) Silver Blaze, a brilliant four year old owned by Colonel

Ross, first favourite for the Wessex Cup, but now gone
mIssmg.
(2) John Straker, a retired jockey, trainer for Colonel Ross of

Silver Blaze and of three other horses; beaten to death on
the moor on the night of Silver Blaze's disappearance, and
found to be carrying an ivory-handled knife made by Weiss
& Co., London.
(3) Ned Hunter, a stable lad whose supper was interrupted on

the night before the disappearance of the horse, by one
Fitzroy Simpson who was seeking information about the
relative merits of Silver Blaze and another horse from the
same stable, Bayard. On the morning of the horse's disap-
pearance he awoke drugged.
(4) Stable lad II.
(5) Stable lad III.
(6) Fitzroy Simpson, a genteel part-time book-maker, now ar-

rested for the murder of Straker.
(7) Silas Brown, the trainer of a neighbouring stables at Cap-

leton; whose footprints are found by Watson and Holmes
next to Silver Blaze's footprints outside the Capleton sta-
ble, where Silver Blaze is found by Holmes (though not by
the police) disguised by horse-fakery.
(8) The dog at Colonel Ross's stable, a dog who did not bark
and did not awaken the stable lads on the night of Silver
Blazes's disappearance.
(9) The abductor of Silver Blaze.
(10) The killer of John Straker.

256 Individuals, Essence, and [dentity
Holmes reflects that character (9) the abductor, must have been famil-
iar to (8) the dog, Otherwise the dog would have barked when the ab-
ductor led Silver Blaze out onto the moor. So (2) John Straker = (9),
The knife suggests that Straker meant to lame Silver Blaze. If so, he
must have intended to lay money against Silver Blaze. Who then is
(10)? Scarcely (6), who was abroad with a leaded stick but had no mo-
tive. The only other possibility is the horse himself. In fright and reac-
tion to the prick of the knife, he kicked and trampled John Straker. So
(1) = (10). In which case character (6) is innocent. Our cast of ten
characters is really a cast of eight.
Is this sort of thing, which is utterly familiar to us, the sort of thing
that the empiricist has in mind? If it is, then the obvious objection
seems to be that what is here in question is deductive reconstruction
from an incomplete narrative of a fuller narrative; and this is the re-
covery of a narrative in which things have an identity already deter-
mined. Epistemically speaking, there is here a triumph, a triumph of
"deduction". But there is no model here for the constitutive determina-
tion of identity out of properties and relations that do not include iden-
tity.
Let this obvious finding be the occasion to reflect that the idea of
supervenience is often said by those who set store by it to be this: that
properties from a certain range of higher level properties should super-
vene on some other range of more explanatorily basic properties. But,
if supervenience is a relationship between classes of properties, then is
not the identity itself of the subjects of such properties to be treated as
something taken already for granted? The supervenientists seem to
suggest this themselves when they say (e.g.) that "it is not possible that
two things should be indiscernible in respect of their lower level prop-
erties without also being indiscernible in respect of their upper level
properties". On this account of the matter, identity itself lies outside
the intended area of supervenience.
To defuse this objection, the empiricist might want to explore the
following suggestion: facts about the identity of objects supervene on
facts about the exemplification of properties and relations other than
the identity of objects. When the claim is put in this way, it is left open
what the exemplification of the other properties and relations comes to.
Maybe these exemplifications are given in feature placing or pointil-
liste style without any particular commitment to any particular ontol-

ogy. Maybe matters of ontology and identity supervene simultaneously
on matters about the distribution ofJeatures.
lO.
There are things I do not understand here. But even from this be-
nighted condition, I am moved to mention an apparent difficulty, a dif-
ficulty long since familiar from discussions of Leibniz's Principle ea-
dem sunt quorum unum alteri substitui potest salva veritate. The Iden-
tity of Indiscernibles, on one standard interpretation of Leibniz, comes
to this: x is identical with y if and only if, for all pure properties cP, x
has cP if and only if y has cP; where pure properties are those that in-
volve or presuppose neither identity itself nor place-, time-, nor indi-
viduation, nor thing-individuation.
This idea of a pure property is right for Leibniz and it is equally
right for our purposes. But then, as Wittgenstein noted (Tractatus
Logico-Philosophicus 5.5302), any such principle will rule out two
objects' having absolutely all their properties in common. So, as Max
Black pointed out, it rules out the logical possibility of a universe con-
sisting of two qualitatively indistinguishable spheres. The wild implau-
sibility of this exclusion seems unfavourable to any supervenience the-
sis. Following on from where Wittgenstein and Black left off, P.F.
Strawson then pointed out, in similar opposition to Leibnizian mona-
dology and the Identity of Indiscernibles, that
It is necessary only to imagine the universe in question being repetitive or

symmetrical in certain ways in order to see that there might be numerically
different points of view from which scenes presented [to an individual
monad] would be qualitatively indistinguishable even though they com-
prehended the entire universe [ .. .]. A very simple illustration will serve.
Think of a chess-board [.. .]. The problem is to provide individuating de-
scriptions of each square, and to do so in terms of the view of the rest of
the board obtainable from each square, [ ... ] the problem cannot be solved
even if the view from each square is allowed to comprehend the entire
board. (Strawson [1959], pp.l22-3.)
When faced with such examples, some people (once upon a time I was
one of them) may not flinch from the thought that it is a presupposition
of our individuative practices that we do not live in such worlds as
Strawson and Wittgenstein have described. Because the presupposition
is global, they believe it can be lived with - pragmatically, say, and/or
as a "framework principle".
If this were the worst of it (I reply) then maybe it would be bear-
able. But matters are even worse than this represents. Is it also a pre-
supposition of our individuative practices, or a framework principle for
them, that there are no fully symmetrical objects? (A symmetrical ob-
ject with exactly matching top half and bottom half has parts that are
indiscernible. It is effectively reduced to half of itself by the Identity of
Indiscernibles as strictly construed. Moreover, an object that is sym-
metrical about all planes which bisect it seems to be precluded alto-
gether. For not only would the top half of such an object be indiscerni-
ble from its bottom half. The left side of the bottom half would be in-
discernible from the right side of the bottom half. And as regards the
residual eighth ... Eventually, we are left with nothing but a line and
then, at the limit, a geometrical point.) Are we to believe that it is a
"framework principle" that no object is symmetrical about all planes
which bisect it?
Strawson's example was devised to embarrass the Leibnizian use
of the Identity of Indiscernibles. But more generally, the thing it seems
to show is that the facts about identity in a given set-up need not su-
pervene on the qualitative and structural properties of the array that
constitutes the set-up. Rather, we may need facts about identity in or-
der to characterize the array itself. If that is right, then no interesting
thesis of supervenience seems defensible. Of course, it is a crucially
important fact about identity, it is of the essence of identity, that it rubs
shoulders with other properties and relations. But that is not enough
for supervenience. Identity is just identity. It cannot be confected or
fixed from other properties and relations.
11.
That is the conclusion that I am led to. But the empiricist's response
may be to say that, with a symmetrical object, we can make the dis-
tinction between one half and the other deictically. A similar response
might have been offered to the WittgensteiniMax Black set-up.
If the empiricist says that, then I entirely agree. But in agreeing I
am not restoring any presumption in favour of the thesis that identity is
some sort of resultant from other properties and relations. Rather the
point that is agreed about deixis seems to me to remove the very need
for that presumption. For it undermines the main reason that philoso-
phers might give themselves for adhering to the Identity of Indiscerni-
bles. Is it not the whole answer to the empiricist's worries about iden-
tity for him to come to see our grasp of identity (and our grasp of
shape and array) as a by-product of our grasp of reference to objects
and the singling out of objects? Identity is not definable. It is as fun-
damental as any notion can be. But that's all right. It's what you would
expect. For even on these terms, the notion of identity is tractable.
12.
To try to render this plausible, I begin by suggesting that we should

recognize a class of primary identity judgments. This class will in the
end be hospitably conceived. But let us begin with a special case - the
primitive case, so to say, of a primary judgement. At least in the
primitive case, the empiricist can have everything that he wants. In the
extensions of this class that add other judgments to the class of primary
identity judgments the empiricist will be invited to see judgments that
concern the very same relation.
Suppose that someone, well knowing his way around and being
located in a part of the world where he is as much at home as it is pos-
sible to be, looks into a field and singles out there a bay horse. Suppose
he watches the horse scratch its back on the gate and then walk to the
opposite comer of the field. Such continuous observation is secure and
it enables this thinker, looking from where he is, to place on the record
the identity judgment: "The bay horse I am watching go to the comer
of the field is the same as the horse I earlier watched scratching its
back on the gate". Or else he can say, pointing to it at the later mo-
ment, "That is the same horse as the horse I saw scratching its back on
the gate". It makes no difference which of these things he says. In ei-

ther case, the relation that is in question here is identity as restricted to
horses. In Principia Mathematica notation (see Volume I, §35), it is
the relation = [horse. Meanwhile, the notion of a horse is a determina-
tion of the concept object of some kind (or object of kind ... ). The con-
tent of the claim we are concerned with seems to be that, if you single
out the horse that scratched itself against the fence and you trace its
life-span forward or you single out the horse in the corner and trace its
life history (however notionally) backward, there's no difference in
what you end up with.
It is natural to put matters so. But let not the impression be created
that any of this is intended as a philosophical analysis of identity. For I
am saying there can be no such thing. Rather, it's the beginning of an
elucidation, an elucidation which would need to be fortified by our
enlarging upon everything which is involved in the business of finding
or recognizing an object under a thing-kind, singling out an object, and
prolonging the singling out of an object into the tracking of the object
and the marking its shape. If our chief question relates to the tractabil-
ity of the notion of identity and what thinkers need to have in order to
deploy their grasp of it, then it is helpful to bring in here Leibniz's ac-
count of that which he called clear (that is practically effective) non-
distinct (i.e. not yet explicit or fully articulate) knowledge. (Cf. Leib-
niz, Gerhardt (ed.) [1978], N, p.422.) Having such clear non-distinct
knowledge (a perfectly ordinary kind of knowledge) of a thing-kind is
a matter of having an understanding (in part practical and recogni-
tional, not necessarily verbal and in part essentially non-verbal) of how
horses look, how they behave, how they move, how they interact with
other things, what tends to happen to them, and so on - an under-
standing that will be practically deployed in tandem with other under-
standings that are coeval with it. See Wiggins [2001], Preamble §1O.
13.
So much for the primitive case and the elucidation of what is involved
in that case. The larger class of primary judgments of identity one as-
sembles by reaching beyond the primitive cases and counting into this
larger class judgments that are answerable to the very same dialectic of
sameness and otherness by which we arrive at our judgment in the case

of continuous observation. As a first fix, but it is only that, let me say
that here, with the cases that go beyond the primitive case, one's cog-
nitive effort is organized around the question of what could have been
revealed, even in cases where nobody did track the things answering to
the terms of the identity, if someone had singled out x, had singled out
y. had either tracked x forward or reconstructed the history of y back-
ward from the time by reference to which it is designated. But this is
only a first fix, and not to be taken seriously. For, as is perfectly noto-
rious, counterfactuals of this kind stir up a mass of questions one nei-
ther wants nor needs to address. (Suppose the object x were allergic, so
to speak, to being tracked by any agent remotely like the person doing
the tracking or attempting to make an identity judgment. Suppose x
were capable of taking evasive action, diving or ducking out of sight.
Suppose the person tracking and making the judgment were simply
incapable of facing up to the actuality of objects remotely similar to x.
Suppose x were Medusa. Under these conditions, nothing at all would
be "revealed".) Rather than try to deal seriously in such counterfactu-
als, it would be much better for us to start to think through the princi-
ples on which one must make judgments of sameness and difference if
sameness is to be the sort of relation that is revealed in the primitive
case. In cases exceeding the limits of the primitive case, one who in-
vokes the identity relation proceeds by applying the very same princi-
ples, in what is recognizably the same dialectic of considerations
counting for and against claims of persistence and identity. That is
what it will take for someone to concern himself with the same rela-
tion. In Sameness and Substance, that was the role of various princi-
ples I called D principles. (There is D(x) for instance, the Only a and
b rule, which says that in questions of a's coincidence (or not) with b
there is no room, constitutively, for reference to objects other than a or
b.) The status of these principles is this: they are principles without
which our attempts at individuation would fail to preserve the formal
properties of identity. I shall say no more of them here, however. For
in this place I want to conclude with three other observations.
14.
First, if we look again at the primitive case based on singlings out and
prolongations of singlings out, then it is worth noting something about
congruence. Simply by virtue of what he finds on the basis of a sin-
gling out, the prolongation of that singling out and the renewal under
changing or changeable conditions of that act of singling out, a thinker
who does these acts in the way that I have described can then declare -
no matter what property l/J is and no matter whether the question of a's
or b's instantiating l/J figured within any inquiry of his into the spatio-
temporal paths of horse a and horse b - that object a has l/J if and only
if object b has l/J. For simply to determine correctly the answer to the
continuity question about the horse, the question (that is) about the
traceability through their life-histories of the horse that scratched itself
and the horse now in the comer, precisely is to settle it that, no matter
what property l/J is, horse a has l/J if and only ifb has l/J.
In epistemological reflection of this criteriological point, the hy-
pothesis now presents itself that it is impossible even in theory to con-
ceive of some way independent of the prior discovery that a = b by
which to establish that a and b have all and only the same properties.
For suppose, simply seeing identity as reflexivity plus congruence, one
were to renounce all elucidations of identity other than those given in
terms of a's and b's complete community of properties. Then how
would one think about the non-permanent properties enjoyed by an
individual a identified with respect to the past and the properties en-
joyed by an individual b identified with respect to the present? One is
only justified in pooling the non-permanent properties of a and b if
there is some other basis for the identity of a and b than their having
all their properties in common. The only basis on which such pooling
is possible (I suggest) is the notion of a certain kind of continuant with
respect to which one can ask 'what is it for an f to persist?' It is this
that makes room for the idea of a sequential history of a thing's doings
and undergoings. Without this idea, little sense will be made of very
much that we actually do with the concept of identity. As Leibniz puts
the point that I too have wanted to insist upon, "By itself continuity no
more constitutes substance than does multitude or number.... Some-
thing is necessary to be numbered, repeated and continued" (Gerhardt
(ed.) [1978], II, 169). Indispensable to our deployment of the deter-

minable ideas of entity, identity and substance is our deployment of
our countless ideas of their determinations.
15.
My second closing point is that the claim I have just made about con-
gruence is not limited to concrete continuants. Consider natural num-
bers, for instance. Suppose that, in making a tally of the fs, Edward
first uses the vocable 'one' then two more vocables. His tally of theft
he gives as 'three'. Suppose Kallias, making a tally of the gs, first de-
ploys 'hen' then uses two more vocables of which the second is 'tria'.
Then on this basis alone, which suffices to secure the numerical iden-
tity, we can say that whatever is true of Edward's tally number is true
of Kallias' tally number. (Or so anyone will say who agrees to make
sense of natural numbers as objects.) No more is needed. In practice
we don't start with congruence, we reach congruence last.
16.
One concluding point. A fair summary of a great deal that I have said
could be condensed into this.
(1) a=b~(3f)(a~b)
There is no question of supplanting the left hand side by the right hand
side and achieving thereby a philosophical analysis. But the right hand
side casts light on the left hand side because it indicates (or reminds
us) how, in practice, we exercise our understanding of identity. We
exercise it by deploying among the other understandings that we have,
our understanding of some determinate or determination of the deter-
minable notion entity of some particular/determinate kind. In deploy-
ing that in the world of objects we exercise the clear indistinct idea we
have of that particular determination. The possibility of such determi-
nations of that determinable is something coeval with the notion of

identity itself.
Let me round out this last contention and focus better upon its
theoretical commitments by saying a word about something that Quine
has said:
When we do propound identity conditions for bodies or persons or classes,
we are using the prior concept of identity in the special task of clarifying
the term 'body' or 'person' or 'class'; for an essential part of the clarifica-
tion of a term is clarification of the standard by which we individuate its
denotata. (see Quine [1964], p.102)
I applaud and admire the last sentence - though I am tempted (because

I think of individuation as heavily involved with the diachronic) to
read more into it than Quine may have intended. I like much less, how-
ever, Quine'S suggestion that the notion of identity is prior, which
must mean that it is prior even to the determinable notion of an entity
of some kind, the notion I think of as bringing with it the logical space
within which may be found countless determinates or determinations
of that determinable. I think such a determinable would be what Witt-
genstein called in the Tractatus a formal concept. If congruence and
reflexivity will not furnish, self-sufficiently and all by themselves, eve-
rything that theory needs in order to understand identity, then I submit
that it is better to think of identity as one of a whole skein of coeval
and equally primitive notions which human thinkers have to catch on
to simultaneously. The grasp of identity, the grasp of the determinable
idea of a thing with a way of being, the grasp of predication, the readi-
ness to assign objects encountered to kinds, the readiness to catch on to
different ways of being and persisting - these things all presuppose one
another. The best thing for theory or philosophy to do here, if it is to
understand the empirical tractability of identity, is to shadow the men-
tal acts of those for whom these ideas have come together and culmi-
nated in a practical mastery that is only mysterious to us by virtue of its
being so familiar to us and so close to us.
Notes
1 I make this claim in Chapter Six of Sameness and Substance Renewed (Wig-
gins [2001]). Several extended passages of the present article appear in another form
in Sameness and Substance Renewed. They appear here by kind permission of Cam-
bridge University Press.
2 As is well known, Saul Kripke has spoken out eloquently against rulings of this
kind. See Kripke [1980]. If his efforts had eradicated completely the underlying
philosophical conceptions that have seemed to sustain such a ruling, then I should not
have begun my speech here.
3 See John Wallace [1964]. See pp. 80ff. John Wallace's original sentence was
"in every forest there grows more than one tree". Timothy Williamson points out to
me that the sentence it would be better for Wallace to consider is "In every forest
there grow many trees". Unlike "more than one", "many" does not reimport identity.
Then (Williamson suggests), in order to make Wallace's point properly secure, let
the quantificational language be treated as possessing a "many" quantifier as primi-
tive. The philosophical significance of Quine's elimination recipe ought not to de-
pend on the question whether "many" is admissible as a logical constant.
COMMENTS ON WIGGINS'S "IDENTITY
AND SUPERVENIENCE"
Edmund Runggaldier
I would like to concentrate on these three issues: (a) on the question

whether identity is primitive; (b) on the thesis of the sortal dependency
of identity; (c) on the belief that we are continuants - even if we do
not know what our sortal determination is.
(a) Wiggins's thesis is that identity relations cannot be reduced to
nor analyzed into more basic or fundamental relations, they do not
even supervene on any such relations: identity is not "some sort of re-
sultant from other properties and relations". One cannot simply see
identity as reflexivity plus congruence!
Thus, Wiggins argues against Quine's attempts to analyze identity:
Quine relies on an idea which is essentially second level. The trouble
with a second-level account of identity, however, is that it involves
quantifying over all first-level properties, including identity (unrecon-
structed identity) and countless other properties that latently involve
identity. Wiggins argues further against the standard attempts at recon-
structing identity on the basis of the identity ofindiscernibles. He sums
up by claiming that identity is as fundamental as any notion can be.
But - surprisingly enough - Wiggins does not explicitly state that
identity is primitive. I see this as a symptom of his philosophical atti-
tude: he does not share a bad kind of foundationalism that stops at
267
A. Bottani et al. (eds.), Individuals, &sence and Identity. 267-271.
claims of primitiveness. That concepts of a certain kind or the corre-

sponding entities do not supervene on others does not mean that there
is no need to explicate or elucidate them. Thus, even when we accept
the thesis that identity is fundamental, we can ask what we mean by
identity statements or why we assume that we understand what others
say when they claim that x is identical with y? Wiggins tackles this
question.
For the ontology in every-day-discourse what counts is mainly
identity in a diachronic sense. We want to know who has done certain
deeds, who is who. However, to elucidate or explicate the content of
diachronic identity statements is much more challenging than that of
synchronic statements. As a matter of fact, many do reject diachronic
identity altogether in favour of various types of continuity or sameness
in quality or form. The notion of continuity seems easier to grasp.
For Wiggins, now, the key to understand and elucidate diachronic
identity lies in our ordinary business of singling out objects and trac-
ing them through time. The corresponding judgments or the mecha-
nism of identification and individuation help us to grasp what we are
after when we make identity statements. Wiggins begins the elucida-
tion with hints at what is involved in the business of finding a thing
under a kind prolonging that singling out and tracking the thing. This
also involves having a practical conception of a sortal or kind.
(b) Wiggins is one of the most successful philosophers - at least in
the German speaking world - in defending the crucial role of sortal
concepts both for identifying and for tracing objects through time. At
the core of his defence stands the thesis of the sortal dependency of
individuation and diachronic identity: for every true identity sentence
there is an identity sentence covered by a substance concept for some
particular kind of thing. The older background of the thesis of the sor-
tal dependency of identity is the Aristotelian conviction, that some-
thing ought to be formed into a substantial mode in order to be real or
actual.
The modem thesis that the very notion of individual or individual
thing presupposes sortal determination can be understood in a conven-
tionalistic but also in a realistic sense.
Edmund Runggaldier, Comments on Wiggins 269
In its realist versions it gets more plausible as soon as we take into

account the temporal aspect of the existence of things: sortal concepts
detennine or express nonns of coming to be, possible change and
passing away of the individuals falling under them. The set of the at-
oms an animal e.g. is composed of at a given time is subject to differ-
ent existence conditions from the organism as a whole. Among arti-
facts e.g. the clump of metal is different from the statue made up of it
even if they cohabit.
In order to trace objects through their life-span we need at least a
sortal rule or an implicit notion of the essence of the thing. Only
against the background of their sortal detennination does it make sense
to speak of them as continuants; only as sortally detennined are they
numerable.
In the opening speech of the Wittgenstein Congress in Austria in
1999 van Inwagen defended the opposite view, the view that it does
make sense to claim that there is a certain number of objects as such in
a given universe. If van Inwagen is right it would make sense to speak
of individuals and thereby of their identity through time independently
of or prior to their sortal detennination. His arguments are sophisti-
cated but convince only against the background of a presupposed uni-
verse of discourse. For the purpose of the elucidation of our notion of
identity as applied in everyday language this presupposes what is at
stake.
Wiggins's attempts at elucidating diachronic identity are based on
our ordinary way of singling out and tracing objects. This amounts to
the effort of uncovering the sortal principles governing the identity of
objects. To continue the elucidation, Wiggins says, one should try to
find out what would have been revealed by the exercise of singling out
each object of an identity statement and tracing its life-span back to the
beginning and/or forward to the end, even in cases where nobody did
track the objects.
(c) Wiggins's strategy and efforts to elucidate identity seem to me

promising. But may I ask: what about those referential acts which do
not presuppose singling out or any kind of identification, like those by
means of the indexical'!,? When I refer to myself by the indexical 'I'
I do not have to identify myself. I need neither descriptive knowledge

nor a conception of my sortal determination.
Why do I ask this question? I think, if people believe in diachronic
identity, they do so mainly because they believe that they themselves
exist through time, in other words, that they themselves are contin-
uants. Especially in discussions with students I had to realize that
many do reject the belief in continuants on the objective level, but on
the level of their attitude to themselves they seem to accept it. They
believe that they themselves go with the flow of time and thus endure
through time. They are firmly convinced that they themselves have
done various deeds in the past and they firmly hope that they them-
selves will live and act in the future.
If this is so, shouldn't the key for explicating or elucidating iden-
tity through time be the belie/that we ourselves are continuants? I do
not know the answer, but I should like to hint at the following prob-
lem: On the one hand there are convincing reasons to account for
identity by means of the sortal determination of individuals and con-
vincing reasons to elucidate identity in the manner Wiggins proposes
it, on the other hand it seems very plausible that the key notion of an
adequate understanding of diachronic identity is the belief in personal
identity which does not rely on any singling out or on any sortal
knowledge.
In fact, in order to predicate various things of me in the past or in
the future I do not have to single myself out in the past or in the future.
Transtemporal predication via the indexical 'I' is much simpler: I do
not have to know how to establish that I am identical with the one
having done the predicated things nor that I am identical with the one
going to do these things in the future. Here the question how I know
that I am the same subject of predication simple does not make sense.
The analogy to the problem of transworld identity with the opposite
views of Plantinga and David Lewis is well known.
In order to believe or at least to be able to believe in the transtem-
poral identity ofmyselfI do not have to know my own sortal determi-
nation either. And who knows what it is? I am - so I hope - a person
and a self. But "Perhaps persons or selves do not constitute a kind of
things, all instances of which share the same identity conditions."
(Lowe [2000], p.265). If I further take into account various thought
Edmund Runggaldier, Comments on Wiggins 271
experiments I see that I could be radically different from what I actu-

ally am. But all this does not destroy my firm belief that it would be
me living through these experiments and existing at different instants
of time.
Does this undermine Wiggins's strategy and attempts to elucidate
identity on the basis of the sortal dependency of Identity? I do not
think so. But it manifests the relative value of his strategy - I suppose.
A way to supplement Wiggins attempts would be to take the first-
person perspective seriously - precisely because of its relevance to the
belief in our own identity.
Wiggins thinks of identity - so it seems - as one of a whole net of
coeval equally basic notions which presuppose one another. Why
should we exclude from them those fundamental notions, we stick to
on the basis of the first-person perspective?
VAGUENESS, IDENTITY AND LEIBNIZ'S
LAW*
Timothy Williamson
When we reflect on the individuation of objects, we soon encounter

hard cases. We are presented with pairs of objects the grounds for clas-
sifying which as identical look no better and no worse than the grounds
for classifying them as distinct; we have no idea how to decide the
question. For instance, the diachronic identity of the ship of Theseus
was already a matter of philosophical debate when Plutarch wrote, and
has still not been settled:
The thirty-oared galley in which Theseus sailed with the youths and re-
turned safely was preserved by the Athenians down to the time of De-
metrius of Phalerum. At intervals, they removed the old timbers and re-
placed them with sound ones, so that the ship became a classic illustration
for the philosophers of the disputed question of growth and change, some
of them arguing that it remained the same, and others that it became a dif-
ferent vessel. (Plutarch, 'Life of Theseus' §23, Scott Kilvert (trans.)
[1960])
If the name'S l' is conferred by ostension at the time of Theseus on

the ship then present, and the name 'S2' is conferred by ostension at
the time of Demetrius of Phalerum on the ship present at that later
time, then it seems unclear whether S 1 = S2; S 1 and S2 are apparently
273
A. Bottani et al. (elis.), Individuals, Essence and Identity, 273-303.
neither clearly identical nor clearly distinct. Examples of this kind

abound.
Someone might describe the case thus:
There are objects of which it is vague whether they are identical. Such
vagueness concerns the objects themselves, not their names. When the
name 'SI' was introduced, it was clear which ship was being pointed at;
likewise when the name 'S2' was introduced. Thus there are objects (the
ships S I and S2) and a relation (identity) such that it is vague whether
those objects stand in that relation. That amounts to vagueness in reality
itself, not merely in our mental or linguistic representations of reality.
Such a description contradicts the usual understanding of vagueness as

located only in representations.
We might doubt whether ships are perfectly representation-
independent objects, for they are artefacts, individuated at least partly
in accord with the intentions of their makers or users; intentions in-
volve mental representations. But similar identity puzzles arise for
natural objects, such as planets, many of which (idealists notwith-
standing) would have had the same existence and nature even if beings
capable of forming representations had never evolved.
Cases of vague identity have also been claimed to occur in quan-
tum mechanics (Lowe [1994]). However, no theory of vagueness has
made any serious contribution to the understanding of problems in
quantum mechanics. It is not as though we understand what is physi-
cally happening in such cases but need a theory of vagueness to de-
scribe it clearly. Quantum mechanics in its present mystifying state has
no obvious morals for the theory of vagueness.
Many philosophers conceive vagueness as a kind of indetermi-
nacy. On such a view, if it is vague whether objects are identical then it
is indeterminate whether they are identical, in roughly the sense that
there is no right answer to the question 'Are they identical?'. Thus re-
ality itself would be indeterminate. That would explain our prolonged
failure to resolve philosophical disputes about identity. Yet Gareth Ev-
ans [1978] and Nathan Salmon [1982] (pp.243-5), have argued for-
mally that it cannot be indeterminate whether objects are identical.
Very roughly: if it is indeterminate whether x is identical with y, then x
has a property that y lacks, because it is determinate that x is self-
Timothy Williamson, Vagueness, Identity, and Leibniz's Law 275
identical, and therefore determinate whether x is identical with x; thus

x is distinct from y, by Leibniz's Law of the indiscernibility of identi-
cals, so it is after all not indeterminate whether they are identical. The
claim of vague identity is self-refuting. In response, numerous attempts
have been made to construct a coherent account of vague identity
within a framework of many-valued logic.
Section 1 of this paper argues, on grounds independent of those
adduced by Evans and Salmon, that many-valued logic is unpromising
as a context for an account of vague identity. Section 2 makes some
more general methodological comments on non-classical treatments of
vague identity. Section 3 traces objections to the Evans argument to a
significant difficulty in formulating an appropriate version of Leibniz' s
Law to govern the logic of identity; but the difficulty can be finessed in
a way which vindicates the spirit of the Evans argument. Section 4
shows how to make sense of a modest sort of vague identity while re-
specting that argument, by using an epistemic account of vagueness.
1.
If identity can be vague and vagueness is a kind of indeterminacy, then

'Yes' and 'No' may be equally inappropriate answers to the question
'Are S1 and S2 identical?'. On this view, the statement 'S1 == S2' and
its negation 'Sl "* S2' have exactly the same truth-status, that is, ex-
actly the same place on the dimension of truth and falsity. Call that
truth-status neutrality. Given that truth and falsity are mutually exclu-
sive and that the negation of a. is true if and only if a. is false and false
if and only if a. is true, it follows (at least by classical metalogic) that
'Sl == S2' is neither true nor false. For if 'Sl == S2' is true and has ex-
actly the same truth-status as 'S1 "* S2' then 'S1 "* S2' is true, so 'S1 ==
S2' is false and therefore both true and false. Similarly, if'S1 == S2' is
false and has exactly the same truth-status as 'S1 "* S2' then 'S1 "* S2'
is false, so'S 1 == S2' is true and therefore again both true and false. A
neutral statement is neither true nor false.
It is far from obvious that the claim that one can say that some-
thing is the case without saying anything true or false can be reconciled
with the fundamental principles governing the notions of truth and fal-
sity (Williamson [1994], pp.187-98; see Andjelkovic and Williamson
[2000] for further issues). However, for the sake of argument we may
bracket this general concern in order to investigate some more specific
problems for many-valued logic as a framework for an account of
vague identity.
The relevant systems of many-valued logic do more than postulate
a neutral status. They treat it as another truth-value, an alternative to
truth and falsity. The two most salient systems are three-valued logic
and fuzzy logic (continuum-valued logic). In three-valued logic, the
truth-values are just truth, falsity and neutrality. Fuzzy logic, by con-
trast, postulates a continuum of degrees of truth from perfect truth to
perfect falsity, with neutrality in the middle; these degrees are typically
identified with the real numbers in the closed interval [0,1], with 1 as
perfect truth, 0 as perfect falsity and 0.5 as perfect neutrality.
The most distinctive feature of many-valued logic is that by defi-
nition it generalizes the truth-tables of two-valued Boolean logic by
postulating a principle of generalized truth-functionality: it computes
the truth-value of a complex sentence formed by applying operators
such as negation, conjunction, disjunction or the conditional to simpler
sentences as a function of the truth-values of those constituent sen-
tences. For example, given that neutrality is a truth-value and that the
negation of the neutral 'SI = S2' is neutral too, generalized truth-
functionality implies that the negation of any neutral sentence is also
neutral. Granted that negation permutes truth and falsity, that is enough
to fix the truth-table for negation in three-valued logic. The easiest
generalization to continuum-valued logic is that if the degree of truth
of a sentence a is x, then the degree of truth of its negation -u is I-x,
just as the probability of """'a is I-x if the probability of a is x. The
phrase "many-valued logic" will be restricted here to systems with
generalized truth-functionality.
Let us focus on the binary connective of conjunction. Since the
conjunction a /\ 13 entails each of a and 13, but entails no more than is
needed to have those two entailments, it is natural in continuum-valued
logic to assign a /\ 13 a degree of truth no higher than the degree of
truth of a and no higher than the degree of truth of 13, but no lower
than is needed to meet those two constraints. Thus the degree of truth
of (l /\ 13 is the minimum of the degree of truth of (l and the degree of

truth of 13.
The same principle can be applied in three-valued logic, where the
truth-values in descending order are truth, neutrality and falsity. As
with negation, the restriction of the rule to the two classical truth-
values gives back the standard two-valued truth-table. For present pur-
poses, the most significant feature of the many-valued truth-table for
conjunction is that a conjunction of neutral sentences is itself neutral.
Any reasonable many-valued table for conjunction will have that ef-
fect, for the repetitive conjunction (l /\ (l should have exactly the same
truth-value as (l itself, in partiCUlar when (l is neutral, so by general-
ized truth-functionality the conjunction of any neutral sentence with
any neutral sentence is itself neutral.
Generalized truth-functionality makes degrees of truth work quite
differently from probabilities. For the probability of a conjunction is
not a function of the probabilities of its conjuncts. Suppose, for exam-
ple, that we are given that each of (l and 13 separately has a probability
of 0.5. What is the probability of (l /\ 13? Our data simply do not de-
termine any answer to the question. It is consistent with our data that 13
is equivalent to (l, in which case the probability of (l /\ 13 is the prob-
ability of (l /\ (l, which is 0.5; but it is equally consistent with our data
that 13 is equivalent to -xx, in which case the probability of (l /\ 13 is the
probability of (l /\ -xx, which is o. Not surprisingly, it turns out that
parallel considerations reveal a grave problem for a many-valued ac-
count of conjunction, and thus for many-valued logic more generally.
This is a version of what Kit Fine ([1975], pp.269-70) has called the
problem of penumbral connections (see also Kamp [1975], p.131,
Williamson [1994], pp.135-8 and Edgington [1996], pp.304-5).
Imagine a symmetrical building with two towers of equal height,
the east tower and the west tower. One false-sounding statement about
the building is this:
(i) The east tower is tall and the west tower is not tall.
For whatever the height of the towers, it is the same; the east tower is
no taller than the west tower. Now suppose that the east tower is a per-
fect borderline case for 'tall'. The grounds for classifying the east
tower as tall are no better and no worse than the grounds for classify-
ing it as not tall; we have no idea how to decide the question. On the
many-valued approach to vagueness, we must therefore evaluate the
sentence 'The east tower is tall' as neutral. Since the west tower is ex-
actly the same height as the west tower, we must evaluate the sentence
'The west tower is tall' as neutral for the same reason. Since the nega-
tion of a neutral sentence is itself neutral, we must also evaluate the
sentence 'The west tower is not tall' as neutral. But then (i) is the
conjunction of two neutral sentences, so (as noted above) the many-
valued approach forces us to evaluate (i) as neutral too. But that is to
misevaluate (i). It is not neutrally poised between truth and falsity;
rather, in the circumstances one should clearly prefer denying (i) (as-
serting its negation) to asserting it. Thus the many-valued approach
bungles the semantics of complex sentences. The objection does not
depend on the peculiarities of any particular many-valued system; it is
endemic to the approach, for the crucial assumptions were just gener-
alized truth-functionality and the possibility of neutrality.
Could a many-valued logician reply that (i) merely sounds false
because, by speaking asymmetrically of the height of the east tower
and the height of the west tower, the speaker generates a false conver-
sational implicature that they are different? No. For someone who de-
nies (i) speaks equally asymmetrically of the height of the east tower
and the height of the west tower, but generates no false conversational
implicature that they are different. The negation of (i) is wholly ac-
ceptable. What is wrong with (i) is that it semantically implies a differ-
ence in height between the two towers, not that there is a difference in
what it semantically implies about their height.
This kind of objection to many-valued logic is neither new nor
esoteric. Many-valued logicians have never answered it in any plausi-
ble way. It is therefore disappointing to find that, without even men-
tioning such criticisms, some philosophers continue to invoke many-
valued logic as though it provided an appropriate framework for an
account of vague identity. Yet we can easily check that the many-
valued mishandling of (i) extends to examples involving identity.
Suppose that I point to the west tower and say:
(ii) That tower = the west tower.
In this context, (ii) is straightforwardly true. One might suppose that

(ii) would suffice to falsify this claim:
(iii) That tower is tall and the west tower is not tall.
By (ii), only one tower is in question, and it is not taller than itself.
But, by reasoning similar to that for (i), since the tower is borderline
for 'tall', the many-valued approach counts (iii) as neutral, not false.
Again, that is a highly counterintuitive result.
Consider the inference from (ii) to the negation of (iii). Since the
negation of (iii) is also evaluated as neutral, the inference has a true
premise and a neutral conclusion. Consequently, it counts as invalid,
whether the condition for validity is that the truth of the premises guar-
antees the truth of the conclusion or that the conclusion is guaranteed
not to be lower in truth-value than every premise. This form of infer-
ence, closely related to Leibniz's Law, is therefore invalidated:
(AI) a =b
-(Fa!\ --,pb)
Yet the example presents no intuitive threat to (AI). The objects in
question are straightforwardly identical, both numerically and qualita-
tively, the property is an ordinary physical one and is attributed in an
ordinary way. The invalidation of (AI) indicates a failure of the ap-
proach, not of (AI).
Someone might try to validate (AI) by defining a valid inference
as one for which the non-falsity of the premises guarantees the non-
falsity of the conclusion, or as one for which the neutrality-or-better of
the premises guarantees the neutrality-or-better of the conclusion. But
those definitions invalidate a contraposed version of (AI):
(A2) Fa !\-.Fb
For under the same interpr,etation as before, (A2) has a neutral

premise and a false conclusion. Yet (A2) should be valid. It codifies
the canonical way in which we establish that objects are distinct.
Let us construct an example that turns on vague identity. Recall
that the names'S I' and 'S2' were introduced as names for the ship
present earlier and the ship present later respectively, without pre-
judging their identity or distinctness. We suppose that it is vague
whether SI and S2 are the same ship. Now make Hobbes' addition to
the story: someone collects the old timbers and, once he has them all,
puts them together in the original arrangement. Call that ship 'S3'. We
may suppose that it is equally vague whether S I and S3 are the same
ship. But it is quite clear that S2 and S3 are distinct ships; they are at
different places at the same time. Hence this holds:
(iv) S2 is not the same ship as S3.
One might suppose that (iv) would suffice to falsify this conjunction:
(v) SI is the same ship as S2 and SI is the same ship as S3.

For identity is a symmetric and transitive relation; SI cannot be identi-
cal with each of two things. Intuitively, (v) is not neutral between truth
and falsity; it is to be denied, not asserted. But the many-valued ap-
proach counts (v) as neutral, because it is a conjunction of neutral
conjuncts. That is yet another misevaluation. Since the negation of (v)
is also evaluated as neutral, the approach invalidates the argument
from (iv) to the negation of (v), whether validity requires that the truth
of the premises guarantees the truth of the conclusion or that the con-
clusion is guaranteed not to be lower in truth-value than every premise.
Consequently, this fonn of inference is invalidated:
(A3) -,b = c
-,(a = b /\ a = c)
Intuitively, the supposed counterexample is again not genuine: the
conclusion of (A3) is as true as the premise, but ineptly classified as
untrue by the many-valued approach.
If someone tried to validate (AI) by defining a valid inference as
one for which the non-falsity of the premises guarantees the non-falsity
of the conclusion, or as one for which the neutrality-or-better of the
premises guarantees the neutrality-or-better of the conclusion, the re-
sult would be to invalidate a contraposed version of (A3):
(A4) a = b /\ a = c
b=c
For under the same interpretation (A4) has a neutral premise and a
false conclusion. But the example provides no intuitive grounds for
denying the validity of (A4). Identity is a one-one relation.
Other cases make the same points about (A3) and (A4). Many
friends of vague identity hold that some ordinary objects have vague
boundaries, while other objects have precise ones. For example, it
might be vague whether the mountain Ml coincides in its boundaries
with the precisely bounded object M2 and equally vague whether Ml
coincides in its boundaries with the slightly different precisely
bounded object M3. While the identity claim 'M2 = M3' is clearly
false, some friends of vague identity regard both 'Ml = M2' and 'Ml
= M3' as neutral. The case has the same logical structure as that of the
ships.
These problems give us good reason to reject any account of vague
identity that uses many-valued logic on those grounds alone, independ-
ently of how it treats the Evans argument.
2.
A few more general methodological comments are in order. The usual

strategy for presenting a non-classical account of vague identity goes
something like this. One presents a rigorous, systematic, non-bivalent
model theory for a formal language some formula of which expresses a
claim of vague identity. One then constructs models of the required
kind in which the formula is true. The result is robust in the sense that
one obtains it even when one reasons classically in the metalanguage.
One then concludes that the informal claim of vague identity is at least
logically coherent, whether or not it is actually true, and therefore can-
not be refuted by a purely logical argument. In particular, one identifies
the step(s) in the Evans argument that can fail to preserve truth in a
model according to one's theory, and announces that the arguments are
fallacious or question-begging. Is this procedure legitimate?
A minimal requirement on the model theory is that it should be
faithful to the intended informal readings of the formal symbols. Sup-
pose that someone claims that it is logically coherent to assert contra-
dictions. To establish his point, he presents a theory in which the for-

mula a /\ 13 is true in a model if and only if a is true in the model or 13
is true in the model; -,a is still true in a model if and only if a is not
true in the model. He points out that, in his theory, a /\ -,a is true in
every model. Evidently, this procedure is worthless as a defence of the
claim that it is logically coherent to assert contradictions, even though
formally his model theory is as rigorous and systematic as you like. For
the theory is not faithful to the intended meaning of /\, conjunction,
treating /\ as though instead it meant disjunction. Similarly, suppose
that someone claims that it is logically coherent to assert that identical
objects differ in their properties. He presents a theory in which the
formula a = b is true in a model if and only if the denotation of a in the
model is distinct from the denotation of b in the model. He points out
that, in his theory, a = b and Fa are sometimes true in a model while
Fb is false in that model. Evidently, his procedure is worthless as a de-
fence of his claim, even though formally his model theory is as rigor-
ous and systematic as you like. For the theory is not faithful to the in-
tended meaning of =, identity, treating = as though instead it meant
distinctness.
We are naturally led to ask whether the non-classical model theo-
ries used in defence of vague identity are faithful to the intended
meaning of =, identity. The answer is that they are not, although the
divergence is of course less gross than in the examples just considered.
If = were interpreted faithfully, the truth-value of a = b in a model
would depend simply on the identity or otherwise of the denotation of
a in the model with the denotation of b in the model: the formula
would be true in the model if they were identical and false in the model
if they were not identical. But that is not what we find. Rather, the
model theory allows us to assign straightforwardly distinct objects as
the denotations of a and b respectively in a model, but still to treat the
formula a = b as half-true or at least not false in that model. For exam-
ple, in the three-valued models constructed by Peter van Inwagen and
by Terence Parsons and Peter Woodruff, a = b is treated as sometimes
neutral rather than false when the denotations of a and b are straight-
forwardly distinct (van Inwagen [1988], p.262, Parsons and Woodruff
[1995], p.180). Similarly, in the fuzzy models constructed by Graham
Priest, a = b is treated as sometimes having a degree of truth higher
than definite falsity (0) when the denotations of a and b are straight-
forwardly distinct (Priest [1998], p.333). These cases are especially
blatant because the authors have formulated their model theories with
care. Other attempts to defend the coherence of vague identity make
their unfaithfulness to the intended reading of = less blatant by de-
scribing the models more sketchily.
Such models may come close to the intended meaning of =, but
that is not good enough. A single model in which a single formula is
assigned a truth-value inconsistent with its intended reading is enough
to misclassify as invalid a principle that is valid on its intended read-
ing. Logic is an exact science; the model theory serves its purpose only
if it is exactly faithful to the intended reading of =. Otherwise, it con-
stitutes no objection at all to logical arguments against vague identity.
Of course, that is not yet to say that such arguments are valid.
The unfaithfulness to intended readings is not an accidental feature
of these non-classical models. Rather, it is essential to the robustness
of the results. The theorists want to prove that there are non-classical
models of vague identity. They fail to prove that conclusion if one can
prove, using classical assumptions in the metalanguage, that their non-
classical models are not models of vague identity - for they have not
shown that those classical assumptions are false. Given classical as-
sumptions in the metalanguage, a model is faithful to the intended
readings of the symbols only if it is in effect a classical model, so a
faithful model of vague identity is a classical model of vague identity.
But the theorists constructed non-classical models of vague identity
just because they conceded that it could not have classical models.
More precisely, on classical metalogical assumptions, either the deno-
tations of a and b in a model M are identical or they are not. If they are
identical and M is faithful to the intended reading of = then a = b is
true in M; if the denotations are not identical and M is faithful then a =
b is false in M. Either way, a = b is bivalent in M. To avoid this result,
the non-classical models treat = unfaithfully.
The preceding argument assumes that the singular terms a and b
have unique denotations in the model. Of course we may expect
something more complicated if the model treats a and b as vague. But
vagueness in the singular terms is not our topic, for it does not consti-
tute vagueness in identity. The question is whether it is vague of some
objects whether they are identical. Vagueness in the sentence a = b an-

swers that question only if the terms a and b themselves are precise. It
is therefore legitimate to assume that they have unique denotations in
the model (compare Lewis [1988] on the original intention of the Ev-
ans argument; contrast Thomason [1982]).
Even an unfaithful model can establish a consistency result, by
showing that some formula does not have every formula as a conse-
quence in some logic. In that sense it is perfectly consistent to reason
with /\ according to the usual rules for disjunction. That sense is far too
weak to be of interest here. Trivially, in some logic a formula that can
express a claim of vague identity does not have every formula as a
consequence, if the logic is not constrained by that reading of the sym-
bols.
Those remarks are not intended to show by themselves that there
can be no faithful model of indeterminate identity, for they do not
show that the metalogic must be classical. Someone might present a
model with classical semantic clauses for the logical constants, in-
cluding =, and by non-classical reasoning in the metalanguage still
avoid conceding that a = b is either true or false in the model (compare
Heck [1998], pp.292-93). But that move would be as controversial as
their non-classical metalogic. The coherence of vague identity would
not have been established.
3.
The standard defences of vague identity fail for reasons independent of

the Evans argument. But what has that argument to teach us?
In order not to be distracted by vagueness in singular terms, we
can formulate the reasoning with objectual variables rather than proper
names flanking the identity sign =. The variables denote objects rela-
tive to assignments. Such an assignment is simply a function from
variables to objects; whether anyone ever specifies the function in
thought or speech, vaguely or precisely, is irrelevant. A version of the
Evans argument with variables, if valid, preserves truth relative to any
assignment.
Let the sentence operator Ll abbreviate 'it is determinate that',
rather than 'it is determinate whether'. As several commentators have
Timothy Williamson, Vagueness, Identity, and Leibniz 's Law 285
noted, Evans's original paper involves an easily corrected confusion

between these two notions (Heck [1998], pp.276-8). 'It is determinate
that P', unlike 'It is determinate whether P', entails 'P'. 'It is determi-
nate whether P' is equivalent to 'Either it is determinate that P or it is
determinate that not P'; 'It is determinate that P' is equivalent to 'P
and it is determinate whether P'. In these structural respects, 'determi-
nate' behaves just like 'non-contingent'. Since 'it is non-contingent
that' is equivalent to 'it is necessary that', the logical behaviour of ~ is
similar to that of the modal operator D. The formula --.~ x = Y /\ --.~--. X
= Y expresses the claim that it is indeterminate whether the objects in
question are identical.
The Evans argument as we are now formulating it involves the in-
ference from ~ x = x and --.~ x = y to --,x = y. This is a contraposed ap-
plication of an inference by Leibniz's Law from x = y and ~ x = x to
~ x = y. As one might expect, much of the controversy about the argu-
ment concerns the legitimacy of applying Leibniz's Law to formulas
containing operators such as ~.
Many different versions of Leibniz's Law have been proposed.
What seems clear is that, properly formulated, the law embodies an
insight absolutely fundamental to our understanding of the logical no-
tion of identity. If x and yare the very same thing, then whatever ap-
plies to x thereby applies to y. To suppose that there are exceptions to
the best statement of the law is to lose one's grip on the topic. On
some canonical formulation, rigorous and free of ad hoc qualifications,
the law is valid. But finding that formulation may not be easy.
Terence Parsons [1987] has emphasized that the Evans argument
contraposes Leibniz's Law and that, in many-valued logic, the validity
of a principle does not guarantee the validity of the corresponding
contraposed principle. For suppose that validity is truth-preservation
and that sentences can be neutral, neither true nor false. If the argument
from a and r:I to y is valid, then y is true whenever a and r:I are; but that
allows y to be false when a is neutral and r:I is true, in which case the
standard many-valued treatment of negation makes --.y true and --.0.
neutral, so the argument from r:I and --.y to --.0. has true premises and a
neutral conclusion, and is therefore invalid. In particular, the validity
of the argument from x = y and ~ x = x to ~ x = y would not guarantee
the validity of that from ~ x = x and --.~ x = y to --,x = y. We have al-
ready seen sufficient reason to reject a many-valued approach to vague

identity. But there is also something more general to say about the
contraposition of Leibniz's Law. Many of the law's most important
applications are to establish conclusions of non-identity. Imagine that
you are trying to work out whether this ruined city is Troy. You must
be able to reach either a positive or a negative answer, depending on
your evidence. How will you reach a negative answer except by find-
ing something that applies to this ruined city and not to Troy? If the
contraposition of the best statement of Leibniz's Law were invalid,
there would be an obstacle of principle to ever resolving identity ques-
tions in the negative. That consequence is scarcely a serious option.
We should therefore assume that on some canonical formulation, rig-
orous and free of ad hoc qualifications, even the contraposition of
Leibniz's Law is valid. But, again, finding that formulation may not be
easy.
The obvious rule to take one from x = y and d x = x to d x = y is a
substitution principle: if t and t* are singular terms, and the formula a*
differs from the formula a at most in having occurrences of t* in some
places where a has occurrences of t (no variable in t or t* being bound
in any of the relevant occurrences in a or a*), then from t = t* and a
one may infer a*. Unfortunately, such a substitution principle is sub-
ject to counterexamples in many languages. Most obviously, there are
quotational contexts; 'Hesperus = Phosphorus' and "Hesperus' begins
with an 'H" are true but "Phosphorus' begins with an 'H" is false.
That there are also modal, temporal and intentional counterexamples is
not as clear as was once assumed. If anything like Russell's theory of
descriptions is correct, definite descriptions are not singular terms and
therefore cannot be substituted for I or t* in the substitution principle
to produce counterexamples in such contexts. Accounts of proper
names and indexicals as directly referential undermine other purported
counterexamples. Nevertheless, the substitution principle clearly fails
in contexts that are not straightforwardly quotational. For instance, one
can define a sentence operator 'distinctly' by stipulating that the result
of prefixing it to a sentence s is equivalent to the conjunction of s with
every sentence of the form 11 "* h, where 11 and h are distinct singular
terms occurring (unbound) in s. Thus, although 'Leningrad' and 'St
Petersburg' are names of the same city, 'Distinctly Leningrad is no
larger than St Petersburg' is equivalent to 'Leningrad is no larger than

St Petersburg and Leningrad *- St Petersburg', which is false, while
'Distinctly Leningrad is no larger than Leningrad' is equivalent simply
to 'Leningrad is no larger than Leningrad', which is true. Conse-
quently, the inference from 'Leningrad = St Petersburg' and 'Distinctly
Leningrad is no larger than Leningrad' to 'Distinctly Leningrad is no
larger than St Petersburg' is invalid, because it has true premises and a
false conclusion. Thus sentences containing 'distinctly' invalidate the
substitution principle, even though 'distinctly' does not create a
straightforwardly quotational context.
Evidently, a substitution principle stated for all sentences of an ar-
bitrary object-language is at the mercy of whatever idiosyncratic se-
mantic devices that language happens to contain. Attempts to add a
qualifying clause to the rule either look ad hoc or risk making its ap-
plication circular; one may be unable to verify that the clause is satis-
fied without assuming the very conclusion the rule is designed to
reach. It is therefore tempting to suppose that Leibniz' s Law should be
canonically stated in metaphysical rather than metalinguistic terms: if x
and yare identical, then every property of x is a property of y and vice
versa. A terminological variation, not adopted here, would be to re-
serve the name 'Leibniz's Law' for the substitution principle, and call
the metaphysical principle something else (Richard [1987]). The meta-
physical version of Leibniz's Law has a quick way with purported
counterexamples: whenever the substitution principle fails, one simply
denies that the relevant formula defines a property. In particular, the
relevant version of the Evans argument assumes that for every assign-
ment A there is a property P ('being determinately identical with x')
such that Ll x = x is true relative to A if and only if the object that A
assigns to x has P and Ll x = y is true relative to A if and only if the
object that A assigns to y has P (P depends on A because it depends on
what object A assigns to x). Evans makes this assumption explicit by
using a device that abstracts a term for a property from an open for-
mula. His argument involves steps of property abstraction and concre-
tion to introduce and eliminate the abstracted term. Predictably
enough, these steps have been challenged. The details of the challenges
do not concern us here, for they exploit semantic assumptions (such as
those of many-valued logic) that have already been rejected. Rather, let
us ask whether the argument really needs to be formulated with refer-

ence to properties.
Consider this argument:
(A5) Brutus stabbed Caesar.
Caesar did not stab Caesar.
Brutus"* Caesar.
Surely (A5) is valid. But imagine a sceptic who raises doubts along
these lines:
Is there really a property P of having stabbed Caesar such that 'Brutus
stabbed Caesar' is true if and only if Brutus has P and 'Caesar did not stab
Caesar' is true if and only if Caesar does not have P? It would be a rela-
tional property; can we be sure that there really are relational properties?
Furthermore, the notion of stabbing is not used in fundamental natural sci-
ence; do non-fundamental notions really stand for properties?
These doubts play on serious issues about the metaphysics of proper-

ties. Nevertheless, (A5) does not seem to depend on their positive
resolution. Whatever the ultimate metaphysics of properties, (A5) is
valid. Thus if a notion of property is to figure in a canonical formula-
tion of Leibniz's Law, it should not be a metaphysically rich notion,
but instead a minimal, pleonastic one. For an object 0 to have a prop-
erty in this sense is simply for something to be true of o. The corre-
sponding version of Leibniz's Law will therefore say something like
this: if 0 and 0* are identical, then whatever is true of 0 is true o~ 0*.
Contrapositively, if something is true of 0 that is not true of 0*, then 0
and 0* are distinct. But this enables us to see how to apply Leibniz's
Law metalinguistically in a way that is not subject to the counterexam-
ples to the original substitution principle.
Suppose that A and A * are identical assignments of objects to
variables, and that a sentence a. is true relative to A; it follows that a. is
true relative to A *. Contrapositively, if a. is true relative to A and not
true relative to A *, it follows that A and A * are distinct assignments.
Now suppose that A and A * differ at most (if at all) in that A assigns
an object 0 to some variable v while A * assigns an object 0* to v. If 0
and 0* are identical, then A and A * are identical, so if a. is true relative
to A, it follows that a. is true relative to A *. Contrapositively, if a. is

true relative to A and not true relative to A *, it follows that 0 and 0*
are distinct, because A and A * are distinct. To be quite explicit, for
future reference:
(CLL) Let an assignment A assign an object 0 to a variable v, an

assignment A* assign an object 0* to v, and A* be exactly
like A in every other way. Suppose that a sentence a. is true
relative to A and not true relative to A *. Then 0 and 0* are
not identical.
The foregoing remarks hold whatever modal, temporal, intentional or

quotational contexts a. may contain, for no singular terms are being
substituted within a. itself. Terms for assignments are being substituted
in the metalinguistic context 'a. is true relative to .. .', but that context
is straightforwardly extensional. Its role is merely to generalize the
context 'a. is true of .. .' to the case in which a. contains many variables,
and we have just noted that substitution in the 'true of .. .' context is
central to a proper grasp of Leibniz's Law. As for the assumption that
if A and A * differ at most in that A assigns 0 to v while A * assigns 0*
then A and A * are identical if and only if 0 and 0* are identical, it is
simply a consequence of the extensional nature of functions.
The claim is not that (CLL) or the uncontraposed variant is the ca-
nonical version of Leibniz's Law. (CLL) is too complex and special-
ized for that. Rather, (CLL) is the result of a particularly unproblematic
application of Leibniz's Law, whatever its canonical version may be.
(CLL) is really just a special case of a trivial theorem of classical
mathematics:
Let f and f* be functions on a domain D such that for some deD,

f(e) = f*(e) whenever d"# e. Suppose that some object x has R to f
but not to f*. Then f(d)"# f*(d).
This theorem is not specially about truth or semantics. In (CLL), the

domain D becomes the set of variables, the relation R becomes truth
relative to an assignment, and f(d) and f*(d) are called 0 and 0* re-
spectively. If we prefer, we can formulate both (CLL) and the theorem
in classically equivalent uncontraposed forms.
Of course, some people claim that vagueness shows that not all
trivial theorems of classical mathematics are true. That claim is usually
premised on many-valued logic, which has already been rejected. Of
course, it might have some other premise. Nevertheless, the aim in the
remainder of this paper will be to explore issues of vague identity un-
der the assumption that classical reasoning may be used in the meta-
language.
Let us apply (CLL) to the claim that it is sometimes neither true
nor false to say of given objects that they are identical. Suppose that
the assignment A assigns the object 0 to the variable x and the object
0* to the variable y. If it is neither true nor false to say of 0 and 0* that
they are identical, then the formula x = y is neither true nor false rela-
tive to A, when = means identity. Let A * be the assignment that differs
from A at most in assigning 0 to y; since A * assigns 0 to x too, x = y is
true relative to A *. Since x = y is by hypothesis not true relative to A,
by (CLL) 0 and 0* are not identical. But then if= really means identity,
x = y should be false relative to A, contrary to hypothesis. Thus the
supposition that it is neither true nor false to say of 0 and 0* that they
are identical is incoherent.
Someone might object that in a context in which we are contem-
plating truth-value gaps for identity statements with uniquely denoting
terms, 'not true' in (CLL) must be replaced by 'false':
(CLL-) Let an assignment A assign an object 0 to a variable v, an
assignment A * assign an object 0* to v, and A * be exactly
like A in every other way. Suppose that a sentence ex is
true relative to A and false relative to A *. Then 0 and 0*
are not identical.
But this modification gets things back to front. Entertain for a moment
the dialetheist idea that a sentence can be both true and false relative to
the very same assignment A#. Let A# assign the object 0# to a variable
v. When we take both A and A* to be A#, and both 0 and 0* to be 0#,
(CLL-) delivers the absurd result that 0# is distinct from itself. If one
accepts (CLL-), one does so on the basis of the anti-dialetheist as-
sumption that if ex is false relative to A * then it is not also true relative
to A *. That is, the reason for accepting (CLL-) is that it follows by the
mutual exclusiveness of truth and falsity from (CLL), which is a spe-
Timothy Williamson, Vagueness, identity, and Leibniz 's Law 291
cial case of a theorem of classical mathematics. (CLL-) is not a special

case of a theorem of classical mathematics; unlike (CLL), it depends
on something specifically semantic, the mutual exclusiveness of truth
and falsity. To endorse (CLL-) without (CLL) is confused. Further-
more, no good reason has been provided to reject (CLL). Of course,
someone who clings on at all costs to the claim that identity statements
with uniquely denoting terms can have truth-value gaps might me-
chanically reject (CLL) simply on the grounds that it is the key premise
of an argument against the claim. Any error can be protected like that.
The point is that nothing has been done to explain how (CLL) could
fail.
We should not take it for granted that indeterminate identity is to
be understood in terms of truth-value gaps. Once we have ~ in the ob-
ject-language, we can apply (CLL) to the earlier formulation -,~ x = Y
/\ -,~--, X = y. Suppose that it is indeterminate whether the objects 0
and 0* are identical. Let A be an assignment that assigns 0 to the vari-
able x and 0* to the variable y. By hypothesis, the formula -,~ x = y is
true relative to A. Let A * be the assignment that differs from A at most
in assigning 0 to y. Since A * also assigns 0 to x, --,~ x = y is not true
relative to A *. Consequently, by (CLL), 0 and 0* are not identical. The
result would be the same even if one used (CLL-) rather than (CLL)
here, for --,~ x = y is false as well as not true relative to A *. Thus it is
indeterminate whether objects are identical only if they are not identi-
cal. Although this is of course a variation on the original Evans argu-
ment, it uses no substitution principle for the object-language.
It would be misconceived to object that the notion of truth relative
to an assignment of objects to variables in the scope of the determinacy
operator ~ is problematic. For unless we can apply that notion to the
formula -,~ x = y, we cannot locate the indeterminacy in the relation
between the objects themselves. The application is licensed by the hy-
pothesis of indeterminate identity itself (see also Heck [1998], pp.278-
81).
What exactly has the argument shown? If it is indeterminate
whether 0 and 0* are identical, then 0 and 0* are distinct. Is this a re-
ductio ad absurdum of the supposition that it is indeterminate whether
o and 0* are identical? Not straightforwardly. In object-language
terms, -,~ x = y entails --, x = y. But --, x = y is not immediately inc on-
sistent with -,~ X = Y /\ -,~-, X = y. What is wrong with the hypothesis

that although it is indeterminate whether 0 and 0* are identical, they
are in fact not identical? The flavour of absurdity comes from the idea
that, once we have worked out that 0 and 0* are not identical, it fol-
lows that it is determinate that they are not identical. If -, X = Y entails
~-, x = y then -,~ x = Y /\ -,~-, X = Y is inconsistent, because -,~ x = y
entails -, x = y. The question is whether to accept the form of inference
from a to ~a.
Recall the kind of situation which the notion of indeterminacy was
introduced to describe. Supposedly, the grounds for asserting some-
thing and the grounds for asserting its negation are evenly balanced,
and there is no room for further considerations to decide the matter. If
the relation between 0 and 0* conforms to this pattern, then the
grounds for asserting that they are identical and the grounds for as-
serting that they are not identical should be evenly balanced, and there
should be no room for further considerations to decide the matter. But,
if it is indeterminate whether they are identical, then the argument
above does decide the matter, in favour of asserting that they are not
identical. The grounds are after all not evenly balanced. Thus the hy-
pothesis that it is indeterminate whether they are identical looks incon-
sistent with what indeterminacy is supposed to be.
As we might expect, the argument leaves it open that for some re-
lation of coincidence, similar but not strictly equivalent to identity,
certain objects neither determinately coincide nor determinately fail to
coincide (Akiba [2000] elaborates such a possibility within a quasi-
supervaluationist framework). But vague coincidence is not vague
identity.
A more formal point can be developed. Those who defend the ver-
dict of indeterminacy in some particular case of 0 and 0* presumably
hold that it is determinate that it is indeterminate whether 0 and 0* are
identical. In symbols, they defend il(-,il x =Y /\ -,~-, X = y) relative to
an appropriate assignment.
We can formulate two plausible principles about the logic of ~,
using the symbol ~ for semantic consequence:
(RK) If aI. ... , an ~ 13 then ~aI. ... , ~an ~ ~13·

(RT) /).a Fa.
(RK.) says that determinacy is closed under semantic consequence; if

there are decisive grounds for the premises of a semantically valid ar-
gument, they combine with that semantic connection to constitute de-
cisive grounds for its conclusion. We allow n to take the value 0, in
which case (RK) boils down to the 'determination' rule that F13 im-
plies F/).13, but only the case when n takes the value 1 is used in what
follows. (RT) is the uncontentious principle that whatever is determi-
nately so is so.
Granted the semantic relation ,/). x = Y F'
x = y already argued
for by appeal to (CLL), we can use (RK) and (RT) to argue by reductio
ad absurdum against the supposition /).(,/). x = y /\ ,/)., X = y) of de-
terminately indeterminate identity, thus:
(1),/).x=yF -,x=y Given

(2) /).,/). x = Y F /)., x =Y 1 (RK)
(3) ,/). x = Y /\ ,/)., X = Y F ,/). x = Y PC
(4) /).(,/). x = Y /\ ,/).-, X = y) F /).,/). x = Y 3 (RK)
H /).,
(5) /).(,/). x = Y /\ ,/).-, X = Y x =Y 4,2
(6) /).(,f!..x=y /\ ,/).-,x=y) F -,f!..x=y /\ ,/).-,x=y (RT)
(7) /).(,/). x = Y /\ ,/).-, X = yH ,/).-, x = y 6PC
(8) /).(,f!..x=y /\ ,/).-,x=y) F /).-,x=y /\ ,/).-,x=y 5,7PC
(9) F ,/).(,/). x = y/\ ,/)., X = y) 8RAA
Thus /).(,/).x = y /\ ,/)., X = y) is an inconsistent hypothesis (see also

McGee and McLaughlin [1995], p.232). It cannot be determinate that it
is indeterminate whether objects are identical.
For some but not all supervaluationists the discharge of a premise
in the use of reductio ad absurdum at line 9 would be problematic
(Williamson [1994], p.151-2). However, they could reformulate the
argument in terms of the unconditional semantic validity of conditional
statements; for example, line 1 would become F ,/). x = y ---+ , x = y.
Under corresponding reformulations of (RK.) and (RT), the revised ar-
gument would be acceptable to those supervaluationists too, and its
bottom line would still be 9. Anyway, line 8 is already bad news for
determinately indeterminate identity.
Evans envisaged a related argument that would reduce the original
indeterminacy hypothesis -,d x = y /\ -,d-, x = y itself to absurdity

within an analogue of the strong modal system S5. In addition to (RK)
and (RT) , such a system would have the S5 rule -,Lla F Ll-,da, and
therefore in particular -,d x = y F Ll-,d x = y. That would combine
with line 2 above to yield -,Ll x = y FLl-, x = y and therefore -,Ll x = Y
/\ -,Ll-, x = y F d-, x = Y /\ -,Ll-, x = y, which gives F -, (-,d x = y /\
-,d-, x = y). The hypothesis that it is indeterminate whether objects are
identical is incoherent. However, the S5 rule is suspiciously strong. In
combination with (RK), (RT) and classical logic, it generates the con-
clusion F LlLla v Ll-,da, which excludes indeterminacy in da. But
any case for first-order indeterminacy corresponds to a case for such
second-order indeterminacy. If it is indeterminate when there ceased to
be a heap, it is equally indeterminate when it ceased to be indetermi-
nate when there ceased to be a heap. Consequently, we should not rely
on S5 reasoning.
We should however note that -,Ll x = Y /\ -,Ll-, x = y can be re-
duced to absurdity on assumptions weaker than the S5 rule. For exam-
ple, the so-called Brouwerian rule -,a FLl-,Lla is quite consistent with
higher-order vagueness (Williamson [1999] has more discussion); one
instance is -, x = y F Ll-,Ll x = y. But we have -,d x = y F -, x = y by
assumption, so transitivity gives -,d x = y F d-,Ll x = y, exactly the
special case of the S5 rule that excluded indeterminate identity. Other
rules short of S5 have a similar effect (McGee and McLaughlin [1995],
p.250). But whether the Brouwerian rule or another of those rules
holds on a reading of d as 'determinately' is far from clear (see also
Williamson [1996], p.l2-13). Let us therefore concentrate on the in-
consistency of d(-,d x = y /\ -,d-, x = y), which does not depend on
such controversial reasoning.
Even if for no objects is it determinate that it is indeterminate
whether they are identical, it does not immediately follow that it is not
determinate that for some objects it is indeterminate whether they are
identical, for Ll3x3y (-,Ll x = Y /\ -,Ll-, x = y) /\ -,3x3yLl (-,d x = y /\
-,d-, x = y) is not an immediately inconsistent combination. Perhaps it
is determinate that there are cases of indeterminate identity even
though there are no determinate cases of indeterminate identity. Nev-
ertheless, on a non-epistemic understanding of indeterminacy, the im-
possibility of determinate cases of indeterminate identity does seem to
detract from the reality of indeterminate identity. Consider a formally

similar case. According to supervaluationists, it is determinate that
some number n is a sharp cut-off point for baldness, in the sense that
having just n hairs on one's head implies baldness but having just n+ 1
hairs on one's head does not (~3n (Bn /\ ...Bn+1», for each sharpening
of 'bald' provides such a cut-off point. However, supervaluationists
regard this as somehow not a serious commitment to a cut-off point,
for they also hold that for no number n is it determinate that n is a cut-
off point for baldness (-,3n~ (Bn /\ -,Bn+1»; the cut-off point varies
across sharpenings. If -,3n~ (Bn /\ -,Bn+ 1) undermines the apparent
commitment in ~3n (Bn /\ -,Bn+ 1) to a sharp cut-off point, then by
parity of reasoning -,3x3y~ (-,~ x = Y /\ -,~-, X = y) should undermine
the apparent commitment in ~3x3y (-,~ x = Y /\ -,~-, X = y) to an in-
determinate identity.
The previous result that any case of indeterminate identity is a case
of non-identity makes the supposedly non-epistemic nature of the in-
determinacy quite hard to grasp (for the general difficulty of elucidat-
ing a non-epistemic conception of indeterminacy see also Williamson
[1995], [1997], pp.216-17). On an epistemicist conception, we are ir-
remediably ignorant in borderline cases. Since there is a truth of the
matter, however unknowable, vagueness does not invalidate classical
logic. Let us explore questions of vague identity within such an epis-
temicist framework.
4.
The Evans argument, once modified to rely on (CLL), involves as-

signments of objects to the variables in the formula ~ x = y, that is, on
an epistemicist reading, to variables within the scope of an epistemic
operator. Such assignments are notoriously problematic. If John knows
that Hesperus is Hesperus but does not know that Hesperus is Phos-
phorus, what is the truth-value of the formula' John knows that x is y'
when that planet is assigned to both variables? Not everyone accepts
that the substitution of coreferring singular terms in propositional atti-
tude contexts does affect truth-value (Salmon [1986]). But since it
certainly appears to, we must proceed with great caution.
We can express the epistemicist reading of ~ by using 'clear' in
place of 'determinate'. The simple reading of ~ as 'it is knowable' is
inadequate, for it would make the formula ,~ex. 1\ ,~,ex. true when-
ever the truth-value of the proposition expressed by ex. could not be
known; but our current interest is not in cases in which the source of
the unknowability has nothing to do with vagueness (undecidable
propositions of mathematics may provide examples; see Williamson
[1995]). The intended epistemicist reading of ~ prescinds from such
irrelevant obstacles to knowledge. The relevant obstacle is, very
roughly, that a vague sentence might express a true proposition on its
actual interpretation I while expressing a false proposition on some
close counterfactual interpretation 1*, where speakers' insensitivity to
the difference between I and 1* makes any belief they express by ut-
tering the sentence too unreliable to constitute knowledge (Williamson
[1994], pp.230-4 has more details). In such circumstances we may de-
scribe 1* as indiscriminable from I. The natural proposal is therefore
that ~ex. is true on an interpretation I if ex. is true on every interpretation
1* indiscriminable from l On such a semantics, what kind of context
does ~ create? We can answer the question by examining the simplest
interesting case.
Consider a propositional language with just two singular terms, the
constants a and b, in which the only atomic sentences are a = a, a = b,
b = a and b = b. There are just three primitive sentence operators, ~, '
and 1\. Other operators are introduced as metalinguistic abbreviations.
A model is an ordered pair of a class of interpretations and a reflexive
symmetric relation of indiscriminability defined over them. A sentence
is valid in a model if and only if it is true on every interpretation in the
model. For most purposes reference to the model can be left tacit. An
interpretation I assigns a denotation to each constant. The sentence 11 =
12 is true on I if and only if I assigns the same denotation to the con-
stants 11 and {z. As usual, ,ex. is true on I if and only if ex. is not true on
I, and ex. 1\ ~ is true on I if and only if ex. is true on I and ~ is true on l
Finally, ~ex. is true on I if and only if ex. is true on every interpretation
indiscriminable from l
On this semantics, = is treated as a logical constant; it is always
interpreted as identity, just as , and 1\ are always interpreted as nega-
tion and conjunction respectively. The logical constants are treated as
precise. In the case of identity, this assumption has been challenged
(Hirsch [1999]). However, identity is a logically salient relation, the
narrowest necessarily reflexive relation. It is implausible that an unno-

ticeable shift in our use of = would make it refer to some other rela-
tion. The assumption is therefore reasonable.
Consider a model with just two interpretations, 1\ and 12 , each in-
discriminable from the other. 1\ assigns a and b the same denotation; 12
assigns them different denotations. Since a = b is true on I\, -,L1-, a = b
is valid; since a = b is not true on 12 , -,L1 a = b is valid. Since a = a and
b = b are valid, so are L1 a = a and L1 b = b. In this model, indiscrimi-
nability is transitive as well as reflexive and symmetric; consequently,
L1 has an S5-like logic. In particular, all sentences of the forms
-,L1a ~ L1-,L1a, L1a ~ a and (L1a /\ L1(a ~ f3» ~ L1f3 are valid, and
whenever a is valid, so is L1a. We could easily complicate the model to
invalidate S5 while retaining the features of current interest.
On this semantics, L1 creates an opaque context, in the sense that
not every instance of the substitution principle t = t* ~ (a ~ a*) is
valid in every model, where a * differs from a at most in having some
occurrences of the constant t* where a has occurrences of t. For exam-
ple, a = b and ~ a = a are true on 1\ but ~ a = b is not. Consequently,
we should not regard a sentence of the form L1 t\ = t2 on an interpreta-
tion I as expressing a proposition about the objects that t\ and t2 do in
fact denote on I to the effect just that they are clearly identical. For
otherwise L1 a = a and L1 a = b would express the same proposition on
I\, and would therefore have the same truth-value. Thus -,L1 a = b /\
-,L1-, a = b on a given interpretation does not express a proposition
about the objects that a and b do in fact denote on that interpretation to
the effect just that it is unclear whether they are identical. It would be
less inaccurate to regard L1a as quoting a, and -,L1 a = b /\ -,L1-, a = b
in particular as expressing a proposition about the names a and b
themselves to the effect that it is unclear whether they denote the same
object. That is vagueness in language, not in the objects to which it re-
fers.
We can introduce variables and quantifiers into the language with
an auxiliary notion of truth on an interpretation relativized to assign-
ments of values to variables. These assignments should be independent
of the interpretations. For otherwise a single assignment A could as-
sign to the variable n relative to each interpretation I the cut-off point
for 'bald' on I; thus Bn /\ .Bn+l would be true on every interpretation
relative to A, so ~ (Bn /\ -J3n+ I) would be true relative to A, so

3n~ (Bn /\ -J3n+ I) would be true. Intuitively, that is the wrong result,
for the formula says that some number is the clear cut-off point for
baldness. Thus assignments should be independent of interpretations.
But then relative to any given assignment either x = y is true on every
interpretation or -, x = y is true on every interpretation, depending on
the identity or non-identity of the objects assigned to the variables x
and y by that assignment. Thus relative to any assignment either ~ x =
y is true or ~-, x = y is true. Consequently, the formula \:Ix\:ly (~x = y
v ~-, x = y) is valid. It does not follow that the schema ~ t] = 12 v ~-, t]
= t2 is valid for closed singular terms, which it certainly is not; since
closed singular terms can be vague in a way in which variables cannot
be, they cannot be substituted for variables within the scope of ~ to
instantiate a universal geneneralization, just as nonrigid designators
cannot be substituted for variables in modal contexts.
Thus epistemicism validates the natural formalization of the claim
that it is never vague whether objects are identical: -,3x3y (-,~ x = Y /\
-,~-, X = y). In that sense, it excludes vague identity (Williamson
forthcoming has more details). That is hardly surprising. Nevertheless,
we want more. For some identity statements constitute borderline cases
while others do not: the relevant singular terms refer to objects. Surely
we can say something about the objects themselves on that basis.
We can say more by exploiting the expressive resources of the
metalanguage. Given epistemicism, one interpretation is correct; it
maps expressions to their actual referents. We can therefore use a no-
tion of absolute truth, truth on the correct interpretation. We can then
define some binary relations between objects 0] and 02 thus:
0] and 02 are presentable as clearly identical <=>
~ t] = t2 is true for some constants I] and 12 denoting 0] and 02 re-
spectively.
0] and 02 are presentable as clearly distinct <=>
~-, I] = t2 is true for some constants I] and 12 denoting 0] and 02 re-
spectively.
0] and 02 are presentable as either clearly identical or clearly dis-
tinct<=>
~ 1\ = 12 V~--, 1\ = /z is true for some constants 1\ and 12 denoting 0\

and 02 respectively.
0\ and 02 are presentable as not clearly identical ~
--,~ t\ = 12 is true for some constants t\ and t2 denoting 0\ and 02 re-
spectively.
0\ and 02 are presentable as not clearly distinct ~
--,~--, t\ = 12is true for some constants 1\ and tz denoting 0\ and 02
respectively.
0\ and 02 are presentable as neither clearly identical nor clearly
distinct ~
--,~ 1\ = 12 /\ --,~--, 1\ = 12 is true for some constants 1\ and 12 denot-
ing 0\ and 02 respectively.
These equivalences define genuine relations; in these contexts the sub-
stitution of coreferring terms for 0\ and 02 is legitimate. We could de-
fine corresponding ternary relations by relativizing the binary relations
to interpretations in the obvious way.
Here are some elementary logical features of the relations. Since
indiscriminability is reflexive, t\ = t2 is true whenever ~ t\ = /z is, so
objects are presentable as clearly identical only if they are indeed iden-
tical. Similarly, since --, 1\ = 12 is true whenever ~--, t\ = 12 is, objects are
presentable as clearly distinct only if they are indeed distinct. Obvi-
ously, objects are presentable as either clearly identical or clearly dis-
tinct if and only if they are either presentable as clearly identical or
presentable as clearly distinct. Less obviously, they are presentable as
neither clearly identical nor clearly distinct if and only if they are both
presentable as not clearly identical and presentable as not clearly dis-
tinct. One half is obvious: if they are presentable as neither clearly
identical nor clearly distinct then they are both presentable as not
clearly identical and presentable as not clearly distinct. For the con-
verse, suppose that 0\ and 02 are both presentable as not clearly identi-
cal and presentable as not clearly distinct. There are two cases. First,
suppose that 0\ and 02 are identical. Since they are presentable as not
clearly identical, --,~ 1\ = 12 is true for some constants 1\ and tz denoting
0\ and 02 respectively. But since 0\ and 02 are identical, --,~--, 1\ = 12 is
also true. Thus --,~ t\ = t2 /\ --,~--, 1\ = 12 is true, so 0\ and 02 are present-
able as neither clearly identical nor clearly distinct. Similarly, suppose

that 01 and 02 are distinct. Since they are presentable as not clearly dis-
tinct, ---,~---, II = 12 is true for some constants II and 12 denoting 01 and 02
respectively. But since 01 and 02 are distinct, ---,~ II = tz is also true.
Thus ---,~ II = 12 /\ ---,~---, II = 12 is true, so 01 and 02 are again presentable
as neither clearly identical nor clearly distinct. QED.
If tl = t2 is any borderline identity statement, and II and 12 denote
the objects 01 and 02 respectively, then 01 and 02 are ipso facIo present-
able as neither clearly identical nor clearly distinct. Of course, the rea-
son may be that II or 12 happens to be vague. For, in virtue of more
precise names, 01 and 02 may also be presentable as either clearly
identical or clearly distinct. Indeed, if they are identical, then they are
presentable as neither clearly identical nor clearly identical only if they
are also presentable as clearly identical. For by hypothesis some con-
stant II denotes 01; but ~ II = II is true, and II denotes both 01 and 02
(this point is of course related to the Evans argument). Even if 01 and
02 are distinct, they may be both presentable as neither clearly identical
nor clearly distinct and presentable as clearly distinct. For example, fix
the reference of the name s by stipulating that it is to denote the highest
small natural number. Suppose that in fact s denotes 17, although for
all we can know it denotes another number; in particular, on some in-
terpretations indiscriminable from the right one it denotes 16. But the
numerals '16' and '17' precisely denote 16 and 17. Since ~ 17 = 17 is
true, 17 is presentable as clearly identical with 17 (that is, 17 and 17
are presentable as clearly identical). Since ---,~ s = 17 /\ ---,~---, s = 17 is
true, 17 is also presentable as neither clearly identical nor clearly dis-
tinct from 17 (that is, 17 and 17 are presentable as neither clearly iden-
tical nor clearly distinct). Since ~---, 16 = 17 is true, 16 and 17 are pre-
sentable as clearly distinct. Since ---,~ s = 16 /\ ---,~---, s = 16 is true, 16
and 17 are also presentable as neither clearly identical nor clearly dis-
tinct. Thus presentability as neither clearly identical nor clearly dis-
tinct, although a genuine relation, is a rather trivial one. It gives little
information about its relata.
A more revealing relation is non-presentability as either clearly
identical or clearly distinct. There are two cases in which 01 and 02 fail
to be presentable as either clearly identical or clearly distinct. In the
first case, no constants denote 01 and 02 respectively. Thus 01 and 02
Timothy Williamson, Vagueness, identity, and Leibniz's Law 301
will not be presentable as neither clearly identical nor clearly distinct.

The issues raised by nameless objects will not be discussed here; al-
though significant, they would take us out of our way. They arise in
acute form only for essentially nameless objects, if such exist (see
Williamson [1994], p.262). In the second case, some constants tl and t2
denote 01 and 02 respectively, but ~ tl = t2 V ~-, tl = t2 is false for all of
them. Since -,~ tl = 12 1\ -,~-, t\ = 12 is true if ~ 1\ = t2 V ~-, t\ = 12 is
false, 01 and 02 will be presentable as neither clearly identical nor
clearly distinct. Conversely, if they are presentable as neither clearly
identical nor clearly distinct then some pair of constants denotes them.
Thus for our purposes the interesting case is that in which objects are
presentable as neither clearly identical nor clearly distinct without be-
ing presentable as either clearly identical or clearly distinct. We noted
above that if 01 and 02 are identical, they are presentable as neither
clearly identical nor clearly identical only if they are also presentable
as clearly identical. Thus the interesting case always involves distinct
objects.
We can illustrate the interesting case in the simple model sketched
above, with just two interpretations II and 12, indiscriminable from
each other, and two constants a and b. Let a denote 0\ on both II and
12 ; let b denote 0\ on II but the distinct object 02 on 12 . Suppose that 12
is in fact the correct interpretation. Since -,~ a = b 1\ -,~-, a = b is true
and a and b denote 0\ and 02 respectively (on the correct interpreta-
tion), 01 and 02 are presentable as neither clearly identical nor clearly
distinct. Since no other constants denote 01 and 02 respectively, they
are not presentable as either clearly identical or clearly distinct.
Of course, such a case might simply manifest the expressive limi-
tations of the language. Perhaps it is sheerly accidental that the lan-
guage does not contain more precise constants t\ and 12 denoting 0\ and
02 for which ~-, tl = t2 is true. To define more satisfying notions, we
should generalize over potential as well as actual ways of denoting the
objects. If the language contains demonstratives, we must at least con-
sider various contexts of their utterance. Since similar issues arise for
vague thought about objects as well as for vague talk, we must con-
sider guises, modes of presentation or individual concepts of the ob-
jects, and elucidate those notions. But there is an opposite risk: if we
generalize too far, we may make all pairs of objects both presentable as
either clearly identical or clearly distinct and presentable as neither

clearly identical nor clearly distinct, so that the interesting case never
arises. Perhaps we can avoid that result by restricting ourselves to cog-
nitively accessible ways of denoting objects in some sense. If so, in the
interesting case the non-presentability of the objects as either clearly
identical or clearly distinct will manifest the cognitive inaccessibility
of all ways of denoting them on which their distinctness is clear. Al-
though that would reflect some cognitive limitation on the thinkers, it
would also reflect something about the objects too, because it would
distinguish them non-accidentally from other pairs of distinct objects.
The task of specifying the appropriate notion of cognitive accessibility
will not be attempted here; the suggestion is left as programmatic.
What has been specified is the overall structure of such an account.
We cannot know of any particular objects that they are not pre-
sentable as either clearly identical or clearly distinct. For otherwise we
could deduce that they were distinct and thereby come to know that
they were distinct, in which case they would be presentable as clearly
distinct, in virtue of whatever ways of denoting them we used; we are
in a position to know only what is clear. Such ignorance will not sur-
prise the epistemicist. We may know that some pairs of objects exem-
plify the interesting case, even if we cannot know which objects they
are. We might even know of particular objects that, unless they are
identical, they are not presentable as either clearly identical or clearly
distinct.
The technique used to define relations between objects from
opaque occurrences of their names can obviously be generalized. For
example, we could substitute mereological relations for identity and
distinctness; 0\ might be presentable as neither clearly part nor clearly
not part of 02 without being presentable as either clearly part or clearly
not part of 02. Any closed sentence «.I\, t2) could be substituted for t\ =
h, although any vagueness in «.I\, t2) not derivative from vagueness in
t\ or t2 is liable to diminish the interest of the corresponding relation
between objects. The logic of the resulting notions will be quite com-
plex. For instance, although L\(a /\ (3) is equivalent to L\a /\ L\[3, objects
may be both presentable as clearly R and presentable as clearly S with-
out being presentable as clearly both Rand S, for the constants needed
to present them as clearly R may differ from those needed to present

them as clearly S.
The relations between objects are defined in the metalanguage.
The move from object-language to metalanguage is usually conceived
as a move from talk about things to talk about words. That conception
is too simple when the metalanguage has the resources for referential
semantics. In the present case, the move from use of object-language
sentences of the forms ~ t\ = tz and ~-, t\ = t2 to mention of them in the
metalanguage is exactly what enables us to define an identity-
involving relation between things rather than a relation of vague
coreference between their names. A fuller exploration of these points
can be set in a framework for discussing wider issues about vagueness
in reality (see Williamson forthcoming for such a framework).
Which of those relations objects have to each other depends on
epistemic factors. In that sense, the relations do not concern only the
mind-independent environment, even when the objects so related be-
long to that environment. Perhaps the relations are extrinsic. Never-
theless, they relate the objects themselves, not linguistic or conceptual
representations of them. Nothing in the borderline cases that inspire
talk of vague identity requires more than epistemicism provides. It can
do justice to those modest phenomena. We have seen reason to doubt
that other views can.
Thanks to audiences at Edinburgh, St Andrews, Bergamo and Syracuse for

helpful comments on earlier versions of this material, and in particular to Dorothy
Edgington for her reply at Bergamo, to Brian Weatherson for his reply at Syracuse,
and to Kit Fine, Richard Holton and Graham Priest for further discussion.
WILLIAMSON ON VAGUENESS, IDENTITY
AND LEIBNIZ'S LAW
Dorothy Edgington
1.
I refer to an object, say a ship, at a time, fl. Call the ship'S 1'. I can see
it, touch it, go for a sail in it. I know what I am talking and thinking
about. The same kind of thing happens at a later time, fl, when I refer
to a ship - call it 'S2'. But, because of transformations which have
taken place, it is, it seems, indeterminate whether or not S 1 is the same
ship as S2.
There are also synchronic cases of apparently indeterminate iden-
tity. I gesture towards a mountain and say 'That mountain is high'.
Some distance away, you gesture towards a mountain and say the same
of it. It can be unclear whether we speak of the same mountain. For we
are at opposite sides of a landmass with a dip in it of such a size and
shape that it is not clear whether the landmass consists of two moun-
tains, or one mountain with a dip in it.
A challenging argument by Gareth Evans [1978] seems to show
that there cannot be objects such that it is indeterminate whether or not
they are identical. Suppose it is indeterminate that SI = S2. It is not
indeterminate that S2 = S2. But now S 1 has a property which S2 lacks
- that of its being indeterminate that it is S2. By Leibniz's Law, it fol-
lows that SI is not S2. If they were identical, there could be no prop-
erty which S 1 has and S2 does not have.
305
A. Bottani et al. (eds.), IndiviLlllllls, Essence and Identity, 305-318.
306 Individuals, Essence and Identity
What is the best account of cases like the above? How exactly
should Evans's (very brief) argument be formulated and is it
irresistible? These are the themes of Timothy Williamson's paper.
There are three kinds of approach to the problem, which I shall
call the semantic, the epistemic, and the realist views. According to the
semantic view, there are no objects such that it is indeterminate
whether they are identical. All material objects have perfectly exact
spatio-temporal bounds. It is our ways of referring to them - our words
and concepts - which are vague. We simply have not decided to which
amongst various overlapping candidate objects 'Sl' or 'that mountain'
refers. There are indeterminate identity statements in this sense: on
some permissible sharpenings of our terms'S 1 = S2' is true; on other
permissible sharpenings 'S 1 = S2' is false. This construal is immune
from Evans's challenge. No objects are being said to stand in the rela-
tion of indeterminate identity. What is indeterminate is to which ob-
jects our singular terms refer. (This point is made by David Lewis
[1988])
The semantic view is set aside at the outset of Williamson's paper.
It is more natural, he says, to hold that the vagueness concerns the ob-
jects themselves, and not merely their names.
When the name 'Sl' was introduced, it was clear which ship was being
pointed at; likewise when the name 'S2' was introduced. Thus there are
objects (the ships Sl and S2) and a relation (identity) such that it is vague
whether those objects stand in that relation.
And:
similar identity puzzles arise for natural objects, such as planets, many of
which [ ... ] would have had the same existence and nature even if beings
capable of forming representations had never evolved (p.274).
The aim of his paper is to find a way to make sense of the phenomenon
of vague identity in terms of a relation between objects.
The epistemic view (which Williamson favours), like the semantic
view, holds that all objects have determinate boundaries. There is a
determinate fact of the matter about when or where S 1, or that moun-
tain, ends. On the epistemic view, unlike the semantic view, I do refer
to one determinate object by 'Sl' or 'that mountain'. But in the cases
Dorothy Edgington, Williamson on Vagueness. Identity. Leibniz's Law 307
at issue, I cannot know what their exact bounds are (and this ignorance
has a special kind of explanation). The quick response to Evans's ar-
gument, from an epistemic perspective, would be that epistemic con-
texts are opaque. From 'It is not known whether a = b' and 'It is
known that b = b' it surely does not follow that a and b are not identi-
cal. However, orthodoxy has it that where there is opacity, the sen-
tences cannot be construed as expressing a relation between the ob-
jects, a and b. Williamson goes on to show how to define genuine re-
lations between the objects to which our terms refer, '01 and 02 are not
presentable as either clearly identical or clearly distinct', etc.. I return
to the details in §4 below.
The realist view, unlike the other two, holds that objects really do
have indeterminate, or fuzzy, boundaries. In the typical case, a material
object - an animal or plant, river or mountain, house or chair - comes
into existence, and goes out of existence, more or less gradually. There
is no precise moment at which it begins or ends. In between, its spatial
boundaries are not completely determinate. In most cases the regions
of indeterminacy are small and untroublesome. In most cases, this in-
determinacy does not give rise to indeterminacy of identity. That an
object fades a little fuzzily into its surroundings does not entail that it
is indeterminate which object it is. Call these 'ordinary fuzzy objects'.
Neither the semantic nor the epistemic view can allow that there
are such objects. For even if they do not give rise to vague identities,
they do give rise to vague statements - about their spatial or temporal
extent, and other things consequent upon that. For instance, 'Ben Ne-
vis has a surface area of at least 20 square kilometres' may be neither
clearly true nor clearly false, as may 'There are spruce trees on Ben
Nevis' if the only spruce trees are on its fringe. The semantic account
of this vagueness is that we have left undecided to which precisely
bounded object 'Ben Nevis' refers. On the epistemic view, these vague
statements do have determinate truth values, but we cannot know
which: 'Ben Nevis' does refer to an object with perfectly exact
boundaries, but we do not know what these exact boundaries are. Nei-
ther of these is as natural as the realist view. This provides some mo-
tive for trying to find an acceptable realist account.
It is what I shall call peculiar fuzzy objects which are threatened
by Evans's argument: cases in which the fuzziness of objects' bounda-
ries make it unclear whether, in a given stretch of space and time, there
is one object or two. Consider our mountain(s). Abbreviate 'that
mountain' said from the right, to 'a', and 'that mountain' said from the
left, to 'b'. The only trees are on the right. 'Determinately, a has trees
on it' is true. 'Determinately, b has trees on it' is false. For it is inde-
terminate whether b stretches as far as the trees. Construing the state-
ments realistically, something is true of a which is not true of b. So, by
Evans's argument, a and b are distinct mountains.
Suppose our original ship, S 1, is being gradually transformed into
another ship, S3, in such a way that throughout the process, there is a
ship there. (Someone offered to buy the original ship, which is of his-
toric importance. The sale was agreed on the conditions that it be re-
placed by a replica, and the replacement be continuous and gradual).
S2 is the ship I refer to around the middle of the process. Is it still S l,
or has it become S3? The realist wants to say: it is indeterminate
whether S2 is the old ship or the new. By Evans's argument, it follows
that it is not the old ship, and not the new ship. So it must be a third
ship. Sl, S2 and S3 are all distinct. Very well, there is a third ship.
When did it begin? Is the ship we have about a quarter of the way
through still Sl, or has it become S2? If we try to say 'it's indetermi-
nate which it is', from Evans's argument it follows that it is neither. So
there is a fourth ship. And so on. If realism about peculiar fuzzy ob-
jects is defensible, fault must be found with Evans's argument.
2.
In §§l and 2 of his paper Williamson criticizes theories which attempt

to vindicate a realist view of vague identity. I have little to say about
his §2. There Williamson argues against those who try to demonstrate
the consistency of the thesis that identity can be indeterminate by con-
structing rigorous formal models for a language in which sentences of
the form 'a = b' may be neither determinately true nor determinately
false. When you look at these models, says Williamson, you find that
straightforwardly distinct objects are claimed to be such that it is in-
determinate whether they stand in this relation; and so, from a classical
perspective, the relation these theorists call '=' is not the identity rela-
tion. (Perhaps it is the relation of partial coincidence.) Of course, that
Dorothy Edgington, Williamson on Vagueness, Identity, Leibniz's Law 309
is not how these model theorists see matters: for them, there are ob-
jects in the model such that it is indeterminate whether they are identi-
cal. But I think Williamson is right that these models can have no sua-
sive force for those who are not already, independently, convinced that
objects can stand in this relation.
In § 1 Williamson criticises attempts to deal with vague identity in
terms of a many-valued logic, whether three-valued or continuum-
many-valued, which operates under the assumption of generalized
truth-functionality: the value to be assigned to (e.g.) a conjunction is a
function of the values of the conjuncts. Call clear truth 1 and clear
falsehood o. On the continuum-many-valued approach a proposition
gets a value the closer to I the closer it is to clearly true, 0.5 when it is
evenly balanced between the two poles. On the three-valued approach,
if we call the third value - neither clearly true nor clearly false - 0.5,
Williamson's criticisms can be stated in a uniform way. On both ap-
proaches, the value to be assigned to the negation of a proposition A is
one minus the value to be assigned to A; and the value to be assigned
to (A /\ B) is the minimum of the values to be assigned to A and to B.
The negation rule causes no problems, but the conjunction rule does.
Suppose Cicero is a borderline case of tall-let v(Cicero is tall) = 0.5.
As Tully is no more or less tall than Cicero, v(Tully is tall) = 0.5. So
v(Tully is not tall) = 0.5; and v(Cicero is tall and Tully is not tall) =
0.5; and v(It's not the case that (Cicero is tall and Tully is not tall» =
0.5. But 'Cicero is Tully' is clearly true. The inference 'Cicero is
Tully; so it is not the case that (Cicero is tall and Tully is not tall)' has
a clearly true premise and a conclusion with value 0.5. No valid argu-
ment should take us from a clearly true premise to a conclusion which
is not clearly true: the argument, on this account, is invalid. Yet the
argument is a trivial variant of Leibniz's Law; and it is intuitively
compelling. Similarly, 'Cicero is tall and Tully is not tall; therefore
Cicero is not Tully' has a premise with value 0.5 and a clearly false
conclusion. Surely, a one-premise valid argument with a clearly false
conclusion must have a clearly false premise. Again, the argument ap-
pears to be invalid on this account; again the argument is intuitively
compelling. Our elementary ways of reasoning with identities breaks
down.
The same problem infects the transitivity of identity. Consider our

ships. S 1 gets transformed gradually into the distinct ship S3. Half way
along the process, we have a ship, S2, such that it is indeterminate
whether it is SI or S3. v(SI = S2) = 0.5. v(S2 = S3) = 0.5. So, on this
approach, v((SI = S2)/\(S2 = S3» = 0.5. But v(SI = S3) =0. Also,
v(--,(Sl = S3» = 1; but v(--,((SI = S2) /\ (S2 = S3») = 0.5. The transi-
tivity of identity appears to be invalid, on this approach.
These criticisms, as Williamson says, are an application to identity
statements of points that have been made previously against general-
ized truth-functionality in many-valued logics for vagueness, originally
by Kit Fine [1975]. I think they are completely convincing. But I do
not think they should be taken as a refutation of many-valued accounts
of vagueness per se.They just show that generalized truth-functionality
for conjunction (and disjunction) is a mistake - a wrong guess, an
over-simplistic idea - and must be abandoned. If we want a many-
valued account, we must look for some other generalization of the
classical two-valued semantics.
I have no brief for three-valued logic. But it is plausible that we
capture something of the essence of vagueness by noting that we can
have a series of cases, differing from their neighbours only minimally,
which take us progressively further from the clearly true, closer to the
clearly false, and we should theorize about the phenomenon in these
terms.
Williamson mentions in passing (p.277) another many-valued
framework applied to propositions, in which the value to be assigned
to (e.g.) a conjunction is not a function of the values assigned to the
conjuncts: the probability calculus. On the next toss, the probability of
heads is 0.5, the probability of tails is 0.5, the probability of not tails is
0.5. But the probability of (heads and tails) is 0, the probability of
(heads and not tails) is 0.5. Why? Because peA /\ B) = peA) x pCB
given A). It is this last factor which varies in the two cases. p(tails,
given heads) = 0, p(not tails, given heads) = 1. In other words, we need
to take into account the fact that there may be relations of positive or
negative dependence between the conjuncts.
The probability calculus, as an abstract structure, has more than
one application: to degrees of closeness to certainty in our epistemic
attitudes; and to objective chances. It may also provide an idealized
Dorothy Edgington, Williamson on Vagueness. Identity. Leibniz's Law 311
model of the right structure for describing the indeterminacy due to

vagueness. I have argued elsewhere (Edgington [1996]) that 'degrees
of closeness to clear truth', idealized as precise, should be given this
structure.
I introduce the term 'verity' as short for 'degree of closeness to
clear truth'. To handle conjunctions, we need the analogue of condi-
tional probability: the verity of B, given the hypothetical decision to
count A as true. Return to the examples above. Cicero is exactly as tall
as Tully. So v(Tully is tall, given that Cicero is tall) = 1. And v(Tully
is not tall, given that Cicero is tall) = O. So v(Cicero is tall and Tully is
tall) = v(Cicero is tall) x v(Tully is tall, given Cicero is tall) = 0.5;
v(Cicero is tall and Tully is not tall) = v(Cicero is tall) x v(Tully is not
tall, given Cicero is tall) = O. Hence v(-{Cicero is tall and Tully is not
tall» = 1. The argument 'Cicero is Tully; so it is not the case that
(Cicero is tall and Tully is not tall)' has a clearly true premise and a
clearly true conclusion. 'Cicero is tall and Tully is not tall; so Cicero is
not Tully' has a clearly false conclusion and a clearly false premise.
Similarly, with our ships: although v(SI = S2) = 0.5 and v(S2 =
S3) = 0.5, v((SI = S2) /\ (S2 = S3» = v(SI = S2) x v((S2 = S3), given
(SI = S2». This last term is 0, so the conjunction gets O. The counter-
examples to the transitivity of identity and to Leibniz's Law have dis-
appeared.
Call the 'unverity' of a proposition (its degree of closeness to clear
falsehood), one minus its verity. All and only classically valid argu-
ments are those which preserve verity in the following sense: there is
no assignment of verities such that the unverity of the conclusion ex-
ceeds the sum of the unverities of the premises. This is the analogue of
an elementary theorem of probability theory. I prove the result directly
for verities in Edgington [1996], pp.307-8. So, if we take this as our
criterion of validity, we can preserve classical logic.
Leibniz's Law, in the form in which it is used in the logic ofiden-
tity for an extensional language, entitles us to infer ¢(b) from a = band
c/J (a). It is valid in this framework. If the premises get value 1, the con-
clusion gets value 1. Hence if v(¢(a» = 1 and v(¢(b» < 1 it follows
that v(a = b) < 1. The contrapositive form is also valid: from ¢(a) and
--,¢(b) we may infer a *- b. Consider our mountain(s). v(There are trees
on a) = 1. v(There are trees on b) = 0.5, for it is not clear whether b
extends as far as the trees. So v(-,(There are trees on b)) = 0.5. It fol-
lows that v(a =F- b) is at least 0.5, hence v(a = b) is at most 0.5.
This framework is non-revisionary about identity. Identity is re-
flexive, transitive and symmetric, and Leibniz's Law, and its contrapo-
sitive, are valid. We still have to face up to Evans's argument. But
there is a many-valued theory of vagueness which is immune to the
objections raised against the generalized-truth-functional variety.
3.
Williamson's §3 asks, which form of Leibniz's Law should Evans's

argument rely upon, if it is to be compelling? In one version it is a
principle which licenses the substitution of singular terms: from a = b
and cM..a), we may infer cM..b). As such, it is not valid in all contexts.
Williamson illustrates this nicely with the artificial operator 'Dis-
tinctly', defined as follows: the result of prefixing 'Distinctly' to a
sentence s is equivalent to the conjunction of s with every sentence of
the form t1 =F- t2, where t1 and t2 are distinct singular terms occurring
unbound in s. Thus, although Leningrad = St Petersburg, 'Distinctly,
Leningrad is no larger than Leningrad' is true, yet 'Distinctly, Lenin-
grad is no larger than St Petersburg' is false. Also, it is widely held that
substitution of singular terms can fail to preserve truth in epistemic
contexts; and perhaps in some modal contexts. From 'It is logically
necessary that Hesperus is Hesperus' and 'Hesperus is Phosphorus' it
is not obvious that it follows that it is logically necessary that Hesperus
is Phosphorus. Arguably, logic is blind to the identity of Hesperus and
Phosphorus. So it would not be very surprising if substitutivity also
failed in the context of the operator t1 'It is determinate that'. If so,
from t1 a = a and a = b it does not follow that t1 a = b. And so from t1 a
= a and -,t1 a = b it does not follow that a =F- b.
We could construe Leibniz's Law as a metaphysical principle: if
objects are identical they have the same properties. This raises the
question how we are to construe 'property'. While, for some meta-
physical purposes, there may be good reason to deny that every predi-
cate picks out a property, it is not clear why, for the working of Leib-
niz's Law, the notion of property should be used in a restricted, narrow
sense. If this emerald is grue and that emerald is not grue we may con-
clude that they are distinct emeralds, whether or not the predicate
'grue' picks out 'real' property. Williamson's own proposal is equiva-
lent to the metaphysical principle on a wide reading of property, such
that every predicate does pick out a property. He construes Leibniz's
Law as a metalinguistic principle, about objects satisfying predicates or
open sentences; equivalently, predicates or open sentences being true
of objects; equivalently, assignments of objects to the variables in open
sentences yielding closed sentences, true, or not, as the case may be.
Thus his (CLL):
Let an assignment A assign an obj ect 0 to a variable v, an assignment A *
assign an object 0* to v, and A * be exactly like A in every other way.
Suppose that a sentence a is true relative to A and not true relative to A *.
Then 0 and 0* are not identical. (p.289)
In terms of (CLL), Evans's argument would run like this: consider the
open sentence 'Determinately x = y'. It is true for an assignment of
objects <0, 0> to its variables. It is not true for an assignment <0, 0*>
to its variables. So 0 "::j:. 0*.
As Williamson says, (CLL) is a consequence of an elementary
theorem of the mathematical theory of functions. I am willing to accept
that it provides the most compelling version of Evans's argument.
Nevertheless, it seems to me arguable that it is irresistible only given
the assumption that the objects in question are all either clearly identi-
cal or clearly distinct. These are, after all, the objects with which
mathematics deals (and with which we all deal, almost all of the time).
You cannot determinately count peculiar fuzzy objects!
Consider the open sentence 'There are trees on x', as it is treated in
the framework I sketched in the last section. The assignment of a to x
yields a sentence with value 1. The assignment of b to x yields a sen-
tence with value 0.5 ~ a sentence which does not get the value 1. Ap-
plying the thinking behind (CLL), it would follow that a and b are dis-
tinct. But all that does follow, in that framework, is that the value to be
assigned to a = b is at most 0.5. It does follow that a and b are not
clearly identical. It does not follow that they are distinct.
So we have a choice. Accept (CLL) and discard the framework as
defective. Or accept the framework; claim that (CLL) is valid when all
the objects involved are either clearly identical or clearly distinct, and
that without that assumption, the conclusion should be weakened to '0

and 0* are not detenninately identical'. It is not obvious that one must
opt for the first alternative.
The framework was applied to a language without a 'detenni-
nately' -operator. But we could introduce ~ to our object language, and
stipulate that ifv(A) = 1, v(~A) = 1 and ifv(A) < 1, v(~A) = O. (I ig-
nore the complication that it might not be clear whether yeA) is 1 or
merely very close - the problem of higher-order vagueness. I need only
consider cases where yeA) is clearly 1, or clearly less than 1). We can
then derive the most important logical properties of ~. Note that ~ A
cannot get a higher value than A. So ~A entails A. Secondly, take any
valid argument. If its premises get value 1, its conclusion gets value 1.
So the result of prefixing the premises and conclusion with ~ is valid.
Leibniz's Law held in the unextended language. So, then, does its 'de-
finitization': ~(a = b); ~(¢(a»; so ~(¢(b». Contraposing, we get
~(¢(a»; .....,~(¢(b»; so .....,~(a = b).
What about Leibniz's Law in the extended language? Is '~(¢(a»,
.....,~(¢(b», so a *" b' valid? Suppose its premises are detenninately true.
That is, detenninately, v(cJ>(a» = 1 and v(¢(b» < 1. If the inference
were valid, this would guarantee that its conclusion is detenninately
true, hence that v(a = b) is O. But all that follows is that v(a = b) < 1,
i.e. .....,~(a = b).
Construe 'a' and 'b' as straightforwardly designating the objects a
and b. Consider the open sentence 'Detenninately, x has trees on it'.
Assigning a to the variable yields the value true. Assigning b to the
variable does not yield the value true. In tenns of the framework I have
proposed, we may conclude only that it is not detenninate that a = b.
Williamson says this:
What seems clear is that, properly formulated, [Leibniz's] law embodies
an insight absolutely fundamental to the logical notion of identity. If x and
yare the very same thing, then whatever applies to x thereby applies to y.
To suppose that there are exceptions to the best statement of the law is to
lose one's grip on the topic. On some canonical formulation, rigorous and
free of ad hoc qualifications, the law is valid (p.285).
I have tried to show that we need not lose our grip on the notion of
identity. We can preserve all the classical laws governing identity
where peculiar fuzzy objects are concerned, except for one small
amendment: within the scope of the 'determinately' operator, from
/).(cfJ(a» and -,/).(cfJ(b» it follows only that -,/).(a = b). If it is determinate
that x and yare the very same thing, then it is determinate that what-
ever applies to x applies to y. If something determinately applies to x
but does not determinately apply to y, it is not determinate that what-
ever applies to x applies to y. So it is not determinate that x and yare
the very same thing.
4.
In his final section Williamson investigates how the epistemic view

can handle vague identity. 'It is not determinate that' having an onto-
logical ring, he prefers the terminology 'It is not clear that': /). is to be
read 'clearly'. 'It is not clear that A' means that A is unknowable, and
moreover, the source of ignorance is of a special kind. It is this:
a vague sentence might express a true proposition on its actual interpreta-
tion I while expressing a false proposition on some close counterfactual
interpretation 1*, where speakers' insensitivity to the difference between I
and 1* makes any belief they express by uttering the sentence too unreli-
able to constitute knowledge. In such circumstances we may describe 1* as
indiscriminable from I. The natural proposal is that /)'U is true on an inter-
pretation I if u is true on every interpretation indiscriminable from I.
(p.296).
A model for a vague language consists of a class of interpretations and

a reflexive symmetric relation of indiscriminability defined over them.
An interpretation I assigns a denotation to each constant. A sentence t1
= t2 is true on interpretation I if and only if I assigns the same denota-
tion to t1 and (2. /).(1 is true on I if and only if (1 is true on every inter-
pretation indiscriminable from I.
Take a model with two interpretations. II assigns the same deno-
tation to a and b, 12 assigns different denotations to a and b, and II and
12 are indiscriminable. The following are valid: /).(a = a), /)'(b = b),
-,/).(a = b), -,/).-,(a = b). Thus, /). creates an opaque context, within
which Leibniz's Law, as a principle of substitution, fails.
Consequently we should not regard a sentence of the form M\ = 12 on an

interpretation I as expressing a proposition about the objects that t\ and 12
do in fact denote on I to the effect just that they are clearly identical. For
otherwise A(a = a) and A(a = b) would express the same proposition [ ... ],
and would therefore have the same truth value. [... ] It would be less inac-
curate to regard Aa. as quoting a., and -,A a = b 1\ -,A-, a = b in particular
as expressing a proposition about the names a and b themselves, to the ef-
fect that it is unclear whether they denote the same object. That is vague-
ness in language, not in the objects to which it refers (pp. 297-298).
Williamson's elaboration of the epistemic view is closely parallel to

the semantic view. The first has its many indiscriminable interpreta-
tions, the second its many permissible interpretations. But now, Wil-
liamson claims, the epistemicist has an advantage to exploit. For he
can say, while the semanticist does not, that one interpretation is cor-
rect. Truth and denotation on the correct interpretation are truth and
denotation simpliciter. And by turning to the metalanguage, we can
define relations between objects. We can say what it is for 0) and 02 to
be presentable as (a) clearly identical, (b) clearly distinct, (c) not
clearly identical, and (d) not clearly distinct, according to whether
there are constants t) and t2 denoting 0) and 02 respectively, such that
(a) Ll t) = t2 is true; (b) Ll-, t) = t2 is true; (c) -,Ll(t) = t2) is true; (d)
-,Ll-,(t) = t2) is true.
If t) = tz is neither clearly true nor clearly false, and t) denotes 0)
and t2 denotes 02, then 0) and 02 are presentable as neither clearly
identical nor clearly distinct. This does not preclude its also being true
that 0) and 02 are presentable as clearly identical, or as clearly distinct.
Indeed, if 0) and 02 are identical, they are presentable as clearly identi-
cal; for in that case, t) denotes 0), t) denotes 02, and Ll(t) = t)).
As Williamson says, we cannot know of any particular objects that
they are not presentable as either clearly identical or clearly distinct.
Suppose we knew that S I and S2 are not presentable as clearly identi-
calor clearly distinct. Well, then we could deduce that they are not
identical - for if they were identical, they would be presentable as
clearly identical, by the names'S!' and'S! '. But then we would know
that they are distinct, and'S!' and 'S2' would present to us two clearly
distinct objects.
Even if 0] and 02 are distinct, and presentable as neither clearly

identical nor clearly distinct, there may be more precise ways of refer-
ring to them so that they are also presentable as clearly distinct. This is
Williamson's example: stipulate that s is to denote the highest small
number. Suppose that in fact (on the correct interpretation) s denotes
17. But for all we know it denotes 16: there are interpretations indis-
criminable from the correct one on which it denotes 16. So -,,tl-, s =
16. Also -,,tl s = 16. So 17 and 16 are presentable as neither clearly
identical nor clearly distinct. But of course, 17 and 16 are also present-
able as clearly distinct - when they are referred to by the numerals' 17'
and '16'. Thus, presentability as neither clearly identical nor clearly
distinct is not a very interesting relation. We should focus on the case
of objects 0] and 02 which are presentable as neither clearly identical
nor clearly distinct, and are not also presentable as clearly identical or
clearly distinct. Only distinct objects can be such, we have seen.
But nor will this case be very interesting if it would be easy to ex-
tend our language, with more precise terms, so that 0] and 02 are pre-
sentable as clearly distinct - if it is just an accident that we have no
precise enough terms to refer to 0] and 02 such that they are clearly
distinct.
To define more satisfying notions, we should generalize over potential as
well as actual ways of referring to objects,
says Williamson. But, he recognizes, there is an opposite risk:

if we generalize too far, we make all pairs of objects both presentable as
clearly identical or clearly distinct, and presentable as neither clearly
identical nor clearly distinct. Perhaps we can avoid that result by restrict-
ing ourselves to cognitively accessible ways of denoting objects. (pp. 301-
302)
If we turn to standard examples of vague identity, the risk Williamson

mentions is real. Consider our ships. S 1 is the ship picked out at t]. S2
is the ship picked out at h. Suppose that in fact they are not identical:
on the correct interpretation, S 1 is a ship which ceased to exist at a
time t before 12. Let s denote the ship at such-and-such location which
ends at t. s denotes S 1. As S2 clearly exists at t2 and s clearly does not
exist at t2, sand S2 are clearly distinct. So our ships, as well as being
presentable as neither clearly identical nor clearly distinct, are present-

able as clearly distinct Consider our mountains. Suppose that on the
correct interpretation they are distinct Then there is no great difficulty
in finding ways of picking them out as distinct objects. So it is far from
obvious that any very interesting or discriminating relation between
objects can be defined in these tenns, when pairs of objects like 16 and
17, as well as our ships and our mountains, can tum out to be present-
able as clearly identical or clearly distinct, and also presentable as nei-
ther clearly identical nor clearly distinct
So, although ingenious, I am not convinced that Williamson's
technique for defining genuine relations between objects reveals any-
thing very interesting about the intuitive cases of indeterminate iden-
tity. If I were an epistemicist, I think I would rest my case with opacity,
and claim that the problem of indetenninate identities reduces to the
puzzle, which goes back to Frege, about epistemic contexts. One can
know a extremely well, easily well enough to know what one is talking
and thinking about Similarly for b. Yet one might not realise that a is
b. Conversely, perhaps one thinks that a is b but they are distinct It is
not definitively settled what is the best philosophical account of this
phenomenon.
Why epistemicism is a minority view is that most of us cannot be-
lieve that our tenns must pick out, unbeknownst to us, perfectly deter-
minate objects. Williamson's variety of epistemicism has much in
common with the semantic view, apart from this claim. If this claim
does not give us a way of making illuminating distinctions between
objects themselves, then (in this context) it is hard to see that it has an
advantage over the semantic view. I have argued that there is also an
uneliminated realist view.
ORIGINS AND IDENTITIES
Graeme Forbes
1. A CRITERION FOR CROSSWORLD IDENTITY
Kripke's thesis of the essentiality of biological origin (Kripke [1972],

pp. 312--4) may be written
(EBO) D(\fx)D(\fy) [O(y originates from x) ~ D(y exists ~ y
originates from X).I
That is, for any possible objects x and y, if y originates from x in some
world, then y originates from x in every world in which y exists. This
thesis strikes an intuitive chord with many, and a number of proposed
justifications for it have been advanced. 2 My own argument for (EBO)
is based on a principle about identity, namely, that for things which in
some good sense come from or are composed of or constructed from
other things ("composite" objects), ungrounded identities and un-
grounded non-identities are to be abjured.
Ungrounded identities of composite objects, if they were possible,
would be cases where an identity holds though there is nothing in
which it consists. For instance, if an individual 0 has physically and
functionally equivalent brain-hemispheres and undergoes successful
hemisphere transplants that produce two new individuals Lefty and
Righty, that 0 = Lefty or that 0 = Righty would be examples of un-
grounded identities. 3 The symmetry makes each identity hypothesis
319
unintelligible in the context of the case, unless we surreptitiously posit

a symmetry-breaker, such as an immaterial soul.
Ungrounded non-identities of composite objects, if they were pos-
sible, would be cases in which we satisfy all conceivable conditions for
the identity of x and y not logically entailing it, but still, x =J. y.
However, not any old condition can be a ground for an identity or
a non-identity. As a first approximation, grounds for identity and non-
identity must be intrinsic, not extrinsic. This means, among other
things, that whether or not x = the (jJ should not tum on the presence or
absence at the relevant time or world of some entity that is causally
isolated from the (jJ.4
My primary argument for (EBO) is that denying it produces un-
grounded identities and non-identities. But a defense in these terms is
only as strong as the alleged principles about identity that (EBO) is
shown to save. An important principle in the present context, one that
we will ultimately qualify in a modest way, is that indistinguishability
of intrinsic nature across worlds suffices for identity. More carefully,
let us write x@w to mean, until further notice, the part of the intrinsic
nature of x at w which does not logically imply the identity of its bearer
(so it may tum out that, at some world, y satisfies x@w though y =J. x).
Then (EBO) preserves the following consequence of the sufficiency
for identity of crossworld intrinsic indistinguishability:
(1) Ifu =J. v, and wt and wt are each distinct from u and v, then ifat
wt, a satisfies x@u, and at wt, b satisfies x@v, then a = b.
Granted, if intrinsic just means non-relational, principle (1) is not very

plausible. No doubt there are cases of non-relational indistinguishabil-
ity that nevertheless involve distinct objects. For instance, if a and b
are, say, identical twins, the non-relational part of the life of either
should be possible for the other. So there are wt, wt, u and v as re-
quired such that the antecedents in (1) hold; but a is not b.
However, there is a broader notion of intrinsic in the literature, ac-
cording to which some relational properties of a thing can be part of its
intrinsic nature. 5 For example, if x is a bicycle and y one of its wheels,
it is intrinsic to x to have y as a part (thought not intrinsic to y to be a
part of x). And it is intrinsic to {x} to have x as a member (though not
intrinsic to x to be a member of {x} ).6 In the same spirit, if x is an or-
Graeme Forbes, Origins and Identities 321
ganism which develops from a single propagule y (the term propagule

was suggested to me for these purposes by Richard Dawkins), it is in-
trinsic to x to develop from y. After all, there is a phase of x's existence
when it is entirely constituted by y (in the case of humans, simplifying
the facts, a zygotehood).7 Extending this to multi-propagule organisms
is a small step, trading entirely constituted by for partially. With in-
trinsic understood in this way, (1) becomes much more plausible. s
Still, for those who are uncomfortable with the broader notion of
intrinsic, there is a way of reading (1) which does not beg any of the
questions addressed in this paper. If interior properties are not intrinsic
in an acceptable sense, they are at least identity-relevant. And the main
issue we face is whether there is any combination of properties at all
that uniquely determines the identity of a satisfier without logically
entailing it. So there is no harm in reading the notation x@u in (1) to
include all the identity-relevant features x possesses at u, including re-
lational ones, as well as those that are intrinsic in a narrow sense, so
long as we continue to exclude anything that logically entails the iden-
tity of a w-satisfier of x@u. Nor need we attempt a definition of iden-
tity-relevant, since our case for (EBO) ultimately hinges on a contro-
versy about what can count as identity-relevant, that is, on the details
of particular candidates.
2. THE ARGUMENT FOR (EBO)
The argument that (EBO) preserves (1), or at least blocks a particular

kind of counterexample to it, is as follows. Suppose (EBO) is false.
Then there are a, WI' w 2 ,PI andp2 (all distinct) such that a's propagule
in WI is PI and a's propagule in W2 is P2' We also suppose that W2 is as
little different as possible from WI' Then there is W3 where a develops
from PI' and also
(2) <>(3x)(3y) (x * y /\ X satisfies a@w3 /\y satisfies a@w 2).
That is, there is w4 (corresponding to <> in (2)) in which x duplicates a

as it is at w3 in every relevant respect, and y duplicates a as it is at w2
in every relevant respect (call W4 a two-candidate world). The crucial
requirement is that the way a is at w 2 and the way a is at w3 are suffi-

ciently different that they are ways for distinct things to be in the same
world. Since a cannot be both of these things, we have a contradiction
with (1), (W4 = wt = wt).
For a more general perspective, say that a property is exclusive iff
necessarily, at most one thing possesses it. To show, given (1), that an
exclusive property P is essential to a, assume it is not and thereby gen-
erate a way for a to be, a@w2' in which a is not P. Then find another
way a@w3 for a to be such that in some world, distinct objects satisfy
a@w 2 and a@w 3. But how do we know that there is such an a@w3? In
any given case the existence of a@w3 is open to intuitive demonstra-
tion. For example, if a is an oak tree and p is its actual propagule-
acorn, then since a@w 2 is as similar as possible to a@the actual world
consistent with q replacing p, a@w3 could be a way for a to be in-
volving developing from p at a different place or time from its point of
origin in w2. This leaves enough "room" for a@w2 to be satisfied by
one thing while a@w 3 is satisfied by another at some two-candidate
world W4'
For a@w3 to leave enough room for co-satisfaction with a@w2 in a
two-candidate world, the exclusive properties of a in W2 must all be
accidental to a. For if some exclusive Q in a@w2 is essential to a, it
must also be in a@w 3, and hence (by definition of exclusive) there is
no world where some x satisfies a@w 2 and some distinct y satisfies
a@w3' Indeed, the exclusive properties of a in W2 must be non-
distributive-all accidental - there must be worlds where a lacks all of
them (and w3 must be such a world). For if not, some exclusive Q is
common to a@w2 and a@w3' again ruling out two-candidate worlds
where a@w2 and a@w3 are satisfied by distinct things. These contin-
gency requirements on exclusive properties might seem onerous. On
the other hand, in the hypothetical case of the oak-tree a and its prop a-
gule acorn p, there does not seem to be any difficulty with the claim
that in w3, a's exclusive properties are all different from those it has in
w2. So we shall assume that the contingency requirements can be met.
Since denying (EBO) leads to a two-candidate world, contradicting
(1), this provides (EBO) with a rationale, that of maintaining consis-
tency with (1). Nor can the sceptic about (EBO) accept a weakened
version of (1), according to which it only applies when there is just one
candidate, with origin, say, or overall intrinsic similarity, as the tie-

breaker in two-candidate worlds. For this means that identity is extrin-
sically determined, the presence or absence of a (typically, causally
isolated) second candidate being the key factor in some cases. 9
The two-candidate world gives us a case of intrinsically un-
grounded numerical difference (or difference ungrounded in intrinsic
and identity-relevant properties - from here on this qualification is to
be understood where appropriate). Since x t y, either x or y is distinct
from a. If it is x, then in view of w3, x t a is intrinsically ungrounded,
and if it is y, then in view of w2 , Y t a is intrinsically ungrounded. On
plausible assumptions, we will also have intrinsically ungrounded
identities, for surely there is some world or other where a satisfies
a@w3 and coexists with some b satisfying a@w2 • This ought to be a
symmetry case as puzzling as the one about Lefty and Righty, but it is
not. The moral appears to be that there is a problem with the alleged
way a could have been, a@w2' in which a has a biological origin dif-
ferent from its actual one.
3. ROBERTSON'S CRITIQUE
In an insightful discussion, Teresa Robertson ([1998], p.743) has re-

cently objected that this argument cannot justify (EBO) specifically,
since many different essentialist principles besides (EBO) preserve (1),
including some very unattractive ones. Suppose that a tree Tactually
grows a branch b. The essentiality of branching (EBR) says that it is
essential to T to grow b.1O We may therefore imitate the reasoning of
the previous section to show that denying (EBR) generates a conflict
with (1) and creates false symmetry cases. At w2' T lacks b, contra-
dicting (EBR). W3 is chosen so that T@w 2 and T@w 3 are different
enough for there to be a world w4 in which they are satisfied by distinct
things x and y. T cannot be both x and y, hence (1) fails. So much the
worse, a defender of (EBR) will say, for the postulation ofw 2, where T
lacks b: T cannot lack b. But it is obvious that we do not want to en-
dorse (EBR), for an actual tree could have ceased to exist before it de-
veloped anyone of its branches.
If we insist on (EBO), we can block the argument for (EBR) just

given. W 2 is supposed to be as similar as possible to WI consistent with
T failing to grow h. If (EBO) is in force, T has the same propagule in
WI' W2 and W 3 . Assuming this propagule cannot give rise to more than
one thing of T s kind in a world,11 a two-candidate world therefore
cannot be generated. (This simply reflects the earlier observation that
the exclusive properties of a@w 2 must be accidental to a if a two-
candidate world is to be produced.)
However, this way of blocking the argument for (EBR) raises two
serious problems for the defender of (EBO). First, one who advocates
(EBR) might well be offering an alternative to (EBO); the principles
are symmetric as far as preserving (1) is concerned, and a defender of
(EBR) can equally object to the argument for (EBO) that if T must
have the same branches in the one-tree worlds and growing a given
branch is exclusive, the two-candidate world that is supposed to em-
barrass the sceptic about (EBO) cannot be generated. Evidently, we
need to find independent grounds for picking one essentialist thesis
over the other. Call this the Bias Problem: of the candidate principles
that do the work of protecting (1), why prefer (EBO) to the others?
Secondly, there is the problem raised by Robertson ([1998], p.
745). Suppose the matter constituting Ts propagule p in WI can in
some way be "recycled" into an intrinsically indistinguishable zygote
that exists later (see further Price [1982]).12 Then a more plausible
biological origin essentialism may not even block the argument for
(EBR). For a more plausible essentialism would allow T to emerge in
W3 from a recycling of p's matter. But in that case, T@W3 does leave
room for co-satisfaction with T@w 2, and we get our two-candidate
world even despite insisting on biological origin essentialism. So to
protect (1), (EBR), or something just as bad, is needed anyway. Call
this the Recycling Problem. I take it first.
4. THE RECYCLING PROBLEM
Perhaps a recycled propagule is numerically identical to the original

propagule. But in addition to making biological origin a non-exclusive
property, this would be unnecessarily controversial, and Robertson
does not assume it. It would be controversial because a propagule, be-
ing an organism, can undergo a complete change of matter, so the re-

cycled propagule might be made of the matter of the original propagule
while the latter still exists. Thus a judgment of identity between the
original and the recycled propagule makes identity extrinsically deter-
mined, since it can only be maintained if a propagule continuous with
the original one does not still exist. 13
So PI f- Pz' However, if Pz is recycled from PI'S original matter and
PI f- Pz, then (EBO) does block the four-worlds argument for (EBR):
the three one-tree worlds across which Ts branches vary use the very
same propagule P for T if (EBO) is in force. And originating from P is
still an exclusive property if recycling can at best produce a q distinct
from p. So as before, no two-candidate world will be available with
which to embarrass the (EBR)-sceptic. We can be brazen about our
(EBR)-scepticism.
Robertson's response to this, one which I would have made myself
(Forbes [1986], pp.7-8) until I saw its consequences spelled out in her
paper, is that in requiring that it be the very same propagule that a one-
propagule organism originates from in every world in which it exists,
(EBO) is simply too strong: we should be willing to allow a recycled
version of that propagule to serve as well. But is it really possible for
me to have originated from a cell which is a molecule-for-molecule
reincarnation of the zygote from which I actually originated, even
though not identical to it? Offhand, it seems possible, but we are no
strangers to the phenomenon of something that initially seems possible
coming to seem impossible on careful reflection. 14 Moreover, I have no
strong intuition about such a recycling case: my strong intuition is that
I could not have originated from your zygote, that is, that Kripke's
judgement about his Queen/child of the Trumans case is correct
(Kripke [1980], p.112). Comparing this ordinary case to a science-
fiction case involving recycling, what is apparently the crucial feature
of the ordinary case is preserved on Robertson's assumptions, namely,
that I originate from different zygotes in different worlds {no doubt
Robertson would just say that the crucial feature missing from the or-
dinary case is that the alternative zygote is a recycled copy of the actual
one).15 Then given that the science-fiction case does not reveal some
fallacy in our thinking about the ordinary case, or enthymeme that
makes the ordinary case a special case, it is reasonable to extend ver-
diets about the ordinary case to the science-fiction case. In addition,

weakening (EBO) to allow recycled substitutions produces an essen-
tialism that fails to protect (1), as we have seen. But preservation of (I)
or something close to it is our only candidate for a deep rationale that
principles like (EBO) could have, and without (1), it is unclear how we
close the door on ungrounded identities and differences such as those
among 0, Lefty and Righty. So there is also a theoretical reason to dis-
allow originating from a propagule in one world and a mere copy of it
in another.
5. EXTRA-STRENGTH HAECCEITISM
Recycling cases pose a more direct threat to a defense of essentialist

theses like (EBO) that depends on the rejection of ungrounded identi-
ties and non-identities for complex things. Tom McKay [1986] has ar-
gued that we can use a pair like {p, q}, q recycled from p, or else a pair
{X, Y} of organisms that p and q give rise to respectively, to construct
strong counterexamples to principles like (1). For it seems that q does
not depend for its existence on p (call this the independence thesis).
And since p and q are things of the same kind, the same possibilities
not involving the other should be open to each. So in addition to a
world where they both exist, for different but overlapping periods,
there are worlds u and v which are factually the same except that p but
not q exists in u, and q but not p exists in v, and there are no other dif-
ferences except those logically entailed by this one. Hence p@u is nu-
merically identical to q@v, but P #- q. (It also follows that particular
identities can hold at two-candidate worlds despite the symmetry of the
two candidates for identity with, say, p.) Note that McKay's case is a
counterexample to (1) whether we construe intrinsic narrowly or
broadly, and would not be affected by any definition of intrinsic or
identity-relevant that we might come up with: regardless of how we
understand 'p@u', so long as it excludes properties which logically
entail the identity of any satisfier, McKay's case is a counterexample. 16
Let us use extra-strength haecceitism for the view that two worlds
u and v can be isomorphic under a function that sends everything that
exists in u to itself, with exactly one exception.17 There is a variant of

my argument in defense of (EBO) which suggests that denying (EBO)
leads directly to extra-strength haecceitism. Let A, which actually
originates from p, originate from q in w. There is no reason why q
could not have given rise to some B distinct from A (suppose in the
actual world it does), so let u be a world where q gives rise to such a B.
Again, it is hard to see why B@u should not be a way A could have
been, so there is v where A satisfies B@u. And apart from the fact that
in u it is B that satisfies B@u while in v it is A, there need be no differ-
ences between u and v beyond those necessitated by this one. This ar-
gument in defense of (EBO) (at least it is such in the eyes of those who
wish to resist ungrounded identities and non-identities) has some ad-
vantages over the four-worlds argument of §2, particularly in connec-
tion with non-standard modal semantics. 18 But it has the same prob-
lems with recycling (as well as bias). Suppose we use the same rea-
soning in defense of (EBR). We have anA which actually grows band,
according to the sceptic about (EBR), a w where A does not grow b.
Let u be a world where a distinct tree B does not grow b. We require
that B@u is a way A could have been, and if we are assuming (EBO)
this means B@u has to include an origin possible for A. We could just
stipulate that at u, B originates from A's actual propagule, but this is a
weak point in the defense of (EBR) if it is only a stipulation. A recy-
cling scenario strengthens the point.
More importantly, arguing that scepticism about (EBO) leads to
extra-strength haecceitism need not embarrass the sceptic if there is
nothing actually wrong with such haecceitism. Maybe McKay has
shown that animus towards ungrounded identities and non-identities is
mere prejudice. But extra-strength haecceitism has some commitments
that seem unattractive (to me). There is a world w where a recycling
process continues without end. Therefore extra-strength haecceitism
implies that there are uncountably many possible worlds which are ex-
actly the same except for the mere identities of certain organisms in
them, one world for each infinite sub-sequence of recycled entities (or
their products) from w. And there are infinitely many worlds where
some dog (a different one for each world) primitively distinct from my
neighbour'S dog leads the dog's life my neighbour's dog actually leads,
and everything independent of that dog's identity is the same as it ac-

tually is.
This proliferation of worlds is startling, and the culprit seems to be
the independence thesis, that q does not depend for its existence on p.
Suppose that in w, a zygote p consists of matter m in configuration c
recycled from an original zygote and a later zygote q consists in the
same matter in the same configuration recycled again (all three zygotes
are m-c zygotes). Assume there is a world u where p is the only m-c
zygote and a world v where q is the only m-c zygote and all else is as
close as possible to u. If we think of a possible world as an unfolding
course of events, this extra-strength haecceitist scenario implies that
there are two courses of events that share an initial segment to the
point where certain molecules are brought together in a certain way in
an event e, but at that point the courses of events branch: in one it is p
that comes into existence as a result of e, and in the other it is q. And
there are no other differences, then or subsequently, other than those
necessitated by this one. This radical transcendence of the identity of
the entity that e brings into existence seems to me to be as difficult to
understand as an alleged identity between 0 and (say) Lefty, rather
than Righty; it is another reason to doubt that q can exist even if p does
not.
6. PREDECESSOR ESSENTIALISM AND INTRINSICNESS
To rule out extra-strength haecceitism we have to say that q needs p,

and more generally that an entity's predecessors in a recycling se-
quence are its predecessors in every world where it exists. 19 One ques-
tion about such predecessor essentialism is its plausibility, though in
these recherche cases it is not easy for modal intuition to get a grip.20
Subsuming the plausibility issue is the question whether or not en-
dorsing predecessor essentialism undercuts (1). 'x@u' now has to be
understood to include information about the number of predecessors, if
any, of x, in the sense of predecessor that means, when x is a propa-
gule, having the same matter configured in the same way, and when x
is a propagule-product, having developed from a propagule with such a
predecessor. Are we to say that the number of predecessors of a prop a-

gule y satistying x@u or a product z of such a propagule is an intrinsic
property of y or z? Or even, weaker though it is, that it is so much as
relevant to the identity of y or z?
To begin with intrinsicness, I will look at some recent discussions
of the intrinsic/extrinsic distinction to see what guidance they offer
about whether p's being the nth m-c propagule in a world is intrinsic to
p. We will see that the proposed criteria either fail to settle the issue
unequivocally, or their delineation of intrinsic/extrinsic is independ-
entlyobjectionable.
Vallentyne [1997] develops the idea that an intrinsic property P is
one with respect to which an object x remains stable (x does not ac-
quire, or lose, P) under arbitrary deletion of parts of the world that are
external to x. Specifically, say that for any x, w, and t such that x exists
in w at t, an x-I contraction of w is obtained by removing as much as
possible of w save x, t, x's location at t, and things not wholly distinct
from x. Then Vallentyne proposes
(3) P is intrinsic =df \iw, t, x: (a) if Px at t in w then Px at t in each
x-t contraction ofw; (b) if --,Fx at t in w then -,Px at t in each x-
t contraction of w.
This criterion makes being the first m-c propagule extrinsic to any p
that has it in w, using (3b): in w, if p is not the first m-c propagule, we
can contrive an x-t contraction of w in which p is the first m-c propa-
gule simply by deleting p's m-c predecessors from w. (3) also makes
being the Jcth m-c propagule, k ~ 2, extrinsic to any p that has it in w,
using part (a) and the same x-t contraction ofw. 21
However, matters are less clear cut if we focus on m, which per-
sists, if in distributed form, through various episodes of constituting
propagules. (3) makes it intrinsic to p to be initially constituted of m,
and if it is also intrinsic to m not to have c-constituted any propagule
before p, we have good reason to count it intrinsic to p to be the first
m-c propagule. 22 Might it be intrinsic to m not to have c-constituted any
propagule before p? The negation of this property is unstable under x-t
contraction so long as we are allowed to contract away the life of m
prior to t «3) does not allow this). But there is some plausibility in the
thought that if at t it was intrinsic to x to be F then at a later t' it should
be intrinsic to x to have been F. So the verdict of Vallentyne's criterion

in application to our cases is not as unequivocal as it first seemed.
Yablo [1999] characterizes the intrinsic/extrinsic distinction
within a modal framework broadly similar to Vallentyne's. Yablo
takes a property P to be intrinsic iff
(4) For any possible x and world w in which x exists, x is stable
with respect to P under augmentation of w; that is, if w' in-
cludes w, Pxw iff Pxw'.
Being the first m-c propagule is evidently extrinsic; for if w is a world
where p is the first m-c propagule, we may simply wrap w in some
further goings- on in which an m-c propagule appears before p. This
also shows that being a subsequent m-c propagule is extrinsic: 'Pxw iff
Pxw" fails when w' contains further goings-on in which an m-c propa-
gule appears before p. On the other hand, as with Vallentyne's crite-
rion, it is intrinsic to p to be initially constituted of m (not a result
Yablo wants); having not previously constituted a c-propagule is one
of m's extrinsic features; but having previously c-constituted a prop a-
gule will be intrinsic unless we are allowed to drop the relevant part of
m's history in choosing a w that is part of w'. On the face of it, this is
not allowed by Yablo's mereological apparatus, which treats the way x
is as a common part of wand w': if pieces of the way x is at w' can be
missing from w, why not pieces that include intuitively intrinsic fea-
tures? Again, it looks as if the applicability of the proposed criterion to
our cases is far from clear cut.
Another objection to Yablo's criterion of intrinsic/extrinsic that
arises in the present context, one that also applies to Vallentyne's, is
that application of it to the properties we are discussing is question-
begging. This is because of a general difficulty accounts of "P is in-
trinsic/extrinsic" have with essential properties if those accounts tum
on the stability of P with respect to x in some crucial class of worlds
where x exists. For if P is essential to x, then inevitably P is stable with
respect to x across the whole range of worlds in which x exists,
whether or not those worlds are related as (3) and (4) require. 23 There-
fore, a property like being the first m-c propagu/e can be shown to be
unstable by accounts based on (3) or (4) only if it is not essential. But
the whole point of appealing to (3) or (4) was to settle a lemma used en
route to inferring that it is essential.
The general difficulty with approaches like (3) and (4), that essen-
tial properties are counted as intrinsic just because they are essential,
motivates the very different approach of Francescotti [1999]. He pro-
poses that a property is extrinsic if it is based on a relation, except, in
effect, if the relation is one of Humberstone's interior ones. So the
problem is to filter these relations out of the group that will give rise to
extrinsic properties. To do this, Francescotti introduces the idea of a d-
relational property of x, which is a property of x consisting in a rela-
tion that is borne to a distinct thing y. P is an impure d-relational prop-
erty of x if there are Rand y, y I- x, such that having P consists in being
in R to y; P is a pure d-relational property if there is a relation R, a sec-
ond-order property X and a quantifier Q such that having P consists in
bearing R to Q of the X's, and furthermore, possibly one of the X's is
not identical to X.24
Being the first m-c propagule would be extrinsic on this account if
it consists in being the earliest of all the m-c propagules, and being a
subsequent m-c propagule would be extrinsic if it consists in being the
later of some two of the m-c propagules, or in originating after the first.
The question is whether there is anything that will be counted as in-
trinsic on this approach. Why not say that squareness is extrinsic, since
to be square consists in being one of the square things? Simply ex-
cluding identity and non-identity from the possible values of R will not
help, since in each case there will be other relations and/or properties
that will do the job, for example, in the present case, being same-
shaped with?5 Nor is it sufficient to exclude being R to one of the X's
from consideration in evaluating X for intrinsicness (France scotti
[1999], p.603), since even if Y I- X, a thing may be R to one of the X's
iff it is R to one of the Y'S.26 It seems that a much greater burden must
fall on the notion of consisting in, allowing us to decide between the
following two accounts of being the first m-c propagule: that it con-
sists in being the earliest of all the m-c propagules, or rather that it
consists in being the first propagule that matter m constitutes in con-
figuration c. The latter, as we noted above, muddies the picture in view
of the intrinsic properties of m. 27
None of the accounts discussed above, of course, were fonnulated

with cases like ours in mind. My conclusion is merely that further de-
velopment is needed before we could trust their verdicts about our
cases.
7. PREDECESSOR ESSENTIALISM AND IDENTITY-

RELEVANCE
For those who think the number of m-c predecessors which a propa-
gule has is extrinsic to it, regardless of whether there is an otherwise
successful account of intrinsic/extrinsic which clearly says so, the ap-
propriate question is whether the number of such predecessors might
be relevant to the identity of such a propagule. We do not need to give
a fonnal definition of relevant to the identity of x; we only need to
make it plausible that certain features count, so that any definition in-
curs an obligation either to so count them, or to explain why the ap-
pearance of plausibility is an illusion. Nor does this move weaken the
case for origin essentialism: principle (1) can play the role it does in
defending EBO just as well if we read the notation x@u to include all
the identity-relevant features x possesses at u, as well as its intrinsic
features. So: are there other cases where number of predecessors, in
some relevant sense, plays an important or central role in detennining
identity?28
One example is the natural numbers. If we have a conception of
them according to which they have no internal structure, then what
distinguishes one natural number from another is its position in the
standard ordering. Position in the ordering consists just in number of
predecessors. It might be objected that this case is of no relevance to
ours, since the composite entities with which we are concerned differ
from numbers precisely in that they do have internal structure, their
intrinsic nature. But of course, the special feature of our crossworld
identity puzzles is that intrinsic nature (along with everything else ex-
cept identity) is factored out by being held constant across the entities
whose identity is to be settled. The only plausible candidate for an es-
sential property that grounds identity which is left in these cases is

number of predecessors.
Salmon has suggested another example closer to home (reported in
Hawthorne and Gendler [2000], p.292). If a ship X evolves ala Ship of
Theseus and a ship Y is built from X's original matter m according to
its plan I, how in the world, he asks, could one have constructed Y
without first having X? Hawthorne and Gendler admit that this ques-
tion has a point, but say that its force depends on thinking of Y as
having a plan that explicitly involves building it from the discarded
pieces of X (Ibid. p.293). But it is much more likely that the force of
the question derives from the thought that if one had built just one m-/
ship, X would preempt Y as a candidate for identity with this ship. If
you are puzzled how the single m-/ ship could be Y rather than X, it is
because you are counting the property of being the first m-l ship as
relevant to the identity of the counterfactual ship. In support of this,
note that there is no comparable puzzle about how the single ship
could be X rather than Y, as Hawthorne and Gendler agree: X could
have been constructed without anything being subsequently built from
its discarded parts.
But Hawthorne and Gendler suggest there are other cases where
we are not inclined to count being the first m-l ship as relevant to
identity. Suppose that actually there is no Y, but much later there is a Z
which by sheer coincidence happens to be an m-I ship. Couldn't such a
Z have existed by itself? Or suppose there is actually just one m-l ship
X, existing at the present time. In a world with two m-l ships, one ex-
isting at the present time, one a thousand years ago, there is little incli-
nation to insist that the earlier ship is X.
Hawthorne and Gendler are surely right about these verdicts, but
the identity-irrelevance of being the first m-l ship is not the only possi-
ble interpretation of them. It is equally conceivable that other identity-
relevant factors are at work which block default identification of the
first m-/ ship with X. The counterfactually-first m-/ ships in their ex-
amples exist at times which are very distant from the period during
which X exists. This seems to me to be no accident, for I find that the
further removed in time the origination of the counterfactual ship from
the origination of X, the weaker the claim X has on it. This suggests a
qualification of (1) to make the sufficient condition of (1) applicable
only when the item a in WI satisfying x@u and the item b in W2 satis-
fying x@v originate close enough in time. Close enough is of course
vague, and there are fundamentally two different ways one might ac-
commodate the vagueness, either through counterpart theory or a
vague accessibility relation. However, in the present context, what is
important is that qualifying (1) in this way concedes nothing to extra-
strength haecceitism. We still have a substantial and interesting suffi-
ciency condition for crossworld identity which, if the reasoning of sec-
tion 2 is cogent, turns out to support substantial and interesting neces-
sary conditions for crossworld identity.
8. BIAS AND THE OPEN FUTURE
Our discussion in §5 of whether only biological-origin-essentialism is

justified by arguments of a certain type allowed the defender of (EBO)
to use (EBO) to disrupt analogous arguments for deviant essentialist
principles. In conclusion, I wish to discuss briefly whether this bias
towards biological origin is itself justifiable. It is certainly more intui-
tive than the likes of (EBR), but that is not a justification. Any kind of
alternative to, or supplement of, (EBO), will classify as essential some
property an entity acquires subsequent to its coming into existence,29 a
property that can play a comparable role to biological origin in
grounding identity. Such a property will be clearly accidental, if for no
other reason than that that very entity could have ceased to exist before
acquiring the property. Can we say what underlies the that very entity,
preferably in a manner that does not rule out extra-strength haeccei-
tism?
I suggest that our ordinary conception of the future as open in a
way that the past is not is playing a role. One account of openness is in
terms of causal influence: present events cannot cause the past to be a
particular way, but they can cause the future to be a particular way.
However, there is a stronger idea of openness, according to which the
past is determinate and the future indeterminate. Model-theoretically,
this idea of openness is captured by branching worlds (see, e.g., Tho-
mason [1984]). For it to be determinate at a time t that A will be the

case, A must be the case in every future branching from t.
Consequently, if a tree T comes into existence at a time t, we
would commonsensically say that it is indeterminate which branches it
will grow, since in some of t's possible futures it grows these branches
and in others, those. But if we adopt (EBR), a deviant essentialist the-
sis which requires a tree T to grow the same branches in every world in
which it exists, this description is incoherent. Whether or not b grows
on the tree with such-and-such an origin at t is indeterminate at t, and
presumably remains indeterminate until either b grows on it or the tree
ceases to exist. But if trees must retain their branches across possible
worlds, there is no single tree with a future that is indeterminate be-
tween growing b and not growing it: there are as many different possi-
ble trees as there are possible outcomes of the branch-growth process.
The problem with this is that we have the tree in front of us now,
before it starts growing branches, and there is only one tree there. 30 It is
currently indeterminate which branches will grow on the tree that is in
front of us now. Therefore it is currently indeterminate which of the
various possible trees it is. This is an indeterminacy in identity even
stronger than that which is countenanced by extra-strength haecceitists,
where the Doppelgangers of my neighbor's dog at least exist in differ-
ent worlds. In the actual world, I can use that dog to refer to exactly
one of the possible dogs, namely, the actual one. But if the future is
open, then by the lights of (EBR), I cannot use that tree (or any defi-
nite description that does not include branch-growth outcomes) to refer
to a unique tree until there are no longer different possible futures dis-
tinguished by branch-growth outcomes (so determinate reference is
possible only to trees in the past!). However, even extra-strength haec-
ceitists can agree with the defenders of principles like (1) that the
identity of an entity that figures in a certain course of events is not in-
determinate in this way; it is not something that is fixed only after the
course of events unfolds past the entity's dissolution. In the extra-
strength haecceitist cases, the indistinguishable worlds with different
dogs are parallel but distinct. This picture does not work for the com-
bination of (EBO) with the open future: if there were parallel but dif-
ferent worlds at t and earlier for each different outcome of the branch-
growth process, there would be no indeterminacy about how things
would go for a certain tree, since each such world unfolds past t into
streams that are the same vis a vis the branch-growth process for the
tree in question.
Could defenders of (EBR) reply that this argument also betrays
bias, since it is only future contingents that are indeterminate, and it is
precisely their view that whether or not a tree grows b is not a contin-
gent matter? Perhaps it is now determinate that b will grow on the tree
in the same way that it is now determinate that all future samples of
water will have chemical composition H20.
Suppose for definiteness that there are exactly two possible trees
which have the same origin as a certain actual tree and are of the same
biological kind, but one of these trees, A, grows b, and the other, E,
does not. Might it be determinate at t and earlier times that the actual
tree is A rather than E, even though nothing has occurred by t that re-
quires or rules out b's growing (nothing that does not prejudge the
tree's identity)? The answer to this question has to be no, if we think
the future is open in any respect. For if we ask how it could be deter-
minate at t that the tree is A, before any branches grow on it, the only
non-question-begging answer is that it is because the tree grows b at
some time after t. But if the present can acquire determinacy in this
respect from the future, why not in every other respect? (Note that such
backwards acquisition of determinacy is not suggested in the chemical
composition case - we do not think it is now determinate that all future
water will have chemical composition H20 just because in the future,
all water has chemical composition H20.)
A broad range of deviant essentialisms seem to be in tension with
the openness of the future as we have understood it. The only essen-
tialist principles about composite objects that sit well with it and that
are consistent with (1) are principles that focus exclusively on the ini-
tial states of those objects and their ancestry. For any allusion at all to
subsequent states will generate the puzzles about determinacy of iden-
tity that we have just described. 31
Notes
I Assume that x is restricted to organisms which originate from a single entity.
2 See, for example, Mackie [1974], McGinn [1976] and Salmon [1982], Kripke
himself gave "something like a proof' of a related thesis about the matter of which a
table is composed, in endnote 56 of Kripke [1972], where it is printed in
"inexplicably garbled" form (Kripke [1980], p,l), An erratic reader of endnotes, I
was unaware of it until it appeared, corrected, in Salmon [1979], By then I had al-
ready devised a similar argument about acorns and oak trees, which appeared in print
in Forbes [1980a,b],
3 For a defense of such primitive identities, see Chisholm [1970],
4 Exclusion of extrinsic grounding follows from Wiggins' slightly stronger Only
a and b principle (Wiggins [1980], p, 96): "[ ... ] if identity is what we want to eluci-
date, [we need] a criterion which will stipulate that for a relation R to be constitutive
of the identity of a and b, a's having R to b must be such that objects distinct from a
or b are irrelevant to whether a has R to b." So R could not include an "absence of a
better candidate" provision. Wiggins and I would both disagree with Mr. Justice Ot-
ton of the Scottish High Court, in the celebrated case of Middlebridge Scimitar Ltd
versus Edward Hubbard. "Mr. Hubbard [ ... ] was granted a court order enforcing an
agreement under which Middlebridge [ ... ] agreed to buy [the Bentley racing car Old
No.1 from him] [ ... ] for £6.8 million [ ... ]. The case centred on whether Mr. Hub-
bard's car was the one which sped the diamond heir Capt Wolf Barnato to victory at
Le Mans in 1929 and 1930 or whether it had undergone so much rebuilding it was no
longer the genuine article. Middlebridge [ ... ] said it had been promised the Le Mans
winner - and the [Hubbard] Bentley was not that car because it had been completely
rebuilt by a master mechanic" (The Scotsman, 28 July 1990). The crucial considera-
tion in his finding against Middlebridge, according to Otton, was that "there is no
other Bentley, extinct or extant, which could legitimately lay claim to the title of Old
No 1 [ ... ]" And they say analytic metaphysics has no practical application.
S Humberstone refers to intrinsic properties in this wider sense as interior. See
Humberstone [1996], p.239-40 for discussion of this sense, attributed to Dunn
[1990], and the whole paper more generally for an instructive discussion of the in-
trinsic/extrinsic distinction.
6 Fine [1994] identifies the asymmetry in this case with essential/accidental, and
rejects the modal account of essential property on the grounds that it does not dis-
criminate DAz.zE {x}(x) and DAz.XEZ({X}). But if the fundamental asymmetry is
intrinsic/extrinsic, we need a further argument that all essential properties must be
intrinsic before this difference can overturn the modal definition.
7Yablo ([1999], p.486) says that "on almost anybody's account," the zygote Z
from which he (YabJo) developed stopped existing before he started, so descending
from Z is extrinsic to him. Perhaps it is well to separate persons and their bodies, in
which case we can still say that it is intrinsic to Yablo's body to develop from z, since
Yablo's body came into existence with z, even ifit took a while for Yablo to occupy
it. Yablo goes on to say that since part of what it takes to be Yablo is to descend from
z, being Yablo is extrinsic to him as well. This is a sense of extrinsic on which I have
no secure grip (though Schoenberg once said that, since no-one else wanted the job
of being Schoenberg, he had to take it on).
8 A counterexample which might be thought to survive the addition of relational

elements to intrinsic natures involves Felix, a cat which exists in u, and Felix-minus,
that portion of Felix in u which lacks a tail. Let v be a world in which Felix is just as
in u except for not growing a tail. Then Felix-minus@u and Felix@v are numerically
identical, thus by (1), so are their satisfiers. But Felix-minus and Felix are distinct
entities in the domain of u, since only one has a tail. To this I would reply that Felix-
minus@u and Felix@v are not numerically identical. For one thing, Felix@v includes
being a cat, while Felix-minus is no cat. Still, some would regard being a cat as ex-
trinsic to each cat (Yablo, Ibid.). But there is still an intrinsic difference between a
natural entity and one which is a mereological abstraction from a natural entity; it is
not clear that Felix-minus is even an organism.
9 The causal isolation is clear enough in this case, since there is no reason why
an organism originating from a propagule p must causally interact with one originat-
ing from a different propagule q. It is a further problem to give a precise account of
causal isolation that is of use in harder cases. Also, something with a certain origin
cannot be made the best candidate for identity with a certain entity simply by throw-
ing in some causal interaction with its rivals: the causal interaction would have to be
somehow in the nature of the case.
10 Robertson's actual example is essentiality of leaf-color. However, if we
choose a non-exclusive property, insisting on its essentiality as an alternative to
(EBO) will not block every counterexample to (1). For example, we can suppose that
the colors of a, x and y in the argument for (EBO) are all the same; we still get a
counterexample to (1). (I assume that growing b is exclusive.)
II Monovular twinhood is not a counterexample to this assumption. Pace Rob-
ertson ([1998], p.735, n.ll) I would say that the propagules from which identical
twins originate are the two daughter cells resulting from the non-standard mitotic
division of the zygote.
12 Despite the obvious echoes of the Ship of Theseus, I think that organisms
contrast with artifacts in important relevant ways. An organism can persist through a
complete change of its matter. But while a ship may undergo repairs at certain times,
so that ultimately there is a ship whose matter is entirely different from the original
ship's matter, I have never seen a good reason to hold that a single ship persists
through such a process, Justice Otton notwithstanding. Fear of vagueness is often the
main motivation; see the discussion of the Mac of Forbes in Forbes [1987].
13 However, it is not so clear that an appropriate causal isolation condition is
met in this case (cf. note 9). So some might try to defend this kind of extrinsic deter-
mination, as is familiar from the Ship-of-Theseus literature; see, e.g., Garrett [1988],
and the response in Mackie [1989].
14 For example, it seems possible that I might have been an identical twin. But
reflecting on the symmetry of mitotic division, the hypothesis that there is a world
where I am one and not the other of a pair of twins seems no better than the hypothe-
sis that 0 is identical to Lefty or else to Righty. See (Forbes [1980a], pp.353-5) for
further discussion of twinning.
15 Is it also crucial that the zygote the Queen could not originate from is some-
one else's, the Trumans' child's? Even if it were a merely possible zygote, I doubt
that that weakens the pull of Kripke's claims. Hawthorne and Gendler [2000] offer an
origin essentialism "lite" (their (21)) which says that there is no world where the ac-
tual Queen comes from the actual Trumans' daughter's actual zygote and the actual
Trumans' daughter comes from the actual Queen's actual zygote. But this is very
much weaker than the intuition which Kripke's discussion promotes, at least in me.
16 Mackie [1987] endorses extra-strength haecceitism, though without the bene-
fit of supporting examples like McKay's. She seems to agree with my verdict about
the Lefty/Righty case, but argues that there is no reason to insist on parallel treat-
ments of transtemporal and transworld identity (Ibid pp.l97-8). But I would say that
identity is identity. If the thing which is F is identical to the thing which was G
(wide-scope tense) requires grounds, then the thing which is F is identical to the
thing which would have been G if. .. (wide-scope modal) must also require grounds.
17 Exactly one is too strong, since one bare difference can give rise to others, if
the primitively distinct entities are parts of other entities. I ignore this complication.
18 Hawthorne and Gendler [2000] raise the interesting and complicated question
of what happens to the defense of (EBO) in a counterpart-theoretic framework, where
it seems that a two-candidate world would just be a world with two counterparts of
some actual entity, which is relatively unproblematic. But the new argument for
(EBO) just given does not use two-candidate worlds, and in a counterpart-theoretic
framework, shows that the counterpart relation would have to hold in some instances
and fail in others even though there is no difference between these instances with
respect to the factors that ground or determine (degree of) counterparthood. This is
no improvement on ungrounded identity. I hope to pursue these issues, including the
Faith-Hope-CharitylPeter-Paul-Mary case from (Hawthorne and Gendler [2000],
p.293), in another paper.
19 See further Forbes [1994a]. Another proposal is that in certain cases there is
no fact of the matter about transworld identity. But this position does not seem to
change the issues in any significant way (though it does complicate the possible-
worlds semantics).
20 Hawthorne and Gendler ([2000], p.293) argue that "[ ... ] the intuitive strength
of the necessity of origins thesis surpasses that of [predecessor essentialism], so if the
project is to generate arguments in favor of the former, it seems best not to invoke the
latter". This might be so if we were trying to explain why (EBO) is intuitive and
thought that a successful non-debunking explanation would have to access explicit
reasons for holding (EBO) and portray (EBO) as inheriting its intuitiveness from
those reasons. But in general, explaining why something plausible is true may require
us to call upon non-obvious lemmas.
21 I interpret not the first to mean the second or later, excluding not at all. Cer-
tainly, if p is not an m-c propagule, contraction will not tum it into one.
22 Here I am assuming that at least for a range of intrinsic properties, if the con-
stituting matter of p has them, so does p (the primary exceptions would be properties
involving p itself).
23 Vallentyne bites the bullet on this issue ([1997], pp.216-7). Yablo modifies
(4) to get round the problem. According to (4) it is the truth-value of P-xw ~ Pxw'
that is criterial; in the revised version (Yablo [1999], p.492) it is the truth-value of
340 Individuals, Essence, and [dentity
Pxw ~ Px'w' that is criterial, where x' is whatever is constituted in w' by the basic
elements of w that make up x (in w). But this makes the criterion harder to apply. For
example, the new notion of part (p.491) allows w to be a part ofw' so long as there is
some concept of sum such that w' is the sum of certain basic elements and w is the
sum of a subset of those elements. It is not obvious that this will keep the shape of x
the same in wand w' unless we make an ad hoc stipulation that only those notions of
sum that do not allow basic elements to arrange themselves into a different shape and
still be the same sum are to be used. Since Yablo wants shape to be intrinsic (p. 480),
such stipulations are apparently needed. And constituting matter m will still come out
intrinsic to the entity e it constitutes, contra Yablo's intentions, unless the same basic
elements can configure themselves as they are in e without thereby forming m. We
will need a special notion of sum to justify this.
24 This is (c*) of (Francescotti [1999], p.604) except that I have used a second-
order variable X for a function from worlds to sets in place of the rigid class-term 'C'
in (c*), which renders the possibly pointless.
25 Francescotti's final version of his criterion says that P is intrinsic to x iff there
are non-d-relational properties such that x's having P consists in its having those
properties. This does not affect the overgeneration problem.
26 It will not help to restrict Y to those values such that necessarily, any Y is an X
and vice versa. Some would say this means Y = X anyway. And it takes us quite far
from the original intuition behind d-relationality.
27 There is a problem with Francescotti's account of consists in (p. 599), which
makes it symmetric, a view perhaps associated with Hegel: a nation consists in its
people, and vice versa.
28 Excluding infinite regresses, essentialism about number of predecessors de-
termines identity in a recycling sequence so long as we are not given two primitively
different starting entities.
29 An exception is the rather special case of spatio-temporal point of origin. My
most recent discussion of this is in (Forbes [1999], §2). I think this special case has to
be ruled out by independent considerations.
30 There are ways of disputing this, but these workarounds are costs of the view
under discussion.
31 This paper includes some parts of a lecture given at the 22nd International
Wittgenstein Symposium (Forbes [1999]). For discussion on that occasion, I thank

David Chalmers, Kevin Mulligan, Peter Simons and Peter Van Inwagen. In preparing
this paper, I have also been helped by Tamar Szabo Gendler, John Hawthorne,
Kathrin Koslicki, Tom McKay, Teresa Robertson, Nathan Salmon and Stephen
Yablo.
FORBES ON ORIGINS AND IDENTITIES
Penelope Mackie
In several writings Graeme Forbes has argued that a version of

Kripke's necessity of origin thesis for biological individuals, repre-
sented in Forbes's contribution to the present volume) by the principle
(EBO) (Essentiality of Biological Origin), can be defended by appeal
to a principle about identity: that the identities and non-identities of
"composite objects" cannot be "bare" or "ungrounded", but must be
appropriately grounded in other properties. (p.319Y
The argument, outlined in §2 of Forbes's paper, appeals to the
principle that if these identities and non-identities are to be appropri-
ately grounded, there must be non-trivial sufficient conditions for the
identities. It is then argued that this sufficiency principle requires that
the essential properties of these individuals include what Forbes here
calls 'exclusive' properties, where a property is exclusive if and only
if, necessarily, at most one thing possesses it. The principle (EBO) is,
in tum, supported by the claim that it meets the demand for exclusive
essential properties in the case of biological individuals. (EBO) is ex-
emplified by the thesis that a biological organism that originates from
one particular "propagule" (as a human being originates from a par-
ticular zygote, or an oak tree from a particular acorn) could not have
existed without originating from that propagule (e.g., that particular
zygote or acorn).
341
In his paper, Forbes discusses three challenges to this argument for

the necessity of origin. One concerns the interpretation of his principle
that identities and non-identities must be appropriately "grounded". In
previous work, Forbes has suggested that the grounding must be con-
fined to properties that are "intrinsic". In § I of his paper he responds to
the objection that his appeal to intrinsic properties is illegitimate.
The second challenge that Forbes confronts is that his requirement
that "composite objects" must have exclusive essential properties (on
which his argument for the necessity of origin relies) must be mis-
taken, since it has unacceptable consequences. The difficulty is illus-
trated here by what Forbes calls "the recycling problem", which ap-
pears to show that in at least some of the relevant cases there are no
suitable candidates to play the role of exclusive essential properties
(§§4-7).
The third challenge is what Forbes calls "the bias problem" (§§3,
8). This rests on the fact that Forbes's argument for (EBO) is, as it
stands, incomplete. It needs to be supplemented by an explanation of
why, of the many exclusive properties that an organism has during its
existence, the properties of its origin that are selected by (EBO) should
be invoked to play the role of its exclusive essential properties. 3
Because of limitations of space, I confine my discussion to the first
two issues, after a brief rehearsal of Forbes's argument for (EBO).
1. THE ARGUMENT FOR (EBO)
As indicated above, Forbes's argument for (EBO) turns on the princi-

ple that identities and non-identities must be appropriately grounded
(the grounding principle), which he takes to require that identities
across possible worlds be grounded in properties that are non-trivially
sufficient for those identities (the sufficiency principle). I use 'non-
trivially sufficient for identity with x' to mean "sufficient for identity
with x without logically entailing identity with x" (cf. Forbes, § I). 4 As
Forbes points out, if a property is to be genuinely sufficient for identity
with some individual x (in the relevant sense of 'sufficient in all possi-
ble worlds') then it must be a property that is "exclusive", in the sense
that "necessarily, at most one thing possesses it" (p.322). It seems an
Penelope Mackie, Forbes on Origins and Identities 343
uncontroversial step from this to the conclusion that if a set of proper-

ties is to be sufficient for identity with x, it must include at least one
property that is exclusive in this sense. A further argument suggests
that if the sufficiency principle is accepted, any set of properties genu-
inely sufficient for identity with x must include some property that is
not only exclusive, but also essential to being x. For without this fur-
ther condition, there is no obvious barrier to the construction of an em-
barrassing "two-candidate" possible world, containing two distinct in-
dividuals each of which satisfies a set of conditions purportedly suffi-
cient for being x (pp.322-323). Forbes argues that since (EBO) attrib-
utes to biological individuals exclusive essential properties to do with
their biological origins, it can be justified by appeal to this argument
(§2).5
2. INTRINSIC PROPERTIES AND IDENTITY -RELEVANT

PROPERTIES
Forbes's grounding principle is more demanding than might at first

appear: not every proposed candidate for grounding satisfies the prin-
ciple as Forbes interprets it. He suggests, as an initial approximation,
that the grounding must be in properties that are "intrinsic" rather than
"extrinsic".
This means, among other things, that whether or not x = the rp should not
tum on the presence or absence at the relevant time or world of some en-
tity that is causally isolated from the rp. (p.320)
The restriction on the grounding principle leads to a corresponding re-

striction on Forbes's "sufficiency principle". Accordingly, Forbes ini-
tially characterizes his sufficiency requirement in terms of the principle
he labels '(1)': to the effect that indistinguishability of intrinsic nature
(across possible worlds) is sufficient for identity across possible
worlds (p.320).
This restriction on the grounding principle (and the sufficiency
principle) is important, since without some such restriction Forbes's
argument for (EBO) will not go through. In particular, if we are per-
mitted to satisfy the grounding principle by making it a component of

every sufficient condition for identity with x that there exists no com-
petitor for identity with x, then Forbes's argument for (EBO) is under-
mined. For we can thereby block the generation of embarrassing 'two-
candidate' worlds without appeal to any principle such as (EBO).
However, as Forbes acknowledges, his appeal to the distinction be-
tween 'intrinsic' and 'extrinsic' properties is problematic.
First, as he points out, if 'intrinsic' means 'non-relational', the
principle of the sufficiency of intrinsic nature for identity is not very
plausible (p.320). Worse, even if it were plausible, this sense of'in-
trinsic' would not serve Forbes's purposes, since the fact that an indi-
vidual a developed from a certain zygote distinct from a surely repre-
sents a relational property of a. In response, Forbes suggests that there
may be a broader sense of 'intrinsic' that is both plausible and suitable
for his purposes, according to which some relational properties are part
of a thing's intrinsic nature while others are not.
However, Forbes does not insist on this point. Instead, he suggests
that it does not really matter whether the term 'intrinsic' is appropriate
to the broader notion that he needs. For we can replace the original
formulation of the sufficiency principle (I) with a formulation accord-
ing to which indistinguishability in respect of "identity-relevant" prop-
erties is sufficient for identity across possible worlds (p.321). This
does not trivialise the principle, says Forbes, as long as it is accompa-
nied by the proviso (which is needed in any case) that the properties in
question must exclude properties that logically entail (and hence are
merely "trivially sufficient" for) identity with the thing in question. For
it is then a substantial (and disputed) claim whether, in the case of an
individual x of one of the sorts that Forbes is discussing, there are any
properties that are sufficient for, without logically entailing, identity
with x: that is, whether x possesses any (non-trivial) identity-relevant
properties at all. 6 And even if it is agreed that x must have some such
(non-trivial) identity-relevant properties, it will be a substantial and
potentially controversial question what these properties are.
However, Forbes's substitution of 'identity-relevant' for 'intrinsic'
in his sufficiency principle (I) makes his argument for the necessity of
origin vulnerable in a way that he does not appear fully to acknowl-
edge. Evidently, this revised version of the sufficiency principle (I)
does not by definition preclude the suggestion - which Forbes needs to

reject - that x's identity-relevant properties may include the presence
or absence of some individual causally isolated from x. Yet, for exam-
ple, in note 9 and the text of §2 that accompanies it, he appears to be
assuming that the revised version of the sufficiency principle still rules
out, by definition, the grounding of identity in the presence or absence
of causally isolated competitors. And this is not so. I return to this
point in §4 below.
3. RECYCLING AND THE SEARCH FOR EXCLUSIVE

ESSENTIAL PROPERTIES
An illustration of what Forbes calls "the recycling problem" is pro-

vided by the (admittedly recherche) possibility that the 'recycling' of
matter could produce intrinsically indistinguishable and matter-sharing
zygotes that come into existence at different times, each of which has
the potential to generate a new individuaP
Forbes admits that the possibility of such recycling presents prob-
lems for his account (§§4-7). The most recalcitrant of these - which he
attributes to McKay [1986] - may be represented as follows (cf.
Forbes, §5). (EBO) is said to rest on the principle that for "'composite
objects '" - "things which in some good sense come from or are com-
posed of or constructed from other things" - identities and non-
identities must be appropriately grounded (pp.319-320). Presumably
zygotes are composite objects in the relevant sense. But if a zygote p is
not identical with a qualitatively similar zygote q that is produced, by
recycling, from the matter of p, how can the non-identity of p and q be
appropriately grounded? If q does not (essentially) depend for its exis-
tence on the previous existence of p (Forbes calls this negative claim
'the independence thesis '), then, since q may apparently be exactly like
p in all relevant respects except for the fact that it is preceded by p, it
appears that any proposed sufficient condition for identity with p will
be satisfied in some possible world by q, contradicting the assumption
that it is genuinely sufficient for identity with p. We shall get the result
that p and q have no exclusive essential properties that distinguish one
from the other, and any properties that distinguish p from q will be ac-
cidental, not essential. If so, any way that p could have been is a way
that q could have been, and this is a counterexample to the sufficiency
principle (1). Further, if the identities of zygotes are not appropriately
grounded, it is plausible to say that the identities of the organisms that
come from those zygotes cannot be appropriately grounded in the
identities of the zygotes. The possibility of recycling, far-fetched
though it is, thus poses a serious problem for Forbes. It seems futile to
suggest that (EBO) respects his grounding principle and his sufficiency
principle unless this problem can be solved.
To meet this difficulty, it seems that Forbes must deny that the zy-
gotes p and q are distinguished only by their accidental properties. An
obvious move is to reject the 'independence thesis' mentioned above,
and attribute to q, as an essential property, the property of being pre-
ceded by p.8 Since the problem can be generalised to include recycling
sequences of any length, a comprehensive solution along these lines
requires what Forbes calls "predecessor essentialism": that "an entity's
predecessors in a recycling sequence are its predecessors in every
world in which it exists" (p.328). Since any essential property of x is a
component of every sufficient condition for being x, predecessor es-
sentialism prevents recycling from generating "two-candidate" worlds.
Of course (if it is to do the work required of it), predecessor essential-
ism must be taken to imply not only that it is essential to the first zy-
gote in a recycling sequence that it have no predecessor, but also that
in the normal case, where there is just one zygote and no recycling, this
zygote's having no predecessor is one of its essential properties.
One difficulty concerning the appeal to predecessor essentialism is
its apparent conflict with the principle that identities and non-identities
must be grounded in intrinsic properties (§6). However, since Forbes
has already conceded that identities may be grounded in properties that
are not, strictly speaking, intrinsic, he takes it that the principal chal-
lenge he has to face is, not that the number of predecessors an item has
is not one of its intrinsic properties (for he concedes that it may not
be), but that it is not even "identity-relevant" (§7).
To illustrate this challenge, Forbes considers examples involving a
sequence of ships each of which is originally constructed of the same
particular matter m according to the same plan I. Call each such ship an
'm-/ ship'. The principle of predecessor essentialism, applied to ships,

implies that if two ships are in fact the first and second such 'm-/'
ships in a recycling sequence, it is essential to the first ship that it be
the first, and to the second ship that it be the second. 9 The principle
also implies that if a ship is in fact the first and only m-/ ship, it is es-
sential to it that it be the first (although not that it be the only) m-l ship.
Forbes acknowledges that predecessor essentialism, and the asso-
ciated version of the sufficiency principle, have counterintuitive con-
sequences (p.333). For example, predecessor essentialism implies that
if a ship X built in the year 2000 is in fact the only m-l ship, then in a
possible world w * in which an m-/ ship is built in 2000, and another m-
/ ship is built a thousand years earlier, the later ship in w * cannot be
identified with X. Moreover, if being an m-l ship with no predecessors
is a sufficient condition for identity with X, it follows that we must
identify X with the earlier of the two m-l ships in w*, rather than the
later one. IO And this verdict, Forbes admits, is very unattractive.
Forbes's response is to suggest that this shows, not that being the
first ship is not identity-relevant, but that "other identity-relevant fac-
tors are at work which block default identification of the first m-/ ship
with X' (p.333). In particular, "the further removed in time the origi-
nation of the counterfactual ship from the origination of X, the weaker
the claim Xhas on it" (Ibid.). He introduces a qualification to the suffi-
ciency principle that is designed to accommodate this intuition, ac-
cording to which it appears that match of time of origin can sometimes
override (or 'trump') sameness of position in a (recycling) sequence in
detennining identity. 11
Forbes's concession appears to amount to this. Two distinct fea-
tures that X has (being the first m-/ ship, and originating in about 2000)
are both 'identity-relevant', in the sense that the possession or non-
possession of these features by an item in a possible world may be
relevant to whether that item is identical with X. Yet neither feature
represents a property that is both exclusive and essential to X, and
neither without the other yields (even in conjunction with the property
of being an m-/ ship) a sufficient condition for being X. 12 However, this
modification to Forbes's sufficiency principle has potentially embar-
rassing consequences, as the following example illustrates.
Suppose that in the actual world (WI) there is just one m-/ ship, SJ,
which comes into existence in the year 1900. Then, given a possible
world W2 in which there are two m-/ ships, one (S2) which comes into
existence in 1895, and a recycled m-/ ship (S3) which comes into exis-
tence in 1905, Forbes's principles imply that it is the first ship in W2
that is identical with SI. \3 If we suppose, in addition, a possible world
W3 in which there are two m-l ships, S4 and ss, one of which (S4) comes
into existence over a thousand years before 1900 (e.g., in 850 A.D.),
and a recycled m-l ship (ss) that comes into existence in 1905, we may
(according to Forbes) identify the second ship with the actual SJ, in
spite of the fact that it lacks the property that SI actually has of being
the first m-l ship.
We thus have two possible worlds, W2 and W3, each of which con-
tains an m-/ ship that originates in 1905. These ships (S3 and S5), in
spite of their similarities, are not the same ship, since one is identical
with SI and the other is not. But now we must ask: what, according to
Forbes, grounds the non-identity of S3 with ss? The only differentiating
feature to which Forbes can appeal is that S3 's predecessor in W2 was,
while S5'S predecessor in W3 was not, built at approximately the same
time as SI was built in WI. However, to treat this feature as what
grounds the non-identity is, on the face of it, a paradigm of the type of
"best-candidate" reasoning that Forbes must reject if his argument for
the necessity of origin based on the grounding principle is not to be
undermined. 14
The only possible defence to this objection is that this version of a
"best-candidate" theory of the identities of composite objects is supe-
rior to the rival versions that Forbes must reject. In the final section of
this paper I consider whether this defence can be sustained.
4. ORIGIN, COMPETITION, AND IDENTITY-RELEVANCE
In response to the objection to his appeal to intrinsic grounding, Forbes

has conceded that the identities of composite objects need not be
grounded in intrinsic properties (Forbes, § 1). In response to the recy-
cling problem, he appears to have conceded that the grounding of the
identities of composite objects does not require exclusive essential

properties (§7). However, to make this second concession is surely to
abandon the original argument for necessity of origin principles such
as (EBO).
However, as I interpret his response to the recycling problem,
Forbes has also suggested a modified account of the grounding of the
identities of composite objects that may be regarded as providing a
novel basis for necessity of origin principles such as (EBO). For ex-
ample, according to the modified account described above, although
the ship s[ has no exclusive essential properties (and, afortiori, no ex-
clusive essential properties that concern its origin) it is nevertheless an
essential property of s[ that it is an m-I ship: that is, a ship originally
constructed from the particular matter m according to plan I. Why so?
Because according to Forbes's modified account, any ship that satisfies
the conditions for identity with s[ must be an m-I ship, even though
many of its other characteristics may vary, including its time of origin
and the number of its predecessors. 15
But how is this crucial restriction on the conditions for identity
with s[ to be justified? We may illustrate the problem as follows, refer-
ring to the example I developed at the end of the previous section. As
we have seen, according to the modified account, the question whether
a recycled m-I ship that comes into existence in 1905 in some possible
world w* is identical with s[ (an m-I ship that actually comes into ex-
istence in 1900) may depend on the time at which another ship in w*-
the predecessor of the recycled ship - was constructed. For simplicity,
let us say that the recycled ship and its predecessor are "competitors"
for identity with S[.16 Since both "competitors" are m-I ships, this case
(call it 'Case 1') involves no violation of a version of the necessity of
origin thesis according to which being an m-I ship is essential to being
St. But now, if 'competition' for identity with s[ is allowed at all, why
should the competition be restricted to m-I ships? For example, why
shouldn't the question whether s) is identical with an m-I ship that
comes into existence in 1905 in w * depend on the properties of a ship
that is constructed in w * in 1895 from matter entirely distinct from m?
We can imagine a case ('Case 2') where, except for this difference in
material origin, the properties that the 1895 ship has in w * are such
that it resembles s) (as s) is in the actual world) much more closely
overall than does the 1905 ship in w*. With what right can Forbes in-
sist that, in spite of this, the 1895 ship in Case 2 is not a genuine
"competitor", and that, if either w* ship in Case 2 is identical with s"
it can only be the 1905 ship?
Perhaps Forbes might respond that in Case 1, the "competitors"
are causally connected with one another, because they share their
original matter m, whereas the putative "competitors" in Case 2 need
have no causal connection with one another at all. However, there are
two problems with this response. First, Forbes has provided no argu-
ment for the claim that an appeal to causally isolated competitors is
illegitimate in a way that that an appeal to causally connected com-
petitors is not (cf. §2 above). Secondly, even if it is accepted that the
distinction between causal isolation and causal connection is relevant
here 17, it is not true that the "competitors" in Case 2 must be causally
isolated from one another. For example, we may fill out the details of
Case 2 in such a way that the 1905 ship is deliberately constructed by
its designer to be a replica of the 1895 ship, with the result that the
ships are causally connected although they do not share their original
matter.
I have claimed that if he concedes that the identities of composite
objects need not be grounded in exclusive essential properties, Forbes
has, in effect, undermined his original argument for necessity of origin
principles such as (EBO). However, he appears to suggest that its place
may be taken by a "best-candidate" account of the grounding of the
identities of composite objects of which the necessity of origin is a
consequence. It remains to be seen whether a justification for the de-
tails of this "best-candidate" account can be provided that does not beg
the question in favour of the necessity of origin principles that Forbes
seeks to defend. I think that this represents a formidable challenge to
this new argument for the necessity of origin, although, of course, I
have not shown that the challenge cannot be met
Notes
I "Origins and Identities". All references are to this paper, unless otherwise indi-
cated.
2 See, in particular, Forbes (1980a] and [1985], Chapter 6. See also Forbes
[1986] and [1994a].
3 See, for example, Robertson [1998], p.741, Yablo [1988], and my [1987],
p.186. Cf. my [1998], pp.64-5.
4 From now on I shall usually drop the explicit qualification 'non-trivially', tak-
ing the qualification to be understood.

S A natural objection is that (EBO) fails to assign exclusive essential properties,
because the phenomenon of twins shows that the property of originating from a par-
ticular propagule (e.g., a particular zygote) is not an exclusive property. Forbes's
response is to say that when a zygote splits to produce 'identical' (monovular) twins,
each of these twins has, as its 'propagule', not the zygote, but the daughter cell from
which it immediately came, and is thus provided by (EBO) with its own exclusive
origin property (Forbes, note 11).
6 The view that Forbes calls 'extra-strength haecceitism' (§5) involves the denial
of the principle that identities across possible worlds must be underwritten by prop-
erties that are non-trivially sufficient for the identities.
7 Cf. Robertson [1998] and McKay [1986].
8 Forbes does not mention an alternative: that p and q might be distinguished by
treating the exact time at which each originates as one of its essential properties. Pre-
sumably he ignores this option because he takes it to be unacceptably counterintui-
tive.
9 Although ships are not biological organisms, questions about their identities
are relevant here, since Forbes's principle (EBO) is alleged to depend on a more gen-
eral principle concerning 'composite objects', of which ships provide an example.
10 The example is due to Hawthorne and Gendler [2000], p.293. One could ac-
cept predecessor essentialism without accepting the associated sufficiency claim: i. e.,
the claim that if an m-/ ship x has a certain number of predecessors, then being an m-/
ship with that number of predecessors is sufficient for identity with x. However, if
predecessor essentialism is to do the work that Forbes requires of it, the additional
sufficiency claim is required.
II As I shall use the expression 'position in a sequence', a thing's position in a
sequence is independent of how many successors it has. For example, a ship that is
the first and only m-/ ship has the same position in a sequence (in the relevant sense)
as a ship that is the first in a sequence of 100 recycled m-/ ships.
12 Some of Forbes's remarks may suggest that he has in mind a different modifi-
cation, according to which predecessor essentialism is retained, but a ship's approxi-
mate time of origin is also one of its essential properties. However, this would not
accommodate Forbes's agreement with Hawthorne and Gendler's claim that there can
be a case where the second m-l ship in a recycling sequence could have been the first
and only m-/ ship (the case of Z; Forbes, p.333).
13 This assumes that a five-year gap is sufficiently small to make the times of
origin "close enough" in the sense Forbes intends (§7, p.334). If not, just rerun the
example with a smaller temporal gap.
14 Cf. §2 above. The best-candidate reasoning is also, of course, congenial to the
counterpart theorist. Cf. Lewis [1986], Chapter 4.

IS Evidently, if this account works for ships, it may be adapted, mutatis mutan-
dis, to provide a basis for necessity of origin principles for other composite objects
that does not require that those objects have exclusive essential properties.
16 To talk of 'competition' for identity is to run the risk of apparent conflict with
the principle of the necessity of identity. However, I hope that my use of this lan-
guage can be interpreted in a way that avoids any genuine conflict with this principle.
17 Elsewhere (Mackie [1987], §7, and Mackie [1989], I have suggested that it
may be more plausible to claim that an identity can depend on the presence or ab-
sence of a causally connected individual than to claim that it can depend on the pres-
ence or absence ofa causally isolated individual. Cf. Forbes, note 9.
Part Four
Time and Persistence

LEIBNIZ, COMPOSITE SUBSTANCES AND
THE PERSISTENCE OF LIVING THINGS
Anthony Savile
One of the several points at which Leibniz's metaphysics appears to

lose contact with common sense and verge on fantasy is where he says
that living things such as men or fishes or plants never truly die but are
in their way no less eternal than those simple substances that are the
fundamental monads of his system (e.g., Monadology §76). Here I
shall argue that, appearances to the contrary, it is plain enough why he
thinks this and also that, strange as it may seem, comparatively little
emendation to his way of thinking is needed before we find an inter-
esting convergence between his way of understanding the natural
world and our own.
What are living things? They are unities consisting of a dominant
monad and an organic body that is particularly tied to it, a body by
means of which the dominant monad perceives the world around it and
through the voluntary control of which it is able to bring about the sat-
isfaction of its desires (Monadology §63). Every monad has to be
dominant over some such body because only if it is will it be possible
for it to enjoy perception and appetition, and monads have to enjoy
those things since they are the sole qualities that simple substances can
possess. Lacking perception and appetition a monad would have no
qualities, and just as qualities cannot exist in the absence of some sub-
355
A. Bottani et aI. (eds.), Individuals, Essence and Identity, 355-367.
stance of which they are qualities, so in the absence of any qualities a

substance could not exist at all (Monadology §8).
The body that a given monad dominates consists of other monads,
although which monads these are will be constantly change from mo-
ment to moment. Leibniz observes that there is Herac1eitean flux here
and that no portion of matter (viz. collection of subdominant monads)
is assigned to a given dominant monad throughout the latter's exis-
tence (Monadology §71). The body that a cabbage or an ant or a man
possesses retains its individual stability on the Leibnizian conception
of things despite the fact that its elements are in flux. To us there is
nothing strange about this since we see ourselves as bodily continuants
of a kind whose existence is unthreatened by the metabolic process.
The point at which we are liable to encounter some difficulty with
Leibniz's introduction of living things is in adequately distinguishing
them from what he calls ''phenomena bene jundata". These latter are
material things which do not qualify as genuine substances in that they
have no true unity as is made manifest by their tendency to dissolve or
come apart. A block of ice, a flock of sheep, a wormy cheese are ex-
amples that Leibniz offers of such things: they are composed of simple
substances all right (there is nothing else that they could be composed
of), and they are real enough in that they borrow their reality from the
reality of their constituting simples, but their unity is a purely conven-
tional or as Leibniz sometimes says, a "mental" matter. That a large
group of monads makes up a single block of ice depends on nothing
more than our determination to count that mass as a single thing, and a
genuine substance, be it simple or complex, has to have a unity that is
natural, and not conventional in the way that these things are.
The difficulty that Leibniz appears to encounter - perhaps it ap-
peared to him that he encountered it - is to figure out how living things
consisting of some dominant monad together with an organic body
could be genuine substances, how they could enjoy any unity that war-
rants us in thinking of them as anything more than phenomena bene
jundata. To take an example that Leibniz discusses in his correspon-
dence with Arnauld, it would seem as though the body that is more or
less under my voluntary control during my lifetime ceases to be so
when I die and persists a while as a corpse l . But in that case it does in-
deed looks as if the dominant monad and its organic body have come
Anthony Saviie, Leibniz and the Persistence of Living Things 357
apart, and hence together lack precisely that unbreakable unity which
alone would entitle a good Leibnizian to account the original compos-
ite living thing a true substantial unity. Setting aside the soul or the
dominant monad, the body itself is surely no more than a conventional
or phenomenal unity, just like the block of ice or the flock of sheep or
the wormy cheese. If this is so, it is bound to strike us that Leibniz's
determination to pick out certainly men and, he is inclined to think,
many other naturally occurring things as truly substantial composites
that are metaphysically quite distinct from aggregative phenomena like
the cheese is under threat. To put it starkly, how could there possibly
be such real things whose high grade, substantial, reality depends on
their possessing a body whose reality itself is no more than low grade
and phenomenal?
When Leibniz considers the example of the cadaver he is quite ex-
plicit that there we do have to do with a phenomenon and not a com-
plex substance. And given that, it may strike us as astute enough on his
part, but ultimately, surely, quite hopeless, to say that living things are
indeed composite substances and that they never truly die. For even if
that might serve to mark out a possible relation between dominant mo-
nads and bodies that is distinct from the relation between the self and
the body that becomes a corpse, it is all too likely that we shall say on
the basis of straightforward observation that such a relation is never
realised, and hence that neither cabbages, nor ants nor even men are
instances of it. The introduction of the everlasting body constantly
conjoined with the indissoluble simple soul like substance is likely to
seem nothing more than a kind of metaphysical wishful thinking. In-
tellectually speaking, it looks quite inapplicable to the actual world.
Such an accusation is mistaken, I think, and it does scant justice to
the subtlety of Leibniz's thought. To set things right we need only re-
flect a little on the idea of a monad's organic body. In our perceptual
engagement with the world we are sensitive to far off events through
the way in which mechanically they impress themselves or leave traces
on other bodies, which in turn transmit that information to yet other
bodies (Monadology §61). A chain of such transmitted impressions
terminates at the point at which a simple soullike monad is immedi-
ately sensitive to information that is mechanically inscribed on a par-
ticular body of matter (Monadology §63). The body of which the mo-
nad is thus immediately aware in perceiving the world around it is its

own body. Similarly, in realizing my desires I bring about changes in
the world around me, often changes that take place at some distance
from me. So in order to satisfy my thirst, say, I have to bring it about
that water flows from a tap in some room in the house some distance
away from me. To do this I have to engineer changes in the world be-
tween myself and the tap, perhaps by getting you to go to the kitchen
and tum the tap on. And this I do through some change in a body over
which I have direct control, exercised on this occasion perhaps by
moving my lips, tongue and mouth as I ask you to fetch me a glass of
water. So, as in the case of perception, desire also can only be satisfied
through the monadic self possessing a body over which it has unmedi-
ated control. Such is the Leibnizian picture.
Now, it is not so very far from here to saying that my organic body
will be whatever body of matter it is that in perception and appetition I
have immediate awareness and control of. This it seems is just what
Leibniz does want to say, since that, and as far as I can see that alone,
is what entitles him to reject the idea of metempsychosis, and with it
Locke's fantasy of the prince and the cobbler exchanging bodies (Cf.
Nouveaux Essais, II.xxvii.14-15). If the dominant monad that is me at
one time dominates a body that we think of as that of a cobbler and
then later on comes to dominate a mass of monads. that act in a more
princely way, the right thing to say will not be that the dominant mo-
nad has come to control a different body, but that the body that the sin-
gle self controls is unchanged in being numerically the same one, but
has adapted itself to qualitative changes in the sorts of desire and per-
ception that are now enjoyed by the one persisting self. Metamorphosis
of this kind there may be (and that Leibniz approves of) but metem-
psychosis, no (Monadology §72).
To get to the point of affirming that the body is everlasting and to
understand Leibniz's motivation for speaking of the cadaver as he does
we have to remind ourselves of the everlasting nature of the simple
substances that make up the world. Their simplicity ensures that they
do not perish or come into being in natural ways. For they have no
parts, and when things (aggregates) perish or come to be in the way of
nature they perish and come to be by dissolution and by combination
of parts. The everlasting character of these simples, however, is not an
Anthony Savile, Leibniz and the Persistence of Living Things 359
absolute metaphysical necessity, for no contradiction is involved in

supposing that within the temporal order of things God might annihi-
late some and create others. That they enjoy temporally unending ex-
istence is however a hypothetical necessity, underwritten by the princi-
ple of sufficient reason, since the world that actually exists is the best
world that it was open to God to have created, and any alteration to its
original constitution that might be wrought by further creation or novel
annihilation would imply that the original set he decided to create was
less good than it might have been. So Leibniz will say that it is a con-
sequence of the principle of sufficient reason that monadic simples are
imperishable.
Now, we have seen that a simple substance must possess an or-
ganic body throughout its existence, and also that whatever functioning
body it possesses and immediately controls at a given time will be the
very same body as it possesses and immediately control at any other
time. Once that is in place we can infer that all simple substances, per-
sisting everlastingly as they do, must possess bodies that themselves
persist eternally. To suppose this is not just metaphysical wishful
thinking; it is the direct consequence of the underlying principles gov-
erning the way in which the natural world is conceived of. Given that,
we can see that Leibniz has no option but to treat the matter of the
corpse, which after all precisely no longer serves the perceptual and
appetitive needs of a dominant monad, as a phenomenal unity. That is,
it is no more than a quantity of matter sloughed off in Heracleitean
fashion by a dominant monad whose new needs enable it to dispense
with its services and for the sake of which needs it must now have
taken on some other quantity of matter in the fulfillment of the internal
principle that governs its development. Seen from this point of view,
the mistake we are prone to make in the ordinary way in which we
think of the dead body is to suppose that it is the very same body as
was previously alive, but we can see now that from Leibniz's point of
view this must be a mistake. The dominant monad that was once
served by that mass of matter is still around and is still necessarily en-
dowed with its own organic body, which may have shrunk or otherwise
changed its demeanor in surprising ways, yet the shrunken body that it
now controls has to be (a stage of) the very same organic body as was
previously made up of the mass of matter that now lies inert upon the
ground. In that situation the corpse cannot possibly be identified as a

late or final stage of the living thing's body. For that reason its inert
existence provides no reason whatever to contest Leibniz's claim that
living things as well as simple substances are impervious to death
when that is taken in the strict sense that implies cessation of exis-
tence.
Has enough been said yet to allow us to fend off the worry that
Leibniz's metaphysics ultimately suffers from having to regard living
things, cabbages, ants and men and so forth, as no more than phenom-
ena, things whose substantial nature is cast in question through essen-
tially depending on material things (viz. their bodies) that enjoy no
more than a phenomenal and conventional unity? Providing only that
we take sufficient care how we use the notion of the organic body I
believe the answer is that it has. What we have to avoid is the tempta-
tion of thinking that the body of any living thing can be identified in-
dependently of its dominating monad. Thus, when Leibniz says that a
living organism is a dominant monad together with its organic body, as
he does at Monadology §63, we should not suppose that that is a dual-
istic observation, as it might be taken to be by anyone brought up
strictly in the Cartesian tradition. For once one does that it will be hard
not to find oneself in the position of one who thinks that the living
body enjoys sufficient continuity and material congruence with the
dead body to count as being at least the very same body, even though it
has changed its state. This we have seen Leibniz wants at all costs to
avoid. 2
To avoid this danger, the organic body is essentially identified in
terms of the dominant monad that it serves, and in consequence cannot
exist except in conjunction with that monad. In this way it forms an
unbreakable unity with it. The dominant monad cannot exist without
its body (for then it would be incapable of perception or appetition),
and equally the body cannot exist except as serving that controlling
monad since it is identified precisely as that body of matter which fa-
cilitates those two fundamental functions in the individual case. But
being an unbreakable unity is precisely what distinguished a Leib-
nizian substance from an ens per accidens or a well-founded phe-
nomenon.
Anthony Savile, Leibni= and the Persistence of Living Things 361
It might seem that there are nonetheless two significant differences

between the primary substances ofLeibniz's ontology and these, to our
eyes, more natural ones, differences that might call in question my in-
terpretation of the latter as genuine substances rather than some lesser
sort of aggregate. In the first place they have parts - which suggests
that their unity is perhaps not so unbreakable after all; and then at the
end of the day, and Leibniz to the contrary, their very perishability
seems to cast that unity in doubt. (Of course I do not mean that Leibniz
will accept this - I have aimed to show how on his principles he need
not - but that issue is one we shall have to confront sooner or later). To
move forward, let me first say something about the implications of
living things having parts.
As far as the parts of a compound substance go, the dominant mo-
nad and its facilitating body, we have just seen that these cannot come
apart. The parts that are at risk of perishing can only be parts of that
body (since after all the dominant monad is simple and has no parts).
Now, of course, my body as ordinarily understood can shed or be de-
prived of its parts, as happens when I lose my hair, graze my knee or
more seriously undergo an amputation. However as far as Leibniz is
concerned those sorts of events, unwelcome though they may be, pose
no threat to the existence of the bodily whole. They are simply in-
stances of Heracleitean flux; the substance casts off (willingly or not)
these material bits and pieces as does a snake sloughing off its skin or
a butterfly breaking out of the chrysalis (Monadology §74). As far as
the organic body goes, that remains a perfect whole under these condi-
tions since it preserves just that material integrity that is needed to en-
able its dominant monad to realise the perceptions and desires that its
inbuilt principle of action requires it to realise. 3 Despite their ontologi-
cal difference, it is instructive to compare these changes of the body to
changes in the qualities that a simple substance undergoes from time to
time. As far as Leibniz is concerned a simple substance, while having
no parts does have qualities or properties that are constantly changing
and whose changes are entirely consistent with the substance's persis-
tence through time. Now a composite substance can undergo changes
of its bodily parts in the form of gain or loss and neither threaten the
identity of the whole. If we assume that it is correct to identify the sub-
stances we all are as individual human beings then no matter what mis-
fortunes may befall us in the physical domain we shall remain whole

human beings, even though the sorts of thing that we can do will be
conditioned by the form that our body has from time to time. That Paul
Wittgenstein could no longer play Ravel's first piano concerto after the
loss of his right arm did not make it any less true that the post-1918
Wittgenstein was as complete a human being as his pre-1918 self. The
loss of bodily parts that composite substances can suffer is thus no
threat to Leibniz's idea that substances unlike other aggregates enjoy
real unbreakable unity.
We may for the sake of contrast reflect on an example of the way
in which a mere aggregate, a Leibnizian phenomenon, does not enjoy
any such unbreakable unity. Consider a flock of sheep. As its members
are sold off by Farmer Jones and not replaced there comes a time when
there is more point to saying that the flock has been dispersed than that
it is still in existence. In this way the loss of parts can imply the disap-
pearance of the flock. Equally, if Jones augments his flock, the addi-
tion to it of new members will after a while encourage us to say that
the flock is a new one despite containing all its original woolly mem-
bers. The addition or subtraction of parts often introduces (quite con-
ventionally of course since we have to do with phenomena, not sub-
stances) new entities. Similar considerations will apply in the case of
other Leibnizian phenomena such as the block of ice or the wormy
cheese.
I said at the start that a relatively slight adjustment to Leibniz's
way of thinking secures an interesting convergence with our own and
that in the light of that his metaphysic can be appreciated as being fun-
damentally less fanciful than we are in the habit of supposing it to be.
Something must be said about this before I can respond to the second
threat I mentioned to his introduction of composite substances. The
issue I am here concerned with is the way in which Leibniz sees com-
posite being necessarily composed of simples, and those simples being
of necessity immaterial ones.
In the first place, we should bear in mind that it is only the com-
mitment to simples, and simples which are indestructible in the natural
course of events that makes organic things for Leibniz naturally inde-
structible. If the dominant monad in any such case were by the exercise
of God's miraculous powers to cease to exist then the associated sub-
stantial whole would also cease to exist, indeed in the case of a rational
intelligence not only would the self have ceased to be, but with that
both the man and the man's body would also have come to grief. What
is left over, the cadaver, is as we know now distinct from the man's
body, so with the disappearance of the one we have the disappearance
of the three (the two elements and the whole).
What makes for the natural indestructibility of simples is precisely
their having no parts, and what makes for the dependence of com-
plexes on simples is the thought that were things otherwise all sem-
blance of genuine unity would be lost. So if it could be put to Leibniz
that it is available to him to think of nature as containing genuine re-
silient (possibly even unbreakable) unities without thereby needing to
fall back on unextended immaterial simples, the way would be open
for him to abandon his claim that nature's compound substances are in
their way no less everlasting than the simples that he finds himself
obliged to posit. The natural eternity of the embodied cabbage or ant or
man could then be quietly dropped, with the clear additional advantage
that we might thereby disengage Leibniz from the taxing task of ac-
counting satisfactorily for the existence of material things in terms of
the aggregation of immaterial elements, a topic with which I have stu-
diously avoided all engagement.
In treating of Leibniz's understanding of bodily identity through
time I have drawn on the idea of a certain material mass serving a cer-
tain function. What makes a particular mass of monads a human body
is that that mass is what the dominant monad needs to draw on in order
to realise with the particular clarity that distinguishes it from other
monads its peculiarly human perceptions and its more or less pressing
human appetitions. 4 What we draw on in determining that the organic
body that is made up of this particular mass of monads is the same as
the organic body made up at some different time of that other mass of
monads is whether those two masses serve the needs of the same
dominant monad in its psychic economy. If the same mass of monads
comes to serve a different master, then it will constitute different bod-
ies at those different times. No mass of monads can immediately and
directly serve two or more masters at once, though there is no impedi-
ment to the material mass that makes up the body of one complex sub-
stance being for a while part of the organic body of another such com-
plex substance.
To my mind, what is suggestive about this is that since that way of
thinking about the body underlies the overall theory, even though
Leibniz never did so it must be open to him to ask whether it might not
be equally well deployed in thinking about those fundamental activities
of the simple substances themselves that in his eyes constitute their
essential properties. That is, if the body is what the dominant monad
needs to enable it to function as a perceiving and appetitive entity,
what is there to stop us, or indeed Leibniz himself, from entertaining
the thought that perception and desire are not so much activities of
immaterial simples as complicated manners of functioning of com-
pounds? If the functioning of the dominant self requires a material
mass of monads to work together in particular ways for it to be able to
perceive the world and entertain and gratify its desires, what is there to
stop us from supposing that perception and appetition themselves
should be understood in terms of the ways in which certain complex
material compounds behave? Developing the very train of thought that
leads Leibniz to the introduction of organic bodies itself raises the
question whether the existence of monadic simples is forced on him at
all.
It is of course plain what Leibniz in fact thinks stops him from
adopting any such stance, and we shall come to that in a minute. But
what I should emphasize here is that his resistance to the idea precisely
does not take the form of saying that to adopt such a course would be
to paint ourselves into an ontological corner from which there is no
escape. That is, it is precisely not to say that abandoning an ontology of
simples would ipso facto be to abandon complexes as well. For if we
think of men and ants and cabbages just as material substances of dif-
ferent sorts, and different in kind by reason of the sorts of activity that
they standardly and systematically go in for, then just as before we can
think of them as unbreakable unities in that as long as they are able to
perform a sufficiently wide range of their species-specific activities no
loss of bodily parts will count as impugning the existence of the
wholes that they are. Moving away from thinking of Wittgenstein as a
particular dominant monad together with his organic body and replac-
ing that with a conception of him as a particular embodiment of a
range of human capacities makes it no more difficult than previously it

was to say that the loss of his right arm in no way threatened the exis-
tence of the whole complex substance that was him. Not for nothing
was it the whole Wittgenstein, not just his arm, for whom Ravel wrote
the 1931 concerto for left hand alone, and so I think that Leibniz could
continue to equate the notions of being, reality and unity (or one) with-
out supposing that in order to do so we must also embrace the exis-
tence of monadic simples. s
Of course Leibniz's fundamental opposition to any such idea as
this stems from the implication that matter might in certain circum-
stances - as in the case of bodies arranged as human bodies are - en-
gage in thought. That he believed was hypothetically (though not ab-
solutely) impossible, just because it seemed to him to be flatly unintel-
ligible how the sort of mechanical interaction that he envisaged condi-
tioning the behaviour of material things could be what thought would
then ultimately reduce to. In that perhaps he was acute enough. How-
ever, we would not need to encourage him to abandon his faith in suf-
ficient reason to temper his hostility to the idea. As far as I can see,
there is nothing in Leibniz's philosophy that forbids him from widen-
ing the range of explanatory devices that apply to the material world
beyond the purely mechanical as soon as we find ourselves able to
draw upon them with clarity and success, and indeed it would do no
harm in our imaginary conversation with Leibniz to draw his attention
to the obscurity that he himself tolerates in the explanatory supposition
that the physical world is answerable in all its aspects to the percep-
tions and desires of its constituent immaterial metaphysical atoms.
Be that as it may, and regarding it now as an open strategy that an
alert Leibnizian might adopt that matter organised in some appropriate
way might indeed be what accounts for perception and desire, we can
address the second of the two questions that I said might appear to
threaten the existence of composite substances as Leibniz understands
them. That is, does not their tendency to perish which we will now ac-
cept once we allow the everlasting dominant monads to drop away in
the attribution to those larger wholes - in particular to men - of per-
ception and desire introduce a threat to their unity? That it does not do
so is in fact already implicit in the way in which I suggested that Leib-
niz can meet the first of the two challenges I raised. For there I said
that as long as a body of matter displays sufficiently wide range of the

species-specific behaviour through the loss (or even gain) of bodily
parts it will enjoy that kind of unbreakable unity that qualifies it as a
Leibnizian composite substance. And now it should suffice to remark
that that idea was cast in such a way to allow such a substance to per-
sist for a while and then for the matter that made it up to cease to dis-
play that species-specific behaviour. So when in our case the mental
and nutritive functions of the body break down, as they eventually will,
then we shall cease to exist. But that is entirely consistent with our
saying that prior to that moment, while those functions were sustained
by a healthily functioning body, we had just the unity that Leibniz took
to distinguish a compound substance from a merely phenomenal ag-
gregate. Our mortality is no threat to that.
Notes
I Draft ofletter to Arnauld (8 December, 1686), in Parkinson (ed.) [1973], p.64.
2 Cf. Nouveaux Essais II. xxvii. 5 : "Si les vegetables et les brutes n 'ant point
d'ame, leur identite n 'est qu 'apparente; mais s'its en ant, l'identite individuelle y est
veritable a la rigeur, quoyque leur corps organises n 'en gardent point", see
Gerhardt (ed.) [1978], V, p.215). Cf. Remnant and Bennett (eds.) [1981], p.232.
3 At New Essays Il.xxvii.6 Leibniz writes: "In fact, an organic body does not
remain the same [«n'est pas de meme»] for more than a moment, it only remains
equivalent. And ifno reference is made to the soul there will not be the same life, nor
a vital entity either. So the identity in that case would be merely apparent." (Remnant
and Bennett (eds.) [1981], p.232)
4 Some hint ofthis is found in Leibniz's saying at New Essays II.xxvii.8 (Rem-
nant and Bennett (eds.) [1981], p.235): "Indeed it does seem to me that we have to
add something about the shape and constitution of the body to the definition of man
when he is said to be a rational animal; otherwise according to my views, Spirits
would also be men." Again in "A Specimen of Discoveries about Marvellous Se-
crets" we find: "So it was sufficient that souls should be given to the lower animals,
especially as their bodies are not made for reasoning, but destined to various func-
tions - the silkworm to weave, the bee to make honey, and the others to the other
functions by which the universe is distinguished" (Parkinson (ed.) [1973], p.84).
5 It is noteworthy that when Leibniz first used the term 'monad' (in correspon-
dence with Johann Bernouilli in 1690) it designated composites such as men or
sheep. (see Gerhardt (ed.) [1978], III, p.542). What drove him to introduce the meta-
physical simples that later went under that name was the thought that only if such
simple things existed could the unity ofthese composites be accounted for. Under my
proposed revision Leibniz could have retained his original usage and bypassed the
introduction of immaterial monads altogether.
ON NATURALISING LEIBNIZ
(A REPLY TO ANTHONY SAVILE)
Richard Glauser
The guiding aim and aspiration of Leibniz' phi-

losophy is to establish a rigorous rational foun-
dation for what he accepted as the fundamental
teachings of ethics and theology.)
I shall not argue against Savile's interpretation of a Leibnizian com-

posite substance - organic body cum dominant monad - as an indivisi-
ble unity. As far as I know Leibniz never explicitly says as much, but
Savile's reading has much to commend it in terms of explanatory
power. So I shall take it as a working hypothesis. All that Leibniz
would have to acknowledge is that there are two distinct ways in which
substances can be indivisible. One way is by being simple. The other
way is picked out by Savile's notion of a composite substance, namely
a substance such that, even though it comprises an indefinitely large
number of monads, each and every one of which, apart from the domi-
nant monad, may be gradually replaced by other monads, the whole
made up at anyone time of such monads nevertheless persists by vir-
tue of the fact that these monads account for the dominant monad's
perceptions and appetitions. All that is required, it seems, is an exten-
sion of the application of the adjective 'indivisible'.
However, Savile does not only interpret Leibniz; he would also
like to reform him. Because the claim that organic bodies last forever
is counter-intuitive, Savile wants to reject it, so as to enable Leibniz to
say that organic bodies, and compound substances, begin and cease to
exist naturally just as we ordinarily believe them to. Now, since or-
369
A. Bottani et aL (eds.), Individuals, Essence and Identity, 369-385.
ganic bodies have to last forever, according to Leibniz, because their

dominant monads do, Savile must say that souls and rational minds are
not eternal either. So he suggests that Leibniz take dominant monads
as being no more than manners of functioning of organic bodies. An
animal or a person would be "a certain body of matter that operates to
maintain itself in existence while performing a number of species spe-
cific tasks: nutrition, growth, reasoning, etc." (Hence, in Savile's revi-
sion there will be no difference between an organic body thus de-
scribed and a composite substance). And this, as Savile acknowledges,
requires that matter should think. The upshot is that souls and minds
will not be substances at all, whether material or immaterial, since they
will be just manners of functioning of organic bodies; nor will they be
simple and naturally everlasting. In fact, there will no longer be any
monads, dominant or not, in Savile's revised universe, for he wants to
"preserve compound substances without simples" . Yet, he believes
that the removal of all monads from Leibniz's best possible world
"would not in fact require the abandonment of his deepest convic-
tions".
I am less confident of this than Savile is. I shall begin by giving a
general view of some of the important features of Leibniz' s system that
Savile's revised Leibniz will be able to accommodate mutatis mutandis
(section 1), and next, of what I believe he will have to abandon (sec-
tions 2 and 3). I shall try to show that, if naturalised Leibniz has to give
up what I think he must, he is not a revised version of Leibniz, but the
author of an altogether different philosophy.
1. WHAT NATURALISED LEIBNIZ CAN RETAIN
The proposed revision can accommodate several cardinal features of

Leibniz's thought. As long as Savile keeps Leibniz's God in place, his
revision will be compatible with the principles of sufficient reason, of
indiscernibles, of continuity, and with the general theory of the best of
all possible worlds. At any rate one can hardly say that the mere identi-
fication of minds with psychic and biological functions of organic
bodies, taken by itself, would prevent such general principles from ap-
Richard Glauser, On Naturalising Leibniz 371
plying. The principles would just apply to things that are different from
those to which they apply in Leibniz. For instance, one would pre-
sumably have to take complete notions to be individual essences of
organic bodies. This means that, in the case of an animal or a person, a
complete notion will have to contain both physical and mental predi-
cates, because one and the same individual organic entity instantiating
the complete notion will have both physical and mental accidents.
Furthermore, one might also go as far as to say that naturalised
Leibniz could retain the general features of Leibniz's account of hu-
man freedom. He is a compatibilist of sorts, inasmuch as he holds that
freedom is compatible both with divine foreknowledge and with pre-
determination. According to the Theodicy, human freedom rests
mainly on three features: contingency, spontaneity and intelligence. 2 If
Leibniz were to adopt Savile's suggestion, he could maintain roughly
the same compatibilist position. I say 'roughly' because, although there
would be no difficulty in retaining contingency and intelligence as they
are discussed by Leibniz himself, his metaphysical account of sponta-
neity would probably have to be abandoned, for reasons to be adduced
below, in favour of our everyday notions of spontaneity and absence of
constraint, which are the ones that count, and suffice, for ordinary
moral and legal purposes. To abandon his metaphysical account of
spontaneity would imply that he give up the claim that "we [i.e. per-
sons] are constrained only in appearance and that in metaphysical
strictness we are in a state of perfect independence as concerns the in-
fluence of all the other created beings".3 (I shall return to this below.)
The real Leibniz can say this as long as persons are dominant monads,
between which there is no real transeunt causation. 4 But it would be
strange for Savile's Leibniz to say that persons are merely bodies with
mental functions and, also, that in metaphysical strictness no person is
ever constrained by any other. Nevertheless, one can grant Savile that
naturalised Leibniz might say, more plausibly, that, although people
qua thinking bodies sometimes are constrained, whenever they are not
and they do act spontaneously they satisfy one of the three necessary
conditions for acting freely. Given that the other two conditions can
also be satisfied, such an adjustment concerning spontaneity would not
jeopardize the essentials of Leibniz's theory of human freedom insofar
as it is relevant to everyday moral purposes.
Nevertheless, I believe that naturalised Leibniz will have to aban-

don at least eight important features of Leibniz's system: his idealism,
his particular brand of platonism, the theory of pre-established mind-
body harmony, natural immortality, natural religion, the pre-
established harmony between nature and grace, important aspects of
the City Of God, and a significant part of his theodicy. In the following
section, I deal with the first three issues.
2. THE ISSUES OF IDEALISM, PLATONISM AND PRE-

ESTABLISHED MIND-BODY HARMONY
First, idealism. Leibniz is an idealist in the following broad sense: he

believes that reality is ultimately made up of minds, that is, souls or
simple substances analogous to souls, and that even if more complex
entities can be understood and explained in terms of mechanics, dy-
namics, and so on, the reality of the more complex entities depends on
that of minds or entities analogous to souls. Yet idealism is precisely
what has been ruled out of Savile's revision. For, if minds are merely
sets of bodily dispositions or functional aptitudes, they no longer enjoy
the kind of basic metaphysical status that Leibniz afforded them. Not
only are minds or souls not substances, but, instead of the reality of
bodies depending on the reality of minds, as Leibniz would have it, it
is the other way round: a mind, i.e. a set of biological and psychic ap-
titudes, ontologically depends on an organic body, just as powers de-
pend on the substance that owns them. And if minds, as construed by
Savile, depend on organic bodies, then material entities will be onto-
logically more basic than minds. Idealism will be hopelessly wrong.
And this, I suggest, does not agree with one of Leibniz's deepest con-
victions. That this was in fact one of his deepest convictions can be
evidenced by Nicholas Jolley, who has argued convincingly in a book-
length study of the New Essays on Human Understanding that "for all
its apparent randomness and lack of direction, [it] is a book dedicated
to defending the idea of a simple, immaterial and naturally immortal
soul". 5
Savile's revision also runs against what one may roughly call
Leibniz's platonism, i.e. the idea that the difference between bodies
and rational minds simultaneously covers at least three contrasts: be-
tween the sensible and the intelligible, the material and the immaterial,
the continually fluctuating and the immutable. With the abandonment
of Leibniz's versions of idealism and platonism the general form and
structure of his system would scarcely be recognisable. 6
The question of the relation between mind and body, just men-
tioned in connection with Leibniz's versions of idealism and plato-
nism, naturally leads to the question of pre-established mind-body
harmony. Leibniz has both a posteriori reasons and a priori reasons
for defending pre-established mind-body harmony.7 His a priori rea-
sons are related to his claim that pre-established mind-body harmony is
a special case of universal mind-mind harmony. To cut a long story
short in an admittedly rough and ready way, bodies are made up of
minds or soul-like substances, so that whatever pre-established har-
mony is to be found between a mind (dominant monad) and its body
will depend on the harmony to be found between the mind (dominant
monad) and the soul-like monads that make up its body. Once minds
are reduced to biological and psychic functions of bodies in accordance
with Savile's reform, any pre-established mind-body harmony that re-
mains will no longer be a special case of pre-established mind-mind
harmony. It will be a special case of body-body harmony, a harmony
between two aspects - mental and physical- of one and the same body.
In sum, mind-mind harmony will be a special case of body-body har-
mony, where the harmony between one mind and another will be taken
as obtaining between two bodies considered together with their psychic
functions. And mind-body harmony will also, but in a different way, be
a special case of body-body harmony, where the harmony between a
mind and its body will be taken as obtaining between two aspects -
psychic and physical - of one and the same body. In both cases, the
rejection of Leibniz's idealism makes body-body harmony come out as
basic, and pre-established harmony is set on its head.
But under Savile's proposed revision how well will such a con-
ception of pre-established mind-body harmony work? I think it will not
work well. In order to show why, let me begin by making two, rather
long preliminary remarks.
Here is my first preliminary remark When presenting pre-

established mind-body harmony in the New System or elsewhere, Leib-
niz does not present it as a solution to a problem concerning causality
as such, causation an sich. His problem is not Hume's. This is evi-
denced by the fact that Leibniz sees no difficulty at all in saying that
God really does cause this world to be. More importantly for our pur-
poses, he sees no difficulty in saying that finite minds really do cause
or produce all their perceptions, desires and other mental states. In fact,
he explicitly connects the two by saying that the way in which finite
minds produce their thoughts is analogous to the way in which God
continually produces and conserves finite substances: both are per-
formed by way of "a kind ofemanation".8 The term 'emanation' shows
that Leibniz believes that real causation depends on a source of activity
from which effects literally flow. To be sure, this part of Leibniz's the-
ory of causation cannot be assimilated to Hume's. "Flowing from" is
foreign to Hume's constant conjunction theory of causation. Emana-
tion is precisely part of what Hume was attacking in the Treatise.
However, the problem that Leibniz did encounter, and which was
related to causality, is the problem of what he called 'transeunt' causa-
tion, i.e. the causality that is supposed to hold between one finite mind
and another, between one body and another, and between a mind and
its body. As the New System shows, the problem was initially that of
mind-body causal interaction, already encountered by Descartes and
Malebranche. (Leibniz will go on to fit his solution to the problem of
mind-body interaction into a general account of so-called transeunt
causation, which also explains what we take to be mind-mind and
body-body causal relations). The solution is in several parts. (1) Finite
minds imitate God in being the sole causes of all their mental states.
This they do by having a metaphysically real source of spontaneous
causal activity from which all their mental accidents emanate. (2)
Since all of a substance's accidents are really produced by itself, and
since real causal over-determination is ruled out of the order that ob-
tains in the best possible world, none of a substance's accidents are
really caused by anything other than the substance that owns them.
Only immanent causation is metaphysically real. Hence, there is no
real mind-mind or mind-body causal interaction. This is why Leibniz
claims, as we saw above, that "in metaphysical strictness we are in a
state of perfect independence as concerns the influence of all the other

created beings". Transeunt causality is not metaphysically real,
whether it be taken as a transmission of species, or of forms, or of
properties, or in the sense of any other kind of real influx or emana-
tion. Hence, the (slightly misleading) metaphor of the mind and its
body as two clocks. But (3) Leibniz is not one to reject ordinary ways
of speaking as long as they can be interpreted according to metaphysi-
cal truth. In this respect he is quite similar to his contemporaries,
Malebranche and Berkeley, who also sought, in different ways, to re-
form our conceptions of what counts as metaphysically real causation.
So, Leibniz says: "Ordinary ways of speaking can still be preserved.
For one may say that when the particular disposition of one substance
provides a reason for a change occurring in an intelligible manner, in
such a way that we can conclude that the other substances have been
adapted to it on this point from the beginning according to the order of
the divine decree, then that substance should be thought of as acting
upon the others in this sense".9 Thus, when Leibniz says, for instance,
that this mind produces a change in its body, what he means is roughly
the following: (i) there is a mental accident of this mind that is really
produced by this mind in the strict metaphysical sense, and in produc-
ing this accident this minds attains a higher degree of perfection; (ii)
there is a change (accident) of this mind's body that is produced by the
body's force, etc.; (iii) this accident of this mind does not really cause
or otherwise produce this body's change (accident); (iv) yet this acci-
dent of this mind can serve to explain (afford a reason for) this body's
change or accident.
In sum, the theory of pre-established mind-body harmony implies
a deep asymmetry between immanent and transeunt causality. The one
is metaphysically real, a kind of emanation, a true producing. The other
is little more than an explanatory relation presumably based on no-
mological concomitance. Statements to the effect that a mind causes
some change in a body, or vice versa, are justifiable far;ons de parler,
acceptable only if translatable into statements that replace 'causing' by
'explaining' or 'giving a reason for' founded on some nomological
regularity.lo Yet the asymmetry is part and parcel of the theory of pre-
established mind-body harmony. I I It is what enables Leibniz to say that
monads harmonise, accord with each other, immanently produce each
their own effects concomitantly yet independently of each other, i.e.

act as though they causally interacted with each other, or with their
bodies, although they do not really do so. If one interprets pre-
established mind-body harmony merely in terms of nomological regu-
larities between mental accidents and physical accidents, leaving the
causal asymmetry out of the picture, one fails to capture what Leibniz
himself considered his definite advance over Descartes and Male-
branche as regards the causation that is really at work in the actual
world. One also fails to account for the fact that Leibniz considers
Shaftesbury's world to be full of universal harmony, yet lacking "my
pre-established harmony".12 So, Leibniz's theory of pre-established
harmony cannot be reduced to universal nomological concomitance.
Now, to a second preliminary remark. Savile's hypothesis would
require Leibniz to have a quite different notion of matter from the one
that he does have. Remember that, according to Leibniz himself, his
notion of what he calls "secondary matter" is not even compatible with
a realist interpretation of Newton's attraction. He believed that attrac-
tion was on a par with scholastic occult qualities.13 Just imagine the
kind of adaptation the Leibnizian account of matter would need to ac-
commodate the claim that matter naturally thinks! Savile says that
Leibniz "thinks it inexplicable [that matter naturally thinks] because he
envisages explicable relations at this level being necessarily mechani-
cal ones. Since we envisage other possibilities than that, Leibniz's re-
sistance here is unlikely to impress us as more than superficially moti-
vated". True, we do envisage other possibilities. But why? Savile
seems to suggest that we do envisage other possibilities - if not en-
tirely, yet in part - because the outlook we have on the mind-body re-
lation and its place in the natural order of things is based on well-
founded scientific beliefs very different from those which were avail-
able at the time. Remember that in roughly the same scientific context,
even Locke, whom Leibniz took to be one of his chief opponents on
the issue, did not claim more than the epistemic possibility that matter
might think. Even Locke held that "the more probable opinion is, that
this consciousness is annexed to, and the affection of one individual
immaterial substance".14 Hence, if one wishes to alter Leibniz's posi-
tion on the mind-body issue in Savile's direction, one should also alter
his contextually well-founded scientific beliefs on the nature of matter.
Let us go along with that, and try to see what happens if naturalised
Leibniz attempts to retain pre-established mind-body harmony includ-
ing, therefore, the causal asymmetry mentioned above.
The metaphor of the two clocks will no longer suggest a harmony
between the mental accidents of a dominant monad and the accidents
of its organic body. Presumably it will suggest a harmony to be found
between the series of mental accidents and the series of physical acci-
dents of one and the same organism. Now, harmony is not identity. So,
if naturalised Leibniz wants to retain some form of pre-established
mind-body harmony, he will probably have to say that the physical ac-
cidents and the mental accidents of one and the same organism con-
stitute two distinct, parallel series, where the members of one series are
neither type-identical nor token-identical with members of the other
series. It seems that we have accident-dualism of one and the same
body. Also, since pre-established mind-body harmony implies causal
asymmetry in the sense introduced above, real transeunt causation is
ruled out. So, naturalised Leibniz will presumably say that the two dis-
tinct, parallel series of accidents of one and the same organic body will
be causally independent of each another. Physical accidents will cause
other physical accidents of the same body; mental accidents will cause
other mental accidents of the same body. But, strictly speaking, mental
accidents cannot cause physical accidents, and vice versa. This is even
stranger than epiphenomenalism for a philosopher bent on some form
of mind-body naturalism. Even if naturalised Leibniz were to hold on
to Leibniz's ideal notion oftranseunt causality, that would not hide the
fact that the mental and physical accidents of an organic body will be
causally inert with respect to each other. Such a position is perhaps not
logically impossible. However, if one really is justified in believing
that matter thinks, such an attempt to salvage pre-established mind-
body harmony is going to seem queer indeed.
Where does the queerness stem from? I suggest it stems from the
assumption that, if naturalised Leibniz really is justified in believing
that matter thinks, then the kind of justification he has for believing
that makes it implausible to hold that the mental and physical accidents
of one and the same body have no real causal relations at all between
each other. For, ask such a naturalistic philosopher, say Hume or
Searle: are mind-body and body-mind causality metaphysically real?
He will reply that they are just as real as any causality is ever likely to
be in this world. That is the point with naturalism: minds and their
states are natural phenomena; so whatever kind of causal relations ob-
tain between physical phenomena also obtain between physical phe-
nomena and mental states, and vice versa, and between mental phe-
nomena. Of course, Hume would add, as naturalised Leibniz would,
that such causality depends on nomological regularity. Fine, but if
mind-body and body-mind causality (based on pre-established regular-
ity) are just as real as any other causality, then Leibniz's causal asym-
metry has been lost. Thus, naturalised Leibniz will not hold that im-
manent causality is metaphysically real whereas transeunt causality is
not. Even if he accepts a distinction between immanent and transeunt
causality, he will put them on a par with each other; in fact he will not
grant the distinction any metaphysical significance, just as Hume does
not. But, again, once causal asymmetry is lost, so is the distinctive
feature of pre-established mind-body harmony. At this point, I submit,
pre-established mind-body harmony will have to give way. And, so,
naturalised Leibniz is not a revised Leibniz but the author of an alto-
gether different philosophy.
Related to this is the question of the real causal source of the dou-
ble series of a body's accidents. Leibniz's pre-established harmony de-
pends on monads' containing within themselves the real causal source
of the entire series of accidents they will ever have. So, if naturalised
Leibniz still wishes to preserve pre-established harmony, he will
probably say that an organic body contains entirely within itself an
autonomous source of causation capable of spontaneously producing
the two parallel series of all physical and mental accidents that are ever
to affect the body during its finite lifetime. But where will such a
causal source lie? Savile does not say. It cannot be attributed to a
dominant monad and to the monads that compose the body, for they
have been rejected. So, would such a causal source be discoverable
empirically, by scientific means, within the body? That does seem un-
likely. If one replies that such a causal source of the accidents of an
organic body lies within the smaller organic bodies of which it is com-
posed, and so on ad infinitum for the immanent causal source of each
of the smaller organic bodies, then it must be shown that this kind of
reply is not an asylum ignorantiae. In any case, is it even remotely
plausible to hold that an organic body has within itself such a source of
causal power? If naturalised Leibniz is to be credited with the kind of
scientifically based naturalistic outlook that makes it plausible to be-
lieve that minds are bodily functions, the same outlook seems to make
it utterly implausible to say that within an organic body lies a source of
immanent causal activity that spontaneously produces the complete,
double series of physical and mental accidents that are to affect the
body during its entire lifetime. Finally, if it is objected that such a po-
sition is tenable for the mere reason that it is not logically impossible,
the reply is that whatever plausibility may have been gained by making
animals' and persons' life spans what we ordinarily take them to be
has been lost by the proposed revision.
3. THE ISSUES OF NATURAL IMMORTALITY, PRE-

ESTABLISHED HARMONY BETWEEN NATURE AND GRACE,
THE CITY OF GOD, NATURAL RELIGION, AND THEODICY
Leibniz holds that pre-established harmony is also to be found between

the physical order of nature and the moral order of grace. IS One impli-
cation of this sort of harmony is highlighted in the Monadology, where
Leibniz explains that
things lead to grace by means of the very ways of nature and that this
globe, for example, must be destroyed and repaired by natural ways at
those times which the government of spirits demands for the punishment
of some and the reward of others"; "[ ... ] sins must therefore carry their
punishment with them by the order of nature, and even by virtue of the
mechanical structure of things; and that noble actions, similarly, attain
their rewards through ways that are mechanical in relation to bodies".16
Let us set aside the issue of the future destruction of the world, and fo-
cus merely on the claim that in the best possible world virtue will be
rewarded and vice punished by entirely natural ways, whatever the
natural ways are. But let us also keep in mind that all rational minds,
which are to be rewarded or punished in due course, are perpetual
members of the City of God, an everlasting moral state comprising all
human minds, angels, and God as prince and legislator. Now, Leibniz
is clear (1) that citizenship in the City of God requires that every finite
rational mind be naturally immortal, and (2) that the natural system of
reward and punishment governing the City of God requires not only
that citizens be naturally immortal, but also that they retain their con-
sciousness, memory and personal identity after what we call their natu-
ral deaths.I7 It is important to recall this, because, even though Leibniz
does go along to a certain extent with the Spinoza-Shaftesbury view
according to which virtue is its own reward, and vice its own punish-
ment, and though he believes, too, that during what we call this life
vice generally causes some unhappiness to the vicious person, and
virtue causes some happiness to the virtuous person, nevertheless he
also believes that, all said and done, by the end of what we call this life
the vicious sometimes end up being happy, and the virtuous unhappy.
That this is a fact shown by experience according to Leibniz (as it will
be later for Kant) is made clear in the Theodicy18 and elsewhere. 19 And
this fact, if it were not "corrected", would be "contrary to order"; it
would violate the pre-established harmony between nature and grace,
and be incompatible with the perfect government of the City of God.
So, how can such a disorderly situation be corrected? Only if rational
minds are naturally immortal, and if unrewarded virtue and unpunished
vice in this life are duly rewarded and punished in what we call an af-
ter-life. That is why Leibniz believes that natural immortality is neces-
sary for due reward and punishment to be naturally, and completely
administered. Although they are not always naturally administered in
full during this life, reason and natural religion justify the belief that
things will be set right in what we take to be an after-life. And even
natural immortality is not enough; permanence of memory, conscious-
ness and personal identity are required, toO.20 Hence, natural immortal-
ity and permanence of memory, etc. are essential to Leibniz's pre-
established harmony between nature and grace and to his theory of the
City of God. (It would be an ad hoc manoeuvre to reply that it is open
to naturalised Leibniz to assert flatly, contrary to the real Leibniz, that
all good and evil deeds are in fact entirely, adequately and naturally
rewarded and punished during this life. In any case, what reason would
naturalised Leibniz have to make such an assertion, since this does not
follow from the best possible world theory, nor from the a priori prin-
ciples of his system?)
Leibniz's conception of natural religion, too, depends on natural
immortality. He rejects the idea that religion should be founded on
revelation alone, and is determined to establish it on reason and nature
as well. Now, the cornerstone of Leibniz's natural religion is the natu-
ral immortality of souls, which can be known by reason. 21 And minds
are naturally immortal, according to Leibniz, if, and only if, they are
immateriaP2 On Savile's proposal, however, minds are bodily func-
tions and so can hardly be immaterial and naturally immortal. On these
four counts - natural immortality, natural religion, pre-established
harmony between nature and grace, and the theory of the City of God -
naturalised Leibniz's position is incompatible with the real Leibniz's
deep convictions.
As I have tried to show, naturalised Leibniz has to give up two
parts of the general theory of pre-established harmony, i.e. mind-body
harmony and the harmony between nature and grace. If so, the general
theory as Leibniz conceived it is crippled. So, then, is a proof of the
existence of God that Leibniz considers one of his most original and
most powerful. 23
But, to return to the question of immortality, one might ask: So
what? Can we not have naturalised Leibniz plus miraculous immortal-
ity and miraculous preservation of personal identity in an after-life?
But that would amount to turning that which Leibniz describes as natu-
ral into something miraculous in naturalised Leibniz. Also, the claim
to a natural religion would be lost, since the belief in immortality could
no longer be justified by reason. Religion would be confined to revela-
tion, which is precisely what the real Leibniz sought to avoid. 24 How,
in any case, could he claim to know that a miracle takes place after
what would be the natural death of rational minds?
Finally, let us reflect that Leibniz's conception of the best possible
world depends on his dispensing with the constant invocation of the
idea of God's miraculous intervention. Leibniz brings this about, not
by rejecting miracles altogether, but by fitting whatever miracles there
are into the general order of the best possible world. 25 If miracles are to
be incorporated in such a general order, they are bound to be few and
far between. Indeed, one of the reasons for which Leibniz was proud of
his theory of pre-established harmony was that, in his opinion, it gave

him a definite advantage over Malebranche, whose occasionalism
Leibniz saw as requiring God to perform miracles continuously. 26
Hence, naturalised Leibniz would have to confront the following alter-
native. Either (1) he holds on to the claim that rational minds are im-
mortal, but only miraculously so, in which case God might have to per-
form a miracle every time some human being naturally dies in order to
be able to see to it that the person is duly rewarded or punished. And
this will contradict the real Leibniz's own conception of the general
order of the best possible world by making miracles continuous. 27 Or
(2) naturalised Leibniz gives up immortality altogether, in which case
even revealed religion will be largely mistaken. In either case, his
losses are pre-established harmony between nature and grace and the
theory of the City of God.
It should be clear by now that the rejection of natural immortality
and the rest will be enough to cripple Leibniz's cherished theodicy.
Consider a central problem for any theodicy, namely moral evil (as
distinguished from both natural evil and metaphysical evil). Leibniz
has several ways to fit moral evil into the general order of the best pos-
sible world, and justify God's permitting it. For example, moral evils
are generally compensated by the occurrence of greater goods; or
moral evils generally cause equal or greater goods to occur. But these
ways of justifying God's permitting moral evil have nothing to do with
justice, i.e. with the punishment of the wrong-doers themselves. So,
Leibniz's theodicy also relies on the further claim that God's justice
prevails over the entire course of the best possible world, even if it is
only in the long run that it prevails completely. God's justice has to do
not only with the punishment of the moral evil brought about by finite
minds, but also with the rewarding of the moral good they bring about.
And inasmuch as God's justice deals with both, this justice is really the
same thing as his legislation within the City of God. But we have seen
that the perfect legislation and justice within the City of God depend
crucially on the natural immortality of all finite rational souls. Hence, I
suggest that once natural immortality has been discarded, one of the
main pillars of Leibniz's theodicy has been destroyed. And it is quite
doubtful that the loss of the pillar will not affect the whole edifice. For
how could Leibniz's theodicy survive the loss of the complete accom-
plishment of divine justice?
Savile says: "there is nothing in Leibniz's philosophy that forbids
him from widening the range of explanatory devices that apply to the
material world beyond the purely mechanical as soon as we find our-
selves able to draw upon them with success", in order to claim that
minds are just bodily functions. In the two last sections I have argued,
to the contrary, that if Leibniz were well justified in accepting that
minds are bodily functions, that they are neither immaterial nor natu-
rally immortal, then ripple effects would undermine enough of the es-
sentials of his system to enforce abandonment of it. 28
Notes
1 Rescher [1993], p.159.
2 Cf. Theodicy, §288, in Gerhardt (ed.) [1978], VI, p.288.
3 A New System of the Nature and the Communication of Substances, § 16, in
Loemker (ed.) [1976] p.458. Loemker translates the French "entraines" by

"determined". I believe one should read "constrained" instead of "determined", for
Leibniz is here contrasting spontaneity with constraint, not freedom with determin-
ism. Cf. also Theodicy, §290-291, Gerhardt (ed.) [1978], VI, p.289-290.
4 I use the term 'transeunt' as does Leibniz; cf. Explanation of the New System
of the Communication of Substances, §17, in Parkinson (ed.) [1973], p. 128.
5 Jolley [1984], p.7.
6 On Leibniz's idealism and his platonism, cf. Brunner [1951]
7 Cf. Explanation of the New System, § 16, Parkinson (ed.) [1973], p.128.
8 Discourse on Metaphysics, §14, Loemker (ed.) [1976], p.311, my italics.
9 New System, §17, Loemker (ed.) [1976], p.459, my italics. It should be added
that in the case of a substance being said to act on other substances, the substance
said to act goes through a real change whereby it attains a higher degree of perfection
than it previously had, whereas the other substances endure a real change leading to a
lesser degree of perfection. Cf. Discourse on Metaphysics, § 15, Loemker (ed.)
[1976], p.313. Of course, these real changes are brought about by immanent causa-
tion.
10 I use the expression 'causal asymmetry' in this sense only, i.e. to express the
contrast at the level of metaphysics between real immanent and, in a sense, unreal
transeunt causation. I do not use the term to express the fact that causal relations have
a fixed direction. I take it for granted that in Leibniz (and naturalised Leibniz) all
causal relations have a fixed direction.
11 Lest there by any doubt on this point, consider that one must distinguish two
claims in Leibniz: (1) the mental states of a Leibnizian substance follow each other in
an orderly series pre-established by God, and explainable by the substance's com-
plete notion, and (2) they do so in fact by being really produced by a source of im-
manent spontaneous causal activity within the substance. The two claims are distinct.
Yet (1) without (2) would make no sense to Leibniz. It would be like trying to say
that all the predicates that can be truly attributed to an individual subject (when the
truth is contingent) are contained in the subject's complete notion, yet it is not the
case that the substance instantiating the complete notion is the metaphysically real
cause of (all) its accidents instantiating the predicates; i.e. it is not the case that (all
of) our thoughts are really produced by us by way of a kind of emanation analogous
to the way by which God produces and preserves us. But (1) and (2) lead to the
causal asymmetry mentioned above, essential to the theory of pre-established mind-
body harmony.
12 Indeed, in his review of Shaftesbury's Characteristics, Leibniz praises
Shaftesbury for his world "all of one piece" and for its "universal harmony". But a
few lines further he adds that it lacks pre-established harmony (Loemker (ed.) [1976],
p.633). At first sight this seems strange, for Leibniz often uses the two expressions as
roughly equivalent. (Cf. Finster, Hunter et alii [1988], pp.136-138). Also, Leibniz
knows, as does any reader of the Characteristics, that Shaftesbury's world is just as
well regulated by causal laws as most other philosophers' worlds are. And he also
knows that all the law-like regularities in Shaftesbury's world have been ordained by
God, i.e. they are pre-established. So, why does it lack pre-established harmony? I
suggest it is because it does not conform to Leibniz's theory of pre-established har-
mony, i.e. it lacks the conjunction of (1) and (2), mentioned above, and therefore
lacks the contrast (asymmetry) between immanent and transeunt causation.
\3 Nouveaux Essais sur l'entendement humain, Preface, Gerhardt (ed.) [1978],
V, p.59.
14 An Essay Concerning Human Understanding II, 27, §25, Nidditch (ed.)
[1975], p.345.
15 Cf., for instance, Monadology, §87, Loemker (ed.) [1976], p.652; Theodicy,
§62, Gerhardt (ed.) [1978], VI, pp.136-137.

16 Monadology, §88 and 89, Loemker (ed.) [1976], p.652.
17 Cf. letters XVI and XXII to Arnauld, Gerhardt (ed.) [1978], II, p.IOO and
p.125.
18 Cf. Gerhardt (ed.) [1978], VI, p.lIO-lll.
19 "Since experience shows us that God, for reasons unknown to us but surely
very wise and based on a greater good, permits many evil persons to be happy in this
life and many good persons to be unhappy, a fact which would not conform to the
rules of a perfect government such as God's if it had not been corrected, it follows
necessarily that there will be another life and that souls will not perish with the visi-
ble bodies. Otherwise there would be crimes unpunished and good deeds unre-
warded, which is contrary to order" (Reflections on the Common Concept of Justice,
Loemker (ed.) [1976], p.564). Cf. Opinion on the Principles of Pufendorf, in Riley
(ed.) [1988], p.66-68. On this whole issue I refer the reader to Grua [1953], Chapter
15.
20 Cf. letter to Malebranche, Gerhardt (ed.) [1978], IV, p.300.
21 Nouveaux Essais sur {'entendement humain, Preface, Gerhardt (ed.) [1978],
V, pp.60-61.
22 Cf. Jolley [1984], p.21.
23 Cf. letter XXII to Arnauld, Gerhardt (ed.) [1978], II, p.115. Cf. also VI,
p.541.
24 Cf. Gerhardt (ed.) [1978], V, pp.60-61.
2S Cf. Discourse on Metaphysics, §7, Loemker (ed.) [1976], pp.306-307.
26 Cf., for instance, Gerhardt (ed.) [1978], VI, p.541.
27 Not to mention that it would hardly befit Leibniz's morally perfect and wise
legislator, a citizen among others in the City, although the most prominent one, to be
obliged to perform miracles in order to see to it that some of his citizens are punished
and others rewarded.
28 I thank David Wiggins for his comments on an earlier version of this paper.
Any shortcomings are my own.
TEMPORAL PARTS AND IDENTITY
ACROSS TIME
Peter van Inwagen
1.
Many philosophers think that 'What is identity across time?' is an im-

portant and meaningful question. I have a great deal of trouble seeing
what this question might be. But, very often, if one cannot understand
a philosophical question, one's best course is to look at some alleged
answers to it; sometimes these answers enable one to see what ques-
tion it is that they are offered as answers to. The following passage by
Michael Tooley is supposed to provide an answer to the question we
are trying to get at.
[W]hat does it mean to say, for example, that the book on the table at time
t is identical with the book on the chair at time t*? One answer is that it
means that the object referred to by the expression 'the book on the table
at time t' is the same object as that referred to by the expression 'the book
on the chair at time t*'. But one need not rest with this superficial account,
since one can go on to ask what is meant by 'the object referred to by the
expression 'the book on the table at time t*". And one very natural answer
is this. The expression 'the book on the table at time t' picks out a certain
spatially and temporally limited part of the world, and it does so either by
picking out an instantaneous slice, of the book variety, which exists at time
t, or else by picking out a relatively small non-instantaneous temporal part,
of the book variety, which occupies a small interval containing time t and
387
A. Boltani et at. (eds.), Individuals, Essence and Identity, 387-411.
then by linking this up with all other slices (or parts) of the relevant sort
which stand in a certain causal relation to the slice (or part) existing at (or
around) time t. (Tooley [1977], pp.97-98)
Let us examine this answer carefully. I shall write as if Tooley ac-

cepted his "one very natural answer," since this answer encapsulates a
point of view I wish to examine and it will be convenient to have
someone to attribute it to; if Tooley is not fully committed to this an-
swer, I apologize to him for taking this liberty. 1
There are two important theses on display in this passage. One is a
thesis about the existence of certain objects, and the other is a thesis
about the relation of certain phrases in our language to those objects.
The first thesis is that there are such things as 'temporal parts' or'tem-
poral slices'. (The difference between very "thin" temporal parts and
temporal slices does not matter much in the present context. I shall talk
mainly of 'slices', but what I shall say could be applied to 'thin parts'
easily enough.) The second thesis is that the way in which "time-
involving" definite descriptions of physical objects like 'the book on
the table at t' relate to their referents should be analyzed or explained
in terms of slices.
One thesis about the referents of time-involving descriptions that
Tooley clearly does not hold is this: that phrases like 'the book on the
table at time t' actually denote slices; that 'the book on the table at
time t' denotes the "t-slice" of the book. Philosophers who hold this
view must either say that sentences like 'The book on the table at time
t is identical with the book on the chair at time t*' must always express
propositions that are, strictly speaking, false, or else they must say that
in such sentences 'is identical with' does not express the idea of nu-
merical identity - two slices being two slices - but rather some relation
of causal or spatiotemporal continuity. 2
Tooley's view of the matter is more artful and does not confront
this awkward dilemma. As Tooley sees it, 'the book on the table at t' is
a name for a certain four-dimensional object, one having slices "of the
book variety" as parts, these parts being bound together into a whole
by (again) some relation of causal or spatiotemporal continuity. As
Tooley sees it, the phrase 'the book on the table at t' means something
like 'the four-dimensional book the t-slice of which is on (the t-slice
Peter van Inwagen, Temporal Parts and Identity across Time 389
of) the table' and, similarly, 'the book on the chair at t*' means 'the
four-dimensional book the t*-slice of which is on (the t*-slice of) the
chair'. And, of course, no particular problems are raised by the asser-
tion that two such descriptions as these might be names for a single
object.
Both these theories about the way in which time-involving definite
descriptions are related to the world can be represented pictorially. The
first theory can be represented like this:
the book on
the chair at t*
o
_____________ 1 _______ 1 ______________ _
o
past t* future
Figure 1
The line drawn beneath this figure is a "time-axis": each point on it

represents a point in time, the left-to-right arrangement of the points
representing the past-to-future arrangement of points in time. The
"books" drawn above the time-axis represent slices "of the book vari-
ety." I have, of course, been able to represent only a few such slices:
the viewer must somehow contrive to imagine that the sequence of
book-drawings is continuous, just as the sequence of points on the line
is. Each book-slice-representation is drawn directly above the point
that represents the point in time it "occupies." Finally, the description-
referent relation is represented by labels bearing the description and
attached to the referent. (I do not mean this device in any sense to rep-
resent the "mechanics" of securing reference. It is meant to be neutral
with respect to theories of what reference is and how it is established.
It is used merely to display the fact that certain phrases denote certain
objects.)
The second theory, Tooley's theory, may be represented by this pic-
ture:
o
_____________ 1 _______ L ______________ _
past t" future
Figure 2
In this picture, the time-axis and the books mean what they meant in
Figure 1. The rectangle represents the boundary of the four-
dimensional object that is what the book really is. The description-
referent relation is again represented by labels, but the labels are fixed
to the "whole" book and not to slices of the book. Moreover, each of
the cords attaching the labels to the "whole" book passes through a
book-slice - the same slice it is attached to in Figure 1 - on its way to
its point of attachment to the book. This feature of the picture is in-
tended to represent Tooley's idea that a description like 'the book on
the chair at t' gets "linked up with" the book via the t-slice of the book.
(The points of attachment of the labels have no significance; I have to
draw them as attached somewhere.)
In my view, these pictures embody grave illusions about the nature
of enduring objects and about the way in which time-involving de-
scriptions apply to their referents and about the kind of facts expressed
by sentences formed by flanking the identity sign with time-involving
descriptions. (Let us call such sentences 'temporal identity-sentences'.)
I believe it is a grave illusion to suppose that there are four-
dimensional objects or that things are somehow composed of

"temporal parts" or "slices" or that the facts represented by temporal
identity-sentences even look as if they were facts about such objects. 1
believe that people who suppose such things as these are the victims of
seductive but incoherent pictures - pictures like Figure 1 and Figure 2,
in fact. In 'Four-dimensional Objects,' 1 presented arguments for the
following conclusion: the thesis that an enduring object is composed of
temporal parts has unacceptable modal consequences. But these argu-
ments were far from conclusive. For one thing - not the only thing -
they presupposed that a counterpart-theoretic account of modality de re
was unacceptable. 1 might in this paper try to plug some of the holes in
the argument of "Four-dimensional Objects." I might, for example, try
to show why 1 believe a counterpart-theoretic account of modality de
re to be unacceptable. 1 will not do this. 1 propose instead first to ex-
plain why 1 find temporal parts hard to understand, and, secondly, to
describe how identity across time "looks" to someone who has no
grasp of temporal parts - to provide a "picture" of identity across time
that is a rival to the pictures presented in figures 1 and 2. Finally, 1
shall show how to draw a third picture, for figures 1 and 2 have at least
one rival, a rival that is far more different from them than they are
from each other.
2.
What are temporal parts supposed to be? - or, if you like, What is
'temporal part' supposed to mean? Many philosophers find these
things, and this phrase, wholly unproblematic. They construct elegant
solutions to various philosophical problems by appealing to temporal
parts, and they seem to assume that these objects enjoy the same meth-
odological rights as numbers or sets: although there are philosophical
problems that could be raised in connection with them (1 dare say they
will be willing to concede that much), temporal parts are well enough
understood that philosophers can appeal to them without incurring any
obligation to interrupt their discussions of personal identity - or what-
ever - to explain them. (The most celebrated example of this sort of
appeal to temporal parts would probably be Quine's "Identity, Osten-

sion, and Hypostasis. "3)
Well, isn't it pretty clear what they are? We have, after all, per-
fectly intelligible names for them. For example: 'St. Paul's in 1850'
and 'Philip drunk'. (The second of these is presumably a name for a
non-connected part of Philip.) And we understand these names per-
fectly well, because we understand perfectly well the sentences in
which they occur: 'St. Paul's in 1850 was dingier than St. Paul's last
year'; 'Philip drunk is rash, but Philip sober is crafty'. And if we un-
derstand the names, we understand the things: 'temporal part' is merely
the general term that covers objects like St. Paul's in 1850 and Philip
drunk.
I do not think this is right, for I do not think that 'St. Paul's in
1850' and 'Philip drunk' are names at all. Let us look at some sen-
tences in which these phrases occur. To start with, let us look at 'St.
Paul's in 1850 was dingy'. In this sentence, 'in 1850' is an adverbial
phrase modifying 'was': When was St. Paul's dingy?; In 1850, at any
rate. 4 It is, therefore, a mistake to think of this sentence as having a
structure like this:
St. Paul's in 1850 was dingy
~b'
su stantIve ~b'
attn utIve ~l
copu a ~d'
pre lcate
adjective ../ '-- adjective..J
'-----------...V y
subject predicate
Its structure is rather this:

St. Paul's in 1850 was dingy
"---y-------J "---y-------J "-y--J "-y--J
substantive adverb copula predicate
adjective
\.
y
)
-
......... --------------
-v---------------_/
subject predicate
Philosophers who do not recognize the adverbial function of phrases

like 'in 1850', and who treat them as if they were adjectives modifying
the subjects of the verbs they in fact modify, are guilty of a fallacy I
like to call adverb-pasting. If the adverb-pasters were given free rein,
all manner of fascinating philosophical problems would be created.
Consider, for example, the sentence
Alice, viewed full-face, is strikingly beautiful, but Alice, viewed in
profile, is aesthetically unremarkable.
Here we have the germ of the problem of cross-perspectival identity:
what is the relation between the strikingly beautiful Alice-viewed-full-
face and the aesthetically unremarkable Alice-viewed-in-profile? And
there is the problem of cross-evaluational identity: what is the relation
between the brilliant Hume-according-to-Professor-A and the doltish
Hume-according-to-Professor-B?
More or less the same points (to revert to the case of temporal ad-
verbs) apply to more complicated sentences, like our 'St. Paul's in
1850 was dingier than st. Paul's last year'. The grammatical structure
of this sentence may be compared with the grammatical structure of
'Condorcet as representative figure of the Enlightenment is more inter-
esting than Condorcet as original thinker'. ("Condorcet is interesting."
"How? In what respects?" "Well, as a representative figure of the En-
lightenment; less so as an original thinker." "St. Paul's was dingy."
"When? At what times?" "Well, in 1850; less so last year.") I trust that
no one will want to say that 'Condorcet as a representative figure of
the Enlightenment' is a name of a certain temporal part of Condorcet, a
part that comprises just those moments at which he was engaged in
representing the Enlightenment, while 'Condorcet as original thinker'
is a name of the part of Condorcet that comprises those moments at
which he was engaged in original thought.
A similar account applies to 'Philip drunk'. As a first approxima-
tion to a correct account of the sentences in which this phrase figures,
we may say that a sentence like 'Philip drunk is rash' is just a fancy
way of saying 'Philip, when he is drunk, is rash', or - even less fancy-
'Philip is rash when he is drunk'. (We may compare 'Philip drunk is
rash, but Philip sober is crafty' to 'Philip the man is rash, but Philip the
politician is crafty'. Lest someone argue that 'Philip the politician' is a
name for a temporal part of Philip, the part comprising just those mo-
ments at which he was engaged in politics, we may stipulate that all
Philip's actions had both a personal and a political component, and

that the utterer of our sentence means by it that Philip rarely thought
about the consequences of his acts for himself and his friends and
family, but always thought carefully about their political conse-
quences.) I say 'as a first approximation' because the rhetoricians have
been at 'Philip drunk'. Availing themselves of the fact that 'Philip
drunk' can look rather like a substantive, they have produced conceits
like this: 'Macedon has two kings, Philip drunk and Philip sober'. But
this sort of trope (doubtless the Greeks had a word for it) is of no on-
tological interest. We should remind ourselves that it may have influ-
enced our thinking and then tum our attention to more profitable mat-
ters.
We cannot, therefore, explain 'temporal part' by saying that it is
"merely the general term that covers objects like St. Paul's in 1850 and
Philip drunk." What other explanations are available? I know of only
one. It is due to David Lewis, and, as one might expect, it repays close
attention.
Some would protest that they do not know what I mean by "more or less
momentary person-stages, or times-slices of continuant persons, or per-
sons-at-times". Others do know what I mean, but do not believe there are
any such things.
The first objection is easy to answer, especially in the case where the
stages are less momentary rather than more. Let me consider that case
only; though I think that instantaneous stages also are unproblematic, I do
not really need them. A person-stage is a physical object, just as a person
is. (If persons had a ghostly part as well, so would person-stages.) It does
many of the same things that a person does: it talks and walks and thinks,
it has beliefs and desires, it has a size and shape and location. It even has a
temporal duration. But only a brief one, for it does not last long. (We can
pass over the question how long it can last before it is a segment rather
than a stage, for that question raises no objection of principle.) It begins to
exist abruptly, and it abruptly ceases to exist soon after. Hence a stage
cannot do everything that a person can do, for it cannot do those things
that a person does over a longish interval.
That is what I mean by a person-stage. Now to argue for my claim
that they exist, and that they are related to persons as part to whole. I do
not suppose the doubters will accept my premises, but it will be instructive
to find out which they choose to deny.
First: it is possible that a person-stage might exist. Suppose it to ap-
pear out of thin air, then vanish again. Never mind whether it is a stage of
any person (though in fact I think it is). My point is that it is the right sort
ofthing.
Second: it is possible that two person-stages might exist in succes-
sion, one right after the other but without overlap. Further, the qualities
and location of the second at its appearance might exactly match those of
the first at its disappearance. Here I rely on a patchwork principle for pos-
sibility: if it is possible that X happen intrinsically in a spatiotemporal re-
gion, and if it is likewise possible that Y happen in a region, then also it is
possible that both X and Y happen in two distinct but adjacent regions.
There are no necessary incompatibilities between distinct existences.
Anything can follow anything.
Third: extending the previous point; it is possible that there might be
a world of stages that is exactly like our own world in its point-by-point
distribution of intrinsic local qualities over space and time.
Fourth: further, such a world of stages might also be exactly like our
own in its causal relations between local matters of particular fact. For
nothing but the distribution of local qualities constrains the pattern of
causal relations. (It would be simpler to say that the causal relations su-
pervene on the distribution of local qualities, but I am not as confident of
that as I am of the weaker premise.)
Fifth: then such a world of stages would be exactly like our own sim-
pliciter. There are no features of our world except those that supervene on
the distribution oflocal qualities and their causal relations.
Sixth: then our own world is a world of stages. In particular, person
stages exist.
Seventh: but persons exist too, and persons (in most cases) are not
person-stages. They last too long. Yet persons and person-stages, like ta-
bles and table-legs, do not occupy spatiotemporal regions twice over. That
can only be because they are not distinct. They are part-identical; in other
words, the person-stages are parts of the persons.
Let me try to forestall two misunderstandings. (1) When I say that

persons are maximal aggregates of person-stages, I do not claim to be re-
ducing "constructs" to "more basic entities", (Since I do not intend a re-

duction to the basic, I am free to say without circularity that person-stages
are aggregates of shorter person-stages.) Similarly, I think it is an infor-
mative necessary truth that trains are maximal aggregates of cars interre-
lated by the ancestral of the relation of being coupled together (count the
locomotive as a special kind of car). But I do not think of this as a reduc-
tion to the basic. Whatever "more basic" is supposed to mean, I don't
think it means "smaller." (2) By a part, I just mean a subdivision. I do not
mean a well-demarcated subdivision that figures as a unit in causal expla-
nation. Those who give "part" a rich meaning along these lines should take
me to mean less by it than they do. (Lewis [1983], pp.76-77)
In a way this is all very straightforward. In another way, it puzzles me

deeply. Let me try to explain my puzzlement. I shall begin by restating
(I hope that is what I am doing) Lewis's explanation. I shall, as Lewis
does, speak here only of the stages or parts of persons, but what I shall
say, like what he says, is easily generalized.
We could easily devise some sort of set-theoretic object to play the
role of the "career" of a person: perhaps a function from moments of
time to sets of momentary properties5 could be called a career, and a
given career could be said to be the career of a given object if its do-
main is just the set of moments at which the object exists and it assigns
to any moment just the set of momentary properties that object has at
that moment. A part of the career of a person or other object would
then be a function whose domain is a subset of the career's domain and
which, in its domain, takes on the same values as the career. It follows
from our definition of 'career' that a part of a career is itself a career,
though not necessarily the career of anything. A Lewis-part of a given
person is an object whose career is part of the career of that person. 6
Since I understand all these words, I understand 'Lewis-part' and
know what Lewis-parts are. In a way. In the same way as the way in
which I should understand talk of "propertyless objects" if I were told
that 'propertyless object' meant 'object of which nothing is true'; in
the same way as that in which I should understand talk of "two-
dimensional cups" if I were told that 'two-dimensional cup' meant
'cup that lies entirely in a plane'. These phrases would not be what one
might call "semantical nonsense" for me; they would not be like 'abra-
cadabara' or 'machine that projects beams of porous light' or 'Das

Nichts nichtet'. But I should hardly care to say that I understood what
someone was talking about (even if it were he who had given me these
definitions) who talked of propertyless objects or two-dimensional
cups, and who, moreover, talked of them in a way that suggested that
he supposed there were such things. For I cannot see how there could
be any such things. In fact, I think I can see clearly and distinctly why
there could not possibly be any propertyless objects or two-
dimensional cups (so defined). And this is very nearly the position I
am in with respect to Lewis-parts. I say "very nearly" because the idea
of a Lewis-part is obviously not an impossible idea, not an idea that
could correspond to no possible reality, for each of us obviously has at
least one Lewis-part: himself. (So much is immediately evident from
the definition.) But what about the "other" temporal parts of persons,
their proper temporal parts, the objects whose careers are proper parts
of the careers of persons? How can I say I cannot see how there could
be any such things in the face of Lewis's argument? I say this because I
do not understand the step in his argument labeled "second." The piv-
otal sentence in the step "second" is: 'It is possible that two person-
stages might exist in succession, one right after the other but without
overlap'. I cannot see how two person-stages could exist "in succes-
sion, one right after the other but without overlap." I will try, in
Quine'S words, to evoke the appropriate sense of bewilderment. God
could, I suppose, create ex nihi[o, and annihilate a year later, a human
being7 whose intrinsic properties at any instant during the year of its
existence were identical with the intrinsic properties of, say, Descartes
at the "corresponding" instant in, say, the year 1625. And if God could
do that, he could certainly create and annihilate a second human being
whose one-year career corresponded in the same way to the 1626-part
of Descartes's career. But could God, so to speak, lay these two crea-
tions end-to-end? (I ignore nice points about open and closed intervals
of time.) Well, he could create, and two years later annihilate, a human
being whose two-year career corresponded to the 162511626-part of
Descartes's career. He could do this, but I do not see what else, what
more, he could do to accomplish the goal of "laying these two crea-
tions end-to-end." What I cannot see is how, if God did this, it could
be that the "two-year-man" would have first and second "halves."
More exactly, I do not see how it could be that the first half of the two-
year-man's career could be the career of anything, and I do not see how
it could be that the second half of the two-year-man's career could be
the career of anything. When I examine the story of the creation and
annihilation of the two-year-man, I do not find anything in it that
comes to the end of its existence after one year: the only thing "there"
(as I see matters), the two-year-man, will not come to an end after one
year; he will, rather, continue to exist for another year. And, in the
same way, when I examine the story, I do not find anything in it that
begins to exist halfway through the story.
These remarks are not meant to be a refutation of Lewis's argu-
ment. They are meant only to identify the point in the argument at
which one philosopher, myself, parts company with Lewis. In identi-
fying this point, I am merely accepting Lewis's invitation: "I do not
suppose the doubters will accept my premises, but it will be instructive
to find out which they choose to deny."
To recapitulate: in virtue of Lewis's explanation of what temporal
parts are, I understand the term 'temporal part', but I do not see how
(in the sense of 'temporal parts' Lewis's explanation has supplied) a
thing could have temporal proper parts. And this is not the end of my
difficulties with proper parts - my difficulties, that is, with under-
standing how temporal parts (understood as "Lewis parts") could have
the features that those who appeal to temporal parts in their philo-
sophical work suppose them to have. When I look at temporal parts
through the lens Lewis's explanation supplies, I see things that seem to
me obviously to have modal properties at variance with the modal
properties that are commonly ascribed to temporal parts. It seems to
me to be obvious that the one temporal part of a thing must be
"modally ductile" and "modally compressible." Consider, for example,
Descartes and his one temporal part, himself. Descartes's one temporal
part - Descartes - existed from 1596 to 1650. The object that is called
both 'Descartes' and 'Descartes's only temporal part' might have ex-
isted for twice as long as it did (it is modally ductile), and it might
have existed for only half as long (it is modally compressible). It seems
obvious to me, moreover, that this object might have had entirely dif-
ferent momentary properties at the corresponding points in its career.
Suppose, for example, that Descartes had been stolen by Gypsies
shortly after he was born; if that had happened, then the object that is
in actuality Descartes's one temporal part would still have existed but
might well never have acquired the property "is able to speak French."
These modal propositions about Descartes's single temporal part seem
to be inconsistent with the modal properties that are ascribed to tempo-
ral parts by those philosophers who believe that there is useful philo-
sophical work for temporal parts to do, for these philosophers, or most
of them, seem to treat temporal parts as things that have their
"temporal extensions" and their careers essentially. Now I may be
wrong about this. I am doing no more than recording an impression. I
cannot point to any passage in which a philosopher has said in so many
words that temporal parts have either their temporal extensions or their
careers essentially. My point is only this: if there are philosophers who
think that temporal parts have their temporal extensions and their ca-
reers essentially, I cannot see how what they believe could be true.
Let us now return to the topic of identity across time. If there are,
as I believe, no temporal parts, or if each enduring thing has only one,
where does this leave us with respect to this notion? Surely it is tempo-
ral parts (or stages, or phases, or whatever) that are the terms of the
cross-time identity relation? If a thing has no temporal parts, or has
only one, what can be meant by the assertion that it exists at different
times?
3.
These questions conflate several issues.

First, whether or not there is such a thing as "identity across time,"
there are certainly what I have called temporal identity-sentences: sen-
tences that consist of 'is identical with' flanked by time-involving
definite descriptions. Tooley's sentence 'The book on the table at t is
identical with the book on the chair at t*' is an example of a temporal
identity-sentence, though, as we shall see, it is a rather special one.
And, of course, what we say when we utter temporal identity-sentences
is often true. Therefore, there are facts that we may call "facts of tem-
poral identity." Does the existence of these facts entail that there is
such a relation as "identity across time"? Well, that all depends on

what one means by 'identity across time'. These words are sometimes
used as a name for a relation that is not identity and which takes proper
temporal parts of enduring things as its terms. This conception is illus-
trated in Figure 1. (But the content of Figure 1 is not exhausted by the
proposition that there is such a relation. Figure 1 also illustrates a se-
mantical thesis: that the time-involving definite descriptions that figure
in temporal identity-sentences denote proper temporal parts.) The
words 'identity across time' are sometimes used as a special name for
identity, a name we call identity by when its terms are the four-
dimensional wholes that some philosophers take enduring objects to be
- rather as 'equality' is a name we call identity by when its terms are
numbers. This conception is illustrated in Figure 2. (But the content of
Figure 2 is not exhausted by the trivial thesis that four-dimensional
objects are identical with themselves. Figure 2 also illustrates a se-
mantical thesis: that the time-involving definite descriptions that figure
in temporal identity-sentences denote four-dimensional wholes and
apply to them only via their proper temporal parts.)
If 'identity across time' means, as the theory represented by Figure
1 says it does, a relation that takes proper temporal parts as its terms,
then there is no such thing, for there are no proper temporal parts. But
whether or not there is such a thing as "identity across time," there are
facts of temporal identity. (Moreover, if there are no proper temporal
parts, then the account of facts of temporal identity represented by Fig-
ure 2 is incorrect, since that account essentially involves proper tempo-
ral parts.)
How, then, should we understand facts of temporal identity? I be-
lieve that the first step towards understanding these facts must be to
dispense with pictures that depict enduring objects as being composed
of parts that are distributed along a time-axis. The next step is to re-
place such pictures with a very simple-minded picture:
on the table at t
the book that was

OD the chair at t*
Figure 3
In this picture, as in the others, the "book-drawing" represents a three-

dimensional object. (But this is dangerously close to a pun. "Slices"
are three-dimensional in the sense that they have an extension of
measure zero along one dimension and a non-zero extension in three
others. The book-drawing in Figure 3 has a non-zero extension in three
dimensions tout court. "Slices" do not endure through time - this is
what Wittgenstein would call a grammatical proposition - but are lo-
cated at a time; books endure.) As in Figure 1, each label is attached to
a three-dimensional object. As in Figure 2, each label is attached to the
same object. (Thus temporal identity sentences, according to the theory
represented in Figure 3, are straightforward expressions of numerical
identity.) Figure 3, unlike its two rivals, assimilates temporal identity-
sentences to other identity-sentences. 'The book that was on the table
at t is identical with the book that was on the chair at t*' differs from
'The book that Bill is reading is identical with the book that Tom is
looking for' in only one interesting respect: in the former sentence, 'is
identical with' is flanked by descriptions whose verbs are in the past
tense, and in the latter sentence it is flanked by descriptions whose
verbs are in the present tense. (Both sentences, of course, are equally
well represented by pictures showing one book twice labeled, which is
what would intuitively seem right for a sentence that pivots about 'is
identical with'.)
The acute reader will have noticed the transition in the preceding
paragraph (and in Figure 3) from Tooley-style time-involving descrip-
tions like 'the book on the chair at t' - verbless ones, that is - to the
time-involving descriptions containing tensed verbs that are the normal
time-involving descriptions of everyday discourse. In some simple
cases, the verb in such descriptions can be dropped: 'The book on the
table at noon was red', is good English, though, I would point out, a
'that was' is present "in spirit" in this sentence, even if it is unpro-
nounced, for the adverbial phrase 'at noon' modifies an understood
'was'. (Philosophers who suppose that such phrases as 'at noon' and
'in 1850' are adjectives describing the location of temporal slices or
whatnot are, as we have seen, mistaken.) In more complicated cases,
there is no possibility of wholly eliminating tensed verbs from time-
involving descriptions, as is shown by 'the car that used to be owned
by the man who will marry the woman who had been the first woman
President' .
There is one sort of temporal identity sentence that cannot be rep-
resented in any very straightforward way by a picture in the style of
Figure 3. I have in mind temporal-identity sentences containing so-
called "phase-sortals." (A phase-sortal is a count-noun such that a
given object may fall within its extension at one time but not at an-
other.) Consider, for example
The surgeon who removed the tumor from my brain is the boy
who once shined my shoes.
Obviously (one might argue) we cannot represent a fact of the sort this
sentence purports to express by a picture of someone twice labeled, for
no one is simultaneously a boy and a surgeon. And it will do the
friends of Figure 3 no good to protest that the 'is' in this sentence is a
rhetorical conceit, the strictly correct copula being 'was', since pictures
in the style of Figure 3 are unable to represent the use ofa tensed iden-
tity-sign.
There are various ways to deal with this problem. One way would
be to say that this sentence is not really an identity-sentence at all, but
a predication. That is, to say that what the speaker of this sentence is
doing is saying of a certain surgeon that he (the surgeon) is now such
that as a boy he shined his (the speaker's) shoes and was, moreover,
the only boy to do so. I shall not explore this avenue. I shall instead
investigate a way of treating this sentence as a real identity-sentence.
But what I am going to say is no more than a proposal for dealing with
phase-sortal identity-sentences. The "predication" analysis, or some
other analysis entirely, may tum out to be more fruitful.
A completely general treatment of phase-sortal identity sentences

as real identities would involve a lot of detail. I will show how to treat
a special class of phase-sortal identity-sentence, those in which phase-
sortals occur only in descriptions of the form
'the' + PHASE-SORTAL + RELATIVE PRONOUN +
PREDICATE.
(Our "surgeon" sentence is of this kind.) I do not think that a correct
treatment of any phase-sortal identity-sentence will differ in any inter-
esting way from the treatment I shall propose for sentences of this spe-
cial kind. I propose that phrases of the form displayed above should be
regarded as abbreviating phrases of the form
'the' + NON-PHASE-SORTAL + RELATIVE PRONOUN +
'when a(n)'+ PHASE-SORTAL + PREDICATE.
For example - assuming that 'human being' is not a phase-sortal - our
"surgeon" sentence can be regarded as an abbreviated version of
The human being who, when a surgeon, removed a tumor from my
brain is the human being who, when a boy, once shined my shoes.
If it is correct to regard our "surgeon" sentence as an abbreviation for
this sentence, then the former sentence presents no difficulties for the
"tag" model for understanding temporal identity-sentences, since there
is no difficulty in picturing a man bearing two labels, each of them
bearing one of the descriptions flanking the copula in this sentence;
and there is no difficulty in supposing that the man so pictured is cor-
rectly labeled.
4.
I turn finally to a nest of interrelated problems about pictorial repre-

sentation that are faced by anyone who holds that facts of temporal
identity are best represented by pictures like Figure 3.
Let us begin with some problems about temporal identity-
sentences whose subject no longer exists. For example:
The dog I owned in 1957 was the dog I owned in 1955.

(A moment ago, I made an imaginary critic say, ' ... pictures in the
style of Figure 3 are unable to represent the use of a tensed identity
sign." In the context in which I made him speak, he had a good point
But 'was' in this sentence is not a tensed identity-sign in the sense that
was there at issue. Its tense is not determined by the fact that 1955 was
earlier than 1957. It merely reflects the fact that the dog I owned in
1955 and 1957 is now dead. If I had wished to state the same fact of
temporal identity in 1958 - when the dog was still alive - I should
have used the sentence 'The dog I owned in 1957 is the dog I owned in
1955'.) One question we might ask is this. What are the tags to be at-
tached to, given that the thing they are supposed to be attached to no
longer exists? But this question confuses the represented object with
the representation. I am not saying that facts of temporal identity are
facts of this form: there are currently existing objects that could be
twice labeled with tags on which time-involving descriptions are in-
scribed. We are not talking about a possible practice of actually tag-
ging objects we could put our hands on; we are talking about drawing
pictures of tagged objects, the purpose of these pictures being to serve
as graphic, intuitive representations of a certain sort of fact. The tags
and the cords are merely diagrammatic representations of semantical
reference. And it just is a fact that we can refer to objects that no
longer exist: the fact that we can refer to Socrates is not a proof of the
immortality of the soul. There mayor may not be a philosophical
problem about how it is that we can refer to objects that no longer ex-
ist, but it does not seem to have much to do with the problem of ana-
lyzing temporal-identity sentences. That is, this problem, if it exists, is
equally a problem for the proponents of any theory of "identity across
time." Therefore, the correct pictorial representation of the fact ex-
pressed by the sentence displayed above is simply a picture of a dog (a
dog that no longer exists; but there can be pictures of dogs that no
longer exist) labeled both 'the dog I owned in 1957' and 'the dog I
owned in 1955'.
But this answer suggests a second and much more interesting
question. What exactly is the picture to show? Suppose my dog lost
her tail in 1956. Will the picture show a tailed or a tailless dog? 'Dog'
is probably not a phase-sortal, but dogs, like most things, can fall un-
der various phase-sortals (like 'tailless dog') and a picture can avoid
representing a dog as falling under a given phase-sortal only by es-
chewing detail; and a picture that eschews all detail is not a picture at
all, but, like the Bellman's map, a perfect and absolute blank.
This problem, the problem of what exactly the picture is to show,
is not a problem that the friends of Figure 3 face only when they are
attempting to generalize the device exemplified by Figure 3 to enable it
to depict those facts of temporal identity that involve objects that no
longer exist. Figure 3 itself presents them with this problem. Suppose
that Tooley's book had got a stain on its cover between t and t*. Shall
our diagrammatic representation of a book include a stain or not? We
can, of course, depict the book so sketchily that we do not have to de-
cide about that. But this tactic will not work in all cases: suppose we
are concerned with a temporal identity sentence about a human being
who is at one time a frail four-year-old girl and at another time a
grossly obese, bearded, six-foot-tall (surgical) male who has lost a leg?
Even a stick-figure has to have a definite number of legs. (And, of
course, someone who, like me, thinks that Danton was at one time a
fetus and at another time a severed head will sometimes find it even
more difficult to draw sufficiently sketchy pictures.) If we considered
only cases like that of the stained book and the girl-man, we should
probably be tempted to say that obviously, the picture should represent
the book (or whatever) as it is at the time the picture is scheduled to be
displayed. Thus, if the book was unstained at t, stained at t*, and now
once again unstained, the picture, if it is to be shown now, should rep-
resent the book as unstained. But we have already seen why this will
not work: there are temporal identity sentences whose terms denote
objects that no longer exist; and, of course, an object that does not now
exist cannot be depicted as it is now. (For that matter, there are iden-
tity-sentences that are not temporal identity-sentences and whose terms
no longer exist. 'The horse Caligula made a consul was the only horse
to hold political office', for example.)
The solution to this problem is a simple one: it does not make any
difference what the picture shows. Remember, the picture is only a
picture. Showing one thing twice-tagged is simply a device for graphi-
cally representing the fact that the descriptions inscribed on the tags
have the same referent. If we want to represent graphically the fact that
'the most famous teacher of Aristotle' and 'the most famous pupil of
Socrates' denote the same man, we have only to draw a man labeled
with tags bearing these phrases. We are no more constrained to draw
him at some particular age or in some particular condition or circum-
stances, than the author of an illustrated history of philosophy is con-
strained to choose a picture of Plato that shows Plato as being of some
particular age or as being in some particular circumstances. The author
of the illustrated history of philosophy knows that he may print a pic-
ture (a detail from the School of Athens, say), that represents Plato as
being - how else? - of a certain determinate age and in certain par-
ticular surroundings, and so on, and that it will be perfectly correct to
label it simply 'Plato'. (This label, by the way, will not be a description
of the picture; or not in the way 'Raphael, 1509-12, The Vatican' is. It
will be a description of the intentional content of the picture, like 'the
mechanism ofa watch' or 'the structure of RNA'. We should think of
the word 'Plato' printed under a picture as applying to the figure in the
picture and not to the picture.) That is, the label 'Plato' is correct tout
court and is not an abbreviation for 'Plato in old age' or anything else.
But if a detail from a picture showing a young man conversing with
Socrates and a detail from a picture showing an old man conversing
with Aristotle can both correctly be labeled 'Plato', then they can both
be correctly labeled 'the philosopher who, in middle age, founded the
Academy', because these two phrases denote the same object.
One minor point about the labeling of pictures. Suppose the author
of an encyclopedia article on General MacArthur accompanied his ar-
ticle with a single photograph of MacArthur, one taken when its sub-
ject was an infant and labeled 'Douglas MacArthur'; suppose the arti-
cle contained no word of explanation of the fact that it was accompa-
nied by a picture of an infant. There would be a lot wrong with that,
but the picture would not be mislabeled. David Lewis has reminded us
that someone's use of a sentence can be faulted on lots of grounds
other than falsity; similarly, someone's use of a captioned picture can
be faulted on lots of grounds other than the incorrectness of the cap-
tion. When I say that it "makes no difference what the picture shows,"
I do not mean to deny the obvious truth that it would be a queer thing
to do to represent the fact expressed by 'The father of Charles II was
the father of James II' by a picture of the infant Charles I twice-tagged.

This would be queer (because it would be wholly unmotivated) even if
both tags were inscribed in this style: 'the human being who, as an
adult, fathered the human being who, as an adult, became the second
king of England to be named Charles'. It would be a queer and unmo-
tivated way to represent that fact, but it would be a perfectly correct
way to represent that fact.
Figure 3, then, is the sort of picture I recommend that you use
when you think about facts of temporal identity. If you use any picture
at all, that is. I do not like pictures in philosophy, despite the fact that I
constantly use them in my own thinking. Even the best attempts to
picture essentially unpicturable things and states of affairs are bound to
be misleading. Even so simple a fact as the fact that 'the most famous
pupil of Socrates' denotes Plato can be represented only very imper-
fectly in a picture. For example, as we have seen, if we represent this
fact by a picture of an appropriately labeled man, our picture will show
far too little (nothing about the mechanics of securing reference, for
example) and far too much (the picture must represent Plato as having
properties that do not figure in the fact that he is denoted by the phrase
inscribed on the label). The mind is curiously unable to let the extrane-
ous features of the picture alone. My mind is, at least. It is only with a
real effort of will that I am able to keep myself from thinking that no
picture can really just be a picture of Plato, that every picture must, at
best, be a picture of Plato-at-some-particular-time. (To think this is to
make the mistake made by those philosophers who held that a geomet-
rical diagram cannot just represent a triangle.) Most of our discussion
of pictures in the style of Figure 3 has had only one object: to help us
to see that such pictures have, of necessity, extraneous features, and to
convince us not to attribute any significance to them.
There is another lesson that must be learned about these pictures.
They do not, by themselves, teach us anything about facts of temporal
identity. Their whole point is supplied by their rivals. If pictures in the
style of Figure 3 are misleading, their rivals are just ruinously wrong.
And these ruinous pictures underlie a lot of our thinking about time
and identity. They do not contribute to our arguments for or against
philosophical theories; they condition what premises we shall find
plausible. Pictures come before argument, and, therefore, pictures can-
not easily be dislodged from our minds by argument. I do not say this
is impossible. After all, I have tried to dislodge Figures 1 and 2 from
your minds by arguing that the temporal parts that figure essentially in
these pictures do not exist. But if your view of "identity across time" is
supplied by one or the other of these pictures, you will set out to find
some premise or inference in my arguments that you do not accept (if
you attend to these arguments at all). And, of course, you will succeed.
No attempt to refute a view that rests on powerful and appealing pic-
tures can hope to succeed unless it supplies a rival picture of its own.
And that is my only reason for asking you to consider Figure 3.
5.
I have called Figure 3 simple-minded, and I have said that it teaches us

nothing about facts of temporal identity. I have said these things be-
cause, in my view, facts of temporal identity are (per se) rather simple-
minded facts, and there is nothing to learn about them. There are facts
of temporal identity that present us with grave metaphysical problems.
There are, for example, facts about the persistence of objects through a
complete change of parts. But such problems are not problems about
facts of temporal identity, any more than problems about causal rela-
tions are problems about relations or problems about mental predicates
are problems about predication. For all I have said, there may be meta-
physical difficulties that infect every (alleged) fact of temporal iden-
tity; if so, these are metaphysical difficulties that are inherent in the
very notion of time and which infect every (alleged) fact that involves
the passage of time. In short, every one of the real problems about time
and identity is either too special or too general to be correctly describ-
able as "the problem of identity through time."
If this is so, why is there so very persistent a conviction among
philosophers that there is such a problem? I am not enough of an his-
torian of philosophy to answer this question. And I think that one
would have to be an historian to answer it. One would have to find the
most primitive, fumbling cases of wonder about the nature of "identity
through time," cases that were clearly differentiated both from wonder
about time itself and from wonder about special problems of identity
like those presented by the Resurrection of the Dead or the Ship of
Theseus. I would hazard a guess, however, that the root of the so-
called problem of identity through time has something to do with what
I have called "adverb pasting."
Here is a famous passage from Locke.
Wherein identity consists - Another occasion the mind often takes of
comparing, is the very being of things, when, considering anything as ex-
isting at any determined time and place, we compare it with itself existing
at another time, and thereon form the ideas of identity and diversity. (Es-
say, Book I, Ch. 27)
I have a hard time resisting the impression that Locke thought that 'it-
self existing at another time' was a name. Or perhaps, since the point I
want to make has nothing to do with pronouns, I should say that I have
a hard time resisting the impression that Locke thought that phrases
like 'Mary existing in 1689' and 'Mary existing in 1690' are names,
and, moreover, names for things that are in some sense two, even if
they are also in some sense one. (If Locke did think this, however, it
does not seem to have done his investigations of substantive problems
about vegetable, animal, and personal identity any harm. My purpose is
not to criticize Locke's whole treatment of his subject.) And it seems
even clearer to me that, if Locke did not accept this thesis, neither did
he reject it. Perhaps the most accurate thing to say is that the idea of
"the two Marys" touched the fringes of his thought so delicately as to
give him no occasion to ask himself what he thought about it. This
judgment of mine is a matter of "feel" and is probably one I, who am
no very experienced reader of seventeenth-century English prose, have
no business making. But I am made uneasy by 'itself existing at an-
other time'; why not 'itself as it had been at another time'? If someone
repeatedly makes judgments like 'Mary was sadder in 1690 than she
was in 1689' and 'Mary was wiser in 1685 than she was in 1680',
there is nothing really wrong with saying of him, "He's always com-
paring Mary as she was at one time with Mary as she was at another."
(But it would be better to say, " ... comparing the way Mary was at one
time with the way she was at another.") If, however, someone says, " ..
. comparing Mary existing at one time with Mary existing at another,"
one begins to wonder if he is not exhibiting at least some tendency to

think of phrases like 'Mary existing in 1690' as names. It is in this ten-
dency, I believe, that the problem of identity through time is rooted.
What was once only a faint tendency in the minds of a few people is
now an established habit of thought, one that is very hard to break,
and, indeed, very hard even to recognize as a mere habit. To the degree
that this habit is persistent, the tendency to think that there is a prob-
lem of identity through time will be persistent, since a philosopher who
thinks that 'Mary existing in 1690 and 'Mary existing in 1689' are
names for objects that are in some sense one and in some sense two,
will very likely want to spell out the relevant senses of 'one' and 'two'.
But this answer to the question 'Why do philosophers persist in
thinking that there is a problem of identity through time?', even if it is
correct, leaves a much more interesting question unanswered. What is
so special about time that philosophers should have a tendency toward
temporal adverb-pasting and no tendency to paste other sorts of ad-
verbs? Why is there no problem of cross-perspectival identity or cross-
evaluational identity? Why is there no tendency to think that 'Alice
viewed full-face' and 'Alice viewed in profile' (or 'Hume according to
A' and 'Hume according to B') are names for objects that are in some
sense one and in some sense two? To this more interesting question I
have no answer.8
Notes
1 The theory I am ascribing to Tooley is very like the theory I called "Theory 2"
in van Inwagen [1990a], pp.245-55.
2 This is the theory I called "Theory 1" in van Inwagen [1990a].
3 Quine [1953], pp.65-79.
4 Achille Varzi has asked me several questions about my analysis of the role of
'in 1850' in the sentence 'St. Paul's in 1850 was dingy'. They could be summed up in
this question: "Granted, 'in 1850' is not an adjective modifying 'St. Paul's'; but is the
only alternative that it is an adverb modifying 'was'? - are there not other possibili-
ties?" According to the traditional grammar I was taught in school, the copula 'was'
is "the verb of' the sentence, 'St. Paul's in 1850 was dingy' and 'in 1850' modifies it.
But a more up-to-date grammar might tell us that when 'to be' functions as a copula,
it does not belong to the grammatical category "verb" (despite the fact that 'to be',
whatever its function, displays the grammatical accidence traditionally definitive of
the category "verb": tense, voice, mood, aspect), but rather to the category "takes an
adjective and makes a verb." According to this view of the matter, the simplest verb
in 'St. Paul's in 1850 was dingy' is 'was dingy' - the past tense of the verb 'to be
dingy' - and that 'in 1850' modifies 'was dingy' (applies to it to produce the com-
plex verb 'was dingy in 1850'). Another up-to-date view of the function of ' in 1850'
in this sentence is this: 'in 1850' is not an adverb (a "takes a verb and makes a verb")
at all, but a sentence-modifier, a representative of the grammatical category "takes a
sentence and makes a sentence"; in the present case, it modifies 'St. Paul's was
dingy'. Which of these three accounts of the function of ' in 1850' is correct- ifthere
is indeed a fact of the grammatical matter - makes no difference for our purposes, for
each account has the consequence that 'St. Paul's in 1850' is not a noun-phrase and
hence does not represent itself as denoting an object. In the text, I presuppose the
traditional view, but the correctness of the traditional view is in no way essential to
my arguments. My arguments could easily be expressed in the terms provided by
either of the "up-to-date" accounts of the function of 'in 1850'.
5 A momentary property is a property an object could have at one time and lack
at another - like being seated and being Socrates' widow, and unlike being de-
scended/rom King David.
6 This definition leaves open the question whether there may be parts of a per-
son's career that are "topologically unsuited" to being the careers of objects. Sup-
pose, by suitable correlation of numbers with moments of time, we associate the do-
main of Descartes's career with the real numbers 0 through 1, inclusive (we should
be able to do this if there was both a first and a last moment of Descartes's exis-
tence). Could the part of Descartes's career whose domain is the rational numbers
between 0 and 1 be the career of an object? A part whose domain is a set that has no
Lebesgue measure? How about some relatively well-behaved (topologically speak-
ing) but non-connected set? - say, one corresponding to March 1610 and Good Fri-
day, 1633? These are questions that we can leave to the friends of temporal parts.
How they are answered is irrelevant to our argument.
7 I ignore Kripkean scruples about whether what was apparently a human being
that was created ex nihilo would really be a human being.
8 This paper first appeared in The Monist, 83 (2000), p. 437-459 and then as
Chapter 8 of van Inwagen [2001]. Some of the early parts of this essay (and a bit to-
ward the end) are taken from my essay "Plantinga on Trans-World Identity," in Tom-
berlin and van Inwagen (eds.) [1985]. Reprinted by premission: copyright © 2000,
The Monist, Peru, Illinois, 61354.
VAN INWAGEN ON TEMPORAL PARTS
AND IDENTITY ACROSS TIME
Andrea Bottani
I shall not argue against the two main theses of van Inwagen's paper:
the idea that persons, cats and trees are three-dimensional entities
lacking temporal parts; and the attack on the 'fallacy of adverb past-
ing', adverb pasting being a certain reading of the semantic role tem-
poral qualifications play in predicative sentences, i.e. the tendency to
attach them to singular terms rather than to predicates. Even if I shall
not discuss these central issues, I shall deny their purported relation-
ship, the idea that four-dimensionalism can be in some way "entailed",
"produced" or "generated" by the fallacy of adverb pasting. My thesis
will be that four-dimensionalism and adverb pasting are quite inde-
pendent of one another. In addition, I shall say something about the
notion of a temporal part. For a major aim of van Inwagen's paper is to
define a clear, neutral concept of a temporal part in terms of which the
thesis that ordinary objects have temporal proper parts can be under-
stood, and so asserted or denied. I will claim that van Inwagen's defi-
nition of 'temporal part' is not acceptable, nor does it make easier for
the enemy of temporal parts to state her case. In my commentary, the
latter point comes first, the former later.
413
Temporal parts and identity across time are matters of heavy philo-
sophical disagreement Some philosophers find it quite easy to grasp
the idea of a temporal part but extremely difficult to understand the
notion of identity across time (at least, the notion of a genuine, nu-
merical identity between things wholly existing at different times).
Other philosophers find it quite easy to grasp the notion of identity
across time but extremely difficult to understand the idea of a temporal
part (at least, the idea of a temporal proper part of an ordinary object
such as a person, a cat or a building). There is a stubborn temptation,
mirrored perhaps in the use of the words 'understand' and 'grasp' in
the above account, to see the disagreement at issue as one concerning
what can or cannot be conceived, rather than what can or cannot be
believed. In particular the enemies of temporal parts have often
seemed inclined to treat the notion of a temporal proper part of an or-
dinary object like a person, a cat or a tree as simply unintelligible -
which, at first sight, would deprive the thesis that ordinary objects
have temporal proper parts of any clear semantic content Neverthe-
less, most enemies of temporal parts firmly believe that events and
processes, such as wars and tennis matches, have proper temporal
parts. And most of them firmly believe that everything - continuant or
event - has at least one improper temporal part (i.e. itself). In order to
have such beliefs, one has to understand the general notion of being a
temporal part of something, and this clearly shows that what is at issue
here is not the intelligibility of some notion or other but the possibility
that some thesis is true: when the enemy of temporal parts says that the
notion of a temporal proper part of an ordinary object is unintelligible,
she is better interpreted as simply saying that it is impossible that per-
sons, cats and buildings have temporal proper parts. Obviously this is
not an argument, nor does it make easier for the enemy of temporal
parts to state her case. For ifit is difficult to prove that a given thesis is
false, it cannot be easier to prove that it is necessarily false.
Peter van Inwagen firmly resists the temptation to settle the debate
on temporal parts by relying to idiosyncratic semantic intuitions. In his
paper, he defines a neutral notion of temporal part in terms of which
the thesis that ordinary objects have temporal proper parts can be un-
Andrea Bottani, Van Inwagen on Temporal Parts and Identity across Time 415
derstood, then he argues against the thesis. Van Inwagen's notion of a

temporal part is defined in terms of the idea of a career, a career being
a function from moments of time to sets of properties. A career is the
career of an object if and only if "its domain is just the set of moments
at which the object exists and it assigns to any moment just the set of
momentary properties that object has at that moment" (the restriction
to momentary properties is important, a momentary property being "a
property an object could have at one time and lack at another"). A ca-
reer X can be said to be part of a career Y if and only if the domain of
X is a subset of the domain of Y and X, in its domain, takes on the
same values of Y (in short, iff X is set-theoretically included in Y). If,
in addition, the domain of X is a proper subset of the domain of Y, then
X can be said to be a proper part of Y. Finally, an object x can be said
to be a (proper) temporal part of an object y if and only if the career of
x is a (proper) part of the career ofy.
Since a career might well be the career of nothing, and this might
be the case of all careers that are proper parts of the career of a given
entity, there might be entities that have no temporal parts, and persons,
trees and buildings might be among them. So, this notion of temporal
part is neutral, leaving us free to assert or deny that ordinary objects
have any temporal proper parts. That an object x has no temporal
proper parts, in this sense of 'temporal part', merely means that the
career of x has no proper part that is the career of an object (in other
words, there is no object such that its career is a proper part of x's ca-
reer, i.e. all the careers that are parts of x's career are careers of noth-
ing). Since in van Inwagen's intention this explanation of 'temporal
part' merely restates Lewis's one, temporal parts in the above sense
are simply called 'Lewis-parts'. In the next few pages, I will argue that
van Inwagen's notion of a Lewis-part is not a good explanans of the
notion of a temporal part. Before the argument, however, some quali-
fications concerning the notion of a Lewis-part.
Van Inwagen's definition of 'career of an object' (and derivatively
of 'Lewis-part') employs the notion of a momentary property - de-
fined as "a property an object could have at one time and lack at an-
other" - and there may be doubts about the exact interpretation of that
notion. For the idea of a momentary property - so defined - seems to
be more naturally intended as relative rather than absolute: given a
property P, it may happen that an object could have P at one time and
lack P at another time while another object could not. In that case, we
should say that P is momentary for the first object but non-momentary
for the second, The property of being seated, for example, is momen-
tary for me but it is non-momentary for trees, for at no moment can a
tree be seated. And it is non-momentary for any stage of me whose ca-
reer has in its domain just one moment (for obviously no such tempo-
rally flat thing could have the property of being seated at one moment
and lack it at another). So, one possible interpretation of van In-
wagen's definition of 'career of an object' is that the career of an ob-
ject x assigns to any time t at which x exists a set of properties such
that: (i) they are momentary for x (i.e. x could have them at one time
and lack them at another); (ii) x has them at t. But this interpretation is
uncharitable, for it gives rise to several unpleasant consequences.
Suppose I am seated at a moment t. And imagine for the sake of
argument that there is a stage of myself in Lewis's sense whose career
has just t in its domain (call it L t , for short). My career would assign to
t a set containing the property of being seated, for that property is mo-
mentary for me, and I have it at t. But the property of being seated
would not be momentary for L t , whatever it may be, for L t could not
have that property at one time and lack it at another. I (Incidentally, this
would be true of any property of L t , so that no property instantiated by
Lt at t would be temporary for Lt , and Lt's career would assign to t an
empty set of properties). Then, under the uncharitable interpretation,
the career of L t would not assign to t a set containing the property of
being seated, and so Lt's career and my career would assign to t differ-
ent sets of properties. Hence, by definition of 'part of a career', Lt's
career would not be part of my career. And Lt. whatever it may be,
would not be a Lewis-part of me in van Inwagen's sense. For the same
reasons, mutatis mutandis, L t would not be a proper Lewis-part of any
object.
I conclude that, under the above uncharitable interpretation of 'ca-
reer of an object', no object could have a proper Lewis-part whose ca-
reer has in its domain just one moment. So, it would be possible for
any object x to have a Lewis-part y such that x ;;f:. y (i.e. y is a proper
part of x) and there is no z such that x = z+y (this would happen when-
ever y's career has in its domain all the moments which are in the do-
main of x's career except one). Far from being temporal parts in
Lewis's sense, Lewis-parts in this uncharitable sense would not even
be parts in the usual mereological sense.
All these consequences can be avoided by taking the expression
'momentary property' occurring in van Inwagen's definition of 'career
of an object' in an absolute rather than a relative sense. It is easy to
derive an absolute notion of momentary property from a relative one,
in one of the usual ways in which absolute notions can be derived from
relative ones: a property is momentary simpliciter if and only if it is
momentary for at least one object (i.e. at least one object can have that
property at one time and lack it at another). Accordingly, the career of
an object x should be conceived as a function which assigns to any
moment t at which x exists just the set of properties that (i) are mo-
mentary simpliciter (i.e. at least one object - not necessarily x - can
have them at one time and lack them at another); (ii) x has them at t. 2 I
assume this is the intended sense of van Inwagen's expression 'career
of an object'. The intended notion of Lewis-part has to be defined ac-
cordingly.
Even in this charitable interpretation of 'Lewis-part', however,
there is reason to doubt that temporal parts in Lewis's sense are just
Lewis-parts in van Inwagen's sense. Strictly speaking, this is only an
exegetical point, but it may be a good way of approaching some non-
exegetical problems. Imagine for the sake of argument that there is a
Lewis-part of me whose career has in its domain the moment t (for
brevity, call it L). And suppose I have at t the momentary property of
being a seated person. Then, L has to have at t the property of being a
seated person too, for L has at t just the momentary properties that L's
career assigns to t. And L's career, being part of my career, has to as-
sign to t just the set of momentary properties that I have at t (among
which, the property of being a seated person). But no stage of me in
Lewis's sense can have that property at any moment. For, according to
Lewis, person stages are not persons, and so they can never be seated
persons (at best, they can be seated person stages). Hence L, though
undoubtedly a Lewis-part of me in van Inwagen's sense, is not a stage
of me in Lewis's sense (it cannot be such a stage, because it comes out
with the wrong properties). Since the same, mutatis mutandis, is true
of any Lewis-part of anything, the point can be easily generalized as
follows: for all x and y such that x is a Lewis-part of y, there is no z

such that z is a stage ofy in Lewis's sense and z is (identical to) x. In
brief, Lewis-parts in van Inwagen's sense, even if they existed, would
not be stages (nor temporal parts) in Lewis's sense.
Perhaps, this conclusion can be avoided by keeping the property of
being a seated person outside the sets of properties the career of an
object x can assign to the moments at which x exists. For example, we
could require that x's career assigns to any moment t at which x exists
neither the set of all the properties x has at t nor the set of all the mo-
mentary properties x has at t, but (say) just the set of momentary and
simple properties x has at t. In this restricted sense of 'career of an ob-
ject', the set of properties that L's career assigns to t would fail to
contain the property of being a seated person. For that property,
though a momentary property of me, is not simple. So, nothing would
still compel us to say that L has at t the property of being a seated per-
son, and no reason would remain to say that L has at t a property that
no stage of a person can have at any time. As a consequence, Lewis-
parts could be unproblematic ally treated as stages (and temporal parts)
in Lewis's sense.
This restriction to simple temporary properties, however, may well
tum out scarcely intelligible, for we seem to have no clear intuition
about the simplicity or completeness of properties, but only, at best, of
predicates. And it seems to be false that simple predicates always ex-
press simple properties. (Take 'eats'. It is simple, but the property of
eating might well be thought of as complex, for it seems that nothing
can eat without being a living organism that assimilates food. This
might have been even more evident if English had no simple predi-
cates to express the property of eating). At any rate, the problem of
distinguishing between simple and complex properties is extremely
hard and it is to be hoped that the notion of a temporal part need not
depend on its solution. 3
I conclude that temporal parts in Lewis's sense cannot be con-
ceived as Lewis-parts in van Inwagen's sense. I do not think this is a
surprising conclusion. For conceiving temporal parts as Lewis-parts in
van Inwagen's sense is tantamount to reducing x's being a temporal
part of y to some sort of relation between the properties of x and those
of y (their reciprocal mereological relations excluded). And it would
seem to me to be very surprising if the mereological relation of being

(temporal) part of could be reduced to a set-theoretical relation be-
tween classes of properties. Even if possible, this reduction would be
useless. For, if our problem is to explain a neutral, transparent sense of
'temporal part' in which we can either affirm or deny the thesis that
ordinary objects have temporal parts, the current mereological notions
are sufficiently general, primitive and clear to do this explanatory job
much better than any artificial concept of 'Lewis-part' .4
I do not believe that reducing temporal mereology to something
else is what Lewis intends to do in the passage van Inwagen quotes. In
any case, no matter whether he intends or not, he ought not. For, ac-
cording to Lewis, we can explain the fact that one object can have dif-
ferent intrinsic properties at different times by assuming it has differ-
ent temporal parts existing at different times with different intrinsic
properties. But if having a part existing at a certain time with certain
properties would mean for an object just having certain properties at
that time, the appeal to temporal parts would explain nothing.
Secondly, whether or not person-stages in Lewis's sense are tem-
poral parts of persons rather than mere co-occupants of some spatio-
temporal regions occupied by persons, does not depend according to
Lewis on their properties (their reciprocal mereologlcal relations ex-
cepted) but on the assumption that persons and their stages cannot
"occupy spatiotemporal regions twice over" (see the seventh step of
Lewis's argument quoted by van Inwagen). And it could not be so if
x's being a temporal part of y consisted in a relation between x's and
y's properties. If, on the other hand, we admit co-occupancy, it may be
surprising how far we can parallel Lewis's solutions without admitting
that persons, cats or trees have any temporal proper parts (obviously,
we would parallel Lewis's solutions only formally, not in spirit). We
could admit the existence of all the person-stages Lewis speaks of (no
matter whether they are scattered or exist "in succession, one right af-
ter the other but without overlap") but treat them as mere co-occupants
of some spatio-temporal regions occupied by persons. The mereologi-
cal sum of all my stages - so treated - would be a process occupying
the maximal spatio-temporal region I occupy, that region being the
sum of all the spatiotemporal regions I occupy as a whole (consider
that on a three-dimensionalist approach there are many different spatio
temporal regions 1 occupy as a whole). That process would not be me.

Perhaps it would be a four-dimensional shadow of me. An alternative
but parallel treatment of temporary intrinsic properties would then be
open: instead of reading a la Lewis the sentence 'I am bent at t' as 'I
have a bent temporal part existing at t', we can read it as 'I have a bent
spatio-temporal co-occupant existing at t'. Instead of reducing a tem-
porary intrinsic property to a relation between a thing and a temporal
part of it, we can reduce that property to a relation between a thing and
a co-occupant of it.
1 do not think this treatment of temporary intrinsics is to be re-
garded as the best possible one (indeed, it has problems 1 refrain from
discussing now). If I have sketched it here it is only to show that x's
being a temporal part of y cannot consist in a certain link between the
properties of x and those of y (their reciprocal mereological relations
excluded). For, whenever there is the required link, we can indiffer-
ently maintain either that x is a temporal part of y or that it is a co-
occupant of the spatio temporal regiony occupies.
Whenever there are any candidates for being temporal proper parts
of ordinary persisting objects and we think that ordinary persisting
objects do not have any temporal proper parts, we can follow one of
two strategies. We can argue that those candidates would have good
credentials if they existed but they do not exist, or instead deny they
have good credentials, no matter whether they exist or not. Van In-
wagen moves mainly in the former direction, for he denies the exis-
tence of any stage of me shorter than me (in order to deny that, how-
ever, he conceives stages as 'Lewis-parts', and 1 have tried to show
they cannot be so conceived). The latter direction seems to me to be
more promising: maintaining that stages, no matter whether they exist
or not, do not have the right credentials to be treated as temporal parts
of ordinary objects such as persons, cats or trees (at best, they are
events and no event can be a proper part of something which is not an
event).
2.
'Adverb pasting' in van Inwagen's sense is a certain reading of the

semantic role temporal qualifications play in predicative sentences. It
is the tendency to attach them to singular terms rather than to predi-

cates. To attach an expression to another is something one does. But
the idea that it is correct to do so is a thesis. So, in different contexts,
adverb pasting can be understood either as an action (or rather, a dis-
position to action) or as a thesis. Both as an action and as a thesis, ad-
verb pasting is, according to van Inwagen, a gross mistake. Also four-
dimensionalism, according to him, is a gross mistake and the latter
mistake is produced by the former.
I agree that adverb pasting and four-dimensionalism are mistakes.
It is about the aetiology of four dimensionalism as produced by adverb
pasting that I would like to say something. It is not completely clear to
me whether the aetiology is to be understood as a logical one (the the-
sis being that the fallacy of adverb pasting entails, and perhaps is en-
tailed by, four-dimensionalism), as a historical one (the thesis being
that four dimensionalism has been historically produced by the fallacy
of adverb pasting), or as a psychological one (the thesis being that it is
the fallacy of adverb pasting that makes psychologically possible or
easier the belief in four-dimensionalism).
Indeed, however, I would not know what to say about the last two
kinds of aetiological explanation. Maybe other factors have been psy-
chologically, sociologically and historically more important than ad-
verb pasting in determining four-dimensionalism. The symmetrical
treatment of space and time in contemporary physics might be one of
them. However, it might be that the problem of explaining the psy-
chological or historical rise of four-dimensionalism is not - after all -
philosophically so important. It might even be a bit disrespectful to the
four-dimensionalist, for nobody likes to see her own beliefs considered
merely in terms of their historical or psychological causes rather than
of their evidential or justificatory support.
The important problem seems to me to be: does adverb pasting
(this semantic account of temporal phrases) entail four-dimensionalism
(the metaphysical thesis according to which any persisting entity has
temporal proper parts)? And conversely: does four-dimensionalism
entail the thesis that adverb pasting is a good semantic account of the
role temporal qualifications play into predicative sentences? In other
words, does the semantic thesis entail the metaphysical one and does
the metaphysical thesis entail the semantic one? The correct answer
seems to me to be doubly negative: neither three-dimensionalism ex-

cludes adverb pasting nor four-dimensionalism excludes the thesis that
adverb pasting is a fallacy. So, the semantic thesis and the metaphysi-
cal thesis seem to me to be as independent as they could be.
Take van Inwagen's example: 'St. Paul's in 1850 was dingy'.
Suppose the subject of this sentence is 'St. Paul's' while its predicate
is 'in 1850 was dingy'. And suppose each sentence of the same form
has a subject of the same form and a predicate of the same form. In
brief, suppose that adverb pasting is a fallacy (incidentally, as I have
said above, I agree that it is). Would this entail that St. Paul's - as well
as any other object - has no temporal proper part? I think not. To see
why not, consider the thesis that ordinary objects have spatial parts. I
assume that most of us agree on that thesis: buildings, trees and seas
have spatial parts. But suppose someone comes in and says: "there are
no spatial parts of an object, the idea that there are is produced by the
fallacy of adverb pasting, the tendency to attach topological phrases
occurring in predicative sentences to their subjects rather than to their
predicates". Well, I think this would be utterly unconvincing. But the
reason why it would be so unconvincing is not that adverb pasting is
more correct in the case of topological phrases than it was in the case
of temporal phrases. On the contrary, the subject of the sentence 'The
Mediterranean sea is deep in front of Genoa' seems to me to be 'the
Mediterranean sea' and not 'The Mediterranean sea in front of Genoa'.
The reason why this would be unconvincing is that most of us (I sup-
pose) have a natural tendency to think that the Mediterranean sea can
satisfy the predicate 'is deep in front of Genoa' (or have the property
of being deep in front of Genoa) only if there is a spatial part of the
Mediterranean sea that is located in front of Genoa, and that part is
deep. But this is not a thesis about the logical form of the sentence 'the
Mediterranean sea is deep in front of Genoa' - and still less a thesis
about the logical role of the topological phrase 'in front of Genoa' . It is
a semantic thesis concerning the meaning of the predicate 'is deep in
front of Genoa'. What makes temporal parts reasonable is the assump-
tion that 'was dingy in 1850' has the same kind of meaning as 'is deep
in front of Genoa': the assumption that something can be dingy in
1850 if and only if it has a temporal part which is located in 1850, and
that part is dingy. This assumption has nothing to do with adverb
pasting. In fact, it can come to the surface only after we have discarded
adverb pasting, i.e. only after we have decided to attach the temporal
phrase 'in 1850' to the predicate 'is dingy' rather than to the subject
'St. Paul's'.
lf this is not enough, consider the thesis that processes like foot-
ball matches have temporal parts. I assume most of us agree on that: it
might be that football matches do not exist at all but, if they exist, then
they must have temporal parts (having temporal parts is the only way
they can exist at all). Well, suppose someone comes in and says: "that
matches have temporal parts is a fallacy springing from adverb past-
ing; we have only to realise that the subject of the sentence 'the foot-
ball match was enjoyable during its first ten minutes' is 'the football
match' - and not 'the football match during its first ten minutes' - to
see that football matches do not have temporal parts". Well, I think
this would simply miss the point. And the point is that a football match
can have the property of being enjoyable during its first ten minutes
only if the first ten minutes of the football match have the property of
being enjoyable. Still, this point can be made only after having at-
tached the temporal phrase 'during its first ten minutes' to the predi-
cate 'is enjoyable' rather than to the subject 'the match' - and so, only
after having discarded adverb pasting.
My conclusion· is that we have no need to subscribe to adverb
pasting in order to think that ordinary persisting objects have temporal
parts. We can believe in four-dimensionalism - and think that there is
an important and meaningful "problem of identity across time" - while
keeping firmly away from adverb pasting. In a nutshell: four-
dimensionalism does not entail adverb pasting.
So, I come to the other direction of the implication: does adverb
pasting entail four-dimensionalism? Does the idea that the sentence
'St. Paul's in 1850 was dingy' is about st. Paul's in 1850 entail that St.
Paul's has temporal parts? One could be tempted to answer negatively,
arguing as follows: that the sentence is about St. Paul's in 1850 does
not exclude that it is about S1. Paul's, for St. Paul's in 1850 might sim-
ply be St. Paul's. After all, the temporal part of St Paul's which occu-
pies the year 1850 might well be an improper part of St. Paul's. If so,
the temporal phrase 'in 1850' would be attached to the name 'St.
Paul's', but nevertheless St. Paul's would have no proper temporal
parts, for it would have only one improper part - St. Paul's itself - and
four-dimensionalism would be false.
Well, I think this would be a bad argument. For, what about the
sentence 'St. Paul's in 1999 was not dingy'? IfSt. Paul's has only one
temporal part - itself - then 'St. Paul's in 1999' denotes exactly what
'St. Paul's in 1850' denotes, i.e. St. Paul's itself. But then, how can
this object be at once dingy and not dingy? If the temporal phrases 'in
1850' and 'in 1999' attach to the name 'St. Paul's', then the terms 'St.
Paul's in 1850' and 'St. Paul's in 1999' cannot be co-referential, for
otherwise the truth value of the sentence 'St. Paul's in 1850 was
dingy' would be insensitive to the replacement of 'in 1850' with 'in
1999' - and we know that it is not so insensitive. That is the reason
why even Tooley's theory does not seem to work: for, according to
Tooley, 'the book on the table at t' and 'the book on the table at t*'
denote the same book (the same set of temporal parts of a book), and
so it is not easy to see how both the sentence 'the book on the table at t
is stained' and the sentence 'the book on the table at t* is not stained'
could be true (how could the same book be both stained and not
stained?).
So, we might conclude: if we attach temporal phrases to subjects
and not to predicates, then we are compelled to think that an expres-
sion resulting from attaching a temporal phrase to a subject denotes a
temporal proper part of what the subject denotes. For otherwise we
could not explain, for example, why 'St. Paul's in 1850 was dingy' is
true and 'St. Paul's in 1999 was dingy' is false. So, the adverb paster
cannot keep away from four-dimensionalism, understood as the thesis
that ordinary persisting objects have temporal proper parts.
Still, however, I think this would be an over-hasty conclusion, for
the temporal phrase 'in 1850' might be attached both to the subject
and to the predicate. This, I assume, is what happens when 'in 1850' is
applied to the whole sentence 'St. Paul's is dingy' (likewise, when we
say that in a possible world W John is rich, what we seem to say is that
the individual that John is in W is rich in W). So, the adverb paster can
keep away from four-dimensionalism if he is careful enough to attach
temporal phrases also to predicates and not only to singular terms. His
position would be as follows.
1. 'St. Paul's in 1850 was dingy' has the following structure: 'St.
Paul's in 1850' is the subject, 'is dingy in 1850' is the predicate.
2. 'St. Paul's in 1999 was dingy' has the following structure: 'St.
Paul's in 1999' is the subject, 'is dingy in 1999' is the predicate.
3. The subjects of the two sentences are co-referential: both denote
St. Paul's.
4. The first sentence is true, the second sentence is false.
5. The reason why the two sentences differ in truth value is that they
apply different predicates to co-referential terms.
Since this position seems to me to be perfectly coherent, I conclude

that adverb pasting (the thesis that temporal phrases in predicative
sentences attach to subjects and not to predicates) does not entail four-
dimensionalism. In fact, adverb pasting and four-dimensionalism are
quite independent theses. At least in this case, we can move independ-
ently on the logical and on the metaphysical level.
I cannot say exactly whether this conclusion is to be read as a criti-
cism of van Inwagen's position, for I am not able to find in his paper
the explicit assertion that four-dimensionalism entails adverb pasting,
or vice versa. And, on the other hand, I have not denied that adverb
pasting can make four-dimensionalism more plausible or easier. All I
have aimed to show is that four-dimensionalism does not logically en-
tail adverb pasting, nor is logically entailed by it.
Notes
1 It might be objected that what is a temporally flat stage of me in the actual
world could be a temporalIy thick object in some other possible worlds. If so, LI
could lose and gain the property of being seated, even if it does not do so in the ac-
tual world. I have no clear intuitions about that matter, particularly if the background
modal semantics is counterpart theory. For the relation of counterparthood is incon-
stant, and I am not sure there cannot be a sense of 'counterpart of in which a tempo-
rally flat stage of me has in some worlds one or more temporally thick counterparts
(though that would seem to me to be very strange). What seems to me to be certain is
the general point that, if objects have stages in Lewis's sense, then there are proper-
ties which are temporary for some objects but not for some of their stages in Lewis's
sense. Take a biological organism of the species homo sapiens and the zygote it was.
Lewis would say that the zygote is a stage of the organism. But the property of being
seated, though temporary for the biological organism, is not temporary for the zy-
gote, for no zygote can be seated at any time (not even, I would dare to say, in coun-
terpart theory). Since the same holds for the property of not being seated, the zy-
gote's career could not be part of the organism's career, and so the zygote could not
be a Lewis-part ofthe organism.
2 May be that all qualitative properties, even dog or animal, are momentary in
this weakest sense. For, if we observed a dog gradually developing into a lycaon, we
would not say that one living organism is dead and another has taken its place (the
same, mutatis mutandis, if we observed a plant gradually developing into an animal
in virtue of its metabolism). If so, the charitable interpretation of 'Lewis-part' would
not be so charitable. For the property of being a person would be momentary sim-
pliciter and so it would be in the set of properties that the career of any person as-
signs to any moment at which that person exists. But the property of being a person
would never be in the set of properties the career of a stage of a person assigns to a
moment, for no person stage is a person at any time. Hence, no stage of a person (no
matter whether person stages exist or not) could be a Lewis-part of any person. After
all, however, there might be at least some qualitative properties which are not mo-
mentary simpliciter (i.e. no object can have them at one moment and lack them at
another), in particular all last sortal properties, such as living organism, and perhaps
person.
3 At any rate, even if the problem could in some way be resolved, the restriction
to simple temporary properties would not solve the problem it is intended to solve.
For consider the property of being old. It does not seem to be a complex property,
and it is undoubtedly temporary (objects can gain it). But, if an object x is old at a
time t, certainly there are many stages of x in Lewis's sense that are not old at t
(some of them may even be born just at t, though all are stages of something that is
old at t). On the contrary, no Lewis-part of x in van Inwagen's sense can exist at t
without having at t the property of being old, for Lewis-parts have to inherit, at any
moment t of their existence, all the temporary properties that the objects they are
Lewis-parts of have at t. Once again, the conclusion is that many Lewis-parts in van
Inwagen's sense are not temporal parts in Lewis's sense. Moreover, in the previous
sentence, 'many' can be replaced by 'all' on the following grounds. Suppose that, for
any precise age an object can be, there is a property X such that any object has X iff
it is exactly that age. And suppose that, for any precise length of time an object will
still persist, there is a property Y such that any object has Y iff it will persist exactly
for that length. If there are such kinds of properties, at any moment any object has to
have exactly one property of the first kind and one property of the second kind. But,
for any property of the first kind and any property of the second kind an object x has
at t, no y can have the same two properties at t and be a proper temporal part of x.
For a temporal part of an object x can be proper only if it is shorter than x. Hence, no
object can have a proper temporal part whose career is part of its career, and no
proper temporal part of any object can be a Lewis-part ofthat object.
4 I have argued above that, when the enemy of temporal parts claims that the
notion of a temporal part is unintelligible, she is better interpreted as simply saying
that it's impossible that ordinary things have any temporal parts. If so, there is no
notion to be in need of explanation here.
CHANGE AND CHANGE-ERSATZ
UweMeixner
Like other sciences, metaphysics must save the phenomena. But, usu-
ally, this can be done in a variety of ways. The problem is to give good
reasons for preferring one way over the other. Take such a familiar
phenomenon as change. Here we find a metaphysical position that, on
the face of it, seems to deny that there is any change at all; I shall call
it supereternalism. Yet, philosophers on the side of supereternalism
have worked hard to make it save the phenomenon of change - with
truly remarkable success. The feeling remains that, nevertheless, a
non-supereternalistic conception of change is more appropriate. This
paper will give reasons for believing this. But before coming to su-
pereternalism and change, I want to discuss something else.
1. SUPERESSENTIALISM AND CONTINGENCY
The rationale for this is to draw what I hope is an illuminating analogy

between the following two pairs of concepts, whose members are ap-
parently rather ill-matched and bound to conflict, the modal pair: su-
peressentialism and contingency, and the temporal pair: supereternal-
ism and change. Structurally, the conceptual relationships are in both
cases rather similar.
427
Consider then the following statement:

(A) If an object had properties different from those it has in fact, it
would be different from what it is in fact.
Prima facie this seems to be obviously true. But reflection reveals an
ambiguity in it. If (A) is read in the sense of
(AI) If an object had properties different from those it has in fact,
it would be qualitatively different from what, qualitatively, it
is in fact,
then (A) is indeed obviously true. But if(A) is read in the sense of
(A2) If an object had properties different from those it has in fact,
it would be numerically different from what, numerically, it
is in fact,
then the truth of (A) is, after all, doubtful. For, in view of the fact that
no object can be numerically different from what it is numerically in
fact (namely, itself), (A2) is nothing else than a little more compli-
cated way of stating that no object can have other properties than it has
in fact. And this, a claim of superessentialism, is of course a doubtful
claim (although, irritatingly, there is also an unintended way of reading
the claim in which it is trivially true).
But it is easy to conflate the doubtful (A2) with the trivial (AI),
the manner of conflating them being exhibited precisely by the am-
biguous sentence (A). Although it is hardly believable, conflation of
qualitative and numerical sameness may be one of the reasons,l why
philosophers like Leibniz and Lewis thought it to be plausible that, in
a sense, every object necessarily has the properties it has in fact - in a
sense, for both philosophers have attempted to preserve our intuitions
of contingency. For them, in a different sense from the sense just men-
tioned (and for Lewis in a vastly more important sense), there are after
all objects that do not necessarily have all the properties they have in
fact. Thus Leibniz and Lewis are both superessentialists who never-
theless want to hold on to contingency.
Before discussing this, let me briefly elaborate on the concepts
employed in (AI) and (A2). Though both are expressed by 'is', quali-
Uwe Meixner, Change and Change-Ersatz 429
tative being and numerical being are vastly different concepts, as can
be seen from the two principles that govern their use:
(Bl) For every object x and every property fof objects: x is quali-
tatively fiffx hasl
(B2) For every object x and object y: x is numerically y iff x is
identical with y.
Given this, (AI) simply says: if an object had different properties from
those it has in fact, it would have different properties from those it has
infact. Nothing could be more certain than this. But (A2), on the other
hand, means: if an object had different properties from those it has in
fact, it would not be identical with the object it in fact is numerically
identical with. This is rather doubtful.
Suppose we nevertheless accept (A2), undaunted by the charge
that we are apparently confusing qualitative and numerical being, the
concepts which Plato laboured to distinguish in the Sophist. How can
we then preserve our intuitions of contingency?
The answer of counterpart theory is very well known: although, in
a certain sense, no object can have other properties than it has in fact,
there is a sense, and it is alleged to be the only truly relevant sense, in
which an object can have other properties than it has in fact. This
sense is given in the following two contingency-descriptions:
(Cl) Let x be an object that does not in fact have the property I
But it is possible that x has f (and it is therefore contingent
that x does not have f), since there is a possible world wand
a counterpart y of x in w such that y has fin w.
(C2) Let x be an object that in fact has the property I But it is pos-
sible that x does not havef(and it is therefore contingent that
x has f), since there is a possible world wand a counterpart y
ofx in w such thaty does not havefin w.
The immediate impression one obtains from (Cl) and (C2) is that the
contingency that counterpart theory is able to provide is nothing but
counterfeit contingency, at best an ersatz for genuine contingency. For
a possibility that is a possibility for an object x only in virtue of its be-
ing an otherworldly fact for another object y, different from x, al-
though in a certain manner similar to x, is really only a vicarious pos-

sibility for x, and thus not a genuine possibility for x at all.
But obvious as this criticism may be, it has not made any impres-
sion on the counterpartists, notably on David Lewis. They persist in
their ways. If the discussion of metaphysical matters is more than an
intellectual game where one never gives up unless one is shown to be
inconsistent, then the reason for this persistence can only be that in the
question of what is genuine contingency, and what is counterfeit, the
answers are not as obvious as for some people, notably myself, they
seem to be. 2
It is instructive to compare contingency according to counterpart
theory with contingency according to Leibniz. Adam, the first human
being, sinned; but normally (normal being here what is theologically
orthodox) one would say that Adam's sinning was not necessary, but
contingent: although he sinned, Adam might not have sinned. Leibniz
was far from denying this. But what does the contingency of Adam's
sinning mean according to Leibniz, who also held that, in a sense,
every object has every property it has necessarily, since he equated an
object's having a property fwith the property fbeing intensionally in-
cluded in the (necessary) individual concept of the object? The answer
is implicit in a letter to Antoine Arnauld from June 16863 : Adam's sin-
ning is contingent since there is a possible Adam who did not sin and
whom God could have made actual instead of the real Adam who
sinned.
Essentially, this a theologically-phrased version of the contin-
gency conception of counterpart theory. The "possible Adams" are
precisely the counterparts of the real Adam, and God's (unexercised)
ability to make actual some possible Adam who did not sin is, in view
of the Leibnizian presuppositions of God's omnipotence and of the
world-implying (or "world-mirroring," as Leibniz says) of each object
(each implies the world it belongs to and in which it has its properties:
it cannot be realized without that world), equivalent to there being a
possible world in which a counterpart of the real Adam (the counter-
part in that world of Adam) did not sin.
Thus the same critical comment fits both Lewis' version of con-
tingency and Leibniz': vicarious contingency is not genuine contin-
gency. Besides being certainly not theologically satisfactory, vicarious
contingency is also not satisfactory ethically. 'Adolf could have done

otherwise' - this, in order to express the ontological basis of the moral
responsibility of Adolf for his deeds, must mean more than that there
is somebody similar to, but not identical with, Adolf, who does other-
wise in some possible world that is appropriately related to the real
world.
2. SUPERETERNALISM AND CHANGE
After having considered an analogous issue, consider now the issue of

superetemalism and change. The following statement is a temporal
parallel to the modal (A):
(D) If an object has at a time properties different from those it has
now, it is at that time different from those it is now.
Prima facie this seems to be obviously true. But reflection reveals an

ambiguity in it. If (D) is read in the sense of
(Dl) If an object has at a time properties different from those it
has now, it is at that time qualitatively different from what,
qualitatively, it is now,
then (D) is indeed obviously true. But if (D) is read in the sense of
(D2) If an object has at a time properties different from those it
has now, it is at that time numerically different from what,
numerically, it is now,
then the truth of (D) is doubtful indeed. For in view of the fact that no
object is ever numerically different from what it is numerically now
(namely, itself), (D2) is nothing more or less than a more elaborate
way of stating that no object ever has other properties than it has now,
which is a claim of supereternalism (or perhaps better, but certainly
longer: supersempiternalism).
Superetemalism flies even more in the face of intuition than does
superessentialism, for superetemalism implies that at least in a sense,
but prima facie not at all in a marginal sense, there is no change, and
hence it appears to be contradicted by everyone's direct experience

which seems to demonstrate the existence of change at every tum.4 But
hardly any superetemalist in the history of philosophy has had the
courage of Parmenides to deny outright that there is change, and
consequently to count for nothing the testimony of experience. 5 A
superetemalist will usually try to avoid the blunt claim that there is no
change. The usual strategy is as follows: "Yes, in a certain sense of the
word 'change' there is no change, but that is not the important sense.
In the truly important sense, which is the sense that I have to offer and
which is all the sense of 'change' we need," says our average
superetemalist, "there is of course change."
What, then, can be the superetemalist's meaning of the word
'change'? One such meaning is encapsulated in the following two
change-descriptions:
(E1) Let x be an object that does not now have the property f -
some property that is appropriate for change (not all proper-
ties are appropriate for change: those which are inappropri-
ate are at most good for so-called "Cambridge changes").
But it will change from not havingfto havingf, since there is
some later time t and a temporal counterpart y of x at t that
has the property f at t.
(E2) Let x be an object that now has the property f - some prop-
erty that is appropriate for change. But it will change from
havingfto not havingf, since there is some later time t and a
temporal counterpart y of x at t that does not have the prop-
erty fat t.
Thus there is a sense in which one can have change, although no object
whatever changes in the sense that it ever has any other properties ap-
propriate for change than those it has now. It is a curious concept of
change, and yet superetemalists could think it to be the adequate con-
cept of change. Indeed they could roundly reject the charge that they
are superetemalists. "Of course," they could say, "we believe in ob-
jects that at other times have different properties than they have now.
Almost every object is such an object." Looking closer, one would
find that the objects that are alleged to have at other times other prop-
erties than they have now, are, in point of fact, taken by the supereter-
nalists to have these other properties merely by analogy, as the owner
of a dog may be said, in a sense, to have a tail, since his dog has a tail.
Undoubtedly, the relationship between the temporal counterparts of an
object and the object itself is in some sense closer than the relationship
between a dog-owner and his dog. This has the consequence that the
supposedly analogous having of a property may be said to be much
more similar to properly having (or possessing) a property than the
analogous having of a tail is similar to properly having a tail. Never-
theless, just as one can comment that the owner of the dog does not,
properly speaking, have a tail, one can also comment, that an object, if
supereternalism is correct, never ever has, properly speaking, other
properties than it has now; it has other properties only by analogy.
Let me briefly indicate how supereternalists can deal with a spe-
cial kind of change, namely coming to be and passing away. This kind
of change poses no special problem for them, but is subsumed under
(El) and (E2), if existence is taken to be a property of objects that is
appropriate for change. Yet, to regard existence as a property of ob-
jects with respect to which they change, immediately implies, even
under (E1) and (E2), that there are objects that at some time do not
exist, that is: it implies temporal possibilism. Therefore, if supereter-
nalists deny temporal possibilism and affirm that all objects exist at all
times and nevertheless do not want to deny coming to be and passing
away, they must explicate coming to be and passing away in a differ-
ent way than by subsuming them under (El) and (E2). Under temporal
actualism, existence, if a property at all, is quite obviously not a prop-
erty that is appropriate for change.
How can they do it? Thus:
(Fl) An object x will come to be if there is no temporal counter-
part of x now, but also a future time at which there is a tem-
poral counterpart ofx. 6
(F2) An object x will pass away if there is a temporal counterpart

of x now, but also a future time at which there is no temporal
counterpart of x. 7
3. PRINCIPLES AND THEOREMS OF TEMPORAL

COUNTERPART THEORY
Let me state some principles for temporal counterpart theory - the the-
ory which is the basis of a superetemalistic conception of change:
(Gl) For every object x there is precisely one time t such that x is
at t a temporal counterpart of x. (Uniqueness of lemporallo-
calion.)
(G2) For every object x and time t there is at t at most one tempo-
ral counterpart of x. (Uniqueness ofa temporal counterpart.)
(G3) For every object x, object y and time I: if y is at t a temporal
counterpart of x and x is different from y, then x is at t not a
temporal counterpart of y. (Restricted asymmetry of the tem-
poral counterpart relation.)
(G4) For every object x, object y and time I: if y is at t a temporal

counterpart of x, then y is at I a temporal counterpart of y.
(Localedness ofbeing a temporal counterpart.)8
(G5) For every object x, object y, time t and time t': if x is at I a

temporal counterpart of x and y is at t' a temporal counterpart
of x, then x is at t a temporal counterpart of y. (Reflection of
Ihe lemporal counterpart relation over time.)
(G6) For all objects x, y and z, times t, (': if x is at I a temporal

counterpart of y, and y is at (' a temporal counterpart of z,
then x is at ( a temporal counterpart of z. (Linking of (empo-
ral counterparts.)
Note that the majority of these principles could not be justified by de-
fining temporal counterparthood as a similarity concept. One might
think of the following definition: x is at ( a temporal counterpart of y
=def X is a t-Iocated object which is (among all the t-located objects)
maximally similar to y, presupposing that for every object x there is
precisely one time t such that x is a t-Iocated object (which presuppo-
sition makes it possible to derive (Gl)). But this definition does not
help to establish (G2): why should not an object have at a time ( two or
more t-Iocated objects that are all maximally similar to it? Nor does it
help to justify (G5): although y is a t'-located object that is maximally
similar to x, and x is a t-Iocated object (that is maximally similar to x),
that does not mean that x is a t-Iocated object that is maximally similar
to y; there may be a t-Iocated object that is more similar than x to y.
And quite clearly, the definition does also not help to establish (G6).
Here are the two main theorems that follow from the stated princi-
ples:
GTl: 'x is an object and y is an object and 3t' (t' is a time and x is
at t' a temporal counterpart of y)" or in other words 'x is
simpliciter a temporal counterpart of y' expresses an
equivalence relation over objects.
Proof
DEFl: R(x, y) =def. X is an object and y is ail object and 3t' (t' is a
time and x is at t' a temporal counterpart of y) [x is sim-
pliciter a temporal counterpart ofy].
Then: (1) For all objects x: R(x, x) (according to (G 1). (2) For all x
and y: R(y, x) ~ R(x, y): Assume: R(y, x); we have: x is at a time t a
temporal counterpart of x (according to (Gl)); hence according to
(G5): R(x, y). (3) For all x, y and z: R(x, y) and R(y, z) ~ R(x, z): As-
sume R(x, y), R(y, z); hence according to (G6): R(x, z).
GT2: If time is linearly ordered, then the set of all (at some time)
temporal counterparts of any given object is timewise line-
arly ordered.
Proof
Let r be any object. Consider the predicate 'Beforer(y, z)' which is
defined as follows:
DEF2: Beforer (y, z) =def R(y, r) and R(z, r) and the time at which y
is a temporal counterpart of itself is before the time at
which z is a temporal counterpart of itself.
Assume that time is linearly ordered, that is, that the before-relation
between times is transitive, irreflexive and linear over times (the latter
means that for all times t and t': t before t', or (' before t, or t = t'). It
immediately follows on the basis of DEFI, DEF2 and (GI): (I) For
every object x: not Beforer (x, x). (2) For every object x, y and z: Be-
fore r (x, y) and Beforer (y, z) :::> Beforer (x, z). What remains to be
proved is: (3) For every object y and z such that R(y, r) and R(z, r): Be-
fore r (y, z) or Beforer (z, y) or y = z.
Assume then: R(y, r) and R(z, r). Assume moreover: not Beforer(Y,
z), not Beforer (z, y). Hence by DEF2, (GI) and the linearity of time:
the time at which y is a temporal counterpart of itself is the time at
which z is a temporal counterpart of itself. By DEFI and the assump-
tions: there is a time t' at which y is a temporal counterpart of r, and
there is a time til at which z is a temporal counterpart of r. Hence by
(G4): y is at t' a temporal counterpart of itself, and z is at til a temporal
counterpart of itself. Hence by (G I) and what has already been de-
duced: t' = til. Hence y is at t' a temporal counterpart of r, and z is at t' a
temporal counterpart of r. Therefore according to (G2): y = z (which
was to be shown).
One may add another principle:
(G7) For all times t and t' and objects x: if there is at t a temporal
counterpart of x and at t' also a temporal counterpart of x,
then there are at all times between t and t' temporal counter-
parts ofx. (Density o/temporal counterparthood.)
(G7) has the consequence that for any object r the equivalence set {x:
R(x, r)} ordered by Beforer (y, z) is isomorphic to a certain time inter-
val ordered by the before-relation between times (restricted to that in-
terval). This is esthetically pleasing, but one may well wonder whether
it is not too restrictive to postulate (G7V
4. SUPERETERNALISM AND CHANGE FOR HIGHER

ORDER OBJECTS
It goes without saying that the superetemalist's world of, properly

speaking, unchanging objects that are related by a temporal counter-
part relation is most strange. How do I, for example, fit into it, or even
ordinary material things? And, really, could the sense described above
in which there is change in the superetemalist's world be even called
'ersatz change'? Hardly. Even the usual superetemalist would admit
that much - and would proceed to making his world somewhat less
strange: to finding a fairly acceptable ersatz for the normal change of
normal objects by considering higher order objects, namely certain
sets of his basic objects. What are considered to be 'normal objects' - I
shall call them 'Aristotelian objects', since they correspond to Aris-
totle'sfirst substances - can be absorbed by these higher order objects
in the following manner.
What I have hitherto simply called 'objects,' I shall from now on
call 'O-objects' (think of the word "object" as replaced by the word '0-
object' in the above principles and definitions of temporal counterpart
theory); certain sets of O-objects I will call' I-objects'. In the supere-
temalist's eyes, I, Uwe Meixner, and other Aristotelian objects are of
course not O-objects, but rather i-objects: each Aristotelian object, ac-
cording to the superetemalist, is a set of precisely the O-objects related
by the (simple, two-place) temporal counterpart relation to a certain 0-
object; it is, in other words (according to GTl), an equivalence set of
that relation: a i-object. The O-objects in the sets that are I-objects are
precisely the (momentary) temporal stages of the I-objects.
Since I-objects are set-theoretic constructions out of O-objects,
one will define the I-objects' having of properties in terms of the
having of properties ofO-objects. As follows:
DEF3: Let x be a I-object, fa property of objects, t a time: x has f
at t =def. there is an element y of x [O-object y in x] such that
y is at t a counterpart of y [y is located at t], and y has fat t
[and this means, since we are talking about a O-object, at
all times].
Change for i-objects can then be described as follows:

(HI) Let x be a I-object that does not now have the property f-
some property that is appropriate for change. But x will
change from not having f to having f, since there is some
later time t at which x does have f.
(H2) Let x be a I-object that now has the property f - some prop-
erty that is appropriate for change. But it will change from

having f to not having f, since there is some later time t at
which x does not have the property f
What is remarkable about (HI) and (H2) is that these change-
descriptions sound just like normal change-descriptions. Moreover, on
the basis that I-objects are equivalence-sets of simpliciter temporally
counterpart-related O-objects and on the basis ofDEF3, (HI) and (H2)
are acceptable to the superetemalist, since I-objects and their having
of properties are mere logical constructions out of O-objects and their
having of properties. Therefore, the admission of a I-object that
changes in the sense of (HI) or (H2), basically, merely amounts to
admitting that some different temporal counterparts of some O-object
have differing properties appropriate for change - which, of course, is
true for a superetemalist.
Thus, the superetemalist can, in large measure, reconstruct our
normal conception of change by simply identifying Aristotelian objects
with certain I-objects. This has the consequence that DEF3 becomes
applicable to Aristotelian objects as well, and the principles (HI) and
(H2) for normal objects can then be immediately derived from (HI)
and (H2) for I-objects. But (HI) and (H2) for normal, Aristotelian
objects simply state our normal conception of future-directed change.
The crucial question is of course: why should Aristotelian objects
- as U.M., this table, this chair, this building - be identified with 1-
objects? The superetemalist has an answer to this question: Aristote-
lian objects correspond one-to-one to the sets of their momentary tem-
poral stages. 10 "A one-to-one correspondence is an opportunity for re-
duction," as David Lewis so aptly remarks in a different context. l1
Hence Aristotelian objects, says the superetemalist, can be, first of all,
identified with the sets of their momentary temporal stages. And in
addition, says the superetemalist, if we assume the following plausible
principles:
(G8) If m is a momentary temporal stage of an Aristotelian object,
then m is a O-object. (Category of temporal stages.)
(G9) If m is a momentary temporal stage of an Aristotelian object
k and z is simpliciter a temporal counterpart of m, then z is
also a momentary temporal stage of k. (Propagation of tem-

poral stagehood by the temporal counterpart relation.)
(G 10) If m' and m are momentary temporal stages of an Aristote-

lian object k, then m' is simpliciter a temporal counterpart
of m. (Joining of temporal stages by the temporal counter-
part relation.)
If we assume these principles, then, says the superetemalist, we have

the theorem
GT3: If M is the set of the momentary temporal stages of an Aris-

totelian object, then there is a O-object y such that: Vz (z is
simpliciter a temporal counterpart of y iff ZEM).
Proof
Let M be the set of the momentary temporal stages of an Aristote-
lian object, say, k. Consider some momentary temporal stage m of k
(there must be such stage, otherwise k would not be an Aristotelian
object). Since m is a momentary temporal stage of an Aristotelian ob-
ject, we have by (G8): m is a O-object.
(1) Assume Z is simpliciter a temporal counterpart of m; hence by
(G9): z is a momentary temporal stage of k, hence zEM.
(2) Assume zEM, hence Z is a momentary temporal stage of k; m is
also a momentary temporal stage of k; hence by (G I 0): Z is simpliciter
a temporal counterpart of m.
But GT3, the superetemalist continues, has the corollary
GT4: If M is set of the momentary temporal stages of an Aristote-

lian object, then M is a I-object.
Proof
Assume M is the set of the momentary temporal stages of an Ar-
istotelian object. Hence by GT3: there is a O-objecty such that: Vz (z is
simpliciter a temporal counterpart of y iff ZEM). Hence there is a 0-
objecty such that {z: z is simpliciter a temporal counterpart ofy} = M.
Hence M is a I-object (according to the definition of I-objects).
Clearly, it is an immediate consequence of GT4 that Aristotelian
objects are I-objects if Aristotelian objects are each identified with the
set of their respective momentary temporal stages (and the supereter-
nalist has already argued that they can be identified with their respec-
tive sets of momentary temporal stages).
5. RESISTING REDUCTION?
Many philosophers would find the absorption of Aristotelian objects

by I-objects attractive. For the superetemalist, of course, this absorp-
tion is a highly desirable achievement that enables him to preserve at
least the letter of normal change descriptions. But, considered apart
from the superetemalist's intuitions, the absorption of Aristotelian ob-
jects by I-objects is far from being a natural option. What, then, makes
it so attractive?
Presumably, it is the lure of reductionism. Here you have a one-to-
one correspondence between Aristotelian objects and certain equiva-
lence sets of O-objects related to each other by the (simple) temporal
counterpart relation. Hence one has the opportunity of identifying Ar-
istotelian objects with those sets, and in doing so, one would dispense
with one extra ontological category and the correlative extra existence
assumption. Is there, in the case at hand, some good reason to resist the
lure of reductionism, so persuasively sanctioned by Ockham 's razor,
which is widely held to be an entirely unproblematic methodological
principle?
Of course, it is prima facie implausible that Aristotelian objects
are sets. I am not a set, I believe. But common sense does not count
much with scientifically-oriented philosophers.
It is possible to attack the basis of the reduction, and some phi-
losophers who are non-reductionists with respect to Aristotelian ob-
jects, for example Peter Geach,12 have, essentially, pursued the fol-
lowing strategy: what strange things, after all, are O-objects, on which
the I-objects are set-theoretically based - the sets with which the Ar-
istotelian objects are to be identified according to the superetemalists.
Apparently, one can give examples of O-objects standing in the tempo-
ral counterpart relation at a time only by referring at the same time to
Aristotelian objects, since one can name a O-object only by collaterally
referring to an Aristotelian object of which the named O-object is a
momentary temporal stage: 'Bill-Clinton-at-t is at t a temporal coun-

terpart of Bill-Clinton-at-t', just as Bill-Clinton-at-t' is at t' a temporal
counterpart of Bill-Clinton-at-t' - it seems one cannot state the exact
objective content of this sentence (the objective state of affairs in-
tended by it) without referring to Bill Clinton, an Aristotelian object.
Moreover, it seems that only because we have a fairly good independ-
ent idea of what Aristotelian objects are and, derivatively, what their
momentary temporal stages are, that we are able to give, via the above
principles (G8) - (GIO), some substantial content to the temporal
counterpart relation (setting (G8)-(GlO) aside, that relation has merely
been characterizedformally).
Yet, one may well ask how seriously the mentioned conceptual
facts, if they are such, are to be taken ontologically. They may well
have more to do with everyday epistemology than with scientific on-
tology. Superetemalists will defend themselves along this line and ar-
gue in the following manner: Nobody doubts that O-objects can pres-
ently only be referred to via Aristotelian objects; that is just the way
our language currently is. Our dependence on Aristotelian objects need
not reveal a deep ontological fact, but in fact merely reflects the
working ontology that human beings have found useful in communi-
cating with each other about their everyday normal environment in the
course of thousands of years. What they have found useful, the
"surface ontology," need not be the ultimate truth of the matter, the
"deep ontology". Indeed, it isn't the ultimate truth of the matter, as
modem science has taught us, and language, specifically its machinery
of reference, will change accordingly, in due time. Thus the supereter-
nalists.
I will not discuss here whether modem science has indeed taught
us the ontologically secondary nature of Aristotelian objects. That is a
claim that is hardly less controversial than the claim that modem sci-
ence has revealed, or rather will reveal, the obsoleteness of so-called
"folk-psychology". I will not offer arguments against asserting the
ontologically secondary nature of Aristotelian objects, and for up-
holding their primacy, but rather concentrate on a quite non-Geachean
(and non-Strawsonian)13 attack against superetemalistic reductionism.
For the sake of the argument, then, let O-objects be accepted, and
their standing in the temporal counterpart relation to each other, and let
this relation be governed by the ten principles given above (with the
possible exception of (G7». Let I-objects be accepted - these set-
theoretical constructions out of O-objects - and let Aristotelian objects
correspond one-to-one to certain I-objects. In a word, let us accept the
whole basis for reducing Aristotelian objects to I-objects. Are there
reasons, then, why we should resist reducing them to I-objects never-
theless, and reasons which are independent of considerations of the
possible ontological primacy of Aristotelian objects, an ontological
primacy which so many philosophers nowadays are unwilling to be-
lieve in?
If such reasons are not to be sought in considerations of the onto-
logical primacy of Aristotelian objects, then such reasons can only be
found in considering some important function Aristotelian objects can
perform, but not their reductive counterparts, the corresponding 1-
objects. At this point, it is again helpful to look at the modal analogy.
6. MODAL CONTINUANTS
The basic idea behind superessentialism is often that a (any) possible

object is, basically, nothing more than the sum (or set) of the proper-
ties it has at its possible world. 14 The basic idea behind supereternalism
is often that a real object is, basically, nothing more than the sum (or
set) of the properties it has at its moment of time in the real world. IS
The crucial principles
(A2) If an object had properties different from those it has in fact,
it would be numerically different from what, numerically, it
is in fact (or in other words: differentfrom itself)
(D2) If an object has at a time properties different from those it

has now, it is at that time numerically different from what,
numerically, it is now (or in other words: different from it-
self)
can be seen to follow from these basic ideas, if 'object' is interpreted

as meaning the same as 'possible object' in (A2), and as meaning the
same as 'real object' in (D2):
Concerning (A2): Assume x is a possible object; hence (according

to the usual basic idea of superessentialism) x is, basically, nothing
more than w(x#x and can be identified with w(x#x, which is the set of
properties x has at its possible world w(x). Since x is nothing more than
w(x#x, but we nevertheless want to talk about the properties x has
(properly speaking and not vicariously) at other worlds, the only
reasonable choice is to proceed on the basis of the following principle:
For all worlds w': w,Mx = w(x#x. Hence: for all worlds w': w,Mx =
realityMx. Hence: for all worlds w': if w,Mx =F realityMx , then x =F x.
Concerning (D2): assume x is a real object; hence (according to
the usual basic idea of supereternalism) x is, basically, nothing more
than t(x),realityMx and can be identified with t(x),realityMx, which is the set of
properties x has in the real world at its moment of time t(x). Since x is
nothing more than t(x), realityMx, but we nevertheless want to talk about
the properties x has (properly speaking and not vicariously) in the real
world at times other than t(x), the only reasonable choice is to proceed
on the basis of the following principle: For all moments of time t': 1',
realityMx = t(x), realityMx . Hence: for all times t': 1', realityMx = now, realityMx.
Hence: For all times t': ifl', realityMx =F now, realityMx , then x =F x.
Conversely, if we hold that some possible object is something over
and above the properties it has at possible worlds, and that some real
object is something over and above the properties it has (in the real
world) at moments of time, then we are denying not only the men-
tioned basic ideas behind (A2) and (D2), but also, quite clearly, these
principles themselves, and therefore superessentialism and supereter-
nalism. For if some possible object is something over and above the
properties it has at possible worlds, then we should conclude that for at
least one possible object x we have w·Mx =F wMx for at least some worlds
wand w'; and if some real object is something over and above the
properties it has at moments of time, then we should conclude that for
at least one real object x we have t',realityMx =F t, realityMx for at least some
times t and t'.
But why should we, in the first place, hold that some possible ob-
ject is something over and above the properties it has at possible
worlds? Or in other words, why should we think that some possible
object is a modal continuant? We should believe this for the following
reason: because there are relational properties of some possible objects
that are not reducible to properties those objects have within or at pos-
sible worlds.
Consider the relation of realizing part of a possible world. This
relation can be either taken causally or cognitively; if it is taken caus-
ally, it means as much as making real or making actual; if it is taken
cognitively, it means as much as cognizing as real or cognizing as ac-
tual. Of course, one can say that a human person x, for example, real-
izes at possible world w part y of w; but the point is that one can say
this only because, and in the conceptual order after, one can simply
say that x realizes part y of w; the latter form of expression is the pri-
mary form, the former merely secondary and, as it were, epiphenome-
nal. Suppose, then, that person x realizes part y of world w; from this,
there follows not merely the triviality that y is actual at w, but that y is
actual simpliciter. Quite clearly, the following relational property of
person x, the property of realizing part y of w, is irreducible (1) to the
properties x has (properly speaking) at possible worlds other than w.
But it is also irreducible (2) to the properties x has (properly speaking)
at w, the reason being that what properties x has at w does not deter-
mine, not even partially, which of those properties are simpliciter ac-
tual or simpliciter real properties of x, while x's partially realizing w
does indeed partially determine this,16 no matter even whether we un-
derstand realizing in the sense of causally realizing or in the sense of
cognitively realizing.
The force of this argument is that possible objects that are true
(namely, literally reality-making) agents and true (namely, literally
reality-cognizing) cognizers have to be modal continuants. Not all ob-
jects, of course, can be plausibly held to be true agents and cognizers,
but persons, certainly, are traditionally conceived to be such. They do
their deeds and they gain factual knowledge, and thereby contribute to
determining what is simpliciter real. Therefore, what is in this world
or in another cannot completely determine what they are, since it does
not even contribute to determining what is simpliciter real. Persons
are, to put it in a word, world-transcendent. 17 And this means, in par-
ticular, that persons are over and above the properties they have at
possible worlds: they are modal continuants.
Hence, unless we want to give up our traditional way of under-
standing ourselves, a way that does not allow that doers and knowers
be absorbed entirely by the field that their doing and knowing is di-
rected at, we have good reason to believe that there are modal contin-
uants, namely ourselves. I am, of course, aware of the fact that the
general philosophical climate in recent years is inimical to the tradi-
tional conception; what I would deny is that endorsing this climate, the
climate of naturalism, is the only rational option.
7. TEMPORAL CONTINUANTS
There remains the question of temporal continuants. Why should we

hold that some real object is a temporal continuant, that is, something
which is over and above the properties it has in the real world at mo-
ments of time? If we have good reason for believing this, then, besides
having a weighty argument against superetemalism, we also have good
reason for believing that some real objects are irreducible Aristotelian
objects, objects that, even though they correspond one-to-one to the
set-theoretically constructed I-objects, are not reducible to them, and
therefore not reducible to the momentary O-objects which the I-objects
are constituted by. IS And if there are irreducible Aristotelian objects,
then there is also genuine change, and not merely ersatz change; and
given a cogent argument for the existence of temporal continuants or
irreducible Aristotelian objects, we would also gain an understanding
of what, in the first place, genuine change truly consists in, and of
what makes it truly different from ersatz change.
This is the argument. It is modeled on the argument for the exis-
tence of modal continuants: We (certain real objects) are not reducible
to the properties we have in the real world at moments of time (and
therefore not reducible to I-objects: certain sets of temporally coun-
terpart-related momentary O-objects that in their turn can, indeed, each
be identified with the set of properties they have in the real world at
their respective moment of time) because we have certain relational
properties that are not reducible to the properties we have in the real
world at moments of time, whether those moments be taken individu-
ally or collectively. Consider the relation of being aware of a moment
of time as present. Suppose person x, anyone of us, stands in that rela-
tion to a certain moment of time, is aware of moment t as present. The

relational property x has, the property of being aware of time t as pres-
ent, is quite obviously irreducible to the properties x has (properly
speaking) in the real world at other moments of time than t. But it is
also irreducible to the properties x has (properly speaking) in the real
world at t, for these properties have nothing whatever to do with t be-
ing present or not: x has those properties in the real world at t, no mat-
ter whether t is present or not; it will still have these very same prop-
erties in the real world at t after t has been present, and it already had
these very same properties in the real world at t before t became pres-
ent. Just as being present is not a temporally located real-world-
property of moment t (in contrast to the property being present at t),
but rather what may be called a fleeting property of it, 19 so being aware
of moment t as present is not a temporally located real-world-property
of x. It is a fleeting property of x, and therefore not reducible to tempo-
rally located real-world properties of x, that is, not reducible to prop-
erties x has (properly speaking) in the real world at moments of time.
Of course, we can also say, if we want to, speaking somewhat improp-
erly, that x has in the real world at t the property of being aware of t as
present; but the point is that we can say this only because, and in the
conceptual order after, we can simply say that x has the property of
being aware of t as present.
8. A METAPHYSICAL PICTURE
Perhaps a simile or conceptual picture can serve to illustrate all these

metaphysical matters at once. Being modal and temporal continuants,
we are outside the realm of all possible worlds, including the real
world. We are looking as if into a landscape that is mainly dusky, but
where a point of light is moving in one direction along a certain path.
We know there are alternative paths out there in the dusk, but here we
have the path the point of light in fact takes. We also know that it de-
pends partially on our own choosing which path the point of light
takes. As the point of light moves on, we see, in a flash, ourselves on
the path, or rather, properly speaking, not ourselves but innerworldly
momentary static representations of us. These momentarily illuminated

representations are always different, yet always representations of us;
each belongs to precisely one of us. Considering those representations,
we say, and it is literally true, that we are changing, whereas, indeed,
our representations do not change, they always have the same inner-
worldly properties (that they are illuminated and then recede into
darkness is a process which does not involve innerworldly properties).
We know, moreover, that each of us is not the sum of all the at-some-
time-or-other-illuminated representations that belong to him or her, but
is only represented, in toto, by this sum; that this sum does not, prop-
erly speaking, change either. Considering the shadowy representations
of ourselves of which we become more or less dimly aware beside the
path that the point of light is taking, we say, and it is literally true, that
we might and even could have been (qualitatively) different. Our rep-
resentations, however, could not have been different; being these rep-
resentations, they could not have had other innerworldly properties
than they have. We know, moreover, that each of us is not even the
sum of all the representations of himself or herself that could, in prin-
ciple, have been illuminated by the point of light; this sum is merely
the total space of one's innerworldly possibilities; and it could not
have been different either (nor could one have had a different one). As
modal and temporal continuants, we are beyond all that and its like:
beyond the innerworldly realm - and yet, undoubtedly, deeply in-
volved with it. 20
Notes
1 There also is a deeper reason, and certainly one that is more credibly a reason
oJLeibniz or Lewis, for their position. See section 6 below.
2 To my mind, Lewis' argument against literal trans-world identity, and mutatis
mutandis against literal trans-time identity, literal identity over time, (see Lewis
[1987], p.199ff, p.202ff, p.2l0) are far from convincing. The arguments have the
following two contestable presuppositions: (1) If literally the same object x existed
wholly at two worlds, respectively: two times, then, unacceptably, it would be a part
of both worlds, respectively times (or in other words: then both worlds, and both
times, would overlap in x). (2) If literally the same objecl x had, as a whole, an in-
trinsic property Jat a world w [time I], and not at a world w' [time 1'], thenJwould,
unacceptably, not be an intrinsic property of x at w [at t] after all, but a relation x

bears to w [to t]. Here, I submit, the most promising denials of (1) and (2): concern-
ing (1), literally the same object may exist wholly at two worlds [times] without be-
ing a part of either of them via its respective representative at each of the two worlds
[times]. Concerning (2), literally the same object x may as a whole have an intrinsic
property, without "relationalizing" it, at one world [time], and not at another, since
its representative at the one world [time] has it, while at the other world [time] there
is no x-representative that has it. (Here 'z being the representative of object x at
world w [at time t]' expresses an intrinsic and essential relation between z, x and w
[t]. For more, see sections 6 to 8 of this paper.)
3 Gerhardt [1978], II, p.47ff.
4 The existence of contingency is only demonstrated by experience to the extent
that it demonstrates the existence of change, and this means that the existence of
contingency is demonstrated by experience, if at all, only for some concepts of con-
tingency, certainly not for all. For example, if an object x wiII change from having
property f now to not having property f later, this means that its having of property f
is not necessary in each sense of that word that implies that what is necessary is al-
ways the case at any time in the future.
5 Another past supereternalist, very different from Parmenides, was Hume. See
footnote 15.
6 Note that newly coming to be should be distinguished from coming to be sim-
pliciter. (Fl) fits the latter.
7 Note that definitely passing away should be distinguished from passing away
simpliciter. (F2) fits the latter.

8 Note that (G3) follows from (G4) and (G2). Note also that (G2*): every object
x is at any time t temporal counterpart of at most one object, cannot be maintained;
for this according to (G4) would imply the undesirable result that every object is at
any time at most a temporal counterpart of itself. (G2*) is also untenable in view of
(05) and (G6) below.
9 (G7) would exclude higher order objects (see below) that "reappear" after a
stretch of time in which they "were not there at all."

10 Note that talking about the temporal stages of an Aristotelian object does not
per se mean that those stages are temporal parts of the Aristotelian object.
II Lewis [1986], p.245.
12 Cf. Geach [1972]. In Meixner [1997], p.203, I am critical of the force of

Geach's criticism.
13 Cf. Strawson [1959], Chapter 1.
14 As is well known, this idea is strongly suggested by Leibniz's philosophy.
15 This idea is clearly visible in the following passage from an outstanding su-
pereternalist: "[O]ur ideas of bodies are nothing but collections formed by the mind
of the ideas of the several distinct sensible qualities, of which objects are composed,
and which we find to have a constant union with each other. But however these
qualities may in themselves be entirely distinct, it is certain we commonly regard the
compound, which they form, as one thing, and as continuing the same under very
considerable alterations. [... ] The smooth and uninterrupted progress of the thought,
being alike in both cases [that of a "succession of related qualities," and that of "one
continued object, existing without any variation"], readily deceives the mind, and
makes us ascribe an identity to the changeable succession of connected qualities."
(David Hume, A Treatise of Human Nature, Book I, Part III, Section III, p.270 of
MacNabb (ed.) [1987].) In the quoted passage, we also find, very apparent, the
blending of numerical with qualitative identity (over time), which is a consequence
of the mentioned idea and which makes Hume implicitly subscribe to the following
superetemalistic principle: if object x has at moment t other qualities than object y
has at moment t', then x*- y. (Note that if 't" is replaced by 't', we have a principle
that is entirely uncontroversial.)
16 To make this more palpable, consider a typical way x qua human person is
involved with world w. For example: x lifts his hand in w. Now, x's partially (caus-
ally) realizing w may well tum this into: x (simpliciter really) lifts his hand (and
thus: lifting his hand is a simpliciter actual property of x). If it does so, then x is
causally responsible (in an absolute sense) for the (simpliciter real) lifting of his
hand; if it does not, x is not causally responsible for it. (The talk of responsibility
here must not blind one to the fact that the crucial point in the argument is not con-
tingency, or freedom of the will, but the imparting of reality or actuality in an abso-
lute sense: world-bound individuals are not up to this.)
17 The world-transcendence of persons and other substances is argued for in
much greater detail in Meixner [1997a].
18 It would be implausible to hold that all Aristotelian objects are irreducible. In
fact, there is no reason to hold that inanimate Aristotelian objects are irreducible.
19 Note the contrast in content between the sentences 'tl is present at II' and 'II
is present' (the first sentence is true at all time, the second only at td. It is also im-
portant for the argument to realize that the latter sentence cannot be synonymous
with the sentence 'II is the time (or belongs to the time) of Ihis utterance'. The rea-
sons for this are: (1) The moment of time th having no extension, cannot be the time
of an utterance (and if it belongs to the time of an ongoing utterance of 'II belongs to
the time of this utterance' or of' II is present', it may nevertheless not be presenl, but,
instead, a later or earlier moment of time). (2) There might not be any utterance at
(incorporating) II at all, while II is nevertheless presenl.
20 In discussion, Kit Fine offered the following illuminating analogy: the build-
ers of a house (realizers of a world) cannot be parts of it. I would merely add: nor can
they be parts of the totality of all possible houses that might be built.
STARTING OVER
Christopher Hughes
1.
A number of philosophers have thought that necessarily, once a thing

has gone out of existence, it is gone for good.! I am unconvinced. In
what follows, I shall consider a number of hypothetical cases in which
a thing arguably comes back into existence. I shall not claim that any
of the cases provides a decisive refutation of the no-two-beginnings
principle; here, as elsewhere in metaphysics, decisive refutations are
hard to come by. But I shall claim that they cast serious doubt on the
principle, and make it incumbent upon its defender to provide an ar-
gument for it. I shall look at one sort of argument, and raise some
doubts its cogency. I shall conclude with some remarks about the de-
fensibility of a weakened version of the no-two-beginnings principle.
2.
It has sometimes been thought that necessarily, when organisms die,

they cease to exist, although they typically leave behind a dead body
or corpse that was a living body before the organism died. It has also
sometimes been thought that it is (at least metaphysically) possible to
bring a body that was once living but is now dead back to life. Sup-
pose both these claims are true. And suppose that, say, a frog dies, and
451
its dead body is (almost immediately) subsequently reanimated. Then

there was a frog at time ti> a frog's corpse at t2, and a frog at t3. Is the
frog that existed at t 1 a different frog from the frog that existed at 13? If
whatever goes is gone for good, yes. But how could frogs have the
same body without being the same frog?
The defender of the no two beginnings principle «(NTB), for short)
could respond to this line of thought in at least two ways. She might
deny that esse viventibus est vivere, and hold that an animal goes on
existing after it dies, as a dead animal, just as a body goes on existing
after it dies, as a dead body.2 If an animal goes on existing after its
death, then the champion of (NTB) can accept that in the re-animation
case we have just one frog and just one body; we have a frog who, in
the uninterrupted course of its existence, goes from life to death and
back again, in much the way it might go from waking to sleeping and
back again. All this would be unproblematic if animals existed until
their corpses crumbled into dust (or the like); but I don't think they do.
Since I have argued for this elsewhere, I shall not rehearse the argu-
ments here. 3
Alternatively, the defender of (NTB) might agree that when the
frog dies, it ceases to exist, and insist that the frog existing at the end
of the story is a different frog from the frog existing at the beginning
of the story. She might concede that the frog body at the beginning of
the story is the same as the frog body at the end of the story, and
maintain that one and the same frog body can be "rented" by different
frogs seriatim (at least as long as there is a gap between the time one
frog's lease expires, and the next frog's lease begins). Or she might
deny that there is anything that goes on existing in an ex-animate state
when a frog dies: although a frog that dies leaves behind a dead body
or corpse, that corpse is not identical to the living body the frog had
ante mortem. 4
In order to explain why I do not find (either version of) the "two-
frogs" view intuitively appealing, I shall start by discussing Eric Ol-
son's views on brainstem replacement.
It might seem that someone could survive having her brainstem
destroyed, and immediately replaced by a perfect duplicate of it. Why
should this be any more problematic than a person's surviving the de-
Christopher Hughes, Starting Over 453
struction of one of her kidneys, and a subsequent kidney transplant?

Well, Olson says:
Despite appearances, it does not seem to be the case that your biological
life continues without interruption when your brainstem is destroyed and
replaced. As soon as your brainstem is destroyed, you lose the capacity to
direct your vital functions. Your individual cells and organs can no longer
work together as a unit in the manner characteristic of a living organism.
What we have is a corpse that merely appears to be alive because it is so
freshly dead. This period of "metabolic anarchy" may seem insignificant
because it is so brief. (How could just a few seconds or minutes matter?)
But suppose Descartes' evil genius annihilates you and replaces you with
a perfect duplicate a thousandth of a second later. Here too the interrup-
tion of your biological life is brief - so brief, in fact, that no one would
suspect that anything out of the ordinary had happened. Nevertheless,
what happens during that thousandth of a second is enough to bring your
existence to an end. I suggest that you cannot survive brainstem replace-
ment for the same reason you cannot survive annihilation and replacement
by a perfect duplicate. 5 (Olson [1997], pp.140-l41)
Olson holds that (i) I am essentially this animal; and (ii) this animal
essentially has this brainstem. (Compare this to the more familiar
Cartesian view that I am essentially this thinker, and this thinker es-
sentially has this mind). Given (i) and (ii), when my brainstem is de-
stroyed, I am destroyed with it; when a new (duplicate) brainstem is
put into my corpse, the resulting person cannot be me, but only my
duplicate. Doubts could be raised about both (i) and (ii), but I shall not
pursue them here. 6 Instead, I shall modify Olson's story slightly.
Suppose a surgeon is carrying out more or less the procedure de-
scribed by Olson. First, she removes my original brain stem. Next, she
gets the duplicate brainstem out of storage, and prepares to hook it up
to the rest of my organs. Unfortunately, in attempting to get the dupli-
cate brainstem from its storage compartment to my body, she drops it,
damaging it severely. Duplicate brain stems are expensive, and the
surgeon does not have a "backup" one. Luckily, she still has my origi-
nal brain stem, which has not yet become damaged. She immediately
puts it back into me, and normal metabolic service is resumed. Be-
cause my body-except-for-my-brainstem was a corpse for such a short
period of time, there is no irreversible damage to any part of my brain

or any of my other organs.
Is the person existing at the end of the episode just described - the
person with my brain and all my other organs - me? It is hard to see
why not. Think of it this way: suppose that at the (futuristic) clinic
where brainstem transplants are carried out, there is also a machine
called a "replicator", which produces a perfect (physical and psycho-
logical) duplicate of a person. The replicator is like a photocopier, in
that it leaves the original person intact, though (briefly) unconscious.
You bring your daughter into the clinic to have her replicated. The
procedure goes without a hitch, and your daughter is put back in bed in
a private room where (you are told) she will regain consciousness in an
hour or less. Unfortunately, a new attendant mistakes your daughter
for a different patient who is scheduled to have a brain stem transplant.
The attendant wheels your unconscious daughter into surgery. The
surgeon (unaware of the attendant's mistake) removes your daughter's
brainstem, and then gets the replacement brainstem out of storage.
When she sees that the other brainstem is the wrong size and shape to
fit into the patient's head, she realizes there has been a terrible mis-
take.
Suppose that at this point the surgeon gets hold of you, and offers
you the following choice. You can authorize the surgeon to immedi-
ately put the original (not yet damaged) brainstem back in your
daughter; the procedure is routine and risk-free (or, at any rate, as risk-
free as is compatible with the ineptitude of the clinic's staff). Alterna-
tively, you can not authorize any further surgical procedures, let the
staff of the clinic destroy your daughter's (ex-animate) body and the
brainstem removed from it, and take home the person the replicator
produced. Wouldn't you choose the first option? For me, the choice is
a no-brainer. If the surgeon puts the brainstem back in your daughter
(and the operation is successful), you will go home with your daugh-
ter; if you do not authorize any further surgery, you will go home with
a replica of her. If this is right, then in my original variant of Olson's
story, the person existing at the end of the botched transplant operation
(with my brain stem, and the rest of my body) is me, and not just a du-
plicate of me.
Is this consistent with (NTB)? I imagine Olson would say it is. Ol-
son (somewhat tentatively) suggests that if an animal has its head sev-
ered, the animal would not go on existing as a headless body, but
would go on existing (for a short time, at least) as a severed head. As
he sees it, this is because the severed head contains the parts of the
brain that direct and coordinate vital functions, and does not, upon
being severed, immediately lose its capacity to direct and coordinate
those functions. 7 I gather from Olson that the parts of the brain that co-
ordinate vital functions are in the brain stem. So, I take it, Olson would
think that if my brain stem is removed from the rest of my body, I go
on existing as a brain stem, for at least as long as the brain stem retains
its capacity to direct and coordinate vital functions. Thus Olson could
say the identity of the person at the end of the botched operation with
the person at the beginning of that operation is compatible with (NTB).
Before the operation, I was a normal human animal with a whole
brain, head, arms, legs, etc. When my brain stem is separated from the
rest of my body, I become a mutilated animal, all of whose parts are
parts of my brain stem. When the brain stem is reattached to what it
was separated from, I become a normal human animal again, with a
normal set of organs.
I have various reservations about this. To start with, supposing
that an animal could survive being "pared down to" the dimensions
(and constitution) of a brainstem, seems rather like supposing that the
central heating system in our house could survive being "pared down
to" the dimensions (and constitution) of a thermostat, or supposing that
a car could be pared down to the dimensions (and constitution) of a
motor. If there is nothing more to a thing than a brainstem, it is hard to
believe that thing could be a (complete, albeit mutilated) animal, rather
than just an (ex) part of an animal. s If (as Olson supposes) I am essen-
tially a (human) animal, and nothing with the constitution of a brain-
stem is a (human) animal, then I cannot survive the separation of my
brainstem from the other parts of me as a brainstem.
But suppose we grant that, when my brainstem is separated from
the rest of me, I am nothing over above my separated brainstem9 • It
seems that we can still tell the story in such a way that I am no longer
living at some time between separation and reattachment. Suppose that
after the brainstem is removed, it momentarily ceases to function, and
becomes (for however briefly) an uncontrolled system at the mercy of

entropy, in the initial stages of decomposition. Medical technicians are
monitoring the state of the brainstem, and immediately "re-start" the
brainstem, quickly enough that it is not significantly damaged. Then
the surgeon reattaches it to the rest of my body. On the assumption that
my life was going on for as long as my brainstem kept functioning,
when the brainstem stopped functioning (even for a fraction of a sec-
ond) my life apparently stopped. If (as Olson supposes) I cannot go on
unless my life does, then I no longer existed, at the moment my brain-
stem stopped functioning. Still, it seems that I do exist, after my brain-
stem is re-started and re-attached to the rest of my organs. Again, sup-
pose your daughter's brainstem has been removed by mistake, and her
(separated) brain stem has been non-functioning for a fraction of a
second. Would you really tell the surgical team that there is no point
re-starting and reattaching the brain stem, because your daughter is
gone for good?
Here someone might say, taking a cue from Peter van Inwagen lO :
Even if my brain stem momentarily stops functioning (in a distinctively
organic, homeodynamic way), so that it (and I) are in entropic free fall,
my life goes on for a bit. That life no longer consists in the sort of large-
scale physical processes it used to consist in (and human lives typically
consist in): it has been, in van Inwagen's words, "squeezed into" various
small-scale physical processes - into the subchemical changes that under-
lie the large-scale physical processes that normally constitute a life. Still,
as long as enough organic molecules retain their integrity, and remain
bonded to each other the right way (because entropy has not yet had
enough time to break down the bonds between the molecules, or the
molecules themselves) my life goes on. When my brainstem is reattached
to the rest of my body, my life regains its normal "size and shape." (see
van Inwagen [1990], pp.146-7)
If this is right, then - even in the version of the story in which my

separated brainstem temporarily ceases to function organically - one
can say that I went on living (and existing) uninterruptedly, from the
time my brainstern was removed to the time it was reattached.
Fair enough: but why suppose that my existing after the reattach-
ment of my brain stem depends on my life going on "at the subchemi-
cal level" at all times between its removal and its reattachment? Why
suppose any that any processes at all need to go on throughout that in-
terval? Suppose that, the time between the removal of my brainstem
and its reattachment is one year and a few minutes. During the period
of a year, my quadrant of the universe undergoes what Sydney Shoe-
maker calls a "local freeze", in which all processes (even subchemical
and subatomic ones) stop. Local freezes are, I take it, physically im-
possible. But my concern is with (NTB), which says that nothing that
exists or might have existed could come back from non-existence. If
local freezes are (metaphysically) possible, and if it is necessary that
whatever ceases to live, ceases to be, then (NTB) is false.
Alternatively, suppose that medical technicians could "take apart"
my brainstem, breaking it down to individual cells. And suppose that
before the individual cells died, the technicians could put all the cells
back together in exactly the way they were put together before the
brainstem was taken apart. I find it very doubtful that my life would go
on during the period that each of the cells that had composed my
brainstem were living their own separate lives. (It has often been sug-
gested that my life was not going on before the primitive streak stage
of my fetal development, precisely because the cells in the clump of
cells whence I came were not then cooperating in a single life, but
living their own separate lives). Still, if the technicians could separate
my brain stem from the rest of me, take it apart, put it back together,
and reattach it to the rest of my body quickly enough, the person ex-
isting at the end of the procedure would be me. Suppose that doctors
knew that I had a small number of cancerous cells in my brain stem,
and that the only way they could discover which cells were cancerous
was to take my brainstem completely apart, and examine the cells that
had constituted it one by one, destroying the (few) cancerous cells,
"reassembling" the others, and then reattaching the reassembled brain
stem to the rest of my body (before it had succumbed in any signifi-
cant way to entropy). If the procedure were medically feasible, I would
not refuse it on the grounds that it could not save my life, because the
person existing at the end of it would be a mere duplicate of me.
If this is right, then someone who wants to endorse both the Aris-
totelian principle that nothing could go on longer than its life and
(NTB) is committed to counterintuitive judgments about what sorts of
operations I could or could not survive. But the proponent of (NBT)

might say:
So much the worse for the Aristotelian principle. Suppose that my in situ
brain stem stopped functioning for just a fraction of a second. I would
(very briefly) cease to live, but I would not cease to exist. If, as a result of
what happened to my brainstem, the loss of my capacity to direct and co-
ordinate my vital functions were irreversible - if I could not ever again di-
rect and coordinate those vital functions - then, perhaps, I would not only
cease to live, but also cease to be. But if the non-functioning is, so to
speak, a hiccup, then I go on for just a fraction of a second without my
life, and subsequently start living again. Turning to the more recherche
cases involving the separation of my brainstem, and/or its subsequent dis-
assembly into its cellular components, here too an individual goes on ex-
isting in a non-living state, as a partially "disassembled" animal. When
my brainstem is separated from the rest of my body, it is tolerably clear
that my life does not go on where the rest of my body is. One might have
doubts about whether it goes on where my brainstem is (as Olson seems to
think) or stops going on. (Compare this to what happens when rivers
'branch' In some cases, it is clear that the course of the river continues in
this (big) branch, not in that (tiny) branch. (Assume that the tiny
"offshoot" of the river does not flow back into the river). In other cases,
where the branching is equal, we would be inclined to say that the river
does not continue in either of its branches. We can imagine intermediate
cases, in which we might wonder whether we should say the river goes on
in this branch, or stops existing here (for the same sorts of reason that an
amoeba that splits stops existing now). Be that as it may, if we do enough
disassembly, it becomes clear that life of the animal has (at least tempo-
rarily) stopped. If I have been disassembled to the point that we have got a
fresh (brainstemless) corpse and a whole bunch of brainstem cells tempo-
rarily living "solo" lives - or the point where all the cells of my body are
temporarily living solo lives - then my life has stopped. And if the sur-
geons reassemble all my cells in the right way, and re-attach them to the
fresh corpse (or to each other) in time, my life will start again. But all this
is compatible with (NTB): in the cases described, I exist uninterruptedly,
first as a normal living animal, then as a partially disassembled and no
longer living animal, and finally as a reassembled and once again living
animal.
The advantage of this response is that it allows us to hold onto (NTB)

without insisting that lives go on, when it is intuitively very doubtful
that they do, or insisting that what would appear to be me (or your
daughter) is only a doppelganger. I suspect, though, that this defense
of (NTB) will not ultimately succeed: it will make it too hard for a
living being to go out of existence. But before I address these ques-
tions, I shall consider some alleged counterexamples to (NTB) in-
volving inanimate objects.
3.
Suppose I drop a cup on a hard floor, and it breaks into many pieces. If
I painstakingly glue the pieces together (in the right way), I'll end up
with a cup. Moreover, it seems that I'll end up with the cup I dropped
and broke. Suppose that last year you lent me the cup I broke into
many pieces yesterday, and glued back together today. You knock on
my door today, and say, "I lent you a blue cup last year. Could I have
it back?" I could hand it to you, and say, truthfully, "Here it is. I'm
very sorry; I dropped it on the kitchen floor yesterday, and I glued the
pieces back together as best I could". I have returned your cup, though
not in the condition in which you loaned it to me.
Suppose, though, that instead of glueing the pieces back together;
I had simply collected them, put them in a box, and handed them to
you. Then, I'm inclined to say, although you lent me a cup, I didn't
return it; I only returned the (ex) bits of it. True, you might say: "I
loaned him a beautiful cup, and he returned it in a thousand pieces."
But, then, in a horror movie, a character might say: "Jones is in the
dustbin in a thousand pieces". What the dustbin contains is surely not
Jones, but things that used to be parts of Jones. Similarly, I lean to-
wards saying, what the box you handed me contains is not my cup, but
bits of china that once were and will again be parts of a cup.
Here is a different case, about which I have somewhat stronger
intuitions. Christ predicted the destruction of the Temple, and it came
to pass. I take it that when the Romans destroyed the Temple, it went
out of existence: it didn't survive its destruction. In this context, there
is some interest in the etymology of the Latin verb destruere, whence
the English terms 'destroy' and 'destruction' come. Destruere means
"un-pile", just as construere means "pile together"; the destruction of a
building consists in putting asunder the stones it is made of, just as the
construction of a building consists in putting together the stones it is
made of. Even in Latin, though, it had something like the wider sense
that "destroy" has in English (one can destroy something by causing it
to cease to exist in all sorts of ways - e.g. by burning it). I presume
that destruere went from meaning something like "demolish" to
meaning something like "destroy" in the current sense because de-
molishing a house (or a temple) destroys it.
When the Romans destroyed the temple, they did not do anything
as drastic as vaporizing each of the stones it was made of (a good bit
of Temple wall is still intact today). Suppose that after the Romans had
knocked down most of the Temple stones, and broken some of those
stones into fragments, the Emperor had had a vision in which God told
him that the destruction of the temple was an abomination in His sight,
and that amends had to be made. Suppose that the emperor had imme-
diately sent a team of his best builders to Jerusalem: their job would be
to find the (now partially dispersed) Temple stones, "glue" them back
together if necessary, and then put them back together in the way they
had been put together before the Romans destroyed the Temple. That
way the Temple the Romans had destroyed would be rebuilt -
"com 'era, dov'era", as Venetians would later say in a different con-
text. If all this had happened, then the building made of the re-
assembled stones would be the building that the Romans destroyed,
just as the cup made of the re-assembled fragments of china is the cup
I dropped and broke.
Examples of this sort could be multiplied. They trade on the fact
that in thinking about the persistence and identity of composita we
have two tendencies. First, we have a tendency to think of separating a
thing's parts from each other as prejudicial to its continued existence. 1I
Of course, it is not a hard and fast rule that separating a thing's parts
causes it to go out of existence: whether or not a certain sort of parts-
separation will result in a thing's non-existence will depend both on
what kind of thing it is, and on what kind of parts they are. When you
take a tent apart (in the usual way), and put the canvas on one shelf,
and the poles on another, it seems that you end up with a ("partially
scattered") disassembled tent, rather than a bunch of ex parts-of-a-tent.
(Suppose that you loaned me your tent in an assembled state, driving it
over on the back of a pickup truck. If you ask for your tent back, I
comply with your request, even if I give it back in a disassembled
state). On the other hand, if you separate the parts of a 500 Lire coin
from each other (in a throughgoing way), you end up with with some
bits of metal that used to be parts of a coin, rather than a
"disassembled coin". Suppose, though, you "take a tent apart" by cut-
ting the tent canvas into many small pieces. Then, I think, the tent does
not survive. (It is similar to a case in which I take an old T-shirt, and
cut it up into lots of small strips to use as cleaning rags. That surely is
enough to spell the demise of my shirt). Like a coin, a tent is not ame-
nable to certain sorts of parts-separation. (The only sorts of things I
can think of that are amenable to any sort of parts-separation are ag-
gregates of simples).
We also have a tendency to think that if you put the same parts
back together in the way they were put together when they constituted
a K, you get a K back. Again, I do not claim that this is hard and fast
rule (it may "go soft" in certain ship of Theseus cases). But we tend to
think that ceteris paribus if you put the original parts back together the
right way, you get the original engine back; if you put the original bits
of china back together the right way, you get the original teacup back,
and so on. Another example of this way of thinking is found in early
Christian thinking about the afterlife.
Early Christians appear to have thought of the afterlife (often - I
do not mean exclusively) in terms of a bodily resurrection, conceived
as the resuscitation of a continuously existing but temporarily ex-
animate corpse. But they were aware that things could not always
work that way: corpses were sometimes devoured by animals, or
burned, or the like. How could those whose corpses had been de-
stroyed rise again? The Church Fathers (often) answered by appealing
to God's ability to "re-assemble" the continuously existing bits that
had composed a person's living body, and subsequently his corpse,
before the destruction of the latter. Thus Irenaeus said that if God
could create the first man and his body, He would have no trouble re-
making our bodies, by putting back together the bits of the body that
had decomposed. And Athenagoras argued that since God knows
where all the particles that used to compose a person's body have
gone, and knows just how to put them back together in order to a get a
person's body back, there is no difficulty about God's resurrecting
those whose corpses have decomposed. This conception of a bodily
resurrection via the reassembly of surviving parts of a decomposed
corpse appears to have antecedents in various (Jewish and Christian)
apocalypses. 12 In both cases, we have the assumption that if God puts
together (even very small) ex parts-of-our-bodies together in the right
way, his reassembling those parts is also his reassembling our bodies.
Of course, someone might say that early Christians and the Church
fathers had an axe to grind: they needed to make room for a bodily
resurrection on the last day, and saw no way of doing so without sup-
posing (counterintuitively) that reassembly could get someone's body
back. It is interesting to note, though, that Lucretius also appears to
have thought that reassembly of a person's long dispersed atoms could
get that person (and his body) back (cf. De Rerum Natura, Book III,
847-851). Given that Lucretius wants his readers to stop worrying
about death and what might lie beyond it, it is rather awkward for Lu-
cretius to grant that (the right sort of) reassembly of atoms might hap-
pen, and, if it did happen, would bring me back.
Some philosophers - among them van Inwagen and Olson -
would be unhappy with the idea that our resurrection could be under-
stood simply in terms of God's reassembling all the dispersed bits of
me in the right way on the last day. Van Inwagen and Olson accept the
Lockean idea that if an organism exists at a certain moment, it exists
whenever and wherever - and only when and only where - its life is
occurring.13 And they think that, whether or not a suspended life can
resume, a life that has been disrupted cannot. 14 So, they would say,
unless my life goes on (possibly in a "squeezed down" state) or is non-
disruptively suspended, God could not get me back on the last day, be-
cause He could not start my life again, whatever He did in the way of
reassembly of the things that used to be parts of me. At best, He could
produce a replica of me.
I do not see this. The teacup that was shattered into ever so many
fragments is the teacup that has now been painstakingly glued back
together. Suppose a dog steps on a powerful mine. Small bits of its
body fly out at high speed in all directions. Then God works a miracle,
putting all the bits of the dog's body (and the dog) back together in just
the way they were just before he stepped on the mine. (If you were
watching the miracle, it would look like what you would see if you ran
a video-tape of an explosion backwards). Why is not the dog whose
parts were reassembled the reassembled dog, just as the cup whose
parts were reassembled is the reassembled cup? The reassemblies are
very different, but does that matter? After all, suppose that it had been
a teacup, rather than a dog, that had been exploded, and God had mi-
raculously put its little bits back together (in the right way). Why
wouldn't we have a miraculously reassembled teacup? Similarly, why
wouldn't we have a miraculously reassembled dog? A defender of van
Inwagen's view might protest that it just is intuitively plausible that,
even if the suspension of an organism's life does not mean that organ-
ism is gone for good, the disruption of its life does. Again, I do not see
this. We have a continuum of cases in the organism ceases to live a
normal life, and it gets harder and harder to get the parts of the organ-
ism that took part in that life back in the same state they were in before
the organism stopped living a normal life. At the easy end of the spec-
trum, we have van Inwagen's frozen organisms, that either go on liv-
ing a "squeezed down" subchemical life, or have their life suspended.
If we start "disassembling" the organism - removing its brainstem, say
- it gets harder to get the parts back in the state they were in just be-
fore disassembly, but it may still be feasible. If we thoroughly disas-
semble the organism - right down to individual cells - I imagine it is
infeasible to get the parts in the state they were in just before disas-
sembly (at least so long as the organism is as complicated as a full-
grown dog). But it may be feasible someday, and if you could do it,
the organism you had at the end of the process of cellular reassembly
would be the organism you had before disassembly. If the organism is
broken up into small enough (non-functional) parts, as happens in the
case of the dog who steps on the mine, then it would take a miracle to
get the dog's parts back in the state they were in just before he stepped
on the bomb. But if God worked that miracle, I do not see why the dog
reassembled from exploded dog parts wouldn't be the dog whose life
had been disrupted by the explosion.

So I want to say that (NTB) makes it too hard for a living being to
go out of existence (just as it makes it too hard for teacups or temples
to go out of existence). If (NTB) is true (and my intuitions about mi-
raculous dog reassembly are right), then blowing up a dog is not suffi-
cient for causing him to cease to exist. For it is consistent with his be-
ing blown up that he be miraculously reassembled at some future time;
and (NTB) implies that if that happens, the dog will have existed con-
tinuously (in a very disassembled state) between the time of the explo-
sion and the time of reassembly. The proponent of (NTB) might say
that our dog ceases to exist upon explosion if but only if he will in fact
never be miraculously reassembled - and similarly that a corpse ceases
to exist upon cremation if but only if God will not reassemble the
corpse on the last day. Surely, though, we do not need to have a view
about what will or will not happen on the last day before we can de-
cide whether the body of Jerry Garcia (whose ashes have been scat-
tered in the Ganges) still exists. Alternatively, the defender of (NTB)
might say that as long as miraculous reassembly is so much as possi-
ble, the thing that could be reassembled still exists (in a disassembled
state). But this seems tantamount to the (implausible) view that the
only way to destroy a thing is to at least partially annihilate it (leaving
not enough parts of it to take part in a possible future reassembly).
4.
Suppose I bring home two qualitatively identical bicycles from the

shop - call them 'A' and 'B' .15 I then disassemble them completely.
Once they are reassembled, I put half of A's parts together with half of
B's parts, and the other half of A's parts together with the other half of
B's parts. (Let us suppose that the frames of A and B can each be dis-
assembled into two pieces). When I have done all this, I shall have two
(assembled) bicycles, which we may call C and D. Is C identical to
either A or B? It seems not. C obviously cannot be identical to both A
and B, since A and B are not identical to each other. But neither could
C be identical to A rather than B, or B rather than A, since it would be
completely arbitrary to identify C with A rather than B, or vice versa.

By the same reasoning, D is not identical to either A or B.
Now I disassemble C and D, and put all the parts that were in A
back together, and all the parts that were in B back together, so as to
get two bicycles - call them E and F - which are perfect duplicates of
A and B, as they were when I brought them home, and are made of ex-
actly the same parts that A and B were when I brought them home. It
seems clear enough that E and F are (respectively) A and B. So I have
at least two bicycles - A and B - at the end of the story. Moreover, I
am inclined to say, I have at most two bicycles at the end of the story.
After all, I came home with two bicycles. How could my swapping
parts around, and then putting them back exactly the way they were ex
ante tum two bicycles into four?
If, however, I end up with just two bicycles (A and B), it would
seem that the bicycles C and D no longer exist. And if C and D do not
exist at the end of the story, then, by parity of reasoning, A and B do
not exist in the middle of the story (when C and D do). In that case, A
and B go out of existence (at or before the time C and D come into ex-
istence) and come back into existence (at or after the time C and D go
out of existence); and we have a counterexample to (NTB).
There are at least two ways a defender of (NTB) might try to dis-
allow this kind of counterexample to it. Taking a leaf from David
Lewis' book l6 , she might insist that there are four bicycles at the end of
the story counting by identity, and explain our reluctance to say there
are four, by supposing that we do not count bicycles by identity. Al-
though I do not want to dismiss this suggestion, it does not seem espe-
cially promising. For one thing, it is unclear to me that we really can
count with propriety by any relation other than identity.17 But suppose
we can. In Lewis' sorts of cases, we get a smaller number of K's than
we would get if we counted K's by identity, by counting four-
dimensional objects as though we were counting temporal stages
thereof (by identity), or counting spatial wholes - say, roads - as
though we were counting spatial parts thereof - say, stretches of road
(by identity). In the bicycle case, we would presumably have to say
something like: in the case under discussion, we count bicycles as
though we were counting assembled bicycles (by identity). To which I
want to reply: why suppose we are counting in anything other than the
usual way? Why not suppose that we are counting by identity, and that
when the parts of a bicycle are dispersed in a certain way - when they
are, as EJ. Lowe puts it, "appropriated" by other bicycles - that bicy-
cle ceases to exist?
A rather different strategy for defending (NTB) from the alleged
counterexample would turn on the idea that 'bicycle' is a phase-sortal
for a kind of thing that ceases to be a bicycle, but does not cease to be,
when its parts are dispersed and appropriated. On one way of devel-
oping this suggestion, 'bicycle' would be a phase-sortal for a set of
bicycle parts. The idea would be that there are four (overlapping) sets
of bicycle parts -A, B, C, and D - in our story. Anyone of those sets
of bicycle parts is a bicycle when and only when its parts are related to
each other in the right way. 18 Nothing I do to the parts of A, B, C, and
D causes anything to go out of or come into existence. It merely causes
this or that individual to become or cease to be a bicycle. I have only
two bicycles at the beginning, in the middle, and at the end of the
story; and the bicycles I have at the beginning and at the end (A and B)
are different from the bicycles I have in the middle of the story (C and
D). But none of this is a threat to (NTB).
Again, I would not want to reject this suggestion out of hand. But I
do not see how to make it work. I doubt that 'bicycle' can be a phase-
sortal for a set of bicycle parts: if it were, bicycles could not survive
the replacement of most or all of their parts. But, waiving this objec-
tion, champions of this way of defending (NTB) are going to have
treat a whole of sortals as phase-sortals. For we can tell the sort of
story we told about bicycles about lots of other things. Suppose I have
a set of fifty cups. All of them become slightly damaged, losing a
(small) chip here or a (small) chip there. I collect all the chips and glue
them together, so that I now have fifty-one cups. Then I break the
fifty-first cup into the fragments I made it from, and glue each frag-
ment back onto the cup it originally belonged to. I end up with only
fifty cups. On the strategy for defending (NTB) under consideration, at
the end of the story my fifty-first cup still exists: although it has ceased
to be a cup, it hasn't ceased to be. I find this very difficult to believe.
Suppose a visitor says to me: "Last time I was here, I drank tea out of
a really strange looking cup. Where is it?". I do not think I could prop-
erly say: "Here, and here, and here .... " [pointing to all the bits in my
fifty cups]. What I am pointing to is not it (that cup you drank tea out
of), but a bunch of bits of china that used to but no longer constitute it.
(Compare this to the case in which a visitor says to you: "Last time I
was here [ten years ago] I met this really difficult teenager. Where is
she?" I could with perfect propriety say "Here she is", pointing to my
- no longer teenage - daughter).
Again, suppose a frog has a billion cells, and I have got a billion
frogs. I take one cell from each frog (making sure to pick the right
kind of cells), and then put them together in such a way that they are
taking part in the same frog-life. I now have a billion and one frogs.
Suppose I subsequently take apart the cells I put together, and put each
of them back in the frog it originally came from, leaving me with just
one billion frogs. On the envisaged defense of (NTB), we would deny
that my going back to the status quo ex ante caused any frog to go out
of existence: it just caused a frog to cease to be a frog. Again, though,
after all the billion cells have been put back in their original frogs, is it
really true that the frog different from those billion frogs is still around
- though it is no longer a frog - in the same way that a sapling planted
by the Thames twenty years ago is still around, although it's no longer
a sapling? I do not think so. In sum: if we block bicycle-type counter-
examples to (NTB) by appeal to the idea that the relevant sortals are
phase-sortals we will end up making it too hard for things (cups, frogs,
and so on) to go out of existence. 19
5.
One way to argue for (NTB) would be to appeal to our propensity, in a

(sufficient) variety of hypothetical cases, to distinguish an individual
who goes out of existence at a time from every individual existing af-
ter that time. It should be clear why I doubt this sort of argument will
succeed: for a good range of cases, it is far from clear that our intui-
tions align with (NTB). Another way would be to argue that if an indi-
vidual x at an earlier time t is identical to an individual y at a later time
t', then some causal condition must be satisfied whose satisfaction is
incompatibile with the interruption of x's existence between t and t'.
For example, someone might argue as follows: if x at tis y at t', then

there must be causal links between states x is in at t and states that y is
in at I' that entail the uninterrupted existence of x between I and I'.
lt seems clear that one cannot argue against the possibility of inter-
rupted existence by appeal to the principle that if x at I is y at I', then
there must be some causal links or olher between states of x at I and
states of y at I'. For in various of the alleged examples of interrupted
existence we have considered, there appear to be causal links between
states of x at I and states of the (allegedly re-existent, allegedly identi-
cal to x) y at I'. (The (allegedly) rebuilt Temple has a certain size and
shape at the time it is rebuilt because the Temple the Romans de-
stroyed had that size and shape just before they destroyed it. The (al-
legedly) reassembled dog has a cool, slightly damp nose right after
being (allegedly) reassembled because the dog who stepped on the
mine had a cool, slightly damp nose right before he stepped on the
mine; and so on). The causal connections in question are indirect, but
no less genuine for that.
Suppose that on the last day all of the particles constituting your
body at a certain time before your death will be reassembled to con-
stitute (to put it neutrally) a body B that is just like your body was at
that time. B will have a certain shape, and its having that shape will
have a causal history. If we go back far enough in time (e.g., before
your conception), that history will not include any events that involve
your body (since it does not yet exist). At some point the causal chain
will include events that involve your body. Thereafter (after the cor-
ruption of your body, say) the chain will no longer include any events
that involve that body (since it no longer exists).
I think that (at least some of the) philosophers who think causal
considerations can be invoked in support of (NTB) have in mind a
principle that goes something like this: if x at I = Y at I', then there is a
pair of states S, S', such that (i) S is a state of x at I, (ii) S' is a state of y
at I', (iii) S figures in the causal history of S', and (iv) the stretch of the
complete causal history of S' between f and f' always involves x (which
is to say, there is no time fi between f and fl, such that the complete set
of causes of S'at Ii includes no x-involving event).20
If this principle is right then - presuming that your body will not
be around (in a scattered state) until the last day - the last day body B
is not the body you have now. Moreover, none of the cases we have
discussed so far are genuine cases of a thing's going out of existence
and subsequently coming back. (If the thing x (frog, cup, temple, bicy-
cle ... ) existing at { went out of existence, and subsequently came back
into existence at {' , there would not be any state S that the thing was in
at that later time whose causal history was x-involving from { right
through to I').
An argument for (NTB) from the causality-persistence principle
just formulated is no more compelling than the principle itself. And the
principle does not look immediately compelling to me. Upon consid-
ering it, my first thought is that it is neither obviously true nor obvi-
ously false; my second thought is to test it against my intuitions con-
cerning various hypothetical cases (e.g. those involving taking things
apart and putting them back together); and my third thought, after
having done that, is to reckon that it is false. More generally, my sus-
picion is that attempts to argue for (NTB) will either depend on princi-
ples too weak to support (NTB), or depend on principles too strong for
us to put any confidence in.
But the main point I want to make here about the causality-
persistence principle under consideration is one lowe to Dean Zim-
merman. Zimmerman points out that someone might accept that a
thing can persist only if the stages of its career are held together by the
glue of (what he calls) immanent causality relations, and still insist on
the possibility of interrupted existence (for living or non-living things):
On the face of it, immanent-causal relatedness among stages of a thing
would seem to be compatible with its making discontinuous spatiotempo-
ral jumps, or even being "temporally gappy." If it is possible for an object
to persist through temporal gaps during which it has no stages, then there
must be suitable immanent-causal relations which cross the temporal gap
between earlier and later stages. But, given that the kind of immanent-
causal connections that normally preserve a Life could cross spatial and
temporal gaps, there is no reason to think that one and the same Life could
not contain spatial jumps or temporal gaps. (Zimmerman [1998], p.384)
This seems right. Suppose we discovered that whenever we put an ob-

ject in a certain place, it appeared to be annihilated in a blinding flash
of light. (Immediately) subsequent investigation would reveal no trace
of the object, or of any of its parts, anywhere at all. But a quarter of an

hour later something would invariably appear on that very spot that
was, to the best of our ability to determine, an absolutely perfect dupli-
cate of the thing that had vanished ten minutes ago. Granted that we
would not necessarily embrace it, wouldn't we at least entertain the
hypothesis that objects put in that place "jumped over" the intervening
quarter of an hour, and that there were direct immanent-causal con-
nections between states of the object at the moment it was put in that
place, and states of the object fifteen minutes later?
To sum up: it is at least initially plausible that persistence over
time requires causal relations between earlier and later states of the
persisting object. If it requires what Zimmerman calls immanent-
causal relations between earlier and later states of the persisting object,
on a certain understanding of immanence, then the alleged counterex-
amples to (NTB) I have discussed fail. But it is by no means clear that
persistence requires immanent-causal relations between earlier and
later states of the persisting object (so understood); and even if it does,
there still may be counterexamples to (NTB) of a quite different sort
than those that I have suggested. One could eliminate this possibility
only by showing that Hume was right about the impossibility of action
at a temporal distance. There is a parallel here between temporal con-
tinuity and spatial contiguity. Some philosophers have thought that
nothing can be a (partite) thing (at a time) unless its parts stand in
causal relations to each other (at that time). This is consistent with the
view that a partite object can be (spatially) "gappy". Similarly, even if
nothing can be an enduring thing (over time) unless its temporal parts
(or, if you prefer, the temporal parts of its history) are "non-gappily"
causally related, this is consistent with the view that a temporally par-
tite thing (or, if you prefer, the history of that thing) is temporally
"gappy".
6.
Zimmerman's remarks on the relation between immanent causation

and interrupted existence are interesting for another reason. In all of
the supposed cases of interrupted existence discussed in the first four

sections of this paper, the thing that goes out of existence and suppos-
edly returns is survived by its parts. Someone who was initially well
disposed to (NTB), but worried about apparent counterexamples to it,
might suggest that perhaps the truth in the neighorhood of (NTB) was
(NTB') Necessarily, nothing that goes out of existence without
leaving behind any parts ever comes back in existence.
Especially if she thought that nothing simple that goes out of existence
could leave behind any parts, she might hold that another truth in the
neighborhood of (NTB) was
(NTB") Necessarily, nothing simple that goes out of existence ever
comes back into existence.
If, however, what Zimmerman thinks is possible is in fact possible,
then (NTB') and (NTB") are false.
Actually, independently of the considerations raised by Zimmer-
man, I have doubts about the above weakened versions of (NTB).
People who think that simples cannot come back have sometimes said
something like this to me:
Suppose that a simple s goes out of existence, and that at some future
time, a duplicate simple s' pops into existence. On what grounds could we
conclude that s = s'? What could make it true that s = s'?
Let us suppose that it would be difficult or impossible to have evi-

dence that s = s'. It does not follow that s ;f. s'. Sydney Shoemaker has
suggested that an absent-minded deity might arrange that a stone tablet
disappear into thin air at a certain time, and subsequently arrange that
a stone tablet exactly like it appears in the same place immediately af-
terwards. 21 If this is possible, it is presumably also possible for a stone
tablet to (uncausedly) just disappear, and for a duplicate of it to (un-
causedly) appear immediately after in the very same place - even if it
would be very difficult to have grounds for supposing that what looked
like the straightforward persistence of the stone tablet was in fact its
uncaused disappearance, followed by its uncaused replacement by a
duplicate.
Similarly, perhaps there are possible situations in which a simple s
goes out of existence, and subsequently returns, even if we could never

have good grounds for thinking so. And in some moods I find it very
difficult to believe it is impossible for simples to come back. I suppose
that what is behind this is a reluctance - of the kind evinced by
Nicholas of Autrecourt and David Hume - to countenance necessary
connections between disjoint contingent states of affairs. Especially if
we think of s as an enduring (three-dimensional) object, wholly pres-
ent at each time it exists, that s exists at t1> that s does not exist at t2,
and that s exists at t3 look like non-overlapping contingent states of
affairs, none, or one, or two, or three of which could be actualized. I
grant, though, that there are reasons to doubt the maxim that there are
no necessary connections between disjoint contingent states of affairs.
Someone who thinks that simples that go are gone for good might
say: forget about the epistemological question, "on what grounds could
we judge that s = s'?" The question is: what could make it true that s =
s'? As Graham Forbes might put it, what could ground this identity? I
do not find this question worrisome, because I am happy with primi-
tive identity and primitive distinctness across worlds, and I do not see
any obvious reason to disallow primitive identity across time. 22
I shall conclude with a suggestion of a simple thing that might go
out of existence (without leaving any parts behind) and subsequently
exist again. According to an Aristotelian tradition, universals exist
only when they are exemplified. Suppose that, as seems perfectly pos-
sible, it is true at t1> false at t2, and true at t3 that something is an in-
stance of universal u. Then - if the Aristotelian tradition is right - u is
a (presumably simple) thing that goes out of existence after t1 (pre-
sumably leaving behind no parts) and comes back into existence by t3.
At least as long as we construe "thing" broadly enough, and assume
that universals are simples that exist when and only exemplified, uni-
versals (should they exist) would be a counterexample to (NTB') and
(NTB").
7.
Is it true that, as the Scholastics would put it, nihil polesl redire
(nothing can come back)? Is something rather like it true? Perhaps; the
case is yet to be made24•
Notes
I Aquinas attributes the view to Aristotle, and accepts a somwehat weakened
version of it himself (see his Commentary on the Sentences, IV, 1, 1, 1). I have often
heard the view attributed to Locke, though I have not been able to find a passage
where Locke explicitly endorses it. I first began to think about the principle years
ago when Anil Gupta suggested to me that it was one of the relatively few Lockean
principles that was true.
2 Fred Feldman has vigorously defended this view in his Confrontations with
the Reaper (cf. Feldman [1992]).
3 See my forthcoming "On the Real (?) Distinction between Persons and Their
Bodies".
4 Eric Olson expresses sympathy for and attempts to motivate this (Aristotelian)
view in Olson [1997], pp.150-52.
5 Olson says that in the Cartesian demon case, "the interruption of your biologi-
cal life is brief." I take it he cannot mean this. For he seems to think that a life cannot
outlast the organism living it (p.l37), in which case the demon who annihilates me
and replaces me with a perfect duplicate a thousandth of a second later terminates
my life, rather than interrupting it.
6 For more on animalism, and Olson's particular version of animalism, see my
[2001].
7 Cf. Olson [1997], p.133.
8 Olson's views appear to imply that if you separate my brainstem from the rest
of me (without destroying either), and separate Daniele Giaretta's brain stem from
the rest of him (without destroying either), and then transplant my brainstem into the
rest of Daniele, and Daniele's brainstem into the rest of me, at the end of the process,
the animal with almost of all my body and my mind will be not me but Daniele, and
the animal with almost all of Daniele's body and his mind will be not Daniele but
me. I find this very hard to believe.
9 I say, "I am nothing over and above my brainstem" rather than "I become my
brainstem", because, after the separation, one could truly say about me, but not of
my brainstem, "that's something that used to weigh about one hundred and forty five
pounds."
10 Van Inwagen suggests that the life of a frozen cat might be "squeezed into"
various small-scale physical processes. Although he says he finds this suggestion

attractive, he allows we might say instead that the frozen cat's life is suspended
rather than disrupted, and that a life that is suspended, unlike a life that is disrupted
may resume. If we say this, we will either have to give up either the esse viventibus
est vivere principle, or (NTB).
11 Of course, if you separate just a tiny part of something from most of its parts,
the thing usually goes on existing without the bit separated from it (e.g. as a chipped
cup); it's a throughgoing separation of part from part that is prejudicial to a thing's
continued existence.
12 See, for example, the Apocalypse of Peter 4: 3-4, "He will command the
beasts and the birds; He will command that they give back all the flesh they have
eaten, because He requires humans to make their appearance." (see Bauckman
[1988], p.272).
13 See van Inwagen [1990], p.145.
14 Ibid., p.147: "We may be confident that the life of an organism which has
been blown to bits by a bomb or which has died naturally and has been subject to the
normal, "room-temperature" processes of biological decay for, say, fifteen minutes
has been disrupted. [ ... ] Ifa life has been disrupted, it can never begin again; any life
that is going on after its disruption is not that life."
15 I briefly discuss the case I am describing in Hughes [1997].
16 See Lewis [1976].
17 See Hughes [1997a], p.65.

18 This might entail being assembled as a single bicycle. But it might only in-
volve not having parts that are "dispersively appropriated."

19 It is not crucial that, in the hypothetical situations discussed in the last two
sections, we could truly say: "That frog [teacup, temple, bicycle] has come back into
existence". For the purposes of challenging (NTB), it is enough if there is a
"sharpened" version of our ordinary concept of frog [teacup, temple, bicycle], or
simply a concept akin to our ordinary concept of frog [teacup, temple, bicycle] such
that, were we to mobilize that concept, we would truly judge that that frog' [teacup',
temple', bicycle'] has come back into existence (where the predicate 'frog" expresses
the sharpened version of or alternative to our ordinary frog-concept).
20 When Dean Zimmerman suggests that persistence requires relations of what
he calls immanent causality between earlier and later states of the persisting object, I
take it he is endorsing a principle in the neighborhood of the one formulated here.
See his remarks on the persistence of a body: "To say that immanent causal connec-
tions are required for the persistence of a body is to say that later states of the body
must be causally dependent, at least in part, on its earlier states. But not just any sort
of causal dependence seems sufficient to give us the kind of immanent causation that
is crucial to the persistence of a body. It is not enough [... ] that the way my body
was at death serve as a blueprint for God's creating a new one at the general resur-
rection. That is causal contribution of a sort; but here the causal chain passes through
God's mind; it does not remain "immanent" with respect to processes going on
within a living human body." (Zimmerman [1998]). As we shall see, Zimmerman
does not argue from the necessity of immanent causality for persistence, to the im-
possibility of uninterrupted existence.
21 Shoemaker [1979].
22 For an attack on the view that identity across time could not be primitive, see
Saul Kripke's unpublished lectures, "Time and Identity."
24 Thanks to Andrea Bottani, Pierdaniele Giaretta, Verity Harte, and Mario

Mignucci.
List of Contributors
Enrico Berti, Dipartimento di Filosofia, Universita di Padova, Italy,

enrico.berti@unipd.it
Andrea Bottani, Dipartimento di Scienze della Formazione e della
Comunicazione, Universita di Bergamo, Italy, abottani@unibg.it
Massimiliano Carrara, Dipartimento di Filosofia, Universita di Padova,
Italy, massirniliano.carrara@unipd.it
David Charles, Christ Church College, Sub-Faculty of Philosophy,
University of Oxford, United Kingdom,
david.charles@chch.ox.ac.uk
Paolo Crivelli, Department of Philosophy, University of Edinburgh,
United Kingdom, paolo.crivelli@ed.ac.uk
Dorothy Edgington, University College, Sub-Faculty of Philosophy,
dorothy.edgington@univ.oc.ac.uk
Kit Fine, Department of Philosophy, New York University, United
States, kf14@nyu.edu
Graeme Forbes, Department of Philosophy, Tulane University, United
States, forbes@mailhost.tcs.tulane.edu.
Pierdaniele Giaretta, Dipartimento di Filosofia, Universita di Padova,
Italy, pierdaniele.giaretta@unipd.it
Richard Glauser, Institut de Philosophie, Universite de Neuchiitel,
Swiss, richard.glauser@unige.ne.ch
Christopher Hughes, Department of Philosophy, King's College
London, United Kingdom, christopher.hughes@kcl.ac.uk
E. Jonathan Lowe, Department of Philosophy, University of Durham,
United Kingdom, e.j.lowe@durham.ac.uk
Penelope Mackie, Department of Philosophy, the University of
Birmingham, United Kingdom, pj.mackie@bir.ac.uk
Mauro Mariani, Dipartimento di Filosofia, Universita di Pisa, Italy,

mariani@fls.unipi.it
Uwe Meixner, Institut fUr Philosophie, Universitat Regensburg,
Germany, uwe.Meixner@mz-net.com
Mario Mignucci, Department of Philosophy, King's College London,
United Kingdom, mario.mignucci@abc.it
Alex Orenstein, Philosophy Program, the City University Graduate
Center, United States, philoren@forbin.qc.edu
Edmund Runggaldier, Institut flir Christliche Philosophie, Abteilung
fUr Metaphysik und Philosophische Gotteslehre, Innsbruck,
Austria, edmund.runggaldier@uibk.ac.at
Anthony Savile, Department of Philosophy, King's College London,
United Kingdom, anthony.savile@kcl.ac.uk
KatheTrettin, Merianstrasse 30, D-60316, Frankfurt am Main,
Germany, kaethe. trettin @t-online.de
Peter van Inwagen, Department of Philosophy, University of Notre
Dame, United States, vaninwagen.l @nd.edu
Achille Varzi, Department of Philosophy, Columbia University,
United States, achille. varzi@columbia.edu
David Wiggins, New College, Sub-Faculty of Philosophy, University
of Oxford, United Kingdom, enquiries@philosophy.ac.uk
Timothy Williamson, New College, Sub-Faculty of Philosophy,
timothy. williamson@philosophy.ox.ac.uk
Stephen Yablo, Department of Linguistics & Philosophy,
Massachusetts Institute of Technology (MIT), United States,
yablo@mit.edu
Bibliography
Adams, R.M., 1994, Leibniz: Determinist, Theist, Idealist, Oxford:

Oxford University Press.
Akiba, K., 2000, "Vagueness as a modality", Philosophical Quarterly
50,359-370.
Alston, W. P., 1958, "Ontological Commitments", Philosophical
Studies 9,8-17 (reprinted in S. Laurence and C. Macdonald (eds.),
Contemporary Readings in the Foundations of Metaphysics, Ox-
ford: Blackwell, 1998,46-54).
Andjelkovic, M., and Williamson, T., 2000, "Truth, falsity and bor-
derline cases", Philosophical Topics 28,211-244.
Armstrong, D.M., 1983, What is a Law of Nature?, Cambridge: Cam-
Armstrong, D.M., 1989a, A Combinatorial Theory of Possibility,
Cambridge: Cambridge University Press.
Armstrong, D.M., 1989b, Universals: An Opinionated Introduction,
Boulder (CO): Westview Press.
Armstrong, D.M., 1997, A World of States of Affairs, Cambridge:
Cambridge University Press.
Aubenque, P., 1978, Etudes sur la Metaphysique d'Aristote, Paris:
Vrin.
Austin, J.L., 1962, Sense and Sensibilia, Oxford: Oxford University
Press.
Austin, J.L., 1968, "'Agathon' and 'Eudaimonia' in the 'Ethics' of
Aristotle", in Moravcsik [1968], 261-296 (reprinted in Austin
[1970], 1-31).
Austin, J.L., 1970, Philosophical Papers, Oxford: Oxford University
Press.
Bacon, J., Campbell, K., and Reinhardt, L. (eds.), 1993, Ontology,

Causality and Mind: Essays in Honour of D. M Armstrong, Cam-
bridge: Cambridge University Press.
Baker, L. R, 1997, "Why Constitution Is Not Identity", Journal of
Philosophy 94, 599-621.
Balaguer, M., 1996, "A Fictionalist Account of the Indispensable Ap-
plications of Mathematics", Philosophical Studies 83, 291-314.
Baldwin, T., 1979, "Interpretations of Quantifiers", Mind 88, 215-240.
Bambrough, R (ed.), 1965, New Essays on Plato and Aristotle, Lon-
don: Routledge and Kegan Paul.
Barnes, J., Schofield, M., Sorabji, R (eds.), 1975, Articles on Aristotle,
vol. I, London: Duckworth.
Barnes, J., 1981, "Proof and the Syllogism", in E. Berti (ed.), Aristotle
on Science. The "Posterior Analytics ", Proceedings of the Eight
Symposium Aristotelicum Held in Padua from September 7 to 15,
1978, Padova: Antenore, 17-59.
Barnes, J. (ed.), 1985, The Complete Works of Aristotle, The Revised
Oxford Translation, Princeton: Princeton University Press.
Barnes, J., 1995, "Metaphysics" in Barnes, J. (ed.), The Cambridge
Companion to Aristotle, Cambridge: Cambridge University Press,
66-108.
Bauckman, R, 1988, The Fate of the Dead: Studies on the Jewish and
Christian Apocalypses, Leiden: E.J.Brill.
Beierwaltes, W., 1972, Platonism us und Idealismus, Frankfurt:
Klostermann.
Bencivenga, E., 1986, "Free Logic", in D. Gabbay and F. Guenther
(eds.), Handbook of Philosophical Logic, Vol 3, Dordrecht: Rei-
del, 1986,373-427.
Bennett, J.F., 1988, Events and Their Names, Oxford: Clarendon
Press.
Berti, E., 1978, "Le probleme de la substantialit6 de l'etre et de l'un
dans la Metaphysique", in Aubenque [1978], 89-130.
Berti, E., 1987, "II problema dell'identita nell'odierna filosofia anglo-
sassone (Strawson, Kripke, Wiggins, Hamlyn)", in V. Melchiorre
(ed.), La difJerenza e l'origine, Milano: Vita e Pensiero, 212-213.
Blackburn, S., 1984, Spreading the Word, Oxford: Oxford University
Press.
Bibliography 481
Blackburn, S., 1992, Essays in Quasi-Realism. Oxford: Oxford Uni-

versity Press.
Blackburn, S., 1998, Ruling Passions: A Theory of Practical Reason-
ing. Oxford: Clarendon Press.
Bochenski, LM., 1968, Ancient Formal Logic, Amsterdam: North-
Holland.
Bondi, H., 1961, Cosmology, Cambridge: Cambridge University Press.
Bonitz, H., 1955 2 , Index Aristotelicus, Graz: Akademische Druk-u.:
Verlagsansta1t.
Brand, M., 1977, "Identity Conditions for Events", American Philo-
sophical Quarterly 14,329-377.
Brown, L., 1994, "The verb 'to be' in Greek Philosophy: some Re-
marks", in F. Brunner (ed.), 1951, Etudes sur la signification his-
torique de la philosophie de Leibniz, Paris: Vrin.
Brunner, F, 1951, Etudes sur la signification historique de la philoso-
phie de Leibniz, Paris: Vrin.
Brunschwig, J.B. (ed.), 1967, Aristote, Topiques, I, Paris: Les Belles
Lettres.
Burge, T., 1974, "Truth and Singular Terms", Nous 3, 309-25, revised
version in M. Platts (ed.) Reference, Truth, and Reality, London:
Routledge and Kegan Paul, 1980, 167-80.
Burgess, J. and Rosen, G., 1997, A Subject With No Object. Strategies
for Nominalistic Interpretations of Mathematics, Oxford: Claren-
don Press.
Burnyeat, M., 1979, Notes on the Book Zeta of Aristotle's Metaphys-
ics, in M. Burnyeat et al. (eds.), Oxford: The Sub-faculty of Phi-
losophy.
Carrara, M., and Varzi, A. c., 2001, "Ontological Commitment and
Reconstructivism", Erkenntnis 55, 33-50.
Casati, R., and Varzi, A.c., 1994, Holes and Other Superficialities,
Cambridge (MA): MIT Press.
Casati, R., and Varzi, A. C., 1999, Parts and Places. The Structures of
Spatial Representation, Cambridge (MA): MIT Press.
Chalmers, D. J., 1996, The Conscious Mind, Oxford: Oxford Univer-
sity Press.
Charles, D., 2000, Aristotle on Meaning and Essence, Oxford: Claren-
don Press.
Chisholm, R., 1970, "Identity Through Time", in H.E. Kiefer and M.

Munitz (eds.), Language, Belief and Metaphysics, New York:
State University of New York Press, 163-182.
Chisholm, R., 1996, A Realistic Theory of Categories: an Essay on
Ontology. Cambridge: Cambridge University Press.
Code, A., 1996, "Owen and the Development of Aristotle's Meta-
physics", in Wians [1996], 303-326.
Davidson, D., 1967, "The Logical Form of Action Sentences", in N.
Rescher (ed.), The Logic of Decision and Action, Pittsburgh: Uni-
versity of Pittsburgh Press, 81-95 (as reprinted in Davidson's Es-
says on Actions and Events, Oxford: Clarendon Press, 1980, 105-
122).
Davidson, D., 1980, "The Individuation of Events", in Essays on Ac-
tions and Events, Oxford: Clarendon Press,163-80.
Davidson, D., 1985, "Reply to Quine on Events", in E. LePore and B.
McLaughlin (eds.), Actions and Events: Perspectives on the Phi-
losophy ofDonald Davidson, Oxford: Blackwell, 172-6.
Della Rocca, M., 1996, "Essentialists and Essentialism", Journal of
Donini, P., 1995, La Metafisica di Aristotele. Introduzione alia lettura,
Florence: La Nuova Italia.
Drewery, A., 2000, "Laws, Regularities and Exceptions", Ratio 13, 1-
12.
Ducasse, C. J., 1942, "Moore's Refutation of Idealism", in P.A.
Schilpp (ed.), The Philosophy of G. E. Moore, Evanston (IL): Li-
brary of Living Philosophers, 225-251.
Dunn, M.J. and Belnap, N.D., 1968, "The Substitution Interpretation
of the Quantifiers", Nous 2 177-185.
Dunn, J.M., 1990, "Relevant Predication II: Intrinsic Properties and
Internal Relations", Philosophical Studies 60, 177-206.
Dummett, M., 1973, Frege. Philosophy of Language, London: Duck-
worth.
Dummett, M., 1976, "What is a Theory of Meaning? (II)", in G. Evans
and J. McDowell (eds.), Truth and Meaning: Essays in Semantics,
Oxford: Oxford University Press, 67-137.
Dummett, M., 1978, Truth and Other Enigmas, London: Duckworth.
Bibliography 483
Dummett, M., 1981, The interpretations of Frege's Philosophy, Har-

vard: Harvard University Press.
Dummett, M., 1991, Frege: Philosophy of Mathematics, London:
Duckworth.
Dummett, M., 1993, The Seas of Language, Oxford: Clarendon Press.
During, I., and Owen, G.E.L. (eds.), 1960, Aristotle and Plato in the
mid-fourth century, Goteborg-Stockholm-Uppsala: Almqvist &
Wiksell.
During, I., 1966, Aristoteles. Darstellung und Interpretation seines
Denkens, Heidelberg: Carl Winter.
Dworkin, R., 1996, "Objectivity and Truth: You'd Better Believe It",
Philosophy and Public Affairs 25,87-139.
Edgington, D., 1980-81, "Meaning, Bivalence and Realism", Pro-
ceedings of the Aristotelian Society 81, 153-73.
Edgington, D.,1996, "Vagueness by degrees", in R. Keefe and P.
Smith (eds.), Vagueness: A Reader, Cambridge (MA): MIT Press,
294-316
Ellis, B., 1999, "Causal Powers and Laws of Nature", in Sankey (ed.)
[1999], 19-34.
Evans, G. , 1978, "Can there be vague objects?", Analysis 38, 208.
Everson, S. (ed.) 1994, Language (The Companions to the Ancient
Thought), Cambridge: Cambridge University Press.
Fales, E., 1993, "Are Causal Laws Contingent?", in Bacon et al. (eds.)
[1993], 121-44.
Feldman, F., 1992, Corifrontations with the Reaper, Oxford: Oxford
University Press.
Field, H.,1980, Science Without Numbers, Princeton: Princeton Uni-
versity Press.
Field, H., 1989, Realism, Mathematics, and Modality, Oxford: Black-
well.
Field, H., 2001, Truth and The Absence of Fact, Oxford: Clarendon
Press.
Fine, A., 1984, "The Natural Ontological Attitude", in J. Leplin (ed.),
Scientific Realism, Berkeley (CA): University of California Press,
149-177.
Fine, K., 1975, "Vagueness, Truth and Logic", Synthese 30, 265-300.
Fine, K., 1994, "Essence and Modality", in J. E. Tomberlin (ed.)

Philosophical Perspectives Volume 8, Atascadero (CA):
Ridgeview Publishing Company, 1-16.
Finster, R, Hunter, G., McRae, R, Miles, M. and Seager, W.E., 1988,
Leibniz Lexikon, Olms: Hildesheim.
Forbes, G., 1980a, "Origin and Identity", Philosophical Studies 37,
353-362.
Forbes, G., 1980b, "Relative Identity and Anti-Essentialism", in Pro-
ceedings of the 4th International Wittgenstein Symposium, Vienna:
Holder-Pichler-Tempsky, 490-492.
Forbes, G., 1985, The Metaphysics of Modality, Oxford: Oxford Uni-
versity Press.
Forbes, G., 1986, "In Defense of Absolute Essentialism", in French,
Uehling and Wettstein (eds.) [1986],3-31.
Forbes, G., 1987, "Is There a Problem about Persistence?", Proceed-
ings of the Aristotelian Society Supplementary Volume 61, 137-
155.
Forbes, G., 1994, Modem Logic: A Text in Elementary Symbolic
Logic, Oxford: Oxford University Press.
Forbes, G., 1994a, "A New Riddle of Existence", in J. Tomberlin (ed.)
Philosophical Perspectives 8: Logic and Language, Atascadero
(CA): Ridgeview, 415-30.
Forbes, G., 1999, "Essentialism Reconsidered", in P. Simons and U.
Meixner (eds.), Proceedings of the 22nd International Wittgen-
stein Symposium, Vienna: Holder-Pichler-Tempsky, forthcoming.
Foster, J., 1982, "Induction, Explanation and Natural Necessity", Pro-
ceedings of the Aristotelian Society 83,87-101.
Francescotti, R, 1999, "How to Define Intrinsic Properties", Noils 33,
590-609.
Frede, M., 1987, "The Unity of General and Special Metaphysics: Ar-
istotle's Conception of Metaphysics", in M. Frede, Essays in An-
cient Philosophy, Minneapolis: University of Minnesota Press, 81-
95.
Frede, M., and Patzig, G., 1988, Aristoteles 'Metaphysik Z', Munchen:
Beck.
Bibliography 485
French, P., Uehling, T., and Wettstein, H. (eds.), 1986, Midwest Stud-
ies in Philosophy XI: Studies in Essentialism, Minneapolis: Uni-
versity of Minnesota Press.
Francescotti, R., 1999, "How to Define Intrinsic Properties", Noils 33,
590-609.
Freddoso, M. and Schuurman R. (eds.), 1980, Ockham, W., Ockham's
Theory of Propositions, Notre Dame (IN): University of Notre
dame Press.
Gaifinan, H., 1975, "Ontology and Conceptual Frameworks", Erk-
enntnis 9, 329-353.
Gallois, A., 1998, Occasions of Identity. The MetaphysiCS of Persis-
tence, Change, and Sameness, Oxford: Clarendon Press.
Garrett, B., 1988, "Best-Candidate Theories and Identity", Inquiry, 79-
85.
Geach, P., 1972, "Some Problems about Time" in Geach [1972], 302-
317.
Geach P.T., 1972, Logic Matters, Oxford: Blackwell
Geach, P.T., 1980, Reference and Generality. An Examination of some
Medieval and Modern Theories, Ithaca: Cornell University Press.
Gerhardt, C.I. (ed.), 1978, Die Philosophischen Schriften von G. W
Leibniz~ Olms: Hildesheim.
Gerhardt C.L (ed.), 1962, Leibnizen's Mathematische Schriften,
Hildesheim: Olms.
Gibbard, A., 1975, "Contingent Identity", Journal of Philosophical
Logic 4, 187-221.
Goldman, A. L., 1971, "The Individuation of Action", Journal of Phi-
losophy 68, 761-774.
Grice, P., 1988, "Aristotle on the MUltiplicity of Being", Pacific
Philosophical Quarterly 69, 175-200.
Grua G., 1953, Jurisprudence universe lie et theodicee selon Leibniz,
Paris: PUF.
Hacker, P.M.S., 1982, "Events, Ontology and Grammar", Philosophy
57,477-486.
Hale, B. and Wright C., 1996, "Nominalism and the Contingency of
Abstract Objects", in M. Schirn, (ed.) Frege: Importance and
Legacy, Berlin: de-Gruyter.
Hambruch, E., 1904, Logische Regeln des Platonischen Schule in der

Aristotelischen Topik, Berlin: Widmannsche Buchhandlung.
Hawthorne, J. and Gendler, T.S., 2000, "Origin Essentialism: the Ar-
guments Revisited", Mind 109,285-98.
Harman, G., 1977, The Nature of Morality, Oxford: Clarendon Press.
Heck, R.G., Jr., 1998, "That There Might be Vague Objects (so far as
Concerns Logic)", The Monist 81, 274-296.
Hellman, G., 1989, Mathematics Without Numbers, Oxford: Claren-
don.
Hintikka, J., 1959, "Aristotle and the Ambiguity of Ambiguity", In-
quiry 2, 137-151.
Hintikka, 1.,1970, "The Semantics of Modal Notions", Synthese, 20,
408-424.
Hintikka, 1., 1986, "Kant on Existence, Predication, and the Ontologi-
cal Argument", in Knuutila S. and Hintikka J. (eds.), The Logic of
Being, Dordrecht: Reidel, 249-267.
Hintikka, 1., 1999, "On Aristotle's Notion of Existence", The Review
of Metaphysics 52, 779-805.
Hirsch, E., 1999, "The Vagueness of Identity", Philosophical Topics,
26, 139-158.
Horgan, T., 1978, "The Case Against Events", Philosophical Review
87,28-47.
Horwich, P., 1998, Meaning, Oxford: Clarendon Press.
Hughes, C., 1997, "Aquinas on Continuity and Identity", Mediaeval
Philosophy and Theology 6, 93-108.
Hughes, C., 1997a, "Same-Kind Coincidence and the Ship of The-
seus", Mind 106, 53-97.
Hughes c., 2001, "Entita Personale e Identita Personale" in A. Bottani
A. and N. Vassallo (eds.), Identita Personale, Napoli: Loffredo,
341-383.
Hughes G.E. and Cresswell M.J., 1996, A New Introduction to Modal
Logic, London-New York: Routledge.
Humberstone, L., 1996, "Intrinsic/Extrinsic", Synthese 108,205-267.
Jackson, F., 1998, From MetaphysiCS to Ethics, Oxford: Clarendon
Press.
Johnston, M., 1992, "Costitution Is Not Identity", Mind 101,89-105.
Bibliography 487
Jolley, N., 1984, Leibniz and Locke: A Study of the «New Essays on
Human Understanding», Oxford: Clarendon Press.
Kamp, H., 1975, "Two Theories about adjectives", in E. Keenan (ed.),
Formal Semantics of Natural Language, Cambridge: Cambridge
University Press, 123-155.
Kaplan D., 1969, "Quantifying in", in D. Davidson and J. Hintikka
(eds.), Words and Objections: Essays on the Work of wv.o.
Quine, Dordrecht: Reidel, 178-214.
Kim, J., 1993, Supervenience and Mind, Cambridge: Cambridge Uni-
versity Press.
Kripke, S., 1972, "Naming and Necessity", in D. Davidson and G.
Harman (eds.), Semantics of Natural Language, Dordrecht: Rei-
del,252-355.
Kripke, S., 1980, Naming and Necessity, Oxford: Blackwell.
Laks A. and Most G. (eds.), 1993, Theophrastus, Metaphysique, Paris:
Les Belles Lettres.
Lambert, K., 1983, Meinong and the Principle of Independence, Cam-
Lejewski, c., 1954, "Logic and Existence", British Journal for the
Philosophy of Science 5, 104-119.
Lejewski, C., 1958, "On Lesniewski's Ontology", Ratio 1, 150-176.
Leszl, W., 1970, Logic and MetaphysiCS in Aristotle, Padua: Ante-
nore.
Lewis, D. K., and Lewis, S. R., 1970, "Holes", Australasian Journal of
Philosophy 48,206-212 (as reprinted in Lewis [1983a], 3-9).
Lewis D., 1976, "Survival and Identity", in A. Oksenberg Rorty (ed.),
The Identities of Persons, Berkeley: California University Press,
1976, 17-40.
Lewis, D., 1983, "In Defense of Stages", appendix b to Lewis [1976]
in Lewis [1983a], 76-77.
Lewis, D., 1983a, Philosophical Papers, Vol. 1, New York: Oxford
University Press.
Lewis, D., 1986, On the Plurality of Worlds, Oxford: Blackwell.
Lewis, D., 1986a, "Events", in Philosophical Papers, vol. 2, Oxford:
Oxford University Press, 241-269.
Lewis, D., 1988, "Vague identity: Evans misunderstood", Analysis 48,
128-130.
Loemker, L. E. (ed.), 1976, G. W. Leibniz, Philosophical Papers and

Letters, Dordrecht: Reidel.
Loux, M. J., 1998, Metaphysics. A Contemporary Introduction, Lon-
don-New York: Routledge.
Lowe, E. J., 1982, "On the Alleged Necessity of True Identity State-
ments", Mind 91, 579-84.
Lowe, E.J., 1987, "Miracles and Laws of Nature", Religious Studies
23,263-78.
Lowe, E.1., 1989, Kinds of Being: A Study of Individuation, Identity
and the Logic ofSortal Terms, Oxford: Blackwell.
Lowe, E.l., 1994, "Vague Identity and Quantum Indeterminacy",
Analysis 54, 110-114.
Lowe, E.1., 1995, "The Truth about Counterfactuals", Philosophical
Quarterly 45, 41-59.
Lowe, E.1., 1998, The Possibility of Metaphysics: Substance, Identity,
and Time, Oxford: Clarendon Press.
Lowe, E.l., 1999, "Abstraction, Properties, and Immanent Realism", in
Tom Rockmore (ed.), Proceedings of the 20th World Congress of
Philosophy, Volume 2: Metaphysics, Bowling Green (OH): Phi-
losophy Documentation Center, 195-205.
Lowe, E.1., 2000, An Introduction to the Philosophy of Mind, Cam-
Mackie, 1. L., 1974, "De What Re is De Re Modality?", The Journal of
Philosophy 71,551-561.
Mackie, P., 1987, "Essence, Origin and Bare Identity", Mind 96, 173-
201.
Mackie, P., 1989, "Identity and Extrinsicness: Reply to Garrett", Mind
98, 105-117.
Mackie, P., 1998, "Identity, Time, and Necessity" Proceedings of the
Aristotelian Society 98, 59-78.
Macnabb D.G.C. (ed.), 1987, David Hume, A Treatise of Human Na-
ture, Book I, Glasgow: Fontana/Collins.
Madigan, A., (ed.), 1999, Aristotle, Metaphysics, Books Band K 1-2,
Oxford: Clarendon Press.
Maier, H, 1969-1970, Die Syl/ogistik des Aristoteles, 3rd ed., 3 vols.,
Hildesheim-New York: Olms.
Bibliography 489
Marconi, D., 1979, "Le Ambigue VirtU della Forma Logica", in Tempo
verbale e strutture quantificate in forma logica, Atti del Semina-
rio, Firenze: Accademia della Crusca, 265-284.
Marcus, R.B., 1962, "Interpreting Quantification", in Inquiry 5, 252-
259.
Martin, C.B., 1993, "Power for Realists", in Bacon et al. (eds.) [1993]
175-86.
Mates, B., 1971, "Leibniz on Possible Worlds", in Frankfurt H. (ed.),
Leibniz, New York: Dobulbeday.
Mates, B., 1972, Elementary Logic, New York: Oxford University
Press.
Matthews, G.B., 1973, "Suppositio and Quantification in Ockham",
Nous 7, 13-24.
McGee, V., and McLaughlin, B., 1995, "Distinctions without a differ-
ence", Southern Journal ofPhilosophy 33, 203-251.
McGinn, C., 1976, "On the Necessity of Origin", The Journal of Phi-
losophy 73, 127-35.
McKay, T. J., 1986, "Against Constitutional Sufficiency Principles",
in French, Uehling and Wettstein (eds.) [1986], 295-304.
Meixner, U., 1997, Axiomatic Formal Ontology, Dordrecht: Kluwer.
Meixner, U., 1997a, Ereignis und Substanz, Paderborn: SchOningh.
Melia, J., 1995, "On What There's Not", Analysis 55, 223-229.
Miller, F.D.M., 1973, "Did Aristotle Have the Concept of Identity?",
The Philosophical Review 82, 483-90.
Moravcsik, J.M.E. (ed.) 1968, Aristotle. A Collection of Critical Es-
says, London-Melbourne: Macmillan.
Munitz, M. (ed.), 1973, Logic and Ontology, New York: New York
University Press.
Neale, S., 1990, Descriptions, Cambridge (MA): MIT Press.
Nidditch, P. (ed.), 1975, John Locke, An Essay Concerning Human
Understanding, Oxford: Clarendon Press.
Oliver, A., 1999, "A Few More Remarks on Logical Form", Proceed-
ings of the Aristotelian Society 99,247-272.
Olson, E., 1997, The Human Animal, Oxford: Oxford University Press.
Orenstein, A., 1978, Existence and the Particular Quantifier, Phila-
delphia: Temple University Press.
Orenstein, A., 1990, "Is Existence What Existential Quantification Ex-

presses?" in R. Barrett and R. Gibson (eds.), Perspectives on
Quine, Oxford: Blackwell, 245-270.
Orenstein, A., 1995, "How to Get Something from Nothing", Pro-
ceedings of the Aristotelian Society 95,93-112.
Orenstein, A, 1999, "Reconciling Aristotle and Frege", Notre Dame
Journal ofFormal Logic 40, forthcoming.
Orenstein, A, 2000, "Plato's Beard, Quine's Stubble and Ockham's
Razor" in A. Orenstein and P. Kotatko (eds.), Knowledge, Lan-
guage and Logic: Questions for Quine, Dordrecht: Kluwer Aca-
demic Publishers.
Orenstein, A, 2000a ,"The Logical Form of Categorical Sentences",
Australasian Journal of Philosophy 78,517-533.
Owen, G.E.L.,1960, "Logic and metaphysics in some earlier works of
Aristotle", in During and Owen (ed.) [1960], 162-190 (reprinted in
Owen [1986], 180-199).
Owen, G.E.L., 1965, "Aristotle and the Snares of Ontology", in
Bambrough (ed.) [1965], 69-95 (reprinted in Owen [1986], 259-
278).
Owen, G.E.L., 1986, Logic, Science and Dialectic, edited by M. Nuss-
baum, Ithaca, Cornell University Press.
Owens, J., 1973, "The Content of Existence" in Munitz [1973], 21-35;
Owens, J., 1978a, The Doctrine of Being in the Aristotelian Meta-
physics, Third Edition, revised, Toronto: Pontifical Institute of
Mediaeval Studies.
Owens, J., 1978b, "The Relation of God to the World in the Meta-
physics", in Aubenque (ed.) [1978],207-228.
Pacius, J.P., 1597, In Porphyrii Isagogen et Aristotelis Organum
Commentarius Analyticus, repro Hildesheim: Olms 1966.
Parkinson, G.H.R. (ed.), 1973, Leibniz Philosophical Writings, Lon-
don: Dent.
Parsons, T., 1987, "Entities without Identity", Philosophical Perspec-
tives 1, 1-19.
Parsons, T., 1987-88, "Underlying States in the Semantica1 Analysis of
English", Proceedings of the Aristotelian Society 88, 13-30.
Parsons, T., and Woodruff, P., 1995, "Worldly Indeterminacy of Iden-
tity", Proceedings of the Aristotelian Society 95, 171-191.
Bibliography 491
Patzig, G., 1960-61, "Theologie und Ontologie in der Metaphysik des

Aristoteles", Kant-Studien 52, 185-205.
Pianesi, F., and Varzi, A. c., 2000, "Events and Event Talk: An Intro-
duction", in J. Higginbotham, F. Pianesi, and A.CVarzi (eds.),
Speaking of Events, New York: Oxford University Press, 3-47.
Plantinga, A., 1974, The Nature of Necessity, Oxford: Clarendon
Press.
Price, M., 1982, "On The Non-Necessity of Origin", The Canadian
Journal of Philosophy 12, 33-45.
Priest, G., 1998, "Fuzzy identity and local validity", The Monist 81,
331-342.
Putnam, H., 1967, "Mathematics without Foundations", The Journal of
Putnam, H., 1975, "The Meaning of 'Meaning''', in Mind, Language
and Reality: Philosophical Papers, Cambridge: Cambridge Uni-
versity Press, 215 -71 .
Putnam, H., 1978, Meaning and the Moral Sciences, London: Rout-
ledge & Kegan Paul.
Putnam, H., 1987, The Many Faces of Realism, La Salle (IL): Open
Court.
Quine, W.v.O., 1940, Mathematical Logic, Cambridge (MA): Harvard
University Press.
Quine, W.v.O., 1948, "On What There Is", Review of Metaphysics 2,
21-38 (reprinted in Quine [1953], 1-19).
Quine, W.v.O., 1953, From a Logical Point of View, Cambridge
(MA): Harvard University Press.
Quine, W.v.O., 1960, Word and Object, Cambridge (MA): MIT Press.
Quine, W.v.O., 1963, Set Theory and Its Logic, Harvard: Harvard
University Press.
Quine, W.v.O., 1964, "Review ofP.T. Geach, Reference and General-
ity" Philosophical Review 73, 100-104.
Reale, G. (ed.), 1993, Aristotele, Metafisica, Milano: Vita e Pensiero.
Remnant, P. and Bennet, J.F. (eds.), 1981, John Locke, New Essays on
Human Understandigs, Cambridge: Cambridge University Press.
Rescher, N., 1993, Leibniz. An Introduction to His Philosophy, Alder-
shot: Gregg Revivals.
Richard, M., 1987, "Quantification and Leibniz's Law", The Philo-

sophical Review 96, 555-578.
Riley, P. (ed.), 1988, Leibniz, Political Writings, Cambridge: Cam-
Robertson, T., 1998, "Possibilities and the Arguments for Origin Es-
sentialism" Mind 107, p. 729-49.
Rolfes E.R. (ed.), 1919, Aristoteles, Topik, Leipzig: Meiner.
Rorty, R., 1979, Philosophy and the Mirror of Nature, Princeton:
Princeton University Press.
Rosen, G., 1994, "Objectivity and Modern Idealism: What is the
Question?", in M. Michaelis and J. O'Leary-Hawthorne (eds.),
Philosophy in Mind, Amsterdam: Kluwer, 277-319.
Rosenkrantz, G., 1998, "The Science of Being", Erkenntnis 48, 251-
255.
Ross, W.D., 1939, "The Discovery of the Syllogism", The Philosophi-
cal Review, 48,251-272.
Ross, W.D. (ed.), 1953, MetaphYSiCS, Oxford: Clarendon Press.
Ross, W.D. (ed.), 1958, Aristotelis Topica et Sophistici Elenchi, Ox-
ford: Clarendon Press.
Russell, B., 1919, Introduction to Mathematical Philosophy, London:
Allen and Unwin.
Russell, B. and Whitehead, A.H., 1927, Principia Mathematica, 2nd
ed., Cambridge: University Press.
Ryle, G., 1931-32, "Systematically Misleading Expressions", Pro-
ceedings of the Aristotelian Society 32, 139-170.
Ryle, G., 1949, The Concept ofMind, London: Hutchinson.
Sainsbury, M., 1999, "Names, Fictional Names, and 'Really'" Pro-
ceedings of the Aristotelian Society, Supplementary 73, 271-286.
Salmon, N., 1979, "How Not to Derive Essentialism from the Theory
of Reference", The Journal ofPhilosophy 76,703-725.
Salmon, N., 1982, Reference and Essence, Oxford: Blackwell.
Salmon, N. U., 1986, Frege 's Puzzle, Cambridge (MA): MIT Press.
Sankey, H. (ed.), 1999, Causation and Laws of Nature, Dordrecht:
Kluwer.
Scott-Kilvert I. (ed.), 1960, Plutarch, The Rise and Fall of Athens:
Nine Greek Lives, London: Penguin.
Bibliography 493
Scott, T.K. (ed.), 1966, J. Buridan, Sophisms on Meaning and Truth,

New York: Appleton Century-Croft.
Shields, c., 1999, Order in Multiplicity. Homonymy in the Philosophy
ofAristotle, Oxford: Clarendon Press.
Shoemaker, S., 1979, "Identity, Properties, and Causality", P. French,
T. Uehling and H. Wettstein, (eds.), Midwest Studies in Philoso-
phy: Studies in Metaphysics 4,321-43.
Shoemaker, S., 1980, "Causality and Properties", in P. van Inwagen
(ed.), Time and Cause, Dordrecht: Reidel, 109-35, reprinted in
Shoemaker [1984].
Shoemaker, S., 1984, Identity, Cause and Mind: Philosophical Essays,
Cambridge: Cambridge University Press.
Shoemaker, S., 1998, "Causal and Metaphysical Necessity", Pacific
Philosophical Quarterly 79,59-77.
Simons, P. M., 1987, Parts. A Study in Ontology, Oxford: Clarendon
Press.
Smith, N.K., 1953, Immanuel Kant, Critique of Pure Reason, London:
MacMillian.
Smith, R. (ed.), 1997, Aristotle, Topics: Books I and VIII with Ex-
cerpts from Related Texts, Oxford: Clarendon Press.
Solmsen, F., 1929, Die Entwicklung der aristotelischen Logik und
Rhetorik, Berlin: Weidmann.
Strawson, G., 1989, The Secret Connexion: Causation, Realism, and
David Hume, Oxford: Clarendon Press.
Strawson, P. F., 1959, Individuals. An Essay in Descriptive Meta-
physics, London: Methuen.
Stroud, B., 2000, The Quest for Reality: Subjectivism and the Meta-
physics of Color, London-New York: Oxford University Press.
Thomason, R., 1982, "Identity and Vagueness", Philosophical Studies
42, 329-332.
Thomason, R., 1984, "Combinations of Tense and Modality", in D.
Gabbay and F. Guenthner (eds.), Handbook of Philosophical
Logic Volume II: Extensions of Classical Logic, Dordrecht: Rei-
del, 135-165.
Thompson, M., 1953, "On Aristotle's Square of Opposition", The
Philosophical Review 62,251-65.
Thomson, J.J., 1971, "The Time of a Killing", Journal of Philosophy

68, 115-132.
Thomson, J.J., 1998, "The Statue and the Clay", Nous 32, 149-173.
Tooley, M., 1977, "Critical Notice of Alvin Plantinga's The Nature of
Necessity", Australasian Journal of Philosophy 55,91-102.
Tricot, J. (ed.), 1950, Aristote, Organon. V: Les Topiques, Paris: Vrin.
Vallentyne, P., 1997, "Intrinsic Properties Defined" , Philosophical
Studies 88, 209-219.
Van Inwagen, P., 1988, "How to Reason about Vague Objects", Philo-
sophical Topics 16, 255-284.
Van Inwagen, P., 1990, Material Beings, Ithaca (NY): Cornell Univer-
sity Press.
Van Inwagen P., 1990a, "Four-dimensional Objects", Nous 24, 245-
255.
Van Inwagen, P., 1998, "Meta-Ontology", Erkenntnis 48, 233-250.
Van Inwagen, P., 2001, Ontology, Identity, and Modality, Cambridge:
Varzi, A. C., 2000, "Mereological Commitments", Dialectica 54, 1-23.
Waitz, T. (ed.), 1844-1846, Aristotelis Organon graece, Leipzig: Teu-
bner.
Wallace, J., 1964, Philosophical Grammar, (Stanford University Ph.D.
thesis), Ann Arbor (MI): University Microfilms Ltd ..
Walton, K., 1993, "Metaphor and Prop Oriented Make-Believe",
European Journal ofPhilosophy 1, 39-57.
Walzer, R.R., and Mingray, J.M. (eds.), 1991, Aristotle Ethica Eude-
mia, Oxford: Clarendon Press.
White, M. G., 1956, Toward a Reunion in Philosophy, Cambridge
(MA): Harvard University Press.
White, N.P.W., 1971, "Aristotle on Sameness and Oneness", The
Philosophical Review, 80, 177-97.
Wians, W. (ed.), 1996, Aristotle's Philosophical Development. Prob-
lems and Prospects, Lanham (Maryland): Rowman & Littlefeld.
Wiggins D., 1976, "Frege's Problem of the Morning Star and the Eve-
ning Star", in M. Schirn (ed.), Studien zu Frege II: Logik und
Sprachphilosophie, Stuttgart-Bad Cannstatt: Frommann-Holz-
boog, 221-255.
Wiggins, D., 1980, Sameness and Substance, Oxford: Blackwell.
Bibliography 495
Wiggins, D., 1994, "The Kant-Frege-Russell view of Existence: to-

ward the rehabilitation of the second level view" in W. Sinnott-
Annstrong, D. Raffman and N. Asher (eds.), Modality, Morality
and Belief, Essays in Honor of Ruth Barcan Marcus, Cambridge:
Cambridge University Press, 93-113
Wiggins, D., 2001, Sameness and Substance Renewed, Cambridge:
Williams, B., 1978, Descartes: The Project of Pure Enquiry, Har-
mondsworth: Penguin.
Williams, B., 1985, Ethics and the Limits of Philosophy, Cambridge:
Harvard University Press.
Williams, S.G., 1996, "Ambiguity and Semantic Theory" in S. Lovi-
bond and S.G. Williams (eds.), Essays for David Wiggins: Iden-
tity, Truth and Value, Oxford: Blackwell, 1996,33-72.
Williamson, T., 1994, Vagueness, London: Routledge.
Williamson, T., 1995, "Definiteness and Knowability", Southern
Journal of Philosophy 33 supplement, 171-191.
Williamson, T., 1996, "The Necessity and Detenninacy of Distinct-
ness", in S. Lovibond and S. Williams (eds.), Essays for David
Wiggins: Identity, Truth and Value, Oxford: Blackwell, 1996, 1-
17.
Williamson, T., 1997, "Imagination, Stipulation and Vagueness", in E.
Villanueva (ed.), Philosophical Issues, 8: Truth, Atascadero (CA):
Ridgeview, 215-228.
Williamson, T., 1999, "On the Structure of Higher-Order Vagueness",
Mind 108, 127-143.
Williamson, T., forthcoming, "Vagueness in Reality", in M. Loux and
D. Zimmennan (eds.), The Oxford Handbook of Metaphysics, Ox-
ford: Oxford University Press.
Winkler, K., 1985, "Skepticism and Antirealism", Mind 94,36-52.
Wisdom, J., 1969, Logical Constructions, New York: Random House.
Wittgenstein, L., 1922, Tractatus Logicus-Philosophicus, London:
Routledge and Kegan Paul.
Wright, C., 1983, Frege's Conception of Numbers as Objects, Aber-
deen: Aberdeen University Press.
Wright, c., 1992, Truth and Objectivity, Cambridge (MA): Harvard
University Press
Yablo, S., 1987, "Identity, Essence, and Indiscemibility", Journal of

Yablo, S., 1988, "Review of The Metaphysics of Modality", The Jour-
nal ofPhilosophy 85, 329-37.
Yablo, S., 1996, "How in the World?", Philosophical Topics 24, 255-
286.
Yablo, S., 1998, "Does Ontology Rest on a Mistake?", Proceedings of
the Aristotelian Society, supp. 72, 229-262.
Yablo, S., 1999, "Intrinsicness", Philosophical Topics 26, 479-505.
Yablo, S., 2000, "A priority and Existence", in P. Boghossian and C.
Peacocke (eds.), New Essays on the a Priori, Oxford: Oxford Uni-
versity Press.
Yablo, S., forthcoming, "Go Figure: A Path Through Fictionalism",
Philosophical Issues.
Zadro, A. (ed.), 1974, Aristotele, I Topici, Napoli: Loffredo.
Zimmerman, D., 1998, "Materialism and Survival" in P. van Inwagen,
and D. Zimmerman (eds.), Metaphysics: The Big Questions, Ox-
ford: Blackwell, 379-86.
Index of Names
Adams, RM. 48n. Boethius 144

Akiba, K. 292 Bonitz, H. 237n.
Alexander 90; 221 Brown, L. 82; 110
Almog, J. 133 Brunschwig, J.B. 245n.
Alston, W. 56; 57; 74n. Burge, T. 139
Andielkovic, M. 276 Burgess, l 61; 74n.
Aquinas, T. 103; 473n. Buridan 138; 144
Aristotle 79-108; 109-126; 154; Burnyeat, N. 107n.
155; 211; 217-238; 239-246;
473n. Carrara, M. 74n.
Armstrong, D. 45n.; 192; 193; Casati, R. 74n.; 75n.
197; 198;205n.; 208; 213 Chalmers, D.l 45n.;
Austin, lL. 80; 81; 93; 98; 110 Charles, D. 106n.
Chisholm, R. 337
Baker, L.R 75n.; 149n. Code, A. 107n.
Baldwin, T. 147 Cresswell, M.l 237n.; 238n.
Barcan-Marcus, R 153; 161n. Crivelli, P. 219; 237n.
Barnes, J. 106; 126; 237n.
Beierwaltes, W. 107n. David, M. 187
Belnap, N.D. 153; 161n. Davidson, D. 55; 199; 205n.
Bencivenga, E. 134; 135; 138 Dawkins, R 321
Bennett, J.F. 74n.; 366n. Della Rocca, M. 71; 75n.
Berkeley, G. 375 Descartes, R. 374
Bernays, J. 249 Donini, P. 107n.
Berti, E. 109-126; 238n. Drewery, A. 205n.
Black,M.257 Dukasse, c.J. 74n.
Blackburn, S. 46n.; 48n. Dummett, M. 9; 12; 45n.; 71;
Bochenski, I.M. 237n. 75n.; 92
Dunn, M.J. 153; 161n.; 337n.

Dworkin R., 45n.; 46n.; 48n. Hacker, P. 74n.
Hale, B. 168; 187n.
Edgington, D. 45n.; 277 Hambruch, E. 237n.
Ellis, B. 205n. Harman, G. 47n.
Evans, G. 131; 133; 149n.; Hawthorne, J. 333; 339n.;
274;275;281;285;287;295; 351n.
305 Heck,R. 284; 285; 291
Hellman, G. 188
Fait, P. 107n. Hilbert, D. 249
Fales, E. 205n.; 207 Hintikka, J. 81; 82; 96; 110;
Feldman, F. 473n. 158; 161n.; 247; 248;
Field, H. 47n.; 170; 171; 173; Hirsh, E. 297
187n. Hobbes, T. 280
Fine, A. 46n. Horgan, T. 74n.
Fine, K. 277; 310; 337n. Horwich, P. 47n.
Forbes, G. 74n.; 205n.; 341- Hughes, G.E. 237n.; 238n.
352;472 Humberstone, L. 331; 337n.
Francescotti, R. 331; 340n. Hume, D. 189; 205; 208; 212;
Frede, M. 98; 99-102; 107n. 374; 378; 448n.; 449n.; 472
110; 238n.
Frege, G. 51; 71; 82; 92; 129; Jackson, F. 45
133; 137; 138; 141; 148n.; 152; Jolley, N. 372; 383n.
165; 177; 185; 187n.; 318 Johnston, M. 75n.
Gaifman, H. 45n.; Kamp,H. 277

Gallois, A. 75n Kant, I. 22; 137; 142; 144;
Garrett, B. 33 8n. 148n.; 154; 155; 158-161; 165;
Geach, P.T. 92; 223; 237n.; 185;380
440; 448n. Kaplan, D. 154; 161n.
Gendler, T.S. 333; 339n.; 351n. Kim, J. 45n.
Gibbard, A. 75n. Kripke, S. 139; 189; 190; 201;
Goldman, A.I. 75n. 205n. 207n.; 265n.; 319; 325;
Gomez-Torrente, M. 187n. 337n.; 339n.; 474n.
Grice, P. 84; 88
Grua, G. 385 Laks, A. 107n.
Gupta, A. 473n. Lambert, K. 134; 135; 138
Index of Names 499
Leibniz G.W.von 48n.; 162n.; Most, G. 107n.

257;260;262;355-368;369- Mulligan, K. 149n.
386;428;430;447
Lejeski, C. 144 Neale, S. 71; 75n.
Lesniewski, S. 142 Nietzsche, F. 134
Lewis, D. 53; 54; 62; 74n.;
270;284; 306; 352n.; 394; 396- Ockham, W. 138; 144; 157
398; 406; 411n.; 415-420; Oliver, A. 74n.
425n.; 426n.; 428; 430; 438; Olson, E. 452; 453; 473n.
447n.; 448n.; 465; 474n. Orenstein, A. 151-161
Lewis, S. 53; 62; Owen, G.E.L. 83; 84; 97; 98;
Locke,J. 358; 376; 409; 473n. 99; 101; 110; 114; 123;
Loemker, L.E. 183n.; 384n. Owens, J. 101; 102; 107n.
Loux, MJ. 74n.
Lowe, J. 72; 73; 75n.; 207-213; Pacius 245n.
270;274;466 Parkinson, G.H.R. 366n.
Lucretius 462 Parsons, T. 60; 74n.; 282; 285;
Parmenides 81; 89; 105; 107
Mackie, P. 337n.-339n. Partridge, C. 74n.
Madigan, A. 90; 91 Patzig, G. 98; 99; 102; 107n.;
Malebranche, N. 374; 375 238n.
Marconi, D. 74n. Peter of Spain 156
Martin, C.B. 205n.; 207 Philo of Alexandria 103
Martino, E. 237n. Pianesi, R. 75n.
Mates, B. 139; 143 Plantinga, A. 205n.; 270
Matthews, G.B. 161 n. Plato 81; 89; 98; 101; 104; 105;
McGee, V. 293 124; 246n.; 429
McGinn, C. 337n. Plutarch 273
McKay, T. 326; 327; 339n.; Price, M. 324
345 Priest, G. 283
McLaughlin, B. 293; 294 Putnam, H. 46n.; 47n.; 188n.;
Meier, H. 237n. 201; 204; 206n.
Meinong, A. 52-57
Melissus, 81 Quine, W.v.O. 57; 61; 79; 127-
Melia, J. 74n. 129; 133; 136; 154;249;250;
Menn, S. 106n. 251;263;264;267;392;397;
Miller, F.D.M. 245n. 41On.; 411n.
Tooley, M. 387; 388; 399; 405;

Reale, G. 238n. 410n.; 424
Remnant, T. 366n. Tricot, J. 88; 92
Rescher, N. 383n.
Richard, M. 287 Vallentyne, P. 329; 330; 339n.
Robertson, T. 323-325; 338n.; van Fraassen l49n.
351n. van Inwagen 64; 74n.; 79; 81;
Rolfes, E.R. 245n. 89; 106n.; 269; 282; 282; 413-
Rorty, R. 46n. 426; 456; 473n.; 474n.
Rosen, G. 45n.; 46n.; 61; 74n. Varzi, A. 41 On.
Rosenkrantz, G. 81
Ross, W.D. 87; 88; 90; 105; Waitz, T. 88; 90
106n.; 237n.; 238n. Wallace, J. 253; 265n.
Russell, B. 37; 51; 55; 57; 61; White, M.G. 53; 58; 237n.
74n.; 82; 187n.; 129; 133; 138; White, N.P.W. 237n.; 245n.
157; 224; 232; 237n.; 238n. Wiggins, D. 72; 110; 127; 130-
Ryle, G. 57; 74n.; 80; 81; 110 134; 137; 148n.; 149n.; 223;
237n.; 238n.; 267-271; 337n.
Williams, B. 48n.; 126
Sainsbury, M. 133; 149n. Williamson, T. 265n.; 305-318
Salmon, N. 131; 274; 275; 296; Winkler, K. 45n.
333; 337n. Wisdom, J. 48n.
Savile, A. 369-386 Woodruff, P. 282
Schiffer, S. 139 Wright, C. 45n.; 92; 168; 187n.
Shields, C. 84; 88; 89; 106n. Wittgenstein, L. 238n.; 257;
Shoemaker, S. 205n.; 207; 457; 258;264;401
471; 474n.
Simons, P. 75n. Yablo, S. 75n.; 330; 337n.-
Smart, J. 205n. 339n; 351n.
Smith, R. 237n.; 245n.
Solmsen, F. 237n. Zadro, A. 88
Strawson, P. 131; 139; 257; Zeno 81
258; 448n. Zimmerman, D. 469; 471;
Stroud, B. 45n. 474n.
Thomason, R. 284; 334

Thomson, J. 75n.; 142
TOPOl LIBRARY
l. A.C. Varzi: An Essay in Universal Semantics. 1999 ISBN 0-7923-5629-2

2. M.E. Vatter: Between Form and Event: Machiavelli's Theory of Political Freedom.
2000 ISBN 0-7923-6533-X
3. E. Bencivenga: Exercises in Constructive Imagination. 2001 ISBN 0-7923-6702-2
4. A. Bottani, M. Carrara and P. Giaretta (eds.): Individuals, Essence and Identity.
Themes of Analytic Metaphysics. 2002 ISBN 1-4020-0548-2
KLUWER ACADEMIC PUBLISHERS - DORDRECHT I BOSTON I LONDON

(Kit Fine (Auth.), Andrea Bottani, Massimiliano CA

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

(Kit Fine (Auth.), Andrea Bottani, Massimiliano CA

Uploaded by

Copyright:

Available Formats

INDIVIDUALS, ESSENCE AND IDENTITY

SPRINGER-SCIENCE+BUSINESS MEDIA, B.V.

ISBN 978-90-481-5988-8 ISBN 978-94-017-1866-0 (eBook)

Printed on acid-free paper

Ali Rights Reserved

Part Three: Identity

PAOLO CRIVELLI / Sameness in Aristotle's Topics 239

Part Four: Time and Persistence

List of Contributors 477

Index of Names 497

What do we do when we do metaphysics? The aim of this introduction

really there is no such a thing. Everything exists in a unique general

In a relativistic and conventionalist mood, Carnap thought that every

lian method. In this book there is an analysis of many philosophical

First of all, we should bear in mind that many analytic philoso-

In order to avoid arbitrariness perhaps it is better to present the pa-

Ontology and Analysis

My aim in this paper is to help lay the conceptual and methodological

Among the most important issues in philosophy are those concerning

genuinely fundamental. The expressivist, for example, may affirm that

appear to avoid the absurdities of skepticism but only by buying in to

Is there any way out of the previous difficulties? Can an intelligible

Another problem concerns the application of the criterion to par-

the given discourse to be true and yet unknowable - or even lacking in

these matters, we need to restore ourselves to a state of innocence in

z is presumably the singular proposition that that couple is married and

possible to acquire an understanding of the reduced sentence on the

identity in terms of the simpler attributes, we do not reductively ac-

4. THE QUIETIST CHALLENGE

order" claims, which merely comment on the metaphysical status of

the B-theorist will want to adopt an Archimedean standpoint in which

contact with an external realm of mathematical facts; and it is often

pable in even the most unproblematic areas of philosophy. Rather I

Metaphysical questions of realism are to tum on questions of a less

We take ground to be an explanatory relation: if the truth that P is

6. SETTLING QUESTIONS OF FACTUALITY

In this and the next section we examine how questions of factualism

the effect that any reasonable disagreement on the factual status of a

(a) any true nonbasic factual proposition is grounded in basic

(b) no nonfactual proposition has a ground consisting entirely of

(d) any nonfactual proposition will contain a nonfactual constitu-

(g) whenever a constituent occurs in a true basic factual proposi-

tion will hold with 'fundamental' (or 'irreducible') in place of 'basic',

then can he not consider the question of when it is appropriate to

in so far as this is plausible, the original proposition should be taken to

should be seen as facing different explanatory demands. For unless

7. SETTLING QUESTIONS OF GROUND

We have shown how to devise a "critical experiment" to test whether

The other main source of evidence is explanatory in character. As

given area, the extended domain might be thought to describe our

The factualist, by contrast, is obliged to make reference to the

about meaning provide an adequate account of the grounds for the

point in favor of an account of our mathematical practice that it allows

We have distinguished between two conceptions of reality - as factual

parative idea of reduction, that should be taken to inform the relevant

On this approach, reduction is to be understood in terms of fundamen-

9. SETTLING WHAT IS FUNDAMENTAL

propositions to be unreal and some to be real. (We do not need to con-

10. THE UNITY OF REALIST METAPHYSICS

We see that questions of factuality and reality are to be answered by

tual. But also if a proposition is factual, then it must be rendered true

When we set ourselves to draw up an inventory of the world - a cata-

objects, or for events, or for whatever categories of entities we counte-