Download as pdf or txt
Download as pdf or txt
You are on page 1of 170

CHANGE, CAUSE AND

CONTRADICTION
Change, Cause and
Contradiction
A Defence of the Tenseless Theory of
Time
Robin Le Poidevin
Lecturer in Philosophy
University of Leeds

Palgrave Macmillan
ISBN 978-1-349-21148-7 ISBN 978-1-349-21146-3 (eBook)
DOI 10.1007/978-1-349-21146-3
© The Scots Philosophical Club 1991
Softcover reprint of the hardcover 1st edition 1991 978-0-333-54286-6
All rights reserved. For information, write:
Scholarly and Reference Division,
St. Martin's Press, Inc., 175 Fifth Avenue,
New York, N.Y. 10010

First published in the United States of America 1991

ISBN 978-0-312-05786-2

Library of Congress Cataloging-in-Publication Data


Le Poidevin, Robin, 1962-
Change, cause, and contradiction: a defence of the tenseless
theory of time I Robin Le Poidevin.
p. em.
Includes bibliographical references and index.
ISBN 978-0-312-05786-2
1. Time. 2. Change. 3. Causation. 4 . Contradiction. 5. Tense
(Logic) I. Title.
BD638.L37 1991
115-dc20 90-19925
CIP
Contents
Acknowledgements ix
Preface X

Introduction 1
Synopsis 8

1 Tense and Change 13


1.1 Two accounts of change 13
1.2 The problem of temporal ontology 17

2 Tense and Contradiction 24


2.1 McTaggart's paradox 24
2.2 Objections and defence 27
2.3 The radical solution 33

3 Temporal Solipsism 36
3.1 Prior's doctrines 36
3.2 'Pathological' cases of diachronic identity 40
3.3 Instants and anti-realism 44
3.4 A paradox of cardinality 48
3.5 The propositional theory of instants 53

4 Temporal Parts 58
4.1 The objections 58
4.2 Understanding temporal parts 60
4.3 The problem of the temporal determinant 67

5 Tenseless Change 76
5.1 Constraints on an account 76
5.2 First attempts 77
5.3 The causal constraint 79

v
vi Contents

6 Causation and Simultaneity 83


6.1 The principle of reciprocity 83
6.2 Historical antecedents of R 84
6.3 The proof of non-simultaneity 86
6.4 Modal properties of R 92
6.5 Time without change 94

7 Causal and Temporal Asymmetry 99


7.1 The priority of time 99
7.2 Causal priority and counterfactuals 103
7.3 Causal priority and probability 110
7.4 Ehring's condition 116

8 A Causal Account of Change 123


8.1 The account 123
8.2 Change and the causal theory of time 129
8.3 Backwards causation 137

Notes 143
Bibliography 148
Index 154
For Kate
Acknowledgements
I have used the quotation from Aldous Huxley's Point Counterpoint,
which appears at the beginning of this book, with kind permission
from Mrs Laura Huxley, Chatto & Windus, and Harper & Row,
Publishers, Inc.
I would also like to thank the editor of Ratio for permission to use
material from my article 'The Principle of Reciprocity and a Proof of
the Non-simultaneity of Cause and Effect'.

ix
Preface
This book was written while I was Gifford Research Fellow in the
University of St Andrews, from 1988 to 1989. While I was there I
was able to try out my ideas on time in seminars, undergraduate
lectures, and a short series of public lectures. I also enjoyed the
company and good advice of my colleagues in the Department of
Logic and Metaphysics: Peter Clark, Bob Hale, Stephen Read,
Roger Squires and Leslie Stevenson. I particularly miss informal
lunchtime discussions with Peter Clark, who unfailingly provided
answers to my naive questions on the philosophy of physics. I
would like to thank the Gifford Lectureship Committee for giving
me the opportunity of pursuing research in such a congenial
environment. I would also like to thank the editors of Macmillan
Studies in Contemporary Philosophy, Alan Millar and Andrew
Brennan, for encouraging me to prepare the book for publication.
I owe a special debt to Murray MacBeath, who read the entire
manuscript in draft and made many detailed and enormously
helpful comments, on the basis of which I was able to attempt the
impossible and prepare a final version. (I should perhaps point out
that he was not able to comment on Chapter 7.4, which had not yet
been written. It may well be that he would have had objections to
it.)
Many of the ideas presented here, especially those of Chapters
2, 3, 4 and 6, are taken from my Cambridge Ph.D. thesis. To Hugh
Mellor, who supervised the thesis from 1985 to its completion, I
owe a very large debt. His numerous writings on time and caus-
ation were - and still are - a constant source of inspiration. Over
the years he has (with characteristic vigour) taught me, not only
the truth about time, but about how good metaphysics can be
done. I also had the privilege of being examined by two other
sources of inspiration: Jeremy Butterfield and Bill Newton-Smith,
who made a number of penetrating criticisms. The influence of
their work on time will be evident throughout the following pages.
I would also like to thank Edward Craig, my first supervisor in
Cambridge, both for his patience with my first attempts at research
and for detailed comments on an earlier version of Chapter 6,
which appeared in Ratio. During my research I was supported,

X
Preface xi

during 1987-88, by a Quatercentenary Research Award from Em-


manuel College. I am very grateful to the Master and Fellows for
this generous assistance.
Moving further back in time, my first tutors at Oxford, David
Charles and Stephen Williams, occupy a special place in the causal
chain leading to this book. I never wrote a single page without
thinking how they would respond to it. Without their constant
encouragement, I would never have embarked upon research in
the first place.
Besides these, I have greatly benefited from talks on a range of
subjects with Matthew Buncombe (especially on backwards and
simultaneous causation), Neil Cooper, Tim Crane (who also gave
technical assistance with the wordprocessing), Dorothy Emmet,
Adrian Moore, W. Z. Souffl~, Richard Swinburne, Nigel Warbur-
ton, and Bernhard Weiss. I have also discussed many of the topics
in this book (though not just those topics) with my wife, Kate
Meakin. I dedicate the book to her.
Introduction
The theme which unites the three elements in the title of this essay
is that of time. In particular, I shall be concerned with two opposing
theories of time, and the consequences each has for an under-
standing of change. 'Change' here includes not only the ordinary
change in the properties of objects - for example the cooling down
of a hot poker withdrawn from the fire - but also the continual
change, real or apparent, in events from being first future, then
present, then past. It is this second kind of change, in particular,
which is the bone of contention between the two theories of time
we shall be discussing. The debate between them has raged for
over eighty years, and it shows no sign of abating. Unfortunately,
however, the participants are rarely at one in the interpretation of
the two theories, so my task in this Introduction is to make clear
what I take to be the essential features of each.
We begin with a distinction of J. E. McTaggart's (1908) between
two ways in which we order events in time. One, which McTag-
gart called the A-series, runs from the distant past to the distant
future. All actual events and objects have a location in this series,
but this location is constantly changing: a present event fust be-
comes past, and then continues to become more distantly past.
Temporally indexical expressions in a token utterance or inscrip-
tion, such as 'now', 'today', 'yesterday', 'next year', 'in three
months' time', serve to locate an event within the A-series, but
(since A-series positions and the reference of temporal indexicals
change) only for the time of the utterance or inscription. Tempor-
ally indexical reference exploits the fact that the token in which the
indexical occurs has itself a temporal location. The other series, the
B-series, is ordered by the relations earlier than, simultaneous with,
and later than. Again, every actual event and object has a location
in the B-series, but there are crucial differences between B-series
and A-series location. Most significantly, the B-series locations of
events do not change. If of two events, p and q, p occurs earlier
than q, then this same relation holds between them from the
viewpoint of any time whatsoever. Expressions such as 'the
Triassic Age', 'the second year of the reign of King John', 'twelve
years before Cromwell's assumption of power', '4 August 1984',
1
2 Change, Cause and Contradiction

serve to locate events in the B-series. That is, they establish the
relative positions of events. In the case of dates, the zero point of
the series is typically defined by some other event (e.g. the official
birth of Christ). B-series reference by some token utterance or
inscription does not depend upon the token's having a location in
the B-series.
Now it is quite uncontroversial that we use both A-series and
B-series expressions in ordering events in time. Indeed, it might
not seem to matter which we use, since the order of events comes
out the same either way, provided we do not make any mistakes.
But the important philosophical question is: which kind of descrip-
tion - A-series or B-series - genuinely reflects the temporal struc-
ture of the world? One is tempted to answer: both. Time consists of
both an A-series and a B-series. However, this view fails to account
for the quite straightforward logical connection between A-series
and B-series statements. If it is true to say that the building of the
Channel Tunnel is present (i.e. is going on), and that the Second
World War is past, then it is also true to say that the Second World
War occupies an earlier position in time than the building of the
Channel Tunnel. The inference is (it seems) a trivial one. Yet how
could it be so, if the two series were completely independent of
each other?
So the A-series and the B-series are not plausibly thought to be
independent of each other: one can be reduced to the other. The
two opposing theories of time I shall be discussing differ over
which direction the reduction should go in. On the first theory, the
B-series is reducible to the A-series. On the second theory, the
A-series is reducible to the B-series. The relevant notion of re-
duction here is that of factual reduction, not of meaning reduction.
That is, putative facts about the A-series are factually reducible to
facts about the B-series if and only if A-series statements, i.e. token
sentences (e.g. 'the mat is now on the cat'), have B-series facts as
truth-conditions (the mat's being on the cat is simultaneous with
that token sentence). Someone who holds this thesis of factual
reducibility does not necessarily want also to maintain that the
meaning of A-series type sentences is to be understood in terms of
B-series type sentences (see Le Poidevin (1988) and (1988b), Ch. 5).
Now I do not need, in this essay, to take a particular stand on the
mutual entailments, or non-entailments, between theses of factual
reduction on the one hand and those of meaning reduction on the
other. But I may as well record my conviction that meaning
Introduction 3

reducibility entails factual reducibility, but not vice versa. I shall


concern myself explicitly only with what I regard as the weaker
thesis, namely factual reduction.
(Stephen Williams has made it clear to me that this characteriz-
ation of Tenseless theory rests upon a particular conception of the
truth-conditions of statements. Someone who saw truth-con-
ditions as tied to the content of what is asserted would naturally
see Tenseless theory - as I have stated it - as committed to
meaning reductionism. Since I want to avoid this consequence,
Tenseless theory could instead be characterized as holding that
tensed facts are supervenient upon tenseless facts.)
It is not easy to find suitable names for these theories which are
usefully descriptive without being floridly metaphorical, 1 but I
hope the following achieve this. I call the first theory Tensed theory,
because it regards tensed (i.e. A-series) sentences as more truly
reflecting the structure of the world. I call the second theory
Tenseless theory, because it states the truth-conditions of temporal
statements in tenseless terms. I shall also find it convenient (fol-
lowing recent tradition, e.g. Mellor (1981)) to call A-series facts
'tensed facts' as a device for identifying a doctrine or account with
Tensed theory. I hope this is an improvement on McTaggart's
colourless terminology. I apologise to anyone who feels strongly
that 'tense' should be applied only to verbs and their inflexions. I
would point out only that the Tensed theorist would hold that
verbal inflexion depends for the most part on tensed facts, not the
other way around.
In order to avoid any ambiguity in what follows, I shall state
what I take to be the essential doctrines of both theories.

Tensed theory is essentially characterized by the following doctrines:


1. All temporal statements- tensed and tenseless- have tensed
(A-series) truth-conditions (Gale (1968); Prior (1957), (1968)). The
A-series, therefore, is connected straightforwardly (as opposed to
via some reductionist analysis) with fact.
2. Any tenseless description of reality is necessarily incomplete
(McTaggart (1908); Dummett (1960); Prior (1959); Lucas (1973)).
This is entailed by 1 but does not entail it, since anyone who
thought that there were both A-series and B-series facts, each
non-reducible to the other, would hold any description which
omitted the A-series facts to be factually incomplete. (In addition,
4 Change, Cause and Contradiction

even a Tenseless theorist could hold that a tenseless description


may be incomplete, but not because it omitted any facts. Such a
description may be epistemically incomplete if tensed beliefs are
essential to any adequate representation of reality (McGinn (1983)).
3. Only the present is real: there are no past and future facts (Prior
(1970)). Past and future statements have truth-value, if at all, only
in virtue of present fact - i.e. such statements have present tense
truth conditions (on this view, of course, 'present fact' is a pleon-
asm). I call this doctrine 'temporal solipsism', by analogy with
ordinary solipsism. It should not be confused with what Wittgens-
tein calls 'solipsism of the present moment', because it does not
deny the existence of presently existing individuals other than
oneself. It does, of course, deny the reality of dead or as-yet-
unborn individuals. Tensed theorists (at least in conversation)
often vigorously oppose this doctrine, which entails 1, 2 and 4. I
shall argue in Chapter 2, however, that anyone who accepts 1 must
also accept 3.
4. The future is ontologically indeterminate, i.e. is just a collection
of (physically circumscribed) possibilities (Broad (1938)). 2 Since this
is entailed by 3, and 3 is (I argue) entailed by 1, the doctrine of an
indeterminate future is essential to Tensed theory. It is in any case
a plausible doctrine, as it has been argued that the notion of the
'closed future' or fatalism is inconsistent with our ability to affect
the future. 4 should be distinguished from physical indeterminism,
which holds that, given the total state of the universe at the
present moment, a number of alternative futures, and alternative
pasts, are physically possible (i.e. compatible with physical law).

Tenseless theory is essentially characterized by the following:


1'. A-series statements have B-series truth-conditions (Russell
(1915); Broad (1921); Smart (1962); Mellor (1981)). For example, the
facts which make a token tensed statement concerning an event or
state of affairs (e.g. 'it was the case that p') true concern the
B-series relation between that event and the state of affairs referred
to (in this case, p obtains at a time earlier than that of the state-
ment). The consequence of this doctrine is a denial of the 'passage
of time' or A-series change.
2'. A-series statements and beliefs reflect our limited epistemic
access to reality (Russell (1915); Butterfield (1984), (1984b)). This
does not entail that such statements and beliefs are not essential
Introduction 5

for successful communication and action - indeed most Tenseless


theorists concede that they are. It is an interesting, and as far as I
know unexplored, question why, if reality is tenseless, creatures
have evolved for whom agency essentially involves tensed rep-
resentation. 2' should not be confused with the view (Grunbaum
(1967), Baker (1979)) that the A-series is mind-dependent.
3'. All times, and hence all the individuals occupying them, are
equally real. This is not to be confused with the obviously problem-
atic doctrine which is sometimes ascribed to Tenseless theory that
all times in some sense co-exist. There must, trivially, be a temporal
separation between different times.
4'. The future is not ontologically indeterminate - though it may
be physically indeterminate, in the senses explained above. Thus
future-tensed statements need be no exception to the law of ex-
cluded middle.

Having characterized the two theories of time, we are now in a


position to see what possible relevance the debate between them
might have for philosophical issues concerning change.
The first, obvious, connection concerns A-series change: the
change undergone by events as they recede into the past. Tensed
theory holds that there is such change, Tenseless theory denies it.
Tenseless theory does not, however, deny the obvious phenom-
enon of the changing truth-value of tensed sentences, but regards
this as a function of the different tenseless truth-conditions of
different token sentences ('It's raining now', uttered at t, will be
true if and only if it is raining at t. Another token of the same type,
uttered at t*, will have different truth-conditions, and so may have
a different truth-value).
The second connection is this: a crucial feature of change is that
it is an essentially temporal phenomenon: it always involves tem-
poral variation in a thing's properties. Put another way, time, not
space, is the dimension of change. Why is this? Any answer to this
must attempt to define the differences between time and space.
Tensed theory has a very clear conception of these differences:
time, but not space, has an A-series. Nothing said so far indicates
what distinguishes time from space for the Tenseless theorist,
however. One of the purposes of this essay is to develop a Tense-
less response to this challenge.
A third issue, related to the above, concerns the nature of the
6 Change, Cause and Contradiction

entity undergoing change. As might be expected, the different


positions concerning the metaphysics of tense are reflected in
different ontologies. This is best illustrated by the contrast between
Quine (Tenseless theory) on the one hand, and Prior (Tensed
theory) on the other. For Quine, the existential quantifier ranges
over all actual objects, past, present and future. His ontology is
that of four-dimensional entities with temporal parts (Quine
(1960). See Lacey (1971)). In contrast, for Prior, first-order quantifi-
cation is restricted to presently existing objects. His ontology is
that of (spatially) three-dimensional entities, sharply distinguished
from events or processes (Prior (1957), (1967), (1968). See Butter-
field (1984). Notoriously, the doctrine of temporal parts conflicts
with certain intuitions concerning change: can the Tenseless
theorist resolve this conflict?
Finally, the question 'could there be time without (ordinary)
change?' has often been posed. Shoemaker (1969) has argued that
it is coherent to suppose that there could be a period of changeless
time. This is not a coherent supposition, however, according to the
viewpoint of temporal reductionism: that times are logical construc-
tions out of events, states of affairs, or propositions about events/
states. What consequences does the outcome of the Tensed/
Tenseless dispute have for reductionism? Among Tensed theorists
Swinburne (1968) 3 rejects reductionism, but Prior (1967) accepts it.
However, it is, I believe, an essential part of doctrine 3 above, and
hence of Tensed theory. In contrast, Tenseless theory is committed
neither to reductionism nor to its converse, absolutism. There
may, however, be advantages for the Tenseless theorist if he
adopts a particular form of reductionism- viz., the causal theory of
time according to which temporal relations are conceived as con-
structions out of causal relations.
A question which is left entirely undecided in the discussion of
the two theories of tense concerns the topology of time: whether it
is, for example, dense or discrete, branching or non-branching,
linear or closed (see Newton-Smith (1980)). Newton-Smith argues
that the debate over rival topologies must be empirical rather than
a priori:

Contrary to a venerable view still fashionable in some quarters,


the investigation of the structure of time is, broadly speaking, an
empirical matter. (Newton-Smith (1980), p. xi.)
Introduction 7

We cannot establish that time of necessity has a given topologi-


cal property by appeal to tense-logical truths. For any such
argument will need to be supported by a proof that such al-
legedly tense-logical truths are indeed tense-logical truths and
this in turn will require a proof that time of logical necessity has
the topological property in question. (Ibid., p. 101.)

In this he finds himself at odds with, for example, Swinburne


(1968). If he is right, then we would not expect the outcome of the
Tensed!fenseless debate, a debate which is largely a priori, to have
direct implications for questions of topology. In addition, different
topologies of time are represented by different temporal logics.
The debate over tense is not so much over rival logics as over the
interpretation of those logics: for example, whether or not the
temporal operators are to be interpreted in terms of A-series
positions.
There is one respect, however, in which questions of topology
do bear upon the theme of this discussion. If Newton-Smith is
right, then models of time which represent it as having non-
standard topologies represent coherent possibilities. In particular,
Newton-Smith has argued that time might have had a closed topo-
logical structure, in which instants of time are related as points on a
circle. The philosophical implications of this fascinating suggestion
have not been fully worked out, and I shall be returning time and
again to the closed time model as a means for showing the limi-
tations of certain arguments - including my own.
As a final prefatory remark: although I do not discuss the
significance of the Special Theory of Relativity for the arguments in
this essay, I wish to make it clear from the outset that the questions
I raise are not rendered incoherent or redundant by the content of
that theory. This point needs to be made, since I talk, for example,
of the fundamental differences between the temporal and spatial
variation in a thing's properties. This somewhat Newtonian vo-
cabulary might appear to conflict with one result of Special Rela-
tivity (on its most common interpretation), namely that the
magnitude of both the temporal and the spatial interval between
particulars may vary from one inertial frame to another. The result,
it seems, is that one cannot speak unqualifiedly of purely spatial
variation or purely temporal variation, as two states of affairs may
be simultaneous in one frame and not in another. What remains
8 Change, Cause and Contradiction

frame-invariant is the space-time interval, s, between particulars,


where this is defined by:

where d is the spatial interval, and t the temporal interval, with


respect to some inertial frame (the units are chosen so that the
velocity of light= 1). When s2 is negative, the space-time interval
is described as time-like; when s2 is zero, the interval is null or
light-like; and when s2 is positive, the interval is space-like. When I
talk of the 'temporal variation' in a thing's properties, therefore, I
want to allow this to be interpreted as the time-like separation
between the components of the variation. Similarly, 'purely spatial
variation' may be interpreted as the space-like separation between
the components of the variation. So when, in Chapter 1, I ask why
change is associated with temporal, not spatial, variation, I can be
taken to be asking why change is associated with time-like, not
space-like, separation. This complicates somewhat the definition
of temporal parts, and I fill in the necessary details in Chapter 4.

SYNOPSIS

The starting point for this essay is an apparent paradox - or group


of paradoxes- concerning change. The initial problem arises from
the commonplace observation that change requires the same object
to exhibit incompatible properties, 4 as for example when a hot
poker drawn out of the fire cools down and eventually becomes
cold. Of course, the apparent contradiction is easily avoided by
introducing time qualifiers: the object exhibits these incompatible
properties at different times. However, this obvious move raises
further questions: what are these 'different times'? On Tensed
theory, they will be A-series positions. On Tenseless theory, they
will be B-series positions. Now this poses a problem for Tenseless
theory, for McTaggart argued that ordinary change entails A-series
change. Further, the Tenseless theorist seems to be committed to a
view of objects as four-dimensional entities with temporal parts,
and so is led into contradiction in accommodating the principle
that the identity of a thing survives change. This problem, which is
avoided by the Tensed theorist, is elaborated in Chapter 1.
The Tensed theorist, however, faces a different contradiction,
Introduction 9

which we owe to McTaggart, concerning the possibility of an


A-series. McTaggart's controversial argument against the A-series,
which has generated more discussion this century than any other
single philosophical problem of time (some writers wanting to
resurrect it, others wanting to give it a decent burial, others - like
myself - taking exception to the numerous obituary notices) is
discussed in detail in Chapter 2. There I provide a reformulation of
it which meets recent objections. I further argue that the only way
out of McTaggart's contradiction for the Tensed theorist is to adopt
A. N. Prior's doctrine of temporal solipsism (see doctrine 3 above).
The elements of temporal solipsism may be found in the pages of
Augustine's Confessions, but its precise articulation involves anum-
ber of logical and ontological doctrines which form the subject of
Chapter 3. I argue there that temporal solipsism faces a number of
difficulties, three of them especially acute, which make it a barely
coherent position. Thus the Tensed theorist finds himself on too
narrow a philosophical ground, and his view, I suggest, should be
rejected.
This puts us in a rather alarming position, for we now have a
defence for premises which, together with an uncontroversial
assumption, entail that time is unreal, as follows:

1. Time entails the possibility of change.


2. Only Tensed theory can account for change.
1, 2- 3. Tensed theory is the correct view of time, if time is
real.
4. Tensed theory is self-contradictory, except on the as-
sumption of temporal solipsism.
5. Temporal solipsism is untenable.
4, s- 6. Tensed theory cannot be the correct view of time.
3, 6- 7. Time is unreal.

The obvious premise to attack, I suggest, is 2, which brings us back


to the Tenseless theorist's difficulties over change. The particular
problem raised for Tenseless theory in Chapter 1 was that it seems
committed to the doctrine that objects have temporal parts, as well
as spatial parts, and this stands in the way of an adequate account
of change. For variation between parts of an object is not genuine
change (it is not the same object which is both F and not F). In
Chapter 4 I ask whether the Tenseless theorist is indeed committed
to temporal parts, and what such a commitment involves. I argue
10 Change, Cause and Contradiction

that Tenseless theory can avoid many of the objections which have
been aimed at the notion of temporal parts.
Nevertheless, the Tenseless theorist must provide an account of
change consistent with a minimal thesis of temporal parts. Some
preliminary analyses of change which meet this requirement are
discussed in Chapter 5, where I consider what further constraints
should be placed on an adequate account of change. At the end of
that chapter I introduce a causal criterion for change; that change
must involve a causal connection between incompatible states of affairs.
Embryonic though this idea is, it faces two major problems:

1. Change takes time, so the incompatible states of affairs con-


stituting the change must be temporally separated. One cannot
just build in the notion of temporal separation to one's account
without making it a trivial matter that time, not space, is the
dimension of change. The causal criterion is supposed to provide a
non-trivial guarantee that change involves temporal separation,
but it cannot (it seems) provide such a guarantee, as there appear
to be cases where cause and effect are simultaneous.
2. An opposing worry is that the notion of temporal priority
must actually be part of any complete analysis - or even criterion -
of causal priority, for otherwise the causal relation would be
merely non-symmetrical rather than asymmetrical. This certainly
excludes cases of simultaneous causation, but then the causal
criterion provides only a trivial explanation of the fact that time is
the dimension of change.

Taking these difficulties in turn: I develop in Chapter 6 an


argument against simultaneous causation, based on the fact that
the possibility of simultaneous causation is incompatible with a
plausible principle of causality. The principle is related to New-
ton's Third Law of Motion, that to every action there is an equal
and opposite reaction. The conclusion of this chapter has a sur-
prising consequence for debates over the possibility of time with-
out change. The second difficulty - that the notion of temporal
priority may have to be built into any account of causal priority - is
tackled in Chapter 7, where I consider a number of accounts of
causal priority, based on counterfactual and probabilistic theories
of causality. I conclude in that chapter that a criterion of causal
asymmetry can be provided which avoids explicit appeal to tem-
poral priority.
Introduction 11

Finally, in Chapter 8, I provide a fuller statement of the causal


account of change, in the course of defending the basic idea against
a variety of objections. I conclude with a consideration of the
prospects for a unified causal account of time and change. Al-
though there are a number of difficulties for the view that temporal
relations are reducible to causal relations, the Tenseless theorist
can still appeal to causality in explaining a number of time's
puzzling features. This leads me in the last pages of the essay to
consider the possibility of a phenomenon which threatens such
explanations: backwards causation.
My aim throughout is to provide, not only an original account of
change, but a vindication of a controversial theory of time.
Tempora mutantur, nos et mutamur in illis.
Anon.

The clock ticked. The moving instant which, according to Sir


Isaac Newton, separates the infinite past from the infinite future
advanced inexorably through the dimension of time. Or, if
Aristotle was right, a little more of the possible was every instant
made real; the present stood still and drew into itself the future,
as a man might suck forever at an unending piece of macaroni.
Every now and then Beatrice actualized a potential yawn.
Aldous Huxley, Point Counter-Point
1
Tense and Change
1.1 TWO ACCOUNTS OF CHANGE

'Times change and we change with them': an undisputed truth,


but one whose real import depends upon the view one takes of the
nature of time. According to Tensed theory its import is that times
change their temporal locations by being first future, then present,
then past, and that this change of times is what enables there to be
change in the objects located within them. In contrast, places do not
change their spatial locations. They can, it is true, be here in one
place and there in another, but then whether they are here or there
is purely a relational matter: it just depends on where we and they
are located with respect to each other. This, for the Tensed theor-
ist, is what distinguishes time from space.
According to the opposing view of time, Tenseless theory, times
do not change in this sense. All that happens is that the world is in
one state at a given time, and in a different state at a later time. On
this second view, the statement that a given event is present (or
past) is made true by that event's being located at a time simul-
taneous with (or earlier than) that of the statement, not by there
being some non-relational property of 'presentness' exhibited
(fleetingly) by either the event or the time at which it occurs. To
this extent, Tenseless theory regards time as analogous to space.
The crucial difference, then, must lie elsewhere.
Now a complete theory of time must also be, or involve, a theory
of change. And in this respect, it has been thought, Tensed theory is
in a much better position than Tenseless theory. Although he did
not put it in such terms, this was one of the central contentions
made by McTaggart in his 'proof of the unreality of time. As he
put it: there cannot be change in the ordinary properties of things
(e.g. a poker losing heat) unless there is also 'A-series change', i.e.
change in the A-series position of objects and events.
Why should anyone hold that ordinary change entails A-series
change? One explanation is very simple, and goes as follows: it is
13
14 Change, Cause and Contradiction

quite uncontroversial that ordinary change must involve time, for


temporal qualifiers are required to resolve an ancient problem of
change. The 'problem', simply, is that change is change in some-
thing: the temperature of a poker, the colour of an apple, the agility
of one's body. So something must survive through the change
(though presumably there are some limitations on this: was it
really Gregor Samsa who woke up one morning to find himself
transformed into a giant beetle?). Yet change also involves the
instantiation of incompatible properties: hot and cold; green and
red; agile and stiff. But how can one and the same thing exhibit
incompatible properties? (Even if one didn't think that change was
change in something, it would still be true that change involved
incompatible states of affairs in a given spatial region.) That is, we
have:

((a is F) & (a is G) & (F and G are incompatible))

which is, prima facie, self-contradictory. The obvious solution is to


qualify the first two conjuncts by temporal terms:

((a is F at t) & (a is G at t*) & -(t = t*))

Now what one takes t and t* to stand for depends upon whether
one opts for Tensed or Tenseless theory. If one opts for Tensed
theory, then t and t* will stand for A-series positions, e.g. four
seconds ago and three seconds ago respectively. In fact, the
Tensed theorist will prefer the more explicit tense logical notation
involving tense operators, the embedded proposition itself being
present tensed:

(It was the case that (a is F) & It is now the case that (a is G))

(In the tense logics of Prior et al., the present tensed operator is not
used, as it is in most cases strictly redundant.) 1 If, in contrast, one
opts for Tenseless theory, then the propositions 'a is F' and 'a is G'
will be taken to be tenseless (i.e. 'is' means 'is at some time'). The
temporal qualifiers 'at t', 'at t*' may be tenseless temporal oper-
ators, or subject qualifiers ('a-at-tis F'), or predicate qualifiers ('a is
F-at-t'). (The implications of a choice among these and other
interpretations are discussed in Chapter 4.)
So the formal description of change will involve temporal quali-
Tense and Change 15

fiers, and for the Tensed theorist these represent A-series pos-
itions, themselves subject to change. So the explanation of why
anyone should hold that ordinary change entails A-series change
is that one plausible view of time, namely Tensed theory, is
committed to this entailment. The Tenseless theorist, of course,
does not hold to such an entailment, but it seems he and the
Tensed theorist can simply agree to differ. (End of explanation.)
This, however, is not the end of the matter. After all, the prima
facie contradiction in 'a is F & a is G' could just as easily be resolved
by the introduction of spatial qualifiers. Of course, this would not
be a description of change, it would be a description of spatial
variation: the variation of a thing's properties from one spatial part
of it to another (the poker may be hot at one end and cold at the
other). But why isn't this change? Why must change involve tem-
poral variation - the variation in a thing's properties from one time
to another? Now in the explanation of this the Tensed theorist may
have an advantage over the Tenseless theorist, for the Tensed
theorist can appeal to the existence of the A-series. Consider an
event such as sand running through an hour-glass. All non-
temporal propositions about the event, such as that it involves the
motion of sand, that the sand is at a certain temperature, are, if
true, true for all times. So no non-temporal fact about the event
changes. What of the temporal facts? If temporal facts are just
B-series facts, then no temporal fact about the event changes
either. For if the event is, when it is occurring, simultaneous with,
say, the smoking of a distant chimney, then the proposition that
these two events are simultaneous with each other is true for all
times. The only changes in temporal facts concerning the event are
changes in A-series facts: the event successively becomes present,
then past, and then more and more distantly past. For there to be
genuine change, then, there must be change in A-series facts. This
is why change involves temporal variation.
To this, it could be objected: 'we can distinguish between change
in things and change in events. There can be change in things
because things can gain and lose properties. But there cannot be
change in events, for these have their properties essentially. In-
deed, events just are changes. So it is just a mistake to say that
change in things involves change in events. If there were change in
events, then this would have to be A-series change. But there isn't.
There is only change in things, and this is adequately represented
by saying that a thing is F at t and G at t*.' This, in essence, is
16 Change, Cause and Contradiction

Russell's view of change. McTaggart objects to this account on the


grounds that temporal variation without A-series change is struc-
turally parallel to purely spatial variation (he compares the B-series
to the series of positions along the Greenwich meridian - see
McTaggart (1927), pp. 14-15). We do not think of mere spatial
variation as change, so what distinguishes time from space in this
respect? The answer must be that time has an A-series, and space
has not- or rather nothing analogous to an A-series. We can, of
course, construct a series of positions running from 'here' to
'there', but there is no ontological significance in such a means of
distinguishing between the positions.
We might ask: 'why is it so important to distinguish temporal
change from spatial variation? After all, the Tenseless theorist
could always insist that the intuitive difference lies only in our
natural prejudice in favour of Tensed theory. Once we see that
Tenseless theory is the correct view of time, we shall come to see
our insistence that only temporal variation is change is merely a
linguistic convention.' The answer to this, I believe, is that it is
important to distinguish quite fundamentally between temporal
and spatial variation because genuine changes are events, and an
event ontology is crucial for certain philosophical positions, such
as relationism (see Chapter 8.2). It is essential that changes, and
hence events, should not include purely spatial variation, for
otherwise such theories will be trivially false. If we cannot give an
explanation of how time enters with genuine change, then there is
an important gap in our theory.
An adequate account of change must be one that excludes, not
only purely spatial variation, but also what Geach (1969) calls
'mere Cambridge change'. As Geach defines it, an object 0 Cambridge
changes if and only if two sentences of the form '0 is F at tt' and '0
is F at t 2 ' differ in truth value. This definition includes genuine
changes, such as a house burning down ('The house is intact at tt'
is true and 'The house is intact at t 2 ' false), but it is wide enough
also to include pseudo-changes such as the price of that particular
house going up. Now going up in price, unlike going up in flames,
does not involve any change in the genuine properties of the
object. Such pseudo-change is called mere Cambridge change, but
from now on I shall, with apologies to Geach, follow current
practice and use 'Cambridge change' to refer just to this pseudo-
change. Now it would be convenient if the crucial feature of the
favoured account of change which excluded spatial variation also
Tense and Change 17

excluded Cambridge change, or at least if the means of excluding


each turned out to be closely related. This would be a parsi-
monious account, and it is worth remarking at this point that the
Tensed theorist's means of dealing with spatial variation doesn't
exclude all Cambridge variation.
For the Tensed theorist, purely spatial variation doesn't count as
change because it doesn't involve a corresponding real change in
locations. This criterion certainly excludes spatial analogues of
Cambridge change, such as one end of a cricket bat's being nearer
the stumps than the other end. But it doesn't exclude temporal
Cambridge change, like the case of the house going up in price, for
this 'change' necessarily involves (if Tensed theory is true) the
required change of (A-series) location: that is, the price was £x and
is now not £x.
I shall suggest in this essay that both McTaggart's challenge, and
the problem of spatial 'change', can be met by the Tenseless
theorist. But first, I want to indicate in more detail the nature of the
problem for Tenseless theory, which goes somewhat deeper than
McTaggart's discussion would suggest. That is, Tenseless theory
seems unable to resolve an apparent contradiction in our intuitive
conception of change.

1.2 THE PROBLEM OF TEMPORAL ONTOLOGY

The problem of change is also a problem of ontology. We have to


reconcile the following three intuitions:

(a) Objects, including people, plants, and animals, change, in at


least the minimal sense of exhibiting different properties at
different times.

This seems to entail:


(b) Objects last through, or are extended in, time: e.g. Bertrand
Russell lived from 1872 to 1970. (This applies even to virtually
'instantaneous' objects, like a minute sample of the highly
unstable element number 103, Lawrencium.)

However:
(c) If we encounter an object at a moment during its lifetime, then
we encounter the object itself, not just a stage of it or its
18 Change, Cause and Contradiction

life-history. In contrast, if we encounter a very large object,


such as an island (by, say, landing upon one of its beaches), we
only encounter a spatial part of it.

We may formulate this last intuition by saying that, although


objects are both temporally and spatially extended, they are wholly
located at each of the many times which constitute their temporal
extension. But they are not wholly located at each of the many
places which constitute their spatial extension. This feature also
distinguishes objects from processes. When we encounter a long
process, such as the eruption of a volcano, we do not encounter, at
a moment, the whole process, but only a temporal part of it. We
could formulate this by saying that processes are not wholly
located at each of the many times which constitute their temporal
extension.
Note a possible source of contradiction here. How can an object
be both wholly located at (which presumably means 'has its entire
temporal extension located at') a given time, as (c) apparently
requires, and extended in time, as (b) requires? Now Tensed theory
can resolve this tension between (b) and (c) by insisting that objects
change their temporal locations, and so have temporal extension in
the sense of having occupied different times in the past from those
they occupy now, and from those they wW occupy in the future.
The object is still just a three-dimensional one (as it has spatial
extension), but it 'moves through' time. This view, of course, treats
temporal extension as being radically unlike spatial extension, and
does so as a consequence of its view of the disanalogies between
time and space. A thing has spatial extension in virtue of occupy-
ing, at any one time, a region of space- a series of places which do
not vary in respect of their hereness in any non-relational sense. An
object cannot therefore be wholly located at each of the many
places which constitute its spatial extension. Rather, it has spatial
parts, whose possession of different properties from each other
constitutes the object's spatial variation.
In contrast, Tenseless theory, it seems, cannot resolve the con-
flict between (b) and (c). An object wholly located at a time is,
according to Tenseless theory, an object with no temporal exten-
sion: a literally instantaneous object. For nothing changes its lo-
cation in tenseless time. That is, an object has all its temporal
locations for all times; there is nothing extra (like the presentness
of its locations) to change. So (c) conflicts with (b), which requires
Tense and Change 19

objects to have more than one temporal location. Further,


(c) conflicts with (a), for instantaneous objects cannot change (to
change is, inter alia, to occupy more than one temporal location).
The tension between (b) and (c) and between (a) and (c) is made
doubly problematic by the fact that (a) seems to require both (b)
and (c). For (a), which asserts that objects have different properties
at different times, surely entails that objects must be locctted at
different times, and so extended; but also that the whole object, not
just a part of it, is located at those different times.
One response to these paradoxes is to give up (c)- or at least one
interpretation of it. If most objects are extended in time, then on
Tenseless theory they have a number of temporal locations, and
have them always. But since objects don't change their tenseless
temporal locations, it is nonsense to say that the whole object is
located at each one of the times that constitute its temporal exten-
sion. Rather, its temporal parts are wholly located at each one of
those times, just as the spatial parts of an object are wholly located
at each one of the places that constitute the object's spatial location
at a time.
Faced with a choice between instantaneous objects on the one
hand and temporal parts on the other, Tenseless theorists have
understandably chosen this response as the less unpalatable of the
two options. But then an intuitive disanalogy between time and
space is lost. Change becomes no more than temporal variation: an
object's exhibiting incompatible properties in its various temporal
parts, and this is structurally parallel to purely spatial variation: an
object's exhibiting incompatible properties in its various spatial
parts. Why then do we count only temporal variation (a poker's
growing colder over time) and not purely spatial variation (a
poker's being hot at one end and cold at another) as change? Or, to
put it in terms familiar from Special Relativity: why is change
associated with time-like separation? It would seem on the face of
things that the Tenseless theorist has no substantial answer to
give. For him it is perhaps just a linguistic convention that only
temporal variation is called change. On the other hand, if we insist
upon a conception of change in which the whole of an object (not
just a part of it) first has a property and then loses it, then the
Tenseless theorist seems unable to accommodate change within
his scheme of things at all. For neither the four-dimensional object,
nor its temporal parts first exhibit and then lose a property: they
have their properties for the whole of their temporal extension
20 Change, Cause and Contradiction

(which in the case of temporal parts is zero). To elaborate; on the


Tenseless theory of time, a temporally extended object cannot be
wholly present at the time at which it exhibits a given property, so
this requirement of change is not fulfilled. However, neither do the
temporal parts of the object change, for they do not exhibit different
properties at different times: they each occupy only one time.
In sum, the case against Tenseless theory is as follows: on the
assumptions of that theory, (b) implies the falsehood of both (a)
and (c). Further, if the 'objects' of (a) and (b) include temporal
parts, then (a) is false and (b) self-contradictory.
Given these difficulties, it would be unfortunate if it turned out
that the only way we can intelligibly conceive of the persistence
through time of objects is by conceiving them as having temporal
parts. Yet just this is the conclusion of an argument of David
Lewis's (Lewis (1986}, pp. 198-204). Lewis's primary interest, in
fact, concerns the modal problem of 'trans-world identity', but he
takes the results of his discussion to generalise to the temporal
case. Let's begin, then, with the modal case.
To fill in the essential background, Lewis is a modal realist: that is,
he takes other, non-actual, possible worlds to exist as concrete
individuals2 - i.e. as much a part of reality as the actual world. This
position, he thinks, is inconsistent with the ordinary view that both
(i) there is genuine trans-world identity: that the same thing can
exist in different worlds; and (ii) not all intrinsic properties3 of
objects are essential to them. He argues for this surprising con-
clusion as follows: if an individual exists at (in a concrete sense}4
more than one world, then these worlds overlap - they have a part
in common. In which case, it would be self-contradictory to say
that the individual is e.g. fat in one world but thin in the other. The
fact that he is part of two different worlds is irrelevant: he cannot
exhibit incompatible properties. To use Lewis's illustration:

You might as well say that the shared hand of the Siamese twins
has five fingers as Ted's left hand, but it has six fingers as Ned's
right hand! This is double-talk and contradiction. Here is the
hand. Never mind what else it is part of. How many fingers does
it have? What shape is it? (Lewis (1986}, pp. 199-200.)

(There is no problem in an individual's having different extrinsic


properties from world to world. For example, Ouspensky may be
universally revered in this world, but not in some other possible
Tense and Change 21

world, w, for there exists in w an individual (who does not exist in


the actual world) who holds Ouspensky in contempt. For different
reasons, a hand might be famous as Ted's left hand- for Ted himself
is famous- but unknown as Ned's right hand.)
But if a thing's non-essential properties are precisely those
which it has in one world but lacks in another, then no object can
have any non-essential intrinsic properties. Clearly, we don't want
to say this, so we must reject overlap between worlds and hence
trans-world identity. The correct view, says Lewis, is to represent
objects as world-bound, but also as having qualitatively similar
counterparts in other possible worlds. These counterparts enter
into the truth-conditions of modal statements about actual individ-
uals. Thus 'Ouspensky might have been a painter' is true if and
only if there exists in some other world an Ouspensky-counterpart
who is a painter.
The temporal parallel is straightforward. We believe (i) that
objects change- i.e. exhibit different intrinsic properties at differ-
ent times, and (precisely because of (i)), (ii) that objects persist
through time by being wholly present at each time constituting
their temporal extension. But if, following intuition, we are realists
about other times, then (i) and (ii) are inconsistent with one
another. For if an individual exists at two times, then these times
overlap: they have a part in common. Consequently, it makes no
sense to say that this individual is F at one time but not F at
another, for this is just to say that the same part is both F and not F.
Therefore, change is impossible. The paradox, however, vanishes
once we realize that (ii) should be rejected, and that we should
conceive of an object's existing at different times in terms of its
having different temporal parts existing at different times. Such
parts, of course, are time-bound. The parallel with modality is not
precise, however, as what makes two temporal parts parts of the
same object will be more than just qualitative similarity (in fact,
they may not be similar at all).
Recently, Graeme Forbes (1987) has objected to Lewis's argu-
ment on the grounds that it relies on the question-begging premise
that worlds are just mereological sums of possible individuals
existing at those worlds (or, in the temporal case, times are just
mereological sums of temporal individuals existing at those times),
but this is only plausible if one already accepts that individuals are
world-bound (or, time-bound). For otherwise, the worlds (or
times) cannot be said to have a part in common. The proponent of
22 Change, Cause and Contradiction

trans-world (or diachronic) identity will of course deny this prem-


ise.
Now Lewis does regard worlds as mereological sums of possible
individuals ((1986), p. 69), although he says nothing about the
temporal counterpart, which has the (possibly unwelcome) conse-
quence that qualitatively indistinguishable times are numerically
identical. 5 But does Lewis need this premise? He certainly does if he
is to talk of overlap between worlds (or times), but he could instead
talk directly of the mereological sum of the occupants of each
world (or time) while leaving it open as to whether worlds (or
times) are to be identified with these sums or not. LetS be the sum
of individuals existing at world w, and S* be the sum of individuals
existing at world w*. Then, if an individual occupies both w and
w*, S and S* overlap ... etc.
Thus reformulated, the argument seems to me to be successful
in the modal case, but needs further elaboration in the temporal
.case. Suppose an object 0 to persist from timet tot* inclusive. Let
St be the mereological sum of individuals existing at t and St" be the
sum of individuals existing at t*. Then it would seem that St and St"
overlap, since they share a part, 0, in common. But whether this is
indeed a genuine case of overlap depends upon which theory of
time is correct. On Tensed theory, 0 changes its position in time, so
which sum 0 is part of is also something which changes. To put it
another way, Tensed theory will interpret St as the sum of individ-
uals presently existing at t. Consequently, Stand St" will not after all
overlap, for it is never true to say both that 0 presently exists at t
and that 0 presently exists at t*. Indeed, the existence of one sum is
actually incompatible with the existence of the other. Therefore
there is no contradiction, for Tensed theory, in saying that 0 is F at
t and not F at t*. (There is no corresponding move in the modal
case simply because the modal realist has no reason to assert that
'actual' and 'possible' mark non-relational distinctions, which
would be the modal analogue of the A-series.)
In contrast, Tenseless theory will interpret St as the set of
individuals tenselessly (i.e. at some time or other) existing at t. In
this case, 0 is indeed a part of both sums, which therefore overlap.
The consequence then goes through that 0 must exhibit all its
intrinsic properties for all times. Since, however, the Tenseless
theorist does not want to exclude temporal variation of intrinsic
properties, he must concede that it is not 0, but rather a temporal
part of 0 which is part of St and another temporal part of 0 which
Tense and Change 23

is part of Sf'. So Lewis's argument has simply led us back to our


earlier conclusion that temporal parts are only inevitable for the
Tenseless theorist.
It seems, then, that only Tensed theory can account for genuine
change. Unfortunately, Tensed theory cannot be the correct view
of time, as I aim to show in the next two chapters.
2
Tense and Contradiction
2.1 McTAGGART'S PARADOX

McTaggart thought that, for time to be real, there must be a


non-relational distinction between past, present, and future: 1 that
is, that there must be a real A-series, non-reducible to B-series
relations between events. However, he also thought that the A-
series was self-contradictory, and this conjunction of premises led
him inevitably (and entirely validly) to the conclusion that time is
unreal. In this chapter, we shall be concerned with this second
premise: that the A-series is self-contradictory. If it is, of course,
Tensed theory fails.
The second premise is really the conclusion of a sub-argument,
which goes as follows (let 'P' represent 'It was the case that'; 'N'
represent 'It is the case that'; and 'F', 'It will be the case that': )2
1. The A-series positions of events are mutually exclusive. Or, as
we might put it, the tensed operators are mutually exclusive:
Pp ~ -(Np v Fp)
Np ~ -(Pp v Fp)
Fp ~ -(Np v Fp)

2. Every event exhibits all A-series positions. Or, to put it in


different terms: each tense operator is applicable to every true
proposition:

(Pp & Np & Fp)

1 and 2 are clearly inconsistent with each other. But must we


accept them both? 1, at least, is surely uncontroversial: if an event
is present, it is not also past or future. In saying that the tense
operators are incompatible, I do not of course wish to rule out the
iteration of tenses: that is, where one tense operator occurs within
the scope of another, e.g. 'PFp': it was the case that (it will be the
24
Tense and Contradiction 25

case that (p)). Premise 2 is apparently guaranteed by the continual


change in A-series positions undergone by every event, object,
and state of affairs. But if we read the conjunctive proposition as
present tensed (as the embedded proposition p is present tensed)
then surely 2 is false. At no time is it the case that some event is
past, present, and future. Rather, it has these different A-series
positions at different times. So any attempt to remove the contra-
diction must of course qualify the terms in 2, by further temporal
terms. The obvious way to do this is to introduce tenseless terms-
i.e. ones which will not change their temporal reference with
context. An example would be '4.40 p.m., 5 July, 1975 AD'. Let 't',
't*' and 't**' stand for such expressions. Then we may alter 2 to
something like:

2'. (Pp at t & Np at t* & Fp at t** & -(t = t* v t = t** v t* = t**))

In doing this, however, we relativize the tensed operators to


particular tenseless times (a more perspicuous notation would
have 'at t' as temporal operators occurring before the tense oper-
ators). In other words, we have made the A-series only a relational
series: the coronation of the Queen is not past absolutely, but past
with respect to, e.g. 1989. And this is what McTaggart wishes to
deny. If time is real, says McTaggart, there must be a non-
relational A-series, whose positions are continually changing, and
this is reflected in the changing truth-value of tensed expressions.
The truth-value of 2' will not change over time, consequently it is
an inadequate expression of the genuine temporal facts.
We can recast McTaggart's point in the terms of the Introduc-
tion. If Tensed theory is correct, then the truth-conditions of all
temporal statements or propositions must be given in tensed, not
tenseless, terms. So the truth-conditions of 2', which is a tenseless
proposition, will be tensed. Consequently, in seeking to remove
the contradiction of the conjunction of 1 and 2, we should qualify
the terms in 2 with tensed terms. And this is surely a very natural
thing to do. We say of a present event that it was future, is present,
and will be past. Symbolically:

2". (FPp & NNp & PFp)

The contradiction is certainly removed, but we have lost some-


thing in removing it. 2, although apparently self-contradictory, at
26 Change, Cause and Contradiction

least expressed the fact that the A-series positions of events, etc.,
change. 2", in contrast, fails to express this. We can think of a spatial
analogue of 2": each one of us is 'here' with respect to our immedi-
ate vicinity, but 'there' with respect to some place outside this
vicinity (we are 'Here'Here and 'There'There). 3 This simply reflects
the fact that there are different spatial perspectives. It does not
suggest that the spatial positions of events change (which, in the
case of instantaneous events at least, they do not). So if we are to
accommodate the inexorable change in A-series positions (the
ever-shifting present, the receding past, the looming future), we
must supplement 2" with:

3. (FPp & NPp & PNp)

And this contradicts 2" (if we accept 1), so we have not resolved the
underlying contradiction in the A-series.
We needn't stop there, of course (although it would save us
wasted effort if we did), for we can represent the fact that 2" is true
now, whereas 3 will be true:

3'. (FFPp & FNPp & FPNp)

But we cannot rest content with just 2' and 3', any more than we
could have rested with just 2'. Again, because of the inevitable
change in the A-series, it must also be the case that:

4. (NFPp & NNPp & NPNp)

which contradicts 3'. And so on ad infinitum. The constant iteration


of tenses, which is all the Tensed theorist has at his disposal, is
useless in dealing with the contradiction. The notion of a non-
relational A-series, then, is self-contradictory.
We are left with a most unfortunate position. For first, the
conclusion of Chapter 1 was that only the Tensed theorist seemed
to be able to account for genuine change in the properties of
things. But second, the Tensed theorist must hold, as an essential
part of his position, that there is a non-relational A-series, and
such a series we have just shown to be self-contradictory. There-
fore, there is no such A-series; therefore there is no change.
McTaggart realized that it was a short step from this to the con-
clusion that time itself is unreal, but he relied on the premise that
Tense and Contradiction 27

time entails (ordinary) change. Now this is a controversial premise,


and it has been denied by Shoemaker (1969) and, more recently,
Newton-Smith (1980), who argue that it is not incoherent to talk of
a period of time without any ordinary change in the properties of
things (see also Chapter 6.5). It is, however, sufficient for McTag-
gart's purposes that time entails the possibility of change. It is true,
even of a world in which there is a temporal 'vacuum' that there
still might have been change during that period. The above proof
shows, not just that change does not occur, but that it could not
possibly occur. Hence, time is unreal.
I said above that premise 1, that different A-series operators are
mutually exclusive, was uncontroversial. For completeness I will
mention, only to set aside, just one circumstance in which this
inoffensive proposition turns out to be false. On the hypothesis of
closed time, in which moments of time are related as points on a
circle, 'past' 'present' and 'future' are not mutually exclusive. On
the contrary, on that hypothesis, any event that is past is also
future, and any present event is both past and future. This can
easily be demonstrated as follows. A tense logic for closed time
may be obtained by adding to Lemmon's minimal tense logic ~
the following axioms (the present-tense operator is omitted here):

FFp - Fp; Gp - p; Gp - Hp

'G' is interpreted 'It always will be the case that', and 'H', 'It
always has been the case that'. From this system (known as PCr) 4
we can prove the following theorems:

Pp - Fp; Fp - Pp; p - Fp; p - Pp

However, this is no help for the Tensed theorist, for 'It is now the
case that p' and 'It is not now the case that p' remain, even on the
hypothesis of closed time, inconsistent, and we can recast McTag-
gart's argument using these expressions instead of the 'F' and 'P'
operators.

2.2 OBJECTIONS AND DEFENCE

Tensed theorists, in the eighty years since its first appearance,


have brought a number of objections to the terms of McTaggart's
28 Change, Cause and Contradiction

argument, which, as Dummett (1960) remarks, gives the appear-


ance of a 'trivial sophism'. Wishing to avoid superfluous repetition
in an area already swamped with a voluminous literature, I shall
limit myself to rebutting some recent objections.
Recall that the obvious reply to McTaggart's premise 2, that
every event exhibited incompatible A-series positions, was that
these positions are not exhibited at the same time. A present event,
for example, was future, is present, and will be past. This obvious
reply won't do, it was argued, for the qualifying terms must (on
Tensed theory) themselves be tensed, so we simply have a series of
compound expressions involving iterated tenses (PFp & NFp etc.),
from which the contradiction is generated. Now in a recent paper
E. J. Lowe (1987) has objected to this iteration of tenses. He denies
that, e.g. 'FNp' should be interpreted as 'It will be the case that (It
is the case that p)', or 'In the future, pis present', but rather 'there
will be a time when the sentence "[p] is present" is true (expresses
a true statement)'. (Lowe (1987), p. 64.) To employ compound
tenses, as McTaggart and his defenders do, is to commit what
Lowe calls the 'indexical fallacy'. On the other hand, to remove
compound tenses is to remove the contradiction they generate, so
the obvious reply to McTaggart that no event is simultaneously past,
present, and future, effectively dissolves the paradox once and for
all.
Lowe's manoeuvre, however, is completely ineffective, for his
(somewhat suspect) meta-linguistic reconstrual of compound
tenses is subject to a regress no less vicious than McTaggart's. For
we can restate the argument in meta-linguistic terms. Let 'NT', 'PT'
and 'FT' stand for 'the assertion that " ... " is now true', 'the
assertion that" ... "was true', and 'the assertion that" ... "will
be true' respectively, where the dots stand for the embedded
sentence. The argument then goes as follows:

5. NT 'Np' -+ -(PT 'Np' v FT 'Np')


6. (NT 'Np' & PT 'Np' & FT 'Np')
The inconsistency of the conjunction of 5 and 6 is not removed by
replacing 6 by:
6'. (NT'NT''Np'" & FT'PT''Np"' & PT'FT''Np"')
for it is also true that
7. (NT'FT''Np"' & PT'FT''Np"' & FT'NT"Np"')
and so on.
Tense and Contradiction 29

The contradiction here is generated among meta-linguistic state-


ments, which ascribe the same truth value to incompatible as-
sertions, rather than incompatible A-series positions to the same
event or state of affairs. And, as in McTaggart's original paradox,
the contradiction can only be removed by reverting to tenseless
meta-linguistic statements. So this is just as effective a proof of the
unreality of the A-series as McTaggart's. 5
A quite different objection concerns McTaggart's view of events.
His picture of events receding into the past is strongly suggestive
of the A-series moving along a static B-series (he himself points
this out in a footnote in The Nature of Existence (1927), pp. 10-11).
Moreover, it heightens the problem of events having incompatible
A-series positions. Now since Davidson's work on the analysis of
action sentences, the theory of an ontology of concrete events has
been seen to be a plausible one. But if instead we reconstrue talk
about events as talk about the properties of things, then there is
nothing which could be said to have incompatible properties. This
treatment of McTaggart's paradox has recently been recommended
by A. B. Levison (1987):

McTaggart's argument against the reality of time's flow presup-


poses that events are atemporal or perhaps sempitemal entities
which have various 'temporal properties'. . . . such as being
past, present, and future. Similarly, he presupposed that events
occupy pre-existing 'temporal positions' successively. In other
words, McTaggart implicitly represents the events of history as
existing forever and changing temporal positions all the
while ....

since the tensed view of time does not require this assumption,
McTaggart has not established that the tensed view of time is
impossible. (Levison (1987), pp. 349-50.)

The view of events that Levison adopts is Prior's. In 'Changes in


Events and Changes in Things' (1962) Prior suggests that we
paraphrase sentences like:

(a) My falling out of a punt has receded 6 years into the past.

as:
30 Change, Cause and Contradiction

(b) It is now 6 years since it was the case that I am falling out of a
punt.

Now as Levison points out, (b) could not count as a paraphrase of


(a), for (a) implies that there is one and one only occasion referred
to, whereas (b) could be made true by any number of failings out of
a punt. So it would be better to say that (a) is, strictly, false, and
that (b) and similar sentences articulate more precisely what we
mean by 'time passes'. But such reconstructions of event-sentences
leave nothing to recede in time, so nothing has incompatible
tensed properties. To that extent, suggests Levison, talk of an
A-series is seriously misleading. All a Tensed theorist needs to
assert is that tensed sentences reflect a structural feature of the
world in a way that tenseless temporal sentences do not.
This is not an adequate reply. McTaggart's argument does not
essentially rely on any ontology of events, or indeed any 'pre-
existing' temporal positions (he actually considers the move of
replacing events with facts, which is what Prior and Levison rec-
ommend). In my reconstruction of McTaggart's argument above,
only propositions and tense operators are involved: the problem
here is how incompatible propositions can have the same truth-
value, not how the same events can have incompatible properties.
In any case, Levison's suggestion that in removing talk of events
there is no longer need for a Tensed theorist to talk of an A-series is
profoundly mistaken: to talk of a non-relational distinction be-
tween past, present, and future is to talk of an A-series - if not of
events, then of times.
If, then, there is a fallacy, or unwarranted assumption, in the
paradox, it must lie elsewhere.
I would now like briefly to consider a complication which the
Tensed theorist might introduce in order to remove the contradic-
tion. This is suggested by George Schlesinger's paper, 'How Time
Flies' (1982). His proposal is that time is two-dimensional, but,
unlike the dimensions of space, these two temporal dimensions
have different properties. The first temporal dimension has a
non-relational A-series, but the second- the 'meta-temporal' di-
mension- does not: it has a B-series only (and hence a purely
relational A-series). This theory, it might be suggested, provides a
literal construal of the intuitive notion of the passage of time. Even
the rate of passage can be given a value, namely the ratio of
difference in time to difference in meta-time. By relativizing pos-
Tense and Contradiction 31

itions in the temporal A-series to positions in the meta-temporal


B-series, Schlesinger seeks to avoid McTaggart's contradiction. But
the manoeuvre (which, after all, has little else to recommend it:
why do we not see time as having two dimensions?) does not
succeed. If an event is present in time only in relation to a tenseless
meta-temporal point, then its presentness is not non-relational. There
is no privileged meta-temporal point - no point which could
(except relative to some point) be called present. Consequently,
there is no privileged temporal point. A non-relational A-series
could only be restored if both dimensions had an A-series. In which
case, we would have two self-contradictory postulates instead of
one. The Tensed theorist, then, will find little comfort in Schle-
singer's model.
In Dummett's (1960) presentation of McTaggart's argument- a
paper which did much to revive interest in the paradox - the
remark is made that 'McTaggart's argument shows that we must
abandon our prejudice that there must be a complete description of
reality.' (Dummett (1960), p. 357.) Now this is an obscure remark. It
is certainly one of the doctrines of Tensed theory that any tenseless
description of reality will be (factually, as well as epistemically)
incomplete. But this is not Dummett's point, which was made in
the context of tensed descriptions. These are incomplete in the
sense that we cannot, by means of them, definitively pick out a
particular time except on a single occasion of use. But then this is
the very nature of tensed descriptions: different tokens will pick
out different times, and so may differ in truth-value. Unless tensed
tokens had this property, they would not adequately reflect the
supposed temporal facts of A-series change. Each token tensed
description will, for the Tensed theorist, give a complete description
of reality for the time of use. It cannot give a true description of the
whole of reality- i.e. temporally extended reality - and is in that
sense incomplete, but why should we expect it to be complete in
this sense?
Perhaps the point is this: if a given description of reality is
complete, then we would expect each token of that description to
have the same truth-value, as it should not be relevant to the
truth-value of such a token where, in reality, the token occurs. It is
in these terms that Mellor (1981) argues against the reality of the
A-series: if there were tensed facts (i.e. a non-relational A-series),
then these would provide the non-token-reflexive truth-conditions
of tensed judgements. They must be non-token-reflexive, for any
32 Change, Cause and Contradiction

truth-condition which involved a specific token of a tensed type


judgement could not be the truth-condition for all tokens of that
type. But there can be no such truth-conditions for tensed types,
for different tokens of any tensed type judgement ('it's now rain-
ing') will typically differ in truth-value. The idea is that facts are
facts regardless of one's position in reality - and this must include
temporally extended reality. Now doesn't this beg the question
against the A-series? Perhaps reality is not temporally extended, in
which case, different (real) tokens of a tensed type would have the
same truth-value, regardless of their position in reality.
Let me develop this point further. Suppose we say, in the words
of the Tractatus, that the world (i.e. reality) is the totality of facts.
Facts are what make particular statements true (or false). What
should we include in any complete list of facts? We should include
the fact that such-and-such a state of affairs is obtaining in such-
and-such a place, no matter how distant that state of affairs is.
What is occurring on Sirius is just as much a fact as what is
happening in Britain (though we may have to allow that this is
frame-relative; see the remarks on Special Relativity in the Intro-
duction). We take reality, in other words, to be spatially extended.
Do we take it to be temporally extended? Intuitively, yes we do.
That is, we include not only present facts but also past facts: the
Boer War is just as much part of reality as the erosion of the ozone
layer. Statements made about the past have a determinate truth-
value (more of this later, in the next chapter). We may, however,
be reluctant to include future fact. We may feel, intuitively, that the
future is as yet unreal. It is not yet a fact of the matter as to what
will happen tomorrow, or next week, or in 100 years' time. Conse-
quently, any statement made about the future will not have a
determinate truth-value. I do not endorse this position, I merely
state it. Many Tensed theorists would, however, applaud such
remarks, and regard the supposed ontological indeterminacy of
the future as the major argument for their position. So, for the sake
of argument, we shall admit only past and present fact. Such facts
will concern, not just the occurrence of events, but the time of the
occurrence of events. That is, there are temporal, as well as spatial,
facts. Precisely what kinds of temporal facts there are will depend
upon which view of time is correct, but we are assuming the
Tensed view here. On the Tensed view, then, such facts will be
tensed facts: 'The Boer War is past', etc. We may distinguish
between present temporal facts and past temporal facts in the
Tense and Contradiction 33

following way: present temporal facts make true statements made


in the present; past temporal facts make true statements made in
the past. So at the time of writing (1989) it is a present, not a past,
fact, that the First World War has been over for exactly seventy-one
years, because if I now say 'The First World War has been over for
71 years', I say something that is true. But if I said the same thing
ten years ago, I would have said something that is false. By the
same token, it is a past fact that the First World War is still going
on, because if someone had said, at some time in the past (any time
between 1915 and 1917, for example) 'The War is still going on',
they would have said something that was true.
We can now see the nature of McTaggart's contradiction. Reality
consists both of present and past fact. It is a present fact that the
First World War is past; symbolically:

Pp

and a past fact that the First World War is present; symbolically:

Np

But these facts cannot both obtain. Yet if there are tensed facts,
they do. So there are no tensed facts.

2.3 THE RADICAL SOLUTION

The obvious (in fact, for the Tensed theorist, the only) move is to
reject past facts. That is, there are only present facts. This, im-
plicitly, is what Tensed theorists are asserting when they insist that
there is no contradiction in supposing the world to be tensed, since
one need only say that a given event was future, is present, and will
be past. To be sure, they say, one gets a contradiction if one adds
that the same event was future, is future, and will be present, but
we needn't add this as it is not true any more. As Prior put it, one
only gets contradiction half the time in McTaggart's argument, so
why should one stop at the level of contradiction, rather than at
the level at which the contradiction is resolved? The answer to
Prior's question is that one does not stop at the level at which
contradiction is resolved because that level includes only present
fact, and we wish to include past fact as well.
34 Clumge, Cause and Contradiction

We can now, perhaps, make sense of Dummett's remark. We


cannot provide a complete description of reality if 'complete'
means 'observer-independent', for what is the correct description
of reality depends upon the time at which one happens to be
located. We cannot therefore hope to ignore the fact that we are
located in time and attempt to adopt a 'God's eye' view in our
descriptions.
This, I believe, makes explicit what is assumed in Tensed theor-
ists' 'resolution' of the paradox: that only present tensed descrip-
tions are to be accepted, and this is because only present facts
comprise reality. Thus stated, it is a puzzling doctrine. What
precisely it involves will be the topic of the next chapter. It is a
position of considerable philosophical interest, not just because of
its immunity to McTaggart's argument, but because it was the
philosophical basis provided for his subject by the founder of tense
logic, Arthur Prior. For the time being, at least, there is a reprieve
for the Tensed theorist.
Just before leaving McTaggart, I would like to show briefly what
relevance the foregoing discussion has to any attempt to apply
McTaggart's argument to the modal case in support of realism
concerning other possible worlds.
The modal version of McTaggart goes as follows: we may classify
the positions of objects and events in logical 'space' according to
whether they are actual or merely possible, thus constructing a
'modal A-series'. However, there arises the following conffict
between two apparently unassailable propositions concerning this
series:

1. Different positions in logical space are mutually exclusive: if an


event or object is merely possible, it is not also actual, and vice
versa.
But:
2. All events and objects are both actual and merely possible.

1 is surely uncontroversial. The justification for 2 is that every


event and object is actual at the world it is located in, and merely
possible at every world it is not located in. Given that the conjunc-
tion of 1 and 2 is self-contradictory, we should accept that there is
in reality no modal A-series, but rather a set of possible worlds
each actual at itself, and none of which is ontologically privileged.
Consequently, we should accept that this world, as one of these
Tense and Contradiction 35

possible worlds, is not ontologically privileged. But since this


world is part of concrete reality, so is every other world. This is
modal realism, and the thought is that if McTaggart's argument
establishes that there is no temporal A-series, its modal analogue
establishes modal realism (or, that modal discourse is incoherent).
Should we just accept the modal consequences, or should we try
yet again to find a fallacy in McTaggart?
In fact the modal argument, like McTaggart's original, is valid,
but in this case we can deny premise 2 above with impunity:
events are not both actual and merely possible. Only one world is
real, and that is the actual world. Other possible worlds are, e.g.
just consistent (but false) stories. Every event which occurs in this
world is actual simpliciter; every other event is merely possible (or
impossible). To assent to 2, in fact, is already to accept modal
realism - to see this, one has only to put the word 'really' before
the 'both'. What independent plausibility 2 has stems from accept-
ance of the 'indexical thesis of actuality' (Lewis (1973)). This is the
thesis that 'the actual world' denotes the world it is uttered or
inscribed in (i.e. not necessarily this world). But this doctrine, as
Stalnaker (1976) has pointed out, is quite consistent with the denial
of the reality of other possible worlds. It is certainly not enough to
establish 2.
The doctrine that only the actual world is real avoids the modal
paradox just as the doctrine that only the present is real avoids
McTaggart's paradox. The modal paradox would only defeat some-
one who both accepted modal realism and thought nonetheless that
there was a non-relational distinction between the actual and the
merely possible - but this would be an odd position indeed.
3
Temporal Solipsism
3.1 PRIOR'S DOCTRINES

A number of philosophers (e.g. Ryle (1953), Mayo (1962)) have


found the doctrine that the future is unreal - that it is simply a
collection of possible worlds- tempting: it fits with a certain notion
of free will (see Denyer (1981)). The view that Arthur Prior adopted
near the end of his life is more radical: he held that the past, too, is
unreal. Prior had no particular name for this position, but we may
call it 'temporal solipsism'. In earlier writings this radical position
is not explicit, and when important component doctrines are
entertained, some doubts are occasionally expressed (Prior (1967),
pp. 170-71). But his last paper is quite unambiguous on the point:

the present simply is the real considered in relation to two


species of unreality, namely the past and the future. (Prior
(1970), p. 245.)

It is this position that I wish to consider in this chapter. It is


fortunate that it was championed by so astute a philosopher as
Prior, for in examining his doctrines we are able to consider a
position of considerable sophistication. I shall therefore be calling
on a number of Prior's writings throughout the chapter, but where
I feel his position needs supplementing if it is to be an adequate
statement of temporal solipsism, I supplement it. It is, after all, the
temporal solipsist, not Prior in particular, who is my target. Apart
from the fact that temporal solipsism represents the only means to
block McTaggart's proof of the unreality of time consistently with
the assumption of a non-relational past, present, and future, it is of
considerable interest in being the philosophical basis of Prior's
pioneering tense logic.
Temporal solipsism is elaborated in a series of doctrines (refer-
ences are to Prior in what follows):
36
Temporal Solipsism 37

(1) The extension of the existential quantifier is restricted to pre-


sently existing objects ((1957), pp. 32-3, 38; (1968), pp. 13,
79-80, 147). As Butterfield (1984) notes, this doctrine by itself
does not entail that past and future objects are not legitimate
domains of quantification, but Prior's view is that there are no
such objects ((1968), p. 147).
(2) Relations obtain only between contemporaries, i.e. objects
which exist at the same time ((1957), p. 31; (1967), pp. 169-70).
Apparent relations between non-contemporaries, such as the
relation of being taller than, in which I (now) stand to my great-
grandfather, must be analysed in terms of some present fact.
(3) Past and future tenses are to be interpreted as sentential
operators on core present-tensed sentences, the present tense
not requiring to be represented by an operator ((1957), pp.
8-10; (1968), pp. 20-22). The import of this is that present truth
is truth simpliciter (we might call this Prior's 'redundancy
theory of the present'). The operators themselves are primi-
tive, and therefore not ultimately analysable as quantifiers over
instants.
(4) Instants are logical constructions out of propositions ((1967),
pp. 79-82, 88-92; (1968), pp. 122-6). An instant, that is, is
equated with all the propositions which would ordinarily be
described as being (contingently) true at that instant. The
content of the propositions concerns the states of affairs taking
place at that instant (mathematical propositions would, of
course, be irrelevant).

Motivating Prior's doctrine (2) above is surely a more general


position, namely reductionist anti-realism with regard to the past
and the future:

(5) Past and future tense statements have only present fact as their
truth-conditions, i.e. what makes a certain statement about the
past true (if it is true) is the evidence that at present exists. (We
need not assume that this evidence is available to us, merely
that it exists - we are not here dealing with a doctrine which
replaces truth with justified assertibility.) This is possible in
virtue of the fact that there are present facts which derive their
character from causal connection with past states of affairs,
and which determine (at least to some extent) the character of
the future. 1
38 Change, Cause and Contradiction

Prior is not explicit in his avowal of this doctrine, though remarks


in (1970) and (1967), p. 170, are suggestive.
I want to suggest that doctrines (1)-(5) above are essential to the
articulation of the temporal solipsist's position. In denying the
reality of past and future times, he must deny the reality of purely
past and future individuals (i.e. those not presently existent, like
Napoleon, and Queen Elizabeth II's great-grandchildren). If (as for
the time being we shall assume) the existential quantifier is to be
tied to existence, then its range must reflect this restriction. (5) is
evidently essential: if there are no past and future facts, then these
cannot be part of the truth-conditions of past and future tense
statements. (5), obviously, guarantees (2). (3) and (4) are also
essential. There is no reason, of course, why the temporal solipsist
should not quantify over instants, provided he is able to eliminate
such quantification in favour of terms which do not commit him to
the existence of past and future times - for it is precisely the
existence of such times that he wishes to deny. I assume here, with
Quine, that a theory which involves ineliminable quantification
over F's is committed to a realist position over F's. The only things
which the temporal solipsist has at his disposal for the construction
of past and future times which guarantee the appropriate speci-
ficity are past and future tense propositions (see section 5). Basic
propositions, therefore, must be tensed, and the tense operators
are required to modify them appropriately. (4), therefore, requires
(3).
Note the immediate consequences of these positions. The extent
to which the principle of bivalence is violated by statements about
the past or future depends, for the temporal solipsist, upon how
much causal determinism he is prepared to allow. In a fully
deterministic universe, given the total state of the universe at a
given time, only one future and one past (i.e. one series of later
and earlier events) is physically possible (compatible with physical
law). In such a universe, all future and past tense statements have
a determinate truth-value, as this is guaranteed by present fact.
But in an indeterministic universe - such as (some think) the actual
one, many statements about the future must for the solipsist lack a
truth-value. Consider, for example, how we observe the gradual
erosion of traces of things and events - the sands of time on which
our own footprints are placed are constantly shifting. To take an
extreme example, it seems likely that present evidence is consist-
ent both with a certain icthyosaur having two ammonites for
Temporal Solipsism 39

breakfast this time 200 million years ago, and with his only having
one. The temporal solipsist has to concede that in such cases there
is simply no fact of the matter. Unappealing as such a doctrine is,
however, we cannot on that basis alone pronounce temporal
solipsism incoherent.
We might ask how past and future tense statements can be made
true by present fact - that is, how the anti-realist analysis would
go. Prior, unfortunately, does not offer an account of this, so we
shall have to supply one ourselves. We may take as our model here
an anti-realist analysis of modal statements (see, e.g. Adams
(1974)), where 'O(p)' is true, if and only if p is a member of a
consistent set of propositions, such that, for any proposition q,
either q or -q is a member of that set. The important feature of this
analysis is that the role of the modal operator here is clearly
defined. Similarly, we should require of any anti-realist analysis of
statements about other times that it clearly define the role of the
temporal terms- in Prior's case, of the tense operators. Exploiting
the causal connections between the past and the present, we may
offer the following analysis, where L is the conjunction of all
physical laws, and S a statement of the total present state of the
universe:

'P(p)' is true iff D((L& S)- p) & -o(L- p) & -o(S- p)

The second conjunct ensures that pis not itself a physical law, and
the third conjunct ensures that p obtains at some time other than
the present. The problem with this analysis is that it fails to
distinguish between the past and future tense operators. That is,
the right-hand side of the biconditional could equally define the
truth-conditions of 'F(p)'. It is in fact hard to see how a full analysis
could avoid reference to instants and the 'before than' relation, and
this conflicts with doctrines (3) and (4) above. I shall return to this
difficulty later.
One conflict with intuition within Prior's scheme concerns his
treatment of individuals who are now dead, such as Napoleon,
Tolstoy, and the Sarajevo assassin. Surely, we want to say, there
are still facts about Napoleon - for example, that he engaged the
Russians at Austerlitz and Borodino? Prior, however, denies this:

it is arguable that before and after an individual x exists there are


no such propositions as [Fx] .... I myself sometimes put it that
40 Change, Cause and Contradiction

at the times in question no such proposition is 'statable'.


This . . . locution has the disadvantage of suggesting that the
difficulty here is simply with our mechanisms of reference; I
want to say rather that there are no facts about x to be stated
except where x exists. (Prior (1968), p. 147.)

How indeed can we say anything of Napoleon if there is nothing


appropriate in the scheme of things to refer to by this name? (We
are not interested in pet animals who happen to be named 'Napo-
leon'). The situation may be compared with failed uses of demon-
stratives: if 'this castle' in my utterance, 'this castle hath a pleasant
seat', fails to refer, then I have not said anything. However, the
cases are not quite analogous. 'Napoleon was a great general',
though not a referential statement, can still be true in virtue of the
(present) fact that it was the case n years ago that there exists an
individual such that x is called 'Napoleon', x is French, x is a
general . . . etc. As Prior puts it, there can be general, though not
singular, facts about the past.
Need the temporal solipsist accept this? Why, for example, can
he not invoke the causal theory of names (Kripke (1972)), in which
singular reference to an individual x is achieved via a series of
causal connections between x (or some naming of x) and my
present use of 'x'? The answer is that this theory is simply not open
to the temporal solipsist, for he denies the reality of the past states
of affairs and/or individuals with which the present event- a token
utterance of 'x' - is supposed to be connected. The causal relation
thus lacks a relatum.
Having made these preliminary remarks, I shall now consider
aspects of Prior's system which lead him into unacceptable pos-
itions. Two of the sections below, although critical of Prior, are
inspired by his own discussions.

3.2 'PATHOLOGICAL' CASES OF DIACHRONIC IDENTITY

In restricting the quantifier to presently existing objects, the tem-


poral solipsist is not thereby committed to an ontology of instan-
taneous objects. And although singular reference is limited to
presently existing objects - as only such objects are real- this need
not imply that what is identified has no temporal extension.
Indeed, it might be thought that, for the temporal solipsist, dia-
Temporal Solipsism 41

chronic identity is peculiarly unproblematic. For him the question


'in what relation must some past object and some present object
stand to each other in order for them to be one and the same
object?' is redundant. Provided we can identify the present object,
we can make it the subject of past tense statements. (Compare
Kripke's remark about 'transworld identity': there is no problem
about how an individual can be in more than one world. Provided
we can identify him in this world, then we can predicate modal
properties of him. See Kripke (1972), p. 268.)
However, recent work by Parfit ((1971), (1984)) has led us to
question strong conceptions of identity. As Parfit puts it, we seem
to be able to 'describe cases in which, although we know the
answer to every other question, we have no idea how to answer a
question about identity.' (Parfit (1971), p. 3.) This might lead us, as
it does Parfit, to reject the strong, supposedly primitive, notion of
identity, in favour of some weaker notion, such as physical or
psychological continuity. Such weaker relations do not exhibit all
of the logical properties attributed to identity. For example, 'psy-
chologically continuous with' need be neither one-to-one nor
transitive, and this is crucial in avoiding the paradoxes which
otherwise arise for pathological cases of identity. By 'pathological'
cases, I mean such phenomena as fission- an individual's dividing
in two (e.g. amoebae)- and fusion- two individuals' becoming one
(e.g. raindrops). Suppose, for example, that x splits to form two
individuals y and z. If we assume that xis identical toy and that xis
identical to z then, since identity is transitive, we must conclude
that y is identical to z. But y is not identical to z: they occupy
different spatial positions and may exhibit other incompatible
properties. However, if we replace identity with physical conti-
nuity, which is not a one-to-one relation, paradox does not result.
Now one of the problems these weaker substitutes for identity
poses for the temporal solipsist is that many of them incorporate as
an essential component causal continuity between non-
contemporaneous individuals, and this the temporal solipsist can-
not accommodate. Causal relations (like other relations) cannot, on
the temporal solipsist's view, hold between non-contemporaneous
objects because at least one of the relata is unreal (only one of a pair
of non-contemporaneous individuals could be presently existent).
These relations have to be reconstrued in terms of present facts
concerning some presently existing (though temporally extended)
individual. However, the view of Parfit and similarly motivated
42 Change, Cause and Contradiction

theorists is precisely that the identity of an object over time is to be


construed in terms of a series of distinct, non-contemporaneous,
individuals. Since their project is incompatible with temporal sol-
ipsism, the greater pressure to adopt their view of identity, the
greater pressure to reject temporal solipsism.
A further problem, raised- to his credit- by Prior himself (1965},
involves a conflict between the possibility of fission and fusion on
the one hand, and tense-logical versions of Leibniz's Law on the
other. Prior was interested in this conflict because he felt that it
ought to be possible to develop a logic which accommodated a
more sophisticated ontology than one of individuals that simply
start and cease to be: one which took into account the fact that
objects may arise out of collections of other objects.
Abbreviating Prior's discussion, the problem may be set out as
follows. The 'weak' form of Leibniz's Law is that, if x and y are
numerically identical, then any property exhibited by x will also be
exhibited by y and vice versa. (The 'strong' form is the principle of
identity of indiscemibles.) Formalizing this:

1. (x)(y)(cj>)(x = y ~ (cj>x ~ cj>y))

If we take this to be a present tense formulation, then it will be


accompanied by its past and future tense counterparts, and there is
no reason to suppose that Leibniz' s Law is temporally restricted.
So, introducing tense operators, indexed by a number of temporal
units which indicates temporal distance from the present moment:

2. P n(X = y} ~ (Pn<I>X ~ P n<I>Y} 2

3. Fn(x = y) ~ (Fncj>x ~ Fncj>y)

Consider 2: a natural reading of this is that, if x and y were


identical in the past, then any past tense proposition true of x at
that time would also be true of y. Among these propositions would
be, not only'- was red', but also the more complex'- was going to
be red', for it is a tense logical principle that

i.e., if it is now the case that cj>x, then it was the case n units of time
ago that it will be the case n units of time hence that cj>x. Conse-
quently, we should also accept that
Temporal Solipsism 43

But to 5 we can oppose the following counterexample: objects x


and y have resulted from the fission, n units of time ago, of a
parent object. Since the division, x and y' s life histories have
diverged somewhat, sox is now <1> andy is now not <j>. Yet it is true
both that Pn(x = y) and that PnFn<l>x, so from 5 we should conclude
that PnFn<I>Y- i.e. y is <j>. Prior's initial response is to cast doubt on
4, which is, he thinks, suggestive of logical determinism. In fact I
believe 4 to be trivial, since on Prior's own scheme, it is guaranteed
by present, not past, fact. But in any case, we can construct a
problem which clearly does not rest upon deterministic premises,
as follows. One supposedly tense logical version of Leibniz's Law
structurally parallel to 5 is

A counterexample to 6 is provided by the case of the fusion of two


individuals, x andy, where prior to the fusion, x was <1> andy was
not <j>.
At the end of his discussion, Prior recommends that Leibniz's
Law be abandoned, or at least seriously modified. Need we accept
this? One alternative strategy is to insist that Leibniz's Law applies
only to bona fide properties, and that time-relativised properties
(like 'was going to be red') are not included in these. 5 is thus a
rather suspicious principle. (More accurately, 5 does not involve
time-relativised properties, but rather treats 'Fn<l>-' as an ordinary
predicate.) In addition, the move to 5 is itself suspicious. From 2
and 4 we should derive, not 5, but

Now this does not at all conflict with the possibility of fission
where the products exhibit incompatible properties, as 'Pn' is
equivalent to 'PnPnFn'· And surely any plausible tense-logical ver-
sion of Leibniz's Law will require only that, for any timet, if x and
yare identical at t, that if xis <1> at t, y is also <1> at t, where '<I>' is not
time-relativised. That is, the antecedent (x = y) and the consequent
(<l>x--+ <j>y) of the Law must have contemporaneous truth-conditions.
This rules out both 5 and 6. We might ask why Prior did not
consider this strategy, but in fact it is not open to him- at least as a
temporal solipsist (it is not clear how far he had moved toward this
44 Change, Cause and Contradiction

position in 1965). For his reductionist anti-realism dictates that


both Pn<l>x and Fn<l>x are made true by present fact. The stricture
that the antecedent and consequent of any tense-logical version of
Leibniz's Law must have contemporaneous truth-conditions does
not, therefore, rule out 5 and 6 for the temporal solipsist. Nor is it
any good to say that they must be about the same time, for if they
are past or future tense propositions, they cannot be about any time
at all, as past and future times do not exist. Consequently, he is
forced either to reject Leibniz' s Law, or deny that fission or fusion
take place. More precisely, to remain consistent with Leibniz's
Law, he would have to say that, e.g., fission involves the demise of
one object and the starting to be of two different ones - a concep-
tion Prior admits is a little simpliste (Prior (1965), p. 190).

3.3 INSTANTS AND ANTI-REALISM

We come now to a more serious problem for the temporal solipsist.


This is that one of his doctrines - that instants are logical construc-
tions out of tensed propositions - creates difficulties for another,
viz. reductionist anti-realism with respect to the past and future.
As we owe the term 'anti-realism' to Dummett, and my use of it
differs somewhat from his, I should make it clear what the differ-
ence is. Dummett (1963) characterizes an anti-realist position con-
cerning a certain class of propositions (e.g. about mental states,
other times, physical objects) as one which holds that propositions
of that class do not have a 'verification-transcendent' truth-value.
That is, we cannot coherently think of them as having a truth-value
in those cases where it would not be possible even in principle to
establish what that truth-value was. Within that category he goes
on to distinguish two kinds of anti-realism. The first holds that
propositions of the 'disputed class' have truth-conditions which
are not of that class. Phenomenalism is an anti-realist position of
this type: it regards statements purporting to be about external
objects as true if and only if some conjunction of counterfactual
statements concerning perceptual experiences is true. This kind of
anti-realism is reductionist. It gives a non-obvious interpretation of
the truth-conditions of certain propositions. The second, weaker,
kind of anti-realism does not involve reductionism - it simply
denies verification-transcendent truth-value for some class of
propositions.
Temporal Solipsism 45

However, because theorists who want to oppose realism con-


cerning certain kinds of proposition are not necessarily anti-realist
in Dummett's sense, I want to use 'anti-realism' just to mean
reductionism. That is, I do not want to tie it to verification. Behav-
iourism (or at least one version of it) is anti-realist in my sense of
the term: it holds that the truth-conditions of statements purport-
ing to be about mental states concern dispositions to various kinds
of behaviour whether or not we could establish the existence of those
dispositions in every case. To avoid confusion, I shall mark Dum-
mett's sense of 'anti-realism', when I want to refer to it, by under-
lining the phrase: anti-realism. It is clear that the temporal
solipsist, as I have characterized him, must be an anti-realist,
though not necessarily an anti-realist over propositions about the
past and future.
Now, is the solipsist's reductionist anti-realism concerning other
times coherent? The first problem to note is that the anti-realist
analysis depends upon there being causal connections between the
present and the past, and between the present and the future. If
this were not so, how could present fact determine the truth-value
of past and future tense statements? Such causal relations relate
non-contemporaneous states of affairs, and hence cannot be counten-
anced by the temporal solipsist (see his doctrine (2): relations
obtain only between contemporaries). However, provided causal
relations are not mentioned in the anti-realist analysis, no doctrine
is explicitly violated- it will just have to remain mysterious what
the mechanism which underlies the analysis is.
This leads us to another difficulty. It is a requirement on any
reductionist theory that the reductive class of propositions - those
directly expressing the truth-conditions of the disputed class - be
intelligible independently of the disputed class of propositions. It
is easy to see why this requirement is necessary. If the terms of the
alleged reductive class had to be analysed in terms of those of the
disputed class, then no genuine reduction would have been
achieved: we would in effect have stated the truth-conditions of
propositions about F's in terms of F's. Behaviourism appears to
meet this requirement: we can talk of dispositions to behaviour
without having to invoke mental states in elucidating such talk.
But reductionism with regard to other times does not meet the
requirement. Consider again the analysis suggested in section 1:

'P(p)' is true iff (D((L & 5)--+ p) & -o(L--+ p) & -o(S--+ p))
46 CJu:mge, Cause and Contradiction

Now I have already remarked that we lack a means to dis-


tinguish between 'It was the case that' and 'It will be the case that'.
The only way we can do this is to exploit the asymmetry between
the past and future and build this into the analysis. But we cannot do
this without reference to the very items we want to analyse, viz. past and
future states of affairs.
But perhaps all is not yet lost. We can distinguish between
statements about past states of affairs and statements about past
times. We can therefore state the truth-conditions of statements
about past and future states of affairs in terms of present fact, and
then go on to explain the asymmetry between past and future in
terms of connections between different times. So circularity is
avoided. However, this way out is blocked for the temporal solip-
sist, for he cannot regard statements about times as independent of
statements about states of affairs. Precisely because he denies the
reality of other times, he is obliged to construct times from prop-
ositions about the past and future. And, on his reductionist views,
these propositions are made true in virtue of present fact. So he is
caught in an explanatory circle: the truth of past and future tense
propositions is explained in terms of present fact, which is explained
in terms of relations between different times, those times being
themselves constructed out of past and future tense propositions.
Further problems are raised by the propositional theory of in-
stants. It has been suggested by Butterfield (1984a) that relationism
(of which the propositional theory is a variant) fails to do justice to
the plenitude of instants comprising the temporal extension of the
world. That is, it identifies instants that are truly distinct (cf.
Lewis's complaints about 'ersatzist' theories of possible worlds
(1986)). Now I suggest that this problem is much more severe if the
propositional theory is combined with reductionist anti-realism
over other times.
The solipsist characterizes times as conjunctions of contempor-
aneously true propositions about events/states of affairs. Conse-
quently, times which are qualitatively indistinguishable are
numerically identical on this theory. Nietzsche's picture of infi-
nitely recurring history would thus be ruled out as incoherent. For
the temporal solipsist, of course, events/states of affairs which do
not presently obtain do not exist, so he needs to resort to re-
ductionism. But since the causal effects of many distant events
become increasingly indirect and overlaid by the effects of sue-
Temporal Solipsism 47

ceeding events, and since the universe (most likely) does not
exhibit determinism in reverse (i.e. a number of different past
histories are compatible with the present state of the universe) it
will no longer be possible to differentiate past states of affairs by
their present effects. So there is very little to differentiate very
distant times - so many of the instants comprising the past will be
conflated, and the temporal extension of the universe will be very
much less than we would imagine it to be (especially if we think it
is infinitely extended into the past: we certainly do not possess the
materials for the construction of an infinite past series of instants).
Consequently, what problems the temporal solipsist has with his
propositional theory of instants are considerably worsened by his
additional (and unavoidable) adherence to reductionism.
The third problem under the heading of this section is this. A
theorist who replaces instants with the supposedly more funda-
mental F's must, ultimately, explain the properties of the time-
series in terms of F' s. Prior is quite explicit about this:

all talk which appears to be about [instants], and about the


'time-series' which they are supposed to constitute, is just dis-
guised talk about what is and has been and will be the case.
(Prior (1968), pp. 122-3.)

One of the features of the time-series is its topology: the fact that it is
branching or non-branching, dense or discrete, linear or closed.
Can these different topologies be represented simply by logical
relations between propositions? Precisely this seems to be guaran-
teed by the existence of a number of tense logical systems, which
arose precisely as a means of representing different temporal
topologies. For example, a logic for non-branching time, system
CL (Cochiarella (1965)), is obtained by adding to the axioms of
Lemmon's minimal tense logic 1<t the following axioms:

FFp~ Fp
(Fp & Fq) ~ (F(p & q) v F(p & Fq) v F(Fp & q))
(Pp & Pq) ~ (P(p & q) v P(p & Pq) v P(Pp & q))

And a logic for dense time can be obtained by adding

Fp~ FFp
48 Clumge, Cause and Contradiction

The question now arises: how is the temporal solipsist to explain


the truth-conditions of the axioms above? His anti-realism with
respect to the past and future commits him to the view that the
truth or falsity of the above axioms are guaranteed by present fact.
But, surely, no purely present fact about the world could determine
which topological structure the time series has. So, again, the
propositional theory of instants is in tension with the solipsist's
anti-realism.

3.4 A PARADOX OF CARDINALITY

At the very end of his important paper, 'Facts and Propositions',


Ramsey proposes a simple argument which has, he thinks, the
surprising conclusion that if there are in fact N things in the world
then it is necessarily the case that there are N things in the world.
The proof, put formally, goes as follows:

Suppose:
1. There are exactly four things in the world, a, b, c, and d.

Given the principle that

2. -(x = y) ---+ o-(x = y)


- i.e. non-identical things are necessarily non-identical, 3 we can
infer:

3. o-((d = a) v (d = b) v (d = c))

Let this complex proposition be represented by 'D'\jld'.

4. D(cj>a---+ 3xcj>x)

This is the modalized principle of existential generalization. Ap-


plying it to d and the property '\jl:

5. D('\pd ---+ 3x'\jlx)

Given the modal axiom D(p---+ q)---+ (D p---+ Dq)- an axiom ofT, 5
entails:
Temporal Solipsism 49

6. D'lj1d - o3x'lj1x

From 3 and 6 it follows that

7. D3x'lj1x

That is, necessarily there is something that is neither a, b, nor c.


This doesn't actually amount to the conclusion that there are
necessarily (if actually) exactly four individuals in the universe, as
Ramsey thought, but merely to the (still surprising) conclusion that
necessarily there are at least four individuals in the universe. To
obtain the conclusion that there are at most four individuals in the
universe, we might assume both 1, above, and, for reductio:

8. ¢(3v3w3x3y3z((v =a) & (w =b) & (x =c) & (y =d) & (z =e))

i.e. it is possible that there exist (in some possible world) five
individuals, a, b, c, d, and e.

The Ramsey argument can then run through for this assumption,
as follows. Since non-identity is necessary non-identity, it is
necessarily the case that e is non-identical with any of a to d. On
the modalized principle of existential generalization, it follows that
it is necessarily the case that there exists an individual which is
non-identical with a, b, c or d. So it is necessarily the case that
there are at least five individuals in the universe. But this contra-
dicts assumption 1, that there are actually only four individuals in
the universe. So if 1 is true, 8 is false. So 1 entails:

9. o-3x-((x = a) v (x = b) v (x = c) v (x = d))

i.e. necessarily, there are at most four individuals in the universe.


(This application of the Ramsey argument, note, is more contro-
versial than the original, in that it rests upon the dubious assump-
tion that one can introduce singular terms for possible
individuals.)
Surprising though the results of Ramsey's argument are, given
that they rest on quite ordinary assumptions about modality (with
the possible exception noted in parentheses above), the argument
has a tense logical analogue. In other words, from the assumption
that there now exist just four individuals in the universe, we can
50 Clumge, Cause and Contradiction

derive the outrageous assumption that there always were and


always shall be just four individuals in the universe. The proof is
structurally identical to the proof above, but turns on tense logical
versions of the principle of existential generalization, the modal
axiom from T (see 4 above) and the necessity of identity. Following
Prior's notation, let 'G' represent 'it will always be the case that',
and 'H', 'it has always been the case that'. We need to accept, for
the temporal version of Ramsey's argument:

Principle TI(for 'Tensed Identity'): -(a = b)~ (H-(a = b)


& G-(a =b))

That is, if it is presently the case that a is non-identical to b, then it


has been and always will be the case that a is non-identical to b.
This does not rule out fusion and fission simpliciter but rather a
particular description of fission and fusion. For example, we can-
not say, when 'a' and 'b' are co-referential, that a will at some point
cease to be identical to b. In addition, we need to accept:

Principles of TEG('Tensed Existential Generalization'):


G(<j>a ~ 3x<j>x); H(<j>a ~ 3x<j>x)

which are the tense logical analogues of the modalized principle of


existential generalization, and, finally:

Principles of TMI('Tensed Material Implication')

G(p ~ q) ~ (Gp ~ Gq); H(p ~ q) ~ (Hp ~ Hq)

which are axioms of I<.. Their conjunction is just the tense ana-
logue of D(p ~ q) ~ (op ~ oq).
Following the tense logical version of Ramsey's argument, we
end up with the conclusion that if there are now just four individ-
uals in the universe, then at no time are there fewer than four
individuals, as follows:

Suppose
10. There are now just four individuals in the universe, a, b, c,
and d.
Then, by TI:
Temporal Solipsism 51

11. G-((d =a) v (d =b) v (d =c)) & H-(d =a) v (d =b) v (d =c))

Let this be represented: G'ljld & H'ljld

Given TEG and 1MI, it follows from 11 that

12. G3x'ljlx & H3x'ljlx

That is, it always was and always will be the case that there is
something non-identical with a, b, and c. So at no time are there
fewer than four individuals in the universe. And assuming that we
can introduce singular terms for future, or past, individuals, 4 we
can construct a reductio of the assumption that at some future time
there will be, or that at some past time there were, more than four
individuals. So either nothing goes in or out of existence, or, every
time something does go out of existence, something else immedi-
ately comes into existence, so as to keep the number of individuals
in the universe constant. Both of these disjuncts are incredible. So
surely there is something wrong with the argument?
The obvious way to disable it is to unrestrict the existential
quantifier so as to include past and future objects in its domain.
The conclusion of the argument is only surprising if we interpret
'3x' as 'there now exists an x'. But if we interpret it as it is generally
interpreted in logic, that is as 'there either did, does, or will exist
an x', then the conclusion is absolutely harmless, for of course the
sum of all past, present, and future individuals does not change
over time. But the temporal solipsist cannot take this step: past and
future individuals are not, for him, legitimate domains of quantifi-
cation. So he must interpret '3x' as 'there now exists an x'.
We might ask, instead, whether TI is a plausible principle. Take
its modal analogue: if a and b are actually non-identical then they
are necessarily non-identical, or, as we might express it, they are
non-identical in every possible world. But this is perhaps mislead-
ing; we should say that in every world in which a and b exist, they
are non-identical. After all, we don't want to assert that, just
because it's necessarily true that -(a = b), a and bare necessarily
existent. Perhaps it would be better to say that a and b are
essentially non-identical. But now if we make this move, Ramsey's
argument collapses, for 3 should now be modified to D(3x(x = d)
- 'ljlx). Since it is not true that d is 'ljl in every world, it cannot be
52 Change, Cause and Contradiction

inferred that there exists in every world an x such that xis '\jl. We
can make an analogous move in the temporal case. That is, rather
than saying that d is '\jl at all times, we should say that at all times at
which d exists, dis '\jl. The argument then fails to go through.
However, d exhibits another property, ;, where; is defined as
follows: xis; if and only if xis such that xis essentially '\jl (i.e. '\jl in
every world in which x exists). Now it is the case that dis; in all
possible worlds, not just in those where x exists, for it is true in all
possible worlds that 3x(x = d) -+ '\jlx. So the argument goes
through as before, this time for the property; instead of '\jl.
Perhaps rather surprisingly, although Prior considers Ramsey's
proof in his essay 'Time and Existence' (Prior (1967), pp. 137-74),
and is in that essay concerned with modal and temporal anal-
ogues, he does not consider the precise temporal counterpart to
Ramsey's proof. His reply to Ramsey is ingenious, but its temporal
counterpart has unfortunate consequences.
The reply (op. cit., pp. 151-2) turns on the distinction between
'It must be (or is necessarily) true that' and the weaker 'It could not
be false that'. The modal operator o is interpreted 'It could not be
false that', which does not entail 'It must be true that'. So in-
terpreted, D'\jld (or o;d), is true, even in situations where the
embedded proposition '\jld (or ;d) is not true - for example in
situations where d did not exist. In such situations, there could be,
as Prior states it, 'no such proposition' as '\jld. So the conclusion of
Ramsey's argument is not that there is some individual in every
possible world which is non-identical with a, b, and c, but rather
that there is no true proposition which denies this. The temporal
analogue of this reply is obvious: it's not always true that there is
an individual which is non-identical with a, b, and c, but at present
there is no true proposition which denies this.
Now this curious result comes about through defining the num-
ber of individuals in the universe by singular reference to those
individuals. But we don't need to do this (at least, it would take
complex argument to show that we do). So we can express Prior's
reinterpretation of the conclusion as follows: it is not true for all
times that the number of individuals in the universe is N, but if
there are now N individuals, then there is no true proposition
which denies that there were and always will be exactly N individ-
uals. But this seems plainly false. Surely even present fact could
guarantee that there used to be more or less, individuals in the
universe. Take the following situation: suppose it is presently the
Temporal Solipsism 53

case that there are just four people in the universe, and suppose
further that (a) they are not siblings (full or half), and (b) they are
exactly the same age as each other. It would follow from these
present facts plus some elementary biological rules that at some
time there were at least eight people in the universe - viz. the
parents of the present four. One might object 'but the fact that two
individuals are not siblings is not just a present fact about them'.
This is true, but the temporal solipsist cannot say this, because he
must hold that the lack of siblinghood between two individuals is
entirely constituted by present fact. Perhaps, then, the temporal
solipsist should hold that it is indeterminate whether two individ-
uals were siblings or not. I imagine that this is genetically most
implausible, but in any case we can present a different case.
Suppose there is just one person in the universe, and he is ter-
minally ill. These present facts make it true that, at some future
time, there will be no people in the universe. It is hard to see how
the temporal solipsist could deny this. So even Prior's way out of
the temporal Ramsey argument leaves one with contradictory
results.
A simpler way out is to qualify the principle of existential
generalization. From the fact that, with respect to a world or time,
xis F, one cannot always conclude that x exists in such a world or
at such a time. For example, consider such predicates as 'famous',
'dead', 'fictional'. So we could tie the principle Fa--+ 3xFx to a
particular domain ofF's (and we should certainly want to exclude
the dubious property ; from this domain). The disadvantage of
this is that, in so doing, we would be abandoning the project of
providing a purely syntactic characterization of existence. For the
temporal solipsist, however, it is the only move left.

3.5 THE PROPOSmONAL THEORY OF INSTANTS

In this section, I want to focus specifically on Prior's conception of


instants as conjunctions of propositions. My contention is, firstly,
that the temporal solipsist is committed to this theory, and sec-
ondly, that the view of propositions on which it depends is inco-
herent.
First then, why must the temporal solipsist regard instants as
propositions? Can he not refuse to entertain any reductionist
theory of instants? And, if he cannot, why can't he regard instants
54 Change, Cause and Contradiction

instead as events, or states of affairs, or facts? He certainly has to


entertain some reductionist theory, because he denies the reality of
the past and the future. Therefore, either he must regard all talk of
past and future instants as contentless, which is absurd, or he
must reconstrue such talk. He can do this in two ways: he can
simply provide the truth-conditions of discourse about instants,
and stop there, or he can go further and identify instants with
some other entity. Let us consider the bolder move of providing
identity conditions (we can always retreat into the weaker position
if problems arise). Suppose the solipsist opts for the identification
of instants with events. The present moment is the collection of
present events, past moments are collections of past events, etc.
The solipsist, however, denies the reality of past and future
events, so this identification is not open to him. A similar problem
prevents his identifying instants with past and future states of
affairs, material objects, facts, and any relations between these. He
obviously cannot identify instants with just present events, states,
objects, or facts, because this would leave him with no means by
which to discriminate between different instants.
The only remotely plausible reductionist strategy open to him is
to identify instants with propositions: an instant is the conjunction
of propositions which would ordinarily be said to be true at that
time. Thus stated, this would appear to rule out the possibility of
time without qualitative change, for such a period would consist of
instants whose associated conjunctions of propositions were
identical. However, this is easily avoided if the relevant prop-
ositions include, not just 'It is the case that ... ',but also 'It was
the case n units of time ago that . . .'. A logic for such propositions
will have to include metric tense operators, and the question that
leaps to mind is, can a semantics for such operators be provided
which makes no reference to instants? Certainly, appeal to instants
allows us the easiest way of interpreting such operators: 'Pn(P) is
true iff 3t(Rn(t, now) & At t(p))'. The solipsist cannot avail himself
of this analysis unless he replaces 't' and 'now' by conjunc-
tions of propositions. And of course, in doing this, he must include
the very propositions he is trying to explicate, viz. 'Pn(p)' and
'Fn(p)'. The solipsist, then, would be better off not resorting to
metric tense operators in the construction of instants.
Turning now to my second contention, it is clear that the prop-
ositional view of instants requires propositions to be tensed- i.e. to
be capable of changing their truth-value from context to context.
Temporal Solipsism 55

(This is equally true of the more modest position which seeks only
to give the truth-conditions of talk about instants in terms of
propositions.) This view of propositions is attributed by Geach to
the Scholastic logicians:

For a Scholastic, 'Socrates is sitting' is a complete proposition,


enuntiabile, which is sometimes true, sometimes false; not an
incomplete expression requiring a further phrase like 'at timet'
to make it into an assertion. (Geach (1949), p. 244.)

Prior quotes this remark of Geach' s when discussing precursors of


tense logic (Prior (1967), p. 15). The example, 'Socrates is sitting',
comes from Aristotle who, as Prior notes elsewhere ((1976),
p. 37-8) does not clearly distinguish between (what we would now
call) 'same proposition' and 'same sentence-type'. On any view of
propositions, 'Socrates is sitting' and 'Socrates is sitting' are tokens
of the same sentence-type. What is at issue is whether they express
the same proposition or not (let us suppose them to have been
inscribed at significantly different times). If they do not, then the
time of the inscription must be part of the content of each prop-
osition - otherwise how can they be distinct?
Supposing that the time of utterance or inscription of a (con-
tingently true or false) token-sentence is part of the content of the
proposition expressed, the natural view to take is that propositions
do not have context-variant truth-value. That is, if two sentence-
tokens express the same proposition, they must have the same
truth-value. However, this view is not forced upon us. In Dum-
mett ((1973), pp. 382-400), we are presented with a view in which
propositions both have times as part of their contents and exhibit (or
may exhibit) context-dependent truth-value (see also Percival
(1989) ). Neither of these views, however, is open to the temporal
solipsist, precisely because he takes times to be conjunctions of
propositions. He cannot without circularity, therefore, hold times
to be part of the content of those propositions.
On the solipsist's picture, then, different tokens of the same
tensed type, even when uttered at different times, express the same
proposition. The consequence is that the very same tensed token
(e.g. this very inscription of 'Socrates is sitting') can be true at one
time and false at another. This makes nonsense of tensed as-
sertion. As Evans puts it, a theory of meaning which was based
upon such a conception of the truth of utterances 'cannot assist the
56 Change, Cause and Contradiction

subject in deciding what to say, nor in interpreting the remarks of


others.' (Evans (1985), p. 349.) To use his example, if we say 'It is
raining', we only rule out dryness for the time of the utterance, not
for the indefinite past or future. But if that very utterance is capable
of becoming false, then we must interpret it as ruling out dryness
at all other times. This means that we could never be in a position
to make any tensed assertions, nor to believe any made by anyone
else. This is clearly absurd.
The idea of a proposition having context-variable truth-value is
not incoherent per se. A harmless example of context-variance
occurs in modal discourse. Suppose I utter a given sentence-token:
'Prior lived for fifty-four years'. With respect to the actual world,
this token is true (expresses a true proposition). But it is only
contingently true. So with respect to some other world, where Prior
lives for sixty-four years, this very same sentence-token is false. No
one could plausibly deny that with respect to this and the other
world the token expresses the same proposition- though there are
some worlds, no doubt (perhaps non-English-speaking ones)
where it expresses a completely different proposition. So worlds
are not part of the content of the proposition expressed by the
token. 5 Consequently, the very same proposition can be true in
one world and false in another. I don't find this at all counterin-
tuitive, but it should be possible to explain why it isn't, when its
temporal analogue is. What is the difference between time and
modality in this respect?
The difference, I think, is this. A sentence-token has a (typically
brief) life-span: the time it takes to utter or inscribe it. It does not
have a location outside that time-span (though in the case of
printed tokens, its causal traces can remain unchanged). Simplify-
ing a little, we say that a token is located at only one time, and we
naturally assess its truth-value with respect to that time. It makes
no sense at all to assess its truth-value with respect to times at
which it is not located. But that same token can be located at many
worlds, besides the actual one. We can therefore assess its truth-
value with respect to those worlds. Since such worlds will differ in
respect of matters of fact, the token will be located in some worlds
where it is false. It makes sense, therefore, to regard its truth-value
as relative to a world, but not to regard its truth-value as relative to a
time. If it was true when it was uttered or inscribed, then it is true
for every other time. This does not mean that other tokens of the
Temporal Solipsism 57

same type cannot be false, merely that they will express different
propositions.
This concludes my study of temporal solipsism. Its most serious
difficulties involve its commitment to the propositional theory of
instants. Not only is this theory in tension with another important
feature of temporal solipsism, viz. its anti-realist construal of past
and future tensed statements, but the theory itself is based upon
an incoherent view of propositions. If this is right, then it is
important for the tense logician to show that his subject does not
have the philosophical basis - in particular, with regard to the
theory of propositions- that Prior took it to have. More import-
antly for our purposes are the consequences for tensed facts. I
argued in Chapter 2 that McTaggart's paradox shows the notion of
tensed fact to be self-contradictory unless one opts for temporal
solipsism. The conclusion of this chapter is that such an option is
an illusory one.
We now face an uncomfortable predicament, for it appears that
the only theory of time which permits genuine change in the world
leads to incoherence. Should we therefore conclude that the notion
of change itself is incoherent? This was McTaggart's conclusion.
The Tenseless account of change as temporal variation in a thing's
properties he thought inadequate, and we saw in Chapter 1 that
there are good reasons for thinking it not only inadequate, but
simply not an account of change at all: Tenseless theory is commit-
ted to a view of objects as having temporal parts, and this (it was
argued) is inconsistent with thinking of objects as subject to
change. It is now time to re-examine this contention, and I shall
begin by considering whether or not there is incoherence in tem-
poral part doctrine itself.
4
Temporal Parts
4.1 THE OBJECTIONS

We must first ask: what are temporal parts? A temporal part is


generally taken to be the intersection of an object and a time or
time-interval (for the purposes of my argument it is to be under-
stood that a 'time' is an unextended temporal point), represented
formally as 'a-at-t'. This is certainly the historical doctrine, as
advocated by Russell ((1914), pp. 105f; (1927), pp. 244, 284) and
Quine ((1960), p. 172). To Quine we owe the notation 'x-at-t'.
Carnap (1958) had introduced a notation, 'Sli{x, y)', interpreted 'x
is a spatio-temporal slice of y', but it never caught on. As first
presented by Russell, the doctrine of temporal parts was seen as a
consequence of physics rather than philosophy. In particular, rela-
tivity theory requires a view of objects as extended in the four-
dimensional manifold. Such four-dimensional objects have tem-
poral, as well as spatial, parts.
However, as Butterfield (1985) has pointed out, temporal parts
are not analogous to spatial parts as they are typically individuated.
On the usual conception of spatial parts, the legs of a table, or the
organs of a body, or the parts of a watch, all count as spatial
parts. It needs no revision in physics to motivate this conception.
Such spatial parts are not the intersections of an object and points
or regions, for objects (and so their spatial parts) can move, and a
spatial part like a chair leg does not lose its identity on being
shifted. This is particularly clear in those cases where spatial parts
are individuated by reference to their functional role, as in the case
of the legs of a table. An object could in principle be disassembled
into such spatial parts, but we do not readily imagine it disas-
sembled into the parts essentially identifiable as 'a-at-s/, 'a-at-s2',
etc. (where 's/ and 's2 ' are spatial locations). This disanalogy
between temporal and spatial parts partly underlies the objections
to temporal part doctrine. But the objections to temporal part
doctrine are not so much aimed at temporal parts per se, but rather
at the thought that ordinary objects possess them. It is never
58
Temporal Parts 59

disputed that events may have temporal parts, for example the
various movements in the performance of a concerto. It is some-
times said that events, unlike objects, lack spatial parts. I shall
have nothing to say about this doctrine here. The point is that we
get our conception of 'part' from considering the spatial parts of
objects and the temporal parts of events, and it is this conception
which is the source of some of the objections to be found in the
literature, as follows:
1. Parts must be causally identifiable independently of the
wholes they are part of (Mellor (1981), p. 133). This is certainly true
of the components of a watch, and the various stages of a war. But
in identifying the putative temporal parts of objects we cannot
avoid referring to that object. So it is never causally explanatory to
account for the existence of one temporal part of an object by
appealing to the existence of another such part, whereas it is
causally explanatory to account for a stage in a war by appealing to
another stage. We might state this difference by saying that div-
isions between spatial parts, or between temporal parts of events,
are empirical, but divisions between temporal parts of objects are
purely a priori. Temporal parts of objects, then, are useless as units
of causal explanation.
2. If objects have temporal parts, then an object existing at one
time cannot literally be identical with an object existing at another.
Different temporal parts must be numerically non-identical. So
diachronic identity should be reconstrued as various relations
obtaining between different objects. This is in fact how Quine
proceeds in 'Identity, Ostension and Hypostasis' (1950). But such a
project cannot succeed, for objects extended in time are logically
and epistemologically more basic than the temporal parts of such
objects. As noted above, one can only refer to an object's temporal
parts via reference to the object itself. And we do not first acquire
the concept of parts and then construct our concept of objects. So
the introduction of temporal parts makes no contribution to our
theory of identity through time.
3. Talk of spatial parts does not threaten an already existing
ontology, but talk of temporal parts of objects does. Our ordinary
ontology, the one presumed by discourse about our sensory per-
ception of the world, is that of three-dimensional objects which last
for a certain period of time (Strawson (1959), Wilson (1955)). This is
not compatible with Quine's proposed redescription of the world
in terms of four-dimensional objects with temporal parts. If we
60 Change, Cause and Contradiction

accepted Quine's revision, we would be introducing a rupture


between phenomenological discourse- discourse about perceptual
experiences - and material object discourse which threatens to
render phenomenological discourse incoherent (Hacker (1982), pp.
Cr-7).
4. If objects were supposed to have temporal parts, then the
distinction between objects and events would collapse. Objects
would simply be extended processes. We know, however, that this
is simply false, as the distinction between objects and processes is
an important part of our representation of the world.
5. For the sake of completeness, we should mention the objec-
tion (Hacker (1982), p. 5) that on Quine's ontology, nothing genu-
inely changes - neither the four-dimensional object nor its
temporal parts. This consequence is simply incredible, so Quine's
ontology should be rejected.
This last objection, which was stated in detail in Chapter 1, is
one of the main themes of this essay. An effective counter to it
requires a theory of change, which I defer to Chapters 5 and 8. The
other objections I shall discuss in the next section.

4.2 UNDERSTANDING TEMPORAL PARTS

Let's begin by considering the disanalogy noted by Butterfield


between temporal and spatial parts. Temporal parts, as character-
ized by Quine, are individuated by reference to a particular time or
time-interval, but they need not be so individuated: consider the
larval stage of the tsetse fly. 1 In this case, the temporal part is
individuated according to its function (the developmental stage).
Such temporal parts do exhibit the temporal analogue of the move-
ment of a spatial part. The movement of a chair-leg across space is
simply a (functionally individuated) spatial part's being at different
places (individuated by co-ordinates) at different times. And there
is a dear temporal parallel to this, namely a (functionally individ-
uated) temporal part- e.g. myself as a child- being at different
places at different times (individuated by dates). This situation
would have obtained if, for example, the first part of my childhood
had been spent in Delhi and the second part in London. My
childhood self can be identified across times just as the chair-leg
can, after movement, be individuated across spaces.
It is admittedly true that, in the literature, there is a tendency to
Temporal Parts 61

identify temporal parts as time-slices, but not to identify spatial


parts as space-slices, and this leads to some mistaken conclusions. 2
It would be interesting to dwell briefly on why, since there are
clearly two ways of identifying parts; philosophers have typically
opted for one way for temporal parts, and the other way for spatial
parts. This tendency, I suggest, has much more to do with the way
we choose to individuate places and times for our convenience
than with the nature of space and time. Lacking a sensitivity to,
and a suitably capacious memory for, numerous micro-events, we
typically individuate times by reference to dates and clock times.
Temporal parts of things are therefore most conveniently individ-
uated by means of reference to these dates and times. Suppose,
however, that we lacked a sensitivity to features of the environ-
ment like the relative spatial positions of objects: we might then
perhaps mark places by means of an elaborate set of co-ordinates -
rather like the ordinance survey grid references, but considerably
more discriminating. Then it would be more natural to mark
spatial parts by reference to such co-ordinates (see Butterfield
(1984b)).
All this, however, is by the way. The functional characterization
of temporal parts - 'my childhood self', 'the caterpillar pupa' - is
inoffensive, and equally available to the Tensed theorist (the Tensed
theorist, recall from Chapter 1, could account for objects' having
temporal extension without regarding them as four-dimensional).
It would be odd to speak of being committed or not committed to
the existence of such parts. The crucial point is that there is
apparently good reason to avoid time-slices. The Tensed theorist
has a way of avoiding them, but the Tenseless theorist, it seems,
has not. We should note that if an ontology of time-slices follows
from the fact that the truth-conditions of temporal indexical state-
ments ('p now') are purely relational, then we must allow an
ontology of space-slices. For space is, uncontroversially, such that
the truth-conditions of spatially indexical statements ('p here') are
purely relational. 3
Even if we limit ourselves to a consideration of time-slices and
their genuine spatial counterparts, space-slices, a disanalogy per-
sists. For temporal parts considered as time-slices can have spatial
extension (it is in fact important for my later purposes that they do)
but spatial parts should not be thought of as necessarily having
temporal extension, as objects (though not their spatial parts) can
move continuously through space.
62 Change, Cause and Contradiction

I now want to consider the objections of section 2 in tum, but in


order to assess them, we should first distinguish three theses
which are often conflated in discussions of temporal part doctrine.
The theses may be characterized as follows:

Minimal thesis: It is not the case that a temporally extended object is


wholly located (i.e. has its entire temporal extension) at each of the
many times which constitute its temporal extension. In this respect
time is like space: a spatially extended object is not wholly located
(i.e. has its entire spatial extension) at each of the many minute
places which constitute its spatial extension.

Formal thesis: In any formal representation of a statement concern-


ing a temporally extended object, the time determinant will qualify
the subject. Thus 'Reggie was fat in 1972' will be represented:
'a-at-t is F' or 'F(a-at-t)'.

Epistemological thesis: Our ordinary concept of a temporally ex-


tended object is derived from that of certain relations between
temporally unextended objects. Temporal parts are 'epistemologi-
cally basic'.

How are these theses related to each other? Note first that the
minimal thesis does not entail the formal thesis, since there are
alternative notations which allow the necessary time determinant
to enter in ways other than those that make it qualify the subject.
At least four other possibilities appear to be available. The time
determinant could enter with:

(i) A sentential operator: At t(a is F)


(ii) The predicate: a is F-at-t
(iii) An adverb: a is t-ly F

Alternatively, the Tenseless theorist could adopt a view of proper-


ties as dyadic relations, with objects and times as relata:

(iv) R(a, t)

The decision as to which view to adopt is not just a decision over


notation, but over genuine philosophical alternatives to an on-
tology of temporal parts. I shall consider the consequences of such
a decision in the next section.
Temporal Parts 63

Second, the minimal thesis does not entail the epistemological


thesis. Indeed, even the stronger formal thesis does not entail the
epistemological thesis: even if temporal parts are ontologically
basic, there is no reason to suppose that they must be epistemologi-
cally basic too. Following Quine's strategy (Quine (1953), Ch. 1),
one shows a certain kind of thing, F's, to be ontologically basic by
showing (a) that it is an essential component in the truth con-
ditions of ordinary discourse (analysed to a suitable degree of
precision), and (b) that there is no tenable reductionist theory ofF's
which construes them as logical constructions out of some other
(perhaps less dubious) kind of entity. Now it is completely im-
plausible to suggest that we are aware from the first of the existen-
tial commitments of our discourse and thought. As (a) suggests,
such commitments may only emerge after a considerable degree of
analysis.
Consider now the objections to temporal part doctrine. Objec-
tion 1 was that the putative temporal parts of objects are causally
non-explanatory. The spatial parts of a watch are clearly defined:
each part can affect and be affected independently of the other
parts. Not so the temporal parts of the same watch: they are always
causally interdependent. The consequence is that temporal parts of
objects are neither logically nor epistemologically basic- we do not
first grasp the idea of a temporal part and then from this construct
the notion of enduring things. Objection 2 was that, if objects had
temporal parts, then there would be no such thing as genuine
diachronic identity, but rather the obtaining of certain relations
between different objects (viz. temporal parts). Given the first
objection, however, such a reconstrual of diachronic identity will
be circular. These two objections are related: they stem from the
view that (some) spatial parts of objects are objects in their own
right (my hand), whereas the temporal parts of those same objects
never are. This has been denied by some temporal part theorists,
who insist that, for example, every temporal part of a table is itself
a table (e.g. Noonan (1985)). But then there is, surely, something
repugnant to the intellect in the idea of a table being composed of
many instantaneous tables each going out of existence and making
way for its successor. 4
I believe that a temporal part theorist who defines his position
just by the minimal thesis can remain untroubled by each of the
objections above. They are effective against an interpretation of the
temporal part thesis as a formal or epistemological thesis. Take
objection 1 above. I have already remarked that the minimal thesis
64 Change, Cause and Contradiction

doesn't entail the epistemological thesis, so the fact that we do not


first grasp the concept of a temporal part and then construct the
notion of enduring things is no problem for the minimal theorist.
To say that things have temporal parts is not necessarily to say that
there are ways of individuating them which don't rely on men-
tioning times. The objection also rests upon a certain conception of
spatial parts. Chair legs may be capable of being individuated
independently of reference to a chair, but space-slices of a chair are
not. In objection 2, it is the formal thesis which causes the prob-
lems. To replace singular terms standing for temporally extended
objects by singular terms standing for temporal parts, as the formal
thesis does, is to treat temporal parts as objects in their own right.
To treat them as such threatens to obscure the concept of dia-
chronic identity by making it a complex relation between temporal
parts.
Consider now objection 3, which pointed to a conflict between
the ontology demanded by phenomenological discourse on the
one hand, and the four-dimensional ontology on the other. I take
this to be another (perhaps more sophisticated) version of intuition
(c) mentioned in Chapter 1.2. This was the intuition that the object
we encounter at a time is the whole object. The more sophisticated
version of this interprets it as a thesis about the object of percep-
tion, namely that we see three-dimensional objects, not four-
dimensional objects. Moreover, we see the whole object, not just a
temporal part. It seems to me, however, that this last statement, at
least, is questionable. What is meant by the phrase 'seeing the
whole object'? One thing it could mean is that we see something
which has temporal extension - though of course we do not
actually see its temporal extension. When we meet an object, say
the adult Russell, in 1912, we see him as an adult aged 40. We do
not see him in addition as a child or an old man. But then what we
actually see could just as consistently be described as only a
temporal part of Russell - that is, something which has temporal
location, but no temporal extension. If we take this line, then we
remove the disanalogy with space: we only see a spatial part - or
rather a number of parts- of the island when we land on it, not the
whole of it. This helps to give content to the minimal thesis of
temporal parts, for we can interpret it as being at least partly about
what we perceive at any one time: we see an object with no (or
perhaps just minimal) temporal extension. How we interpret what
confronts us in perception is determined by our stance vis a vis
temporal parts, not the other way around.
Temporal Parts 65

Perhaps objection 3 should be rephrased so as to concern, not


the ontology of phenomenological discourse, but rather the on-
tology of intentional thought. That is, when we meet Russell, the
object of thought cannot be only a temporal part, for as Moore (1918)
pointed out, the object of thought (and reference) may always be
greater than what is in the perceptual field. However, if this is a
reason to give up the notion of temporal parts, then we should also
give up the analogous notion of spatial parts, since here there is a
precise analogy with space (Moore was in fact thinking of the
spatial case): however we encounter the island, we can think of the
island as a whole, not just (say) this part of a beach. We would only
be impressed by this objection if we had accepted the epistemo-
logical thesis of temporal parts.
Given these remarks, it will come as no surprise when I say that I
regard Tenseless theory as committed only to the minimal thesis of
temporal parts, and therefore can escape objections 1-3. Objection
4, however, is rather different. If it is correct, then it threatens even
the minimal thesis. The objection is that, if objects are really
four-dimensional entities with temporal parts, then the important
distinction between objects and processes vanishes. Traditionally,
proponents of temporal part doctrine seem to have swallowed this
supposed consequence quite readily. Here is Russell on the topic:

Instead of a permanent piece of matter, we now have the


conception of a 'world-line', which is a series of events con-
nected with each other in a certain way. (Russell (1927), p. 244.)

And later, Quine:

physical objects, conceived . . . four-dimensionally in space-


time, are not to be distinguished from events or, in the concrete
sense of the term, processes. (Quine (1960), p. 171.)

But what argument is there for the supposed entailment between


the doctrine of temporal parts on the one hand and the identifi-
cation of objects with processes on the other, other than the fact
that two distinguished philosophers held both these positions?
One argument might go as follows: 'the temporal parts of an object
(if such there be) would be causally related to each other. How-
ever, since only events are related as cause and effect, the four-
dimensional object is nothing other than a series of causally related
events, in other words, a process'. But why accept that only events
66 Change, Cause and Contradiction

can be related as cause and effect, if 'events' here implies change?


Surely two unchanging states of affairs can be related as cause and
effect - it is certainly true that conditional relations can hold
between such states (If A had not obtained then B would not have
obtained). Indeed, many cases of static causation are put forward
as examples of simultaneous causation, such as a ball's causing a
depression in the cushion upon which it is resting (see Chapter 6).
In any case, there are differences between objects and processes
which surely survive the reconstruction of the former as four-
dimensional. Take the case of an object which remains completely
changeless for the whole of its existence - how could this possibly
be identified with a process, which involves continuous change?
Further, processes are said to possess their properties essentially,
whereas objects could lack some (if not all) of their properties
without ceasing to be those very objects. It is hard to see why such
a difference would be threatened by a thesis about temporal parts.
I therefore beg to resist a distinguished tradition and retain the
intuitive distinction between objects and processes within the
temporal part camp.
That completes my survey of the objections to temporal parts.
Before moving on, I want to consider how our discussion of parts
is affected by Special Relativity. So far, I have been talking of
temporal parts and spatial parts (considered as space-slices). How-
ever, on Special Relativity, a more fundamental notion is that of a
space-time part, defined as the intersection of an object with a
space-time point. This can replace our intuitive notion of a spatial
part, for this was just the notion of a temporally unextended
space-slice. Since I still want to retain the distinction between
temporal and spatial parts, I propose to redefine temporal parts as
follows (readers may choose whether to adopt this or the standard
definition; it does not affect later argument). LetS be the collection
of space-time points which constitute the space-time extension of
some object 0. We define a spatially maximal subset, S', of S as a set
which is such that all the members of S' are both also members of S
and are separated from each other by a space-like interval, and
there is no point which is (a) within S, and (b) not within S', and (c)
separated by a space-like interval from any of the members of S'.
Clearly S will contain a number of such sub-sets. A temporal part of
0 may then be defined as the intersection of 0 and some maximal
sub-set of S. The point of introducing the notion of a spatially
maximal sub-set of 0 is that we can retain the intuitive idea that a
Temporal Parts 67

temporal part of x, x-at-t, has the same spatial extension as would


ordinarily be attributed to x at t.
I said earlier that the minimal thesis of temporal parts did not
entail the formal thesis, because there are alternative notations to
the temporal parts notation. The Tenseless theorist who wishes to
avoid some of the objections to temporal parts may then avail
himself of these alternative notations. It is still incumbent upon me
to demonstrate that there are indeed workable alternatives to a
temporal part notation. This I shall now do.

4.3 THE PROBLEM OF THE TEMPORAL DETERMINANT

Crucial to a fully formalized theory of change is a satisfactory


formal treatment of the way in which temporal terms modify
predicative sentences. I shall not come down firmly in favour of
one notation, but I shall indicate some grounds for preference.

(i) Temporal operators


In Prior's tense logic, the time-determinants enter as tense oper-
ators on core present tense propositions (see Chapter 3). Given the
Quinean contention that the primitive logical terms of a theory
should reflect the metaphysical outlook of that theory, a natural
development of a logic for Tenseless theory would be the inclusion
of tenseless temporal operators with quantifiers ranging over times
(compare David Lewis's quantified modal logic (1968)). So one way
of resolving the prima facie contradiction between 'a is F' and 'a is G'
(where F and G are incompatible predicates) would be:

(Let it be assumed for all succeeding formulations in section 3 that


t1 is non-identical to t2 .) This way of introducing the time determi-
nant has recently been criticised by Mark Johnston (1987), who
thinks that if such an account is to be successful,

then it had better be that at most one of the abstract represen-


tations [t1 and~] correspond to the concrete reality that makes
sentences simply true or simply false. For suppose that both did
so correspond. Then it would be simply true that z is G and that
z is not G and we would not have advanced. (Compare saying
68 Change, Cause and Contradiction

that Jones is crooked according to the Times and honest accord-


ing to the News and that both papers correctly represent the
concrete facts of the matter.) (Johnston (1987), p. 114.)

This objection could equally be directed against a Tensed theorist


who, like Prior, introduces the time-determinants by tense oper-
ators. Prior himself avoids the objection because he assigns a
unique ontological status to a single time, viz. the present (Prior
(1970). (See Chapter 3)). The present tensed 'Fa' corresponds
directly to reality, and so is made true simply by a's being F. In
contrast, 'It was the case that Ga' does not directly correspond to
concrete reality. It is made true, not by the state of affairs at some
past time, but by the state of affairs at the present time. There is no
Tenseless analogue of Prior's position, of course, as there would be
no sense in assigning a unique ontological status to a tenselessly
individuated time as a general thesis.
Johnston's objection is likely to strike us as odd as it stands. 'At
t1(Fa)' and 'At tAGa)' cannot concern the same part of reality other-
wise the temporal operators would be redundant. What the oper-
ators do, surely, is to select different time-slices of reality, of which
it is true, respectively, that 'Fa' and 'Ga'. However, this raises a
question about what is identified by 'a' when it occurs within the
scope of a temporal operator: is it the temporally extended (four-
dimensional) object, or a time-slice of that object? If the former,
then 'At t1 (Fa)' and 'At t 2 (Ga)' do indeed attribute incompatible
properties to the same object. If the latter, then there is an implicit
time determinant qualifying the subject, and so a more perspicu-
ous formulation would be:

which appears to leave the temporal operators redundant. Hence,


a logic of temporal operators is not, it seems, an alternative to a
logic of temporal parts. However, we may invoke the distinction
made in the previous section between the minimal and formal
theses of temporal parts to block Johnston's objection. 'At t(Fa)'
succeeds in ascribing a property to a part of a without introducing a
term which essentially denotes that part (in the scope of a different
operator, it would denote a different part). Thus, contra Johnston,
it is the temporal operator notation which makes the temporal part
notation redundant.
Temporal Parts 69

(ii) Relativised predicates


According to the second alternative to temporal part logic, the time
determinant enters with the predicate:

This formulation is recommended by Wilson (1955), who suggests


that 'substance talk' (i.e. talk of three-dimensional continuants)
can be derived from 'space-time talk' (i.e. talk of four-dimensional
space-time objects) by allowing the time determinant to 'cross the
copula' from the subject to the predicate. Wilson's assumption is
that 't' is tenseless, and it is this that Prior objects to in his
discussion of Wilson's paper (Prior (1976), p. 80). If the time
determinant is tenseless, says Prior, then there is no difference
between qualifying the predicate and qualifying the subject (i.e.
introducing a temporal part notation). Prior illustrates this by way
of analogy:

A person, call him Owen, who is colour-blind, might see this


leaf as green when it is in fact red. This leaf, we might then say,
looks green to Owen . . . some philosophers would want to
replace this by 'is green-to-Owen', abolishing the explicit refer-
ence to looking or seeming in the same way as Wilson abolishes
the explicit tense. But as in the other case, the supposed abol-
ition is only a disguise. 'To Owen' is intelligible as an answer to
the question 'To whom does it look green?'; as a description of
the kind of green the leaf is, 'green-to-Owen' is just nothing at
all, unless, indeed, we mean by calling it 'green-to-Owen' that
the leaf is green where it is turned towards Owen - green on that
side of it. And we might give 'green-in-August' a similar sense-
we might mean that that part of the leaf which is green in
August . . . is green; but then it cannot really be a leaf that we are
talking about ... (Prior (1976), pp. 80-81).

The argument here is somewhat obscure. It involves two anal-


ogies with time-relativized predicates, one of which is clearly
inappropriate. Prior's intention in this paragraph is to show that 'a
is green-at-t' is just a notational variant of 'a-at-tis green'. But if 'a
is green-to-Owen' is a fair analogue of 'a is green-at-t', then the
perceptual analogue of its supposed notational variant, 'a-at-t is
70 Change, Cause and Contradiction

green' must be 'a-to-Owen is green', where 'a-to-Owen' is presum-


ably either a sense-datum, or a spatial part of a (i.e. the part turned
towards Owen). If the former, then the statement has little, if
anything, to do with a. So Prior must presumably be read as
attempting to place the proponent of relativized properties (e.g.
Wilson) in a dilemma: either 'a is green-at-t' fails to ascribe a property
to a at all, or, if it succeeds, it only does so because it ascribes a
non-relativized property, green, to a temporal part, a-at-t.
Let us take a look at both horns of this dilemma.
We may concede to Prior that being green-to-Owen (in the sense
of looking green to Owen, but perhaps not to others) is not, or at
least not necessarily, a way of being green, but this does not make
it no kind of property at all. It is the property of being such that
when Owen views it under certain conditions he has a certain kind
of experience. Now this is not very far, in fact, from certair.
philosophers' definition of colour, especially those philosophers
impressed by the primary/secondary quality distinction. Don
Locke, for example, defines the actual (rather than merely appar-
ent) colour of an object 'as the colour seen by a normal percipient
in standard conditions, i.e. in broad daylight.' (Locke (1967),
p. 99). Now suppose 'Owen' was just another name for 'normal
percipient in standard conditions', and not for a colour-blind
person. Then 'green-to-Owen' would just be the conventionalist's
definition of green. I do not say that the conventionalist analysis of
colour and other secondary qualities is right, but the possibility of
such an analysis shows that Prior cannot simply offer as an unde-
fended assumption the view that 'green-to-Owen' cannot be the
description of a property. So it won't do, by itself, as an argument
for the view that 'green-at-t' isn't intelligible as a description of a
property. One final point: Prior's intention may be to show merely
that 'green-at-t', like 'green-to-Owen', is not an intrinsic property.
But an intrinsic property of x is one whose description does not
involve essential reference to any material object or state of affairs
outside x. Now green-to-Owen is clearly not an intrinsic property
in this sense, but it is far less clear that green-at-t can't be an
intrinsic property, for a time is not (at least outside a particular
relationist account) a ordinary material object or state of affairs.
If, on the other hand, 'green-to-Owen' is interpreted in a spatial,
rather than in a phenomenological, sense, then it is at least a
legitimate analogue of 'green-at-t'. It would be even more clearly
Temporal Parts 71

so if 'Owen' were replaced by a term denoting a place, as in


'green-at-s'. Is 'F-at-s(a)' just a notational variant of 'F(a-at-s)'?
Consideration of the following case would suggest not. Consider a
plank, a, which is situated between places s0 and s2 (its midpoint
being at s1 ), and which exhibits property F throughout its spatial
extension. At a given time (which for the purposes of simplicity we
will not include in the notation) it is true that

It is also true that, in the rival notation,

1'. F(a-at-s1)

Moreover, 1 and 1' would seem to have the same truth-conditions.


Indeed, if a given relativized predicate formulation is logically
equivalent to the corresponding relativized subject formulation, it
follows that the truth-conditions of one will be the truth-conditions
of the other. But the case we have imagined just now provides an
instance where supposedly equivalent formulations do not share
the same truth-conditions. For as we have described the case it is
true that

where S3 is a place as far from s2 as s2 is from s 11 so that the three


places form a straight line. It is also true that

2'. -F(a-at-s3 )

Ex hypothesi, 2 and 2' are equivalent, but do they share the same
truth-conditions? Arguably not, for 'a-at-s3 ' is a non-denoting
term, whereas 'a' in 2 is not. The case has a straightforward
temporal analogue: an object a exists from to to ~' and is then (just
before ~) destroyed. During its entire existence, it is F. We can
therefore say that

and also, but for different reasons,


72 Change, Cause and Contradiction

But we can also say, since 'a-at-~' fails to denote, that for any
property G, it is not the case that a-at-~ is G. Whereas we cannot
say that for any property relativized to~' it is not the case that a is
F-at-t3 • For suppose the property is that of being famous-at-~: this
is an extrinsic property, and a of course can be famous at~' even if
it doesn't exist at that time. But a-at-t3 can be nothing at all, not
even famous. So here is a case where a statement containing a
relativized subject (representing a temporal part) is true, but its
supposed equivalent, containing a relativized predicate, false.
What can be salvaged from Prior's criticism is that allowing the
time determinant to qualify the predicate obscures the type ident-
ity between F-at-t1 and F-at-t2 • This is certainly true. A related
objection is that there can apparently be change in the temporally
relativized properties of an object without there being any change
in what we would regard as the intrinsic properties of that object.
Just by being temporally extended, that is, a leaf can 'change' from
being green-at-t1 to being green-at-~. This objection, however,
does not strike me as decisive. If an account of change were to put
itself forward as a complete account, then the objection that it
attributed change to situations where there was none would in-
deed be decisive. But a logical notation for temporal attributions of
properties is not in itself a complete account of change. It will be a
further challenge to our final account of change to exclude bogus
change from one temporally relativized property to another.
(iii) Properties as relations
On the account put forward by Mellor (Mellor (1981), pp. 110-12),
properties are construed as two-place relations whose relata are a
subject and a time. (The properties may be regarded as three-place
if one also includes places among the relata.) So a's being F and G
at different times may be represented:

This is objected to by Johnston, as follows: on the relational


account of properties, objects which would normally be considered
duplicates in having the same composition, size, shape, colour,
temperature, etc., are nevertheless judged to have distinct proper-
ties in virtue of occupying different positions in time (and space).
'Hence only confusion can come from building in either spatial or
Temporal Parts 73

temporal position as one of the relata of predicates supposed to


express intrinsic properties.' (Johnston (1987), p. 113.) But this
objection is suspect. It is just not the case that token spatial or
temporal position is built in to the relata of type predicates express-
ive of properties (which is the point Johnston is surely making
here). It is certainly true that the relational account must attribute
to the duplicates different token properties, but then any plausible
account must do this. The account is only in trouble if it attributes
different type properties to the duplicates, but it is unclear why it
should be committed to this. Why should '~(a, t1 )' and 'RF(a, ~)'be
thought to concern different type properties? The times here are
surely irrelevant in the characterization of the type property, as
indeed is the object which stands as the other relatum.
Perhaps the main prima facie difficulty with the relational account
is that it seems counterintuitive - 'Simply incredible', Lewis
((1986), p. 204) calls it, because he takes it to be denying the
existence of intrinsic properties. But it needn't deny their exist-
ence. If we characterize type-intrinsic properties as those whose
individuation does not involve essential reference to particulars,
then a type property like 'red' can still be intrinsic on the relational
account, for its individuation does not involve essential reference
to a particular time, as we saw above. I suspect that Lewis's
incredulity stems from the same mistake as the one Johnston
makes.
However, there are two genuine problems for the relational
account - one of them quite serious - and I discuss them below.
(iv) Adverbs and predicate modifiers
The account which Johnston himself favours lets the time deter-
minant enter with an adverb. This is suggested by modal state-
ments such as 'a is actually F but possibly G.' The precise temporal
analogue of this would be the tensed 'a is presently F but was/will be
G.' The tenseless version is the less natural

a is t 1-ly F & a is ~-ly G


This certainly appears to fit the bill. If we think of ordinary
adverbs, like 'slowly' and 'grotesquely', we can reflect that run-
ning slowly and running grotesquely are ways of running, and that
they are ways of running that are performed by individuals, not
parts of individuals, and that 'Augustus is running slowly' does not
contradict 'Augustus is running grotesquely'.
74 Change, Cause and Contradiction

This account of time determinants is (with minor notational


differences) identical to that presented by Romane Oark, in his
paper 'Concerning the Logic of Predicate Modifiers' (1970). There
are good reasons for preferring this account over the foregoing
accounts, for it satisfies two criteria for notation involving temporal
(and other) qualifiers. The first criterion, suggested by Kenny
((1963), Ch. VII) and taken up by Davidson (1967), is that a
notation should show as valid ordinary inferences from 'Reggie
was lurking grotesquely behind the conservatory at four o'clock' to
'Reggie was lurking at four o'clock' and to 'Reggie was lurking',
etc. The second criterion, considered by Prior ((1967), p. 170), is
that a notation should preserve the surface features of sentences in
which temporal expressions qualify singular terms, e.g. 'Reggie
was fatter in 1972 than Edna in 1980'.
Kenny's criterion raises difficulties for a relational account of
properties. If 'Reggie was lurking behind the conservatory at four
o'clock' is formulated using a three-place relation:

(a) L(a, s, t)

and 'Reggie was lurking at four o'clock' is formulated using a


two-place relation:

(b) L*(a, t)

then the validity of the inference from (a) to (b) has simply been
obscured by the notation. Similarly, the case of Reggie's being
fatter in 1972 than Edna in 1980 poses a problem for the relational
account, necessitating the introduction of peculiar entities, such
as, in this case, girths, or degrees of fatness, represented here by a
and~:

Fat(a, Reggie, 1972) & Fat(fl, Edna, 1980) & Greater(a, fl)

The same move will have to be resorted to by the temporal


operator account. It is not a satisfactory move, as when we make
comparisons between people at different times (or the same person
at different times) we are surely talking directly about ordinary
objects, not such strange entities as girths, or heights, or person-
alities. Now the temporal part notation, for all its faults, at least
retains this intuition. Admittedly, the temporal part view has one
Temporal Parts 75

comparing temporal parts of persons, rather than persons, but at


least temporal parts are, unlike girths, concrete objects. So we
have:

Fatter(Reggie-in-1972, Edna-in-1980)

Now Oark has pointed out that in order to meet Kenny's criterion,
one does not have to resort to Davidson's ontology of events
('there exists an event such that e is lurking and e is grotesque and
e is an act of Reggie's ... etc.') One can, instead, treat adverbs and
temporal and spatial modifiers as predicate modifiers:

At(Behind(Grotesquely(Lurks(a))p)t

From which one can validly infer:

At(Lurks(a))t

Recently, Butterfield and Stirling (1987) have shown how predi-


cate modifiers can be introduced into a temporal logic in order to
deal with Reggie, Edna, and their respective degrees of bulk. Their
preferred notation would provide the following formulation:

(1972xReggie) & (1980xEdna) & Fatter(x, y)

In Clark's notation, the formulation would be:

Fatter((At(Reggie)1972), (At(Edna)1980)

Thus the introduction of predicate modifiers allows us to capture


the surface features of such temporal comparisons. Since it also
gives us an elegant way of dealing with Kenny's problem without
having to follow Davidson in appealing to an ontology of events, it
seems to be a suitable notation for the Tenseless theorist in search
of a formal representation of his theory of change.
However, a Tenseless theory of change is what we still lack. So
without further delay, let us turn to this task.
5
Tenseless Change
5.1 CONSTRAINTS ON AN ACCOUNT

Let us begin by rehearsing what we should require of an adequate


account of change.
(a) It should, first and most importantly, be a Tenseless account. It
must not invoke the 'passage of time', a non-relational A-series, or
any other tensed fact. We left such things behind, once and for all,
at the end of Chapter 3.
(b) It should explain, in a non-trivial way, why time, not space, is
the dimension of change. That is, it should not define change even
partly in terms of temporal variation in a thing's properties. The
fact that change involves time should rather follow as a substantial
consequence of a non-temporal account.
If both (a) and (b) can be met, we shall have some conception of
how the Tenseless theorist should represent the difference be-
tween time and space.
(c) It should exclude Cambridge change - i.e. whatever brings
about variation in the truth-values of tokens of the same sentence-
type without an associated real change in the object those tokens
refer to; for example, the 'change' in Ouspensky when he became
famous. Ideally, it should exclude this pseudo-change in a way
which is at least related to its means of excluding purely spatial
variation. It would then be clear what particular characteristic of
real change is lacked by all other kinds of variation.
(d) It should be at least consistent with the minimal thesis of
temporal parts. This follows from (a) above and the argument of
Chapters 1.2 and 4.2 that Tenseless theory is committed to tem-
poral parts. The account may, but need not, explicitly refer to
parts. If it does, it should not explicitly link change with variation
in temporal parts.
These points have already been made. We might now make
some additional requirements:
(e) It should explain why things, but not events or processes,
change. Of course, if the account simply defines change in terms of
76
Tenseless Change 77

variation in a thing's properties, then we can trivially exclude


change in events and processes. But we should allow for cases of
change which are not changes in the properties of particular things
at all (consider the change brought about by the collision and
consequent annihilation of a particle and anti-particle). So we
should aim to provide, in addition to an account of changes in
things, an account of changes in regions of space. This second,
more general account, must be able to provide some account of
why, although events/processes are changes, there is no such thing
as change in an event/process. A consequence of meeting this
requirement would be a refutation of the common suggestion that
Tenseless theory- in particular, theories which adopt temporal
parts - must regard things as extended processes. If things change
and processes don't, they must have different identification con-
ditions.
(f) It should explain the asymmetric nature of change: the fact that
change is change from something to something else. Again, re-
membering requirement (a), this asymmetry cannot simply be
explained in terms of the asymmetry of temporal priority.
(g) It should not be tied to the standard (i.e. linear) topology of
time. It may not be obvious that this is a requirement of any theory
of change, but it will emerge in the next chapter that certain
accounts of change (including the embryonic account offered at the
end of this chapter) fail for worlds where time has a closed struc-
ture. This requirement is also intended to capture the intuition that
whether a thing is changing or not does not logically depend upon
facts about the structure of the time-series.
Unless I'm very much mistaken, no theory of change yet pro-
posed meets all these criteria (and some meet only one or none).
My contention is that they can be met, and in meeting them I shall
be led into explanations of important features of the world which
might be thought to have little connection with change.

5.2 FIRST ATTEMPTS

We may take as our starting point Russell's definition of change,


from The Principles of Mathematics:

Change is the difference, in respect of truth or falsehood, be-


tween a proposition concerning an entity and a time T and a
78 Change, Cause and Contradiction

proposition concerning the same entity and another time T',


provided that the two propositions differ only by the fact that T
occurs in the one where T' occurs in the other. (Russell (1903),
p. 469.)

It is clear that this definition fails to meet our requirements. To


begin with, the introduction of times into the definition makes it a
merely trivial truth that time, not space, is the dimension of
change. A further problem concerns the reference to 'an entity'
and 'the same entity'. Is Russell referring to the whole, temporally
extended entity, which is first F and then not F? If so, it cannot be a
truly Tenseless definition, for, as was argued in the previous
chapter, Tenseless theory is committed to the minimal thesis of
temporal parts. Now it may be that Russell intends his definition to
be neutral, and indeed his definition is one of change in general
rather than change in an object. Nevertheless, Russell must be
thinking primarily of change in objects, for otherwise why would
he mention an entity at all? If we are interested in providing
conditions for change in objects, then, to be completely explicit, we
should phrase it in terms of parts of objects. (This does not, as
might appear, invoke the formal thesis of temporal parts, for
statements of natural language which mention parts may be re-
formulated in terms of, e.g. predicate modifiers. Natural language,
however, lends itself more easily to talk of parts than of predicate
modifiers.) We may remark, finally, that it is prima facie odd to give
a definition of change in terms of propositions and truth-values,
and in what follows, I drop the reference to them.
In accommodating these points, we may begin with something
like the following:

Object 0 changes iff for some incompatible properties F and G,


one part of 0 is F, and another part is G

There could hardly be a broader definition of change than this. It


allows purely spatial variation (since 'part' in the definition is
neutral between spatial and temporal part), and it allows in Cam-
bridge changes, such as a piece of antique furniture 'changing' as a
result of its price going up. This is not surprising. Russell's defi-
nition was one of those Geach had in mind when he coined the
phrase 'Cambridge change' (so called because it 'keeps on occur-
ring in Cambridge philosophers of the great days.') 1 Although only
Tenseless Change 79

real changes satisfy Russell's definition (apart from mythical A-


series change), the moment we turn it into an analysis of change in
objects, we shall naturally have something which is satisfied by
pseudo-change. A further complication is the fact that not all kinds
of change involve a single entity, or the same entities throughout
the change, or indeed any entities at all. Cases of change not
covered by the definition above include the annihilation of an
object, the fusion of two objects, the passage of light through a
vacuum, the mixing of a number of gases. In response to such
non-standard changes, Newton-Smith has proposed the following
disjunctive definition:

A change occurs if and only if some object has at some time some
property and later that object lacks that property or some region
of space is characterizable in such and such a way . . . and later
that same region of space is not so characterizable. (Newton-
Smith (1980), p. 15.)

The first two remarks concerning Russell's definition apply here


too, of course. A simple way to include non-standard changes
without recourse to a disjunctive account is that of replacing refer-
ence to objects, spatial regions, and type properties, by reference
to spatio-temporal regions and instantiations of properties - to
give, in other words, a property-, rather than an object-orientated
account. This will be resisted by those who hang on to the intuition
that change essentially involves some continuing entity which
both undergoes and survives the change, but given the cases
above, this intuition may not be well-founded. In one sense, the
Tenseless theorist is committed to rejecting the intuition anyway,
since he wants to allow that variation amongst temporal parts
counts as genuine change. I shall elaborate a property-orientated
account later. I now want to canvass a proposal for excluding
spatial variation.

5.3 THE CAUSAL CONSTRAINT

Consider the purely spatial variation in temperature between the


ends of a poker. At any given instant, there is no causal connection
between the current temperature at one end of the poker and the
temperature at the other end, and this, I want to say, is what
80 Change, Cause and Contradiction

prevents spatial variation from being change. If there is a genuine


change in an object from its being in one state to its being in
another, there is also a causal connection between the two states, where
'causal connection' may be defined as follows: 'for any two states a
and b, a is causally connected to b iff there is a causal chain c1 • • •
ex such that c1 is a state of affairs partly constitutive of a and ex is a
state of affairs partly constitutive of b, and such that for each n
(where n>l), en is causally dependent upon cn_1 .' I call this the
causal criterion of change.
I do not wish to say that the earlier state in any change is what
we would ordinarily call the cause of the later state. For example, if
a moving ball suffers a change in its velocity when it strikes
another object, then it is natural to say that the cause of the change
in its velocity is the intervention of the other object, rather than the
ball's earlier velocity. Nevertheless, its earlier velocity is relevant in
determining its final velocity: it is a causally relevant factor.
Further, if the ball had not struck the object, then its final velocity
would have been different, and we justify this counterfactual by
reference to the earlier state of affairs. So the initial states of things
in any change figure in a causal explanation of the later states of
things. Now there are certain cases of genuine change which seem
to provide direct counterexamples to this. For example, suppose a
white wall is painted red. This is a genuine change in the colour of
the wall, but there is, it would seem, no causal connection between
the later colour of the wall and the earlier colour. Surely the wall
would have been red after painting even if it hadn't been white
before. However, it is undeniably the case that an essential con-
stituent of the wall's earlier state of affairs (and indeed of its later
state of affairs) is simply the wall's existence, and there is certainly a
causal connection between the wall's simply being there at the
earlier time and its being there at the later time. Admittedly, as was
mentioned earlier, there are cases of change which do not involve
changes to a single continuant. But if all we have in a particular
case is two causally unconnected states of affairs, what motivation
is there for considering the case to be one of change?
The causal criterion of change may be compared to Reichen-
bach's (1924) causal account of time order. 2 The difference is that
Reichenbach invokes possibilia: event E1 is later than E2 iff it is
physically possible that there be a causal chain S 1 • • • sk, such that si
is causally dependent upon si_1, and such that s 1 and sk are coinci-
dent, respectively, with E1 and E2 • I am somewhat suspicious of
Tenseless Change 81

this modal criterion, not because it is a modal criterion (and so


raises questions of ontology), but because the question arises of
what actual fact grounds this physical possibility. It is very easy to
reply that the relevant fact is that the two events are temporally
separated, and so involve oneself in an explanatory circle. So I
prefer the stronger criterion of actual causal connection for genuine
change, without committing myself to such a criterion as a necess-
ary, rather than merely sufficient, condition for time order (see
Chapter 8 for further discussion).
The causal criterion of change, then, guarantees that time, not
space, is the dimension of change, for there must be a temporal (or
time-like) separation between causally connected events or states
of affairs. In fact, I want to say something stronger than this,
namely that if A is causally prior to B, then A is earlier than B. 3
Even this simple criterion of change, however, faces two major
objections.
The first is this. It is controversial simply to assert that there
must be a temporal separation between cause and effect, for there
appear to be cases where cause and effect are simultaneous. Con-
sider, for example, an electric current passing through a wire. It is
surely quite plausible to regard this as simultaneously causing the
presence of a magnetic field around the wire. To rule out such
cases by fiat is unacceptable. The causal criterion for change could
only be fully justified on the basis of an a priori argument against
the possibility of simultaneous causation. It is not enough, note, to
say simply that, as a matter of fact, simultaneous causation does
not occur, for an adequate account of change must extend to
purely possible change.
The second objection is that the causal criterion is really no
improvement on Russell's account, which simply builds the notion
of temporal separation into the definition. For many analyses of
causality essentially involve reference to temporal priority: that is,
they make it a necessary condition of a's causing b that a be prior to
b. Judging by the difficulties philosophers have had in attempting
to explain the asymmetric nature of the causal relation, it seems
reasonable to conclude that causal priority must be defined (in
part) in terms of temporal priority. If this is right, then the causal
criterion of change does indeed essentially involve (implicit) refer-
ence to temporal separation, and so makes trivial the assertion that
time is the dimension of change.
Of course, no one could accept both of these objections. If causal
82 CJumge, Cause and Contradiction

priority entails temporal priority, then simultaneous causation is


impossible. However, each objection seems plausible, and each
should be met. I attempt to meet them, in tum, in the next two
chapters.
6
Causation and
Simultaneity
6.1 THE PRINCIPLE OF RECIPROCITY

In this chapter I shall present a simple proof that causes are never-
arguably, could never be- simultaneous with their effects, based
upon a principle widely accepted but never, as far as I have been
able to discover, commented upon in philosophical treatments of
causation. I shall call it the Principle of Reciprocity! (hereafter 'R'),
and it may roughly be stated as follows: a necessary part of any
cause is itself affected as a direct result of that cause's bringing
about its effect. To the question 'affected by what?', the answer is
'affected, not by the effect itself, but by that upon which the cause
acts in order to produce the effect.' I shall consider concrete
examples which should make this rather abstract statement
clearer. I hold that the supposition of simultaneous causation is
inconsistent with R.
Plausible cases in the literature in which cause and effect appear
to be simultaneous are numerous. They include: the moving of a
locomotive engine causing the movement of the carriages attached
to it; one end of a see-saw's going down causing the other end to
go up; the increasing of a gravitational field causing the bending of
a light beam; and Kant's famous example of a leaden ball's causing
the depression in the cushion upon which it is resting. 2 My interest
in such cases is that they threaten my causal criterion of change
(Chapter 5.3), but they also appear to provide direct counter-
examples to otherwise plausible theories of time which make
causal priority the criterion of temporal priority (see Chapter 8.2).
So any argument which succeeded in demonstrating the universal
non-simultaneity of cause and effect would be, I take it, of no small
significance.
Quite apart from its role in the proof I shall offer, R is of both
historical and philosophical interest. So I shall begin by tracing
very briefly some historical antecedents of R before going on to a
more precise formulation of the principle.
83
84 Change, Cause and Contradiction

6.2 HISTORICAL ANTECEDENTS OF R

A principle related to R as I shall eventually formulate it is one


Jonathan Bennett calls the 'Balance Principle'. 3 As he states it, it is
the principle that 'in any fully intelligible causal transaction some-
thing gains what something else loses'. (Bennett (1974), p. 59.) Now
this could be one of two quite different principles governing the
behaviour of two continuants x and y which causally interact with
each other: (A) for some value ofF, x becomes more F while y
becomes less F, or that x becomes F while y ceases to be F; (B) x
acquires some or all of that very same F-ness which y had. Prin-
ciple (B), which requires the transfer of a token property, is re-
garded by Bennett as being clearly false but more prominent than
(A) in the history of ideas.
Bennett quotes Berkeley's appeal to the 'old known axiom:
Nothing can give to another that which it hath not itself (Berkeley
(1713), p. 331) as being suggestive of the Balance Principle. But
there is nothing in Berkeley's remarks to suggest that the cause
loses anything in bringing about its effect. The suggestion is merely
that cause and effect must always have properties in common. In
the context of his remark Berkeley is in fact concerned to discredit
the notion that 'unperceiving' matter can be the cause of our
perceptions. A clear reference to the Balance Principle, however, is
contained in Locke's distinction between ordinary causal transac-
tions and those that are genuine paradigms of power. The former
are merely energy transfers, as when a moving ball strikes another
and so causes the second to move while slowing down itself. In
contrast, the sun has power to melt wax without transferring any
of its properties to the wax. Genuine powers for Locke are propen-
sities to originate changes in properties (Locke (1689), Il.xxi.4).
Leibniz very briefly considers formulation (B) of the Balance
Principle in his thorough examination of Locke, the New Essays on
the Human Understanding, only to dismiss it as a fantasy. As he
admits, however, it isn't clear that Locke's treatment of ordinary
causal transactions involves (B) rather than, say, (A). (B) is also
criticised in the Monadology, but as part, it seems, of a more general
thesis about the possibility of causal interaction:

Monads have no windows through which anything could enter


or depart. Accidents cannot be detached from substances and
march outside of substances, as the sensible species of the
Scholastics once did. (Loemker (1956), p. 643.)
Causation and Simultaneity 85

This would seem to rule out any causal interaction. However, in a


letter to De Voider, Leibniz writes:

Who will deny . . . that one substance is modified by the in-


tervention of another, as when a body is repelled by some
obstacle in its path? (Ibid., p. 524.)

Hide Ishiguro interprets Leibniz, not as denying causal interaction,


but as rejecting a certain conception of it, namely one embodying
(B). For Leibniz, the correct view of causal interaction is that of a
cause's being temporally contiguous with an independent (though
divinely co-ordinated) change in a substance, in accordance with
the hypothesis of pre-established harmony (Ishiguro (1972), pp.
112-13). But this is not so much a redescription of causal interaction
as a complete rejection of it. Leibniz's point is surely this: genuine
causal interaction would involve the movement of token properties
from one substance to another. Such movement, however, is
impossible, given the nature of monads. (The notion is in any case
incoherent- see Leibniz (1704), pp. 171-2.) Consequently, genuine
causal interaction cannot take place. If this is not Leibniz's reason-
ing, then any mention of (B) in a discussion of interaction would
seem to be something of a red herring.
Leibniz's views on causation did not prevent him from develop-
ing an early conservation principle, of which (A) is a consequence
(but perhaps only for certain values of F). Principles of conser-
vation are in fact entirely in keeping with his insistence on on-
tological parsimony. In the New Essays Leibniz properly criticised
Locke for his assertion that when a moving ball strikes another, the
first loses as much motion as it communicated to the first. But
Locke's suggestion is not quite as wide of the mark as it seems.
'Motion' here is more likely to mean mass multiplied by velocity,
rather than just velocity. What is rather conserved is 'motive force'
(vis motrix), defined as mass multiplied by the square of the vel-
ocity. This principle had been developed by Leibniz in his Brief
Demonstration of a Notable Error of Descartes. The error in question
had been Descartes' own conservation principle, namely that the
'quantity of motion' (where 'motion' = m x v) in the world is
conserved. It seems likely that Locke adopted this (at the time)
generally accepted principle. Leibniz's rejection of it caused some
heated debate. 4
Formulation (B) of the Balance Principle seems, then, to have
been discredited on philosophical grounds early in the eighteenth
86 Change, Cause and Contradiction

century. This is appropriate, as (B) is partly a philosophical prin-


ciple. (A), in contrast, is a wholly empirical principle, cast into
doubt by scientific advances in the early nineteenth century. The
experiments of Rumford and, later, Joule helped to establish that
certain forms of energy (a term suggested in 1807 by Thomas
Young in place of Leibniz's 'vis viva') could be transformed into
others: for example, kinetic into thermal. The type property which
y gains need not, therefore, be the very same as that which x loses.

6.3 THE PROOF OF NON-SIMULTANEITY

I now want to move from the Balance Principle to a correct one,


viz. R, which captures the spirit of the former. Consider first a
variant on the first part of (A) above:

(1) For any A and B, if A's being Fat timet causes B's being Gat
t' then A is no longer F at t'

This is intended to cover a variety of cases. The domain of 'A' and


'B' includes whole continuants, which typically figure in cases of
causal interaction. But the domain also includes parts of continu-
ants, regions of space, electro-magnetic fields, and so forth. The
justification for (1) involves appeal to principles of conservation: in
any causal interaction there will be some parameter which is
conserved throughout the interaction, for example, linear and
angular momentum, electrical charge, or, more generally, mass-
energy. F and G, the properties undergoing reciprocal change,
may, but need not be, instances of this parameter. This much will
at least be the case: changes in F and G will be governed by the
constraint that the relevant parameter must be conserved. Now, if
the appropriate conservation principle is not to be violated,
changes in B must be accompanied by simultaneous changes in A.
For example, if a cold silver spoon is placed in a cup of hot water,
the heating-up of the spoon must be simultaneously accompanied
by the cooling-down of the water. If the changes were not simul-
taneous, then there would either be too much, or too little, thermal
energy in the system.
(1), however, is false. Firstly, not all cases of causation involve
change. There are 'static' cases, such as Kant's ball resting upon,
and causing the depression in, the cushion. So the principle will
Causation and Simultaneity 87

have to be limited to those cases where B was not G immediately


before t'. Secondly, there are perfectly legitimate descriptions of
causal interactions where the stated cause is unaffected by its
bringing about its effect. In such cases, the properties involved will
not be subject to some conservation principle. Consider the follow-
ing substitution instance of (1):

(1 ') If Anne's being angry at t causes Brian's being amused at t',


then Anne ceases to be angry at t'

There is no reason to expect that this will turn out to be true.


Clearly, not all properties undergo change as a result of causal
interaction. But even if we limit (1) to certain kinds of property, we
are still confronted, thirdly, by cases where A's being F is a distant
cause of B' s being G (in the sense of there being an extended causal
chain connecting the two states). In such cases, A may well have
returned to being F by the time B starts to be G, for A may for other
reasons be undergoing cyclical alteration.
We might deal with (1') by requiring a complete account of the
cause. Clearly, in any complete account of how Anne's anger
causes Brian's amusement, there must be reference to properties
which do change. Some of these changes will be accidental, but
others will be intrinsic to the causal process. For Anne's anger to
cause Brian's amusement in this case, he had to perceive her
condition. If he visually perceived her, then a necessary part of the
causal process was that light struck Anne's body, where part of it
was absorbed, causing changes in her surface temperature. The
reflected light then struck Brian's retina where it was partially
absorbed, causing changes in the visual pigments of the retinal
cells, etc. More generally, the relevant properties will be those
exhibited (unlike anger) at the micro-level, and which in addition
will be among those whose behaviour is constrained by physical
laws, including conservation principles. Neither of these charac-
teristics of the relevant properties, however, need be made explicit
in the formulation of R. R itself is not a conservation principle, nor
indeed a physical law, but a formal principle motivated by conser-
vation principles.
We now turn to cases of distant causation which involve cyclical
change in A. One possible means which suggests itself is to limit R
to cases of immediate causation. But this would be unfortunate: if
immediacy is defined in terms of the absence of any state of affairs
88 Change, Cause and Contradiction

between cause and effect, there may be no genuine cases of


immediate causation. That is, causation may be dense in the sense
of there being, between any two causally connected events or
states of affairs, a third event or state of affairs (see Emmet (1988)).
What I propose, therefore, is a more neutral solution. Although in
distant causation, A may have gone back to being F by the time B
becomes G, it will still be the case that at some point in the causal
chain between cause and effect, A will cease to have been F as a
result of its bringing about one of the events in the chain. Taking
all the above into account, then, we are left with the following:

(2) If a causal interaction with some item A at time t causes B' s


being Gat t' (where B was not G immediately before) then it
is a necessary part of the causal interaction (which any
complete account of A's causal action will include reference
to) that A is F, for some F such that A is not Fat some time
between t and t'

The case is still an idealised one, as A and B have to be considered


as an isolated system. Otherwise, there is no reason why some
external agency should not prevent A's becoming not F (for example,
in the case where a spoon is put into a cup of hot water, the
warming-up of the spoon need not be accompanied by the
cooling-down of the water if the cup itself is being heated from
without). We have to assume that the total energy of the isolated
system is conserved. However, I will not spend time in building
this into an already bloated principle, as the cases I shall consider
clearly obey it as it stands. I shall therefore take (2) to be the correct
statement of R, even though there is an element of reciprocity in
causal interaction which (2) does not capture.
Now, although the formulation is neutral as to whether t = t' or
not, contradiction clearly follows if t is identified with t'. In other
words, it follows from R that causes cannot be simultaneous with
their effects. For suppose the cause of B's starting to be G at t is
simultaneous with its effect; then the relevant necessary part of the
interaction, A's being F, will obtain at t. But, on (2) above, A must
cease to be F between the time of the cause and the time of the
effect. On the supposition of simultaneity, this time too will bet.
So, at t, A is both F and not F.
Consider the following example of a moving billiard ball A's
striking another billiard ball B, which until then was stationary. As
Causation and Simultaneity 89

a result, both balls are in motion. At the moment of the encounter,


A's momentum was MA. After the encounter, A's momentum is
MA. and B's is Ms. Clearly, MA. must be less than MA: A cannot
have brought about movement in B without a reduction in its own
momentum (otherwise momentum is not conserved). Now A's
momentum at the moment of the encounter is a necessary part of
the cause of B' s subsequent movement. So if cause and effect are
simultaneous, then A's having momentum MA must be simul-
taneous with B' s having momentum Ms. But if B has momentum
Ms, then A's momentum must beMA.. So A's momentum, at the
time of the effect, must be both MA and not MA. Since this is a
contradiction, causation cannot be simultaneous in this case. What
happens rather is that at the time of the encounter, A's momentum
is MAl and at times thereafter, MA..
The defender of simultaneous causation may object that I have
begged the question here by assuming that it is the different states
of affairs of A and B which are causally related. Correctly con-
strued, he will say, the cause is the change in momentum of A, and
the effect is the change in momentum in B. Since the principle of
conservation of momentum requires these changes to occur simul-
taneously, we actually have a clear case here of simultaneous
causation. I think this way of representing the situation, however,
is demonstrably incorrect. Causes, it is said (and rightly so) make a
difference: their presence or absence is relevant to the outcome.
Now, I ask: is the change in momentum of A relevant to the
outcome? Certainly the first part is, i.e. A's initial momentum. If
A's initial momentum had not been MA, then B' s final momentum
would not have been Ms. But the second part of the change, A's
final momentum, is not relevant to the outcome as far as B is
concerned. What determines B's final momentum is A's momen-
tum at the moment of impact, not at times thereafter. We could have
intervened as soon as A struck B and prevented A from suffering a
change in momentum - this need have no effect on B whatsoever.
The relevant cause, then, is not A's change in momentum.
Not all cases of causation, however, involve change. Let us now
consider in detail Kant's often-cited case from the Second Analogy
of a ball's resting upon a cushion (Kant (1787), A203). The position
of the ball causes the depression in the cushion, although no
change is taking place. There seems to be no conflict with R here:
the principle appears to be inapplicable in this case. Indeed, it
seems quite plausible to view this as a case of simultaneous
90 Change, Cause and Contradiction

causation. This is Kant's own view of the matter: 'If I view as a


cause a ball which impresses a hollow as it lies on a stuffed
cushion, the cause is simultaneous with the effect.' (Ibid.) This and
like instances pose a prima facie problem for Kant, who is concerned
in the Second Analogy to provide a causal criterion for temporal
order, based on the assumption that causes always precede their
effects. Now the proponent of successive causation may concede
that the state of the ball is simultaneous with the state of the
cushion while insisting that the ball's position at t causes the
cushion's compression, not at t, but at times later than t. Conse-
quently, he can say that if the ball had been removed at t, then the
cushion would have been decompressing at times later than t. This
seems to be Kant's own solution: 'If the cause should have ceased
to exist a moment before, the effect would never have come to be.'
(Ibid.) It is only on a certain conception of cause that causation can
be regarded as simultaneous here. If for example we regard the
relevant causes as a continuant (in this case the ball) rather than a
state of affairs obtaining at a particular time, then many instances
of causation will tum out to be simultaneous. I take it that this is
what Kant intends when he says The great majority of efficient
natural causes are simultaneous with their effects.' 5
However, the Kantian response is not decisive. The proponent
of simultaneous causation will, and indeed must, deny Kant's
counterfactual. He will insist that if the ball's position at t causes
the cushion's compression at t, then if the ball had not been there
at t, the cushion would not have been compressed (at least to that
extent) at t.
How can we decide the issue here? I suggest that whatever one
says about the static case, the same must hold for the dynamic case
in which the ball is first placed upon the cushion. So if the resting
ball's position brings about its effects on the cushion simul-
taneously, then at the time at which the ball was first placed on the
cushion is the time at which the cushion is already compressing.
The argument for this is in the form of a reductio, as follows:
suppose that (i) the ball brings about its effects on the cushion
successively while sinking into it, but (ii) simultaneously when it
comes to rest, at t. By (i), the cushion's compression at tis caused
by the ball's state, not at t, but at times earlier than t, when the ball
is still sinking into the cushion. But this contradicts (ii), which
asserts""that the cushion's compression at tis caused by the ball's
state at t. One cannot hold (i) and (ii) together without assuming
Causation and Simultaneity 91

that at some time, the cushion's compression is overdetermined-


that it has two independent and separately sufficient causes. The
argument can be repeated for times later than t ad infinitum. I
assume that we do not wish to allow that actual cases of causal
overdetermination of effects by their causes occur (an assumption I
shall almost immediately argue for), in which case we must con-
clude either that the ball brings about its effects on the cushion
simultaneously both while sinking into the cushion and when
resting on it, or that the ball brings about its effects successively in
both cases. If there was no such first moment when the ball came
into contact with the cushion (suppose them always to have been
in contact) then we can infer the following counterfactual 'Had
there been a time tat which the ball first came into contact with the
cushion, then the cushion would have started to compress (if
causation is successive) or already been compressing (if causation
is simultaneous) at t'.
What allows me to assume that causal overdetermination does
not (more: cannot) occur? The answer is that it is incompatible with
an assumption I have already called upon: that causes are, in the
circumstances in which they occur, necessary conditions for their
effects, i.e. causation sustains counterfactuals. Suppose now that E
is overdetermined by two independent conditions, C and C*.
Neither of these can be causes, since neither is, in the circum-
stances, necessary for E (if C had not occurred, then C* would still
have brought about E, and vice versa).
We can now construct an argument against the view that Kant's
example is genuinely one of simultaneous causation: first, consider
the case where the ball is first placed on the cushion. Here, the
cause is the ball's coming into contact with the uncompressed
cushion at time t, and exerting a force of value F upon the cushion;
the effect is the compression of the cushion at time t'. As the
cushion compresses, it exerts a greater force upon the ball, which
then begins to decelerate. By Newton's third law of motion, the
ball now exerts a reciprocally greater force upon the cushion. The
change in the force exerted by the ball is the change in a necessary
part of the causal process required by R. Now if cause and effect
are simultaneous in this dynamic case, then at t the ball both is and
is not exerting a force of value F upon the cushion, which is
absurd. So causation is not simultaneous in the dynamic case. But,
as the reductio above showed, if causation is simultaneous in the
static case that Kant envisaged, where the ball is at rest, then it is
92 Change, Cause and Contradiction

simultaneous in the dynamic case. So, by modus tollens, it is not


simultaneous in the static case.
(In dealing with the dynamic case above, I have not appealed
directly to any conservation principle, but there is in fact an
interdependence between Newton's laws of motion and conser-
vation principles: the principle of conservation of momentum
follows directly from Newton's second and third laws; in addition,
the holding of Newton's third law in the dynamic case is entailed
by the principle of conservation of mass-energy.)

6.4 MODAL PROPERTIES OF R

I submit that the foregoing argument can be applied to any other


case of static causation. From R, then, we can infer that no cause,
even a static one, is ever simultaneous with its effect. In fact, we
can infer something stronger than this, namely that simultaneous
causation is physically impossible, for violation of R would involve
violation of physical laws. Can we infer something stronger still? It
is my hope that we can, for the causal criterion of change proposed
in the previous chapter is intended as a logically, not merely
physically, necessary condition of change. If simultaneous caus-
ation remains a logical possibility, then the causal criterion fails to
rule out purely spatial variation for some worlds.
It is not easy to see how we could decide the issue either way. I
am suspicious of a simple appeal to intuition, for are we so sure
that possible worlds whose laws of physics diverge significantly
from ours are nevertheless worlds where genuine causality exists?
To be sure, influential analyses of the causal relation, such as
Mackie's 'inus' condition analysis and Lewis's counterfactual
analysis, allow for causality in worlds where our laws of physics do
not obtain, but this is not to say that our concept of cause does not
presume a background of physical assumptions, not captured by
these schematic analyses. Indeed, such accounts, as I shall argue in
the next chapter, face difficulties which can be resolved by building
reference to reciprocal effects into the analysis of causation. In
doing this, I turn R into a logically necessary principle. This is
obviously a controversial move on my part, but it is not purely ad
hoc: I depend upon it in dealing with the 'problem of epiphen-
omena' (see Chapter 7, sections 2 and 4).
A problem still remains, for I am inclined to believe that time
Causation and Simultaneity 93

only has its topological structure as a matter of contingent fact. In


particular, I consider it to be at least logically possible that time
might have had a closed structure. The states of a closed time
universe, assuming there to be no causal discontinuities, will form
a causal loop: one state can be both a cause and an effect of some
other state. Moreover, any state S will be causally connected to
itself, in the sense of there being a series C11 c2 • • • • ex, where en is
dependent upon cn_ 1, and where S = (or is partly constituted by)
c1 = ex. There is no reason why causation in such a world would
not obey R. But because S is both simultaneous with itself (it could
not fail to be) and a (distant) cause of itself, this constitutes a case of
simultaneous causation. So even the logical necessity of R appar-
ently fails to guarantee the logical impossibility of simultaneous
causation.
This particular example, note, does not threaten the causal
criterion of change proposed in Chapter 5.3, for change involves
incompatible states of affairs, and this example does not. However,
in order to generate a problem for the causal criterion we have only
to suppose that the closed-time universe exhibits spatial variation
at a given time: a is Fat s1 and G at s2 • Since this is part of the causal
loop, a's being Fat S 1 will be amongst the causes of a's being Gat s2
(Figure 1).
I shall postpone a resolution to this problem for my account of
change until Chapter 8; it would be inappropriate in a discussion of
the modal properties of R. The important point for our present

/
I
I

I
I s1 F
a-....,
II ' \
tJ I I
I t1
I I
' I
\, I
'
'' -... ___ _ /
/
/

Figure 1
94 Clulnge, Cause and Contradiction

interests is that the case shown here is not a genuine case of


simultaneous causation. True enough, the cause and the (distant)
effect are simultaneous, but the series of causal relations in virtue
of which cause and effect are connected extend over a temporal
interval. Genuinely simultaneous causation would involve rela-
tions which did not extend over an interval. That is, not only
would cause and effect occupy the same instant, but all the inter-
mediate causes would do so too. So entertaining the possibility of
closed time, I submit, does not commit us to the possibility of
simultaneous causation.

6.5 TIME WITHOUT CHANGE

To end this chapter, I want to consider an interesting consequence


of R in regard to the much-debated question 'Could there be time
without change?' Shoemaker (1969) and Newton-Smith (1980)
have argued for the possibility of temporal vacua (periods of time
without change) by elaborating circumstances under which we
would be justified in predicting the occurrence of a vacuum. Such a
strategy is clearly appropriate for any theorist who is sceptical
about verification-transcendent truth. I shall not discuss the argu-
ments here. Rather, I want to tie in the debate about vacua to our
discussion of causality. My contention is that we are committed to
the possibility of temporal vacua by our ordinary assumptions
about causality.
The essence of the argument is as follows:

1. Causality entails time.


2. Causality does not entail change.
Therefore: 3. Time does not entail change.

I shall cast the argument in terms of a rather bleak fiction. Suppose:

1. It is possible that there should exist an absolutely un-


changing world, w. (It is at this stage a completely open
question whether w has a temporal structure or not.)

Suppose further:

2. In w, a large boulder on a cliff edge is resting upon a


Causation and Simultaneity 95

small stone, which is so placed as to prevent the boulder


from rolling down the cliff.

3. No cause is simultaneous with its effect.

2, 3 ~ 4. The various causes and their corresponding effects de-


scribed in 2 are temporally separated.

4 ~ 5. w has a temporal structure.

1, 5 ~ 6. It is possible that there should exist an absolutely un-


changing world which has a temporal structure: i.e.
temporal vacua are possible.

I presume that no one, at least at this stage, will object to my


characterizing the boulder and stone as being in causal contact. As
we have seen, cases of static causation - those which do not
involve change - abound, especially in the writings on simul-
taneous causation. If further argument is needed, reflect that the
boulder and stone are the subjects of non-trivial counterfactuals of
the form 'if the stone had not been there, then, under the circum-
stances, the boulder would not have been in its present position'.
Such counterfactuals typically characterize causal contexts, al-
though I would not want to go as far as saying that counterfactual
dependence is a sufficient condition for causal dependence (see
Chapter 7.2). In this particular case, however, it is hard to see what
grounds the counterfactual if not causal relations between the
stone and boulder. Precisely how one characterizes the cause(s)
and effect(s) in this case needs some care. Obviously, if we allow
that w is a world with a temporal structure, then the presence of
the stone is contemporaneous with the presence of the boulder.
But, as I argued earlier, this does not mean that there is simul-
taneous causation here. Rather, it is the presence of the stone at
some time which is the cause of the boulder's staying where it is at a
later time, or later times. The truth of R requires that the network
of causes and effects must exhibit some temporal structure.
What, now, of 3? The contention of this chapter is that it is at
least a physically necessary truth. If that is all it is, then w had
better be a physically possible world, otherwise the argument
collapses. I see no reason why absolutely changeless worlds are
not physically possible, although it is true that we shall have to
96 Change, Cause and Contradiction

make certain assumptions about them. For example, we would


have to assume that, amongst its other pleasures, w has a tempera-
ture of absolute zero. For otherwise, the molecules of the stone and
boulder would be in constant motion, and w would not, after all,
be an absolutely changeless world. Further, the planet on which
the boulder and stone are resting must not be orbiting a star or
some other planet: it must be at rest. But this is not incompatible
with the assumption that (very weak) gravitational forces are being
exerted on the boulder. I would, I confess, prefer not to make
physical assumptions of this kind, but if my criterion of causal
connection in Chapter 7.4 is correct, then R is a logically necessary
principle, and so we need only assume that w is a logically possible
world.
The move from 3 to 4, however, needs further support. It might
be objected: 'Even if we accept that simultaneous causation is
logically impossible, it does not follow that the causes and effects
in w are temporally separated - they may not be temporally related
at all.' Consider again the fact that the causal relation sustains
counterfactuals. These are of the general form

(C) If it had not been the case that p at t, then it would not have
been the case that qat t*, where t* is later than t

Such a counterfactual is true at w- indeed must be true if w sustains


causal relations. But such a counterfactual would make no sense if
w were in fact a completely timeless world. We can put this point
in terms of Lewis's analysis of counterfactuals: (C) is true at w iff at
the closest worlds to w where p does not obtain, q does not obtain
either. 'Closeness' here is defined in terms of physical law and
particular matters of fact. 6 Now it is plain from this that (C) could
not be true at a timeless world, for its closest worlds will themselves
be timeless, and so not worlds where the antecedent of (C) is true.
So if w sustains causal relations, it must also sustain temporal
relations.
But perhaps this appeal to Lewis's analysis tells against my
argument, too. For one of the counterfactuals we may want to
assert in supporting the contention that the stone and boulder are
in causal contact is this: 'if the stone were removed then the
boulder would fall down the cliff'. This, on Lewis's account, is only
true if, at the closest worlds in which the stone is removed, the
boulder falls down the cliff. But such worlds will be worlds in
Causation and Simultaneity 97

which, unlike w, there is change, and the difference between


changing worlds and w is too great for them to be part of the
truth-conditions of any counterfactuals true at w. Consequently,
we just can't truly assert of the stone and boulder in w that if the
stone were removed, the boulder would fall down the cliff. One
could reply: so much the worse for Lewis's analysis, for we do
want to assert such a counterfactual of w. But if we reject Lewis's
analysis, then my grounds for asserting that w is genuinely a
temporal world are undercut. Instead, I concede that the worlds
closest to w must all be changeless ones, but also insist that the
relevant counterfactual is not 'if the stone were removed, the
boulder would fall', but rather 'if the stone were not in that
particular position, the boulder would not be in that particular
position'. The truth-conditions for the second counterfactual need
involve no worlds except changeless ones.
My argument for w' s having a temporal structure may cause us
to reconsider the apparently uncontroversial premise 2 (or at least
the assertion that the boulder and stone are in causal contact). It
may not be enough just to point to cases of causation which don't
involve change, for we have shown that such cases require a
temporal context, and this temporal context may be precisely what
is lacking in timeless worlds. So it is still open for the 'no time
without change' proponent to deny the possibility of causation in
the absence of change. However, if he denies that the boulder and
stone are in causal contact, then he robs himself of the explanation
of why the boulder is in the position that it is in, or of why the
stone is slightly depressed, and a host of other features of the
situation. So this is not a plausible way out of the argument.
To sum up, we can respond to the argument of this section in
three ways:
(a) the 'Heraclitean' move: changelessness is logically impossible,
and so a fortiori, there cannot be causal relations in completely
changeless worlds.
(b) it is logically possible that there should be changeless worlds,
but no true counterfactual (in particular no counterfactual
grounded in causal relations) applies to them, so there cannot
be causal relations in completely changeless worlds.
(c) time can exist in the absence of change.
I find (a) totally implausible, and (b) entails that no causal
explanation is possible in the total absence of change, and this too I
find implausible, but perhaps less so than (a). By far the least
98 Change, Cause and Contradiction

implausible, in my view, is (c). So the truth of R, together with the


assumption that static causation is possible, provides an argument
for the possibility of temporal vacua. This means that (if we reject
(a) and (b)) my final account of change, relying as it does on R, has
the surprising consequence - which purely formal analyses of
change do not - that it is not incoherent to suppose that there can
be time without change. Further, since a fully worked-out view of
Tenseless theory must include an account of change, it follows that
on my view Tenseless theory takes a stand over what is usually
represented as a quite separate issue in the philosophy of time: the
reductionism/absolutism debate. For if there can be time without
change then a certain reductionist theory of time which holds
times to be logical constructions out of actual events and their
relations must be mistaken.
7
Causal and Temporal
Asymmetry
7.1 THE PRIORITY OF TIME

It is a commonplace that Hume's analysis of the causal relation in


terms of the constant conjunction of the relata makes it a symmetri-
cal relation. To introduce the asymmetry, he had to stipulate that
causes precede their effects. In so doing he begged the question
against simultaneous and backwards causation. This, I think, is an
undesirable consequence, for although we saw in the previous
chapter that there are no actual cases of simultaneous causation,
this is not just a trivial truth. The question that concerns me in this
chapter is: however undesirable Hume's stipulation of temporal
priority is, can any account of causal priority avoid it? If not, then a
causal theory of change is a less attractive possibility than at first it
seemed. For the hope was that, in defining change in terms of
causation, such an account would have as a non-trivial conse-
quence the result that time is the dimension of change. Such a
hope is dashed if Hume's stipulation is inevitable.
Perhaps, we might add, not completely dashed. Time order may
indeed enter into the necessary conditions for causal order. But
this is not to say that it enters into the necessary conditions for
causal connection. The relation of being causally connected, unlike
that of being causally prior, is symmetrical. If we can define change
in terms of just causal connection, rather than causal priority, then
we can surely remain untroubled by the inevitability of Hume's
stipulation. I think this is a possible move, but there are some
problems with it. First, an account of change should include an
account of what determines the direction of change: change is
always from some state to another. A causal account of change will
naturally examine the direction of change in terms of the direction
of causation. Second, the suggested way out takes it for granted
that an understanding of causal connection is independent of an
99
100 Change, Cause and Contradiction

understanding of causal priority: that an account of priority can be


achieved simply by 'tacking on' a clause to an already existing
account of connection. This may be true if one is prepared to accept
a sufficiently weak sense of connection. For instance, a sense
which allows two states to be causally connected if they have a
common cause, or common effect. But this is too weak a sense of
connection for the account of change I want to offer. In the sense
that account requires, a and b are causally connected if and only if
either a is causally dependent upon b, or b is causally dependent
upon a. Here the notion of causal connection clearly depends upon
that of causal priority. Since I do in fact offer a criterion of causal
connection which does not involve reference to causal priority, it is
only the first of these problems which troubles me. In addition, as
will become clear in the final chapter, the Tenseless theorist can
exploit the asymmetry of causation to good effect provided he has
not defined it in terms of temporal asymmetry. It is therefore incum-
bent upon me to show that Hume's stipulation is not inevitable.
Before considering in detail two sophisticated and justifiably
popular accounts of causation, I want to examine a quite general
argument to the effect that Hume's stipulation is inevitable. Such
an argument is suggested by a passage in Newton-Smith (1980),
which starts with an epistemological problem: how do we come to
know that two states of affairs are causally related? The kind of
causation relevant here is sometimes called property causation, re-
ported by general causal statements ('Sumo wrestling causes obes-
ity'; 'Doing philosophy causes neuroses'). Since causal explanatory
relations can exist between types of event or state, as well as
between type properties, a more general name- and one I shall
adopt - for the kind of phenomenon reported by general causal
statements is 'type causation'. Contrasted with this is token causation,
which concerns relations between particular events or states of
affairs, reported by singular causal statements ('the letter's arrival
caused his heart-attack'; 'the Sarajevo assassination was a cause of
the First World War'). It was a central contention of Hume's that
we cannot in general detect isolated cases of token causation. 1
Knowledge of a particular instance of a causing b would have to be
grounded in the knowledge that A's (in general) cause B's. In other
words, knowledge of token causation is parasitic upon knowledge
of type causation. This is a controversial assumption (it is rejected-
though without much in the way of supporting argument - by
Anscombe (1971)), but we will let it pass for the time being. To go
Causal and Temporal Asymmetry 101

back to Newton-Smith's question, we should phrase it thus: how


do we gain knowledge of type causation? The familiar Humean
answer is: by observation of constant conjunctions of A's with B's.
Now we recognize these as constant conjunctions because the
constituent events/states of affairs are spatio-temporally contiguous.
That is, we recognize them, in part, by their temporal relations. This
suggests that the notion of temporal order is epistemologically
prior to that of causal order. As Newton-Smith puts it:

Without such a notion [of temporal ordering] we have no way of


getting our experiential field organized into constant conjunc-
tions at all. That is, without this, we might equally group all
instances of event type E1 together and group all E2 type events
together rather than forming pairs of events of type E1 and type
E2 • (Newton-Smith (1980)), p. 204.)

We do not have to accept Hume's view of causation in order to


obtain this result. Perhaps, contra Hume, we are (always) aware of
token causation independently of prior experience of conjunctions,
but since such awareness will, in part, be of contiguous events, we
still depend upon a notion of temporal order (a) to recognize the
conjunction as a conjunction, and (b) to decide which is the cause
and which the effect.
This seems convincing, but what follows from it? If, as one kind
of anti-realism (see Chapter 3) has it, the truth-conditions of certain
statements are constituted by the conditions under which we
would typically recognize them to be true, and if causal statements
are of this kind, then time-order must enter into the truth-
conditions of causal statements. This is not quite a verificationist
move, as it is not being asserted that evidence- or verification-
conditions are part of the meaning of causal statements. Providing
the necessary and sufficient conditions for something is not necess-
arily to give a conceptual analysis. But even this more modest
position is enough for us to move from the proposition that time
order is epistemologically prior to causal order to the conclusion
that time order determines causal order, not vice versa.
In general, we seem to be able to distinguish between the
evidence and what it is that the evidence evidences. However,
perhaps we can only do this in those cases where occasionally we
have direct access to the evidenced. In the case of the past, we do
not have such direct access - we have access only to memories and
102 Change, Cause and Contradiction

other causal traces. In the case of other minds, similarly, we have


no direct access - we have access only to (typically linguistic)
behaviour and (occasionally) brain-states. And the object of
Newton-Smith's argument was to show that we do not have direct
access to the causal relation. Causality can, as it were, only be
perceived through a temporal filter.
Even in such cases, however, the realist will insist on a distinc-
tion between truth-conditions and verification-conditions. So how
much we can infer from the epistemological priority of time order
depends in large part upon the outcome of the debate between
realism (or rather realism) and anti-realism. My sympathies, for
what they're worth, are firmly on the side of the realist, but I do
not want to get entangled with this dispute here. Instead, then, I
shall cast doubt on the attempt to establish the epistemological
priority of time.
Against this attempt can be matched an equally compelling one
to establish the opposite conclusion. Consider the question, how
do we perceive temporal order? Mellor ((1981), (1985)) has argued
that causation is the mechanism by which we perceive temporal
order. The relevant causal relation here is not between the events
perceived, but between the perceptions themselves. Suppose we
see an event - say a car going through on a red traffic-light, and
soon after, another event - say the owner of the car smiling
obsequiously at a policeman. Then the perception of the smiling
driver may be accompanied by a memory of the red traffic-light.
But the perception of the red traffic-light is not accompanied by a
memory of the smiling driver. One mental state, then, can causally
influence the other, and this is an asymmetrical relationship.
Because of this, we judge the perception of the car going through
on red to be earlier than the perception of the encounter of the
driver with the policeman. We do this for other- perhaps most -
kinds of events. And unless we are given any qualifying informa-
tion about the events themselves, for example, that the events are
very very distant, then we will judge there to be an asymmetric
temporal relation between the events themselves. Now we have
only to add the supposition that we must be aware of the causal
order of the perceptions in order to judge their temporal order, to
reach the conclusion that we judge the temporal order of events on
the basis of the causal order of our perceptions. So time on this view
is seen through a causal filter. This argument, then, purports to
establish that the notion of causal order is epistemologically prior
to that of temporal order.
Causal and Temporal Asymmetry 103

The two arguments, having incompatible conclusions, cannot


both be right. Newton-Smith would insist that the perception of
causal order among the perceptions would itself depend upon the
prior perception of temporal order. But both arguments seem
equally compelling, and hence (since incompatible) equally con-
troversial. The moral of this conflict, I suggest, is that we should
not place much emphasis on considerations of epistemological
priority in dealing with the logical priority of time or cause. I
propose, then, to deal with this latter issue directly.

7.2 CAUSAL PRIORITY AND COUNTERFACTUALS

The first account of causation I shall discuss is the counterfactual


analysis, as developed by Lewis ((1973), (1979)). As he points out,
the roots of this idea go back to Hume. 2 The intuition which lies
behind the counterfactual analysis is that token causes are, in the
circumstances in which they obtain, necessary conditions for their
effects. One feature of this analysis is that it is able to account for
cases in which we want to assert a causal connection between this
particular a and this particular b without wanting to assert a causal
explanatory relationship between A's and B' s in general. Perhaps
the arrival of this particular letter caused Uncle Julius's heart-
attack, but the arrival of letters in general does not cause heart-
attacks.
I shall, for the purposes of this section, follow Lewis is making
the causal relata events (although there is no reason that I can see
which prevents the counterfactual account being extended to
states of affairs). Let 'c' and 'e' denote particular events, and 'O<c>
and 'O<e>' represent the propositions that c and e obtain respect-
ively. Then e depends causally upon c iff the following con-
ditionals are true:

'- o~ -' is to be interpreted 'if it had been the case that-, then it
would have been the case that -'. Clearly o~ is not truth-
functional- if it were, then the second conditional, being counter-
factual, would be trivially true. It is the second, counterfactual,
conditional which is the linch-pin of Lewis's analysis, and I shall
concentrate just on that in what follows. The truth-conditions of
counterfactuals are defined by Lewis as follows: p o~ q is true iff q
104 Change, Cause and Contradiction

holds at the closest possible worlds at which which p holds. 3


'Closeness' here is defined in terms both of particular matters of
fact and of physical law. These, as Lewis says, 'trade off' against
each other. A possible world which shares all its physical laws with
the actual world may yet not be sufficiently similar to the actual
world, for the purposes of counterfactual analysis, if it differs
greatly in terms of matters of fact. Conversely, two worlds which
are identical with respect to matters of fact may be very different
with respect to physical law.
The assumption motivating the counterfactual analysis is a de-
terministic one: the state of affairs at one time places rigid con-
straints on the state of affairs at a later time. If the world were
indeterministic, then there would be no guarantee in all cases that,
had the cause not occurred, the effect would not have occurred. All
one would be entitled to say is that, had the cause not occurred,
then the probability of the effect's occurring in the circumstances
would have been very very small.
Since the assumption of determinism has serious consequences
for the counterfactual analysis of causal asymmetry, it will be
useful to provide a clear statement of the kinds of determinism
relevant here. Using established terminology, I shall say that the
universe is futuristically deterministic if and only if, given the total
state of the universe at any given time t, only one course of events
lying in the future oft is physically possible (i.e. compatible with
physical law). Conversely, the universe is historically deterministic if
and only if, given the total state of the universe at t, only one
course of events lying in the past oft is physically possible. The
world is two-way deterministic if and only if it is both futuristically
and historically deterministic. Determinism is in fact often defined
in a time-symmetric way, e.g. 'if two distinct worlds have precisely
the same physical laws, then they cannot perfectly match each
other with respect to particular matters of fact at any time: such
matching is either total or completely absent (total matching would
be identity)'. It is an interesting question- and one we shall briefly
come back to - whether the world could be futuristically determin-
istic without also being historically deterministic.
Let us now consider the question whether or not the counterfac-
tual analysis explains why causation is asymmetric. If c had not
obtained, then e would not have obtained. But if e had not
obtained, then c could still have obtained, since c may not have
been sufficient to bring about e. But let us consider a case in which
Causal and Temporal Asymmetry 105

c cannot fail to bring about e. In such a case the following counter-


factual obtains:

-O<el D-+ -O<cl


Here, then, the counterfactual relation between the events
(strictly, between the corresponding propositions, but for elegance
of exposition we may forget this nicety) will be symmetric. Since
such cases are perfectly intelligible, it would appear that the
counterfactual analysis makes causation at best a non-symmetrical
relation. But since causation is asymmetric, this would mean
simply that the counterfactual analysis fails as an analysis,
although it could be represented as a useful (though fallible)
criterion of causal priority.
The simplest way of dealing with this problem - Lewis calls it the
problem of effects - is to require temporal priority as well as counter-
factual dependence. However, not only do we wish to avoid this if
possible, it doesn't solve another problem, the problem of epiphen-
omena. Suppose c is a necessary condition of f, and both a neces-
sary and a sufficient condition for e, where e precedes f. Suppose
further that e is not cause off (Figure 2).
c<lllli;;;"" _ _ _ _ _ _ _ _ ......,.. e

''
' '------------------~ f

to -t

Figure 2

We therefore have the following counterfactual relationships:

from which, since counterfactual dependence is a transitive rela-


tion, it follows that

-o<eJ D-+ -o<fl


So f is both counterfactually dependent upon, and temporally
successive to, e. Yet ex hypothesi, e is not the cause of f. So causal
dependence apparently involves more than counterfactual depen-
dence plus temporal succession.
It would be an advantage if the means of dealing with the
106 Change, Cause and Contradiction

problem of effects also dealt with the problem of epiphenomena.


Lewis thinks he has a way of doing this. 'The proper solution to
both problems', he writes, 'is flatly to deny the counterfactuals that
cause the trouble.' (Lewis (1973), p. 170). But how can he deny
that, if c is a sufficient condition for e, then e is also a necessary
condition for c? The answer lies in Lewis's characterization of
closeness (see above). The closest worlds to our own do not all
share our physical laws, and among these closest worlds where e
does not occur will be those where c does happen, but fails to bring
about e: c does not determine e in worlds whose physical laws
differ so:Qlewhat from our own. So e turns out, after all, not to be a
necessary condition of c. Contrast this result with the one we
would get if we insisted on keeping the laws constant- i.e. if we
allowed only those worlds which shared our laws as being suf-
ficiently 'close'. On the counterfactual assumption that e does not
occur, then, in a deterministic world, the whole history of that
world would be different from that in which e did occur. The past
determines the present, so if the present were different, the past
would have to be different 'and nothing guarantees that the
change could be kept negligible except in the recent past' (ibid., p.
171). So worlds with the same physical laws as this one, but where
e fails to occur, would- or might, depending upon the nature of e
- differ too much in terms of matters of fact from the actual world
to be c,.onsidered sufficiently close for the purposes of assessing
counterfactuals. Better, therefore not to keep the laws constant.
This deals in one stroke with the problem of epiphenomena,
because it rejects the crucial premise that if e had not occurred, c
would not have occurred.
Lewis's move, it seems to me, cuts both ways. If e is not a
necessary condition for its cause, c, how can we be confident that c
is a necessary condition fore? If we keep the laws constant, then,
on the counterfactual supposition that c did not obtain, the whole
of the past would have to have been different. Consequently, in
considering the closest worlds in which c does not occur, Lewis
says that we should not insist on keeping the laws constant. But
since it is the actual laws which guaranteed that in the circumstances e
could not have occurred without c, what reason do we any longer have
for supposing that, in the absence of c, e would not have obtained?
None. On Lewis's conception of closeness, we cannot say defi-
nitely that e is counterfactually dependent upon c. So, again, the
asymmetry is lost.
Causal and Temporal Asymmetry 107

We will get a better understanding of counterfactuals, or at least


a possible worlds interpretation of counterfactuals, I submit, if we
take the closest worlds to be physically possible. That is, if we take
them to share our laws. Degrees of closeness can then be deter-
mined quite precisely by the degree of fit between matters of fact.
We need then no longer worry about the vagueness introduced by
the 'trade off' between matters of fact and physical law. Of course,
given two-way determinism, such worlds will have histories which
do not perfectly match that of the actual world at any stage, but this
does not mean that they will therefore be different in all respects.
The relevant worlds will be those which match the actual causal
conditions to the extent that, in the absence of the actual cause c,
no other possible cause of e is available. So c comes out as a
necessary condition of e. Unfortunately, e will sometimes come
out as a necessary condition of c. So we are back with a non-
symmetrical rather than an asymmetrical analysis. What this sug-
gests to me is that it is mistaken to try to locate the asymmetry of
the causal relation in the counterfactuals themselves, even though
the counterfactual analysis may be the right one for causal connec-
tion in a deterministic world. I consider where else the asymmetry
might be located below.
Before I turn to that, I want to consider a case which appears to
support Lewis's view of asymmetry in counterfactual dependence.
Consider the following counterfactuals:
(1) If I had challenged Fujiwara to a sumo match yesterday I
would have lost.
Reason: I cannot sumo-wrestle at all, and Fujiwara is a top-ranking
wrestler (even though he's having a bad season because of his
injured ankle). So he would certainly have beaten me.
(2) If I had challenged Fujiwara to a sumo match yesterday, I
would have been a top-ranking wrestler.
Reason: the only circumstance in which I would could possibly
have challenged Fujiwara would have been one in which I myself
was a top-ranking wrestler.
(3) If I had challenged Fujiwara to a sumo match yesterday, I
would have won.
Reason: because of (2) above, and the fact that Fujiwara is having a
bad season with his injured ankle. In such circumstances, since I'm
in perfect health, I would certainly have won.
108 Change, Cause and Contradiction

Now it is counterfactuals of kind (1), rather than (2) or (3), that we


tend to assert. But why? The reasons all seem impeccable, so all (if
any) are true. Lewis (1979) denies this (he considers a different, but
related, example). He would accept (1), but would reject both (2)
and (3). His reason is that 'seldom, if ever, can we find a clearly true
counterfactual about how the past would be different if the present
were somehow different' (Lewis (1979), p. 32). In contrast, we can
very often say how the future would be different if the present were
different. Now (1) is a proposition about how the counterfactual
occurrence of an event would have made a difference to the future
of that event, and so is acceptable. Now (3) does the same, but isn't
acceptable because it depends on (2), which is a proposition about
how the counterfactual occurrence of an event would have made a
difference to the past of that event. This is problematic, on Lewis's
view, because counterfactual dependence runs from effects to
causes, not causes to effects (and, in this case at least, effects lie in
the future of their causes). So here is a case where our intuitions
seem to converge- (1) is acceptable but {2) and (3) aren't- and
Lewis has an explanation of it, based upon his theory of counter-
factual asymmetry. So doesn't this support his view of counterfac-
tual asymmetry?
No, because there is a much simpler explanation of the differ-
ence between (1) and the other counterfactuals. This difference
doesn't lie in the truth-values of (1), (2) and (3): since the reasons in
favour of each seem to me to be pretty conclusive, they are all
equally as likely to be true. Rather, the difference lies in our choice
of background assumptions. The reason we choose to assert (1)
rather than (2) or (3) is not because (2) and (3) are suspicious, but
simply because we are more likely to make an assumption that p
where p is part of the background conditions of (1), but not (2) or (3),
namely that I am not a sumo wrestler. This is a natural assump-
tion, not just because I actually find it hard to imagine myself as a
sumo wrestler, but because in asserting counterfactuals, we tend
to keep the background conditions as similar as we can to that
actual state of affairs, even when we are considering quite unlikely
counterfactual situations. The actual background conditions in this
case will include, not only the fact that I am not a sumo wrestler,
but also the fact that I am in perfect health, and (though this makes
no difference) that Fujiwara is not.
Consider again Lewis's remark: if earlier situations determine
later ones, then we know that if the present had been different, the
past would have been different in some way; but we may have
Causal and Temporal Asymmetry 109

some difficulty in saying precisely how the past would be different.


Let us suppose that the difficulty is not just epistemological, not
just to do with the limited information we have about physical law,
but has some basis in reality. Then we might appeal, not to an
asymmetry in counterfactuals, but to an asymmetry in determin-
ism. That is, we might have futuristic determinism, but not histori-
cal determinism: the present determines the future, but does not
determine the past; there are a number of physically possible pasts
but only one physical possible future, etc. We might dub this view
'Open Past, Closed Future', in pleasant contrast to the Rylean
'Closed Past, Open Future'. 4 Now asymmetric determinism can
provide us with a very simple account of causal asymmetry. Let L
be the conjunction of physical law, S a statement of the back-
ground conditions obtaining at the time of the cause, and S' a
statement of the background conditions obtaining at the time of
the effect. Then c is causally prior to e iff

o ((O<c> & S & L) ~ O<e>) & -o(S ~ O<e>) & -o(L~ O<e>) &
-o(O<c> ~ O<'e>)) and -o((O<e> & S' & L) ~ O<c>)

So we return to a question raised earlier: is it intelligible to assert


futuristic determinism and yet deny historical determinism? Ear-
man ((1986), pp. 130-32) has shown that, with respect to laws that
are time-reversal invariant, futuristic and historical determinism
stand or fall together. (A law A. is time-reversal invariant iff, for any
process p, if pis allowed by A., then the reverse of pis also allowed
by A..) Somewhat simplified the proof goes as follows. Let two
worlds, wi and w 2, share some set of physical laws, L. Then these
worlds are historically L-indeterministic if it is compatible with L
that their respective states of affairs should match each other
perfectly at time t, but diverge at times prior to t. If the laws of L
are time-reversal invariant, then L does not distinguish between
the actual order of states of affairs in each world and the reverse
order. Consequently, if, given matching at t, L permits divergence
of wi and W 2 prior tot, it also permits divergence after t (Figure 3).
However, not all laws are time-reversal invariant- anyone who
thought they were, incidentally, could not consistently hold that
the world was deterministic, for physical laws would not then
distinguish between one order of events ABCDE and the reverse
order EDCBA. So for all that has so far been said, it is consistent to
hold futuristic determinism while rejecting historical determinism.
However, this is not to say that this is the correct analysis- or
110 Change, Cause and Contradiction
w, . . . . . . .

.- >----------

---------<
t --
.....

Figure 3

indeed even the correct criterion - of causal asymmetry. One


consequence of the analysis is that causal loops are ruled out a
priori, and this is a consequence to be regretted, since it also rules
out a closed topology for time. More seriously, it might be thought
that we should allow for causal indeterminism in our analysis of
causal asymmetry. There are two reasons for this: one is that the
world may indeed be indeterministic; the other is that, even if the
world is deterministic, this is surely only a logically contingent
fact, whereas it is not merely contingent that the causal relation is
asymmetric. So let us move on now to an account of causation
which accommodates the possibility of indeterminism.

7.3 CAUSAL PRIORITY AND PROBABILITY

The probabilistic theory of causation is essentially a theory of


indeterministic causation. This does not mean that it cannot ac-
commodate deterministic causation, but it can regard simply as
unusual cases causes which are necessary or sufficient, or necess-
ary and sufficient, conditions for their effects. The intuitive belief
behind the probabilistic approach is that causes do not typically
determine, but only make more probable, the occurrence of their
effects. There are a number of variants on this basic assertion, and I
refer the reader to Suppes (1970), Cartwright (1979), Skyrms (1980)
and Salmon (1981). I shall be content to consider just one version of
the theory (no question is thus begged), on which c causes e iff
Causal and Temporal Asymmetry 111
P(e/c) >P(e/-c)

That is, the probability of e given c's occurrence is greater than the
probability of e given c's non-occurrence. On this account, there is
no requirement that causes make their effects highly probable
(though, again, there are different views on this). The analysis is
generally represented as an analysis of type causation ('looking at
Cubist paintings gives one vertigo'), but there is no reason why it
cannot also be offered as an analysis of token causation ('the
conversation with Kripke caused Eric to abandon the Fregean
account of reference'). The point, however, needs handling with
some care. Obviously, one cannot talk of increasing the probability
of a token event- consider the sense of saying 'the conversation at
breakfast made it more probable that this very instance of the
event-type "nasty happenings in the woodshed" would occur'.
Rather, the conversation at breakfast made it more probable that an
instance of the event-type 'nasty happenings, etc.' would occur. So
in saying that the probabilistic analysis can be offered as an analy-
sis of token causation, I mean simply that the probabilistic analysis
can provide an account of the truth-conditions of singular, as well
as general, causal statements.
A final prefatory remark: as with counterfactuals of the sumo
wrestling variety, one needs to specify the background conditions
carefully if probabilistic assertions are to come out true. I shall take
it for granted in what follows that these background conditions are
written in, though we shall come back to the question of how they
are to be specified.
Does the probabilistic analysis ensure asymmetry in token caus-
ation? No. The 'making more probable' relation is only non-
symmetric. To take an extreme example, suppose a cause c to have
been a sufficient condition for its effect, e, and this remains true
under any circumstances, since c is powerful enough to override
any other factors. Then P(d-e) = 0, for e is also a necessary
condition for c. Suppose in addition that e could in the circum-
stances only have been caused by c, so that P(c/e) = 1. Conse-
quently P(c/e) > P(c/-e), which on the above analysis entails that e
causes c. So the probabilistic analysis faces its own version of the
problem of effects. We need not have assumed such an extreme
case, in fact: even with probabilities between 1 and 0, the making
more probable relation fails to be asymmetrical.
The probabilistic analysis, like the counterfactual analysis, also
112 Change, Cause and Contradiction

faces a version of the problem of epiphenomena. Suppose that c


causes both e and f, and that in the circumstances there was no
alternative cause either of e or of f present, so that the probabilities
are as follows:
P(de) > P(d-e); P(f/c) > P(f/-c)
Then, if the difference between P(de) and P(d-e) and between
P(f/c) and P(f/-c) respectively is sufficiently great, it may also be
the case that P(f/e) > P(f/-e). So e, on the analysis, is a cause off,
which ex hypothesi it is not.
One approach to the problem of epiphenomena in this case is to
regard c as 'screening off' an association between e and f. The
intuitive idea here is that, although e appears to make f more
probable, it in fact makes no difference to the probability off on the
assumption that c obtains. So a fuller analysis would run as follows:
c causes e iff
(a) P(e/c) > P(e/-c)
and (b) there is no X such that
P(e/c & X) = P(e/X)
P(e/c & -X) = P(e/-X)
The proposal deals with the problem of epiphenomena in part by
excluding those probabilistic connections between events which
are attributable only to a common cause of those events. However,
we do not always want to exclude common causes. Take the causal
sequence, abc, where a causes b and b causes c. It is natural to say
here that a is also the cause of c (the fact of there being an
intermediate cause does nothing to weaken this: if causation, like
time, is dense then there will always be intermediate causes be-
tween any two causal relata). Sob and c have a common cause.
Assume rather that a is a necessary and sufficient condition for b
and b is a necessary and sufficient condition for c. Then on the
proposal above, this entails that a does not cause c, for P(da & b) =
P(c/b) = 1, and P(da & -b) = P(d-b) = 0. (The assumption of
deterministic causation here is essential.) It seems, then, that the
problem of epiphenomena cannot be dealt with so easily. I shall
leave the matter for the time being until I discuss my own proposal
for dealing with it in the next section.
Turning back to the problem of effects, I want now to canvass a
proposal of Papineau's (1985). As he puts it:
Causal and Temporal Asymmetry 113

the asymmetry of causation derives from the fact that the back-
ground conditions together with which causes determine their
effects are independent of each other, whereas the same does
not hold of the background conditions together with which
effects 'determine' their causes. (Papineau (1985), p. 280.)

We may ignore the implication of two-way determinism in this


remark, as the proposal can be applied equally to indeterministic
causation. The suggestion is that there is no significant probabilis-
tic relationship between the conditions in the context of which the
cause brings about a certain effect - the initial background condi-
tions, but that there is a probabilistic relationship between the final
background conditions. This is because the final background con-
ditions are brought about, at least in part, by the cause. (Because of
this, it is perhaps not entirely appropriate to make a distinction
between 'the effect' and 'the final background conditions' - both
are effects of the same cause. So I'll just refer to 'the final condi-
tions'.) The 'screening-off' condition, if a suitable one can be
formulated, would prevent the probabilistic relation among the
final conditions being construed as a causal relation.
I think this is an important insight of Papineau's, and his gener-
alization certainly seems to cover standard cases of causation. Take
the case of a haystack about to be set fire to. The initial background
conditions are: the dryness of the haystack, the presence in abun-
dance of oxygen, and the combustible nature of the hay. The cause
is the dropping of a lighted match into the haystack. There is no
probabilistic relation between each of these conditions. The final
conditions are: the evolution of heat, the presence of carbon
dioxide, the absence of oxygen, the presence of carbon. Oearly,
there is a probabilistic relationship between the final conditions -
they are all effects of the process of burning. So the 'initial' and
'final' conditions need not in this case be distinguished by the
temporal relations they stand in to cause and effect, but simply by
their internal probabilistic relationships.
However, we seem to be able to construct plausible cases in
which the initial conditions are not independent of each other, and
these are counterexamples to Papineau's proposal. Such cases
arise where the initial conditions themselves have a common
cause. Consider the following dismal story. A man learns of the
death of his sister. The trauma of this produces long-term damage
to his heart and digestive system. Thereafter, he eats only irregu-
114 Change, Cause and Contradiction

larly and sleeps poorly. His depression causes him to smoke and
drink heavily. He takes to going for long walks in the rain, which
bring on a succession of minor respiratory illnesses, producing
further strain on the heart. Then, one night, burglars break into his
house wearing alarming Halloween masks and threaten him with a
shotgun. The shock brings about heart failure and death.
In this depressing case, the initial background conditions -
considered as those obtaining at the time of, but not including, the
burglary - are probabilistically related to each other, and this is
because they have a common cause, namely his learning of his
sister's death. Now this story is not so bizarre. Similar cases occur
frequently (consider the combination of conditions leading to sui-
cide in many instances). So independence of initial background
conditions is a not a universal feature of causal interactions.
Papineau does in fact consider this kind of objection, and deals
with it with the materials of his original proposal. Even where the
background conditions have a common cause, he suggests, that
common cause will itself have been caused in a context of back-
ground conditions between which there will be no probabilistic
relationship. But as Ehring (1987) notes in his discussion of Papi-
neau, why shouldn't it be possible for there to be a series in which
there is a common cause of the background conditions however far
back one goes? A further difficulty is this: even if the initial
background conditions do not have a common cause, they do have
(or at least contribute towards) a common set of effects. And this,
in some cases, will be enough to create a probabilistic relation
between the initial background conditions: a background condition
a increases the probability of a number of effects e 1 • • • en which,
in turn, increase the probability that some antecedent background
condition b obtained. If the increases in probabilities are suf-
ficiently great, then a also increases the probability of b. So, far
from its being the case that probabilistic relations are absent from
the set of background conditions, such relations will often be a
mark of them.
A quite different qualm we may have concerning Papineau's
whole approach is this: why should the causal priority of a par-
ticular event be determined by a relationship between the back-
ground conditions? Is it plausible that the asymmetry of the causal
relation be extrinsic to the causal relata? We shall come back to this
question before long.
Is there some other asymmetry in the probabilistic relationships
Causal and Temporal Asymmetry 115

of causes and effects which one can exploit? It may be that there is,
but at this point I want to change the direction of our enquiry
because of some general doubts concerning the probabilistic
analysis of causal connection- i.e. doubts which arise before the
question of asymmetry is tackled. These doubts are well expressed
by Cartwright (1979). Central to the whole probabilistic analysis is
the contention that causes make the probability of their effects
higher than those probabilities would have been in the absence of
those causes. But is this generally true? Consider two apparent
counter-examples:

(1) Smoking causes heart disease. But suppose smoking is also


correlated with under-eating, which reduces the probability of
heart-disease. Suppose further that under-eating is more effec-
tive in helping to prevent heart-disease than smoking is in
helping to bring it about. Then, because of this correlation,
smoking actually reduces the probability of getting heart-disease.

(2) Suppose we have two pieces of radioactive material in front


of us: one is a piece of uranium, the other, of polonium. We
select one piece at random and place it in front of a geiger
counter. The probability that the counter will reach a certain
number n of clicks in a given time is much higher if we select the
polonium than if we select the uranium, as polonium is the more
radioactive. Let us suppose that we select the uranium and the
counter does reach n clicks. It is still the uranium causing the
counter to have reached n clicks, even though its being put in
front of the counter in the circumstances reduces the probability
of the actual effect. (Cf. Salmon (1977), p. 64.)

Now the kinds of causal statement in these two examples are of a


different logical type. 'Smoking causes heart disease' is a general
causal statement, whereas 'the uranium caused the counter to
reach n clicks' is a singular causal statement. Since we are in this
chapter dealing only with token causation, it would be acceptable
for a proponent of the probabilistic analysis to respond to (1) as
follows: 'It is not invariably true that causes of a certain type raise
the probability of effects of a certain type, but I do want to insist
that a token cause, in the circumstances of its occurrence, raises the
probability that an effect of a certain type will occur.' However,
this response is threatened by counterexamples like (2).
116 Change, Cause and Contradiction

Neither example, in fact, requires us to give up the view that


causes raise the probability of their effects, but they do require us
to qualify it. In order to reveal the true probabilistic relationships
between events we need to exclude all other causally relevant
factors. In the first example, smoking reduces the probability of
heart-disease in the general population but increases it in the
population of exercisers. Similarly, in the second example, ura-
nium decreases the probability of the counter's reaching n clicks in
the actual circumstances, characterized in part by the conditional
'if uranium isn't selected, polonium will be', but increases it in
circumstances where this conditional is false (and no other radioac-
tive substance is present). To accommodate this, Cartwight pro-
poses the following relationship between causality and probability:
c causes e if and only if c increases the probability of e in every
situation which is causally homogeneous with respect to e. (Cart-
wright (1979), p. 423.) 'Causally homogeneous' situations are those
where every causally relevant factor other than c is held fixed.
This, of course, requires that the causally relevant factors be
identified. As she points out, her biconditional could not possibly
count as an analysis of causality, since that very notion appears in
the right-hand half of the biconditional. Evidently, we shall have to
look elsewhere for an account of causal connection, and perhaps
that account will also provide the basis of causal asymmetry.
That is one moral of the story. Another is that, although the
probabilistic account of causation cannot provide an analysis of
causal connection, it might still hold the key to an account of causal
asymmetry once we have found the right analysis of causal con-
nection. Generalizing this: perhaps our mistake has been to at-
tempt to find asymmetry within an account of causal connection,
instead of looking for a condition for asymmetry which can simply
be imposed upon a variety of accounts of causal connection. And
ideally, this condition should be entirely neutral as to the question
whether causation is deterministic or indeterministic. Such a con-
dition is the subject of the next section.

7.4 EHRING'S CONDillON

In this final section of the chapter, I shall present a condition


(Ehring's) which determines causal direction in those cases where
a causal connection has already been established. This deals with
Causal and Temporal Asymmetry 117

the problem of effects. I shall then make a proposal of my own


concerning a necessary and sufficient criterion of causal connection
which deals with the problem of epiphenomena. The result is an
account of causality which, as we shall see in the final chapter, is
applicable just as much to closed time as to linear time.
Let us begin with the supposition that we have established a
full-blown causal connection, not just some weaker relation (e.g.
counterfactual or probabilistic), between c and e. We know that
one is causally dependent upon the other, but we don't know
which. How do we establish the direction of causation here?
Ehring (1982) has suggested the following criterion: if there is a
causal connection between c and e, then there will be some further
factor f, upon which the connection between c and e is counterfac-
tually dependent. c is then causally prior to e iff f is causally
connected to e but not to c.
Some examples will show this to be a plausible condition: (a) my
placing a piece of sodium in a beaker of water causes the sodium to
burst into flame on the surface of the water. The causal connection
here is counterfactually dependent upon the molecular constitu-
tion of the sodium which is the physical basis of its disposition to
ignite in contact with water. The molecular constitution is causally
connected to the sodium's bursting into flame, but is not causally
connected to the sodium's being placed in water; (b) my becoming
bored with a wireless programme causes me to switch the wireless
off. The causal connection is counterfactually dependent upon
there being a device manipulation of which cuts off the current to
the set (viz., the on/off switch). The existence of this device is
causally connected to the switching off of the set, but is not
causally connected to my becoming bored.
Some strengthening of Ehring' s condition will be necessary, for f
might on occasions be connected to the cause, as well as the effect.
The molecular constitution of this piece of sodium may have been a
causally relevant factor in my forming a belief that it would ignite
on contact with water and from thence to an intention to drop it in
water ('Well, the results of the spectroscopy test certainly indicate
that this is sodium, so let's just drop it in this beaker of water to
make sure shall we?'). Again, having just read a popular novel of
J-P. Sartre, perhaps I become fixated on the wireless switch and
have an attack of Nausea, which causally contributes to my bore-
dom and consequent decision to switch off the programme. All
this is possible, so we need to strengthen the condition that f is (in
118 Change, Cause and Contradiction

fact) causally connected to e, to the condition that there is some


law-like connection between f and e. Any causal connection be-
tween f and c would be just accidental. 5
The account will have to be refined further, as Ehring realizes,
because in cases where c is in the circumstances a sufficient
condition for some effect, e*, other thane, the condition will make
e causally prior to c. This is because e* will also be a necessary
condition for c: if e* had not occurred then, since c is in the
circumstances sufficient fore*, c could not have occurred either. If
c itself is counterfactually dependent upon e* then, a fortiori, the
causal connection between c and e is counterfactually dependent
upon e*. But e* is causally connected to c, not toe, so on Ehring's
unrefined condition, e turns out to be causally prior to c. How can
he avoid this result? His proposal is as follows: let S be the set of
events/states of affairs/dispositions, etc., upon which the causal
connection between c and e is counterfactually dependent. We
now place the following restriction on S: there is no member of S
which is counterfactually dependent upon anything except
another member of S. (Evidently, Sis rather large.) How does this
rather puzzling condition help? Well, e* in the case above will be
counterfactually dependent upon some other event/state, etc., call
it g. A diagram will help to make the relationships clear (Figure 4).

g -- -- - -- - - - e*
...... ......

...... ......
c~-------------------- e

Figure 4

g1 let us say, is one of the causes, along with c, of e*. e* is a member


of S - i.e. the set of things upon which the causal connection
between c and e is counterfactually dependent. Now, if the exist-
ence of e* is really going to make e causally prior to c, then g had
better be a member of S too. However, if g had not occurred, then e*
would not have occurred. But it is not the case that c would not
have occurred either, for c is only sufficient for e* in the circum-
stances - circumstances which include g. In the absence of g, c is
not a sufficient condition for e*. So c could still have caused e.
Consequently, the causal connection between c and e is not
counterfactually dependent upon g, and so g is not, after all, a
Causal and Temporal Asymmetry 119

member of S. But, as we said earlier, e.. only establishes the causal


priority of e over c if g is a member of S. Therefore, e.. does not here
establish that e is causally prior to c. We can easily verify that
Ehring' s expanded condition is satisfied by genuine cases of causal
priority. Take the case in the diagram (Figure 5).

9-----t.
.......
.......
' ""'-
C--------- ------------ e

Figure 5

Here the causal connection between c and e is counterfactually


dependent upon f, and f is causally connected to e but not to c. In
addition, f is counterfactually dependent upon g. g itself is a
member of S (the set of conditions upon which the causal connec-
tion between c and e is counterfactually dependent), for if g had
not occurred, f, and hence e, would not have occurred either. So
Ehring' s condition entails what is indeed the case: that c is causally
prior to e.
Ehring elaborates his account still further and ends up with a
disjunctive analysis of causal asymmetry, but I will not pursue the
issue further, as we now have a solution to the problem of effects.
One remaining doubt that one might have, however, is that the
asymmetry of the causal relation is made extrinsic to the causal
relata- i.e. a property whose obtaining or not obtaining depends
upon the relationships holding between those relata and other
events or states of affairs. Ehring recognizes this, but is quite
untroubled about it: he considers the causal priority relation to be
'circumstantial' in character. In this he finds himself in agreement
with Mackie. Nothing will be lost, however, as far as our immedi-
ate purposes are concerned, if we proceed more cautiously and
treat Ehring's condition as an infallible criterion, rather than an
analysis, of causal asymmetry. One might say that there is nothing
to an analysis of x over and above the provision of necessary and
sufficient conditions for x. But if we want to resist the idea that
causal asymmetry is circumstantial, then we would not regard as
an analysis an account which (so viewed) made it circumstantial,
while being perfectly happy to accept a circumstantial criterion.
Ehring' s condition can now be imposed upon a number of
accounts of causal connection. If, however, we impose it upon the
120 Change, Cause and Contradiction

probabilistic account of causality, we get an account of causal


asymmetry which is very similar to Papineau's, and that, as we
have seen, does not work. I shall now end this chapter by propos-
ing a criterion for causal connection. In the context of the central
argument of this essay, of course, I do not actually need to do this:
all I set out to do in this chapter was to show that Hume' s
stipulation did not have to be resorted to. However, the suggestion
I want to make links in important ways the themes of the previous
chapter with those of the next.
My suggestion, roughly, is this: a necessary condition of two
items being in causal contact is that they should obey the Principle
of Reciprocity. This, recall, was defined as follows:

If a causal interaction with some item A at time t causes B' s being


G at t' (where B was not G immediately before) then it is a
necessary part of the causal interaction (which any complete
account of A's causal action will include reference to) that A is F,
for some F such that A is not F at some time between t and t'

This version of the principle is actually quite unsuitable for our


present purposes, for firstly it mentions causality on the RHS,
secondly it only applies to causal situations which involve change,
and thirdly it is actually satisfied by some causally unrelated states
of affairs. Recall that we remarked in Chapter 6 that there is an
element of reciprocity in causal interactions which the above ver-
sion does not completely capture. Consider again the paradigm
cases of causal interaction: a cold spoon being placed in hot water;
one billiard ball striking another, stationary, ball. Here the change
(cooling down/gain in momentum) in one item (water/second ball)
must be simultaneously accompanied by a reciprocal and opposite
change (heating up/loss in momentum) in the other item (spoon/
first ball). The two changes must be simultaneous in such cases
because these kinds of interaction are governed by conservation
principles. So, with such cases in mind, let us define 'intimate
causal interaction' as follows:

(a) Two items A and B are in intimate causal interaction iff:


(a) for some times, t and t', A is Fat t, B is Gat t', and B is
not Gat t,
and (b) given the actual relations which obtain between A and
B in respect of properties F and G throughout the
Causal and Temporal Asymmetry 121

interval t-t', it is physically necessary that A is not F at


t'.

Oearly, 'F' and 'G' will be properties whose behaviour is governed


by conservation principles. Causal interaction between items
which involve such properties will therefore involve reciprocal
variation in these items' properties, not just as a matter of fact, but
as a matter of physical necessity. It is physically necessary that the
spoon should heat up as the water cools down. But (a) deals only
with cases involving change. What of causal interactions which
don't involve change? Are they not constrained by similar prin-
ciples - viz. about what would happen if change were involved?
They are, and to accommodate such principles, I propose we
amend (a) to:

(p) Two items, A and B, are in intimate causal interaction


iff:
(a) for some times, t and t', A is Fat t and B is G at t',
and either (b) B is not G at t, and, given the relations which
actually obtain between A and B in respect of
properties F and G throughout the interval t-t', it
is physically necessary that A is not F at t',
or(c) B is Gat t, but if it had been the case that B was not
G at t, then, given the relations which actually
obtain between A and B in respect of properties F
and G, it would have been physically necessary
that A would not have been F at t'.

The causal connection for which this is a criterion is indeed of a


peculiarly intimate kind. The criterion is not satisfied by states of
affairs which are causally connected merely in the sense of having
a common cause, or common effect, or where one state is a
(sufficiently remote) causal ancestor of the other. My contention is
that all genuine cases of causal connection involve intimate causal
interactions as defined by (p). More precisely, c and e are causally
connected iff there is a chain s1 • • • s.., where s1 is a constituent of c
and SJ. is a constituent of e, and where s,. is intimately causally
connected with s,._ 1 and with Sx+t· States of affairs can be said to be
intimately causally connected if the items essentially involved in them
are in intimate causal interaction. We now have a criterion of
causal connection between two states of affairs which deals with
122 Change, Cause and Contradiction

the problem of epiphenomena and upon which we can impose


Ehring' s condition in order to determine priority. Our criterion,
note, entails that the Principle of Reciprocity holds as a matter of
logical necessity, and hence that simultaneous causation is logi-
cally impossible.
We have now, I submit, met the two main objections to the
causal criterion of change raised at the end of Chapter 5 (the
possibility of simultaneous causation and the threat of triviality),
and so we can turn finally to a fuller account of the Tenseless
theory of time and change.
8
A Causal Account of
Change
8.1 THE ACCOUNT

At the end of Chapter 5, we were left with the following brief


analysis of change: object 0 changes iff a part of 0 is F and another
part of 0 is G, where F and G are incompatible properties, and
there is a causal connection between O's being F and O's being G.
This analysis employed what I called the causal criterion of change.
The argument of Chapter 6 showed that the causal criterion
guarantees that this variation between 0' s parts will involve time
(since simultaneous causation is impossible), and the argument of
Chapter 7 showed that the guarantee here is not a trivial one, in
that causal priority does not have to be defined in terms of tem-
poral priority. Ehring's condition can be used to establish the
direction of causation without appealing to temporal direction. It is
now time to develop this rather schematic theory of change and to
defend it against a number of apparent counterexamples.
The first objection I want to consider is this: the causal criterion
does not, by itself, distinguish between purely temporal variation
and the persistence of purely spatial variation. There are numerous
cases of causal connection between different spatio-temporal parts of
an object exhibiting incompatible properties which yet do not consti-
tute changes. Consider a tall and relatively massive object standing
on its end. Its own weight brings about varying degrees of com-
pression throughout its length. Thus, at any one time, different
spatial parts of the object exhibit incompatible properties. Suppose
that this variation persists through to a later time. Although there
is no change here, there is nevertheless the required causal connec-
tion between the state of one spatial part at one time (i.e. one
spatio-temporal part), and the state of another spatial part at a later
time. To exclude this case, we must insist that the relevant varia-
tion is either just between temporal parts, defined as in Chapter 4.2
123
124 Change, Cause and Contradiction

(i.e. as the intersections between an object and spatially maximal


subsets of the set of space-time points constituting the extension of
that object), or just between spatial parts, i.e. variation between
space-time parts which does not involve temporal variation. The
causal criterion can then be brought in to exclude variation which
is just between spatial parts.
Consider next the Cambridge change in the late Prof. Wittgen-
stein which occurs when Dr X considers him, at t 1, to be the
greatest twentieth-century philosopher, and at t 2, to be the second
greatest twentieth-century philosopher (his position in X' s world-
view having been usurped by Jacques Derrida). Whether Wittgens-
tein is alive at t 1 or not, this is clearly only a Cambridge change in
him. Now one way of drawing the distinction between genuine
change and Cambridge change is to do it in terms of the distinction
between extrinsic and intrinsic properties. This latter distinction is
by no means without difficulties, but as an initial attempt we might
pursue a suggestion (which he later rejects) by Ronald Hoy (see
Hoy (1984)). This is that intrinsic properties are independent, where
independence is defined as follows: 'property F is independent iff
it is possible that (there exists a contingent object which is F and
there is no other contingent object)'. However, on this view of
intrinsicness, the property of being the only object in the universe
comes out as intrinsic, which pretty clearly it isn't. 1 A definition
somewhat similar to Hoy's, but avoiding this problem, goes as
follows: 'F is an intrinsic property of x iff x' s being F does not
logically depend upon either (a) the actual existence or non-existence
of any other object or event, or (b) the actual obtaining of any states
of affairs located outside the (spatio-temporal) boundaries of x. I
stress 'logically', because I want to allow that the exemplification of
certain intrinsic properties in objects will be causally dependent
upon other objects, events or states of affairs. On this definition,
'red' is intrinsic, and 'expensive' extrinsic, as one would expect.
Now genuine change is change in intrinsic properties, for example
changes in temperature. Cambridge changes, on the other hand,
involve changes in extrinsic properties, for example the change
from being famous to being forgotten. Putting the various compo-
nent definitions together, we may offer an object-orientated ac-
count of change as follows:

Object 0 changes iff:


(1) there are two incompatible properties F and G such that one
A Causal Account of Change 125

part of 0- 0 1 - exhibits intrinsic property F and another part


of 0 - 0 2 - exhibits intrinsic property G,
and where 0 1 and 0 2 are either different temporal parts
or different spatial parts of 0,
and
(2) 0/ s being F and 0 2' s being G are causally connected via a
chain c1 • • • cx, where c1 is a constituent of 0/ s being F and
cx is a constituent of 0 2' s being G, and such that, for each n
(where n>1), cn is causally dependent upon Cn- 1 •
The equivalent property-orientated account, suitable if we wanted
to include changes which did not involve changes in an object,
would go:
Change occurs in some spatia-temporal region Q iff:
(1) there are two incompatible and intrinsic properties F and G
which are instantiated in Q, and their instantiations, f and g,
occur either in different spatial regions,
or in different temporal regions,
and
(2) the states of affairs of which f and g are part are causally
connected by a chain ... etc.
(Spatial and temporal regions of Q can be defined along the same
lines as spatial and temporal parts of objects.)
In Chapter 1, I said that a parsimonious account of change
would be one which used a single criterion (or, failing that, two
intimately related criteria) to exclude both cases of purely spatial
variation and cases of Cambridge change. I noted then that the
Tensed theorist's account of change could, by insisting on the
non-relational character of pastness, presentness and futurity,
exclude purely spatial variation, but could not, by the same means,
exclude Cambridge change. Similarly, I have offered a Tenseless
account which has had to employ two quite distinct criteria to deal
with the two kinds of problem case, viz., an appeal to causal
relations (to exclude purely spatial variation) and an appeal to the
intrinsidextrinsic property distinction (to exclude Cambridge
change). Is a more parsimonious account available?
· In answer to this, I would like briefly to pursue another way of
excluding Cambridge change which employs a causal criterion.
This is suggested by Mellor's discussion of change in Real Time. Of
Cambridge change he remarks that
126 Change, Cause and Contradiction

What is wrong with all these cases is that the apparent change
involved has no immediate contiguous effects. Naturally there
are effects . . . But the immediate effects of these events are
nowhere near the things in which they are alleged to be changes,
and this is why we deny them that location. (Mellor (1981),
p. 107.)

If Ouspensky becomes famous today as a humourist rather than as


a philosopher, then that it is a Cambridge change in him, but the
causal effects of this change are in the people who admire him, or
find him funny, not in Ouspensky (for he is dead). Mellor's
proposal, therefore, is that the criterion of genuine change in x be
that it have effects which are spatia-temporally contiguous to x.
But this in fact does not exclude all Cambridge changes. Although
my denigrating Chirac may have no effects contiguous to him, it
could have had such effects: I could be ranting in his face, for
example. My becoming annoyed with him is only a Cambridge
change in him, but its effects might be right next door to him
(perhaps I stamp on his foot). We will therefore have to modify
Mellor's criterion. I propose a modal version of it. The difference
between genuine change and Cambridge change is that the former
must, of physical necessity (i.e. in virtue of physical laws), have
spatia-temporally contiguous effects, whereas the latter need not.
Incorporating it into the rest of the definition (for brevity I give just
the object-orientated definition):

Object 0 changes iff:


(1) there are two incompatible properties, F and G, such that one
part of 0- 0 1 - has property F, and another part- 0 2 - has
property G, etc.
and
(2) it is physically impossible for this variation in 0 to have no
causal effects spatio-temporally contiguous to 0 1 and 0 2,
and
(3) Ox's being F and 0 2's being G are causally connected via a
chain ... etc.

The second clause obviates the need to specify in the first that F
and G are intrinsic properties. The advantage of this account is that
the means of excluding purely spatial variation - viz. (3) - and the
means of excluding Cambridge change- viz. (2) - can be seen to be
closely related: they are both causal criteria.
A Causal Account of Change 127

The next problem I wish to discuss is that the object-orientated


account of change proposed above can apply processes (simply
substitute 'process' for 'object'), but processes are not thought to be
subject to change. Those theorists who deny temporal parts to
objects, and believe that this is what distinguishes spatial from
temporal variation, can appeal to the fact that processes have tem-
poral parts to explain why they don't change. Now this explana-
tion is not open to the Tenseless theorist, as he is committed to the
minimal thesis of temporal parts. But neither can he appeal to the
causal criteria which exclude Cambridge change and spatial varia-
tion, for processes do satisfy the causal criteria.
The problem here is not so much over the adequacy of an
account of change as of the distinction between objects and pro-
cesses. This distinction may turn on the observation that, while
both objects and processes may be said to exhibit properties,
processes have them essentially, whereas objects have them only
contingently. E.g. a poker is only contingently at, say, 80 degrees
Cat a given time. But the process of the poker's cooling from 110
degrees C to 30 degrees C could not have involved a different
object, or temperatures higher than 110 and lower than 30 degrees
C. So we can say that a necessary condition for x' s changing is for x
to exhibit variation in its contingent properties, not its essential
ones. This is not as arbitrary as it seems, for genuine change of
essential properties involves loss of identity, since an essential
property is one which x must have in order for it to exist. Therefore
processes, which have all their properties essentially, cannot by
definition change, if change in x requires x to survive the change.
So the relevant clause of the object-orientated account should be
amended to:

(1) there are two incompatible properties, F and G, such that one
part of 0 - 0 1 - contingently exhibits property F, and another
part - 0 2 - contingently exhibits property G, etc.

Finally, we come to the problem that was postponed from


Chapter 6. One of the requirements we made on an adequate
account of change, a rather unusual one, was that it should not
presuppose any particular topology of time. The point of this is
twofold: in the first place, I want to respect the suggestion that
debates between various topologies of time should, in the end, be
settled by empirical, not a priori, means; in the second place, it is
surely intuitive that whether or not something has changed is quite
128 Change, Cause and Contradiction

to

0 is Fat s0 _

'
I
/ .,... - 0 is G at s1
' \
\
I I
(
t3 I l t,
\ I
\ I
\ I
\. /
" .- /
/

Figure 6

independent of which topological structure time has. However, in


taking the causal criterion to guarantee that the constituents of
change are temporally separated, we are assuming that time is
linear. In closed time, it would be possible for the components of
purely spatial variation to be causally connected, even given the
impossibility of simultaneous causation. Consider again the case
mentioned in Chapter 6.4 (Figure 6).
At to, 0 exhibits variation between its spatial parts. The chain of
causes beginning with 0' s being F at s0 runs around the causal loop
and eventually connects with 0' s being G at s 1 • Our criterion of
change is therefore satisfied by a case of purely spatial variation.
The answer in this case, I believe, is simply to build the require-
ment that change be independent of topology into the causal
account of change. That is, it is not enough that there be a
connection between 0' s being F and 0' s being G. We should
require in addition that the existence of a causal connection is not
dependent upon the topological structure of the time series in
which 0 is located. In the case illustrated above the causal connec~
tion clearly relies on there being a causal loop. In linear time, such
a loop could not exist, so the causal connection here is dependent
upon time's topology. (In saying that closed causal loops could not
exist in linear time, I am assuming the impossibility of - certain
cases of - backwards causation. This needs further argument, and
this is supplied at the end of the next section.)
We now have, I submit, an account of change which satisfies all
A Causal Account of Change 129

the constraints of Chapter 5: it is a Tenseless analysis which


excludes in a non-trivial way purely spatial variation and Cam-
bridge change, it distinguishes objects from processes, it is com-
patible with (indeed, explicitly embraces) the minimal thesis of
temporal parts, it can account for the direction of change in terms
of the direction of causation, and finally it makes change indepen-
dent of temporal topology. We can now consider what place it
might have in a broader theory.

8.2 CHANGE AND THE CAUSAL THEORY OF TIME

The essence of Tenseless theory is very simple: it denies the


existence of tensed fact. There is very strong support for this view
of time: its rival runs into incoherence whichever way it turns.
Stronger support for any theory is hard to imagine. But the demise
of Tensed theory leaves a large gap behind, in that an intuitive
explanation of a number of time's essential features has been lost.
To fill this gap, we need to construct a much richer (and hence
more contentious) Tenseless theory than the one which merely
provides the truth-conditions of tensed statements in tenseless
terms. By placing the notion of cause in a central position, I believe
we can achieve this richer Tenseless theory. I have concentrated in
this essay on the Tenseless account of change. This is of evident
importance, not just because any adequate theory of time must
also manifest a theory of change, but because it was a crucial
premise in McTaggart's proof of the unreality of time that change is
impossible in a Tenseless world. Now a theory which characterizes
change as a causal phenomenon is ideally suited to the Tenseless
viewpoint: it can explain the crucial features of real change without
recourse to tensed fact. Such a theory also links, to no small
advantage, two common means of characterizing the difference
between time and space, namely (1) that time is the dimension of
change, and (2) that time is the dimension of causality- a proposi-
tion familiar from discussions of Special Relativity.
But the Tenseless theorist can press his advantage still further.
He can use causality to explain other aspects of time which seem to
require Tensed explanation. For example: 'why do we only observe
the present (or, more precisely, the immediate past) but never the
future?' Answer: since later states of affairs cannot affect earlier
states of affairs we cannot be affected by the future. But to observe
130 Change, Cause and Contradiction

something is to be causally affected by it. Therefore, we cannot


observe the future. This familiar answer is essentially correct,
though the discussion of backwards causation at the end of this
chapter will show that it needs some qualification.
A further question that is often asked is this: 'if the distinction
between past and future is only a relational one (i.e. is constituted
by B-series, not A-series, fact), as Tenseless theory holds, then
how can the past be determinate and the future not?' If Tenseless
theory is correct, then one kind of indeterminacy is certainly ruled
out: the future cannot be ontologically indeterminate or 'open', that
is, it cannot just consist of a collection of possible worlds. Future
fact is just as much part of reality as past and present fact. But the
future is certainly epistemologically indeterminate: beliefs about the
future cannot be invested with the same certainty as can beliefs
about the past. This links up with the point about observability: we
know about the past because we have observed it, but we cannot
observe the future, and this difference has a causal basis. Causa-
tion may enter at another level, too: the future may be causally
indeterministic: it is physically possible for there to be more than
one causal outcome of the present state of the universe. However I
have reservations about this. If the future is causally indeterminis-
tic, then it is likely that the past is too: more than one history of the
universe could account for its present state. So invoking causal
indeterminism fails to explain the difference between the future
and the past, unless it is only the future which cannot be deter-
mined by the present.
Perhaps the most vexed question in the philosophy of time,
however, is this: what constitutes the direction of time? The Tense-
less theorist would make his account almost irresistibly attractive if
he could explain this, too, in terms of the direction of causation.
Such a promise is offered by the causal theory of time, and although
this theory requires (and has received) detailed and lengthy treat-
ment, I will end this essay with some brief remarks on the pros-
pects for a unified causal theory of time and change.
The weakest causal theory of time holds simply that the direc-
tion of causation between two events fixes the direction of time
between them: if A causes B, then A is earlier than B. A stronger
theory holds that temporal priority is reducible to causal priority.
This is generally what is meant by 'the causal theory of time'. Since
we readily imagine temporally separated events which yet are not
causally connected, causal theories tend to define temporal re-
A Causal Account of Change 131

lations in terms of actual and possible causal relations. Such a theory


is proposed by Reichenbach:

E2 is later than E1 if and only if it is physically possible for there to


be a chain s11 s2 • • • sk, such that for each i, from 1 to k-1, Sj is a
cause of s;+ll and such that E1 coincides with s1 and E2 with~·
(Reichenbach (1924), p. 22.)

Two events so related are described as being causally connectible.


Events which are not causally connectible are described as indeter-
minate with respect to time-order. 2 Since we want to allow the possi-
bility of a branching topology for time, we should perhaps qualify
this: events which are not causally connectible are either indetermi-
nate as to time-order or do not belong to the same time-series.
There have been, of course, a number of elaborations and improve-
ments of Reichenbach's original account, 3 but it contains enough
components for the purposes of my rather modest discussion.
One advantage of linking the causal account of change to the
causal theory of time is that it becomes an analytic truth that time is
the dimension of change. This is, intuitively, the status we assign
to that particular truth. However, since we already have as a
consequence of the account of change in Chapter 8.1 and that of
causality in 7.4 that the truth is a logically necessary one, we may
not be much inclined to strain ourselves further in making it
analytic too. However, if we do, then we are not, as it might first
appear, violating the requirement that our theory of change should
not make the connection between time and change a trivial one.
For the temporal nature of change would follow from the conjunc-
tion of the causal account of change and a substantive theory of
time, not just as a trivial consequence of the causal theory of
change alone.
A further advantage of the combined view is this. Suppose two
incompatible states of affairs, cj> and 'II, are not causally connectible
in Reichenbach's sense. That is, it is not possible for there to be a
causal chain between them. Then on the causal account of change
it follows that cj> and 'II could not constitute a change in any object
or region. Now we might find this result puzzling: why should it
not be at least possible for any two events to constitute a change, if
they are incompatible? The causal theory of time provides an
answer: if two events are not causally connectible, then either they
are indeterminate as to time order, or they do not belong to the
132 Change, Cause and Contradiction

same time series. Plainly, temporally unrelated events cannot


possibly constitute a change, and since change is always from
something to something else, how can the components of the
change be indeterminate as to time-order?
I take it that the causal theory of time, thus characterized, is a
philosophical thesis. Sklar (1974) distinguishes the causal theory
considered as a scientific thesis from the theory considered as a
philosophical thesis. 4 The difference is that the first considers it an
empirical truth that temporal relations are constituted by causal
(including causal connectibility) relations, the second, as Sklar
represents it, regards it as a conceptual truth: that is, temporal
relations are simply defined in terms of causal relations. I have my
qualms about this characterization. One could hold that causal
relations entered into the truth-conditions of ascriptions of tem-
poral relations while leaving it open whether the meaning of tem-
poral ascriptions should be given in terms of causal relations or
not. What the philosophical causal theory of time must hold is that
it is a logically necessary truth that temporal relations are constituted
by causal relations. Its scientific counterpart, presumably, regards
this same truth as merely physically necessary. That is, it allows that
time could have different realizations: in this world, it happens to
be the network of (actual or possible) causal connections. Evi-
dently, anything which would threaten the scientific thesis would
also threaten the philosophical thesis, but not necessarily vice
versa. Sklar goes on to assert that the causal theory is in tension
with quantum physics. It may be rather surprising to find that a
metaphysical position should tum out to be in tension with phys-
ics, but there is a clear precedent: Tensed theory is in tension with
the relativity of simultaneity found in orthodox interpretations of
Relativity theory. 5 I shall limit myself here to rebutting two
interesting and relatively undiscussed philosophical objections to
the causal theory.
In The Cement of the Universe John Mackie raises the following
objection to the causal theory of time:

these relations [of causal and temporal priority] seem to have


different logical structures. The direction of time characterizes
the dimension as a whole, whereas the direction of causation
characterizes each process or sequence on its own. (Mackie
(1974), p. 162.)
A Causal Account of Change 133

Mackie's grounds for this somewhat obscure assertion are consti-


tuted by a case in which it is established for some set of events that
their time order is either ABCDE (i.e. A occurs first and E last) or
EDCBA (i.e. E occurs first and A last). Now, fixing the time order
for just one pair of events will decide between these two disjuncts.
If, for example, Cis before D, then we know the time order for the
whole series: it is ABCDE. However, fixing the causal order for any
pair will not be sufficient on its own to decide the causal order of
the whole series. For example, if one knows just that B causes C,
one has still a number of possible permutations of the causal order
left: BCAED; ADEBC; DBCEA, etc. Indeed, some pairs may be
totally unrelated, so that the series is fragmented into two or more
parts: BC, AED; BC, A, E, D.
This case, however, fails to reveal any logical difference between
temporal and causal relations at all, for we can easily set up a
similar case. Let the causal order of the series of events be either
ABCDE or EDCBA. Then fixing the temporal order of just one of
the pairs will not (on Mackie's assumptions) fix the temporal order
of the whole series. But fixing the causal order of just one of the
pairs will fix the causal order for the whole series. Granted, there
may be a difference between time and causality - and these cases
may illustrate it (although one would have to have accepted the
difference already to find the cases convincing), but this difference
is not a logical one: for otherwise we would not have been able to
construct the mirror-image of Mackie's case.
Consider Mackie's point again, phrased slightly differently: fix-
ing the direction of time for just one part of the universe fixes it for
the rest of the universe, whereas fixing the direction of causation for
one part of the universe does not fix the direction of time for the
rest of the universe. Now why should we accept this? The causal
theorist of time will surely retort that, as causal order fixes tem-
poral order, then, if the first part of Mackie's (rephrased) assertion
above is true, fixing the causal order for just one part of the
universe will fix the direction of time for the rest of it. What is
certainly true is that causal relations are less pervasive than tem-
poral relations: two events can be temporally related without being
causally related, but the more sophisticated causal theory we have
been considering defines temporal relations in terms of causal
connectibility. To defeat this view, Mackie would have to show
that causal connectibility is less pervasive than time.
134 Change, Cause and Contradiction

Let us pursue this issue further. We could quite readily imagine


the universe split into different regions which are not causally
connectible to each other: they are completely causally isolated.
But there is more resistance to the idea that these same regions are
not temporally related to each other in any way. And we can back
up this intuition, for Special Relativity requires spatially related
points to be temporally related also (their spatial separation in any
frame is defined by the space-time interval between them and
their temporal separation in that frame. Iff cannot be ascribed a
value, then neither can d 2). Consequently, disjoint time-series
cannot belong to the same universe, if a universe is thought of as,
among other things, a spatial array. However, contra certain wri-
ters (Quinton (1962) Swinburne (1965)) there is nothing demon-
strably incoherent about a model for non-unified time. To put it
another way, the unity of time is not (as I suspect Mackie believes)
a necessary truth. So although this universe (conceived of as
spatially unified) could not contain disjoint time-series, this world
(i.e. the parts of logical space where every actual fact obtains) could
contain disjoint time-series. So the causal theorist need not be
threatened by the intuition that not every pair of events need
exhibit causal connectibility: such events can be part of the same
world, but not part of the same universe, if the universe is essen-
tially temporally unified.
A second objection (Newton-Smith (1983), p. 120) to the causal
theory of time goes as follows. Consider the (by now) familiar case
of world in which time is closed. Here the relations 'before' and
'after' are symmetrical: each instant is both before and after every
other. It also appears that, in such a world, causal relations must be
symmetrical, at least if the closed time world involves causal loops.
There could, perhaps, be closed time worlds which did not exhibit
a causal loop: such worlds would contain instants which were
causally disjoint in one direction. In such a case, causation could be
asymmetrical. But in a world which did involve a causal loop (i.e.
in which no pair of instants was causally disjoint in either direc-
tion), causation would be symmetrical. Since, however, time could
still have a direction in such a world, the direction of causation
cannot determine the direction of time.
The reply to this objection is that, of course, if we assume that
the causal relation is transitive, then in a universe which consisted
of a giant causal loop of events, it will also be symmetrical.
Newton-Smith clearly assumes that any causal connection that we
A Causal Account of Change 135

.,..,--~- -
direction of causation
.......
'
I
/
/
'\
I \
I \
I I
\

no causation
Figure 7
might take as fixing the direction of time will be transitive. How-
ever, an interesting feature of what we called 'intimate causal
connection', defined in Chapter 7.4, is that it is non-transitive.
Consequently, if we can identify intimate causal connections in a
closed time world we can, with the aid of Ehring' s condition,
establish the direction of causal influence. Such a relation, there-
fore, need not be symmetrical in a closed time world.
The causal theory of time, it seems, can resist these attempts to
refute it. However, I am not prepared to endorse it here because I
suspect that the causal theory of change I have offered cannot be
combined with a really interesting causal theory of time. Reichen-
bach's analysis of time order employed a condition of causal
connectibility between actual events. Now if this implies that time
order between instants cannot be defined except in terms of rela-
tions between the events which occupy those times, then his
account conflicts with the conclusion of Chapter 6, that there can
be time without change- i.e. in the total absence of events. To
accommodate this, the causal theorist of time might replace refer-
ence to events by reference to states of affairs. But in a world in
which no change takes place, the states of affairs remain the same
from one time to another. So one cannot individuate the different
states of affairs between which a causal connectibility relation is
supposed to exist. Now one solution is to introduce explicit refer-
ence to instants:
t and t* are distinct instants iff
either there exists or it is possible that there exist two events, e
and e*, where
e occurs at t and e* occurs at t*, and where
e and e* are causally connectible
136 Change, Cause and Contradiction

The problem with this is that reductionist theories of temporal


relations, of which the causal theory is an instance, should in
consistency attempt to reduce temporal points as well. The connec-
tion here can be expressed in the following way: two instants are
non-identical if and only if there is a temporal relation of earlier/
later than between them (for simplicity, we will discount the
possibility that time might be discrete). On the causal theory of
time, such relations are wholly constituted by causal connectibility
relations. But these latter relations can obtain only between events
or states of affairs. So, on the causal theory of time, the necessary
and sufficient conditions for two instants' being non-identical are
given by relations between the contents of those instants. Now the
absolutist, who thinks of time as being logically independent of its
contents, will reject this.
It appears, then, that the causal theorist, in order to avoid such a
conflict, will naturally espouse a reductionist view of instants. One
reductionist analysis is Prior's theory of instants as propositions.
We saw in Chapter 3 that such a theory rests upon an incoherent
view of propositions. Another reductionist analysis is the relationist
one:

't' denotes a time such that Rn(e, t) iff 03x(Event(x) & Rn(e, x))

That is, there is a time which is n units temporally distant from


event e iff it is possible that there should exist an event which is n
units temporally distant from e. Notice that this analysis depends
upon the actual existence of events in a temporal world. Again,
this conflicts with our earlier conclusion that there can be time in
the absence of events. If the relationist is to accommodate this, he
will have to allow that the existence of a time in a changeless world
depends upon purely possible events. Since for any world it is
always possible that there should be events, then on this analysis
(a) every world is a temporal world - so the existence of time is a
necessary, not contingent, truth; (b) every world is infinitely tem-
porally extended, since for any event in that world, it is always
possible that there be another n units after (or before) it, where n is
infinitely large. I find both these consequences completely unac-
ceptable. To avoid them, the relationist must, it seems, construct
his times out of actual and physically (not merely logically) possible
events. This involves a rejection of determinism, for it is not
A Causal Account of Change 137

physically possible for there to be events in an absolutely change-


less and deterministic world.
This is not the place to embark upon a full discussion of relation-
ism, but I have said enough to make it seem doubtful whether
relationism, or any other reductionist analysis, can move in the
confined space it is restricted to by my account of change and
causality. Since I believe the causal theory of time is committed to
some form of reductionism, the prospects for a unified causal
theory of time and change seem bleak.
However, I do not want to end on a negative note. It is import-
ant for the Tenseless explanations, with which I began this section,
of some of the fundamental features of time that the following
relatively modest proposition be true: that if A causes B, then A is
earlier than B. I shall end this essay with some arguments for this
proposition.

8.3 BACKWARDS CAUSATION

On my account of causal priority, simultaneous causation is logi-


cally impossible. Consequently, if A causes B, then either A is
earlier than B or B is earlier than A. To rule out the second
conjunct, it would seem that we need to argue against the possi-
bility of backwards causation. This topic has led to a sizable
literature, 6 and I do not want to reassess the arguments here. One
discomfort I feel with any position which regards backwards caus-
ation as impossible simpliciter is that it rules out a priori the causal
loops that obtain in some closed time worlds. Since I regard
debates concerning the topology of time as empirical in nature, I
want to allow for the existence of (some) closed causal loops. In
such cases, any instance of forwards causation is ipso facto a case of
backwards causation. So any proof of the impossibility of back-
wards causation must be careful not to conflict with such cases.
The first argument I want to present relies upon the premise that
causation is deterministic: that a cause is, in the circumstances in
which it occurs, always sufficient for its effect. We can move from
this premise to the result that backwards causation is impossible in
linear time given two further assumptions: (1) every event has an
antecedent cause, and (2) causes are, in the circumstances in which
they occur, necessary for their effects. Consider now a putative
case of backwards causation:
138 Change, Cause and Contradiction

(i) B < ------------ C

Here B is caused by a later event, C, not by any earlier event. That


is, nothing prior to t 1 is causally relevant in bringing about B. This
however conflicts with assumption (1), that every event has an
antecedent cause. So case (i) is ruled out. If backwards causation
obtains then, it must, it seems, be in cases of kind (ii):

(ii) A ------------> B < ------------ C


~ ~ ~

Here, B is brought about both by an earlier event, A, and by a later


event, C. But this case, too, is impossible, for on the assumption of
deterministic causation, A and C are separately sufficient causes of
B: that is, they overdetermine B. Overdetermination of effects by
their causes, however, is ruled out by assumption (2): that causes
are necessary conditions for their effects. In case (ii) above, neither
A nor C is necessary for B - if one had not occurred, then the other
would still have brought about B.
Perhaps, however, we have not considered all the possible
cases. Consider the following: 7

r--------------------------------~
I I
I V
(iii) A B < ---------- C

Here A causes B only via C, so both A and C are necessary


conditions for B. Neither (1) nor (2) appear to be violated. We could
exclude such a case by strengthening (1) to something like (1)':
every event e has an antecedent cause c, and the route by which c
causes e involves no event which is not itself antecedent to e. At
this point, one begins to get the feeling (if one hadn't had it before)
that the assumptions upon which this argument against back-
wards causation is based are too strong for it to be interesting. One
might well think that causation is deterministic, and think more-
over that this is a necessary truth, but why should one subscribe to
(1), let alone (1)'? Perhaps all events have causes, but why must
these be antecedent to their effects? One would only find this
plausible if one had already rejected the notion of backwards
A Causal Account of Change 139

causation. In any case, since I have not tied my account of causality


to determinism, I would rather appeal to an argument which
applied much more generally.
The second argument attempts to establish the conclusion that it
would never be rational to adopt any singular causal hypothesis 'A
caused B' where B in fact preceded A. 8 The argument rests upon
the idea that many causal situations are in principle repeatable and
that this is how we generally test causal hypotheses. Suppose for
example our hypothesis is that 'A caused B in circumstances C'.
Then we test the hypothesis by bringing about (A & C) a certain
number of times and record the result. If B occurred not at all or
relatively rarely, then in the absence of further explanations of why
B did not occur, we would be inclined to reject the hypothesis.
Now this method of testing isn't appropriate for hypotheses where
backwards causation is held to occur ('backwards hypotheses'},
because in such cases B is supposed to have occurred before (A & C)
is brought about. However, one can attempt to falsify the back-
wards hypothesis by trying to bring about (A & C) in situations
where one knows B has not occurred. If one is successful in a
sufficient number of instances, then this is powerful evidence
against the backwards hypothesis, which should then be rejected.
However, it would still seem to be possible to replace the hypoth-
esis with another backwards hypothesis which states the necessary
background conditions in more detail: 'A causes Bin circumstances
(C & D)'. This can then be subject to falsification as before. Now
unless this process of falsifying hypotheses and replacing them
with more elaborate ones stops somewhere - i.e. unless one
reaches a point at which (A & X) cannot be brought about in
situations where B did not occur - it would seem that one could go
on for ever putting forward backwards hypotheses of increasing
detail. Eventually, however, the hypotheses will be so detailed
that the conditions they describe will no longer be repeatable, and
so not capable of being tested. It would clearly be irrational to
adopt a hypothesis just because it could not be falsified in this
sense: that one cannot set up an experiment to try and falsify it.
Perhaps the process does reach a point where one cannot bring
about (or can only rarely bring about) the putative cause in the
absence of the putative effect. There is, in other words, a prob-
abilistic relationship between (A & X) and B. This result is compati-
ble, not only with the backwards hypothesis that A causes B in
circumstances X, but also the forwards hypothesis that B (the earlier
140 Change, Cause and Contradiction

event) is the only cause of (A & X). Now given that the causal relation
is necessarily asymmetric, these two hypotheses are rivals: they
cannot both be true, for if they were, the causal relation between
(A & X) and B would be symmetrical. So which hypothesis is it
rational to adopt? If we have any reason to suppose that B has
antecedent causes which explain the occurrence of B to one's
satisfaction (if, for example, bringing about these causes is a good
way to bring about B), then the hypothesis that (A & X) caused B is
idle. Now if it is always possible to point to antecedent conditions
which explain B's occurrence, then adopting the forwards hypoth-
esis that B is the only cause of (A & X) will always be more rational
than adopting the rival backwards hypothesis that A caused B in
circumstances X. So, on the assumption that these antecedent
conditions are never absent, the result of any test of a backwards
hypothesis will either directly falsify that hypothesis, or equally
support a rival forwards hypothesis whose adoption is on other
grounds more rational than adoption of the backwards hypothesis.
The problem here is that this assumption - that every event has an
antecedent cause - is, as we noted above, a very strong one, and
precisely the kind of assumption that the proponent of backwards
causation is likely to reject.
The third argument is not so much an argument as a demand for
explanation. If we allow the possibility of backwards causation,
then we deny any logical connection between the direction of
causation and the direction of time. We are then faced with the
question: what, in any given instance of causality, determines the
direction in which the causal influence will go? It might seem that a
trivial answer to this question is sufficient: causality is just a
collection of relations which obtain between states of affairs. The
existence of temporal relations in the world is one brute fact, the
existence of causal relations is another. Between them these facts
provide the (logical) explanation of the temporal direction of vari-
ous instances of the causal relation. We cannot, however, stop at
this answer, because it fails to explain one obvious feature of
causation - that there is (at least) a massive bias in favour of
forwards causation. It cannot therefore be just an arbitrary matter
as to which temporal direction causal influences will take. Since, in
accepting the possibility of backwards causation, we have denied
ourselves a logical explanation of the connection between the
arrow of causation and the arrow of time, we must appeal to an
empirical explanation. In other words, the explanation, when we
A Causal Account of Change 141

have it, will be of the form: the physical conditions cl . . . en


under which two items are related as cause and effect make it
overwhelmingly likely that the cause will precede the effect in
time. Here is a possible example of such an explanation: 'the
causing of one state of affairs by another always involves an
increase in entropy. More precisely, considering the cause and
effect as a closed system, then the move from cause to effect in that
system involves an increase in the entropy of that system provided
that the system is not interfered with by external influences. Now
it is overwhelmingly unlikely (although physically possible) that
the entropy of that system will decrease over time for any time-
interval, however small. Therefore, it is overwhelmingly unlikely
that any effect will precede its cause.'
Take the first sentence of this sample 'explanation': 'the causing
of one state of affairs by another always involves an increase in
entropy'. This is a generalization from many instances of causality.
But since these instances are of forwards causality, where is the
guarantee that backwards causality does not provide a counter-
example to the generalization? Perhaps backwards causality al-
ways involves a decrease in entropy, in which case the explanation
fails. In similar fashion, doubts are cast upon the second part of the
explanation: 'it is overwhelmingly unlikely . . . that the entropy of
that system will decrease over time ... '. This generalization is
based on observations of systems all of which involve (at least for
the most part) forwards causation. The thermodynamic arrow (i.e.
the direction in which entropy increases) may well be tied to the
causal arrow: it is bringing something about that creates greater
disorder. Perhaps in worlds where there is no directed causality,
there is no well-defined thermodynamic arrow. Now suppose that
the first part of the explanation is correct, that all causation in-
volves increase in entropy. We would then have reason to doubt
the second part - systems in which there was a large degree of
backwards causation would show decrease in entropy over time. So
again the explanation fails.
The doubts concerning this example generalize to other explana-
tions of the form characterized above. The question was: why is
backwards causation so rare? The form of the answer consisted of
two generalizations: one about the physical nature of causality, the
other about the consequences that nature has for the temporal
order of cause and effect. At least the first, and perhaps the second
too, of these generalizations rests upon observation of forwards
142 Change, Cause and Contradiction

causality. There is no guarantee that instances of backwards caus-


ation will not provide counterexamples to one or other of the
generalizations. So, ironically, the very possibility of backwards
causation undermines the physical explanation of why backwards
causation is so rare. This argument does not show that backwards
causation is impossible, but it does I think show that the possibility
of backwards causation raises an unanswerable physical puzzle.
The proponent of backwards causation can only avoid the puzzle if
he shows that backwards causation is, after all, much more per-
vasive than we took it to be.
Perhaps the strongest argument for taking the direction of
causation to fix the direction of time, however, is precisely the fact
that we can thereby explain time's many puzzling features: why
time is the dimension of change, why we cannot see the future,
why the future is physically indeterminate (if it is), why time
appears to flow, etc., etc. All these advantages are lost once the
possibility of backwards causation is seriously entertained. We
thus have a strong pragmatic reason not to entertain it.
None of the above (ultimately inconclusive) arguments touches
the backwards causation encountered in closed time, for in closed
time (and only in closed time) any instance of forwards causation is
also an instance of backwards causation in a way which is entirely
compatible with the asymmetry of the causal relation. We should
then state our axiom as follows: backwards causation never occurs
in linear time. On this may rest much of the Tenseless theorist's
distinctive view of the world.
Notes
Introduction
1. Various names for the theories have been introduced. Lloyd (1978) and
Butterfield (1984) use the terms 'tensers' and 'detensers' to describe the
proponents of the two theories. Mellor (1981) uses the somewhat
military terms 'tensed camp' and 'tenseless camp'. Gale (1968) speaks
of 'A-theory' and 'B-theory'. Less neutrally, writers sometimes call the
theories 'the dynamic theory' and 'the static theory', thus revealing
their Tensed inclinations.
2. Broad changed his views on time. The representative sources for the
three different positions he successively adopted are: (1921), (1923) and
(1938) respectively. See Mundie (1959) for a full account of this develop-
ment.
3. Swinburne's version of Tensed theory differs somewhat from my
characterization. See his (1990). It is not clear to me that his characteriz-
ation is coherent.
4. Commonplace or not, it has long been of interest to philosophers. See
e.g. Aristotle, Metaphysics Z, 7.

1 Tense and Change


1. Prior argued that the introduction of a present tense operator would
create difficulties for tense logic as such an operator could not attach, as
the past and future tense operators do, to propositions (Prior (1957), p.
10). But see Evans (1985a) for criticism of Prior's argument.
2. I do not mean to imply that this is the only brand of modal realism.
Stalnaker (1976) adopts a modal realist view in which there are irre-
ducibly modal properties of the actual world, viz. 'ways this world
might have been'.
3. The problems surrounding the definition of 'intrinsic' and 'extrinsic'
are discussed in Chapter 8.1.
4. I intend a contrast here with the sense in which an individual'exists at'
(i.e. according to) different biographies of him.
5. This would rule out the possibility of time without change and the
possibility of cyclical history. It also raises the awkward question of
how temporal parts are to be individuated if not by reference to times.

2 Tense and Contradiction


1. This is not quite right. Broad ((1938), p. 317) points out that McTaggart
seems drawn to the view that A-series distinctions would, if real, be
relational - that is, they would mark relations between an event and
something else (see especially McTaggart (1927), sections 326-8). But
this 'something else' cannot be another event, or a moment, for
143
144 Notes
temporal relations between events and moments are unchanging. To
see the A-series as relational in this sense would be to reduce it to the
B-series. Since A-series positions change, the other relatum would have
to be such that the temporal relation in which an event stood to it was
itself changing. But then it is difficult to see what possible relatum could
have this property. McTaggart took this to be a subsidiary argument
(although not a decisive one) against the possibility of an A-series.
2. In using tense operators, I do not state the argument as McTaggart
stated it. McTaggart's own version involves expressions like 'event e is
past'. This has evoked the criticism (first voiced, I think, by Broad- see
(1938), p. 314) that the contradiction depends upon reading the 'is' as
non-temporal, which of course it is not. By avoiding McTaggart's
formulation I bypass this objection.
3. For an account of the difficulties involved with the iteration of spatial
indexicals, see MacBeath (1988). The point in the text could be made
instead using spatial operators: 'It is here the case that', etc.
4. For a summary of this and related logics, see Prior (1967), pp. 63--4;
(1968), p. 122; and McArthur (1976), p. 81.
5. Further comments on Lowe's paper are made in Le Poidevin and
Mellor (1987).

3 Temporal Solipsism
1. Compare Augustine's remark 'when we describe the past correctly, it is
not past facts which are drawn out of our memories but only words
based on our memory-pictures of those facts'. (Augustine (398), p. 267.)
2. I omit the universal quantifiers in this and succeeding formulations.
3. Oearly, the principle relies on 'x' and 'y' being treated here as rigid
designators.
4. This is precisely the move Prior would have objected to, for he allowed
singular reference only to presently existing individuals (Prior (1970)).
However, this still leaves the temporal solipisist with the first part of
the argument, which leads to the conclusion ihat at no time can there
be fewer than four individuals in the universe.
5. This needs some qualification. It has been suggested (e.g. by Stalnaker
(1981)) that the content of a proposition is just a set of possible worlds.
To take account of this, the point I wanted to make in the text should be
put as follows: even if possible worlds are part of the content of the
proposition expressed by a token utterance u, the content of what is
expressed by u will not necessarily vary from one world to another.

4 Temporal Parts
1. That is, the larval stage of the fly itself, which I take to be distinct from
the corresponding stage of the fly's life-history. Some writers (e.g.
Loizou (1986), pp. 12-16) have assumed that to regard objects as having
temporal parts is to treat them as extended processes. As will become
clear later on in this section, I see no good reason for making this
assumption.
Notes 145

2. Butterfield (1985) identifies this tendency in Richard Taylor's (1955)


otherwise careful discussion of temporal and spatial analogues. This
influential paper may be the source of a widespread practice.
3. Perhaps this is not so uncontroversial. Both Sosa (1983) and Swinburne
(1990) accept the existence of 'spatially tensed facts'. This makes me
wonder whether their understanding of Tensed theory entirely co-
incides with mine.
4. It is not entirely clear to me that the notion of 'going out of existence',
redolent as it is of temporal passage, is appropriate in the context of
Tenseless theory. This is not to say (although it has been said- see
Godfrey-Smith (1977)) that the Tenseless theorist cannot accommodate
birth or death in his scheme of things.

5 Tenseless Change
1. There is a pleasant ambiguity in Geach's words. Cambridge change as a
concept crops up in the writings of both Russell and McTaggart. But
these philosophers are also undergoing Cambridge change when they
are, for example, written about.
2. The precise statement of Reichenbach's theory of time order is given
and discussed in Chapter 8.2.
3. See Chapter 8.3 for a defence of this principle, which is a crucial one for
the Tenseless theorist.

6 Causation and Simultaneity


1. Despite the choice of name, I do not intend to invoke here Kant's 'Law
of Reciprocity or Community' from the Third Analogy (Kant (1787),
B257).
2. For a discussion of these and other putative cases of simultaneous
causation, see Brand (1980).
3. The roots of R go back further than this. See the very suggestive
passage in Aristotle, Physics, III, 201"19-23.
4. See David Papineau, 'The Vis Viva Controversy', in R. S. Woolhouse
(ed.), Leibniz: Metaphysics and Philosophy of Science, Oxford (1981), pp.
139-56. Papineau notes: 'In Brevis Demonstratio Leibniz did not ex-
plicitly say that "motive force" should be taken as proportional to
velocity squared: but of course this would follow directly from the
estimate of "force" in terms of height.' (p. 145.)
5. Kant (1787), A203 (italics mine). 'Efficient causes' is Kemp Smith's
translation of 'wirkenden Ursachen'. The use of the Aristotelian term,
which refers to the continuant(s) in the causal conditions, is, I assume,
deliberate.
6. I have, however, some doubts about Lewis's characterization of close-
ness. For these and a fuller account of Lewis's analysis of counterfac-
tuals, see Chapter 7.3.
146 Notes

7 Causal and Temporal Asymmetry


1. The exceptions are admitted in l.iii.8 of the Treatise: 'Tis certain, that not
only in philosophy, but even in common life, we may attain the
knowledge of a particular cause merely by one experiment, provided it
be made with judgement ... '. (Hume (1739), p. 104.) Hume explains
this by saying we can only attain such knowledge against a background
of experience of constant conjunctions of other (though perhaps re-
lated) kinds of events. The habit of the mind to pass from cause to
effect, formed by past experience of conjunctions, transfers itself to the
case of a single conjunction. The explanation is not entirely satisfactory.
2. ' ... we may define a cause to be an object followed by another ....
where, if the first object had not been, the second never had existed.'
(Hume (1777), p. 76.)
3. Note the plural: 'worlds'. In Stalnaker's (1968) very similar analysis, it is
assumed that there is a uniquely close world. Lewis considers this an
unjustified assumption.
4. The contrast is not exact. Ryle was concerned with the ontological
indeterminateness of the future, not its causal indeterminacy. I do not
know what he thought concerning the causal determinacy of the past.
5. In (1987) Ehring re-casts his criterion as follows: 'C is causally prior to E
iff f is normally causally connected toE but not C.' The 'normally' here
suggests that he considers the connection between f and E to be
non-accidental. But it also suggests, curiously, that it need not always
obtain.

8 A Causal Account of Change


1. I owe this counterexample to Jeremy Butterfield.
2. The fact that Reichenbach defines 'a and bare indeterminate as to time
order' as 'neither a nor b is later than the other' suggests that it
corresponds to our intuitive conception of 'a and bare simultaneous',
but it is better construed as 'there is a null interval between a and b'.
3. See van Fraassen (1970) Ch. 6, Grunbaum (1973) Ch. 7, and Reichen-
bach's later account, in The Direction of Time.
4. Sklar does, however, have some doubts about a clear distinction be-
tween the scientific and the philosophical theory. See Sklar (1974), p.
332.
5. See Putnam (1967) and Mellor (1974).
6. The loci classici on this topic are: Flew (1954) and Dummett (1964).
7. I am grateful to Roger White for suggesting this case, and to Peter
Millican for suggesting a means of dealing with it.
8. This argument is closely related to Mellor's own argument against
backwards causation, in (1981) Ch. 10. The differences between his
presentation of the argument and mine are: (i) he clearly regards causal
asymmetry as a consequence of his argument against backwards caus-
ation, whereas I regard it as a crucial assumption of the argument,
requiring independent support; (ii) he takes his argument to rule out
closed time: 'Cyclical universes are as impossible as backward caus-
Notes 147

ation, for the same reason.' (Ibid., p. 187.) I think that the kinds of
closed loops made possible by closed time are untouched by his
argument. Susan Weir has shown that certain crucial premises of his
fail on the assumption of closed time (see Weir (1988)).
Bibliography
ADAMS, R. (1974) 'Theories of Actuality', in Loux (1979), pp. 190-209.
ANSCOMBE, G. E. M. (1971) 'Causality and Determination', in Sosa
(1975), pp. 63--81.
AUGUSTINE, St. (398) Confessions, trans. R. S. Pine-Coffin (Harmonds-
worth: Penguin, 1961).
AYER, A. J. (1950) 'Statements about the Past', in Ayer (1954), pp. 167-90.
AYER, A. J. (1954) Philosophical Essays (London: Macmillan).
BAKER, L. R. (1979) 'On the Mind-dependence of Temporal Becoming',
Philosophical Studies, 39, pp. 341-57.
BENNETT, J. (1971) Locke, Berkeley, Hume: Central Themes (Oxford: Claren-
don Press).
BENNETT, J. (1974) Kant's Dialectic (Cambridge: Cambridge University
Press).
BERKELEY, G. (1713) 'Three Dialogues between Hylas and Philonous', in
A. C. Fraser (ed.), The Works of George Berkeley (Oxford: Clarendon
Press), Vol. I, pp. 261-360.
BRAND, M. (1980) 'Simultaneous Causation', in van Inwagen (1980), pp.
137-53.
BROAD, C. D. (1921) 'Time', in J. Hastings (ed.), Encyclopaedia of Religion
and Ethics (Edinburgh) pp. 334--45.
BROAD, C. D. (1923) Scientific Thought (London).
BROAD, C. D. (1938) An EXIlmination of McTaggart's Philosophy, Vol. II Part
I (Cambridge: Cambridge University Press).
BUTTERFIELD, J. (1984) 'Prior's Conception of Time', Proceedings of the
Aristotelian Society, 84, pp. 193-209. .
BUTTERFIELD, J. (1984a) 'Relationism and Possible Worlds', British Jour-
nal for the Philosophy of Science, 35, pp. 101-13.
BUTTERFIELD, J. (1984b) 'Seeing the Present', Mind, 93, pp. 161-76.
BUTTERFIELD, J. (1985) 'Spatial and Temporal Parts', Philosophical Quar-
terly, 35, pp. 32-44.
BUTTERFIELD, J. and STIRLING, C. (1987) 'Predicate Modifiers in Tense
Logic', Logique et Analyse, 117-18, pp. 31-50.
CARNAP, R. (1958) Introduction to Symbolic Logic, trans. W. H. Meyer and
J. Wilkinson (New York).
CARTWRIGHT, N. (1979) 'Causal Laws and Effective Strategies', Nous,
13, pp. 419--37.
CLARK, R. (1970) 'Concerning the Logic of Predicate Modifiers', Nous, 4,
pp. 311-35.
COCHIARELLA, N. B. (1965) Tense and Modal Logic, Ph.D. Thesis, UCLA.
DAVIDSON, D. (1%7) 'The Logical Form of Action Sentences', in David-
son (1982), pp. 105-22.
DAVIDSON, D. (1982) 'Essays on Actions and Events (Oxford: Clarendon
Press).

148
Bibliography 149

DAVIES, M. (1981) Meaning, Quantification, Necessity; Themes in Philosophi-


Cill Logic (London: Routledge &: Kegan Paul).
DENYER, N. C. (1981) Time, Action and Necessity (London: Duckworth).
DUMMETI, M. A. E. (1960) 'A Defence of McTaggart's Proof of the
Unreality of Time', in Dummett (1978), pp. 351-57.
DUMMETI, M.A. E. (1963) 'Realism', in Dummett (1978), pp. 145-65.
DUMMETI, M. A. E. (1964) 'Bringing about the Past', in Dummett (1978),
pp. 358-74.
DUMMETI, M.A. E. (1969) 'The Reality of the Past', in Dummett (1978),
pp. 35~74.
DUMMETI, M.A. E. (1973) Frege: Philosophy of Language (London: Duck-
worth).
DUMMETI, M. A. E. (1978) Truth and Other Enigmas (London: Duck-
worth).
EARMAN, J. (1986) A Primer on Determinism (Dordrecht: D. Reidel).
EHRING, D. (1982) 'Causal Asymmetry', Journal of Philosophy, 79, pp.
761-74.
EHRING, D. (1987) 'Papineau on Causal Asymmetry', British Journal for the
Philosophy of Science, 38, pp. 81-7.
EMMET, D. (1988) 'How Close can a Cause Get to its Effect?', PhilosophiCill
Quarterly, 38, pp. 455-70.
EVANS, G. (1973) 'The Causal Theory of Names', Proceedings of the Aris-
totelilln Society, Supplementary Volume, 47, pp. 187-208.
EVANS, G. (1985) Collected Papers (Oxford: Clarendon Press).
EVANS, G. (1985a) 'Does Tense Logic Rest Upon a Mistake?', in Evans
(1985), pp. 343-63.
FINDLAY, J. N. (1941) 'Time: A Treatment of Some Puzzles', Australasilln
Journal of Philosophy, 19, pp. 216-35.
FLEW, A. (1954) 'Can an Effect precede its Cause?', Aristotelilln Society,
_ Supplementary Volume, 38, pp. 27-62.
FORBES, G. (1987) 'Is there a Problem about Persistence?', Proceedings of
the Aristotelilln Society, Supplementary Volume, 61, pp. 137-55.
GALE, R. M. (1962) 'Tensed Statements', Philosophical Quarterly, 12, pp. 53-9.
GALE, R. M. (1967) (ed.), The Philosophy of Time (New York).
GALE, R. M. (1968) The Language of Time (London: Routledge &: Kegan
Paul).
GEACH, P. T. (1949) Critical Notice of Julius Weinberg, Nicolaus of Autri-
court, Mind, 58, pp. ~5.
GEACH, P. T. (1965) 'Some Problems about Time', Proceedings of the British
ACildemy, 51, pp. 321-36.
GODFREY-SMITH, W. H. (1977) 'Beginning and Ceasing to Exist', Philo-
sophiCill Studies, 32, pp. 393-402.
GODFREY-SMITH, W. H. (1978) 'Prior and Particulars', Philosophy, 53, pp.
335-42.
GRUNBAUM, A. (1967) 'The Status of Temporal Becoming', in Gale
(1967), pp. 322-54.
GRUNBAUM, A. (1973) PhilosophiCill Problems of Space and Time (Dordrecht:
D. Reidel).
150 Bibliography

HACKER, P.M. S. (1982) 'Events and Objects in Space and Time', Mind,
92, pp. 1-19.
HELM, P. (1975) 'Are Cambridge Changes Non-events?', Antllysis, 35, pp.
140--44.
HINCHLIFF, M. (1988) A Defense of Presentism, Ph.D. Thesis, Princeton.
HOY, R. C. (1978) 'Becoming and Persons', Philosophical Studies, 34, pp.
269--80.
HOY, R. C. (1984) 'Inquiry, Intrinsic Properties and the Identity of Indis-
cemibles', Synthese, 61, pp. 275-98.
HUGHES, G. and CRESSWELL, M. (1972) An Introduction to Modal Logic
(London: Methuen).
HUME, D. (1739) A Treatise of Human Nature, (ed.) L. A. Selby-Bigge
(Oxford: Clarendon Press), 2nd Ed. revised P.H. Nidditch, 1978.
HUME, D. (1777) Enquiry Concerning Human Understanding, (ed.) L. A.
Selby-Bigge (Oxford: Clarendon Press), 3rd. Ed. revised P.H. Nidditch,
1975.
ISHIGURO, H. (1972) Leibniz's Philosophy of Logic and Language (London:
Duckworth).
JOHNSTON, M. (1987) 'Is there a Problem about Persistence?', Proceedings
of the Aristotelian Society, Supplementary Volume, 62, pp. 107-35.
KANT, I. (1787) Critique of Pure Reason, trans. N. Kemp Smith (London:
Macmillan, 1933).
KENNY, A. J. P. (1963) Action, Emotion and Will (London: Routledge &
Kegan Paul).
KRIPI<E, S. (1972) 'Naming and Necessity', in D. Davidson and G. Har-
man (eds), Semantics of Natural Languages (Dordrecht: D. Reidel), pp.
253-355.
LACEY, H. M. (1971) 'Quine on the Logic and Ontology of Time', Austra-
lasian ]ourntll of Philosophy, 49, pp. 47-77.
LEIBNIZ, G. W. (1704) New Essays on the Human Understanding, trans. J.
Bennett and P. Remnant (Cambridge: Cambridge University Press,
1981).
LE POIDEVIN, R. (1988) 'Time and Truth in Fiction', British ]ourntll of
Aesthetics, 28, pp. 248-58.
LE POIDEVIN, R. (1988a) 'The Principle of Reciprocity and a Proof of the
Non-simultaneity of Cause and Effect', Ratio (New Series), 1, pp. 52-62.
LE POIDEVIN, R. (1988b) Time and Solipsism, Ph.D. Thesis, Cambridge.
LE POIDEVIN, R. and MELLOR, D. H. (1987) 'Time, Change, and the
"Indexical Fallacy"', Mind, 96, pp. 534--8.
LEVISON, A. B. (1987) 'Events and Time's Flow', Mind, 96, pp. 341-53.
LEWIS, D. (1968) 'Counterpart Theory and Quantified Modal Logic',
]ourntll of Philosophy, 65, pp. 113-26.
LEWIS, D. (1973) Counterfactuals (Oxford: Blackwell).
LEWIS, D. (1973a) 'Causation', in Sosa (1975), pp. 180-91.
LEWIS, D. (1979) 'Counterfactual Dependence and Time's Arrow', in
Lewis (1986), pp. 32-52.
LEWIS, D. (1983) Philosophical Papers, Vol. I. (Oxford: Oxford University
Press).
LEWIS, D. (1986) Philosophical Papers, Vol. IT. (Oxford: Oxford University
Press).
Bibliography 151

LEWIS, D. (1986a) On the Plurality of Worlds (Oxford: Blackwell).


LLOYD, G. (1978) 'Time and Existence', Philosophy, 53, pp. 215-28.
LOCKE, D. (1967) Perception and Our Knowledge of the External World
(London: Allen & Unwin).
LOCKE, J. (1689) An Essay Concerning Human Understanding, ed. P. Nid-
ditch (Oxford: Clarendon Press, 1975).
LOEMKER, L. E. (1956) Leibniz: Philosophical Papers and Letters (Chicago).
LOIZOU, A. (1986) The Reality of Time (London: Gower).
LOMBARD, L. B. (1986) Events; A Metaphysical Study (London: Routledge
& Kegan Paul).
LOUX, M. (1979) (ed.), The Possible and the Actual (Cornell).
LOWE, E. J. (1987) 'The Indexical Fallacy in McTaggart's Argument for the
Unreality of Time', Mind, 96, pp. 62-70.
LOWE, E. J. (1987a) 'Reply to Le Poidevin and Mellor', Mind, 96, pp.
539-42.
LUCAS, J. R. (1973) A Treatise on Time and Space (London: Methuen).
McARTHUR, R. (1976) Tense Logic (Dordrecht: D. Reidel).
MACBEATH, M. (1988) 'Dummett's Second-Order Indexicals', Mind, 97,
pp. 113-16.
McGINN, C. (1983) The Subjective View (Oxford: Oarendon Press).
MACKIE, J. (1974) The Cement of the Universe (Oxford: Clarendon Press).
McTAGGART, J. MeT. E. (1908) 'The Unreality of Time', Mind, 18, pp.
457-84.
McTAGGART, J. MeT. E. (1927) The Nature of Existence, Vol. II, ed. C. D.
Broad (Cambridge: Cambridge University Press).
MASSEY, G. (1969) 'Tense Logic! Why Bother?', Nous, 3, pp. 17-32.
MAYO, B. (1962) 'The Open Future', Mind, 71, pp. 1-14.
MELLOR, D. H. (1974) 'Special Relativity and Present Truth', Analysis, 34,
pp. 74-8.
MELLOR, D. H. (1980) 'The Self from Time to Time', Analysis, 40, pp.
5~2.
MELLOR, D. H. (1981) Real Time (Cambridge: Cambridge University Press).
MELLOR, D. H. (1985) 'Fixed Past, Unfixed Future', in B. Taylor (ed.),
Contributions to Philosophy: Michael Dummett (Nijhoff).
MOORE, G. E. (1918) 'Some Judgements of Perception', Proceedings of the
Aristotelian Society, 19, pp. 1-29.
MUNDLE, W. C. K. (1959) 'Broad's Views about Time', in P. Schilpp (ed.),
The Philosophy of C. D. Broad (La Salle, Illinois), pp. 353-74.
NEWTON-SMITH, W. H. (1980) The Structure of Time (London: Routledge
& Kegan Paul).
NEWTON-SMITH, W. H. (1983) 'Temporal and Causal Asymmetry', in
Swinburne (1983), pp. 105-21.
NOONAN, H. W. (1985) 'A Note on Temporal Parts', Analysis, 45, pp.
151-2.
PAPINEAU, D. (1985) 'Causal Asymmetry', British Journal for the Philosophy
of Science, 36, pp. 273-89.
PARFIT, D. (1971) 'Personal Identity', Philosophical Review, 80, pp. 3-27.
PARFIT, D. (1984) Reasons and Persons (Oxford: Oarendon Press).
PERCIVAL, P. (1989) 'Indices of Truth and Temporal Propositions', Philo-
sophical Quarterly, 39, pp. 190-99.
152 Bibliography

PRIOR, A. N. (1957) Time and Modality (Oxford: Clarendon Press).


PRIOR, A. N. (1959) Thank Goodness That's Over', Philosophy, 34, pp.
12-17.
PRIOR, A. N. (1962) 'Changes in Events and Changes in Things', in Prior
(1968), pp. 1-14.
PRIOR, A. N. (1965) 'Time, Existence and Identity', Proceedings of the
Aristotelian Society, 66, pp. 183-92.
PRIOR, A. N. (1967) Past, Present and Future (Oxford: Oarendon Press).
PRIOR, A. N. (1968) Papers on Time and Tense (Oxford: Clarendon Press).
PRIOR, A. N. (1970) 'The Notion of The Present', Studium Generale, 23, pp.
245-8.
PRIOR, A. N. (1976) The Doctrine of Propositions and Terms, eds P. T. Geach
and A. J.P. Kenny (London: Duckworth).
PRIOR, A. N. (1976a) Papers in Logic and Ethics, eds P. T. Geach and A. J.P.
Kenny, (London: Duckworth).
PRIOR, A. N. and FINE, K. (1972) Worlds, Times and Selves (London:
Duckworth).
PUfNAM, H. (1967) 'Time and Physical Geometry', in Mathematics, Matter
and Method, (Cambridge: Cambridge University Press) 1975, pp. 198-205.
QUINE, W. v. 0. (1950) 'Identity, Ostension and Hypostasis', in Quine
(1953), pp. 65-79.
QUINE, W. v. 0. (1953) From a Logical Point of View (Harvard University
Press).
QUINE, W. v. 0. (1960) Word and Object (Cambridge, Massachusetts).
QUINE, W. v. 0. (1966) The Ways of Paradox (New York).
QUINE, W. v. 0. (1969) Ontological Relativity and Other Essays (Columbia
University Press).
QUINTON, A. (1962) 'Spaces and Times', Philosophy, 37, pp. 130-47.
RAMSEY, F. P. (1927) 'Facts and Propositions', in D. H. Mellor (ed.),
Foundations (London: Routledge&: Kegan Paul, 1978).
REICHENBACH, H. (1924) Axiomatik der Relativitischen Raum-Zeit Lehre
(Braunschweig: Viewig).
REICHENBACH, H. (1948) Elements of Symbolic Logic (New York).
RESCHER, N. and URQUHART, A. (1971) Temporal Logic (Springer-
Verlag).
RUSSELL, B. (1903) The Principles of Mathematics (London: Allen&: Unwin).
RUSSELL, B. (1914) Our Knowledge of the External World (London: Allen &:
Unwin).
RUSSELL, B. (1915) 'On the Experience of Time', The Monist, 25, pp.
212-33.
RUSSELL, B. (1927) The Analysis of Matter (London).
RYLE, G. (1953) Dilemmas (Cambridge: Cambridge University Press).
SALMON, W. (1977) 'Statistical Explanation', in Salmon, W. (ed.), Statisti-
cal Explanation and Statistical Relevance (Pittsburgh University Press,
1977), pp. 29-87.
SALMON, W. (1981) 'Probabilistic Causality', Pacific Philosophical Quar-
terly, 61, pp. 59-74.
SCHLESINGER, G. N. (1982) 'How Time Flies', Mind, 91, pp. 501-23.
Bibliography 153

SHOEMAKER, S. (1969) 'Time without Change', Jounuzl of Philosophy, 66,


pp. 363--81.
SKLAR, L. (1974) Space, Time and Spacetime (University of California Press).
SKYRMS, B. (1980) Causal Necessity (Yale University Press).
SMART, J. J. C. (1962) 'Tensed Statements', Philosophical Quarterly, 12, pp.
264-5.
SOSA, E. (1975) (ed.) Causation and Conditionals (Oxford: Oxford Univer-
sity Press).
SOSA, E. (1983) 'Consciousness of the Self and of the Present', in J. E.
Tomberlin (ed.), Agent, Language and the Structure of the World; Essays
presented to Hector-Neri Casteneda, (Hackett Publishing Company, India-
napolis).
STALNAKER, R. (1968) 'A Theory of Conditionals', in Sosa (1975), pp.
164-79.
STALNAKER, R. (1976) 'Possible Worlds', Nous, 10, pp. 65-75.
STALNAKER, R. (1981) 'Indexical Belief', Synthese, 49, pp. 91-122.
STRAWSON, P. F. (1959) Individuals (London: Methuen).
SUPPES, P. (1970) A Probabilistic Theory of Causality (Amsterdam).
SWINBURNE, R. G. (1965) 'Time', Analysis, 26, pp. 189-91.
SWINBURNE, R. G. (1968) Space and Time (London: Macmillan).
SWINBURNE, R. G. (1983) Space, Time and Causality (Dordrecht: D.
Reidel).
SWINBURNE, R. G. (1990) 'Tensed Facts', American Philosophical Quarterly
(forthcoming).
TAYLOR, R. (1955) 'Spatial and Temporal Analogies and the Concept of
Identity', Jounuzl of Philosophy, 52, pp. 599-612.
van FRAASSEN, B. (1970) An Introduction to the Philosophy of Space and Time
(New York: Random House).
van INWAGEN, P. (1980) (ed.), Time and Cause: Essays Presented to Richard
Taylor, (Dordrecht: D. Reidel).
WEIR, S. (1988) 'Closed Time and Causal Loops: a Defence against Mel-
lor', Analysis, 48, pp. 203-9.
WILLIAMS, S. G. (1986) 'On the Logical Possibility of Time without
Change', Analysis, 46, pp. 125-8.
WILSON, N. (1955) 'Space, Time and Individuals', Jounuzl of Philosophy,
52, pp. 589-98.
YOURGRAU, P. (1986) 'Time and Actuality: The Dilemma of the Privi-
leged Position', British Jounuzl for the Philosophy of Science, 37, pp. 405-17.
ZEMACH, E. (1979) 'Time and Self', Analysis, 39, pp. 143-7.
Index
A-series, 76, 129, 130, 143-4 Brand, M., 145, 148
belief, 4, 5 Brennan, A., x
change in, 1, 5, 8, 13-17, 18, 25, Broad, C. D., 4, 143-44, 148
26 Buncombe, M., xi
defined, 1 Butterfield, J. N., x, 4, 6, 37, 46,
fact, 3, 15 58, 60, 61, 145, 146, 148
mind-dependence of, 5
paradox concerning, 24-35 Cambridge change, 1fr17, 76,
reducibility of, 2-3, 4, 24 78-9, 124-6, 127, 129, 145
relationality of, 24, 25, 26, 30, cardinality, paradox of, 48-53
76, 125, 130 Carnap, R., 58, 148
A-theory, 143 Cartwright, N., 110, 115-16, i48
absolutism, 6, 98, 136 causal connectibility, 131, 133-34,
Adams, R. M., 39, 148 135-6
adverbs, 73-5 causal connection, 41, 45, 80-1,
Anscombe, G. E. M., 100, 148 99-100, 11fr17, 123-4
anti-realism, 37--40, 44-8, 57 criterion of, 120-2
anti-realism, 45, 101, 102 intimate, 120-1
a priori, 6, 59, 81, 127, 137 causal constraint, 79-82
Aristotle, 55, 143, 145 causal criterion of change, 10, 80,
Augustine, St, 9, 144 81-2, 92, 93, 122, 123-9
Ayer, A. J., 148 causal explanation, 80, 100, 103
causal homogeneity, 116
B-series, 130 causal loops, 128, 134, 137, 14fr7
defined, 1-2 causal priority, 81-2, 99-122,
fact, 3 130-35, 146
reducibility of, 2, 3, 144 causal statements, see general
truth-conditions of tensed causal statements; singular
tokens, 4, 5; see also A-series, causal statements
relationality of causal theory of change, 10-11,
B-theory, 143 123-42
background conditions, 108, causal theory of names, 40
113-14 causal theory of time, 6, 129-37
oackwards causation, 11, 128, 129, as philosophical thesis, 132
137-42, 14fr7 as scientific thesis, 132
backwards hypotheses, 139--40 causation,
Baker, L. R., 5, 148 asymmetry of, 81-2, 99-122,
Balance Principle, 84-6 130-1, 140
behaviourism, 45 backwards, see backwards
Bennett, J., 84, 148 causation
Berkeley, G., 84, 148 efficient, 91, 145
bivalence, principle of, 38 immediate, 87-8
branching time, 47 parts of an object and, 59

154
Index 155

probabilistic theory of, see continuity


probabilistic theory of causal causal, 41, 93
priority of movement, 61
property, 100 psychological, 41
relata of, 65--6 see also denseness
simultaneous, see simultaneous Cooper, N. L., xi
causation conventionalist theory of colour, 70
static, see static causation counterfactual theory of causal
token, 100, 103, 111, 115 priority, 103-10, 111-12
transitivity of, 135 counterfactuals, 10, 66, 80, 90, 91,
type, 100-1, 103, 111 95, 96-7, 117-19, 138, 145
change, 120--1 see also counterfactual theory of
asymmetry of, 77 causal priority; modality
causal theory of, 10--11, 123-42 counterparts, 21, 22
direction of, 77, 99 Craig, E. J., x
in events, 15, 29, 76-7 Crane, T., xi
problem concerning, 8, 14, Cresswell, M., 150
17-20, 21, 22-3 cyclic history, 46, 143
Tenseless theory and, 13, 14-17, cyclic time, see closed time
1&-23, 57
Tenseless theory of, see under Davidson, D., 29, 74, 75, 148
Tenseless theory Davies, M., 149
in things, 1, 15, 29, 76-7, 78; see denseness
also object-orientated account of causality, 88, 112
of change of time, 47, 112
time and, 5, 19 Denyer, N. C., 36, 149
time without, 27, 94-8, 135, 136, Descartes, R., 85
143 detensers, 143
two kinds of, 1, 13-23 determinism
unreality of, 57 causal, see determinism,
see also A-series; time physical
Charles, D. 0. M., xi futuristic, 104, 109-110
Oark, P. J., x historical, 104, 109-110
Oark, R., 74, 148 logical, 4, 38, 43, 112
closed time, 7, 27, 77, 93-4, 110, physical, 4, 38, 47, 104, 106,
117, 12&-9, 146-7 116, 130, 136, 13&-9
closeness, of worlds, 96-7, 104, two-way, 104, 107, 113
106-8, 145 De Voider, 85
Cochiarella, N., 47, 148 diachronic identity, 17
colour, conventionalist analysis of, 'pathological' cases of, 40--4
70 spatial parts and, 58
common cause, 100, 112, 113, 114, temporal parts and, 59, 63, 64
121 transitivity of, 41
connectibility, causal, 131, 133-4, Dummett, M. A. E., 3, 28, 31, 34,
135--6 44-5, 156, 149
conservation principles, 86, 87, dynamic theory, 143
92
contiguity, 101, 126 Earman, J., 109, 149
156 Index

effects, problem of, 105-6, 111, Gale, R. M., 3, 143, 149


112-14, 117 Geach, P. T., 16, 55, 78, 145, 149
Ehring, D., 114, 123, 149 general causal statements, 111,
account of causal asymmetry, 115
116-22, 146 Godfrey-Smith, W. H., 145, 149
Emmet, D., xi, 88, 149 Grunbaum, A., 5, 146, 149
energy, 86, 87, 92, 120, 121
entropy, 141 Hacker, P., 60, 150
epiphenomena, problem of, 92, Hale, R. L. V., x
105-6, 112, 117, 122 Helm, P., 150
epistemological priority, Hinchliff, M., 150
of cause, 102-3 homogeneity, causal, 116
of time, 101-2, 103 Hoy, R. C., 124, 150
ersatzism, 46 Hughes, G., 150
Evans, G., 55-6, 143, 149 Hume, D., 99-101, 103, 120, 146,
events 150
change in, 15, 29, 76-7
ontology of, 16, 29, 30 incompleteness, 3--4, 31
Prior's treatment of, 29-30 indeterminacy, 4 see also
reductionism and, 54 determinism
temporal parts of, 59, 65-6 independence, of properties, 124
time's flow and, 29 indexical fallacy, 28
see also processes; reductionism indexical thesis of actuality, 35
existential generalization, 50, 53 indexicality, 1, 55-7, 61, 144
existential quantifier, 37, 38, 40, 51 see also A-series; indexical
extrinsic properties, 20-1, 53, 72, fallacy; indexical thesis of
120, 124-5, 143 actuality; tense
initial conditions, 113--14
final conditions, 113--14 instantaneous objects, 19, 40 see
Findlay, J. N., 149 also temporal parts
fission, 41, 43--4, 50 instants, 37, 38, 39, 135-6
Flew, A., 146, 149 see also propositional theory of
Forbes, G., 21-2, 149 instants
forwards hypotheses, 139-40 intrinsic properties, 20, 21, 22, 70,
free will, 36 72, 73, 124-5, 143
fusion 41, 43--4, 50 'inus' condition, 92
future Ishiguro, H., 85, 150
asymmetry with past, 46,
10S-10, 129-30 Johnston, M., 67--8, 72-3, 150
epistemological determinacy of, Joule, J., 86
130
fact, 32, 37 Kant, I., 83, 86, 89-92, 145, 150
individuals, 38 Kemp Smith, N., 145
ontological determinacy of, 4, 5, Kenny, A. J. P., 74, 75, 150
32, 130, 146 Kripke, S., 40, 41, 150
physical determinacy of, 142,
146 Lacey, H. M., 6, 150
see also A-series; determinism; Lawrencium, 17
past; present; tense laws, 118
Index 157

see also determinism; physical monads, 84-5


possibility Moore, A. W., xi
Leibniz, G. W., 84-5, 86, 145, 150 Moore, G. E., 65, 151
Leibniz's Law, 42-4 Mundie, W: C. K., 143, 151
Lemmon, E. J., 27
Le Poidevin, R., 2, 144, 150 Napoleon, 39-40
Levison, A. B., 29-30, 150 necessity
Lewis, D., 20-3, 35, 46, 67, 73, axiom in T concerning, 48, 50
96-7, 103-110, 145, 146, 150, of identity, 48, 50, 51
151 Newton, Sir 1., 7, 12
Lloyd, G., 143, 151 second law of motion, 92
Locke, D., 70, 151 third law of motion, 10, 91-2
Locke, J., 84, 85, 151 Newton-Smith, W. H., x, 6-7, 27,
Loemker, L. E., 84, 85, 151 79, 94, 100-3, 134-5, 151
logical priority, 103 Nietzsche, F., 46
Loizou, A., 144, 151 non-existent individuals, 39-40,
Lombard, L. B., 151 51-3, 71-2
Loux, M., 151 non-simultaneity of causation,
Lowe, E. J., 28-9, 144, 151 proof of, 86-92
Lucas, J. R., 3, 151 Noonan, H. W., 63, 151

McArthur, R., 144, 151 object-orientated account of


MacBeath, M., x, 144, 151 change, 79, 125, 126-7
McGinn, C., 4, 151 objects
Mackie, J. L., 92, 119, 132-4, 151 contrasted with events,·58-9
McTaggart, J. MeT. E., 1, 3, 8-9, four-dimensional, 58, 59--60, 65
13, 16, 17, 24-35, 57, 129, of perception, 64
143-4, 145, 151 temporal parts of, see under
McTaggart's paradox, 8-9, 13, temporal parts
24-35, 57, 129, 143-4 three-dimensional, 59--60
Massey, G., 151 of thought, 65
Mayo, B., 36, 151 ontology, 6, 59, 60, 61
Meakin, K. A., xi Ouspensky, 20-1, 76, 126
Mellor, D. H., x, 3, 4, 31-2, 59, overdetermination, 91, 138
72-3, 102, 125--6, 143, 144,
146-7, 151 Papineau, D., 112-14, 120, 145,
mereology, 21, 22 151
meta-time, 30-1 Parfit, D., 41, 151
Millar, A., x past
Millican, P. J. R., 146 asymmetry with future, 46,
modal realism, 20, 22, 35 108-10, 129-30
modality, 20,22,48-9, 56,73 causal isolation of, 129 see also
anti-realist analysis of, 39, 46 backwards causation
McTaggart's paradox and, 34-5 fact, 32-3, 40
Principle of Reciprocity and, individuals, 38, 39-40
80-1 physical determinism and, see
time order and, 92-4 determinism, historical
momentum, 88-9, 92 times, 46
see also energy unreality of, see temporal
158 Index

solipsism propositional theory of instants,


see also A-series; future; present; 37,44, 46--8,53--7,136
tense Putnam, H., 146, 152
'pathological' cases of diachronic
identity, 40-4 quantification, 6, 37, 40, 51
perception, 64-5, 69-70 quantified modal logic, 67
Percival, P., 55, 151 quantum physics, 132
phenomenalism, 44 Quine, W. v. 0., 6, 58, 59-60, 63,
physical determinism, see under 65, 67, 152
determinism Quinton, A., 134, 152
physical possibility, 80-1, 92,
95--6, 104, 120-1, 126, 131, 132
possibilia, 80 R, see Reciprocity Principle of
possible worlds, 20, 21-2, 34-5, Ramsey, F. P., 48-53, 152
56, 94-8, 144 Read, S. L., x
counterfactuals and, 104, realism, 38, 102
106-110, 136-7, 144 realism, 102
'ersatzist' theories of, 46 Reciprocity, Principle of, 10,
see also modality 83--98, 120-2, 145
predicate modifiers, 14, 73--5, 78 defined, 86--8
pre-established harmony, 85 historical antecedents of, 84--6,
present 145
fact, 4, 13, 33--4, 37--8, 48 modal properties of, 92-4
reality of, see temporal solipsism in proof of non-simultaneity of
redundancy theory of, 14, 37 causation, 8S-92
uniqueness of, 68 time without change,
see also A-series; future; tense consequences for, 94-8
Prior, A. N., 3, 4, 6, 9, 33, 34, recurring history, 46, 143
69-70, 72, 143, 152 reducibility
events, treatment of, 29-30 factual, 2-3
instants, theory of, 53--7, 136 meaning, 2-3
temporal solipsism, 36-40, 42-4, see also A-series; B-series
47, 52-3 reductionism, 6, 37--8, 44--8, 53-4,
tense logic of, 67--8, 144 98, 136-7
probabilistic theory of causal reference, 40
priority, 110-16, 120 Reichenbach, H., 131, 135, 146,
probability, 10, 110-16 152
processes, 18, 60, 65--6, 76-7, 127, causal theory of time order,
144 80-1, 131-5, 145
properties, 69-70 relationism, 16, 46, 70, 98, 136-7
of processes, 66 relativity of simultaneity, 7, 32,
relational theory of, 62, 72-3, 74 132
see also extrinsic properties; Relativity, Special Theory of, 7--8,
intrinsic properties 19, 32, 58, 66-7, 129, 132, 134
property-orientated account of relativized predicates, 69-72
change, 79, 125 Rescher, N., 152
propositions, scholastic theory of, rigid designation, 144
5.>-7 Rumford, Count, 86
Index 159

Russell, B. A. W., 4, 16, 58, 65, TEG, principle of, 50


145 TI, principle of, 50-1
account of change, 77-9, 152 TMI, principle of, 50-1
Ryle, G., 36, 146, 152 Taylor, R., 145, 153
temporal asymmetry, 99-122
Salmon, W., 110, 115, 152 temporal determinant, 14-15, 25,
Schlesinger, G., 30-1, 152 62, 67-75
screening off, 112, 113 temporal ontology, 6, 17-23
Shoemaker, S., 6, 27, 94, 153 temporal parts, 8, 9-10, 17-18,
simultaneous causation, 10, 66, 19-23, 57, 58-75, 79, 123-5,
81, 82, 83-98, 99, 122, 123, 127, 129, 144-5
128, 145 alternative logics to, 67-75
singular causal statements, 111, epistemological thesis of, 62-5
115, 139 formal thesis of, 62, 63, 64
Sklar, L., 132, 146, 153 minimal thesis of, 62-7, 76
Skyrms, B. 110, 153 temporal solipsism, 4, 9, 33-5, 144
Smart, J. J. C., 4, 153 temporal variation, 5, 7, 8, 15,
Sosa, E., 145, 153 16--17, 19, 57, 123-4, 126
Souffle, w. z., xi tense, 1, 3, 4, 5, 24-7
space iterated, 24-5, 26, 28-9
change in regions of, 77 propositions and, 54
time, contrasted with, 5, 8, 13, token-reflexivity of, 3-21, 54
18, 26 see also A-series; indexicality
see also spatial operators; spatial tense logic, 7, 57, 143
parts philosophical basis of, 34
space-like separation, 8 principles, 7, 42, 50-1
space-time interval, 8, 134 systems: CL, 47; Kt, 27, 47, 50;
space-time parts, ~7 PCr, 27
spatial operators, 144 tense operators, 14, 24-5, 37, 39,
spatial parts, 9, 18, 19, 58--67, 42-4, 50, 143
70-71, 78, 123-5, 128 exclusivity of, 24-27
spatial variation, 7, 8, 15, 16--17, metric, 42, 54
19, 78, 79-80, 93, 123-4, 126, Tensed Existential Generalization,
128-9 Principle of, 50
spatially tensed facts, 13, 18, 145 tensed fact, 3, 32-3, 129, 145
Squires, J. E. R., x see also A-series
Stalnaker, R., 35, 143, 144, 146, 153 Tensed Identity, Principle of, 50-1
static causation, 66, 86--7, 89-92, Tensed Material Implication,
94-8, 121 Principle of, 50-1
static theory, 143 Tensed theory
Stevenson, L. F., x change and, 5, 9, 13-17, 18, 23,
Stirling, C., 75, 148 125
Strawson, P. F., 59, 153 contradiction in, 8-9, 24-7
sumo wrestling, 100, 107-8, 111 conflict with STR, 132
supervenience, 3 defined, 3-4
Suppes, P., 110, 153 temporal determinants and, 68
Swinburne, R. G., xi, 6, 7, 134, temporal parts and, 61
143, 153 see also Tenseless theory
160 Index

Tenseless theory, 9, 10, 11 theory


causality and, 100; see also Tenseless theory of, see
change, causal theory of Tenseless theory
change and, 8, 1~23, 57, 98, topology of, see topology of time
145 two-dimensional, 30-1
change, theory of, 76--82, 122, unity of, 131-2, 134
12~ unreality of, 9, 13, 24, 26-7, 129
defined, 4-5 without change, 6, 10, 27, 94-8,
temporal determinants and, 135, 136, 143
67-8 time-like separation, 8
temporal parts, commitment to, times, 44, 46, 55, 56, 77-8, 135-7
61-7 change of, 13
see also, Tensed theory see also instants
tensers, 143 token-reflexives, see indexicality;
time tense
branching, 47 topology of time, 6-7, 27, 47-8,
causal theory of, 129-37 77, 93, 110, 127-9, 137, 142
change and, 1, 5, 9, 10, 1~23, Tractatus, The, 32
25, 26 trans-world identity, 20-2, 41
closed, see closed time
contingent existence of, 136 universe, spatio-temporal unity of,
dense, 47 132-4
dimension of causality, 129 Urquhart, A., 152
dimension of change, 5, 78, 81,
129, 131 vacua, temporal, 27, 94-8
direction of, 130-5, 140-1 van Fraassen, B., 146, 153
discrete, 136 van Inwagen, P., 153
disjoint, 131-2, 134 verification, 44-5, 101, 102
distinctions in, see A-series; vis motrix, 85, 145
B-series; tense vis viva, 86, 145
events and, see events
flow of, 4, 25, 26, 29, 30, 76,
Warburton, N., xi
142, see also A-series
Weir, S., 147, 153
-like variation, 8
Weiss, B., xi
linear, 117, 128, 137, 142
White, R. M., 146
order, 10, 80-1, 131-5
Williams, S. G., xi, 3, 153
priority of, 99-103
Wilson, N., 59, 69, 70, 153
Relativity and, 7-8, 19
Woolhouse, R. S., 145
-reversal invariance, 109
space, comparison with, 5, 8,
13, 16, 18, 26, 64, 76, 129 Young, T., 86
symmetry, 104 Yourgrau, P., 153
tense and, see tense
Tensed theory of, see Tensed Zemach, E., 153

You might also like