Conjoining Meanings Semantics Without Truth Values Paul M Pietroski Full Chapter PDF

Conjoining Meanings: Semantics
Without Truth Values Paul M. Pietroski

Visit to download the full and correct content document:
https://ebookmass.com/product/conjoining-meanings-semantics-without-truth-values-
paul-m-pietroski/
More products digital (pdf, epub, mobi) instant
download maybe you interests ...
English Prepositions: Their Meanings and Uses R. M. W.

Dixon
https://ebookmass.com/product/english-prepositions-their-
meanings-and-uses-r-m-w-dixon/
God in Moral Experience - Values and Duties Personified

Paul K. Moser
https://ebookmass.com/product/god-in-moral-experience-values-and-
duties-personified-paul-k-moser/
No Truth Without Beauty: God, the Qur’an, and Women's

Rights 1st ed. 2022 Edition El-Ali
https://ebookmass.com/product/no-truth-without-beauty-god-the-
quran-and-womens-rights-1st-ed-2022-edition-el-ali/
Power System Protection, 2nd Edition Paul M. Anderson
https://ebookmass.com/product/power-system-protection-2nd-
edition-paul-m-anderson/
Meanings as species First Edition. Edition Richard
https://ebookmass.com/product/meanings-as-species-first-edition-
edition-richard/
POW/MIA Accounting 1st ed. Edition Paul M. Cole
https://ebookmass.com/product/pow-mia-accounting-1st-ed-edition-
paul-m-cole/
Men, Masculinities, and Earth: Contending with the

(m)Anthropocene 1st Edition Paul M. Pule (Editor)
https://ebookmass.com/product/men-masculinities-and-earth-
contending-with-the-manthropocene-1st-edition-paul-m-pule-editor/
Inquisitive Semantics 1st Edition Ivano Ciardelli
https://ebookmass.com/product/inquisitive-semantics-1st-edition-
ivano-ciardelli/
Beyond Semantics and Pragmatics Gerhard Preyer
https://ebookmass.com/product/beyond-semantics-and-pragmatics-
gerhard-preyer/
OUP CORRECTED PROOF – FINAL, 15/3/2018, SPi
Conjoining Meanings
CONTEXT AND CONTENT

SERIES EDITOR : François Recanati, Institut Nicod
Other titles in the series:

The Inessential Indexical
On the Philosophical Insignificance of Perspective and the First Person
Herman Cappelen and Josh Dever
Fixing Reference
Imogen Dickie
Propositional Content
Peter Hanks
The Mirror of the World
Subjects, Consciousness, and Self-Consciousness
Christopher Peacocke
Assessment Sensitivity
Relative Truth and its Applications
John MacFarlane
Context
Robert C. Stalnaker
Conjoining Meanings
Semantics Without Truth Values
PAUL M. PIETROSKI
1
3
Great Clarendon Street, Oxford, OX DP,
United Kingdom
Oxford University Press is a department of the University of Oxford.
It furthers the University’s objective of excellence in research, scholarship,
and education by publishing worldwide. Oxford is a registered trade mark of
Oxford University Press in the UK and in certain other countries
© Paul M. Pietroski 
The moral rights of the author have been asserted
First Edition published in 
Impression: 
All rights reserved. No part of this publication may be reproduced, stored in
a retrieval system, or transmitted, in any form or by any means, without the
prior permission in writing of Oxford University Press, or as expressly permitted
by law, by licence or under terms agreed with the appropriate reprographics
rights organization. Enquiries concerning reproduction outside the scope of the
above should be sent to the Rights Department, Oxford University Press, at the
address above
You must not circulate this work in any other form
and you must impose this same condition on any acquirer
Published in the United States of America by Oxford University Press
 Madison Avenue, New York, NY , United States of America
British Library Cataloguing in Publication Data
Data available
Library of Congress Control Number: 
ISBN ––––
Printed and bound by
CPI Group (UK) Ltd, Croydon, CR YY
Links to third party websites are provided by Oxford in good faith and
for information only. Oxford disclaims any responsibility for the materials
contained in any third party website referenced in this work.
Contents
Preface vii
Chapter zero: Overture 

Chapter one: Locating meanings 
Chapter two: Introducing concepts 
Chapter three: Invention and satisfaction 
Chapter four: Truth or understanding 
Chapter five: Events and framing 
Chapter six: Massively monadic, potentially plural 
Chapter seven: Minimal semantic instructions 
Chapter eight: Reprise 
References 
General Index 
Index of names 
Preface
It’s hard to say when I started writing this book. In , I was in the early stages of
writing a book that was beginning to address similar issues. That activity initiated a
scattered process that resulted in this product. The descendant bears little resem-
blance to its ancestor. But at some point, there was a book that was this one, whose
main thesis is that meanings are instructions for how to access and assemble concepts
of a special sort. One implication of my view is that ‘book’ does not have an
extension, and that semanticists don’t need to worry about what books are, or
when they come into existence. So let me turn to the more important matter of
thanking some people who provided help, along the road that led to this progress
report, and apologizing to the many people who provided help that I have forgotten
or never recognized. I have benefited from many academic villages and their citizens,
in ways which guarantee that much of what I think is due to them, in ways I cannot
trace. But I do remember some of my debts.
The largest is to my friend and colleague, Norbert Hornstein. I admired his Logic
as Grammar before I got to know its author. But references to Norbert’s written
work, and our collaborative efforts, fail to capture his deeper influence. We have been
talking about this stuff, usually over lunches, for most of this century. Norbert
provided insightful comments on many drafts. More importantly, he has been an
unfailing source of encouragement, good advice, and general camaraderie. I initially
met Norbert via Jim McGilvray, a friend and former colleague at McGill University.
I learned a lot from Jim, who regularly challenged my youthful fondness for truth
conditional semantics. At McGill, I was also part of a reading group that included
Jim, Dave Davies, and two linguists—Mark Baker and Brendon Gillon—who make
appearances in the pages below. In retrospect, I see being part of that group as a
formative experience with regard to how I currently think about meaning, syntax,
and truth.
Going further back, to teachers at MIT, I was ridiculously lucky. George Boolos,
Jim Higginbotham, and Richard Larson each had a major impact that is reflected in
several chapters. And no version of this book would exist if not for Noam Chomsky.
It will be obvious that his work plays a central role in mine, and that several aspects of
my proposal are due to him. Noam showed me a fruitful way of thinking about
philosophy of language and linguistics. I am enormously grateful for this, and for his
generous support. Also, his examples are pretty good.
In –, I went back to Cambridge as a fellow in the Mind, Brain, Behavior
program at Harvard. During a dream year there, I attended terrific lab meetings in
Psychology, taught a seminar and met wonderful students in Philosophy, and worked
through some material that eventually became chapters in a reading group organized
by Cedric Boeckx—an early supporter of this project and a consistently stimulating
lunch partner—all while living in Boston’s North End (post Big Dig) and getting to
watch the Red Sox win the World Series again. Thanks to Marc Hauser, who made
viii Preface
the visit possible, Susan Carey, Elizabeth Spelke, Bernhard Nickel, Susanna Siegel,
Jake Beck, and Dennis Ott; though listing makes me sure I’m forgetting.
Between  and , I had the chance to present newer versions of the ideas at
the Center for Advanced Study in Theoretical Linguistics in Tromsø, the Language
Research Group at Durham University, the Center for the Study of Mind in Nature in
Oslo, Beihang University in Beijing, and the Pontifical Catholic University of Rio
Grande do Sul in Porto Alegre. Thanks to the many who arranged for and partici-
pated in those extended discussions; special thanks to Peter Svenonius, Gillian
Ramchand, Wolfram Hinzen, Robyn Carston, Francis Lin, and Ana Maria Tramunt
Ibanos. A more recent trip to the CSMN in Oslo was partly arranged by Terje
Lohndal, a former Maryland student who became a collaborator and friend. The
emerging connection to Norway has been an unexpected delight. They even turned
on the Aurora Borealis one night in Tromsø.
In June of , I gave a series of “Context and Content” lectures at the Institut
Jean Nicod. Many thanks to François Recanati for the invitation and the conversa-
tions it led to with him and others, including Pierre Jacob, Philippe Schlenker, Kit
Fine, and whoever happened to be passing through in a given week. I’ve never gotten
through fewer slides in talks, given the discussions that would break out in the first
two minutes, or had more fun presenting material. If forced to pick a place in which
this book was born, I would have to say it was Paris. Preparing the talks—in the
delightful apartment that François had suggested, or just wandering from one café
to another—led to a draft that I tried out with the Foundations of Semantics Group
at the University of Konstanz in . Thanks to Brendan Balcerak Jackson for
organizing that visit and a series of conversations that included Magdalena Balcerak
Jackson and Irene Heim. Those experiences led to a revised draft, which was
subjected to trial by graduate seminar and eventually became a version that went
off to OUP.
The help did not end there. Peter Momtchiloff, patient as always, sent the
manuscript to a pair of referees who together provided the most valuable reports
I have encountered. They were kind, encouraging, and just plain right about some
things that needed to be fixed. That led to a major rewrite, which took a while to
settle. John Collins, who deserves a special category, offered incisive, informed, and
intelligent comments at several stages. The graduate student philosophy and linguis-
tics group (PHLING) at Maryland provided valuable feedback on the penultimate
draft in the fall of , when I was able to devote time to getting it finished, thanks to
the goodwill of my chairs, Bill Idsardi and Chris Morris.
More generally, the University of Maryland provided the daily setting that made
the book happen. I’m also grateful for the occasional sabbatical support that let me
spend time elsewhere, including a favorite spot in northern New Mexico where many
pages were written. But being at UMD, in linguistics and philosophy, is what led to all
this. The initial prompt was some seminars led by Norbert and Juan Uriagereka,
whose blackboard got well used while we talked in his office. Their students would
come into my office, wanting to know how semantics was related to Chomsky’s
minimalist program. So we talked. When Howard Lasnik joined the group, watching
him teach syntax helped me understand the subject in a new way. Being at Maryland
Preface ix
also allowed me to learn about psycholinguistics, on the job, by collaborating with

Stephen Crain and then Jeff Lidz. After meeting with Stephen to discuss the latest
experiments involving Kermit, Grover, and anaphoric dependence, I could go talk
with Georges Rey—who I have known and admired since my days in grad school—
about conceptions of intentionality.
Jeff connected me to Justin Halberda at Johns Hopkins; and the three of us have
spent a lot of time thinking together about how meanings are related to mental
representations. Our meetings and joint work have been deeply satisfying, intellec-
tually and personally, in ways I couldn’t have imagined without Jeff and Justin. They
and a series of amazing students—Tim Hunter, Darko Odic, and Alexis Wellwood—
helped me see how experimental tools really could, given the right team, be brought
to bear on old questions about logically equivalent logical forms. In a related vein, Jeff
Horty’s work on Frege played an important role in my thinking about lexicalization,
as did Peter Carruthers’ proposals about the role of the human language faculty in
integrating modules. Valentine Hacquard and Alexander Williams, colleagues in
semantics, were great conversational partners who provided sympathetic ears along
with needed doses of facts and skeptical questions. This book is very much a
Maryland product.
I can’t list all the students who helped by asking penetrating questions, and the
audiences who had me rethinking issues on the way home after presentations at
colloquia or conferences. But special mention is due to Chris Vogel, who helped with
the index and in other ways via many thoughtful discussions connected with his
dissertation. Visits to the University of Arizona turned out to be particularly valuable;
thanks to Massimo Piatelli-Palmarini, for much over many years, and to Heidi
Harley and Tom Bever for conversations in Tucson and Maryland.
Many of the ideas in this book were first presented in public at Rutgers University
in colloquia, or at events caused by Ernie Lepore. I met Ernie at a conference when
I was still teaching at McGill and just starting to write papers in philosophy of
language. He became a generous mentor and invited me to the initial philosophy/
linguistics workshops that he organized, back when they were small local affairs.
Through those growing sessions, I got to meet or reconnect with terrific philosophers
and semanticists, including many who I need to thank for comments and discussions
that led to many deletions of errors and additions of good points: Zoltán Szabó, Jason
Stanley, Rich Larson, Peter Ludlow, Jim Higginbotham, Michael Glanzberg, Rob
Stainton, David Braun, Steven Gross, Elisabeth Camp, Bob Matthews, Frankie Egan,
Kirk Ludwig, Thony Gillies, Roger Schwartzschild, Kai von Fintel, Veneeta Dayal,
Ken Safir, Barry Schein, and Barry Smith. The two Barrys, along with Ernie and
Martin Davies, have been especially important in shaping my thinking about Donald
Davidson’s project and its relevance for cognitive science. In terms of intellectual
influence, my debt to Jerry Fodor is huge, as will be clear from the text.
I was an undergrad at Rutgers. So it has been a real treat to visit regularly, most
recently on the university’s th birthday. (Another implication of my view about
linguistic meaning is that proper nouns like ‘Rutgers’—‘Barry’, ‘Balcerak Jackson’,
etc.—do not have denotations, and that it’s not my job to worry about when my alma
mater began.) I’m writing these words in a transitional phase, thrilled by knowing
x Preface
that I’ll soon be joining the faculty there, while also feeling real gratitude for my time
at Maryland.
Through it all, my partner and best friend, Susan Dwyer, offered sanity checks and
some of the best conversations about the important issues. In her work, Sue juggles
so many balls that I get tired just watching. But she found ways to provide lots of
patience, support, and love.
In short, this book reflects the various neighborhoods I have been fortunate to
hang out in. A lot of resources that were not mine have been invested in this project.
Thanks to all.
Chapter zero
Overture
Some animals undergo dramatic changes after birth. Caterpillars go through a pupal
stage and emerge with scaled wings. Humans acquire languages of a special sort. Such
changes can transform an animal’s lifestyle. Butterflies flutter. People talk. In talking,
we use expressions that are meaningful, pronounceable, and remarkably combinable.
As children, we acquire languages whose expressions have these striking properties.
This book is about our linguistic expressions, their meanings, and how these mean-
ings are related to human thought.
I think meanings are composable instructions for how to access and assemble
concepts. More specifically: lexical meanings are instructions for how to fetch
concepts that can be combined in certain systematic but limited ways; phrasal
meanings are instructions for how to build monadic concepts (a.k.a. mental predi-
cates) that are massively conjunctive. On this view, meanings are recipes for how
to make mental representations of a special kind. For example, the meaning of
‘red dot’ is a recipe for how to assemble a monadic concept from ingredients
that can be accessed via the lexical items ‘red’ and ‘dot’. In defending this account
of what meanings are, I’ll argue that meanings do not determine extensions, and
that ordinary sentences do not have truth conditions. But words like ‘meaning’
and ‘concept’ get used in many ways. So let me start by saying a little about the
topic and my terminology.
. Human linguistic meanings

Many things are meaningful in some sense. Smoke often indicates fire, and a red light
may be a signal to stop. In some contexts, a shrug can convey a lot of information;
see, e.g., Grice (, ). But linguistic expressions are meaningful in an inter-
esting combinatorial way.
. Distinctively composable

A speaker of English might use ‘red’ to report the color of a nearby apple. This word
can also be used along with others to assert that some distant apples are not red, to
ask whether blood is red, or to call for a certain bucket of paint. Similarly, utterances
of ‘dot’ need not be symptoms of dots. Words do not merely signify. Words are
significant in a way that is interwoven with how they can be combined. The meanings
of ‘red’ and ‘dot’ are intimately related to the meaning of ‘red dot’, which has
 . Overture
something in common with the meaning of ‘green line’, which somehow depends on
the meanings of ‘green’ and ‘line’. In general, each meaningful expression is a
constituent of boundlessly many others. So whatever linguistic meanings are,
humans can connect many of them with pronunciations, and thereby create expres-
sions that can be used—systematically—to talk about many things.
These expressions, of a language like spoken English or ASL, seem importantly
different from any analogs in animal communication systems. We can say that
forager bees use a dance language to indicate locations of food. But even if bees
thereby communicate intentions, their dances do not have meanings of the kind
exhibited by phrases like ‘food over there’. The languages that children naturally
acquire also differ from various notational systems that adults have introduced, for
scientific purposes, by explicitly formulating rules for how to generate and interpret
certain symbols. Such notation can be useful in many ways. But expressions of an
invented mathematical language may not be meaningful in the ways that ordinary
words and phrases are. Likewise, human linguistic expressions may lack a kind of
significance that is exemplified by bee dances, or by invented formulae whose
interpretations are stipulated. We can and should allow for many types of languages,
whose expressions may be significant in sundry ways.
For these purposes, let’s say that something counts as a language if it connects
interpretations of some kind with signals of some kind. If this conception of
languages is too generous, that will do no harm. The spoken or signed languages
that children can acquire as native languages, given ordinary human experience, can
be described as special cases. It will be convenient to have a term for these distinct-
ively human languages. I’ll call them Slangs. But whatever we call them, Slangs
connect interpretations of a special sort with signals that are spoken or signed.
I will use ‘meaning’ and ‘pronunciation’ to talk about these human interpretations
and signals. Given this terminology, Slangs connect meanings with pronunciations.
So whatever meanings are, Slangs connect them with pronunciations.
. Pronounceable instructions

This leaves room for many proposals about what Slangs are, and how these child-
acquirable languages are related to other things (e.g., concepts, expressions, and
aspects of environments that speakers share). As discussed in chapter one, I think
Slangs are biologically implementable procedures that generate expressions; see
Chomsky (, ). But this is not a matter of definition. I will argue against
other conceptions of Slangs, in order to stress that meanings are connected with
pronunciations via expression-generating procedures that children can acquire.
I think children also have languages of thought that connect certain interpretations—
perhaps mind-independent objects, properties, or possibilities—with mental symbols.
We don’t know how such symbols are related to meanings. But many animals represent
things. When a dog chases a rabbit, the dog tracks its moving target. And a child can
represent the rabbit, as an entity distinct from the dog, without having a word that can
be used to talk about either animal. Indeed, in acquiring the word ‘rabbit’, a child may
well connect a prior representation of rabbits with a pronunciation. Though not
everything that gets linked to a pronunciation is a meaning.
. Meanings are not concepts 
I’ll argue that lexical meanings are instructions for how to access mental symbols
that may be introduced in the course of acquiring a lexicon. In acquiring words like
‘rabbit’ and ‘chase’, a child may use available representations to create new concepts
that exhibit a distinctive representational format. Moreover, as a child matures, a
lexical item that was initially linked to a single concept may become an address for a
family of accessible concepts. On this view, Slangs play an important role in cogni-
tion, not just in communication. So my proposal is deeply mentalistic. But my use of
‘meaning’ and ‘language’ is more neutral.
Others can hypothesize that the meaning of ‘rabbit’ is a set of rabbits, or a
mapping from possible worlds to sets of rabbits, and that Slangs are sets of
meaning-pronunciation pairs; see, e.g., Lewis (a). As discussed below, I think
this abstraction from psychology has outlived its utility. But these are empirical
issues, not matters for stipulation. We have to figure out what meanings are, even
if we describe them as things that Slangs connect with pronunciations.
For present purposes, we can remain neutral about what pronunciations are.1 One
can start with the idea that spoken Slangs connect meanings with certain “sounds,” in
some sense that abstracts from many dimensions of variation across acoustic signals,
which may be produced by sopranos from Sydney or baritones from Brooklyn. But
refining this idea, say in terms of phonological features, calls for investigation as
opposed to stipulations regarding the operative notion of sound. It is even less clear
what Slangs connect with pronunciations. Though it may be useful to locate my
proposal about meanings relative to some more familiar alternatives, and hint at
some arguments that are developed in later chapters.
. Meanings are not concepts

One historically important claim is that meanings are concepts. A more recent
suggestion, discussed in section three, is that meanings are extensions of idealized
concepts. I think the older idea is wrong, and the newer one is worse. I’ll return to the
proposed alternative in section four.
. Concepts: Symbols vs. contents

Like ‘meaning’, ‘concept’ gets used in many ways. I use the ‘c’-word to talk about
mental representations of a special sort; see also Margolis and Laurence ().
Peacocke () and many others use ‘concept’ to talk about Fregean senses, or
other abstract entities that exhibit logical relations. In ordinary speech, the word
plays various roles depending on the context. But along with Fodor (, ,
, ) and many others, I want to stress that animals often represent aspects of
their environments by employing mental symbols that are composable in the follow-
ing sense: each can be combined with some others to form complex representations.
Prima facie, many animals use such symbols to think about food, nests, distances,
1
Though unsurprisingly, I tend to think of pronunciations as instructions for how to articulate. Halle
(, p. ) describes such instructions—executed by certain anatomical structures of the vocal tract—as
“choreographic scores” that tell each anatomical “dancer” what to do and when; see §. of chapter seven.
 . Overture
directions, partners, predators, events, causal relations, social dominance, etc. I will
use ‘concept’ to talk about composable mental symbols that can be used to think
about things.
As discussed in chapter two, the relevant notion of “thinking about” allows for
concepts that are not as systematically composable as Slang expressions. There may
be minds, perhaps equine, that can form concepts of delicious grass and scary rabbits
without being able to form concepts of scary grass and delicious rabbits. However, to
think about a rabbit is to think about it as a thing of some sort (e.g., as a rabbit, or as
something that was chased by a dog). In this sense, concepts have contents that can
be described as ways of thinking about things; see Evans (). A concept that can
be used to think about something as a rabbit, whatever that amounts to, has a content
that we can gesture at by talking about the concept type RABBIT. An instance of this
type is a mental symbol that can be used to think about a rabbit as such, or to classify
something—perhaps wrongly—as a rabbit; see Fodor (). A concept of the type
RABBIT-THAT-RAN, which can be used think about something as a rabbit that ran, is
presumably a complex mental symbol whose constituents include an instance of
RABBIT. A thought can be described as a sentential concept that lets us think about (some
portion of) the universe as being a certain way. Thoughts of the type A-RABBIT-RAN can be
used to think about the world as being such that a rabbit ran.
Classifying concepts in terms of contents allows for instances of RABBIT that differ
in other respects. Thinkers may employ mental symbols that are formally distinct
concepts of this type. We can distinguish ‘•’ from ‘⊗’ even if in some language, both
symbols indicate the same operation of multiplication. Perhaps a conceptual content
can also be mentally encoded in diverse ways. That said, distinct minds may often use
the same symbols. Children who encounter similar rabbits may acquire a common
instance of RABBIT. Likewise, instances of RED-DOT may include a common concept of
conjunction, even if the instances of RED and DOT are diverse. For present purposes,
we can be agnostic about how multiply realizable concept types are.2
We can also distinguish two versions of the idea that meanings are concepts:
(i) each meaning is a concept type like RABBIT or RED-DOT; (ii) each meaning is itself a
concept, and hence a token of some such type. On the first view, speakers of the same
language can connect a pronunciation π with a meaning μ via distinct concepts of the
type μ. On the second view, such speakers connect π with distinct meanings/concepts
of type μ. I think this contrast is largely terminological, and that in any case,
meanings are neither concepts nor concept types.
. One meaning, a family of concepts

In my view, a meaning can be the Slang correlate for two or more conceptual
contents. This phenomenon is often described in terms of polysemy. My aim is not
to provide an account of polysemy. But the phenomenon suggests that a meaning
need not correspond to a single concept.
2
As Piatelli-Palmarini () discusses, our immune systems respond to new viruses with the same
antibodies. Of course, utterances/inscriptions of ‘rabbit’ can be spatiotemporally distinct, as can events of
using any particular instance of RABBIT. And let’s not worry, until §., about concepts like RABBIT-STUFF.
. Meanings are not concepts 
Consider the noun ‘book’ and a familiar contrast between two ways of thinking
about some corresponding things: as spatially located objects that can carry infor-
mation; or as chunks of information that can be encoded in many ways. Given
examples like ‘book that he defaced’ and ‘book that he plagiarized’, it seems that
‘book’ can be used to access concepts that have different contents, whatever we say
about the ontology of books. So let’s distinguish BOOK:VEHICLE from BOOK:INFO. If
neither type is the meaning of ‘book’, perhaps this lexical meaning lets us access
instances of either type. Some speakers may also have a neutral concept of books. But
in my view, the meaning of ‘book’ is not a concept, not even a permissive one.
Similarly, I don’t think there is a noun ‘circle’ whose meaning is a concept that
applies to certain ideal figures governed by geometric theorems and to certain
perceptible figures made of chalkdust on blackboards. We can use ‘line’ to speak of
Euclidean lines, fishing lines, telephone lines, waiting lines, lines in faces, lines of
thought, etc. We can use ‘door’ to access a concept of certain impenetrable objects, or
a concept of certain spaces that can be occupied by such objects. In later chapters,
I discuss ‘country’ and ‘France’, which can be used to access concepts of terrain or
political institutions. Polysemy is ubiquitous. So often, the meaning of a word seems
to be an aspect of a Slang expression that lets us access members of a certain family of
concepts.
To be sure, homophony is also common. The pronunciation of ‘bank’ can be used,
with different meanings, to talk about financial institutions or river edges. Following
the practice of lexicographers who distinguish ‘1bank’ from ‘2bank’, we can distin-
guish $BANK from ▽BANK. To take another example, I recently learned that the
pronunciation of ‘fish’ can be used to talk about a flat plate of metal that has been
attached to a beam, or across a joint, as an added source of support (especially as a
temporary repair to a damaged mast or spar). This was an instance of acquiring a
new lexical item, ‘2fish’. I didn’t learn that the more familiar noun ‘1fish’ has an
expansive meaning that can also be used to talk about both trout and certain flat
metal plates.
Perhaps some alleged examples of polysemy are really cases of distinct words
connecting the same pronunciation with different meanings. Maybe we use ‘1door’ to
talk about things that can block the doorways we speak of by using ‘2door’. Though
given doors to success, and windows of opportunity, polysemy appears to outrun
homophony. Some cases may be hard to classify, especially given the possibility of
metaphorical usage. But one generation’s metaphor can be another generation’s
polysemy. So even allowing for homophones and metaphors, it seems that a typical
word meaning corresponds to more than one concept.
The adjective in ‘bare skin’ shares its pronunciation with some words that get
spelled another way. But ignore the nouns that are used to talk about ursine animals
or pessimistic denizens of Wall Street. There is at least one verb meaning that we can
use to talk about certain episodes or states of carrying or supporting something.
We speak of bearing gifts, weight, malice, pain, a likeness to relatives, children,
names, and false testimony. Citizens may bear arms; things can bear watching; etc.
Dictionaries typically describe this variation partly in terms of homophony—two or
three verbs, each listed with a few subsenses—but not in terms of many separate
 . Overture
verbs used to talk about carrying or supporting or having or yielding or tolerating or

meriting something. Listing subsenses for a single word is, in my view, a way of
reflecting the fact that one lexical meaning can be the Slang correlate for a cluster of
concepts. However many meanings there are for ‘bear’, there seem to be more
corresponding concepts.
Prima facie, a pronunciation π can be connected with a meaning μ that is more
flexible than any concept that is accessed via π and/or μ. I think the pronunciation
bʊk is connected with a meaning that is sometimes used to access an instance of BOOK:
VEHICLE, and sometimes used to access an instance of BOOK:INFO. Of course, no single
example establishes that polysemy is a real phenomenon, distinct from homophony
and metaphorical usage. But there are many examples. So rebutting a few doesn’t
make it plausible that each lexical meaning is conceptually univocal.
It is also worth remembering that in acquiring lexical items, a child cannot treat
each encountered use of a pronunciation as the first encountered use of a new
expression. If a child has already connected the pronunciation bæŋk with an instance
of $BANK, and then hears a speaker using bæŋk to talk about something having to do
with a river, this might lead the child to distinguish ‘1bank’ from ‘2bank’. But there is
no reason to think that a typical child tries, or should try, to maintain a lexicon in
which each entry is linked to exactly one concept. Given that Slang expressions are
often used polysemously, it may be useful—e.g., for scientific purposes—to invent
languages whose expressions are univocal, in that each expression connects a per-
ceptible signal with exactly one concept. But the words that we naturally acquire may
be well suited to ordinary thought and talk among primates who can think about
things in many ways.
. Other kinds of conceptual equivocality

Whatever we say about homophony and polysemy, we need to be careful when using
Slang expressions (perhaps written in a special font) to talk about concepts. If an
instance of DOT is a concept with which one can think about something as a dot,
then thinking about something as a red dot may well require an instance of DOT. But if
an instance of RABBIT is a concept with which one can think about a rabbit as such,
then thinking about some stuff as rabbit may not require an instance of RABBIT.
Someone might acquire a mass-concept of rabbit in the way that many of us acquire a
concept of tofu—in response to some stuff on a plate—and only later acquire a
corresponding count-concept that can be used to think about a rabbit as a potential
source of rabbit. Such a person might initially think that rabbit is a vegan-friendly
substitute for chicken. Someone else might think that tofu comes from free-range
tofus, and that he once saw a tofu. Such a person might often use ‘tofu’, misguidedly,
to access a count-concept.
Let’s say that instances of RABBITC and TOFUC are count-concepts, while instances of
TOFUM and RABBITM are mass-concepts; see chapter two. This makes it vivid that an
instance of PREFERS-TOFUM-TO-RABBITM need not include an instance of TOFUC or
RABBITC. One can think about tofu and rabbit (as such) without thinking about
anything as a tofu or a rabbit. We can also posit types like [RABBITC STUFFC/M]M and
[TOFUM THINGM/C]C; where the slashed subscripts indicate concepts that combine with
. Meanings are not concepts 
a concept of one sort to form a concept of another sort. By hypothesis, an instance of

[RABBITC STUFFC/M]M is a complex mass concept. We can think about some stuff in this
complex way—i.e., as some stuff from a rabbit, or perhaps stuff from one or more
rabbits. But one need not think about rabbit in any such derivative way. Prima facie,
instances of RABBITM and TOFUM can be atomic concepts. And one needn’t be confused
to have an atomic mass-concept of rabbit. We can think of rabbit in this simple way,
if only to entertain and ascribe false thoughts according to which some rabbit on a
certain plate is textured soybean.3
The count/mass contrast reminds us that if a theorist uses ‘RABBIT’ to offer a
hypothesis about the meaning of ‘rabbit’, then we need to know which concept (type)
RABBIT is supposed to be, and which expression ‘rabbit’ is supposed to be. We can and
should distinguish the count noun ‘rabbit+CT’, which can be pluralized, from the
homophonous root noun ‘rabbit–CT’. I prefer to describe ‘rabbit–CT’ as a root noun
that is neither singular nor plural, rather than a mass noun.4 But whatever our termi-
nology, the meaning of ‘rabbit–CT’ seems to be neither RABBITM nor [RABBITC STUFFC/M]M
nor any instance of either concept type. Likewise, in my view, the meaning of ‘rabbit+CT’
is neither RABBITC nor [RABBITM THINGM/C]C nor any instance of either concept type.
The meanings are more neutral than any concept.
This neutrality might be described as a kind of polysemy. We can use ‘rabbit–CT’ to
access an instance of RABBITM or an instance of [RABBITC STUFFC/M]M. Perhaps concepts
of these types are coextensive, while instances of BOOK:VEHICLE and BOOK:INFO are not.
But two concepts can apply to the same stuff, yet provide different ways of thinking
about it. Likewise, a thinker might sometimes use ‘triangle’ to access a concept with
which one can think about the relevant figures as three-sided, and sometimes use the
same word to access a concept with which one can think about these figures as having
three angles that sum to  . If you know that Paderewski was both a musician and a
3
Faced with stuff from a rabbit, one can think that it is textured soybean; and this does not require an
instance of RABBITM or RABBITC. But the point concerns “de dicto” thoughts that we might correctly ascribe to
someone who thinks there are no rabbits; cp. the thought that tofu is not stuff from one or more tofus. Even
if RABBITM is necessarily coextensive with [RABBITC STUFFC/M]M, a thinker can have an atomic mass concept
of rabbit.
4
The relevant nouns include ‘furniture’ and ‘jewelry’. But furniture isn’t uncountable stuff, at least not
in the way that rabbit/beef/water is; jewelry includes countable rings and necklaces. For discussion, see
Gillon (), Bale and Barner (). We can think of beef as stuff from one or more cows; see Pelletier
(), who discusses David Lewis’s idea of a “universal grinder.” But ground tables and chairs would not
be typical furniture. Speakers of English use ‘hair–CT’ and ‘spaghetti–CT’ (cp. ‘a hair+CT’ and ‘a noodle+CT’)
to talk about the same stuff/things that speakers of French/Italian describe with plural count nouns
(‘cheveux+PL’, ‘spaghetti+PL’). And while sand might seem uncountable, until you focus closely, a pound of
gravel seems like a pound of pebbles. So it can be misleading to say that nouns of the form ‘ . . . –CT’ are mass
nouns, as if the grammatical property of being neither singular nor plural is tied to mass-concepts. Drawing
on Gillon, one can say instead that root nouns are neutral in this respect, and that the count feature adds a
restriction: count nouns cannot be used to fetch mass-concepts, which do not support counting; see §. of
chapter six. On this view, while ‘furniture–CT’ can be used to fetch a concept of things that are countable,
using a bare count noun will often suggest a mass-concept (apart from special locutions like ‘hunting
rabbit–CT’). For now, let’s not worry about the further complication that ‘rabbit’ can also be used to talk
about fur, or a jacket that incorporates such fur in a salient way; see Recanati (). If ‘rabbit+CT’ can also
be used to fetch a concept of a certain species, then the count noun is polysemous in yet another way.
 . Overture
politician, you might connect his name with two ways of thinking about him; cp.
Kripke (a). But the point of such examples is not to insist that a word is
polysemous if it is connected to concepts that have distinct contents. My claim is
that an account of what meanings are, and how they compose, must allow for cases of
conceptual equivocality that are not cases of homophony. And while this phenom-
enon is often illustrated with nouns or verbs, I think that modifiers also tend to be
conceptually equivocal.
Consider ‘green paint’, ‘green bottle’, and ‘green apples’. The adjective ‘green’ can
combine with a root noun or a count noun, singular or plural. Absent reasons for
recapitulating these grammatical distinctions for adjectives, it seems that the same
lexical item ‘green’ can be used to describe (i) some paint that is uniformly green,
(ii) a bottle that might be described as tinted but transparent, (iii) an apple that is
green only on the surface, or (iv) some bottles or apples, each of which is green.
Moreover, ‘green’ is not limited to an adjectival form. Some greens (or shades of
green) are greener than others. In this sense, ‘green’ is strikingly flexible in terms how
it can be used in combination with other expressions.
One can posit instances of GREENM as mass-concepts, instances of GREENC as count-
concepts, and even instances of GREENM-OR-C as neutral-concepts that can combine
with instances of either PAINTM or BOTTLEC. But in my view, the meaning of a flexible
word is not a correspondingly flexible concept. As discussed in later chapters, I have
nothing against concepts that are neutral along various dimensions. I think that
humans often use meaningful expressions to introduce concepts that are more like
words, in various respects, than the concepts we share with other animals. But if the
meaning of ‘green’ is used in the course of introducing an instance of GREENM-OR-C,
then presumably, the word meaning differs from the concept and its content.
Instead of saying that meanings are concepts, we can say that each lexical meaning
is an instruction for how to access a concept from a certain address, which may be
shared by a family of concepts. On this view, ‘green’ can be used to express a concept,
but there is more than one concept that can be expressed with the word. The meaning
of ‘green’ is apt for varied uses, including acts of describing some paint, some apples,
or both. The meaning of ‘green apples’ can be described as an instruction for how to
build a concept from lexically accessible ingredients. So the phrase need not, and
apparently does not, correspond to exactly one concept.5
In general, I think a Slang expression Σ can be used to access/build/express a
concept C that is less flexible than Σ—in terms of what Σ can be used to talk about,
and how it can combine with other expressions, compared with what C can be used
5
We also need a verb/noun distinction, since one can 1vfish for 1nfish and 2vfish a mast with a 2nfish. It is
tempting to specify the meaning of ‘1vfish’ as TRY-TO-CATCH-1FISH-IN-A-STANDARD-WAY. But someone can be
fishing without trying to catch anything; see Fodor and Lepore () on the hazards of decomposition. So
why think any one concept of 1vfishing applies to cases of fishing for trout, fishing hats from holes, fishing
for complements, etc? And even if speakers connect ‘1vfish’ with a suitably permissive concept type, it’s
hard to see how any such content could be defined in terms of 1FISH. But ‘1vfish’ and ‘1nfish’ are not mere
homophones; cp. ‘1vfish’ and ‘2nfish’. We need a conception of meaning that doesn’t commit us to
diagnosing lexical flexibility in terms of homophony, decomposition, or permissive concepts that are
exactly as combinable as the corresponding lexical items.
. Meanings are not extensions 
to think about and how it can combine with other concepts—since Σ might be used
to access/build/express a related but distinct concept C0 . As we’ll see, this flexibility
can promote a kind of cognitive integration. For certain purposes it can be useful to
regiment ordinary thought and talk in ways that approximate a scientific ideal of
using each expression as a perceptible sign of exactly one concept. Infants may also
link pronunciations to concepts in one-to-one fashion, at least initially. But as a child
acquires many words and concepts, partly in response to varied uses of Slang
expressions, a lexical item that provides an address for one concept may attract
related concepts. In this way, I claim, a lexical item can become polysemous. So we
shouldn’t expect the meaning of ‘green’ or ‘book’ or ‘book that I bought’ to be any
particular concept. We can instead describe the meaning of a Slang expression as an
instruction that can be executed in more than one way.
. Meanings are not extensions

In other chapters, I offer some further reasons for not identifying meanings with
concepts. But for these introductory purposes, let me turn to the idea that meanings
are extensions of ideal concepts that actually have extensions. I suspect that typical
concepts don’t have extensions, if only because of vagueness; see Sainsbury (),
cp. Williamson (). Though let’s suppose that each instance of POSITIVE-INTEGER
applies to a given entity, regardless of context, if and only if that entity is a positive
integer. Then instances of this type have an extension—{x: x is a positive integer},
a.k.a. {x: x is a natural number that zero precedes}. A set can be the shared extension
of distinct conceptual contents. So we can invent languages that connect signals with
extensions, leaving room for variation in how users enforce such connections. For
example, some thinkers might link the symbol ‘P’ to a certain set via instances of
POSITIVE-INTEGER, while others link ‘P’ to the same set via instances of NATURAL-
NUMBER-THAT-ZERO-PRECEDES. But in my view, Slang expressions do not have exten-
sions. I’ll argue that given polysemy, meanings do not even determine extensions.
Moreover, if each Slang expression has an extension, there is an obvious initial
objection to identifying meanings with extensions.
. Coextensive meanings

To use a standard example, ‘creature with a heart’ and ‘creature with a kidney’ have
different meanings. Yet the creatures with hearts are the same animals as the
creatures with kidneys. In general, it seems that if Slang expressions have extensions,
then many pairs of expressions have the same extension but different meanings. Alas,
replies to this simple point can get complicated.
Perhaps there could have been animals with hearts but no kidneys. So one might
say that there are “possible worlds” that include such animals, and that meanings are
sets of a special sort—i.e., functions that map each possible world onto some set, like
the set of things with hearts at that world. On this view, meanings are extensions of
ideal concepts that can be used to think about certain mappings from possible worlds
to extensions. But the objection remains. At every possible world, the numbers  and
 are the prime factors of , which is the second perfect number; and at every
 . Overture
possible world,  and  are the first two prime numbers apart from the successor or
predecessor of the fourth positive integer. Nonetheless, the meaning of ‘prime factors
of the second perfect number’ differs from the meaning of ‘first two prime numbers
apart from the successor or predecessor of the fourth positive integer’.
One might respond by saying that atomic meanings are extensions of some kind,
and that complex meanings are structured entities, built from atomic meanings.
On this modified view, phrasal meanings can be composite things that determine
extensions, and phrases with different meanings can be coextensive.6 But given words
like ‘unicorn’ and ‘ghost’, which seem to have atomic meanings, the initial objection
remains if there are no possible worlds at which there are unicorns or ghosts.
Similarly, if ‘woodchuck’ and ‘groundhog’ differ in meaning, the objection remains
if there are no possible worlds at which the woodchucks differ from the groundhogs;
cp. ‘Hesperus’ and ‘Phosphorus’. In §., I’ll argue that it doesn’t help to insist on the
possibility of unicorns, ghosts, and woodchucks that are not groundhogs. But first,
let me say why I follow Kripke () in thinking that there are no possible worlds
that include such things. For especially in linguistics, Kripkean intuitions are often
ignored when theorists invoke possibilities to salvage some version of the idea that
meanings are extensions.
Perhaps there could have been uni-horned horse-like animals. Maybe there are
such things on some actual planet in a distant galaxy. But if our uses of ‘unicorn’ have
been causally independent of any such creatures, possible or actual, one can’t just
assume that these unicornish things are unicorns as opposed to simulacra. We can
sensibly talk about possible horses that were never actually born—distinguishing
them from horse-like creatures of some other kind—since we can think about things
as animals of the same kind as actual horses. So we can sensibly talk about possible
scenarios in which an animal of that kind (i.e., a horse) is born with a horn, or
surgically altered to make it look like a unicorn. But if there are no actual unicorns
about which we have had some thoughts, then no possible thing is relevantly like
some actual unicorn that we have thought about. So it’s not at all clear that we can
sensibly talk about possible unicorns.
Perhaps many possible worlds include creatures that are born looking unicornish
in some respects. But if these possible creatures are diverse—in terms of their
appearance, history, and constitution—are they all unicorns? If not, what makes
some of them genuine albeit nonactual unicorns? Absent some plausible account of
what would distinguish possible unicorns from mere lookalikes, I doubt that there is
such a distinction. So I don’t think there could have been unicorns, given that there
are none. (I’ll return to the related idea that if a “kind-concept” actually applies to
nothing, then it applies to nothing at every possible world.)
Similarly, I don’t think there are ghosts at any possible world. So in my view,
‘unicorn’ and ‘ghost’ do not have distinct extensions. In which case, these words do
6
Perhaps some lexical items also have complex meanings. Examples may include nouns like ‘bachelor’
and ‘triangle’, along with irregular verbs like ‘brought’ (cp. ‘carried’, whose past tense is expressed with a
morpheme).
. Meanings are not extensions 
not have extensions that are their meanings. Given a domain of things called worlds,
one can model meanings with functions that map each world w onto a set of things
that exist at w, given a specified sense of ‘exist at’. The question, however, is what such
models are models of. There are many ways of representing the fact that ‘unicorn’ and
‘ghost’ have different meanings. For example, we can represent words by pairing their
pronunciations with certain functions, subject to the requirement that words with
different meanings be represented with different functions. This doesn’t tell us what
the meanings are. But if such models invariably associate the pronunciation of
‘unicorn’ with a function that maps each world w onto a set that is either empty or
full of things that are not unicorns—and likewise for ‘ghost’—this suggests that
meanings are not extensions.
Identity claims bolster this suspicion. Woodchucks are groundhogs, members of
the species Marmota monax, a.k.a. whistlepigs. Yet it seems that speakers of English
can connect the pronunciations of ‘woodchuck’ and ‘groundhog’ with different
meanings, just as they can connect the pronunciations of ‘woodchuck’ and ‘rabbit’
with different meanings. Imagine someone who acquires ‘groundhog’ in an encoun-
ter with a small whistlepig (kg, cm), and later acquires ‘woodchuck’ a thousand
miles way, in an encounter with a very large whistlepig (kg, cm). Such a speaker
might well think that woodchucks differ from groundhogs. But it doesn’t follow that
there are possible worlds at which woodchucks are not groundhogs.
A theorist can choose to represent the meanings of ‘woodchuck’ and ‘groundhog’
with functions that map at least one world w onto different sets. But my intuitions
suggest that at least one of these sets will not be the set of woodchucks/groundhogs at
w, even if the members of both sets are furry animals. Theorists can try to argue that
these Kripkean intuitions reflect some error, as opposed to correct judgments
regarding how Slang expressions can be used to describe possibilities. Though absent
reasons for thinking that the intuitions are unreliable, it seems that ‘woodchuck’ and
‘groundhog’ do not have distinct extensions. We can invent languages in which
analogs of these words connect pronunciations with distinct mappings from worlds
to sets, perhaps in ways that reflect different similarity metrics; see Lewis (,
), discussed below. But even if it would be useful to speak such a language for
certain purposes, this doesn’t show that the meanings of ‘woodchuck’ and ‘ground-
hog’ are such mappings.
This point is reinforced by an even humbler Kripke-style example. Suppose that
two dice are thrown. One lands on the left and comes up , as the other lands on
the right and comes up . Some of the people watching introduce ‘Louie’ as a name
for the die on the left; and they use ‘Ralph’ to talk about the other die. Others
introduce ‘Trey’ as a name for the die that came up three; and they use ‘Deuce’ to
talk about the other die. Eventually, there might be consensus that Louie is Deuce,
and that Ralph is Trey. Though at least initially, many of those watching the dice
may not have realized that the one that landed on the left also came up , or that
the one that came up  also landed on the right. So prima facie, ‘Louie’ and ‘Deuce’
have different meanings, as do ‘Ralph’ and ‘Trey’. But this observation about
some words doesn’t make it plausible that there is a possible world at which
Louie isn’t Deuce.
 . Overture
If the dice are fair, then there are worlds at which Louie—the die that actually
landed on the left and came up —landed on the right and came up . But those are
not worlds in which Louie is Ralph. They are worlds in which Louie is Louie, Louie is
Deuce, and Deuce came up . Likewise, I don’t think there are worlds in which the
woodchucks are not the same animals as the groundhogs. The words ‘woodchuck’
and ‘groundhog’ differ in meaning, but not because there are worlds at which the
whistlepigs are distinct from the woodchucks and/or the groundhogs.
I haven’t yet said anything about what possible worlds are. But like Kripke, I take
them to be ways the universe could be, independent of how we think about things; see
also Stalnaker (, ). Given this conception of worlds, as possible configur-
ations of the things surrounding and including us, Louie (a.k.a. Deuce) is such that
there are no worlds at which that die exists yet fails to be Deuce. Though as Kripke
notes, one need not adopt any particular conception of modality or meaning to find
the following generalization plausible: given any entity e and any entity e0 , if e is
identical to e0 , then there is no possible world at which e is distinct from e0 . Given this
generalization, which seems to be a logical truism, there are no worlds at which
Deuce (a.k.a. Louie) is a thing distinct from Louie. And if a proposal about meanings
requires false ancillary assumptions about possible worlds, then we should look for
another proposal.
I readily grant that talking about possibilities can be relevant when discussing
meanings. Kripke noted that certain hypotheses—e.g., that the meaning of a proper
noun like ‘Aristotle’ or ‘Louie’ is the meaning of some associated description—seem
less plausible if we remember that words can be used to describe scenarios that are
not actual. But appealing to non-actual possibilities is an odd way to maintain that
meanings are extensions, as opposed to mental representations of some kind. If the
meaning of a word is not whatever set of things that the word happens to be true of,
why think the meaning is a mapping from each possible world w to whatever set of
things that the word happens to be true of at w? Once we agree that Slang expressions
need not connect pronunciations to actual things, it seems contrived to insist that
these expressions connect pronunciations to possible things. I think that invoking
possible unicorns is contrivance on stilts.
In later chapters, I’ll address the idea that while meanings are not extensions,
knowing what Slang expressions mean is a matter of knowing in a special way what
the expressions are true of. For example, one can hypothesize that understanding
‘woodchuck’ is a matter of (linguistically) perceiving that this expression is true of
an entity e if and only if e is a woodchuck, and that even if a speaker of English
knows that woodchucks are groundhogs, she can understand ‘woodchuck’ without
perceiving that this lexical item is true of an entity e if and only if e is a groundhog.
Similarly, one can say that speakers of English perceive that ‘unicorn’ is true of e if
and only if e is a unicorn, but not that ‘unicorn’ is true of e if and only if e is a
ghost. I think this turns out to be an indirect way of identifying meanings with
concepts like WOODCHUCK, GROUNDHOG, and UNICORN. But others have suggested that
if we posit enough things, then meanings can be identified with extensions after all.
So let me digress, mainly for specialists, to say why I think that positing more
things doesn’t help.
. Meanings are not extensions 
. Metaphysical detour: Outlandish worlds don’t help

People are fallible. So even for those who justifiably believe that Louie was Deuce, it
will seem logically possible that Louie and Deuce are distinct. This can make it
tempting to say that there are possible worlds at which Louie is distinct from Deuce.
More generally, one might embrace the following generalization: if it is logically
possible that P, then there is a possible world at which P. This can make it tempting
to characterize meanings in terms of logical possibilities that can be described as
possible worlds, some of which are worlds at which there are unicorns and wood-
chucks that are not groundhogs. But while logical possibility can be modeled in terms
of things that can be called possible worlds, I don’t think this provides any plausible
defense of the idea that (atomic) meanings are extensions.
Consider our dice. For all I said thus far, it is logically possible that they are ten-
sided. But if Louie is six-sided, I don’t think there is a possible world at which that die
came up . Similarly, the possibility of not knowing that Louie is Deuce doesn’t imply
a world at which Louie isn’t Deuce. Ignorance is a poor basis for drawing conclusions
about how things might have been, just as it is a poor basis for drawing conclusions
about how things are.7
As Frege (b) discussed, logical possibility has something to do with how
things can be presented to rational thinkers. Prima facie, this is why the following
inference is not valid: Louie is Louie; so Louie is Deuce. The invalidity is not evidence
of possible worlds at which the premise is true and the conclusion is false, at least not
if possible worlds are ways things could be independent of how we think about them.
One can use ‘possible world’ to talk about other things—e.g., sets of thought contents
that are consistent and “maximal” in some specified sense. So one can speak of
worlds that include the thought that Louie is a disguised groundhog, and Deuce is a
disguised unicorn. But that is no reason for thinking that such worlds can be mapped
onto nonempty sets of unicorns, ghosts, or groundhogs that are not woodchucks.
Again, it’s important to distinguish theoretical models from the represented phe-
nomena. We can model logical possibility with invented languages whose sentences
include formulae, like ‘=(α, α)’ and ‘=(α, β)’, that can be assigned various interpret-
ations. For example, ‘α’ and ‘β’ might be constants, each of which can be interpreted as
standing for any element of some domain that includes Louie/Deuce and other things.
But suppose that ‘=’ must stand for the relation of identity. Then ‘=(α, α)’ is true on
every permitted interpretation, while ‘=(α, β)’ is true on some of these interpretations
and false on others. So even if ‘α’ and ‘β’ happen to be interpreted as labels for the same
thing, we can say that an inference from ‘=(α, α)’ to ‘=(α, β)’ is invalid because some
interpretations make ‘=(α, α)’ true and ‘=(α, β)’ false. But even if we use ‘possible world’
to talk about possible interpretations for the pronunciations of ‘Louie’ and ‘Deuce’, it
doesn’t follow that there are possible worlds at which Louie isn’t Deuce.
7
Limited knowledge—perhaps expressed with ‘For all I know, Louie might not be Deuce’— can fail to
rule out worlds at which Louie isn’t Deuce, but also fail to rule in such worlds. Relatedly, a priori knowledge
is a not a sure sign of necessity. One can know a priori that Louie landed on the left, even though there are
possible worlds at which Louie landed on the right.
 . Overture
Likewise, if ‘G’ and ‘W’ are predicates, we can say that an inference from ‘Gα’ to
‘Wα’ is invalid, regardless of which interpretations the predicates happen to have.
From this perspective, it doesn’t matter if ‘G’ and ‘W’ are interpreted as analogs of
‘green’ and ‘white’, ‘groundhog’ and ‘woodchuck’, or ‘prime factor of the second
perfect number’ and ‘first or second prime that is not the successor or predecessor of
the fourth positive integer’. Correlatively, the possibility of assigning various inter-
pretations to invented symbols doesn’t show that there are possible worlds at which
the set of groundhogs differs from the set of woodchucks. And we don’t need to posit
such worlds to account for why replacing ‘groundhog’ with ‘woodchuck’ is not
logically valid. Two words can be used to express distinct concepts that turn out to
be extensionally equivalent, even across possible worlds.8
Lewis () offers an alternative that is, in my opinion, spectacularly implausible.
He proposes a radical form of nominalism according to which possible worlds—ways
things could be—are universes that really exist, though not in any region of our
spacetime. According to Lewis, each of these worlds is an isolated totality of things
like us and the things around us. Indeed, each of us has endlessly many “counter-
parts,” each existing in the same way that we do. And while we are causally
disconnected from our simulacra, they determine the truth or falsity of our modal
claims. On this view, Louie could have come up six if and only there is a world at which
a counterpart of Louie came up six. The things we call actual are simply the things in
our universe: we have counterparts who use ‘actual’, with equal right, to talk about the
things in their worlds; and for Lewis, these possible things are as real as actual things.
I find it hard to take this proposal seriously. But I won’t try to rebut Lewis’s reasons for
positing his pluriverse. Here, I just want to note two points. First, if this is the best way
to defend the idea that meanings are extensions, then perhaps we should adopt a
different hypothesis about the place of meanings in nature. Second, I still don’t see why
we should think that any totality of things includes unicorns, woodchucks that are not
groundhogs, or something that is Louie but not Deuce.9
8
For related reasons, I think model-theoretic conceptions of logical possibility are overrated; see
Etchemendy (). We might use ‘G’ and ‘W’ in representing an invalid inference from a premise
about groundhogs to a conclusion about woodchucks. But if the question is why the represented inference
is invalid, it doesn’t matter that we could have used ‘G’ and ‘W’ to represent gophers and wombats. Even if
premises and conclusion are themselves (ideal?) formulae that admit various interpretations, and validity is
a matter of preserving truth relative to every admissible interpretation, the real work would involve
specifying the relevant formulae—and accounting for the range of admissible interpretations, rather
than just stipulating that the interpretations of “logical items” are fixed.
9
The actual situation—in which Louie/Deuce came up  and landed on the left, while Ralph/Trey came
up  and landed on the right—differs from the superficially similar but nonactual situation in which Louie/
Deuce came up  and landed on the right, while Ralph/Trey came up  and landed on the left. If we don’t
enforce this distinction, and count possibilities in terms of surface appearances, we can describe the
possible outcomes (ignoring position) with locutions like “one die came up , and the other die came up
.” But then instead of partitioning the possibilities into  equally likely cells, we’ll be partitioning them
into  cells,  of which are twice as likely as the other . Even if that is coherent, it’s not how we naturally
think about the space of possibility regarding fair six-sided dice. Like Stalnaker (), I think it’s better to
forego Lewis’s nominalism and regard possible worlds as modal properties of the universe: ways the totality
of things could be; where one of these ways is the way the universe actually is.
. Meanings are not extensions 
If there are Lewisian totalities that include creatures that look like we would expect
unicorns to look, those creatures are as irrelevant to our word ‘unicorn’ as any similar
inhabitants of some actual planet in a distant galaxy. The question was whether the
meaning of ‘unicorn’ determines a distinction between unicorns and mere lookalikes,
despite the absence of encountered examples. Positing more lookalikes doesn’t help.
Lewis defends appeal to totalities that include unicorns—groundhogs that are not
woodchucks, Louie but not Deuce, etc.—by combining his outlandish metaphysics
with an implausible conception of how Slang expressions are used to describe
possibilities. The idea is that given a context, ‘unicorn’ (‘groundhog’, etc.) is associ-
ated with a similarity metric that determines a mapping from each world w to a set of
things that count as unicorns (groundhogs, etc.) at w relative to that context. This
allows for variation in what makes something sufficiently unicorny (groundhoggy,
etc.) for the purposes at hand.
We can invent expressions, including ‘Lunicorn’, that work this way. Though we
can also stipulate that ‘Kunicorn’ applies only to things of the same kind as paradig-
matic unicorns, with the consequence that if there are no such paradigms, then
‘Kunicorn’ does not apply to anything in any world. One can say that ‘Lunicorn’ is a
more ideal expression, since the logical possibility of Lunicorns corresponds to
worlds at which there are Lunicorns, and that an ideal language would include ‘L’-
analogs of ‘groundhog’ and ‘woodchuck’. But one can’t also assume that meanings or
extensions are preserved by “translations” from Slangs to ideal languages.
For some purposes, it can be useful to replace instances of ‘it is logically possible
that there is at least one Φ’ with invented sentences of the form ‘◇∃x[Φ(x)]’. Yet
choices remain. If ‘◇’ is glossed as ‘there is a possible world at which’, then ‘◇∃x
[Kunicorn(x)]’ is false, while ‘◇∃x[Lunicorn(x)]’ is true and in this respect like the
thought that unicorns are logically possible. But in other respects, ‘Lunicorn’ is unlike
‘unicorn’ and the concept UNICORN, which seems more like ‘Kunicorn’. There is a
tension between using ‘◇’ to reflect logical possibility, of thoughts we actually
express, and glossing ‘◇’ in terms of possible worlds. Unsurprisingly, no translation
into an alien language will be perfect for all purposes. And the mere logical possibility
of unicorns doesn’t ensure that some things, at home or abroad, are unicorns.
. Polysemy redux

We were, recall, considering an obvious objection to the idea that meanings are
extensions—viz., that expressions with different meanings can have the same exten-
sion. Considering replies can lead to complicated discussions about possible worlds
and how they can be correctly described. But in my view, the net result is that if Slang
expressions have extensions, these extensions are not the meanings that the expres-
sions connect with pronunciations. Moreover, the antecedent of this conditional can
be challenged. And if some expressions have meanings but not extensions—unlike
ideal concepts, which have extensions but not meanings—that’s another reason
for denying that meanings are extensions. So the phenomenon of polysemy is
again relevant.
One can insist that either (i) ‘book’ is a lexical item whose extension includes many
bodies of information as well as the corresponding vehicles, or (ii) the pronunciation
 . Overture
of ‘book’ is homophonous as between two or more words, each with its own
extension. But I think both options are less plausible than saying that ‘book’ can be
used to access more than one concept. And as discussed in section two, the apparent
examples of polysemy are diverse: ‘circle’, ‘line’, ‘window’, ‘country’, ‘France’, several
words with the pronunciation of ‘bear’, etc. So even if some cases are plausibly
diagnosed in terms of homophony or inclusive extensions, this doesn’t motivate
diagnosing every case in one (or both) of these ways. Absent plausible specifications
of the alleged extensions, it seems that meanings tend to be conceptually equivocal,
with the result that Slang expressions are flexible devices that do not have extensions.
In my view, there is nothing special about the meanings of ‘unicorn’ and ‘ghost’.
These words are like ‘book’ and ‘green’ in not having extensions. The concepts BOOK:
VEHICLE and BOOK:INFO apply to actual things, while UNICORN and GHOST do not. But
customers who look for the unicorns in a toy store might find them next to the
dragons; and the guests at a costume party can include some ghosts, unicorns, and
dragons. We have concepts like PRETEND-GHOST that apply to actual things. So my
claim is not that ‘ghost’ and ‘unicorn’ have empty extensions. I think these words
illustrate the general point that words don’t have extensions. If we ignore polysemy,
the relevant point is that there are no ghosts or unicorns. If we don’t ignore polysemy,
the case against identifying meanings with extensions is even stronger.10
In §., I noted that lexical meanings are flexible in ways that go beyond the usual
examples of polysemy. Even if the root noun ‘rabbit–CT’ has an extension that
includes the extensions of both the mass-concept RABBITM and the count-concept
RABBITC, it isn’t clear what the extension of ‘green’ would be, given that the adjective
can combine with nouns like ‘paint’, ‘bottle’, ‘apples’, and ‘fish–CT/fish+CT/fish+PL’.
A related concern, discussed in the chapters below, is that the meaning of a word is
often context sensitive in subtle respects.
Again, the ways in which paint can be green seem different than the ways in which
bottles or apples can be green, as if ‘green’ can be used to access concepts that differ in
content. Moreover, these concepts seem to differ with regard to which aspects of
contexts matter with regard to whether or not the stuff or things count as green. One
can still speculate that ‘green’ has an extension—perhaps a set of ordered pairs <k, s>
such that k is a context, in some technical sense, and s is a set of things that are green
relative to k. But absent a plausible specification of this alleged extension, why think
the meaning of ‘green’ determines any such set? A meaning may provide access to
certain concepts, each of which may be context sensitive in its own way.
10
We can grant that some theorists sometimes use ‘unicorn’ to express the technical concept LUNICORN.
But if ‘unicorn’ can also be used to express the concept KUNICORN, then it seems like contrivance to insist
that the Slang expression has a meaning that maps some contexts onto the extension of LUNICORN and other
contexts onto the extension of KUNICORN. If we assume that words like ‘possibly’ have extensions, then
perhaps we should specify the meanings of such words in terms of a suitably generic notion of world that
allows for special cases corresponding to metaphysical and epistemic modalities; cp. Kratzer (, ).
But in my view, theorists should not posit (things that include) unicorns in order to accommodate
correct uses of ‘Possibly/Perhaps/Maybe unicorns exist’ or ‘There may be unicorns’; and likewise for
squarable circles.
. Meanings are not extensions 
In this respect, the interesting and much discussed word ‘I’ can be a distraction.
Each speaker of a Slang may be a thinker who has concepts of two special types, EGO
and SPEAKER; where the former can be used to think about oneself in a first-personal
way, and the latter can be used to think about the speaker in a given context as such.
Neither of these concepts is the meaning of ‘I’. It is often inappropriate to access an
instance of EGO upon hearing someone else use ‘I’; and prima facie, one can use ‘I’
without thinking of oneself as the speaker. So especially after Kaplan (b, ), it
can be tempting to say that the meaning of ‘I’ is a function that simply maps each
communicative context onto its speaker. One might then conclude that the meaning
of ‘green’ is also the extension of some context-sensitive concept. But we shouldn’t
assume that ‘I’ provides a good model for all species of context sensitivity.
The meaning of ‘I’ may be a relatively permissive instruction that can be executed
by accessing any concept of the relevant user. In contexts where I use ‘I’, such an
instruction might be executable by directly accessing an instance of EGO. In contexts
where an unknown speaker uses ‘I’, the same instruction might be executable by
accessing an instance of SPEAKER. In contexts where my wife Susan uses ‘I’, the same
instruction might be executable by accessing a concept with which I think about her
as that very person (viz., my wife Susan), perhaps via some recognition that she is
speaking. This would be yet another kind of conceptual equivocality, even if the
meaning of ‘I’ is partly specified in terms of a concept like SPEAKER. This isn’t yet an
argument that the meaning of ‘I’ is an instruction as opposed to an extension. But
reflection on ‘I’ hardly shows that all conceptual equivocality can be plausibly
diagnosed in terms of the hypothesis that meanings are extensions.
More generally, given that a single meaning can correspond to several concepts, it
can be tempting to think that meanings are somehow less removed from the mind-
independent things that we think and talk about. But we shouldn’t conclude that
meanings are extensions of concepts. There is another diagnosis of conceptual
equivocality. In my view, meanings are doubly removed from the external world:
we use meanings to access and assemble concepts that let us think about things in
certain ways; and a lexical meaning can connect its pronunciation with a family of
polysemously related concepts. Putnam () suggested that meanings provide a
“coarse grid over use.” I like the grid metaphor. Though it might be better to say
that meanings provide coarse grids over human concepts, and that uses of Slang
expressions depend on these expressions having meanings that can be used to build
concepts that can be used in many ways.
. Communication without extensional equivalence

I deny that Slang expressions have extensions. Nonetheless, these expressions are
often used in episodes of successful communication. So like Strawson () and
many others, I deny that successful communication requires expressions that have
extensions. In §., I’ll address Putnam’s () influential and apparently contrary
claims about the meaning of ‘water’. But first, let me stress that communicative
success comes in degrees.
If you don’t know which meaning a speaker of English is expressing with the
pronunciation of ‘pen’, you might not know whether she is talking about writing
 . Overture
implements or animal enclosures. And speakers who use this pronunciation to

express different meanings, in a single conversation, might mistakenly think that
they disagreed (or agreed) about the cost of something. But speakers can also use one
of the meanings to formulate a point of substantive disagreement (or agreement)
regarding a certain writing implement, as if that meaning connects its pronunciation
with certain objects that are not animal enclosures. Similarly, the meaning of ‘star’ can
be used to formulate tendentious hypotheses about the nature of certain celestial points
of light—as if that meaning somehow embraces the stars, whatever they turn out to be.
Such examples can make it seem that meanings at least determine extensions.11
On the other hand, speakers can connect the pronunciation of ‘pen’ to the same
meaning without determining a set of pens that they are talking about. If each
speaker uses a polysemous word that lets her access a concept of writing implements,
as opposed to animal enclosures, they resolve the homophony the same way and
thereby avoid blatant miscommunication. In this sense, they connect a common
pronunciation with a common meaning. But it doesn’t follow that any speaker used a
word that has an extension, much less a word that can be modified by ‘green’ to form
a phrase that has a certain set of green writing implements as its extension.
As Travis (, ) discusses, speakers can agree about what ‘green pen’ means,
yet not be in sync with regard to what they would count as a green pen. Likewise, they
might have different standards for what counts as expensive, or at least expensive for
a green pen. So even if speakers connect the pronunciation of ‘the green pen was
expensive’ to the same sentential meaning, they need not connect this pronunciation
with a meaning that determines an extension—e.g., a truth value, or a set of situations
in which the sentence would be true—much less an extension that could be deter-
mined by an impartial referee who knew the relevant facts.
Perhaps variation in standards is another kind of conceptual equivocation. But in
any case, speakers can agree about what expressions mean, yet fall short of a certain
ideal for communication. For some purposes, it might be nice to describe an entity
and a set of green pens—resolving any uncertainty about which entity and which set
we are talking about—and then have a debate about whether or not the entity is an
element of the set. For many other purposes, this would be horrifically tedious.
Though whatever the merits of creating situations in which linguistic expressions
are used in ways that approximate extensional equivalence, creating such situations
takes work. To achieve this kind of communicative precision, conversational partners
need to do more than just agree about which pronunciations go with which meanings.
Given that humans can use a single pronunciation to access different concepts,
there is always some risk of speakers “talking past each other.” Resolving homo-
phonies helps. But one person might use ‘book’ in saying that the content of a certain
tome is interesting, while someone else uses ‘book’ to say (in a disagreeing way) that
the tome is boring. Even if such disputes are not substantive, they need not be verbal
11
This leaves room for views like Putnam’s, according to which at least some meanings are multi-
dimensional entities that comprise extensions and other things, like pointers to paradigm cases. Or perhaps
competent speakers are expected to know, or at least assume that many other speakers will assume, certain
bits of “common sense;” see, e.g., Higginbotham ().
. Meanings are not extensions 
as opposed to conceptual. Meanings provide a framework for a kind of communi-

cation that is distinctively human and enormously valuable. But meanings are not
prophylactics against miscommunication. Even in the absence of any special compli-
cations, connecting pronunciations with the same extensions seems like a quixotic task,
or at least much harder than connecting pronunciations with the same meanings.
However, one might think that substantive disagreements about the things or stuff
that we can talk about—e.g., disagreements about the nature of rabbits, stars, or
water—tell in favor of the idea that meanings determine extensions. A concept of
rabbits can be used to form, and perhaps endorse, eccentric thoughts that can be
expressed with the word ‘rabbit’. We can entertain the thought that rabbits are robots
from Mars; that they are ideas in the mind of God; that stars are luminous rabbits, set
in the heavens to honor their deeds; or that rabbits are stars sent down from the
heavens to scare horses. Similarly, people who know what ‘star’ means can have
different views about what things like Polaris and Sirius are: holes in the canopy; fires
attached to the celestial sphere; effects of nuclear fusion; etc. We describe the sun as a
star, and not as a planet, though for many humans, this description would have
seemed bizarre, since the sun is among the “wanderers” whose position in the sky is
not fixed relative to the background stars. But if speakers can use ‘star’ to advance
very different hypotheses about the same things, this does suggest that speakers can
connect ‘star’ with distinct concepts that are extensionally equivalent.
If a child can acquire an instance of STAR, then an educated speaker may have
acquired an instance of a more scientific type, STAR*, and learned that every star is a
star*. (Perhaps this is how one learns what stars are.) It doesn’t follow that every
star* is a star. But even if STAR has the same extension as STAR*, and a child acquires
‘star’ by connecting its pronunciation with an instance of STAR, the acquired word
might become polysemous—as the child acquires a concept of the gold stars that
indicate praise for homework, a concept of the stars awarded by a critic to a star-
studded movie, and (eventually) a scientific concept of things like Polaris and Sirius.
Moreover, if we focus on the shared extension of STAR/STAR* and ignore the other
concepts, then I think ‘star’ is an unusual case from which we should not generalize.
It’s worth being explicit about this, since there is an important point about concepts in
the vicinity, with implications for the study of Slangs and meanings. To advertise: we
can use ‘star’ to express a concept that applies to celestial objects of a certain kind,
which competent speakers (theorists included) may misdescribe; but we can also use
‘meaning’ to express a concept that applies to interpretations of a certain kind, which
competent speakers (theorists included) may misdescribe.
We can discover that the sun is a star as opposed to a planet, and that meanings
are polysemous instructions for how to access and assemble concepts, as opposed to
concepts or extensions.
. Meanings, natural kinds, and hard water

When a child perceives things like Polaris and Sirius, and hears them described as
stars, she may acquire a count-concept that is also a kind-concept in the following
sense: the concept applies to (and only to) things that are similar, along some
dimension, to the initial examples; the relevant dimension of similarity is determined
 . Overture
by the nature of the paradigms, as opposed to any associated appearances or

descriptions; and so the acquired concept applies to the sun, along with the light
sources responsible for the Milky Way, but not to quasars. The familiar idea is that a
kind-concept of stars is a concept with which we think about stars as things that share
a common nature, regardless of how much one knows about stars. We can discover
this nature empirically. But substantive conceptions of stars can also be wildly
incorrect. And it’s easy to misapply a kind-concept of stars, since some non-stars
seem more starry than many visible stars.
Spelling out the details is hard, though not my job. My claim is that even if words
like ‘star’ are initially paired with kind-concepts that have extensions, meanings don’t
determine extensions. Again, kind-concepts do not preclude polysemy. But even
ignoring gold stars and movie stars, a scientific concept that an adult can express with
‘star’ may have a somewhat different extension than a kind-concept that a child
naturally acquires. As an analogy, consider the various concepts that have been
proposed as scientific correlates of the ordinary word ‘planet’; some but not all of
these concepts exclude Pluto.12 Still, let’s suppose/pretend that there is an ordinary
concept STAR:KIND whose extension is shared by any concept that can be accessed with
‘star’. Then ‘star’ is not a counterexample to the claim that meanings at least
determine extensions. But the question is whether the general claim is plausible.
And in this context, it is impossible to avoid discussion of ‘water’, given Putnam’s
() thought experiment involving Twin Earth.
My own view is that Putnam misapplied the moral of his example. I think ‘water’ is
a complicated case that tells against the idea that words have extensions; see Pietroski
(a). I won’t rehearse the full argument here. But since Putman’s claims have been
influential, I do want to stress—following Chomsky (b, b)—that we use
‘water’ in many ways, and that focusing on natural kind uses does not make it
plausible that the word has an extension.
Putnam asked us to imagine a planet like Earth except that in place of HO, Twin
Earth has a chemically distinct substance, XYZ, that is superficially similar and
equally potable. He then invited us to share his judgment that the watery stuff on
Twin Earth is not water, infer that the extension of ‘water’ is limited to samples of
HO, and then generalize to broader claims—e.g., that meanings “ain’t in the head”
no matter how you “cut the pie.” But let’s slow down. I grant that ‘water’ can be used
to access a kind-concept that applies to, and only to, samples of HO; where some
stuff can count as HO, despite trace impurities. In which case, ‘water’ can be used to
access a concept—an instance of the type WATER:KIND—that does not apply to samples
of XYZ (a.k.a. twin-water). Let’s also grant that in this sense, many conceptual
contents are individuated externalistically; cp. Burge (, ). Nothing yet
follows about the individuation of word meanings, if only because a word can be
used to access more than one concept.
12
Infants may assume that many things have essences; see, e.g., Keil (). But a child-acquirable
concept of woodchucks may not have the same extension as the corresponding biological concept
MARMOTA-MONAX.
. Meanings are not extensions 
We can speak of polluted water, salty water, hard water, etc. In some places, water
from taps can lead to health threats that are not due to trace impurities. This already
suggests that ‘water’ and WATER:KIND are relevantly different. A little investigation
bolsters the point. I happen to have a well in an area where the ground water has a
high mineral content. Experts at “National Testing Laboratories, Ltd.” analyzed a
sample of the stuff that comes from my well, and according to their “Quality Water
Analysis,” that stuff is about . percent HO. These experts reported that the water
from my well is very hard. (Six parts per thousand is a lot.) But ocean water of
average salinity is only about . percent HO. The water contents of some other
familiar substances are also listed below, data from the United States Department of
Agriculture; where for these purposes, water does indeed contain only trace impur-
ities beyond HO.13
Club Soda .
Diet soda, not cola .
Tea .
Diet Cola .
stuff from my well .
Coffee .
Espresso .
ocean water, avg. salinity .
Michelob Ultra .
Bud Light .
Distilled vinegar .
Diet Coke has a higher percentage of HO than my well water, which has the
granularity of coffee. So if the stuff from my well counts as HO, why doesn’t Diet
Coke count as water? Tea is even closer to pure HO. An espresso has a higher
percentage of HO than typical samples of ocean water. Distilled vinegar and Bud
Light are much closer to ocean water—by an HO modulo impurities test—than
ocean water is to HO. Perhaps some impurities are especially polluting; though see
Malt () for a study of actual judgments by ordinary speakers. And why ignore
the fluoride in tap water, yet stress the neither-hydrogen-nor-oxygen components of
clear diet soda?
I note these facts in part because of Chomsky’s (b) illuminating example.
Suppose cup- is filled from the tap. It is a cup of water, but if a tea bag is dipped into it, that is
no longer the case. It is now a cup of tea, something different. Suppose cup- is filled from a tap
connected to a reservoir in which tea has been dumped (say, as a new kind of purifier). What is
in cup- is water, not tea, even if a chemist could not distinguish it from the present contents of
cup-. . . . In cup-, the tea is an “impurity” in Putnam’s sense, in cup-, it is not, and we do not
have water at all (except in the sense that milk is mostly water, or a person for that matter). If
cup- contains pure HO into which a tea bag has been dipped, it is tea, not water, though it
13
See https://ndb.nal.usda.gov/ndb/nutrients.
 . Overture
could have a higher concentration of HO molecules than what comes from the tap or is drawn
from a river (pp. –).
As Chomsky observed, there are ordinary uses of ‘water’ such that what counts as
water in this human sense is a complicated matter having to do with sources and
intended purposes. We can also use ‘water’ to talk about the stuff itself, and the
common nature—given that there is one—exhibited by samples of pure water.
Borrowing an Aristotelian distinction, Chomsky notes that we often think about
aspects of reality in terms of function as well as form. Putting the point in terms of
conceptual contents, we might say that instances of WATER:FUNCTION let us think about
some stuff as water in a familiar human sense—viz., a sense in which the stuff in
Chomsky’s cup- and the stuff from my well is water, while the stuff in his cup- and
Diet Coke is not. By contrast, instances of WATER:KIND let us think about some stuff as
water in another familiar sense that Putnam highlighted.
This provides a simple diagnosis for why the following argument is obviously
invalid: water is HO; the water from my well has a high mineral content; so the HO
from my well has a high mineral content. In my view, the following argument is also
invalid: water is HO; so ‘water’ is true of an entity e if and only if e is a sample of
HO. The premise is about water, while the conclusion is about a word. If the word is
polysemous, then the premise is true only if the polysemy is resolved in favor of
scientific usage. So even if the conclusion is true when restricted to scientific usage, it
is a fallacy to drop the restriction and conclude that however the polysemy is
resolved, ‘water’ is true of all and only samples of HO.14
In short, I think ‘water’ is yet another example of a conceptually equivocal word
whose meaning does not determine an extension. And if ‘water’ can be used to access
an instance of WATER:FUNCTION, I see no reason to insist that the watery stuff on Twin
Earth cannot be described as water. For what it’s worth, my own intuition is that
people on Twin Earth can quite literally water their lawns and take a drink of water
from a garden hose. But context matters. In situations where chemistry is important,
uses of ‘water’ are more likely to access instances of WATER:KIND; and Putnam’s
thought experiment explicitly makes chemical composition relevant. Chomsky’s
thought experiment reminds us that ‘water’ has other uses. I also see no reason to
insist that when children acquire ‘water’, they connect this word with an instance of
WATER:KIND as opposed to an instance of WATER:FUNCTION. So I don’t think ‘water’
14
Putnam initially says that whatever meanings are, they are not “in the head” and also things that
determine extensions. (Chomsky agrees, as do I.) But Putnam then slides into assuming that meanings
determine extensions. He concedes that talk of words having extensions is a “very severe” idealization
(p. ). Yet he goes on to say that two familiar “assumptions” are “not jointly satisfied by any notion, let
alone any notion of meaning” (pp.–): (i) “knowing the meaning of a term is just a matter of being in a
psychological state;” and (ii) the meaning of a term determines its extension. Putnam says that (i) is to be
understood in terms of methodological solipsism: no psychological state “presupposes the existence of any
individual other than the subject to whom that state is described.” He then takes (ii) as a premise to argue
against (i). Yet he concludes (p. ), “The traditional problem of meaning splits into two problems. The
first is to account for the determination of extension. Since, in many cases, extension is determined socially
and not individually, owing to the division of linguistic labor, I believe this problem is properly a problem
for socio-linguistics.” But why think ‘water’ has an extension, much less one that is determined socially?
. Meanings are not extensions 
lends support to the idea that ‘star’ provides a paradigmatic example of how word
meanings are related to kind-concepts.
In the big scheme of things, stars are remarkably uniform. Animals of the same
species are also quite similar. Though even with regard to ‘dog’, it is far from obvious
that the biological facts determine a set of animals that corresponds to correct
application of the word. It’s hard to distinguish dogs, wolves, and coyotes in terms
of their biology. But I think we have a concept of dogs that excludes coyotes yet
allows for tame wolf-crosses. We may also use ‘dog’ to access a kind-concept whose
extension is determined by the relevant biological facts, whatever they are, making it
reasonable to defer to biologists about which animals this concept applies to. Though
if we have such a concept, then prima facie, the meaning of ‘dog’ is conceptually
equivocal in a way that tells against the idea that ‘dog’ has an extension. And if we
remember that things may be less uniform than some philosophers would like,
further thought experiments come to mind.
Suppose that on Fraternal-Earth, dopplegangers of our scientists discover—to
their great surprise—that all the stuff they call ‘mud’ has the same chemical structure.
Those scientists could acquire a corresponding kind-concept and use their word
‘mud’ to express it. Other speakers might defer to these experts. There is, in effect, a
Platonic form of Frat-mud, all of which would count as Earth-mud. But why think
the Frat-Earth scientists, who grew up as our children do, couldn’t use their word
‘mud’ to talk about our diverse samples of mud? Why think their word has a
restrictive meaning just because their mud is especially uniform? Prima facie, the
Frat-Earth scientists supplement their ordinary concepts with a new kind-concept.
Similarly, in my view, we can have a kind-concept of water that is not the only
concept we can access with ‘water’.15 I have dwelled on this point for two reasons:
First, Putnam () has been so influential that it can be hard to a get a hearing for
internalist conceptions of meaning; cp. McGilvray (), Stainton (). Second,
while Putnam rightly highlighted the importance of “natural kind uses” of words,
he ignored the possibility that ‘meaning’ can be used to access a kind-concept of
the interpretations—whatever they are—that Slangs connect with pronunciations.
We have to figure out what these meanings are and how they related to other things
like concepts, extensions, and truth. And theorists must be prepared to discover
that Slangs connect pronunciations with interpretations that do not determine
extensions.
15
Kripke’s () points apply, in my view, to certain concepts rather than nouns; cp. Evans ().
This allows for the logical possibility that stars are, like pieces of jade, varied examples of more than one
kind of stuff. Likewise, it is logically possible that most rabbit is textured soybean, and that some rabbit is
made from cotton; cp. note  above. Nebulae actually turned out to be things of diverse sorts. But after the
discovery of galaxies, astronomers refined at least one concept of nebulae in a way that excludes things like
the so-called Great Nebula in Andromeda. Prima facie, the meanings of ‘star’ and ‘planet’ do not differ in
kind. But given the history of how ‘planet’ has been used, and how Pluto has been classified, it seems
plausible that ‘planet’ can be connected with distinct conceptual contents (even for a speaker who knows
the relevant facts). And prior to astronomers developing the current technical concept of a planet, it’s not
obvious that they were using extensionally equivalent concepts.
 . Overture
Speakers of English are free to use ‘meaning’ as a word for talking about things
that determine extensions. But like most words, ‘meaning’ is polysemous. So in the
spirit of Putnam’s observation that theoretical disagreements can run deep, we
should ask if there is a natural phenomenon that we can think about with a kind-
concept of meaning—i.e., a concept that lets us think about some things that we call
meanings, and think about them in a theoretically neutral way that lets us offer
hypotheses about what these meanings are. And there is indeed a natural phenom-
enon of humans acquiring distinctive languages that connect meanings with pro-
nunciations in specific ways. In the next chapter, I’ll discuss some relevant facts that
can help us zone in on the meanings in question. But while the details cannot be
known in advance, we should have suspected all along that inquiry is required to
reveal the nature of stars, water, and meanings.
. Meanings as modest Begriffsplans

In thinking about what meanings are, one important question is whether meanings
compose. I think they do. If two Slang expressions are combined to form a third, then
in my view, the meaning of the third expression has parts that include the meanings
of the first two. As discussed in chapter one, this is controversial, in part because
many theorists take meanings to be more like sets than concepts. The set {x: x is a red
dot} does not have parts that include {x: x is a dot} or {x: x is red}. But a typical
concept of red dots is presumably complex, perhaps of the type +(RED, DOT); where ‘+’
signifies some concept that can combine with two others, of the right sort, and
thereby form a conjunctive concept that has three constituents. I think the meaning
of ‘red dot’ is likewise an instruction that includes the meanings of ‘red’ and ‘dot’.
If phrasal meanings have lexical meanings as constituents, then the relevant
combinatorial operations are unlike set-theoretic intersection, and more like oper-
ations of concatenating symbols or connecting pieces of a jigsaw puzzle; see, e.g.,
Frege (a). This highlights a series of questions about the typology of meanings
and any corresponding concepts: how many types of meaning are there; what are the
atomic and non-atomic types; what kinds of concepts can be accessed via atomic
meanings and built via non-atomic meanings? In later chapters, I argue for a spare
typology according to which all non-atomic meanings are instructions for how to
assemble concepts that are conjunctive and monadic—i.e., predicative as opposed to
relational—while atomic meanings are instructions for how to access concepts that
are monadic or dyadic. In this section, I advertise some of the details, mainly for
specialists.
. Instructions that compose

An instruction for how to build something can be complex in ways that mirror the
complexity of things built by executing the instruction. Many recipes—e.g., for how
to make dough—do not have this property. But imagine a series of numbered boxes
and the following instruction: get something from box ; get something from box ;
link them both to a connector from box . Carrying out this tripartite instruction will
yield a tripartite object. Though depending on the contents of the boxes, following
. Meanings as modest Begriffsplans 
the instruction twice might yield very different products; and depending on the
available connectors, choices from boxes  and  may have to be made judiciously
in order link them. Indeed, there is no guarantee that the instruction can be executed,
given the materials at hand. But the instruction exhibits a kind of structure that is
reflected in any objects that are created in the specified way. I think meanings are
instructions of this sort.
Let ‘μ(Σ)’ stand for the meaning of expression Σ. To a first approximation, I think
that μ(‘red dot’) is the tripartite instruction Join[μ(‘red’), μ(‘dot’)]; where any instance
of the general form dJoin[μ, μ]e is an instruction for how to build a concept by
conjoining the results of executing the subinstructions, μ and μ. On this view,
executing μ(‘red dot’) can be a way of building an instance of +(RED, DOT) and thereby
building a concept of red dots.
The idea that expressions can pair signals with directions for use, or instructions
for how to build representations of some kind, is far from new; see, e.g., Strawson
(), Hobbs (), Espinal (), Cann et al. (). Programming languages
provide obvious examples. As discussed in chapter three, I draw more heavily on
Frege (, , ). But since I don’t borrow the aspects of his work that
semanticists have tended to adopt, a preview may be useful.
Frege held that each sentence of his Begriffsschrift, an invented “Concept Script,”
mirrors the logical structure of the thought that the sentence expresses. This thought,
a potential premise/conclusion, is said to be the sense of the sentence. The sense of a
sentence is composed of senses expressed by constituents of the sentence. Each sense
presents a corresponding denotation, which may be a truth value, or an entity in
some specified domain, or a function.16 Frege argued that thoughts exhibit a kind of
function-argument structure, as opposed to classical subject-predicate structure. But
typically, the value of a function given an argument isn’t composed of the function
and the argument. So while senses compose, denotations don’t.
Frege also maintained that each denotation is presented by many senses, making it
possible to formulate informative identity claims like ‘=(h, p)’; where ‘h’ and ‘p’ are
symbols that present a certain celestial object as, respectively, the Evening Star and
the Morning Star (a.k.a. Hesperus and Phosphorus). In ideal cases, each user of the
Begriffsschrift links each atomic expression to the same sense, thereby associating
each sentence with the same thought. In practice, however, different users may think
about the same denotation in different ways.
Two people might both use ‘v’ as a label for Venus, but link ‘v’ to different senses,
perhaps because they use different concepts to think about Venus. Or they might
represent the relation of identity in different ways. So they might agree that ‘=(h, v)’ is
true, and that it implies ‘Planet(h) Planet(v)’, yet not associate ‘=(h, v)’ with the
same thought. One can insist that such people are using different languages, or that at
least one of them is using a shared language incorrectly. But thinkers could use
Fregean expressions as instructions for how to build mental representations of
16
The words ‘thought’ and ‘sense’ are polysemous. Fregean thoughts (Gedanken) are idealized pro-
positions, not mental representations; and he took senses (Sinnen) to be publicly available ways of
presenting denotable things (Bedeutungen).
 . Overture
denotations, while allowing for some variation in the representations that get built,
given agreement about which Begriffsschrift sentences are true and which arguments
are valid. And one can imagine further departures from the Fregean ideal.
In particular, a language might allow for atomic expressions that are denotation-
ally equivocal in certain limited ways, with the result that two uses of a complex
expression (to construct a mental representation) may not present the same denota-
tion. A sentence of such a language might be used twice to construct two thoughts
that do not have the same truth value. But complex expressions of the language
could still be viewed as “Begriffsplans” that provide instructions for how to build
mental representations from a stock of atomic elements. For some purposes, a
pseudo-Fregean language of this sort might provide a useful model for Slangs,
whose expressions are often polysemous. But in my view, there is a deeper respect
in which Begriffsschrift expressions reflect ideal thoughts as opposed to Slang
expressions.
As discussed in chapter three, complex expressions of Frege’s invented language
can be generated by combining simpler expressions, or by abstracting a constituent
from a complex expression. Moreover the relevant form of abstraction is permissive
in a way that makes it possible to generate expressions of boundlessly many types. At
the risk of overburdening an introduction, I want to briefly discuss this somewhat
technical but important point, which will figure prominently in my argument that
composition of meanings is restricted far more severely.
Given the Fregean sentence ‘Planet(v)’, abstraction on ‘v’ yields ‘Planet(_)’, with
the blank indicating a gap that corresponds to expressions of the same type as ‘v’.
Given the same sentence, abstraction on ‘Planet’ yields ‘__(v)’, with the longer blank
indicating a gap that corresponds to expressions of the same type as ‘Planet’. So if
‘Planet(v)’ is an instance of the truth-evaluable type <t>, and ‘v’ is an instance of the
entity-denoting type <e>, then ‘Planet’ can be described as an instance of the abstract
type <e, t>. In terms of Frege’s favored metaphor, saturating ‘Planet(_)’ with an
entity-denoter yields a truth-evaluable sentence. Instances of type <e, t> correspond
to functions from entities to truth values—e.g., the smallest function that maps each
entity in a given domain to truth or falsity, depending on whether or not that entity is
a planet. So the abstracted expression ‘__(v)’ can be described as an instance of a
“higher” type <<e, t>, t> that corresponds to functions from <e, t>-functions to truth
values. Instances of such types can also be introduced. For example, we can stipulate
that ‘∃[__]’ indicates a function that maps each <e, t>-function to truth or falsity
depending on whether or not that <e, t>-function maps something to truth. Then the
Fregean sentence ‘∃[Planet(_)]’, in which ‘Planet(_)’ saturates ‘∃[__]’, is true if and
only if something is a planet.
More generally, given expressions of any types <α> and <β>, Frege’s Begriffsschrift
allows for expressions of the type <α, β>. So given expressions of types <e> and <t>,
there are expressions of boundlessly many types. Indeed, as we’ll see, expressions that
exhibit millions of types can be easily generated with just a few rounds of recursion.
Contemporary semanticists often describe Slangs as Fregean in this respect. By
contrast, I follow Frege in thinking that he needed to invent a language whose
expressions exhibit this kind of interpretive typology.
. Meanings as modest Begriffsplans 
This point is interwoven with Frege’s work on the logical foundations of arith-
metic. Given classical conceptions of grammar and logic, as developed by Aristotle
and the medieval logicians, proofs by arithmetic induction seem different in kind
from valid syllogisms. As Kant discussed, the a priori character of classical logic was
rooted in the “analytic” idea of a predicate “containing” a subject, while the a priori
character of geometry and arithmetic seemed to have some other “synthetic” basis;
see Friedman () for helpful discussion. Frege responded by offering a broader
conception of logic along with an invented language that let him represent proofs by
induction as instances of valid inference forms, while also representing the “axioms”
of arithmetic as reducible claims that are not logically independent.17
Frege assumed that sentences of languages like German (English, Greek, etc.)
exhibit subject-predicate structure, as opposed to the function-argument structure of
ideal thoughts. Though he also held that we somehow use ordinary sentences to
indicate thoughts. The idea was that a thought content can be “dimly grasped,” in
some natural way, and then re-presented in a more logically perspicuous format that
highlights inferential relations to other contents—many of which we might entertain
only via the new formal notation. But this raises further questions. How can we use
ordinary sentences to even gesture at thoughts that are mirrored by Begriffsschrift
sentences? And if Slang expressions have their own meanings, how are the meanings
of complex expressions related to the meanings of words? In particular, if phrases like
‘chased every rabbit’ and ‘precedes every positive integer’ are predicates that can
combine with subjects to form sentences, then one wants to know how the meanings
of such phrases are compositionally determined. As discussed in later chapters,
the answer may require a conception of grammar according to which Slangs are
not sources of subject-predicate structure; cp. Chomsky (). But that raises the
question of why appeals to such structure seem natural.
I think we can make progress by viewing meanings as “modest Begriffsplans” that
diverge from the Fregean ideal in two respects. First, a Slang expression is an
instruction for how to build a concept, but not any particular concept (or even a
concept that has a certain extension). Second, meanings exhibit very few types—
perhaps just two—with the most important type being predicative. The idea will be
that predicates play a central role in the systems of judgment that interface with
Slangs, which let humans employ evolutionarily ancient cognitive capacities in new
ways. I assume that many animals can form Subject-Predicate thoughts, which can be
viewed as sentences generated by internalized procedures that are often called mental
languages. But these old procedures may provide only limited ways of building
17
Frege’s language was designed to represent higher-order relations, like the “transitive closure”
relation exhibited by the relations indicated with ‘is an ancestor of ’ and ‘is a parent of ’. Even given relative
clauses like ‘who begat Seth’ and ‘who Adam begat’, English does not let us abstract on ‘begat’ to form a
clause (e.g., ‘which Adam Seth’) that would apply to each relation that Adam bears to Seth. Similarly, even if
‘ancestor’ and ‘parent’ are analogs of ‘AncestorOf(_, _)’ and ‘ParentOf(_, _)’—expressions of type <e, <e, t>>—
we cannot introduce a verb of type <<e, <e, t>>, <<e, <e, t>>, t> and use ‘Ancestor transits parent’ to mean
‘TransitiveClosureOf[AncestorOf(_, _), ParentOf(_, _)]’. I’ll say that again, more slowly, in chapter three.
 . Overture
complex predicates, while Slangs let humans build boundlessly many predicates that
are systematically combinable.
As we’ll see, this proposal draws heavily on Frege’s idea that a language can be
used to re-present conceptual contents that were initially presented in a different
format. Frege did not invent his Begriffsschrift merely as a tool for connecting certain
interpretations with public signifiers. He used the invented notation to introduce
symbols that exhibited a distinctive representational format that was useful for
certain computational purposes; see Horty (). But while Frege wanted logically
perspicuous representations, and so invented a language designed to be unlike Slangs
in important respects, I’ll argue that Slangs let us build mental representations whose
format is logically boring though useful in other respects.
. A monadic basis

In my view, Aristotle and the medieval logicians were right to think that predicates
play a central role in thought, its linguistic expression, and our natural sense of valid
inference. I’ll say more about the relevant notion of predication in the next chapter.
But intuitively, a predicate classifies things, into those that meet a certain condition
(e.g., being a rabbit) and those that do not. Anything that meets the condition
satisfies the predicate, which applies to anything that meets the condition. We can
invent perceptible predicates. Though for now, let’s focus on predicative concepts,
like instances of RABBIT. I assume that many animals have such mental predicates.
Given some things to think about, a predicate may apply to each of those things, or
to nothing. But these are just special cases of classifying. Likewise, even if a predicate
applies to exactly one thing, it differs from any denoting (or “singular”) concept with
which one can think about that thing as a particular individual. A predicate also
differs from any concept, like an instance of ABOVE, with which one can think about
some things as exhibiting a certain relation. But there are at least two ways to blur the
distinction between predicates and relational concepts.
One can say that all such concepts classify: monadic predicates classify individual
things; dyadic/triadic/n-adic predicates classify ordered pairs/triples/n-tuples of
things. Alternatively, one might describe a monadic predicate M as expressing a
certain dyadic relation that entities can bear to truth values—i.e., a relation that each
entity e bears to truth or falsity depending on whether or not M applies to e. Likewise,
one might describe a dyadic predicate D as expressing a certain triadic relation that
pairs of things can bear to truth values, etc. In Frege’s Begriffsschrift, all unsaturated
expressions are relational in this sense. But even if logically ideal predication is
relational as opposed to classificatory, there seems to be a psychological distinction
between relational and classificatory concepts, even if we speak of monadic/dyadic/n-
adic predicates.18
18
If an instance of BETWEEN is a predicate that applies to ordered triples <x, y, z> such that x is between y
and z, and an instance of ABOVE is a predicate that applies to ordered pairs <x, y> such that x is above y, then
non-monadic predicates must be distinguished from predicates that are collectively (but not distributively)
plural in another sense of applying to some things without applying to any one of them. Consider FORMED-
A-TRIO, instances of which might apply (collectively) to three people—perhaps identical triplets—one of
. Meanings as modest Begriffsplans 
I assume that unboundedly many Slang expressions correspond to monadic

predicates. Here, I am not just thinking about phrases like ‘red dot’. A noun can
also combine with a relative clause—as in ‘doctor who saw a lawyer’, or ‘lawyers who
a doctor saw’—to form a phrase that corresponds to a conjunction of predicates.
Such combination is often described in neo-Fregean terms, by saying that (subject to
grammatical restrictions) two expressions of type <e, t> can be combined to form a
third; where things of type <e> can be plural entities (e.g., sets of lawyers) or events
(e.g., episodes of seeing a lawyer). This invites the hypothesis that words like ‘dot’ and
‘doctor’, along with relative clauses, are used to access Fregean concepts that relate
entities to truth values. But I suspect that the accessed concepts are genuinely
monadic, and that the corresponding Slang expressions are instances of a basic
type <M>. These expressions have meanings that are instructions for how to access
or build mental predicates, which classify things.
Slangs provide several ways of creating complex expressions of type <M>. But one
way is to simply (ad)join two expressions of type <M>. Suppose that Σ and Σ0 are
such expressions, with meanings μ(Σ) and μ(Σ0 ). Then Σ^Σ0 —the result of combin-
ing Σ and Σ0 —is an expression of type <M> whose meaning, μ(Σ^Σ0 ), is an instruc-
tion for how to build a monadic concept by conjoining two concepts obtained by
executing μ(Σ) and μ(Σ0 ). If the meaning of ‘doctor’ is used to fetch a monadic
concept, and the meaning of ‘who saw a lawyer’ is used to build a complex concept of
the same type, then the meaning of ‘doctor who saw a lawyer’ can be described as an
instruction to execute the constituent instructions and conjoin the results.
Many conjunctive operations are conceivable. For example, we have grown accus-
tomed to an ampersand that can connect an open sentence that has two unbound
variables with an open sentence that has three unbound variables to form an open
sentence that has four unbound variables, as in ‘Sxy & Byzw’. In my view, combining
meanings of type <M> invokes a simpler operation that is limited to monadic
concepts: given two such concepts, joining them creates a third that applies to an
entity e if and only if each of the two constituent concepts applies to e. This operation
has set-theoretic and truth-theoretic analogs.19 But I posit “M-junction” as an
operation that applies to certain mental representations, not the represented things.
And because this operation is limited to predicates, applying it does not require
variables or any other mechanism for establishing which saturatable position is
linked to which; see chapter three.
Relative clauses offer another way of constructing predicates. The meaning of ‘who
a doctor saw’ seems to be a result of performing some operation on a sentential
meaning like that of ‘a doctor saw her/him/them’. This makes it tempting to say that
expressions of type <t> can be converted into expressions of type <e, t>. But this neo-
Fregean hypothesis about Slangs goes far beyond the mere claim that there are truth-
evaluable thoughts and entity-denoting concepts. It describes relative clauses as
whom is between the other two. But let’s delay discussion of this complication and the question of how
triplets can form a trio that does not form triplets.
19
Higginbotham () spoke in terms of modification and identifying saturable/bindable positions.
Heim and Kratzer () say that one expression of type <e, t> can restrict or modify another.
 . Overture
expressions of a non-basic type, characterized in terms of abstracting entity-denoters

from sentences, which are described as expressions of a basic type that is character-
ized in terms of truth. I will defend an alternative view, according to which sentences
and relative clauses are special cases of type <M> expressions.
As we’ll see, it isn’t clear what sentences of a Slang are. But sentence meanings can
be described as instructions for how to build “polarized” predicates that are satisfied
by each thing or by nothing; cp. Tarski (). This suggestion can be encoded in
terms of two operations on monadic concepts. Given any such concept M, applying
the operation “UP” yields a polarized concept, ⇑M, that applies to each thing if
M applies to something; applying “DOWN” yields a polarized concept, ⇓M, that
applies to nothing if M applies to something. For example, an instance of ⇑RABBIT
applies to everything or nothing, depending on whether or not there is a rabbit. An
instance of ⇑RABBIT^HOPS applies to everything or nothing, depending on whether or
not something is both a rabbit and a thing that hops; and likewise, ⇑HOP^DONE-BY-A-
RABBIT applies to everything or nothing, depending on whether or not something is
both a hop and done by a rabbit. But the idea is not that ⇑ is a quantifier that binds a
variable, as in ‘∃e[Rabbit(e)]’, much less a quantifier of the Fregean type <<e, t>, t>.
A polarized concept is still a predicate.20
From this perspective, a relative clause is a tool for converting a polarized
predicate into a more complex predicate that can apply to some but not all things.
If only for simplicity of initial exposition, let’s assume that ‘who a doctor saw’ has the
following grammatical structure: [who [a doctor saw t]]; where ‘who’ has been
displaced from the embedded sentence, leaving a coindexed trace. I’ll argue that the
meaning of the embedded sentential expression, which contains an index, is an
instruction for how to build a polarized predicate that is correspondingly indexed.
Borrowing from Tarski, the idea will be that a predicate containing one or more
indices can apply to each thing or to nothing relative to an assignment of values to the
indices. For example, relative to any such assignment A, a polarized predicate that
contains exactly one index might apply to each thing or nothing, depending on
whether or not a doctor saw whoever (or whatever) A assigns to that index. The
meaning of a relative clause can then be described as an instruction for how to build a
correspondingly de-polarized and de-indexed predicate, which might apply to an
entity e if and only e was seen by a doctor.
Spelling out the proposed de-polarizing/de-indexing operation, “M-abstraction,”
requires a review of (i) Tarski’s notion of satisfaction and (ii) how assignments of
values to indices are related to the notion of abstraction made explicit in Church’s
() lambda calculus; see chapter three. But the same review is required to
understand how expressions of type <t> could be converted into expressions of
type <e, t>. Once the formal background is clear, I’ll argue that the Fregean typology
is neither needed nor wanted in an account of how relative clauses are related to their
embedded sentential expressions, and that the requisite operation of abstraction is
best characterized as one that builds predicates from polarized predicates. This has
20
As discussed in chapter seven, the logic of “truth tables” is easily recast in these terms.
. Meanings as modest Begriffsplans 
implications for quantificational constructions, like ‘chased every rabbit’. But here
too, I’ll argue that Fregean typology is neither needed nor wanted, and that the
requisite operation of abstraction both applies to and yields mental predicates.
In short, Slangs let us access and assemble monadic concepts that can be con-
joined, indexed, polarized, and used as bases for a limited kind of abstraction. I think
Slangs also let us access some dyadic concepts, in a limited way that has a dramatic
effect. Indeed, much of this book is devoted to showing how dramatic the effect
can be.
. Modest relations

I readily grant that expressions of a Slang can be used to assemble and express
thoughts whose constituents include some relational concepts. For example, speakers
can think and say that every rabbit was faster than most of the dogs that like to chase
rabbits. This invites a pair of empirical questions: what kinds of relational concepts
can be fetched via lexical meanings; and to what degree can phrasal meanings be used
to build relational concepts?
My strategy for addressing these questions will be to formulate and defend a
specific proposal, according to which Slangs are minimally relational. I suspect that
while lexical meanings are often used to access dyadic concepts, they are rarely if ever
used to access triadic concepts, and that phrasal meanings are (almost?) always used
to assemble monadic concepts. So I will be suggesting a very spare meaning typology
for Slang expressions: lexical items exhibit two basic types, <M> and <D>, suggesting
monadic and dyadic concepts; complex expressions are of type <M>, allowing for
sentences and relative clauses as special cases, as sketched in §. above. This requires
an additional composition principle, according to which expressions that exhibit the
two basic types can be combined to form an expression of type <M>. But otherwise,
the proposed system for constructing concepts is fundamentally geared to predicates.
Indeed, the suggestion will be that allowing for a little dyadicity can greatly enhance
the utility of an otherwise boring system that generates only predicates.
I can’t establish, even for a single Slang, that every construction can be plausibly
described in these spare terms. But I’ll argue that the simple typology accommodates
textbook examples, along with many others, in attractive ways that often yield
explanations of otherwise puzzling facts. So I think we should at least explore the
option that Slangs are minimally relational, and that Slang expressions exhibit very
few types. If my initial proposal is too spare, it is relatively easy to relax the
constraints and allow for more non-monadic expressions.
By contrast, Frege’s Begriffsschrift is unlikely to provide a good model for Slangs,
despite frequent claims to the contrary. As noted above, Frege designed his language
to allow for expressions of (boundlessly many) types that are apparently not exhib-
ited by Slang expressions. But the issue here is not just about the “higher” types that
Frege discussed in the context of his proposals about how arithmetic is related to
logic. The important point, stressed in chapter three, is that Frege’s logic was
fundamentally relational as opposed to predicative. So his Begriffsschrift was designed
to accommodate relational notions of any adicity. Yet as discussed in chapter six, a lot
of evidence suggests that Slangs abhor even triadic concepts.
 . Overture
For example, we say that Jupiter is between Mars and Saturn—not that Jupiter
betweens Mars Saturn—as if we need to circumlocute, using ‘between’ along with
‘and’ to form an oddly complex predicate, instead of expressing a triadic concept
more directly. We can say that Judy gave Martin the soap. But a verb can—and in my
view, ‘gave’ does—appear in ditransitive constructions without expressing a triadic
concept. We can also say that Judy sold/threw/kicked Martin the soap, or that she
sold/threw/kicked/gave the soap to Martin. More generally, I’ll argue that despite
some constructions that might suggest otherwise, lexical meanings are either dyadic
or monadic. If this is even roughly correct, then instead of viewing Slangs as pale
imitations of a Fregean language that was invented for certain scientific purposes,
perhaps we should view Slangs as modest extensions of a more Aristotelian model in
which predicates play a central role; cp. Sommers (), Parsons (, ),
Schein (, , ), Pietroski (a), Pratt-Hartmann and Moss (),
Moss ().
There are many ways to extend the predicate-generator sketched in §. above.
But one simple way is to add an operation that lets an expression of the dyadic type
<D> combine with an expression of the monadic type <M> to a form another
expression of type <M>. Recall that given two monadic concepts, the operation of
M-junction yields a third such concept that applies to an entity e if and only if each
of the two constituent concepts applies to e. Characterizing an analog operation of
D-junction is relatively easy: given a dyadic concept and a monadic concept, this
operation yields a concept that applies to an entity e if and only if e bears the dyadic
relation to some entity that the monadic constituent applies to.
For example, D-joining an instance of ABOVE with an instance of RABBIT yields a
monadic concept that can be represented as shown below.
∃[above(_, _)^rabbit(_)]
The monadic concept is targeted by the “internal” slot of the dyadic concept, and
also targeted for a kind of closure, so that the resulting concept applies to an entity e
if and only if e is above a rabbit. But the idea is not that ‘∃’ binds a variable, as
in ‘∃y[Above(x, y) & Rabbit(y)]’, much less that ‘∃’ is an instance of the Fregean
type <<e, <e, t>>, <e, t>>. As discussed below, that would allow for analogs of
‘∃x[Above(x, y) & Rabbit(y)]’, and ‘∃x[Above(x, y) & Rabbit(x)]’.
My suggestion is that Slangs employ a simpler operation that makes it possible to
link a two-slot (doubly unsaturated) concept to a one-slot (predicative) concept in a
particular way, thereby forming a complex predicate with the following features: its
one unsaturated slot corresponds to the “external” slot of the dyadic concept; and it
applies to things that bear the dyadic relation to something that meets the condition
imposed by the constituent predicate. This book is largely an attempt to see how
much can be explained by identifying meanings with instructions for how to access
and assemble concepts that exhibit this spare typology.
. Meanings as modest Begriffsplans 
. Idealizations idealize

If meanings are instructions for how to build concepts, then the hypothesis that each
meaning is a concept can be viewed as a simplification that abstracts away from
polysemy. One can imagine minds that are disposed to connect each lexical item with
exactly one concept. Faced with the input that a typical child receives, such thinkers/
speakers might acquire a similar lexicon that exhibits more homophony. Or perhaps
they would ignore much of the usage that leads to conceptual equivocality in minds
like ours. I don’t deny that a capacity to acquire Slangs might be combined with
independent constraints that preclude conceptually equivocal expressions. And for
some purposes, it may be useful to pretend that our minds impose such constraints.
But we shouldn’t conclude that each Slang expression connects its pronunciation
with only one concept.
Put another way, the hypothesis that meanings are concepts can viewed as a severe
idealization concerning the minds that can use meanings to assemble concepts. If we
assume that for each Begriffssplan, there is only one concept it can be used to build—
i.e., that executing the instruction always yields a concept of the same type—then
each Begriffssplan can be identified with the corresponding concept type. This leaves
room for conceptual equivocality across the members of a speech community. But as
noted above, one might give up the idea that successful communication requires that
speakers use the same meanings/concepts.
Alternatively, one might say that meanings are extensions of idealized concepts.
This hypothesis can be viewed as an idealization that abstracts away from many cases
of polysemy by imposing an external constraint on how expressions can connect
their pronunciations with concepts. But the proposal allows for a limited kind of
conceptual variation across (and within) speakers. One can imagine communities
that try to promote and maintain the external constraint. And for some purposes, it
may be useful to pretend that meanings are extensions, or that understanding the
expressions of a Slang is a matter of knowing what the expressions are true of. But we
shouldn’t conclude that each Slang expression connects its pronunciation with an
extension or a class of extensionally equivalent concepts. Ideal minds might be able to
use Slangs this way. But it doesn’t follow that our minds are ideal, much less that
meanings themselves enforce the constraints that certain minds might impose.
As I will stress in the next chapter, we should also be cautious when adopting the
idealization that members of a community have acquired the same language—or that
they use the same words, or words that have the same meanings. There may be as
many English languages as there are speakers of English. And as noted in section three
above, speakers who resolve homophonies in the same way may still talk past each
other in various respects. One can adopt a more demanding notion of “same meaning,”
and then ask whether speakers who communicate to some degree do so by using
expressions that have the same meanings. But in my view, focusing on communication
is not a useful way to get at the natural phenomenon of linguistic meaning. (I think it’s
more helpful to focus, at least initially, on how meanings compose.)
One can say that ‘rabbit’ is the English translation of ‘lapin’, the French translation
of ‘rabbit’. In this sense, the two words have the same meaning; and one can ask in
 . Overture
virtue of what this is so. Similarly, one can ask in virtue of what my word ‘rabbit’ is
the homophonic translation of yours. If the meaning that I connect with the
pronunciation of ‘rabbit’ is an instruction for how to access a concept from a certain
lexical address of mine, and the meaning that you connect with a very similar
pronunciation is an instruction for how to access a concept from a certain lexical
address of yours, one might want to know what makes my word ‘rabbit’ the analog of
yours. But in my view, we shouldn’t expect deep answers.
One relevant fact, I assume, is that the words in question are so often used to
access concepts of rabbits and rabbit (and not, say, concepts of dogs and dog).
Indeed, concepts like RABBIT-KIND may play an important role in letting us coordinate
language use, in ways that often allow for smooth translation. We can say that ‘lapins’
is the French word for rabbits, and that my word for rabbits sounds a lot like yours.
This may be a kernel of truth lying behind that idea that meanings determine
extensions; cp. Putnam’s () metaphor, noted earlier, of coarse grids. But it
may also be relevant that conceptual equivocality can facilitate communication. If a
lexical item can provide access to a range of concepts, then a sympathetic listener
may be able to adopt and use another speaker’s concepts, at least temporarily. If we
meet someone who thinks that tofus have visited Hesperus but not Phosphorus, then
we might try to think about tofu and Venus in the ways that he does, even if we reject
the thought that he endorses. Or the conceptual adjustment might be mutual. Instead
of arguing about whether a certain pen is green, we might agree on definitions for
SEMIGREEN and SUPERGREEN, and agree that only the first of these concepts applies to the
pen. This might be a special case of letting a lexical item become more conceptually
equivocal than it already was; cp. Wilson and Carston () on “ad hoc” concepts.
So a lexical item that I use to access certain concepts might count as the analog of a
lexical item that you use to access somewhat different concepts, in part because we
could each make our addresses more polysemous in a way that would let us access the
same concepts.21
That said, there may be many cases in which there is no determinate answer to the
question of whether or not two speakers have connected a certain pronunciation with
the same meaning—or a certain meaning with the same pronunciation—absent a
stipulation about what counts as sameness for the purposes at hand. My suspicion,
pace Quine (), is that this doesn’t matter. It may be tempting to say that meanings
are what expressions with the same meaning have in common, and then ask what it is
for an expression Σ to have the same meaning as an expression Σ0 . But I think that’s a
distraction. If we want to know what meanings are, we should try to figure out what
Slangs connect with pronunciations, and delay questions about how meanings are
related to our ordinary talk of expressions having the same meaning. Ordinary talk is,
at best, an imperfect guide to the natural phenomena we sometimes talk about.
21
See also Carston (, ). Davidson () allows for cases of changing one’s language on the fly;
see Ludlow () for related discussion. But I don’t think that making a word more polysemous should be
assimilated to cases of connecting pronunciations with new meanings, as in Davidson’s example
of someone connecting Mrs. Malaprop’s pronunciation of ‘derangement of epitaphs’ with the meaning
of ‘arrangement of epithets’.
Another random document with
no related content on Scribd:
bracelets are made of them. It is one of the most beautiful trees in
the country when in flower, in which state it remains for some time.
The Artecum, or Araticu, of which there are various kinds, is
similar to the ambuzo in size, in the entwining of its branches, and in
having an ash-coloured rind. The leaves, which it casts off annually,
are almost round, and of a beautiful green, and varnished on both
sides. Its flower is like a fig, of a yellow colour, which opens in three
equal portions, as if they were parted with a knife, are thick, of a
reddish hue, and in the form of a shell; below each portion there is a
smaller shell, not so thick, and externally white, inclosing a button in
the shape of a pine, which becomes a fruit of the size and form of
the largest pear; the pulp is white or yellow, soft and full of pips; few
are good, but they are generally eaten.
The Barriguda, (Big-bellied,) so called in consequence of its
trunk being thicker at the middle than at the ground, and otherwise
denominated Arvore da lan, (Wool-Tree,) has the rind covered with
round thorns, the leaves are almost elliptical, and from three to four
inches in length. The flower begins with the figure of a little round fig,
smooth, without an eye, having a short and thick pedicle; this fig has
the length and form of an acorn, being always of a green colour, it
opens at the extremity in three parts, from whence issues a bud
similar to that of cotton, round, and two inches long, which unfolds in
five petals three inches in length. It has five capillaments surrounded
with a pistil, which is delicate and terminates in a little red ball, the
whole six attached to a receptacle at the bottom of a cup, where a
fruit is formed with some resemblance to a cucumber, which, when it
opens, is full of fine white wool, which the wind blows away if not
immediately taken; it cannot be spun, but it is used for stuffing beds.
There are three sorts of Brazil Wood;—Brazil Mirim, which is
the best; Brazil Assu, or Rozado, so called from its trunk being
higher, more upright, of less girth, and the dye which is extracted
from it of less consistency and more rosy, hence arises its second
name. The Brazileto differs little from the Assu in size, in the form of
its trunk and tuft, and affords little dye. The Brazil Mirim has a larger
trunk, the rind greener and finer, the thorns smaller and thicker, the
leaf smaller, and the wood more of a purple hue. The bark of the
whole is smooth and the thorns begin at the shoot and extend to the
point of the branches. The flower of the Mirim is white and very
small; the wood, immediately on being cut, if put to the tongue, is
sensibly bitter, which, on becoming dry, it loses, turning to an
agreeable sweetness. These trees grow on rocks as well as in
plains. The wood is heavy and excellent for building; put into water it
will remain sound, it is said, eternally; put into the fire it breaks in
pieces, but does not produce any smoke. It is said, that the most
proper periods for cutting this timber for its dye is at the time of the
new moon in winter, and when in the fourth quarter in summer. A
portion of vitriol and lime, or ashes thrown into the decoction of the
Brazil wood forms a black dye.
Cacauzeira, or Cocoa-Tree, which generally does not exceed
the size of a small tree, almost always has many trunks, and extends
its branches horizontally and incliningly; the leaves resemble those
of the largest chestnut; its fruit, which grows only upon the trunks
and thickest branches, is a large oblong capsule, in the shape of a
mellon, with a very hard rind, which encloses forty to fifty almonds in
a white, viscous, and sweet substance. Of these almonds chocolate
is made.
Cafeeiro, or Coffee-Tree, transplanted from Arabia, and
prospering admirably in the strong and fresh soils of this country, is a
branchy shrub, with the leaves opposite, which are smooth, pointed
at both extremities, and larger than those of the laurel; the flower is
white, entire, and tubulous on the lower part, and cleft into five lancet
points on the upper, with the same number of capillaments attached
to the angle of the divisions, and a pistil to the receptacle. The fruit is
a berry, somewhat resembling a cherry, attached to a thick and short
pedicle. It is exceedingly fruitful, and grows to the greatest size when
planted under the shade of large trees; but the fruit of those trees
exposed to the sun is of the best quality.
Cajaty is a shrub with a very thick rind, black, and furrowed with
the first bark, the leaf differs little from the laurel, and affords a yellow
fruit the size of the sloe, having an agreeable taste and perfume,
with seed like the orange, and attached to the extremity of the
branch by a long and delicate pedicle.
Cajazeira, or Cajaza-Tree, is high, with a round tuft and small
leaf, its fruit is like the acorn, yellow, insipid, and with a large stone.
The Cajueira, or Cajue-nut Tree, which grows in sandy soils, is
a small tree with a crooked trunk, a round tuft, with the branches
raised horizontally and inclined; the first rind affords a sort of black
dye; the wood a yellow one; the leaves are almost round; the flower
in bunches; its singular fruit is of the size and figure of a long round
pepper, with a fine smooth skin, tinged with red or yellow, and
sometimes with both, with a white spongy substance, half sweet,
without stone or seed; the taste is not agreeable, although many
persons are partial to this fruit; it has at the extremity a nut of an ash-
colour, containing a white and oily substance, and not unlike the
chestnut when roasted. Excellent sweetmeats are made of the fruit,
and also a lemonade and even wine of its juice.
Calumby, or Calumbo-Tree, is small, with little tuft, having
exceedingly small leaves, which close when the sun sets and open
in the morning. There are male and female, both armed with thorns
like the bramble; the male has a very small flower, resembling a
paint-brush, and its wood is green, hard, compact, and very heavy,
being only serviceable for joiners’ work, as scarcely one is found,
even amongst the most slender, without being hollow. The female
has a flower resembling the chestnut-tree, and its timber, although
solid, is not so heavy, and is waved with green and white. The fruit of
both are flat berries.
Candea (Lamp) is a crooked shrub, with a large stock, the leaf is
generally white on the under side; the wood, when dried, affords a
good light without smoke, and saving much oil to the poor of the
certams, who put a fire-brand of it into the wall, where it lasts for a
long time, giving a flame like a flambeau; hence comes its
denomination. When put into the ground it keeps many years.
Caneleira, Canela or Cinnamon Tree, was transplanted from
Asia, and was cultivated with some care at the commencement of
the colonization, but was afterwards destroyed, by royal order, that it
might not interfere with the oriental trade; this error was soon
discovered. At present the increase of those which spring from roots
is preferred; the quality of the soil in which it should be planted
depends upon the vegetable. The best Asiatic cinnamon-trees grow
in dry soils.
Colla is a tree introduced from Africa, of medium size, with
leaves rather long, pointed, and shining on both sides; the flower is
white and in bunches. It fructifies in a pod.
Cutezeira, or Cuteza-Tree, is small, with horizontal branches,
and the leaves rather long, smooth, wide, rounded at the point, and
pointed at the base; it affords a large oval fruit with a hard shell, of
which, when parted, is made cuyas, or cups.
Gamelleira, or Gamella-Tree, is of considerable girth, having a
round and thick leaf; it extends roots from the branches to the
ground; its shade is sought after.
Geremma is a small thorny tree with very small leaves, which
daily close from the influence of the sun; the flower resembles that of
the chestnut; of its wood charcoal is made for forges.
Goyabeira, or Goyaba-Tree, is a shrub with a smooth rind, the
leaf almost round, the fruit, like a pear, is yellow and odoriferous, the
pulp is rubicund, with a great quantity of small and round seed in the
centre. A preserve much esteemed is made of it.
Guabirabeira, or Guabiraba-Tree, is one of the largest fruit-trees
in the country; its leaf is a little less than that of the peach, and
almost of the same shape; the flower is white; the fruit has the size
and form of a pear, and is eaten in the same way as sorvas or
service-berries.
Herva de Cobra (Serpent-Herb) is a small plant resembling a
little the fern in its foliage, the flower is small and yellow, with five
petals, and its fruit is a very little berry. The name arises from its
virtue in curing the bite of a snake by placing the leaves pounded
upon the wound and giving the patient the juice to drink.
The Jabuticabeira, or Jabuticaba-Tree, is small and slender,
with smooth bark; the leaves, which are varnished on both sides, but
not of the same green, vary in form upon the same branch. It flowers
upon the trunk only, beginning at the ground, and extending to the
branches, that are thick; the fruit is generally of the size and colour of
a cherry, and of a good flavour; a strong liquor is distilled from it.
The Jambeira, or Jamba-Tree, whose size is proportioned to the
quality of the ground upon which it grows, has a pointed leaf, with
dark green on the upper side, and clear green on the under; the
flower has four small petals in the form of a shell, with numerous
long and upright capillaments, and a pointed pistil still longer; its fruit
resembles an apricot, has a fine flavour, and smells like a rose.
The Jaqueira, or Jaca-Tree, transplanted from Asia, and
prospering only in the tropics, is a large tree, with a round tuft, a thick
trunk, large leaves rounded at the extremity and pointed at the base,
varnished on both sides, the upper one of dark green. Its fruit, which
only grows upon the trunk and main branches, is very large, (some
forty pounds weight,) of an oblong form; the skin is green, rough-
grained, and thick; the white pulp is fibrous, and impregnated with
viscous milk; but it has another pulp, sown promiscuously with a sort
of almond, less fibrous, without milk, and sweet, and which is the
part eaten.
Jatuba, which has not this name in all the provinces, fructifies in
husks like the tamarind-tree.
Jenipapeiro, or Jenipapo-Tree, is of good height with the trunk
erect, the tuft round and of medium size; the leaf is similar to that of
the chestnut, very thick, and of a dark green; it is never without fruit,
which is the size of an apple, the skin tenuous, a little harsh, and of
an ash colour, the pulp is white, and the interior full of seed. They
remain from one year to the next upon the branches, which wholly
shed their foliage, and only change when the tree puts forth its new
leaves, and when already the new fruit, for the following year, is of
good growth. Its wood is preferred for the shafts of the sege, or
cabriolet.
Jiquitiba is a tree of considerable girth and affords a nut.
The Joazeiro, or Joaza-Tree, which grows in sandy soils, is of
the size of a middling olive-tree; its tuft is round and thick; the leaves,
which it scarcely sheds, are round and carefully protected; its wood
is white, and its ashes are a good substitute for soap. The flower is
in small and round bunches; the fruit is of the size of a cherry,
oblong, yellow, disagreeable to the touch, and with a stone (full of
smaller ones) which is difficult to divest of the pulp when the fruit is
not half dried. It is aliment for some quadrupeds, and for the jacu and
other birds, although few are well flavoured.
Mangabeira, or Mangaba-Tree, is of medium size, with small
pointed leaves, and a flower like the jasmine; the fruit is round and of
various sizes upon the same branch, with a yellow and greenish rind;
the pulp white, extremely soft, with various seeds covered with
down; the wood, the leaves, the flowers, and the fruit, when parted,
distil a clammy and white milk.
The Mangue is a small tree, with a smooth rind, thick and
varnished leaves. It prospers only upon the sea-coast or the margins
of salt rivers.
The Mangueira, or Mangua-Tree, transplanted from Asia, and
prospering only in the torrid zone, is a bulky tree, with a leafy tuft,
having a long leaf a little narrow and pointed. The fruit is the size of
an apple and a little flat, the skin similar to a greengage, green or
yellow, and tinged sometimes with red; when divested of the skin,
which has a turpentine taste, the pulp is juicy and delicious, although
unpleasantly full of fibres attached to the stone.
Mozes is a tree of medium height, slender, with little tuft,
remarkable for its foliage, which does not differ from the fern; its
flower is white, and its ashes are good for the lixivium of soap.
Mucory is a large tree of excellent timber; its fruit is of the size of
a sloe, yellow, and odoriferous, of very fine flavour, and has a large
stone.
Muricy is a shrub, with large thick and harsh leaves, rounded at
the end and pointed at the base; the flowers are in bunches,
beginning with yellow and turning to a carnation colour; the fruit is
very small, with little flavour. There is another called muricy-bravo,
(or wild,) differing in the flower, which is white, and in the leaves,
which are much less elliptical and varnished on both sides.
Oyty is a middling tree, of good timber, with fruit well flavoured,
and of the colour and form of a pear, with a large stone, which, when
ground or scraped, and used as a beverage or as a clyster, is an
efficacious remedy against diarrhæas.
The Palm-tree of Dates, which is so abundant in Asia and
Africa, are introduced only, and very partially, in the province of Rio
de Janeiro.
Amongst the multiplied species of Palm-trees that denominated
Tucum or Tycum is particularly remarkable; its trunk is thorny,
slender, and of proportionable size; its leaves differ a little from the
common resemblance observed amongst all the other palm-trees;
from its fibres a flax is made that is a little harsh, but as lustrous as
silk, without any appearance of the coarsest description of flax, and
which, from its strength, is generally consumed in making fishing-
tackle. It is well adapted for making a certain sort of lace.
Pindahiba is a handsome tree, and of proportionable size
according to the quality of the soil in which it grows; its wood is light;
its leaves are lancet, one inch in width, and from three to four in
length; it fructifies in very small bunches, and its berries are
sometimes used as peppers.
Piquiha, is a medium-sized tree, affording fruit like the quince,
with a thick and hard rind, and full of a gray liquor, very sweet and
cooling, with some seed like those of an apple.
Pitangueira, or Pitanga-Tree, which reaches the size of a plum-
tree when planted in good ground, but generally not exceeding the
size of a middling shrub in the woods; its leaf resembles that of the
myrtle; the flower is white and small, with a great number of
capillaments; the fruit is the size of an unripe cherry, of a scarlet or
purple colour, and rather sour. An agreeable spirit is distilled from it.
Quinaquina, the Jesuit or Peruvian Bark, was discovered about
three centuries ago in Peru, and met with only a few years since
near the heads of the river Cuiaba; it is a high tree, nine inches in
diameter; the leaves are round at the base and pointed at the end,
glossy and of a beautiful green above, and striped with a brilliant
dark green in the half near the base. The flowers, which are in
bunches at the extremity of the branches, are shaped like a funnel,
with the edge parted into five lancet forms, and shorter than the tube,
hairy, green in the middle, bounded with white, and fringed at the
borders. The pistil is white, and surrounded with five capillaments,
within the tube of the flower. When the flower falls the cup swells at
the middle, and takes the shape of an olive, changing into a fruit,
whose numerous seed, which are long, thick, of a green colour, and
flat at the edge, are enclosed in two lodgements, divided by a double
membrane. Thus a tree so useful to mankind is propagated
abundantly.
Amongst the Resin-Trees are the Angico, which produce the
gum-copal; those that produce mastick, benzoin, and storax;
amongst those that distil balsam are the cabureigba, better known by
the name of Balsam of the Holy Spirit, the cupahybu, or capivi, and
the cumaru.
Amongst the medicinal plants is noted sarsaparilla, ipecacuhana,
jalap, butua, purging cassia, quassia, aristoloquia, or hart-wort,
cahinana, Jesuit’s bark of the country, ginger, capeba, commonly
called herb of St. Luzia, from its great virtue on application to
diseases of the eyes. It is said that a surgeon of Rio de Janeiro, in
the year 1784, by only using the juice of this plant, in the course of
three months, restored the eye of a soldier to its former state, which
had been injured by the point of a bayonet. Marvellous cures are
related of this vegetable, which is said to regenerate the sight;
experiments have been made by perforating the eye of a cock with
sharp instruments, and on applying the juice or even the milk of this
herb it is asserted that the eye has been cured in a few hours, and
the sight restored. The curucu, whose juice, when drunk with water,
is an efficacious stiptic for a bleeding at the mouth. There is also
betony, ground-ivy, but very different from that of Europe, with a leaf
resembling the rosemary, and a small white flower in a species of
artichoke; the herva ferro (iron herb); the herbs mercury, eurucucu,
and mallows; orelha d’onça, (ear of the ounce,) generally two feet in
height, the leaf like a heart, flat, and hairy on both sides, of a pearl
colour, and as flexible when dried as when green; the plant called
hervachumbo; and many others.
Sapucaya is a high tree of good timber, with a leaf similar to that
of the peach; the bark, softened, produces a tow for caulking
vessels. Its produces a very large spherical nut, full of long almonds.
For their extraction nature has formed an orifice at the extremity four
inches in diameter, covered with a lid of the same size, which has
over it an outer rind similar to that of the whole nut, and of which it is
necessary to strip it in order to find the entrance. The monkeys, by
instinct, shake off this species of cocoa-nut when ripe, and with a
stone, or hard piece of wood displace the lid and eat the almonds.
St. Caetano is a delicate plant, resembling that of a water-
melon; its fruit is a species of small cucumber and thorny; it opens in
three portions when ripe, exhibiting some small seed similar to those
of the pomegranite. It is the sustenance of birds, who, carrying its
seed, propagate it in all parts. This plant is applied to various
domestic purposes, and augments the properties of soap in its
ordinary use; on this account it was transplanted from the coast of
Guinea, where it is called Nheziken, and being planted near a chapel
of St. Caetano, took the name of that saint.
Tababuya is a tree remarkable for the lightness of its wood, of
which scarcely any thing is made besides corks and floats for
fishing-nets; it resists all instruments except such as are used for
cork¬cutting.
Taruman is a shrub with lancet leaves of unequal size; the tea of
these leaves have a diluent effect upon stones in the bladder.
Theu is a delicate sipo or plant of long and flexible shoots,
scarcely exceeding the thickness of a hen’s quil, but of extraordinary
growth, always winding round other larger plants and trees. I have
seen them so firmly entwined round orange-trees that the prosperity
and fructification of the tree was impeded by them; its leaf is
exceedingly small, resembling that of the broom; the root is nearly
two yards in length, having a strong smell, and operating as an
emetic, and is an approved remedy against the venom of snakes.
A great diversity of piratical trees or plants are observed in the
Brazil, fixed to the bark or body of others, and nourished alone by
their substance. In some parts there are divers species of climbers
which rise to the top of the highest trees, sometimes
unaccompanied, at other times twisted spirally with another of the
same, or of a different species. Occasionally these prodigiously long
cords have four, six, or more legs, or shoots.
Tinguy is a small tree with the branches and leaves alternate;
the latter are small and lancet. The bark and leaves well pounded,
and put into lakes, &c. cause the fish to die, from becoming soon
intoxicated with it.
The Urucu does not in general exceed the size of a large shrub;
the leaves are in the form of a heart, and the flowers in bunches with
fine petals a little purpled, a pistil, and a great number of
capillaments; the fruit is a capsule, a little flat and pointed, of the size
of a large chestnut, and of a green colour, composed of two valves
or folds, covered with fine soft thorns, and lined with a membrane
that encircles a large quantity of small seed, having over them a
green substance which, when diluted in water, affords a precious
dye. The Indians are not ignorant of this, and use it to paint their
bodies.
Vinhatico is a high and straight tree of yellow wood, and
fructifies in pods with beans.
There are a variety of edible roots:—potatoes of various kinds;
mandioca, of which is made the usual bread of the country; its plant,
of which there are various sorts, is a shrub of one or more stems; the
root, after being scraped and reduced to flour, is pressed until
exhausted of its abundant juice, which is generally poisonous, and is
ultimately toasted in a large earthern or copper vessel over a furnace
till it becomes dry; this vegetable prospers well only in substantial
soils: it is planted in little mounds of earth, by putting into each half
the stock of the plant, which is a span in length. The aypim is a
species of mandioca, whose root is boiled or roasted. The mendubim
is a plant of little growth, with leaves similar to the French bean,
producing beans at the root with a gray skin, which encloses from
one to three small seed. The potatoe do ar, a creeping plant, without
a flower preceding it, produces a fruit of irregular form, without stone
or seed, is covered with a thin and green skin, and has the taste of
the potatoe.
Besides the fruits mentioned there are many others, amongst
which are the pine, or atta, the size of a quince, with a white, soft,
and savoury pulp; the conde, which is of the size of the preceding,
with the pulp equally soft, but not so white; the mammao is larger,
with a smooth and yellow skin, and the pulp of the same colour; the
pitomba; the mocuge; of the banana, whose length exceeds many
times its diameter, there are three sorts, the whole having a thick
skin, and clustered upon one stalk; the plant which produces them is
of considerable growth, without either branches or wood in the trunk;
the leaves are very long, slender, and smooth, with proportionable
width, and the back fibres very thick; the trunk is formed of the
leaves firmly woven together, being two or three yards in height,
round, erect, and inflexible, the leaves branching out from it at the
top. The pine-apple, resembling a pine, with various leaves in the
eye, is of delicious flavour and aromatic scent; the plant from whose
centre it issues is very similar to the aloe. The muracuja is of the size
of an orange, oblong and regular, with a thick and hard skin, green
on the outside and white within; it is full of gross and rather sour
liquid, containing seeds similar to those of the melon.
The sugar-cane, mandioca, tobacco, and the matte-plant, are all
indigenous, and now cultivated to a great extent with considerable
advantage, furnishing many lucrative branches of commerce.
The indigo-plant and opuncia are met with almost in all parts; the
first, which only prospers in strong soils, is yet cultivated but in few
provinces. There are a diversity of peppers; that of Malabar, which
only thrives in substantial and fresh soils, has been cultivated only
within a few years.
The plant commonly called malicia de mulker (woman’s malice) is
a creeping and thorny twig, with very small foliage, whose little
leaves obtain their opposite one’s, when they immediately adhere,
so that the twig is encircled, and remains in this state for a
considerable time.
THE END.
DIRECTIONS FOR THE BINDER.
1. Don John VI. and his Attendants to front Page
the
Title
2. Map of the Brazil to front 1
3. View of the Western Side of the Bay of Rio 9
4. Custom-House Negroes 10
5. Convent of St. Antonio 52
6. Convent of St. Theresa, Part of the Aqueduct, and a Sege 53
7. Convent of Ajuda 54
8. Pillars and Scenery near the Source of the Aqueduct 56
9. House at the Bottom of the Orange-Valley 57
10. A Miner, from the Province of Minas Geraes 61
11. Bella Fonta, the Shacara of J. E. Wright, Esq. 62
12. Palace of St. Christovao 63
13. A free Negress, and other Market-Women 71
14. Nightman, Water-Carrier, Washerwoman 74
15. The Casa of a Padre, in Campinha 85
16. The Rio Exchange, Trapiche, Grass-Waggon, and Gallows 96
17. A Captain of Militia, in the Province of Rio Grande do Sul 116
18. A Paulista and a Brazilian Mendicant 176
19. A Map of the Comarca of Sabara 277
20. An Officer of Cavalry, in Minas Geraes, and a Hermit 283
21. Botocudo Indians about to cross a River 299
22. A Brazilian Sedan-Chair, and a Person begging for the
Church 336
23. A Brazilian Sesta, or Afternoon Nap 346
24. A Jangada, or Catamaran, near Pernambuco 357
25. A Mattuto returning from Pernambuco 385
26. Style of Houses at Poço de Penella 388
27. Negroes impelling a Canoe with the Vara, and Scenery at
Ponta de Cho 389
28. The House of the Senhor d’ Engenho de Torre, near
Pernambuco 391
29. The Sugar Engenho de Torre, and a Plan of its Interior 392
30. The Site and Remains of Fribourg-House, formerly the
Residence of Prince Maurice, of Nassau 393
MARCHANT,
Printer,
Ingram-Court, Fenchurch-Street.
FOOTNOTES:
[1] Eight successive days after a festival.
[2] “The said small vessels found a reef with a port within, very
fine and very secure, with a very large entrance, and they put
themselves within it.” Also, “and all the ships entered and
anchored in five and six fathoms, which anchorage within is so
grand, so beautiful, and so secure, that there could lie within it
more than two hundred ships and men of war.”
[3] “We showed them a grey parrot, which the captain had
brought with him; they took it immediately into their hands. We
showed them a sheep, they took no notice of it. We showed them
a fowl; they were afraid of it, and would not put their hands upon
it.”
[4] “I kiss the hands of your Royal Highness from this secure
port of your island of Vera Cruz. To-day, Friday, first day of May,
1500. P. V. de Caminha.”
[5] “L’altro giorno che fu alli dua di Maggio del derto anno 1500
l’armata fece vela pel camino per andare alla volta del capo di
Buona Speraza. Li quali comincioro no a piangere, et gli huomini
di quella terra, gli confortavano, et mostravano havere di loro
pieta.”—Ramuzio.
[6] “We arrived at the Cape of Good Hope at Easter-tide, and
there met with good weather. We continued the voyage, and
arrived at Besengue, near Cape Verd, where we met with three
caravels, which the King of Portugal had sent to discover the new
land, which we had found in going to Calcutta.”—From the
relation of the voyage of Cabral in Ramuzio.
[7] “The relations of A. Vespuccius contain the recital of two
voyages, which he made upon the same coast (the Brazil), in the
name of Emanuel, King of Portugal; but the dates are false, and it
is in that which the imposture consists; for it is proved by all the
cotemporary testimonies, that at the time which he names he was
employed upon other expeditions.”—Hist. Gen. des Voyages,
tom. 14. liv. 6. c. 9. Paris, 1757.
[8] This is a corroboration of Vincente Yanez Pinson’s voyage,
as the character of the Indians near those two same places where
he landed are described to be similar.
[9] “The King, Don Manuel, extremely attached to Vespuccius,
gave him the command of six ships, with which he sailed on the
10th of May, 1503, and passed along the coast of Africa and
Brazil, with the intention of discovering a western passage to the
Molucca Islands, (afterwards discovered.) “After entering the Bay
of All Saints, he navigated as far as the Abrolhos and the river
Curababa. As he had provisions only for twenty months, he took
the resolution of returning to Portugal, where he arrived on the
18th of June, 1504.”—Murery Dicc. Grand. Paris, 1699.
[10] “Ordered immediately to be prepared another armament
of caravels, which he delivered to C. Jacques, a fidalgo of his
house, and with the title of Captain Mor, ordered him to continue
in this undertaking of discovery. The armament sailed, and
pursuing the voyage, arrived upon the coast, sounding bays and
rivers, erecting pillars with the Portuguese arms upon them. He
entered a bay to which he gave the name of All Saints, and after
all necessary diligence he returned to Portugal.”—Cunha.
[11] “In the same year, 1503, Gonsallo Coelho was sent to the
land of St. Cruz, with six ships, with which he left Lisbon on the
10th day of the month of June. In consequence of not having
much knowledge of the land, four were lost, and the other two
brought to the kingdom merchandise of the land, which then were
no others than red wood, which was called Brazil, and parrots.”—
Goes.
[12] An overlooker of negroes.
[13] To kiss his hand.
[14] No one had suffered the penalty of death for a long period
before July 22, 1819, when a wretched criminal was hanged. He
had stabbed five or six men, and it was said his punishment
would have been commuted for transportation to Africa, had he
not imbrued his hands in the blood of a pregnant female, whom
he stabbed mortally.
[15] These parishes in the Brazil are considerable districts of
country, and take their names generally from the chief or mother
church.
[16] Mr. Langsdorff has recently (since his return to Europe)
presented to the National Institute of Paris the head of a
Botocudo Indian, who had been in his service for some years. He
sent this Indian to St. Helena to collect insects, which commission
he executed satisfactorily, and died soon after his return to the
Brazil.
The Botocudos are really the Aymores of the province of Porto
Seguro, and received their present name from the Portuguese, in
consequence of their custom of perforating the ears and lips, and
introducing pieces of wood as ornaments, which they imagine
render them more genteel and comely.
[17] This denomination originated in Portugal, where the judge
was prohibited from having any jurisdiction within his native town.
Hence it is applied to judges in the Brazil, universally, without the
capital.
[18] The Torres are two great and contiguous rocks, upon the
coast, in 29° 40′ south latitude.
[19] Sumacas are from ten to nearly two hundred tons burden.
[20] Bandeira is a name given in the Brazil to an indeterminate
number of persons, who provide themselves with arms,
ammunition, and every thing requisite for their subsistence, with
the project of entering the territories of the Indians, for the
purpose of discovering mines, exploring the country, or chastising
the savages; the individuals who compose those companies, call
themselves bandeirantes, and their chiefs, certanistas.
[21] Pantanos signifies marshy, and is the name given to the
swampy islands at the mouth of the Tocoary, which are
submerged at the floods.
[22] “By the treaty of limits, liberty was given to these Indians
to remain in the country or to pass to the other side of the
Guapore, which latter they did by the persuasions of the Spanish
curates. The mission of St. Roza, which consisted of four hundred
Indians, and removed in 1754, was near a waterfall, where D.
Antonio Rolin ordered a fort to be constructed. That of St. Miguel,
which was in a plain near the river, and had six hundred Indians,
was removed in 1753; and that of St. Simao, which comprised
more than two thousand inhabitants, and was removed in 1752,
existed near a river a little below the situation of the Pedras.”
Extract from the History of the Journeys of the Lieutenant-General
of Matto Grosso. M.S.
[23] Passo is a picture or image representing some of our
Saviour’s sufferings.
[24] The interior districts of provinces are so called.
[25] A bowl into which diamonds are put when found, from
whence they are taken, weighed, and registered daily.
[26] He acquired the appellation of Caramuru, which signifies
“a man of fire,” on the occasion of his first discharging a musket in
the presence of the astonished Indians.
[27] “Where there was a Portuguese, who said he had lived
twenty-five years amongst the Indians,” proving Correa’s
shipwreck to be in 1510.
[28] The province is designated, as well as the city, by the
Portuguese word for bay, from the great importance justly
attached to its fine bay.
[29] “Sepulchre of Donna Catharine Alvarez, Lady of this
Captaincy of Bahia, which she, and her husband, James Alvarez
Correa, a native of Vianna, gave to the Kings of Portugal, and
erected and gave this chapel to the patriarch St. Bento. The year
1582.”
[30]
The Don of the land is arrived,

The one hundred and fifty are ended.
[31]
The Don of the land is on the main,

The one hundred and fifty remain.
[32] The precarious nature of the tobacco trade is evidenced

by a shipment made of this article from Bahia to London, where
on its arrival it was unsaleable; the agent, therefore, reshipped it
for Gibraltar, drawing for the freight and charges he had paid. On
its arrival in Gibraltar, it was there equally unsaleable; and, after
being deposited some time in a warehouse, it was discovered to
be rotten, and condemned by government to be thrown into the
sea; after which the agent there had to draw also for the freight
and charges that he had paid, in which was included the expense
of throwing the tobacco into the sea. The shippers, no doubt,
considered the termination of this speculation as unpalatable as
the element to which it was finally consigned.
[33] One thousand reas, or a milrea, is now worth about five
shillings, and varies according to the exchange.
[34] “Oh! what a beautiful situation for founding a town.”
Hence its name of Ollinda.
[35] Some writers have stated that this Duarthe Coelho
Pereyra served as a military man in India; but Duarthe Coelho, of
whose military exploits there Barros and Farria both speak, had
not the surname of Pereyra. He died by the hands of the Moors of
the island of Sumatra, after having suffered shipwreck at the
mouth of the river Calapa, in 1527.
[36] The Dutch armament, commanded by Admiral Hervey
Zonk, consisting of sixty-four vessels, of various sizes, and eight
thousand men, landed, on the 15th of February, 1630, on the
beach of Pau Amarello, three leagues north of Ollinda, by the
direction of Judea Antonio Dias, who had resided many years in
the country and acquired a large fortune, with which he
established himself at Amsterdam. In 1654 the Dutch evacuated
the captaincy.
[37] A capibara, the animal from which this river takes its
name, is now in Exeter Change.
[38] Ollinda, although the head of a comarca, being commonly
considered, with Recife, to constitute the city of Pernambuco,
they will be described together.
[39] A Brazilian term for the Indian.
[40]
Erected
under
the Illustrious Government
of the
President and Council
in the year 1652.
[41] Some of these people are also called certanejos,
inhabitants of the certams, or interior.
[42] I have been informed, since my return to England, that a
clergyman had arrived at Pernambuco.
[43] Great River of the North; as there is Rio Grande do Sul,
(Great River of the South,) and which must occasion some little
confusion, both being called Rio Grande: it would be better to
designate this St. Roque, the cape being even a more
conspicuous object than the river.
[44] The Spaniards, from whom the French took this portion of
territory, always recognized the river Oayapoek and the Vincent
Pinson as the same river; and near its mouth a marble stone was
erected, by order of Charles V. to serve as a limit between his
conquests and those of the Portuguese.
[45]
Conserva o tardo impulso por tal modo,

Que em poucos passos mette um dia todo.
Cant. vii. p. 57.
[46] Inhabiting the calingas.

[47] Inhabiting the campos, or plains.
[48] Correiçao is a term applied to a judge going out to travel
through the district of his jurisdiction.
Transcriber’s Notes:
1. Obvious printers’, punctuation and spelling errors have been corrected

silently.
2. Where hyphenation is in doubt, it has been retained as in the original.
3. Some hyphenated and non-hyphenated versions of the same words have

been retained as in the original.
*** END OF THE PROJECT GUTENBERG EBOOK A HISTORY OF
THE BRAZIL ***
Updated editions will replace the previous one—the old editions

will be renamed.
Creating the works from print editions not protected by U.S.

copyright law means that no one owns a United States copyright
in these works, so the Foundation (and you!) can copy and
distribute it in the United States without permission and without
paying copyright royalties. Special rules, set forth in the General
Terms of Use part of this license, apply to copying and
distributing Project Gutenberg™ electronic works to protect the
PROJECT GUTENBERG™ concept and trademark. Project
Gutenberg is a registered trademark, and may not be used if
you charge for an eBook, except by following the terms of the
trademark license, including paying royalties for use of the
Project Gutenberg trademark. If you do not charge anything for
copies of this eBook, complying with the trademark license is
very easy. You may use this eBook for nearly any purpose such
as creation of derivative works, reports, performances and
research. Project Gutenberg eBooks may be modified and
printed and given away—you may do practically ANYTHING in
the United States with eBooks not protected by U.S. copyright
law. Redistribution is subject to the trademark license, especially
commercial redistribution.
START: FULL LICENSE

Conjoining Meanings Semantics Without Truth Values Paul M Pietroski Full Chapter PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Conjoining Meanings Semantics Without Truth Values Paul M Pietroski Full Chapter PDF

Uploaded by

Copyright:

Available Formats

Conjoining Meanings: Semantics

Without Truth Values Paul M. Pietroski

English Prepositions: Their Meanings and Uses R. M. W.

God in Moral Experience - Values and Duties Personified

No Truth Without Beauty: God, the Qur’an, and Women's

Power System Protection, 2nd Edition Paul M. Anderson

POW/MIA Accounting 1st ed. Edition Paul M. Cole

Men, Masculinities, and Earth: Contending with the

Inquisitive Semantics 1st Edition Ivano Ciardelli

Beyond Semantics and Pragmatics Gerhard Preyer

CONTEXT AND CONTENT

Other titles in the series:

Chapter zero: Overture 

also allowed me to learn about psycholinguistics, on the job, by collaborating with

. Human linguistic meanings

. Distinctively composable

. Pronounceable instructions

. Meanings are not concepts 

. Meanings are not concepts

. Concepts: Symbols vs. contents

. One meaning, a family of concepts

. Meanings are not concepts 

verbs used to talk about carrying or supporting or having or yielding or tolerating or

. Other kinds of conceptual equivocality

. Meanings are not concepts 

a concept of one sort to form a concept of another sort. By hypothesis, an instance of

. Meanings are not extensions 

. Meanings are not extensions

. Coextensive meanings

. Meanings are not extensions 

. Meanings are not extensions 

. Metaphysical detour: Outlandish worlds don’t help

. Meanings are not extensions 

. Polysemy redux

. Meanings are not extensions 

. Communication without extensional equivalence

implements or animal enclosures. And speakers who use this pronunciation to

. Meanings are not extensions 

as opposed to conceptual. Meanings provide a framework for a kind of communi-

. Meanings, natural kinds, and hard water

by the nature of the paradigms, as opposed to any associated appearances or

. Meanings are not extensions 

. Meanings are not extensions 

. Meanings as modest Begriffsplans

. Instructions that compose

. Meanings as modest Begriffsplans 

. Meanings as modest Begriffsplans 

. A monadic basis

. Meanings as modest Begriffsplans 

I assume that unboundedly many Slang expressions correspond to monadic

expressions of a non-basic type, characterized in terms of abstracting entity-denoters

. Meanings as modest Begriffsplans 

. Modest relations

. Meanings as modest Begriffsplans 

. Idealizations idealize

The Don of the land is arrived,

The Don of the land is on the main,

[32] The precarious nature of the tobacco trade is evidenced

Conserva o tardo impulso por tal modo,

[46] Inhabiting the calingas.

1. Obvious printers’, punctuation and spelling errors have been corrected

2. Where hyphenation is in doubt, it has been retained as in the original.

3. Some hyphenated and non-hyphenated versions of the same words have