Paul Horwich - Truth-Clarendon Press (1999)

TRUTH
TRUTH
SECOND EDITION
PAUL HORWICH
CLARENDON PRESS · OXF O R D

1 998
Oxford University Press, Great Clarendon Street, Oxford OX2 6DP
Oxford New York
Athens Auckland Bangkok Bogota Buenos Aires Calcutta
Cape Town Chennai Dar es Salaam Delhi Florence Hong Kong Istanbul
Karachi Kuala Lumpur Madrid Melbourne Mexico City Mumbai
Nairobi Paris Silo Paolo Singapore Taipei Tokyo Toronto Warsaw
and associated companies in
Berlin Ibadan
Oxford is a registered trade mark of Oxford University Press
Published in the United States

by Oxford University Press Inc., New York
©Paul Horwich 1990 and 1998
First edition published by Basil Blackwell Ltd. 1990

Second edition 1998 Oxford University Press
The moral rights of the author have been asserted
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.
Within the UK, exceptions are allowed in respect of any fair dealing for the
purpose of research or private study, or criticism or review, as permitted
under the Copyright, Designs and Patents Act, 1988, or in the case of
reprographic reproductions in accordance with the terms of the licences
issued by the Copyright Licensing Agency. Enquiries concerning
reproduction outside these terms and in other countries should be
sent to the Rights Department, Oxford University Press,
at the address above
This book is sold subject to the condition that it shall not, by way
of trade or otherwise, be lent, re-sold, hired out or otherwise circulated
without the publisher's prior consent in any form of binding or cover
other than that in which it is published and without a similar condition
including this condition being imposed on the subsequent purchaser
British Library Cataloguing in Publication Data

Data available
Library of Congress Cataloging in Publication Data

Data available
ISBN 0-19-875224-5
ISBN 0-19-875223-7 (Pbk.)
I 3 5 7 9 10 8 6 4 2
Typeset by Invisible Ink

Printed in Great Britain
on acid-free paper by
Biddies Ltd, Guildford and King's Lynn
To Samira
Preface to the Second Edition
The differences between the present work and the first edition of
Truth have three sources. First, some new objections to the minimal
istic theory have appeared in the last eight years, and I attempt in what
follows to reply to them, particularly to the doubts raised by Anil
Gupta, Hartry Field, Mark Richard, Donald Davidson, Bernard
Williams, Scott Soames, Michael Devitt, Crispin Wright, and Paul
Boghossian. Second, several easily correctable mistakes have come to
my attention, along with quite a few points at which my original for
mulations could be simplified and strengthened. And third, I now
have a worked-out version of the view of meaning that minimalism
requires: namely, a use theory. This point of view and its implications
have been incorporated here, although somewhat sketchily: a full ac
count of it can be found in another book, Meaning, published in con
junction with this one.
The upshot of these forces for change is that nearly every section
of the book has been revised to one degree or another, especially the
parts concerning propositional structure (Question 2), the 'liar' para
doxes (Question 1 0), the grounds for concluding that truth has no un
derlying nature (Question 1 4), the intrinsic value of truth (Question
1 9), negation (Question 25), vagueness (Question 28), utterance truth
(Question 34) and the concept of 'correspondence with facts'
(Questions 35 and 36). In addition, there is now a Postscript, which
restates the essence of deflationism about truth and responds to the
above-mentioned critics. Besides these philosophers I would also like
to thank Philippe de Rouihlon, Jesus Mosterin, Marcello Pera,
Gabriele Usberti, and Albert Visser for their suggestions about how
the first edition could be improved.
Preface to the First Edition
Perhaps the only points about truth on which most people could agree
are, first, that each proposition specifies its own condition for being
true (e.g. the proposition that snow is white is true if and only if snow
is white), and, second, that the underlying nature of truth is a mystery.
The general thrust of this book is to turn one of these sentiments
against the other. I want to show that truth is entirely captured by the
initial triviality, so that in fact nothing could be more mundane and
less puzzling than the concept of truth.
This rough idea has been expressed by a fair number of eminent
twentieth-century philosophers-including Frege, Wittgenstein,
Ramsey, Ayer, Strawson, and Quine-and I certainly claim no origi
nality for advocating it. But in spite of these impressive endorsements,
the so-called 'redundancy theory of truth' remains unpopular; and
this, I believe, is because a full case for it has never been made. The
purpose of the present essay is to fill that gap. I have tried to find the
best version of the idea-which I call 'minimalism'-give it a clear
formulation, deal with a broad range of familiar objections, and in
dicate some of its philosophical consequences. I hope that this treat
ment will help the deflationary view of truth finally gain the
acceptance that it deserves. Not only would this be good in itself, but
the effect on many surrounding issues would be quite beneficial. For
despite its reputation for obscurity the notion of truth is constantly
employed in philosophical theory. One is tempted to rely on it in de
scribing, for example, the aims of science, the relations of language to
the world, the character of sound reasoning, and the conditions for
successful planning. Yet only in light of an adequate account of truth,
and an appreciation of what the notion may and may not be used for,
can such ideas be fully understood and evaluated.
My plan is as follows. I begin (Chapter 1) by presenting the mini
malist conception, and in the following chapters I refine and defend
x Preface to the First Edition
it in the course of responding to 39 alleged difficulties (which are set
out in the Contents). In Chapter 2, I start to deal with these objec
tions, answering questions about what is required of an adequate the
ory of truth and distinguishing minimalism from other deflationary
accounts. In Chapter 3, I argue that the role of truth in laws of cog
nitive science is nothing more than a display of its minimalistic func
tion and that it should not lead us to expect there to be a theoretical
analysis of truth. I then turn to the use of the concept in philosophi
cal theory and suggest that this is often a cause of confusion: gener
ally the issues are independent of truth, and its introduction merely
muddies the waters. This is illustrated in treatments of scientific real
ism (Chapter 4) and of problems in meta-semantics and in the phi
losophy of logic (Chapter 5). My assumption throughout is that
propositions are the bearers of truth and, for those readers not com
fortable with this idea, Chapter 6 offers some arguments in its favour.
Finally, in Chapter 7, I address the feeling that truths are what corre
spond to reality, and I determine the extent to which this intuition may
be squared with the minimalist perspective.
In order to explain this conception of truth I have had to say some
thing about various other matters such as reference, meaning, belief,
logic, vagueness, realism, and the notion of proposition, and I have
sometimes taken positions in these areas without providing adequate
support for them. I hope that the reader will sympathize with the de
sire to keep this book focused and short, and will agree that the
sketchiness of some of these discussions is justified by that end.
The point of view articulated here is a development of some ideas
in my 'Three Forms of Realism' (published in Synthese in 1 982),
which was in turn a reaction to various writings of Michael Dummett
and Hilary Putnam. Although I disagree with their conclusions about
truth, I have benefited from the depth and ingenuity of their thought.
It was only against this rich background that my own contrasting po
sition could be elaborated. Another debt is to Hartry Field, with
whom I have had several conversations about truth in the last few
years. I'm afraid I don't remember if either of us ever convinced the
other of anything, but I do remember coming away from our meet
ings knowing that I had been helped a great deal. Anyone interested
in the issues addressed here should read his essay, 'The Deflationary
Conception of Truth' . In addition I would like to thank Ned Block,
who saw many of my drafts and, as always, supplied lots of reason
able advice; Marcus Giaquinto, who never quite accepts anything I
Preface to the First Edition xi
say and thereby gets me to think of better arguments; George Boolos
and Dick Cartwright, who helped me grapple with the foundations of
logic and the early views of Moore and Russell; Jerry Katz, Tom
Kuhn, and Massimo Piattelli-Palmarini, who pushed me to sort out
my thoughts about propositions; Tyler Burge, Frank Jackson, and
Bob Stalnaker, whose various sensible observations saved me from
several wrong turns; and fellow deflationists Arthur ('The Natural
Ontological Attitude') Fine and Mike ('Do We (Epistemologists)
Need a Theory of Truth?') Williams, who are thinking and working
along similar lines and with whom I have enjoyed many fruitful and
friendly discussions. I wrote the penultimate draft of this book while
I was in France in the Autumn of 1 988, and I would like to thank the
members of the Centre de Recherche en Epistemologie Appliquee for
their hospitality, and the United States National Science Foundation
for financial support during that period. The analytic philosophers in
Paris with whom I discussed the project-especially Dick Carter,
Pierre Jacob, Frarn;ois Recanati, Dan Sperber, and Bill Ulrich�pro
vided an excellent intellectual environment, and their acute and in
formed criticism has made this essay much less unsatisfactory than it
would otherwise have been.
Contents
1 The Minimal Theory 1

A Sketch of the Minimalist Conception 1
The Space of Alternative Theories 8
Summary of Alleged Difficulties 12
The following is a list of the questions and problems regarding

minimalism to which replies and solutions will be proposed in the course
of this essay
2 The Proper Formulation 15

1 . 'Of what kinds are the entities to which truth may
be attributed?' 16
2. 'What are the fundamental principles of the mini
mal theory of truth?'
17
3 . 'It seems unlikely that instances of the equivalence
schema could possibly suffice to explain all of the
great variety of facts about truth. ' 20
4. 'The minimal theory must b e incomplete, fo r i t says
nothing about the relationships between truth and
affiliated phenomena such as verification, practical
success, reference, logical validity, and assertion. '
23
5 . 'Even i f the minimal theory is, in some sense,
"adequate" and "pure", it is nevertheless unsatis
factory, being so cumbersome that it cannot even be
explicitly formulated. '
25
xiv Contents
6. 'If there were really no more to a complete theory
of truth than a list of biconditionals like "The
proposition that snow is white is true if and only if
snow is white", then, since one could always say "p"
rather than "The proposition that p is true", it
would be inexplicable that our language should
contain the word "true": there would be no point in
having such a notion. ' 31
7. 'The minimal theory fails to specify what are meant
by attributions of truth. It fails to provide necessary
and sufficient conditions for the applicability of the
truth predicate. ' 33
8 . ' I s the minimalist conception concerned with truth
itself or with the word "true"?' 36
9. 'Even if we grant that, as predicates go, the truth
predicate is highly unusual-even if we grant that
its function is to enable us to say certain important
things while avoiding new forms of quantifi-
cation-it surely does not follow that being true is·
not a genuine property. ' 37
1 0. 'If the equivalence schema is relied on indiscrimi-
nately, then the notorious "liar" paradoxes will
result. ' 40
3 The Explanatory Role of the Concept of Truth 44

1 1 . 'Truth has certain characteristic effects and causes.
For example, true beliefs tend to facilitate the
achievement of practical goals. General laws such
as this call for explanation in terms of the nature of
truth. Therefore there must be some account of
what truth is, going beyond the minimalist story,
that provides a conceptual or naturalistic reduction
of this property. ' 44
1 2. 'Another lawlike generalization is that beliefs ob-
tained as a result of certain methods of inquiry tend
to be true. Again this suggests that the minimalist con-
ception overlooks truth's causal/explanatory nature. ' 46
Contents xv
1 3 . 'A further explanatory role fo r truth lies in the fact
that the truth of scientific theories accounts for
their empirical success. ' 48
14. 'Even if all our general beliefs about truth are de-
ducible from the minimal theory (suitably aug-
mented), this does not imply that no deeper analysis
of truth is desirable; for one might well hope to find
something that will show why it is that the equiva-
lence schema holds. ' 50
4 Methodology and Scientific Realism 52

1 5 . 'Doesn't the deflationary perspective-the renunci-
ation of a substantive notion of truth-lead in-
evitably to relativism: to the idea that there is no
such thing as objective correctness?' 52
1 6 . 'Isn't the minimalist perspective in some sense anti-
realist? Does it not deny that scientific theories are
intended to correspond to a mind-independent
world?' 53
1 7. 'Is it notobvious that the nature of truth bears di-
rectly on the structure of reality and the conditions
for comprehending it? Surely, "truth" and "reality"
are semantically inextricable from one another; so
how could one's position in the realism debate be
divorced from one's conception of truth?' 56
1 8 . 'If, as the minimal theory implies, "truth" is not defmed
as the product of ideal inquiry, why should we believe
that an ideal inquiry would provide the truth?' 60
1 9 . 'How is i t possible, given the minimal theory, for
truth to be something of intrinsic value, desirable
independently of its practical utility?' 62
20. 'How can minimalism accommodate the idea of
science progressing towards the truth?' 63
2 1 . 'From the perspective o f the minimalist conception
of truth, it is impossible to produce an adequate
justification of scientific methods. ' 64
xvi Contents
5 Meaning and Logic 68
22. 'As Davidson has argued, understanding a sen

tence, say, "Tachyons can travel back in time", is a
matter of appreciating what must be the case for the
sentence to be true-knowing its truth condition.
That is to say, one must be aware that "Tachyons
can travel back in time" is true if and only if
tachyons can travel back in time. Therefore it is not
possible to agree with the minimalist claim that this
knowledge also helps to constitute our grasp of "is
true". For in that case we would be faced with some
thing like a single equation and two unknowns.
Rather, if knowledge of the truth conditions of
"Tachyons can travel back in time" is to constitute
our understanding that sentence, then this know
ledge would presuppose some pre-existing concep-
tion of truth. ' 68
2 3 . 'What about falsity and negation?' 71
24. 'As Frege said, logic is the science of truth; so surely
our accounts of truth and logic should be, if not
identical, at least bound up with one another. Yet the
minimal theory does not even enable one to prove that
the principle of non-contradiction is true. ' 73
25. 'Minimalism cannot be squared with the role that
the notion of truth must play in the foundations of
logic-in justifying one logic over another. ' 74
26. 'How can truth-value gaps be admitted?' 76
27. 'Doesn't philosophy require truth-value gaps in
order to accommodate such phenomena as non
referring names, vagueness, the emotivist concep-
tion of ethics, etc.?' 78
28. 'It is obvious that many predicates--for example,
"blue", "small", "bald", "heap" ---d o not have
definite extensions; and when such predicates are ap
plied to certain objects the result will surely be propo-
sitions with no truth value. ' 78
Contents xvii
29. 'There is a substantive issue in meta-ethics as to
whether evaluative utterances purport to assert
truths or whether they are merely expressions of
feeling; but this question would be trivialized by
minimalism.' 84
6 Propositions and Utterances 86

30. 'Propositions are highly dubious entities. It is un-
clear what they are supposed to be, and their very
existence is controversial. Would it not be better,
therefore, to develop a theory of truth that does not
presuppose them-by assuming, for example, that
utterances are the primary bearers of truth?' 86
3 1 . 'The case for propositions assumes the adequacy of
a certain logical analysis of belief-one that con-
strues the state of belief as a relation between a per-
son and a kind of entity, the content of the belief.
But this assumption is plagued with familiar
difficulties and appears to be mistaken. ' 90
32. "'The proposition that p is true iffp" can be thought
to capture our conception of truth only if truth is
not already presupposed in the very idea of a
proposition. But this requirement may well be vio-
lated. For a central component of the notion of
proposition is lodged in the statement of identity
conditions for propositions-the conditions for two
utterances to express the same proposition. But this
is an idea one might plausibly explain in terms of
the intertranslatability of the utterances, which, in
turn, must be construed as their having the same
truth conditions. And if the concept of truth is
needed to say what propositions are, then a theory
of truth cannot take propositions for granted. ' 93
3 3 . 'The "use theory" o f meaning implies that propo-
sitions don't exist. For if translation is a matter of
resemblance in use, then it is not a transitive rela-
tion, and so there can be no such things as "what
intertranslatable utterances have in common". ' 95
xviii Contents
34. 'Many philosophers would agree that if proposi
tions exist then propositional truth would be cov
ered by something like the equivalence schema. But
they might still maintain that the truth of an utter
ance consists in its "correspondence with reality",
or some other substantive thing. Thus, it is for ut
terances that the deflationary account is controver
sial, and this position has received no elaboration
or defence. ' 98
7 The 'Correspondence' Intuition 1 04

35. 'Is it Dot patently obvious that the truth or falsity of
a statement is something that grows out of its rela-
tions to external aspects of reality?' 1 04
36. 'Is it not equally clear that, contrary to minimalism,
statements are made true by facts to which they cor-
respond?' 105
37. 'Certain cases of representation (e. g. by maps)
clearly involve a correspondence-a structural
resemblance-to what is represented. So is it not
reasonable to expect some such relation in linguis-
tic representation also?' 108
3 8 . 'The minimal theory fails to show how the truth of
a sentence depends on the referential properties of
its parts. ' 1 10
39. 'The great virtue of defining truth in terms of ref
erence is that the account may be supplemented
with a naturalistic (causal) theory of the reference
relation to yield, in the end, a naturalistic and scien-
tifically respectable theory of truth. ' 1 13
Conclusion 1 18
Postscript 1 20
Bibliography 1 47
Index 1 55
1
The Minimal Theory
A Sketch of the Minimalist Conception
'What is truth?' , we sometimes ask-but the question tends to be

rhetorical, conveying the somewhat defeatist idea that a good answer,
if indeed there is such a thing, will be so subtle, so profound, and so
hard to find, that to look for one would surely be a waste of time. The
daunting aura of depth and difficulty which surrounds this concept is
perfectly understandable. For on the one hand the notion of truth
pervades philosophical theorizing about the basic nature and norms
of thought and action--e.g. ' truth is the aim of science'; ' true beliefs
facilitate successful behaviour'; ' truth is preserved in valid reasoning' ;
'to understand a sentence i s t o know which circumstances would
make it true'; 'evaluative assertions can be neither true nor false' . So
insight into the underlying essence of truth promises, by helping us
assess and explain such principles, to shed light on just about the
whole of our conceptual scheme. But, on the other hand, this very
depth can suggest that in inquiring into the nature of truth we have
run up against the limits of analysis; and, indeed, it will be widely
agreed that hardly any progress has been made towards achieving the
insight we seem to need. The common-sense notion that truth is a
kind of 'correspondence with the facts' has never been worked out to
anyone's satisfaction. Even its advocates would concede that it
remains little more than a vague, guiding intuition. But the tradi
tional alternatives-equations of truth with 'membership in a coher
ent system of beliefs', or 'what would be verified in ideal conditions',
or 'suitability as a basis for action'-have always looked unlikely to
work, precisely because they don't accommodate the 'correspon
dence' intuition, and this air of implausibility is substantiated in
straightforward counterexamples. Hence the peculiarly enigmatic
2 The Minimal Theory
character of truth: a conception of its underlying nature appears to
be at once necessary and impossible.
I believe that this impression is wholly wrong and that it grows out
of two related misconceptions: first, that truth has some hidden
structure awaiting our discovery; and secondly, that hinging on this
discovery is our ability to explain central philosophical principles
such as those just mentioned, and thereby to solve a host of problems
in logic, semantics, and epistemology.
The main cause of these misconceptions, I suspect, is linguistic
analogy. Just as the predicate 'is magnetic' designates a feature of the
world, magnetism, whose nature is revealed by quantum physics, and
'is diabetic' describes a group of phenomena, diabetes, characteriz
able in biology, so it seems that 'is true' attributes a complex property,
truth-an ingredient of reality whose underlying structure will, it is
hoped, one day be revealed by philosophical or scientific analysis.
The trouble is that this conclusion-which we tend to presuppose in
the question, 'What is truth?'-is unjustified and false. An expression
might have a meaning that is somewhat disguised by its superficial
form-tending, as Wittgenstein warned, to produce mistaken analo
gies, philosophical confusion, and insoluble pseudo-problems. The
word 'exists' provides a notorious example. And we are facing the
same sort of thing here. Unlike most other predicates, 'is true' is not
used to attribute to certain entities (i.e. statements, beliefs, etc.) an
ordinary sort of property-a characteristic whose underlying nature
will account for its relations to other ingredients of reality. Therefore,
unlike most other predicates, 'is true' should not be expected to par
ticipate in some deep theory of that to which it refers-a theory that
articulates general conditions for its application. Thus its assimila
tion to superficially similar expressions is misleading. The :role of
truth is not what it seems.
In fact, the truth predicate exists solely for the sake of a certain
logical need. On occasion we wish to adopt some attitude towards a
proposition-for example, believing it, assuming it for the sake of
argument, or desiring that it be the case-but find ourselves thwarted
by ignorance of what exactly the proposition is. We might know it
only as 'what Oscar thinks' or 'Einstein's principle'; perhaps it was
expressed, but not clearly or loudly enough, or in a language we don't
understand; or-and this is especially common in logical and philo
sophical contexts-we may wish to cover infinitely many proposi
tions (in the course of generalizing) and simply can't have all of them
The Minimal Theory 3
in mind. In such situations the concept of truth is invaluable. For it
enables the construction of another proposition, intimately related
to the one we can't identify, which is perfectly appropriate as the
alternative object of our attitude.
Consider, for example,
(1) What Oscar said is true.
Here we have something of the form
(2) x is F,
whose meaning is such that, given further information about the
identity of x-given a further premise of the form
(3) x = the proposition that p
-we are entitled to infer
(4) p.
And it is from precisely this inferential property that propositions

involving truth derive their utility. For it makes them, in certain cir
cumstances, the only appropriate objects of our beliefs, suppositions,
desires, etc. Suppose, for example, I have great confidence in Oscar's
judgement about food; he has just asserted that eels are good but I
didn't quite catch his remark. Which belief might I reasonably
acquire? Well, obviously not that eels are good. Rather what is
needed is a proposition from which that one would follow, given
identification of what Oscar said-a proposition equivalent to
(1 * ) If what Oscar said is that eels are good then eels are good,
and if he said that milk is white then milk is white, . . . and so
on;
and the raison d'etre of the concept of truth is that it supplies us with
such a proposition: namely (1 ).
To take another example, suppose we wish to state the logical law
of excluded middle:
( 5) Everything is red or not red, and happy or not happy, and
cheap or not cheap, . . . and so on.
Our problem is to find a single, finite proposition that has the intu
itive logical power of the infinite conjunction of all these instances;
and the concept of truth provides a solution.
( 6) Everything is red or not red
is known to be equivalent to
( 6*) The proposition that everything is red or not red is true.
And similarly for the other instances. Thus the infinite series of uni
versal disjunctions may be transformed into another infinite series of
claims in which the same property, truth, is attributed to all the mem
bers of a class of structurally similar propositional objects. And in
virtue of that form the sum of these claims may be captured in an
ordinary universally quantified statement:
(5*) Every proposition of the form: (everything is F or not F) is
true.
It is in just this role, and not as the name of some baffling ingredient
of nature, that the concept of truth figures so pervasively in philo
sophical reflection. 1
What permits the notion of truth to play that role is simply that,
for any declarative sentence
1 Notice that one could design an alternative way of putting the things that we
actually express by means of the truth predicate. With the introduction of sentence
variables, predicate variables, and substitutional quantification, our thoughts could be
expressed awkwardly as follows:
( 1 * *) For any sentence such that Oscar claimed that it, then it,
or in logical notation
(1 * * *) (p)(Oscar claimed that p � p);
and
(5**) Given any predicate, a thing is either it or not it,
or
(5***) (F)(x)(Fx v -Fx) .
However, the variables 'it', 'p', and 'F ' are not the usual kind which replace noun
phrases and refer to objects. Rather, 'F' must be construed as a 'pro-predicate', and 'p'
as a 'pro-sentence' . Moreover, generalization with respect to these variables cannot be
understood in the usual way as saying that every object has a certain property, but
must be construed as asserting the truth of every legitimate substitution instance. Thus
(1 * *) means intuitively that any result of substituting an English declarative sentence
for 'p' in 'Oscar claimed that p � p' is true.
The advantage of the truth predicate is that it allows us to say what we want with
out having to employ any new linguistic apparatus of this sort. It enables us to achieve
the effect of generalizing substitutionally over sentences and predicates, but by means
of ordinary variables (i.e. pronouns), which range over objects. See Chapter 2, the
answer to Question 6, for discussion of this point.
(4) p
we are provided with an equivalent sentence

(4 * ) The proposition that p is true,
where the original sentence has been converted into a noun phrase,
'The proposition that p', occupying a position open to object vari
ables, and where the truth predicate serves merely to restore the struc
ture of a sentence: it acts simply as a de-nominalizer. In other words,
in order for the truth predicate to fulfil its function we must acknowl
edge that
(MT) The proposition that quarks really exist is true if and only
if quarks really exist, the proposition that lying is bad is
true if and only if lying is bad, . . . and so on,
but nothing more about truth need be assumed. The entire conceptual
and theoretical role of truth may be explained on this basis. This
confirms our suspicion that the traditional attempt to discern the
essence of truth-to analyse that special quality which all truths sup
posedly have in common-is just a pseudo-problem based on syntac
tic overgeneralization. Unlike most other properties, being true is
unsusceptible to conceptual or scientific analysis. No wonder that its
'underlying nature' has so stubbornly resisted philosophical elabora
tion; for there is simply no such thing.
This sort of deflationary picture is attractively demystifying. 2
Nevertheless, it has not been widely accepted, for it faces a formida
ble array of theoretical and intuitive objections. My aim in this book
is to work out a form of the approach that is able to deal with all the
alleged difficulties. Some of them expose genuine deficiencies in cer
tain versions of the doctrine and reveal the need for a better formula
tion of the deflationary position. But most of the complaints have
simply been given more weight than they deserve. Indeed, I tend to
think that the approach has been underrated more because of the
sheer number of objections to it than because of their quality. Put in
2 More or less deflationary views about truth are endorsed and defended (in vari
ous forms and to various degrees) by Frege ( 1 8 9 1 , 1 9 1 8), Ramsey ( 1 927), Ayer ( 1 935),
Wittgenstein ( 1 922, 1 953), Strawson ( 1 9 5 0), and Quine ( 1 970) . In recent years the idea
has been developed by Grover, Camp, and Belnap ( 1 975), Leeds ( 1 978), the present
author ( 1 982a), A. Fine ( 1 984), Soames ( 1 984, 1 997), Field ( 1 986, 1 994), M. J.
Williams ( 1 986), Loar ( 1 987), Baldwin ( 1 989), and Brandom ( 1 988, 1 994).
more positive terms, my plan is to provide a highly deflationary
account of our concept of truth-but one that can nevertheless
explain the role of the notion in scientific methodology and in science
itself, and enable us to find answers to such questions as: In what does
our grasp of truth consist? Why is it practically useful to believe the
truth? Can there be, in addition, any purely intrinsic value to such
beliefs? Does science aim and progress towards the truth? How does
our conception of truth bear on the nature of various types of fact
and on our capacity to discover them? Is truth an explanatorily vital
concept in semantics or in any of the empirical sciences? I shall start
by giving what I believe is the best statement of the deflationary point
of view. Because it contains no more than what is expressed by
uncontroversial instances of the equivalence schema,
(E) It is true that p if and only if p,
I shall call my theory of truth ' the minimal theory', and I shall refer to
the surrounding remarks on behalf of its adequacy as ' the minimalist
conception ' . With a good formulation in hand, I want to show that
the standard criticisms of deflationary approaches are either irrele
vant or surmountable, to display the virtues of the theory in compar
ison with alternatives, and, by answering the above questions, to draw
out the implications of minimalism for issues in semantics, psychol
ogy, and the philosophy of science. For the sake of simplicity and
conformity with natural language I begin by developing the account
of truth for propositions. However, I shall go on to argue that the min
imalist conception applies equally well to the 'truth' of utterances,
mental attitudes, and other types of entity.
It might be thought that minimalism is too obvious and too weak
to have any significant philosophical implications. Let me try, in at
least a preliminary manner, to quell this misgiving. The real proof, of
course, will be in the execution of the project. We should start by dis
tinguishing (very roughly) two types of 'philosophical implication'
that might be expected. First, there are general principles involving
truth: for example, the fact that verification indicates truth, and that
true beliefs are conducive to successful action. And, secondly, there
are solutions to philosophical problems such as the paradoxes of
vagueness and the issue of scientific realism. According to the mini
malist conception, the equivalence schema, despite its obviousness
and weakness, is not too weak to have significant philosophical
implications-at least within the first category. On the contrary, our
thesis is that it is possible to explain all the facts involving truth on the
basis of the minimal theory. This may indeed appear to be a rather
tall order. But remember that most of the interesting facts to be
explained concern relations between truth and certain other matters;
and in such cases it is perfectly proper to make use of theories about
these other matters, and not to expect all the explanatory work to be
done by the theory of truth in isolation. When this methodological
point is borne in mind it becomes more plausible to suppose that the
explanatory duties of a theory of truth can be carried out by the min
imal theory.
As for the second class of 'philosophical implication'-namely,
solutions to problems-one would expect these to flow, not from the
minimal theory as such (i.e. instances of the equivalence schema), but
rather from the minimalist conception (i.e. the thesis that our theory
of truth should contain nothing more than instances of the equiva
lence schema). Philosophical questions are typically based on confu
sion rather than simple ignorance. Therefore an account that makes
plain the character of truth will permit a clearer view of any prob
lems that are thought to involve truth. The account itself may well
never entail, or even suggest, any solutions. But, in so far as it eluci
dates one of the sources of confusion, it will help us to untangle the
conceptual knots that are generating the problems, and thereby facil
itate their solution. In the limiting case, a conception of truth can
achieve this result by enabling us to see that, contrary to what has
been generally presupposed, the notion of truth is not even involved
in the problem. The recognition that truth plays no role can be vital
to achieving the clarity needed for a solution. Thus, to put the matter
somewhat paradoxically, the relevance of a theory of truth may lie in
its import regarding the irrelevance of truth. We shall see, I think,
that this is very often the situation. Consider, for example, the debate
surrounding scientific realism. It is commonly assumed that truth is
an essential constituent of the problem; one sees reference to 'realist
conceptions of truth' and to 'anti-realist conceptions of truth'; and
questions about the meaning of theoretical assertions, our right to
believe them, and what it would be for them to be true, are all lumped
together as components of a single broad problem. This intertwining
of philosophically puzzling notions is why the realism issue has
proved so slippery and tough. What I am claiming on behalf of the
minimalist conception of truth is not that it, by itself, will engender
realism or anti-realism; but rather that it will make it easier for us to
see that the central aspects of the realism debate have nothing to do
with truth. By providing this clarification of the main problems, min
imalism will take us a long way toward being able to solve them.
The Space of Alternative Theories
It will help us to focus on what is at stake in accepting the minimalist

conception of truth if I contrast it with some of the well-known
alternatives.
Correspondence
First, there is the venerable notion that truth is the property of corre
sponding with reality. In its most sophisticated formulations this has
been taken to mean that the truth of a statement depends on how its
constituents are arranged with respect to one another and which enti
ties they stand for. One strategy along these lines (Wittgenstein, 1 922)
is to suppose that a statement as a whole depicts a fact whose con
stituents are referents of the statement's constituents, and that the
statement is true if and only if such a fact exists. Another strategy
(Austin, 1 950; Tarski, 1 958; Davidson, 1 969) is to define truth in
terms of reference and predicate-satisfaction without importing the
notions of fact and structure. Either way, these correspondence the
ories further divide according to what is said about reference. For
example, one might suppose, with Wittgenstein ( 1 922), that it is sim
ply ip.describable; or, with Field ( 1 972) and Devitt ( 1 984), that refer
ence is a naturalistic (causal) relation; or, with Quine ( 1 970) and
Leeds ( 1 978), that it is merely a device for semantic ascent. From our
minimalist point of view, the last of these ideas is along the right
lines-reference and truth being parallel notions-although, as we
shall see, it is a mistake to explain truth in terms of reference.
Coherence
The second most popular view of truth is known as the coherence

theory. A system of beliefs is said to be coherent when its elements
are consistent with one another and when it displays a certain overall
simplicity. In that case, according to the coherence theory, the whole
system and each of its elements are true. Thus truth is the property of
belonging to a harmonious system of beliefs. This line was urged by the
idealists Bradley ( 1 9 1 4) and Blanshard ( 1 939), embraced by Hempel
( 1 935) as the only alternative to what he regarded as the obnoxious
metaphysics of correspondence, and resurrected for similar reasons
by Dummett ( 1 978) and Putnam ( 1 98 1 ) (as the 'verificationist' or
'constructivist' theory) in their identification of truth with idealized
justification. What has struck most philosophers as wrong with this
point of view is its refusal to acknowledge what would appear to be a
central feature of our conception of truth, namely the possibility of
there being some discrepancy between what really is true and what we
will (or should, given all possible evidence) believe to be true. For it
seems quite conceivable that there are facts we are not capable of dis
covering (for example, that there are infinitely many stars, or that
every even number is the sum of two primes); conversely, our evidence
might on occasion point incontrovertibly toward some conclusion
that happens to be false; so truth and verification are not the same.
Pragmatism
In the third place we have the so-called pragmatic theory of truth

devised by James ( 1 909) and Dewey ( 1 938), and recently elaborated
by Rorty ( 1 982) and Papineau ( 1 987). Here truth is utility; true
assumptions are those that work best-those which provoke actions
with desirable results. However, just as in the case of the coherence
theory, one must guard against taking a strong correlation for an
identity. Although there is indeed an association between the truth of
a belief and its tendency to facilitate successful activity, the tightness
of the association should not be exaggerated. After all, actions based
on true belief can none the less work out badly. Moreover, the link, to
the extent that it holds, is surely something to be explained, not
merely stipulated in the very definition of truth.
Unanalysable quality
Fourthly-perhaps the least attractive conclusion-there is the one

time thesis of Moore ( 1 899, 1 9 1 0/1 1 ) and Russell ( 1 904) that truth is
an indefinable, inexplicable quality that some propositions simply
have and others simply lack-a fundamental property of which no
account can be provided. 3 This gives a sense of impenetrable myste
riousness to the notion of truth and can be the resort only of those
who feel that the decent alternatives have been exhausted.
These traditional approaches do not typically impugn the correct

ness of the equivalence schema,
(E) (p) is true iff p,4
but question its completeness. They deny that it tells us about the
essential nature of truth, and so they inflate it with additional con
tent in ways that, I will argue, are, at best, unnecessary and, at worst,
mistaken. To explain this point a little further it is useful to imagine
various dimensions on which alternative accounts of truth may be
characterized-each dimension varying with respect to some form of
theoretical commitment.
1 . An account may or may not be compositional-it may or may not

define the truth of an utterance or a proposition in terms of the
semantic properties of its parts. For example, a theory inflated in this
way might involve the principle,
(T/R) 'a is F' is true iff there exists an object x such that ' a' refers
to x and 'F' is satisfied by x.
The minimalist policy is not to deny such principles relating truth,
reference, and satisfaction, but to argue that our theory of truth
should not contain them as axioms. Instead, they should be derived
from a conjunction of the theory of truth and quite distinct mini
malist theories of reference and satisfaction.
2. One may or may not insist on a conceptual analysis of truth, a

specification (in philosophically unproblematic terms) of the content
of every statement employing the concept. Minimalism offers no
such definition, and denies the need for one.
3 For an examination of this view as it appears in the early writings of Moore and
Russell see Cartwright ( 1 987).
4 I shall write '(p)' for 'the proposition that p', and 'iff' for 'if and only if'.
3 . An account may or may not suppose that there is some substan
tive, reductive theory of truth-some non-definitional analysis spec
ifying the underlying property in which truth consists. In the context
of a compositional account, the parallel issue is whether reference
and satisfaction are complex relational properties-according to
some philosophers, reducible to certain causal notions. The minimal
ist denies that truth, reference, or satisfaction are complex or natu
ralistic properties.
4. One may or may not attempt to formulate a non-trivial, finite the

ory of truth itself-a succinct body of statements about truth that
can be tacked on to our other theories (in physics, mathematics, etc.)
to enable the deduction of everything we believe about truth.
According to minimalism, there is no such thing. We can say what is
in the basic theory of truth-an infinity of biconditionals of the form
'(p) is true iff p'-but we can't formulate it explicitly because there are
too many axioms.
5. One may or may not propose an account which inextricably links

truth with other matters: for example, assertion, verification, refer
ence, meaning, success, or logical entailment. Minimalism involves
the contention that truth has a certain purity-that our understand
ing of it is fairly independent of other ideas.
Thus my account will take the less theoretically loaded view with
respect to each of these dimensions of commitment. The theory of
truth it proposes involves nothing more than the equivalence schema;
it is non-compositional; it denies that truth and reference are com
plex or naturalistic properties; and it does not insist on an eliminative
account of truth attributions. In this way minimalism aims for a max
imally deflationary theory of truth, which, though complete, has no
extraneous content-a theory about truth, the whole of truth, and
nothing but truth.
I should stress that the minimalist critique of the correspondence,
coherence, constructivist, pragmatist, and primitivist accounts of
truth is not that they are false. On the contrary, it seems quite likely
that carefully qualified, true versions of each of them could be con
cocted. The main objection is rather that none can meet the explana
tory demands on an adequate theory of truth. Specifically, none
provides a good account of why it is that instances of the equivalence
schema are true. Minimalism involves a reversal of that explanatory
direction. We shall find that on the basis of the equivalence bicondi
tionals it is easy to see why, and in what form, the traditional princi
ples hold. Indeed, every fact about truth can be naturally derived
from those biconditionals. Therefore it is they that should constitute
our basic theory of truth.
Summary of Alleged Difficulties
Objections to deflationary approaches have concerned six related

topics.
The Proper Formulation. It has been no easy matter to provide even a

prima facie plausible version of such a theory of truth-something
that meets the normal methodological standards of fidelity to obvi
ous fact, simplicity, explanatory power, etc., and that is not falsified
by the 'liar' paradoxes.
The Explanatory Role of the Concept of Truth. The concept of truth

is apparently employed in certain forms of scientific explanation (e. g.
to help account for the contribution of language use to the achieve
ment of practical goals), and it has been argued on this basis that
deflationism must be missing something-namely, the naturalistic
character that provides truth with its causal properties.
Methodology and Scientific Realism. A natural (realist) view of sci

ence is that it aims for, and gradually progresses towards, the truth
a goal that exists independently of our capacity to reach it, and that
we value partly for its own sake, independently of any practical
benefits that might accrue. This position would seem to require a sub
stantial notion of truth-a conception of just the sort that the
deflationary point of view eschews. In other words, any deflationary
account of truth would seem to entail an anti-realist perspective on
science.
Meaning and Logic. A further body of objections concerns the role of

truth in semantics, and the ability of any deflationary approach to
explain this role. For example, it is usual to analyse understanding in
terms of knowledge of truth conditions, to use the concepts of truth
and reference to show how the meanings of sentences depend on the
meanings of their parts, to suppose that truth must be a central con
cept in the appraisal of alternative rules of inference, and to treat var
ious semantic phenomena (e.g. vagueness, empty names, expressive
utterances) by exploiting the idea that a proposition might be neither
true nor false. It is commonly assumed that deflationary theories of
truth are precluded by these demands.
Propositions and Utterances. Propositions can appear to be such

obscure and bizarre entities that it may seem undesirable to base an
account of truth on the schematic principle
(E) The proposition (p) is true iff p
which presupposes them. At the same time, the natural deflationary

account of truth for utterances, the disquotational schema
(D) Any utterance of the sentence 'p ' is true iff p,
has difficulty with indexical expressions (try 'I am hungry'), foreign

languages (' Schnee ist weiB '), and indeed with all sentence-tokens
whose truth or falsity depends on the context in which they are pro
duced.
The 'Correspondence ' Intuition. The idea that a representation is

made true by its correspondence to reality has great intuitive appeal,
yet there appears to be no room for any such conception within the
deflationary picture.
Each of these topics receives separate treatment in the following

chapters. However, they need not be taken in their order of appear
ance. Readers uncertain about what exactly the minimal theory is
should certainly not miss Chapter 2. But otherwise one can proceed
directly to Chapter 3 where some influential arguments against mini
malism are rebutted and a case in favour of it is made, or Chapters 4
and 5 where its implications are examined. Anyone wary of proposi
tions should not delay long before looking at Chapter 6 where I hope
their concerns will be assuaged. And those philosophers who are
fond of the correspondence theory of truth should perhaps not wait
until the end before reading Chapter 7 and seeing that most of their
intuitions may be accommodated.
I have organized the above-mentioned areas of criticism into 39
specific questions and objections. In what follows I shall articulate
these problems in more detail and, in each case, sketch what I think is
an adequate response. What will emerge, I hope, is a view of truth
that is clear, plausible, and fairly comprehensive.
2
The Proper Formulation
The conception of truth to be defended in this essay is similar

in spirit to other deflationary accounts that have appeared during
the past hundred years or so, maintaining, in one way or another,
that truth is not a normal property and that traditional investiga
tions into its underlying nature have been misconceived. None of
these accounts, however, has won over very many adherents, and the
vast majority of philosophers either still subscribe to some form
of correspondence, coherence, pragmatist, or primitivist picture,
or else think that no decent theory has yet been made available.
One cause of dissatisfaction with deflationary proposals in the
literature is that they are not described fully or precisely enough
to be properly evaluated. For instance, it isn't always said whether
the theory concerns the nature of truth itself, or merely the mean
ing of the word 'true'. Secondly, and exacerbating the evaluation
problem, there is a tendency to omit explicit statement of what a
satisfactory account is supposed to do. The adequacy conditions
for a theory of truth are left unclear. A third common defect of
deflationary views is their commitment to certain blatantly
implausible theses: for example, that being true is not a property
at all, or that every instance of ' "p" is true iff p' is correct. And,
in the fourth place, objections are often left standing that could
in fact be rebutted: for example, that the theory fails to say what
truth is, and that it cannot be reconciled with the desirability of
truth. The purpose of this chapter is to reach an exact charac
terization of the minimalist conception and, whilst doing so, to show
how to deal with some of the problems that have notoriously
afflicted previous deflationary proposals.
16 The Proper Formulation
1. Of what kinds are the entities to which truth may be

attributed?
The list of candidates includes: (a) utterances-individual sounds

and marks located in particular regions of space and time (e.g.
Oscar's saying the words 'I am hungry' at midday on 1 January
1 988); (b) sentences-types of expression in a language: syntactic
forms that are exemplified by particular utterances (e.g. the English
sentence 'I am hungry'); (c) statements, beliefs, suppositions, etc.
individual, localized actions or states of mind (e.g. Oscar's state at
midday of believing that he is hungry); (d) propositions-the things
that are believed, stated, supposed, etc.; the contents of such states
(e.g. that Oscar was hungry at midday on1January1988). I shall fol
low ordinary language in supposing that truth is a property of
propositions. 1 Thus, if we agree with Oscar, we attribute truth to
what he said, to the proposition he asserted. Evidently the sentence
type of English that he used is not true; for that very sentence-type
is used on other occasions to make false statements. Nor would one
normally characterize the noises he made, or his belief state, as true.
These entities are more naturally described as 'expressing a truth'
and 'being of a true proposition' . No doubt we do attribute truth to
statements, beliefs, suppositions, and so on; but surely what we have
in mind is that the propositional objects of these linguistic and men
tal acts are true, and not the acts themselves.
Most of the time I will conform to this way of speaking. To some
extent this decision is non-trivial; for it involves a commitment to
the existence of a breed of things called 'propositions'. However this
commitment, though controversial and in need of some defence (to
be supplied in Chapter 6), is much less substantial than it might seem
at first. For it presupposes very little about the nature of proposi
tions. Granted, minimalism entails that the notion of proposition
not depend on the notion of truth. For the minimalist's direction of
conceptual priority is the other way round: in so far as our concept
1 In light of the locution 'It is true that p' , it might be thought that a theory of
the truth predicate would have to be supplemented with a separate theory of the truth
operator; but this is not so. We can construe 'It is true that p' , on a par with 'It is true
what Oscar said', as an application of the truth predicate to the thing to which the
initial 'It' refers, which is supplied by the subsequent noun phrase, 'that p'.
The Proper Formulation 17
of truth is constituted by our acceptance of instances of 'The propo
sition that p is true iff p', we must already be capable of grasping
propositions. But this requirement leaves open many possibilities. As
far as the minimal theory of truth is concerned, propositions could
be composed of abstract Fregean senses, or of concrete objects and
properties; they could be identical to a certain class of sentences in
some specific language, or to the meanings of sentences, or to some
new and irreducible type of entity that is correlated with the mean
ings of certain sentences. I am not saying that there is nothing to
choose amongst these answers. The point is rather that the minimal
theory does not require any particular one of them. So that some
one who wishes to avoid commitment to 'propositions' of any
specific sort need not on that score object to the conception of truth
that will be elaborated here.
Moreover, the view that truth is not strictly speaking attributable
to utterances, or to linguistic or mental acts, is not substantial, and
nothing of importance in what follows will depend on it. If some
one holds that an utterance may be 'true', in a certain sense, then he
can simply regard my claims about the property of expressing truth
as claims about 'truth' in his sense. Similarly for those who think
that a truth predicate may be applied to acts of asserting, states of
believing, etc.
2. What are the fundamental principles of the minimal

theory of truth?
The axioms of the theory are propositions like

(1) ((Snow is white) is true iff snow is white)
and
(2) ((Lying is wrong) is true iff lying is wrong);
that is to say, all the propositions whose structure is
(E*) ((p) is true iff p). 2
2 This claim will be modified slightly in the answer to Question 1 0 in order to

accommodate the 'liar' paradoxes.
In order to arrive at this 'propositional structure' we can begin
with any one of the axioms and note that the sentence expressing it
may be divided into two complex constituents. First there is a part
that is itself a sentence and which appears twice. In the case of ( 1 ),
this is
(3) ' snow is white' .
And second there i s the remainder o f the axiom-formulation
namely, the schema
(E) '(p) is true iff p'.
Now we assume that if a complex expression results from the appli
cation of a schema to a sequence of terms, then the meaning of the
expression is the result of applying the meaning of the schema to
the sequence of the terms' meanings. In particular, since, as we have
just seen, the sentence
(1 *) '(Snow is white) is true iff snow is white'
is the result of applying (E) to (3), then the proposition expressed
by (1 *) is the result of applying what is expressed by schema (E) to
what is expressed by (3). That is to say, the axiom ( 1 ) is the result of
applying the propositional structure (E*) to the proposition
(3*) (snow is white).
Similarly, if (E*) is applied to the proposition
(5) (Lying is wrong),
it yields
(2) ((Lying is wrong) is true iff lying is wrong).
Indeed, when applied to any proposition, y, this structure (or func
tion) yields a corresponding axiom of the minimal theory, MT. 3
3 I am employing the convention that surrounding any expression, e, with angled

brackets, '( and ') , produces an expression referring to the propositional constituent
' '
expressed by e.
It might be argued that the two sentence-tokens in each MT axiom-formulation
do not have the same content as one another, since the first occurs in an opaque con
text (after 'The proposition that . . ) and the second does not, and, consequently,
.'
In other words, the axioms o f M T are given by the principle
(5) For any object x: x is an axiom of the minimal theory if
and only if, for some y, when the function E* is applied to
y, its value is x.
Or in logical notation
that these axioms cannot really be regarded as the various results of applying a
single one-place function (the propositional structure) to the various propositions. If
this is correct (which I doubt) then we must proceed differently.
One alternative is to characterize the axioms of the minimal theory as anything
that is expressed by instances of the sentence schema
(E) '(p) is true iff p'.
However, the theory cannot be restricted to instantiations of (E) by English sentences;
for presumably there are propositions that are not expressible in current English, and
the question of their truth must also be covered. So further 'equivalence axioms' are
needed, one for each unformulatable proposition.
Although we cannot now articulate these extra axioms (any more than we can
articulate the propositions they are about), we can nevertheless identify them. One
way of doing this is by reference to foreign languages. We can suppose that the the
ory of truth includes whatever is expressed by instances of translations of the equiv
alence schema: e.g. instantiations of
(E-f) '(p) est vrai ssi p'
by French sentences, instantiations of
(E-g) '(p) ist wahr gdw p'
by German sentences, and so on, for all languages. If it were assumed that every
proposition is expressed in some language, then this would do. But we want to allow
for the existence of propositions that are not yet expressible at all. To accommodate
these we might suppose that every proposition, though perhaps not expressed by any
actual sentence, is at least expressed by a sentence in some possible language. And we
can then regard the theory of truth as whatever would be expressed by instances of
translations of the equivalence schema into possible languages.
However, once the need to refer to possible languages has been acknowledged, we
can see that there was no reason to have brought in actual foreign languages. For we
can make do with our own language supplemented with possible extensions of it. In
other words, we can characterize the 'equivalence axioms' for unformulatable propo
sitions by considering what would result if we could formulate them and could instan
tiate those formulations in our equivalence schema. Thus we may specify the axioms
of the theory of truth as what are expressed when the schema
(E) '(p) is true iff p'
is instantiated by sentences in any possible extension of English.
Alternatively, instead of identifying the axioms indirectly in terms of how they
would be expressed, we can solve the problem by directly specifying the propositional
structure which all and only the axioms have in common. This is the strategy adopted
in the text.
(5*) (x)(x is an axiom of MT H (3y)(x = E* (y))) . 4
The minimal theory has several striking features-features that
might at first be regarded as grounds for dissatisfaction with it. In
the first place it does not say explicitly what truth is: it contains no
principle of the form ' (x)(x is true iff . . . x . . . )', or 'What makes a
proposition true is its having characteristic P' . And so one might
suspect that certain general facts about truth could not be explained
by the theory. Secondly, it does not mention phenomena such as ref
erence, logical validity, assertion, and the aim of inquiry-notions
whose relation to truth one might have thought any decent theory
should describe. And, thirdly, although we have been able to char
acterize the axioms of MT (as the propositions of a certain form)
we cannot explicitly formulate the theory-for two independent
reasons. In the first place the number of axioms that we have the ter
minology to formulate is too great; there are infinitely many, and
though each one of them can be expressed, it is not possible to write
down the whole collection. In the second place there are many
propositions we cannot express in current terminology. And for
those the corresponding equivalence axioms are themselves inex
pressible-although, as we have seen, it is none the less possible to
indicate what they are.
In the following few sections we shall examine our justification
for concluding that MT is nevertheless the best theory of truth, and
we shall see why the peculiar features of the theory should not be
held against it.
3. I t seems unlikely that instances of the equivalence

schema could possibly suffice to explain a l l the
great variety of facts about truth.
The primary test of this (and any other) theory is its capacity to
accommodate the phenomena in its domain. That is to say, if our
4 Patrick Grim pointed out to me that the minimal theory cannot be regarded as
the set of propositions of the form ((p) is true iff p); for there is no such set. The
argument for this conclusion is that if there were such a set, then there would be dis
tinct propositions regarding each of its subsets, and then there would have to be
distinct axioms of the theory corresponding to those propositions. Therefore there
theory i s a good one, i t will b e able t o account fo r all the facts about
truth. Let me give three examples of the sort of explanation that
minimalism can provide.
(I) From 'What Smith said was true' and 'What Smith said was that
snow is white', it follows that 'Snow is white' . Given the minimal
theory (MT) this fact can be explained as follows:
1. What Smith said is true.
2. What Smith said = (Snow is white) .
 3. (Snow is white) is true. [from 1 ,2]
4. (Snow is white) is true iff snow is white. [MT]
 5. Snow is white. [from 3,4]5
(II) If one proposition implies another, and the first one is true, then
so is the second. Here is a minimalist explanation:
1 . Logic provides us with facts like
[dogs bark & (dogs bark � pigs fly)] � pigs fly,
that is, with every fact of the form
[p & (p � q)] � q.
would be a 1-1 function correlating the subsets of MT with some of its members.
But Cantor's diagonal argument shows that there can be no such function. Therefore,
MT is not a set. In light of this result, when we say things like '(A) follows from the
minimal theory', we must take that to mean, not that the relation of following from
holds between (A) and a certain entity, the minimal theory; but rather that it holds
between (A) and some part of the minimal theory-i.e. between (A) and some set of
propositions of the form ((,p) is true iff p).
5 In order to explain why 'Possibly, snow is white' follows from 'What Smith said
is possibly true' and 'What Smith said is that snow is white', we must assume, not
merely statement 4, but rather
Necessarily, (snow is white) is true iff snow is white.
Thus it might seem that the axioms of the theory of truth should be strengthened
and taken to consist of modal propositions of the form
(Necessarily, (,p) is true iff p).
An alternative strategy, however-and one that I prefer-is to keep the theory of
truth un-modal and simple, and instead derive the necessity of its axioms from a sep
arate theory of necessity, specifying, in general, what makes a proposition not merely
true but necessarily true. It might be supposed, for example, that the necessary truths
are distinguished by being explanatorily fundamental. In that case, given our argu
ment to the effect that MT is explanatorily basic, it would follow that its axioms are
necessary. Thus we might obtain the necessity of instances of the equivalence schema
without having to build it into the theory of truth itself.
2. Therefore, given MT, we can go on to explain every fact of
the form
[(p) is true & (p � q)] � (q) is true.
3 . But from the nature of implication, we have all instances of
((p) implies (q)) � (p � q)
4. Therefore we can explain each fact of the form
[(p) is true & (p) implies (q)] � (q) is true.
5. And therefore, given MT, we get each fact of the form
([(p) is true & (p) implies (q)] � [(q) is true]) is true.
6. But it is a peculiar property of propositions that any
general claim about them-any characterization of all
propositions-is made true by the infinite set of particular
facts associating that characteristic with each individual
proposition. 6
7 . Therefore, in light of 5 and 6, we can explain the general
fact:
Every proposition of the form, ([(p ) is true & (p)
implies (q)] � [(q) is true]), is true.
-
(III) We would be inclined to endorse the following thesis: 'If all
Bill wants is to have a beer, and he thinks that merely by nodding
he will get one, then, if his belief is true, he will get what he wants. '
This fact would be explained as follows:
We begin with the suppositions
1 . Bill wants (Bill has a beer);
2. Bill believes (Bill nods � Bill has a beer).
In addition, we can make the normal assumption (an instance of
the 'practical syllogism') about the relation between Bill's belief,
desire, and action:
3. [Bill wants (Bill has a beer) &
Bill believes (Bill nods � Bill has a beer)]
� Bill nods; [premise]
:. 4. Bill nods. [from 1 ,2,3]
6 See sect. 5 of the Postscript for further discussion of this premise.

Now let us assume, for the sake of argument,
5 . Bill's belief is true.
That is to say,
6. (Bill nods � Bill has a beer) is true. [from 2,5]
Also we have, from the theory of truth,
7. (Bill nods � Bill has a beer) is true iff
Bill nods � Bill has a beer; [MT]
:. 8 . Bill nods � Bill has a beer; [from 6,7]
:. 9. Bill has a beer. [from 4,8]
But again from the theory of truth,
1 0. (Bill has a beer) is true iff Bill has a beer; [MT]
:. 1 1 . (Bill has a beer) is true; [from 9, 1 0]
:. 1 2 . Bill gets what he wants. [from 1 , 1 1 ]
And, as we shall see in Chapter 3 (in the answer to Question 1 1 ),
this sort of explanation may be universalized to show in general how
true beliefs engender successful action.
According to the minimalist thesis, all of the facts whose expression

involves the truth predicate may be explained in such a way: namely,
by assuming no more about truth than instances of the equivalence
schema. 7 Further explanations of this sort, dealing with a range of
philosophically interesting facts about truth, will be given as we pro
ceed. These explanations will confirm the minimalist thesis that no
account of the nature of truth, no principle of the form ' (x) (x is true
iff . . . x . . . )', is called for.
4. The m i n i mal theory must be i ncomplete, for it says

nothing about the relationships between truth and
affiliated phenomena such as verification, p ractical
success, reference, logical val i dity, and assertion.
A theory of any phenomenon, X, is a collection of principles (i.e.

axioms and/or rules); and the theory is good to the extent that it
7 This is a slight exaggeration. As Anil Gupta ( 1 993b) has pointed out, the equiv
alence axioms cannot explain why, for example, Julius Caesar is not true. It would
seem that a complete theory of truth will require, in addition to MT, the axiom '(x)(x
is true � x is a proposition)' .
captures all the facts about that phenomenon in the simplest possi
ble way. It won't do merely to produce some set of important facts
about X and call that the theory. Nor would it suffice even if every
fact about X were explicitly listed. Rather, the understanding that
we want requires some account of explanatory relationships. We
have to locate the most basic facts regarding X, from which all the
others may be explained. Of course we don't expect our theory of
X to do the explanatory work all by itself. It does not follow solely
from the theory of electrons that electrons are smaller than ele
phants; we need a theory of elephants too. Our goal, then, is to find
a simple theory of X, which, together with our theories of other mat
ters, will engender all the facts.
Sometimes it will turn out, for certain phenomena, X and Y,
that we cannot separate two distinct theories, one for X and one for
Y: the simplest adequate body of principles we can find concerns
both X and Y. Consider, for example, a geometric theory about
points, lines, angles, etc. This cannot be split up into a theory of
points, a theory of lines, and so on. Sometimes we are forced to
acknowledge that certain theoretical phenomena are, in this way,
inextricably entangled with one another. And this is a significant fact
about such phenomena. But when this is not so, where distinct
theories of X and Y can be given, then they should be given.
Otherwise, a misleading illusion of interdependence is conveyed
and the cause of simplicity and explanatory insight is poorly
served.
For this reason it seems to me not merely legitimate but impor
tant to separate, if we can, what we say about truth from our theories
of reference, logic, meaning, verification, and so on. No doubt there
are interesting relationships amongst these matters. But in so far as
we want to understand truth and the other phenomena, then our task
is to explain the relationships between them and not merely to recog
nize that they exist. We must discover the simplest principles from
which they can all be deduced: and simplicity is promoted by the
existence of separate theories of each phenomenon. Therefore it is
quite proper to explain the properties of truth by conjoining the
minimal theory with assumptions from elsewhere. (Note, for
example, the use of extraneous premises in the explanations in the
previous section-drawn, in those cases, from psychology, logic, and
the theory of propositions.) The virtue of minimalism, I claim, is
that it provides a theory of truth that is a theory of nothing else, but
which is sufficient, in combination with theories of other phenomena,
to explain all the facts about truth. 8
5 . Even if the m i n i mal theory is, i n some sense,

'adeq uate' and 'pure', it is nevertheless u nsatisfactory,
being so cumbersome that it cannot even be expl icitly
formulated .
Presented with the minimal theory of truth one's first instinct, no

doubt, is to imagine that we can surely improve on it and capture
the infinity of instances of the equivalence schema in a compact for
mulation. However, there does not have to be any succinct, explana
torily adequate theory of truth, and I shall be arguing that in fact
there isn't one. Such a theory would encapsulate the properties of
truth in a finite body of principles which would generate everything
true of truth, including, at the very least, infinitely many instances
of '(p) is true iff p ' . Moreover, if it is to be explanatorily adequate,
the theory would have to subsume all these facts without the use of
notions that are themselves mysterious and unexplained. But how
might this be done?
One natural suggestion is the single principle
(6) (x)(x is true iff {3q} (x = ( q) & q)),
where the curly brackets indicate substitutional quantification over
the sentences of English. But this idea fails for a couple of reasons.
In the first place, the use of substitutional quantification does not
square with the raison d'etre of our notion of truth, which is to
enable us to do without substitutional quantification.9 In the second
place, the notion of substitutional quantification would itself
require theoretical elucidation. But what kind of elucidation could
be given? It would be circular to rely on the standard explanation
which is couched in terms of truth. 10 Alternatively, one might try to
8 Although the minimal theory characterizes truth in relation to propositions, it

is not a joint theory of these two phenomena; for we can give a prior theory of propo
sitions which makes no reference to truth (see Chapter 6). Thus truth is the only pre
viously unexplained concern of the minimal theory.
9 See Chapter 1, n. 1 , and Question 6 of the present chapter.
1 0 ' {:Jp } ( . . . p . . . ) ' means 'Some sentence formed by replacing the "p " in
"
. . . p . . . " with a sentence of (some extension of) English is true'.
characterize the notion of substitutional quantification via a
specification of the rules of inference that govern it: for example, a
version of 'universal instantiation', which is the schematic rule
{q}( . . . q . . . )
(7) .
. . .p . . .
However, this cannot be formulated as a generalization over every

sentence 'p' , viz.
{q} ( . . . q . . . )
(8) {p } .
. . .p . . .
For there would then be no way of getting from that principle to
instances such as:
{q} ( . . . q . . . )
(9) .
. . . snow is white . . .
Nor, again on pain of circularity, could we construe (7) as the claim
that every instance of the schema preserves truth. The only alterna
tive would be to recognize that the apparently single rule (7) is in
fact an infinite collection of rules; one for each 'sentence in context'
that can be put in place of ' . . . p . . . ' . But then we are embracing
an unformulable theory after all. Nothing has been gained; yet
something (i.e. the use of truth to dispense with substitutional
quantification) has been lost. Thus it seems that the best overall the
ory will not involve a definition of truth in terms of substitutional
quantification. 1 1
Another tempting approach, again designed to avoid the need for
infinitely many axioms, is to formulate the theory of truth as the sin
gle proposition
( 1 0) Every instance of ((p) is true iff p) is true.
It is clear, however, that this will not do. For it would enable us to
deduce, for example,
11
For further discussion of the policy of explaining truth in terms of substitu
tional quantification see Grover, Camp, and Belnap ( 1 975), Baldwin ( 1 989), and
Brandom ( 1 994), who embrace it, and Forbes ( 1 986), who rejects it. Similar difficul
ties beset the prospect of explaining truth in terms of other forms of quantification
into sentence positions.
(1 1) ((Snow is white) is true iff snow is white) is true.
But we would have no licence to get from there to the conclusion
that
(1) (Snow is white) is true iff snow is white.
To do this we would need the schematic rule of inference
(p) is true
(E#) ,
: p
.
which, as we have just seen, must be regarded as an infinite collec

tion of separate rules.
Inspired by Tarski ( 1 958), one might think that the solution to
our difficulty is to be found by defining the truth of a whole propo
sition in terms of the reference of its parts and how those parts are
put together. But this is a vain hope. Truth and reference are closely
affiliated notions, and so a theory that characterized truth in terms
of reference but gave no account of reference would be unsatisfac
tory. But any attempt to provide such a theory re-encounters pre
cisely the problems with which we are now struggling. For just as
the theory of truth must subsume everything like
(1) (Snow is white) is true iff snow is white,
so any decent theory of reference would have to subsume the fact
that
( 1 2) The propositional constituent expressed by the word
'Aristotle' refers (if at all) to Aristotle,
and so on. It might be thought that in the case of reference this
problem may be solved easily-by simply listing the referents of each
of the finitely many primitive terms in our language. But this is not
so. Just as our understanding of truth goes beyond the list of
presently formulable instances of the equivalence schema and tells
us that any new sentences could also be instantiated, similarly, our
conception of reference goes beyond a knowledge of the referents
of our current primitive vocabulary. 1 2 It covers a potentially infinite
12
This point is stressed by Max Black in his critique of Tarski's theory (Black,
1 948).
number of new terms. Consequently, we are pushed into formula
tions such as
( 1 3) (x)(y)(x refers to y iff {::l d} (x = 'd' & y = d)),
where the substitutional variable, d, ranges over singular terms in
possible extensions of our language. Thus we find ourselves relying
again on substitutional quantification and the need to explain it with
an infinite number of rules; so the reduction of truth to reference
has turned out to be futile.
Let me make some further points about the search for a finite
axiomatization of MT. In the first place, problems exactly like those
I have just mentioned arise for the project of explaining truth in
terms of predicate satisfaction. We would need to add a theory of
satisfaction that could encompass all facts like
( 1 4) The predicate 'blue' is satisfied by blue things;
and once again no finite list could suffice. An adequate theory would
have to contain infinitely many propositions of the form
( 1 5) The propositional constituent associated with the predi-
cate 'F' is satisfied by, and only by, things that are F.
Therefore concern about the infinite character of the minimal theory
of truth cannot be assuaged by explaining truth in terms of satis
faction.
Secondly, these conclusions do not tell against Tarski's own
project, in so far as it aspired merely to explicate a notion of 'true
in-L' for certain highly artificial languages, L. Each of these
languages has a fixed stock of primitives, so it is possible to expli
cate 'refers-in-L' and 'satisfies-in-L' with finite lists of principles.
Our project, however, is in certain respects more ambitious than
Tarski's. We are aiming for a theory of 'being true'-a property
which is attributed to propositions regardless of how or whether
they are expressed. Similarly we are looking for a theory of 'express
ing truth'-a property we may attribute to an utterance regardless
of the language in which it is couched. I have been considering the
possibility that someone might hope, in defining 'true', to exploit the
strategy that Tarski used in his definition of 'true-in-L'; but this will
not work-or so I have argued. 1 3
1 3 For further discussion o f Tarski see the answer t o Question 3 8 .

Thirdly, it might be thought that the difficulty in obtaining a
finite, compositional theory of truth stems from the implicit assump
tion that propositions are constructed, as Frege said, from the senses
of words-which are entities that require some theory of reference
and that such problems do not arise if propositions are instead con
structed, as Russell proposed, from concrete objects and properties.
For in that case we can say
(ET) (x)(R)(S)[x is the proposition consisting of the n-place
relation, R, and the sequence, S, of objects �
(x is true B S exemplifies R) ] .
But although this may be fine as far as it goes, it does not go far
enough. For it would have to be supplemented with a theory of
exemplification; and here is where the old troubles will emerge. We
might be able to derive from (ET)
(1 * * ) (Snow is white) is true iff snow exemplifies whiteness.
But in order to get the conclusion that
(1) (Snow is white) is true iff snow is white
we would need the schema
x exemplifies R 1 -ness iff R 1 x,
and in general we need schemata for each higher value of n as well,
that is,
(Ex) (x, y) exemplifies Rrness iff R2xy,
(x,y,z) exemplifies R3 -ness iff R3xyz,
. . . and so on.
And this apparatus is no less infinitary than the minimal theory.
Thus it isn't any easier to give a finite theory of truth if we focus on
Russellian, 'concrete' propositions.
An alternative idea involves the conception of a proposition as a
set of possible worlds. It may seem that the right definition of truth
for such propositions is not the equivalence schema but rather the
principle that
(W) x is true = the actual world is a member of x. 1 4
14 Suggested by Harty Field ( 1 992).

But again this offers no advantages over the minimal theory. For we
would have, for example,
( 1 * * *) (Snow is white) is true iff the actual world is a mem

ber of (snow is white).
And we are assuming the schema

(P) (p) = {w I P in w} .
Therefore we could infer
( 1 * * * *) (Snow is white) is true iff snow is white in the actual

world.
But at this point we are stuck. In order to derive the MT axiom we

would need
Snow is white in the actual world iff snow is white.
And to get all of the axioms we would need every instance of the
schema
(MTA) p in the actual world iff p.
Thus finite axiomatization is not achieved by the explanation of

truth in terms of actuality.
Finally, notice that no help is to be found by looking in the direc
tion of traditional theories of truth such as the coherence and
pragmatic approaches, or by entertaining some other way of identi
fying truth with a naturalistic property. For whatever property, F, is
associated with truth, we will be able to explain instances of the
equivalence schema only to the extent that we can explain instances
of the schema
( 1 6) (p) is F iff p.
And this infinite theory will be no easier to encapsulate than the min
imal theory.
I conclude that we should not expect to contain all instances of
the equivalence schema within a finite formulation: an infinity of
axioms is needed. And since this would seem to be an unavoidable
feature of any adequate theory of truth, it should not be held against
MT. Therefore we must acknowledge that the theory of truth can-
not be explicitly formulated. The best we can do is to give an implicit
specification of its basic principles. 1 5
6. If there were really no more to a complete theory of

truth than a l i st of biconditionals l i ke 'The p roposition
that snow is white is true if and only if snow is white' ,
then since one could always say 'p' rather than 'The
p roposition that p is true', it would be inexplicable that
o u r language should contain the word 'true': there
would be no point i n having such a notion .
This argument has already been dealt with; but it is often raised
against what are sometimes called 'redundancy' accounts of truth,
so let me repeat my response. First, the fact that the only applica
tions of truth expressly contained in the theory are within proposi
tions of the form
(E * ) ((p) is true iff p)
does not imply that the theory covers only those cases in which truth
is attributed to an articulated proposition. For suppose 'Einstein's
law' refers to the proposition (E = mc2). In other words,
( 1 7) (E = mc2) = Einstein's law.
In that case the theory of truth, which applies in the first instance
to
( 1 8) (E = mc2 ) is true,
must apply indirectly to
( 1 9) Einstein's law is true,
from which 'is true' cannot be removed. And it is from its role in
such sentences that the truth predicate gets its value. To see this, con
sider how we would manage without it. We would have to put the
matter roughly as follows:
1 5 For further discussion of why we should not expect MT to be explained in

terms of more fundamental axioms, see the answer to Question 14.
(20) (x)(If Einstein's law is the proposition that x, then x).
But this could not be construed in the usual manner. For, given the
usual conventions of quantification, that sentence is ill-formed in
two distinct ways: the second occurrence of 'x' is in an opaque con
text, beyond the reach of normal quantification; and a variable that
ranges over objects appears in sentential positions. In order to avoid
these incoherences it would be necessary to introduce a new form of
quantification-substitutional quantification-that could legiti
mately govern opaque contexts and sentence positions. That is to
say, we need a quantifier,
(2 1 ) {p } ( . . . p . . . ) ,
whose meaning i s not
(22) Every object, p, satisfies ' . . . p . . . ',
but rather
(23) Every grammatical substitution of a declarative sentence

of English in place of 'p' in ' . . . p . . . ' yields a truth.
But such a quantifier, with its special syntactic and semantic rules,
would be a cumbersome addition to our language. The point of our
notion of truth is that it provides a simple alternative to this appa
ratus. For, as I mentioned in Chapter 1 , the truth predicate allows
any sentence to be reformulated so that its entire content will be
expressed by the new subject-a singular term open to normal
objectual quantification. In other words, 'p' becomes '(p) is true'.
Therefore, instead of
(24) {p } (p -7 p),
we can say
(25) (x)(If x is a proposition of the form 'p -7 p', then x is

true).
Instead of
(26) {p } (If Einstein's law is the proposition that p, then p ),
we can say
(27) (x)(If x is a proposition of the form (If Einstein's law is
the proposition that p, then p), then x is true),
which is logically equivalent to
(28) (x)(If x = Einstein's law, then x is true) .
And, in general, instead of the substitutionally quantified
(2 1 ) {p } ( . . . p . . . ) ,
we can make do with the ordinary, objectually quantified

(29) (x)(If x is a proposition of the form ( . . . p . . . ) , then x
is true).
I am not suggesting, of course, that the truth predicate was intro
duced deliberately to perform this useful function. But I am sup
posing that its usefulness, as just described, is what explains its
presence. For if it were not valuable at all, it would presumably fall
out of use; and as for alternative functions that it might have, there
simply aren't any plausible candidates.
7. The mini mal theory fails to specify what are meant by

attributions of truth . It fails to provide necessary and
sufficient conditions for the appl icabil ity of the truth
predicate.
The second part of this point is quite correct, but does not justify
the initial complaint. For it is not the case that a satisfactory char
acterization of the meaning of a predicate must take the form of
necessary and sufficient conditions for its correct application-i.e.
an explicit, eliminative analysis. A definition of that sort is merely
one particularly simple way of specifying the use of a word; but we
should be open to more complex ways of doing it. So the present
objection presupposes needless restrictions on what sort of
definition of 'true' is needed. Once these implicit constraints are
loosened, the minimalist account will no longer seem inadequate.
I can perhaps clarify this response by distinguishing some differ
ent forms that a definition of 'true' might be thought to take. In the
first place one might offer an atomic definition: that is, a definition
of the familiar form
(30) 'true' means ' . . . ',
supplying a synonym that would permit us to eliminate the word
'true' in a uniform way from every context in which it appears. An
example of an atomic definition is the definition of 'bachelor' as
'unmarried man' . The pragmatists' identification of truth with util
ity has this character.
In the second place, and a little more modestly, one might offer
a contextual definition: that is, a set of rules that would allow the
conversion of any sentence containing the word 'true' into a syn
onymous sentence that does not contain it. A well known example
of this style of definition is Russell's theory of definite descriptions
(Russell, 1 905) :
(3 1 ) 'The F is G' means 'Some G is the same as every F',
which reduces the definite article, 'the', to the notions of predicate

logic-specifically, 'some', 'every', and 'the same as' . A partial
account of truth along these lines would be contained in the schema
(E!) 'It is true that p' means 'p' .
Thirdly, one might abandon the attempt t o provide the sort of

account that would enable the word 'true' to be eliminated, and aim
instead for implicit definition: that is, a set of principles involving the
truth predicate, our commitment to which fixes its meaning. For
example, it is sometimes said that the axioms of any geometry
implicitly determine the meanings of the terms 'point' and 'line', at
least as they are used when proving theorems of that geometry. An
account of the meaning of 'true' along these lines would be given
by the substitutionally quantified principle
(E+) {p } (x)(x = (p) � (x is true � p)).
Finally, one might deny that the meaning of the truth predicate
can be captured in our commitment to any definite body of prin
ciples. One might hold that the use-hence the meaning-of 'true'
is given by regularities with a more complex structure than simply:
'We accept 'A".' An example of this sort of use definition is the idea
that our conception of number is determined by the disposition to
accept Peano's axioms, including infinitely many instance of the
induction schema
(32) { F(O) & (n)[F(n) � F(n + 1 )] } � (n) [F(n)] .
Another example is provided by a certain account of the meaning

of counterfactual implication, namely,
(33) 'If p were true, then q would be true' is assertible to
degree x iff it is known that x is the empirical tendency
for q to be true in circumstances in which p is true and in
which all the facts causally and conceptually independent
of not-p still obtain.
This rule characterizes a certain sense of 'If . . . , then . . . ' by spe
cifying the appropriate level of confidence for any such conditional,
and without involving any principles, in the material mode, relating
counterfactual dependence to other aspects of reality. 1 6
I would suggest that the truth predicate belongs in this final cat
egory. Our understanding of 'is true'-our knowledge of its mean
ing-consists in the fact that the explanatorily basic regularity in our
use of it is the inclination to accept instantiations of the schema
(E) 'The proposition that p is true if and only if p',
by declarative sentences 1 7 of English (including any extensions of
English) . Thus for a normal English speaker it consists (a) in his dis
position to accept
(MT) 'The proposition that snow is white is true iff snow is
white', 'The proposition that I am hungry is true iff I
am hungry', 'The proposition that Paris is beautiful is
true iff Paris is beautiful', . . . ,
and (b) in the fact that his acceptance of other sentences contain
ing the truth predicate is explained in terms of (a) .
1 6 For elaboration of this account of counterfactual dependence see Horwich,

1 987.
1 7 Such sentences may be identified by their meanings. Moreover it is required
that their two tokens express the same proposition. If meaning and propositions were
then to be defined in terms of truth we would have a vicious circle. However, I argue
in my answers to Questions 22 and 32, and in sect. 3 of the Postscript, that meaning
and proposition may be explained in terms of aspects of use (including acceptance)
that do not presuppose the notion of truth.
The minimalist account of what it is to know the meaning of the
truth predicate does not provide an analysis and does not enable us
to specify in non-circular terms the content of attributions of truth.
This is precisely what distinguishes it from traditional approaches.
But it may be none the less a perfectly acceptable account of what
our understanding the truth predicate consists in, just so long as it
is capable of explaining all pertinent linguistic behaviour-all our
ways of deploying the term 'true'. 1 8 The question, in other words,
is whether we can explain, on the hypothesis that the equivalence
schema governs some person's use of the truth predicate, why, for
example, that person should endorse an inference from 'What Oscar
said is true' and 'What Oscar said is that eels are good' to 'Eels are
good'; and, in general, why he uses the truth predicate in just the
way that he does. This is the adequacy condition for a theory of the
meaning of the truth predicate; and, judging by the examples in the
answers to Question 3, the minimalist account would appear to sat
isfy it.
8. Is the m i n i mal ist conception concerned with truth itself

or with the word 'true ' ?
I t i s concerned with both-and other intimately related matters as

well. However it is important to separate the different questions it
addresses. Specifically we should distinguish between:
1 . A theory of the function of the truth predicate;
2. A theory of what it is for someone to understand the word
'true';
3 . A theory of the meaning of the word 'true';
4. A theory of what it is to have, or grasp, the concept of truth;
5. A theory of truth itself
1 8 See sect. 2 of the Postscript for further discussion. Notice that since minimal
ism does not provide an explicit definition of truth, it superficially resembles Moore's
view that truth is an 'inexplicable quality' (Moore, 1 899). The important difference
between the two accounts, however, is that minimalism nevertheless purports to give,
by means of the equivalence schema, a complete account of truth and of what our
grasp of it consists in, whereas on Moore's view it is impossible to shed any light on
these matters and truth remains impenetrably mysterious.
According to the minimalist conception, the function of the truth
predicate is to enable the explicit formulation of schematic general
izations. Our understanding of the word is constituted by the prac
tice of using it to perform that function-a practice whose basic
regularity is an inclination to accept instances of the equivalence
schema, '(p) is true iff p'. The concept of truth (i.e. what is meant
by the word 'true') is that constituent of belief states expressed in
uses of the word by those who understand it-i.e. by those whose
use of it is governed by the equivalence schema. And the theory of
truth itself-specifying the explanatorily fundamental facts about
truth-is made up of instances of that schema. Thus, the minimal
theory of truth will provide the basis for accounts of the meaning
and function of the truth predicate, of our understanding it, of our
grasp upon the concept of truth, and of the character of truth
itself. 1 9
9. Even if we grant that, as predicates go, the truth

p redicate is highly u nusual-even if we grant that its
special function is to enable us to say certain i mportant
things while avoiding new forms of quantification-it surely
does not fol l ow that being true is not a
genuine property.
Quite right. And it is not part of the minimalist conception to main

tain that truth is not a property. On the contrary, 'is true' is a per
fectly good English predicate-and (leaving aside nominalistic
concerns about the very notion of 'property') one might well take
this to be a conclusive criterion for standing for a property of some
sort. What the minimalist wishes to emphasize, however, is that truth
is not a complex or naturalistic property but a property of some
other kind. (Hartry Field ( 1 992) suggests the term 'logical
property'.) The point behind this jargon is that different kinds of
property correspond to different roles that predicates play in our lan
guage, and that unless these <;lifferences are appreciated, we will be
tempted to raise questions regarding one sort that can legitimately
1 9 See sect. 5 of the Postscript for more detailed discussion of the answer to this
question.
arise only in connection with another sort. A familiar example of
this phenomenon derives from the predicate 'exists' . Another, more
controversial, case is the conflation of normative and descriptive
properties. According to minimalism, we should, for similar reasons,
beware of assimilating being true to such properties as being
turq uoise, being a tree, or being made of tin. Otherwise we will find
ourselves looking for its constitutive structure, its causal behaviour,
and its typical manifestations-features peculiar to what I am call
ing 'complex' or 'naturalistic properties'. We will be puzzled when
these expectations are inevitably frustrated, and incline to the
conclusion that the nature of truth is profoundly obscure-perhaps
even incomprehensible.
As I have indicated, some philosophers hold that no predicate
refers and that properties do not exist; and, of course, from that
nominalistic point of view the particular question 'whether truth is
a property' does not arise-at least, in those words. However the
underlying issue is still with us in the form of whether or not appli
cations of the truth predicate engender statements about the propo
sitions to which it is applied. The thesis that they do distinguishes
the present view from certain more radical formulations of
deflationism-those according to which it is a grammatical illusion
to think that
(34) X is true
makes a statement of any kind about the proposition X. For
example, it was suggested by Frege ( 1 89 1 , 1 9 1 8), Ramsey ( 1 927), and
Ayer ( 1 93 5 , 1 936) that the forms
(3 5) p
and
(36) It is true that p
yield the same sense no matter what declarative English sentence is
substituted for 'p' . This is often referred to as 'the redundancy the
ory of truth' and it evidently conflicts with the view advanced here
which associates a definite propositional constituent to the truth
predicate-a constituent which is part of one of these propositions
but not of the other. Similarly, from the present perspective·we are
rejecting the idea due to Strawson ( 1 950) and Ayer ( 1 963) that the
truth predicate is not used to give descriptions or make statements
about the things to which it is applied, but that it is used instead
to perform quite different speech acts: endorsing, agreeing, conced
ing, etc.
The trouble with the 'redundancy/performative' conception is
thatit cannot be squared with obvious facts about the character and
function of truth. It addresses only cases like
(37) (Snow is white) is true,
in which the truth predicate is attached to an explicitly articulated
proposition. And it maintains, with a certain prima facie plausibil
ity, that the whole sentence has the same sense as the constituent
(38) Snow is white.
But notice that such uses of truth have no great value: we could eas
ily do without them. And when we turn to genuinely useful attribu
tions, as for example in
(39) Oscar's claim is true,
the theory has nothing to say about its sense, except that the logical
form is supposedly not what it would seem to be: i.e. not
(40) X is F.
Consequently, the redundancy theory is quite unable to account for
the inference from (39) and
(4 1 ) Oscar's claim = the proposition that snow is white
to
(37) The proposition that snow is white is true,
and hence to
(38) Snow is white
-which is precisely the sort of reasoning on which the utility of our
concept of truth depends. 20 Thus the redundancy/performative
20 Similar objections to the redundancy theory have been made by Tarski

( 1 943/4), Thomson ( 1 948), Cohen ( 1 950), Ziff ( 1 962), and Ezorsky ( 1 963). A redun
dancy theorist might attempt to explain the above inference by first analysing 'X is
true' in terms of substitutional quantification as
(SB) {::3p } (X = (p) & (p) is true) (cont. )
theory must be rejected. No doubt we do perform all kinds of speech
act (such as agreeing and conceding) with the truth predicate. But,
as Warnock observed ( 1 964), it is best to say that we do so by (not
instead of) making a statement-that is, by attributing the property,
truth, to the proposition in question. Just as the assertion
(42) Your article is brilliant
may be intended to achieve an effect beyond conveying accurate
information, so one might well have some ulterior purpose in mind
in saying
(39) Oscar's claim is true,
yet none the less be making a statement about Oscar's claim: i.e.
attributing a property to it.
10. If the equivalence schema is rel ied on indiscrimin

ately, then the notorious 'liar' paradoxes wil l resu lt.
Indeed-and for that reason we must conclude that permissible

instantiations of the equivalence schema are restricted in some way
so as to avoid paradoxical results. To see this, let '#' abbreviate 'THE
PROPOSITION FORMULATED IN CAPITAL LETTERS IS
NOT TRUE' . Then assuming for the sake of argument that
(43) (#) is true,
and given the pertinent instance of the equivalence schema, namely
(44) (#) is true iff #,
we can infer
and then taking this to be synonymous with

(ST) {::Jp } (X = (p) & p).
However, although this strategy is clearly motivated by certain redundancy-theoretic
intuitions, it departs radically from that theory in associating a definite content with
the truth predicate. We should distinguish between the redundancy theory, accord
ing to which ' X is true' says nothing about X, and theories according to which it does
say something about X-but something that should be analysed in terms of substi
tutional quantification. The latter theories face their own set of difficulties, as we saw
in the answer to Question 6.
(45) #,
whose subject, said t o b e not true, turns out t o b e the proposition
(#). Therefore we have
( 46) (#) is not true,
contradicting our initial assumption. But the alternative fares no
better. For if we assume instead that
(4 7) (#) is not true,

then, given modus tollens applied to the equivalence schema, it fol
lows that
(48) - #,
which says of (#) that it is not not true. Therefore
(49) (#) is true.
Thus we have deduced that
(50) (#) is true iff (#) is not true.
In order to block the derivation of this contradiction our options
are: ( 1 ) to deny classical logic-specifically, either modus ponens,
modus tollens, double-negation elimination, or Leibniz's law (the
indiscernibility of identicals); (2) to deny (a la Tarski) that the con
cept of truth can be coherently applied to propositions, such as (#),
which themselves involve that concept; (3) to deny that the sentence
in capital letters succeeds in formulating a proposition; or (4) to
reject certain instances of the equivalence schema-including the
one obtained by substituting '#' into it.
But ( 1 ) cuts too deep; (2) also smacks of overkill; and (3) goes
against the fact that, for any condition C, one might happen to
believe that the proposition meeting that condition is not true
which (since any object of belief is a proposition) would imply that
'The proposition meeting condition C is not true' expresses a propo
sition. And this will be so even if it happens to turn out that the
proposition it expresses is the one meeting C; so '#' does express a
proposition. Therefore the only acceptable solution is (4): only cer
tain instances of the equivalence schema are correct.
We know that this restriction need not be severe. It need have no
bearing on the propositions of science-the vast majority of which
do not themselves involve the concept of truth. The problem of giv
ing a constructive account of exactly how far one can push the equiv
alence principle without engendering paradox is the subject of a
great deal of contemporary research (e.g. Tarski, 1 958; Kripke, 1 975;
Gupta, 1 982; McGee, 1 99 1 ; Gupta and Belnap, 1 993) and will not
be addressed in this book. Given our purposes, it suffices for us to
concede that certain instances of the equivalence schema are not to
be included as axioms of the minimal theory, and to note that the
principles governing our selection of excluded instances are, in order
of priority: (a) that the minimal theory not engender 'liar-type'
contradictions; (b) that the set of excluded instances be as small as
possible; and-perhaps just as important as (b)-(c) that there be a
constructive specification of the excluded instances that is as simple
as possible. 2 1
I should emphasize that my intention in these remarks is not to
disparage constructive attempts to deal with the paradoxes, or to
suggest that our knowledge about truth is not deficient in the
absence of such an account. My point is merely that there are man
ageable and philosophically fruitful problems of truth that are inde
pendent of the search for a constructive solution to the paradoxes:
first, to outline a theory of truth; second, to specify what we mean
by the truth predicate; third, to explain its role in our conceptual
scheme; and fourth, to say whether there is some theory of the
underlying nature of truth. There is no reason to suppose that the
minimalist answers that are advanced in this essay could be under
mined by any particular constructive solution to the paradoxes-so
we can temporarily set those problems aside.
The object of this chapter has been to specify the adequacy condi
tions for a complete account of truth, to suggest that these desider-
2 1 Anil Gupta has pointed out to me that the need to restrict instantiation of the
equivalence schema is somewhat in tension with the minimalist thesis about the func
tion of our concept of truth-namely that it enables us to capture schematic gener
alizations. For, in so far as 'p' is not invariably equivalent to '(p) is true', then a
generalization of the form 'Every instance of schema S is true' will not invariably
entail every instance of S; nor will it always be justified or explained on the basis of
those instances. For example, 'Everything he says is true' (i.e. 'Every instance of "If
he says that p, then p" is true') does not entail 'If he says that #, then #', and is not
partially justified or explained on the basis of that conditional. However, such prob
lem cases are few and far between; so the utility of truth as a device of generaliza
tion is not substantially impaired by their existence.
ata are satisfied by a certain deflationary conception of truth, called
'minimalism', and to make sure that this proposal is not confused
with various superficially similar views, such as Tarski's and the
redundancy/performative account. The axioms of the minimal the
ory are all the propositions of the form, ((p) is true iff p)-at least,
those that don't fall foul of the 'liar' paradoxes. And (as Anil Gupta
has pointed out) there is one further axiom to the effect that propo
sitions are the sole bearers of truth. We found some reason to believe
that such a theory-weak as it is-is nevertheless strong enough to
account for the conceptual utility of truth, and to explain the facts
in which truth is a constituent. And we saw that the single unat
tractive feature of the theory-its infinite list-like character-is not
mitigated by accounts of truth in terms of reference or substitu
tional quantification. Thus we have gone some way towards justify
ing the minimalist conception: the view that the minimal theory is
the theory of truth, to which virtually nothing more should be
added.
But many problems remain. For one thing, our entire discussion
has taken for granted that truth is a property of propositions; and
those philosophers suspicious of propositions will find it hard to
swallow that aspect of the view. This issue is the focus of Chapter
6. I have placed it towards the end because it is something of a
digression, and anyone who is already comfortable with propositions
can manage perfectly well without it.
Another widely felt objection to the deflationary view of truth is
that it cannot be squared with the explanatory role of the notion of
truth; and I shall attempt in the next chapter to provide further sup
port for minimalism by showing where this argument goes wrong.
The basis for the objection is the idea that any law of nature relat
ing various properties can be explained only by reference to theories
that specify the underlying character of the properties involved. For
example, in order to say why all emeralds are green we need to know
what it is to be an emerald and what it is to be green. And similarly,
it is argued, in so far as the notion of truth is employed in the for
mulation of general laws, we are going to need a substantive theory
of what truth is in order to explain these laws. I want to suggest, on
the contrary, that truth appears in explanatory generalizations in
precisely the role identified by the minimalist conception, and that
the equivalence axioms are quite sufficient to account for them.
3
The Explanatory Role of the

Concept of Tr uth
1 1 . Truth has certain characteristic effects and causes.

For example, true beliefs tend to facil itate the achievement
of practical goals. General laws such as this call for
explanation i n terms of the natu re of truth . Therefore there
m ust be some accou nt of what truth is, going beyond the
m i n i mal ist story, that provides a conceptual or empi rical
reduction of this p roperty. (Putnam, 1 978; Field, 1 972,
1 986; Devitt and Sterelny, 1 989)
As we shall see, truth does indeed enter into explanatory principles,

but their validity may be understood from within the minimal
theory.
Consider in the first place those of a person's beliefs of the form
(1) (If I perform action A then state o f affairs S will b e real-
ized).
The psychological role of such beliefs is to motivate the performance
of A when S is desired. When this process takes place, and if the
belief involved is true, then the desired result will in fact obtain. In
other words, if I have belief ( 1 ) and desire S, then I will do A. But
if my belief is true, then, given merely the equivalence axioms, it fol
lows that if I do A then S will be realized. Therefore, by modus
ponens, S will be realized; I will get what I wanted. Thus it is easy
to see how the truth of beliefs of the kind in question may contribute
to the fulfilment of goals. (A formal version of this explanation was
given in the answer to Question 3 .) Moreover, such beliefs are more
likely to be true if they are inferred from true premises; and very lit
tle of what we believe can be definitively excluded from the prospect
of entering into such inferences as a premise. Therefore it is clear, in
The Explanatory Role of the Concept of Truth 45
general, how true beliefs contribute to practical success. Nothing
beyond the minimal theory is called for to explain this phenomenon.
It is worth noting three features of our argument. In the first
place, it does not imply that true beliefs are always beneficial. That
would be a mistake, since there are obviously circumstances in which
a false belief will happen to produce the best outcome and circum
stances in which the truth would be too costly to be worth finding
out. The argument purports merely to articulate a certain mech
anism by which true beliefs engender beneficial results, and does not
deny the existence of other mechanisms, which may operate simul
taneously, by which a true belief will have bad consequences. In the
second place, our explanatory demonstration of the beneficial con
sequences of true belief is based on facts that are easily and widely
recognized. Therefore it may be transformed into an account of why
we should want our beliefs to be true-why we aim for the truth.
And in the third place, notice that the essential line of explanation
is unaffected by the recognition of more complex and realistic pat
terns of deliberation than those we have been assuming. For exam
ple, suppose that we really act according to the principle of utility.
In other words, given the choice between actions A and B, we per
form the one with the greatest expected value, calculated by means
of the formula,
(2) V(x) = [ V( S 1 ).B(S 1 /x)] + [ V(S2).B(S2fx)] + . . .
(where Si, S2 , • is an exhaustive set of mutually exclusive possible
• •
outcomes; V(S 1 ), V( S2), are the values that the agent places on
• • •
them; and B( S 1 /x), B(S2/x), . . . are his degrees of belief that each
outcome will obtain given the performance of action x). To the
extent that the degrees of belief are near the truth (i.e. to the extent
that B(Sj /x) is high if x would in fact bring about Sj, and low if it
wouldn't) then the expected value of each action will be close to its
actual value (i.e. to the value of what would in fact occur if it were
performed); and therefore the decision is more likely to be objec
tively correct. 1 Moreover, as we saw with respect to the simpler
1 To see this, suppose that the actual consequence of act x would be Si . and con
sider the possibility of having had degrees of belief, B*, that were closer to the truth
than B are. What this means is that B*(S1/x) is closer to 1 than B(S1/x) is; but for all
46 The Explanatory Role of the Concep t of Truth
model, these beneficial consequences of truth (in the case of beliefs
of the specific kind involved in deliberation), indicate that there is
value in the truth of any premises from which those beliefs might be
inferred.
We may conclude that the explanatory role that the concept of
truth plays in the general principle, 'True beliefs facilitate successful
behaviour', may be completely understood via the minimal theory.
This fact tends against various anti-minimalist positions. First, it
should assuage the concerns of philosophers (such as Dummett
( 1 959) and Wright ( 1 98 8 , 1 992)) who think that no deflationary con
ception of truth could do justice to the fact that we aim for the truth.
Secondly, it shows that the presence of truth in such general prin
ciples gives no reason to suppose (with Putnam ( 1 978)) that the
property of truth has any sort of underlying structure. And third it
undermines the pragmatist's impulse to ensure by definition the role
of truth in successful activity.
1 2.Another lawl i ke generalization is that beliefs obtained

as a result of certain methods of i n q u i ry tend to be true.
Again this suggests that the m i n imalist conception
overlooks truth's causal/explanatory natu re.
We are now turning from the effects of truth to its causes. Beliefs
are sometimes reached in such a way as to inspire particular
confidence in their truth, and in such cases our confidence is usually
vindicated. Consider, for example, observations of the colours of
ordinary objects in good light. Reports of such observations are
n ;t i , B*(Snlx) is closer to 0 than B(Snlx) is. Thus the expected value of x would have
been
V*(x) =V(S 1 )(B(S 1 /x) + e2 + e3 + e4 + . . . ) + V(S2/x)(B(S2/x) - e2) +
V(S3'x)(B(S3/x) - e3) + . . .
=V(S 1 )B(S 1 /x) + (e 1 + e2 + e 3 + . . ) V(S 1 ) + V(S 2)B(S2/x) - e2 V(S2) +
.
V(S3)B(S3/x) - e 3 V(S3 ) + . ..
which is between V(S 1 ) (the objective value of x) and V(x) (its expected value rela
tive to the degrees of belief, B) . Therefore degrees of belief closer to the truth imply
expected values closer to objective ones.
generally correct; and beliefs arrived at by deductive and inductive
inference from observational premises are often correct. The ques
tion is: why is this so? Why are beliefs regarding certain domains,
when resulting from certain methods of inquiry, so strikingly
reliable? And won't the answer reveal something important about
the nature of truth?
For any observation sentence, ' O', such as 'That's red', 'The
needle coincides with the spot marked with a "3"', etc., there are cir
cumstances, C(' O'), that we take to be particularly conducive to the
accurate determination of its truth value, and there are other cir
cumstances in which, though the sentence may nevertheless be
asserted or denied, there is thought to be a much higher risk of error.
One of the things we are trying to explain is why it is that every
instance of
(3) ' O' would be affirmed in C(' O') iff% O*
is true (where 'p iff% q ' means 'The probability of q given p and of
p given q are both very high', and where ' O* ' is our way of formu
lating the proposition expressed by ' O') .
At the most superficial level the explanation is quite straight
forward. It is a biological fact that humans can be educated, and a
social fact that some of them are educated, to say 'That's red' when
and only when something red is present, providing the light is good,
eyes are open, etc. That is why
(4) 'That's red' would be affirmed in C('That's red') iff%

something red is there.
For analogous reasons, though certainly not the same reasons,
( 5) 'That's green' would be affirmed in C('That's green') iff%

something green is there.
And so it will go, for each instance of
(3) ' O' would be affirmed in C(' O') iff% O* .

Each instance has its own explanation, though some of the instances
share some explanatory antecedents. Taken together, these explana
tions show why every instance of (3) is true; or, in other words, why
observation reports made in good conditions tend to be true. The
minimal theory of truth is perfectly adequate.
48 The Explanatory Role of the Concept of Truth
Once we have explained why, in certain circumstances, observa
tional beliefs tend to be true, it is not difficult to see why inferred
beliefs are also reliable. Truth functions of observational statements
(e.g. 'That's red or green') tend to be true because they are reached
by deduction from observational beliefs that tend to be true. Beliefs
in generalizations of the form
(6) All As are B
(where 'A' and 'B' are observation terms) are reliable because they
are reached on the basis of the observation of many diverse A s that
are B and of no As that are not B-and it so happens that the world
is uniform in this respect. Finally, consider how we might account
for the reliability of certain scientific instruments. Suppose a device,
I, is designed to discover whether the state of affairs in some
domain, S, is Si , S2, . . . , Sb . . . , or Sn ; and suppose this is done
by noticing whether the observable output of the instrument is O i ,
02 , . . . , Ob . . . , o r On , and then inferring the presence o f the cor
responding state-i.e. inferring Sk from Ok. An explanation of the
reliability of I might proceed from the following premises:
(7a) The use of instrument I will give rise, for some k, to the
belief that Ok obtains;
(7b) If we believe that Ok obtains then Ok probably does
obtain;
(7c) There is a high nomological correlation between Ok and
Sk;
(7d) If we believe that Ok obtains, then we infer that Sk obtains.
From these we can infer that instrument I will probably give rise to
true beliefs concerning the domain S. Again, nothing beyond the
minimal account of truth is needed here.
1 3. A fu rther explanatory role for truth l ies in the fact that

the truth of scientific theories accou nts for the i r
empi rical success. (Putnam , 1 978)
No doubt we often do explain the success of a theory by reference

to its truth or approximate truth. We say such things as:
(8) The Special Theory enables accurate predictions because it
is true,
and
(9) The electron microscope works so well because the theo

ries on which it is based are true.
It remains to be seen, however, whether such explanatory statements

provide a reason for thinking that truth has a hidden naturalistic
structure, or whether they can be perfectly well accommodated by
the equivalence axioms.
Of course, I urge the latter position. Consider the situation in
which we know explicitly which theory we are talking about; and
suppose its formulation is not very long or complicated. Suppose,
for example, that the theory is simply
( 1 0) Nothing goes faster than light.
In that case, rather than saying
(1 1) The theory that nothing goes faster than light works well
because it is true,
we could equally well have said
( 1 2) The theory that nothing goes faster than light works well
because nothing goes faster than light.
No further explanatory depth is achieved by putting the matter in

terms of truth. None the less, use of the truth predicate in this sort
of context will often have a point. What it gives us is a certain econ
omy of expression, and the capacity to make such explanatory
claims even when we don't explicitly know what the theory is, or
when we wish to generalize, e.g.
( 1 3) True theories yield accurate predictions.
But these are precisely the features of truth that are central to the
minimalist conception. Clearly they can provide no reason to go
beyond it.
50 The Explanatory Role of the Concept of Truth
1 4. Even if all o u r general beliefs about truth are

deducible from the m i n i mal theory (suitably augmented),
this does not i m ply that no deeper analysis of truth is
des i rable ; for one m ight wel l hope to find something that
wi l l show why it is that the equivalence schema holds.
(Papineau , 1 99 1 )
We can certainly entertain the possibility that the minimal theory is

susceptible of explanation via some deeper account of truth.
However, there is excellent reason to suppose that in fact there is no
such deeper theory.
In the first place, the initially tempting strategies for developing
such a theory are not workable. In the answyr to Question 6 we tried
analysing truth in terms of non-standard (sentential) quantification,
in terms of reference and satisfaction (a la Tarski), in terms of prop
erty exemplification, or in terms of 'actuality', but found it impos
sible to arrive at an account capable of explaining or improving
upon the minimal theory.
In the second place, the equivalence schema is a priori and con
ceptually basic. In these respects it is analogous to the fundamental
laws of logic and arithmetic where there is no expectation of a reduc
tive analysis or any other kind of deeper explanation.
Thirdly, in the domain of a posteriori fact it is reasonable to
expect reductive explanations, because the behaviour of a physical
system is the causal consequence of the properties of its parts. But
such considerations do not apply in the a priori domain. Thus, the
minimal theory of truth does not cry out for explanation in the way
that a posteriori theories do. 2 Consider, for example, the account
of chemical valance (discussed by Field ( 1 972)) which consists in
simply listing the valances of each element:
( 1 4) (x)(y)(x is the valence of y iff x = + 1 and y = potassium,
or x = - 2 and y = sulphur, or . . . ) .
I n this case there is a reason t o expect further reduction. For there
are laws of nature about valence-laws about the relationship
between the valences of elements and the proportions in which they
combine-that are not explained by the list ( 1 4). And any a pos-
2 See Leeds, 1 978 for discussion of this point.

teriori lawlike generalization calls for explanation on pain of look
ing like a sheer coincidence. However the minimal theory does not
itself contain such laws; and it is conceded that every general fact
about truth may be explained by the minimal theory. Thus nothing
should lead us either to expect or to desire a deeper explanation.
Fourthly, the equivalence axioms could be explained only by prin
ciples that are simpler and more unified than they are-principles
concerning propositional elements and the conditions in which truth
emerges from combining them. But the single respect in which the
body of minimal axioms is not already perfectly simple is that there
are so many of them-infinitely many; and no alleged explanation
could improve on this feature. For there are infinitely many funda
mental propositional constituents to take into account; so any char
acterization of them will also need infinitely many axioms.
Conceivably, such a theory might none the less obtain unifying
(hence explanatory) credentials by accounting not only for the
equivalence axioms, but also for the phenomena in some other non
truth-theoretic domain. But this prospect is pure fantasy. We have
no idea of what realm of fact could be related to truth in such a
way. 3
It has been shown in this chapter that the existence of various sci
entific, explanatory generalizations, couched in terms of truth, does
not call for an analysis of truth-a theory of its underlying struc
ture. This is because such laws may be wholly understood on the
basis merely of the equivalence axioms, and because any explana
tion of these propositions is highly unlikely. In the next chapter I
turn from the scientific to the philosophical use of the notion of
truth and argue, in a similar vein, that its scope and value are cap
tured by the minimalist conception.
3 Notice that these considerations relate not merely to truth, but also to reference
and satisfaction. We should expect no deeper analyses of any of these semantic phe
nomena than are provided by their minimal theories (sketched in the answer to
Question 39).
4
Methodology and Scientific

Realism
A deflationary attitude towards truth is inconsistent with the usual

view of it as a deep and vital element of philosophical theory.
Consequently the many philosophers who are inclined to give the
notion of truth a central role in their reflections on metaphysical,
epistemological, and semantic problems must reject the minimalist
account of its function. Conversely, those who sympathize with
deflationary ideas about truth will not wish to place much theoret
ical weight on it. They will maintain that philosophy may employ
the notion only in its minimalist capacity-that is, as something
enabling the formulation of certain generalizations-and that
theoretical problems must be resolved without it. The latter point of
view is what I will be trying to sustain in the present chapter and in
the one immediately following. Here I shall try to show that the
realism/anti-realism issue (together with various related questions in
the philosophy of science) have nothing at all to do with truth, and
that a failure to recognize this fact has stood in the way of clear
thinking about those matters. And in the next chapter I shall argue
more or less the same point with respect to a range of questions in
semantics and in the philosophy of logic.
1 5. Doesn't the deflationary perspective-the renu nciation

of a substantive notion of truth-lead inevitably to
relativism: to the idea that there is no such thing as
objective correctness?
The claim that truth is not a complex or naturalistic property-that

it is 'unreal' or 'insubstantial', in the sense advocated by minimal
ism-must not be confused with the idea that truths are unreal, or,
Methodology and Scientific Realism 53
in other words, that no sentence, statement, or belief is ever true.
The latter view might arise from an extreme form of relativism in
which it is supposed that truth is 'radically perspectival' or 'con
textual' or something of the sort. But this kind of theory is not at
all affiliated with the minimalist conception of truth. On the con
trary, the two philosophical positions tend to be opposed to one
another. For it can be precisely the association of the truth predi
cate with some beefed-up, highly esteemed metaphysical or
epistemological property-i.e. the substantiality of truth-that leads
to the conclusion that nothing ever quite manages to be absolutely
true. And conversely, the minimalist position, in so far as it makes
it easy to suppose that every proposition, or its denial, is true,
implies that relativism (in at least one popular formulation) is unten
able. Thus the 'insubstantiality' of truth is in no way tantamount to
the nonexistence of truths. 1
1 6. Nevertheless, isn't the mini mal ist perspective i n some

sense anti-real ist? Does it not deny that scientific
theories are i ntended to correspond to a
m i nd-independent world?
Debate over the question 'What is realism?' can easily take on the
aspect of an empty, pointless, terminological wrangle. One philo
sopher will identify the position with, say, an aversion to reduction
ism; another will complain that certain pro-reductionist positions
(e.g. that the mind is merely the brain (materialism), that mathe
matical facts are merely logical facts (logicism)) are not intuitively
anti-realistic and that certain positions that are anti-realistic (e.g.
that the numbers are a human invention (mathematical intuition
ism)) are non-reductionist; the critic may then propose an alternative
definition of 'anti-realism'-say, rejection of the principle that every
proposition is either true or false; but this again will fail to satisfy
everyone's intuitions about when to apply the label. Thus the process
continues interminably, so that one is left wondering-since after all
1 An argument that the minimalist conception of truth leads to relativism has

been given by Putnam ( 1 98 1). However, objections that I consider to be conclusive
have been levelled against it by Michael Williams (1 986).
54 Methodology and Scientific Realism
the words 'realism' and 'anti-realism' are terms of art-whether
there is any genuine problem here at all. Wouldn't it be best to dis
tinguish explicitly various senses of 'realism' corresponding to each
of the alternative proposals, so we can begin to focus our attention
on the real questions: namely, which of these so-called 'realist'
positions are correct and which are not?
I think that this tempting point of view is in fact mistaken.
Realism and anti-realism are definite and interesting philosophical
stances, and the issue of what exactly they are is a substantial one.
Our impression to the contrary comes partly from the fact that the
answers that are usually suggested bear such little resemblance to
one another, and partly from the fact that even with respect to indi
vidual philosophical positions (for example, behaviourism and intu
itionism) there is no consensus about whether they should be
counted as realist or not. But the explanation of all this divergence
of opinion, it seems to me, is not that there is no correct definition
of realism to be found, but rather that the definitions usually pro
posed are of completely the wrong sort.
What I have in mind can be brought out by reviewing some well
known facts about how natural kinds-for example, diseases-are
properly defined. Notorious difficulties arise if one tries to charac
terize a disease in terms of observable symptoms, for there will
always be some that are not manifest in a few people who none the
less have the disease, and other characteristic conditions that some
times occur in the absense of the disease. The familiar moral is that
the right way to specify the criterion for having a disease is to iden
tify the underlying causes of its symptoms, rather than the symp
toms themselves. It seems to me that this moral applies pretty well
to our questions about the nature of the realism/anti-realism issue.
What accounts for the endless squabbling about it is that we have
been focusing our attention on the symptoms of realism and anti
realism. Not surprisingly, no definition in those terms can work.
What we must do instead is think about what it is that leads people
to adopt the positions we are inclined to regard as realist or anti
realist. Thus we will see what the basic conflict between realism and
anti-realism really is. And the positions that we intuitively classify
as realist or anti-realist-e.g. that physical objects are constructs
from experience (phenomenalism), that electrons are fictional
entities (instrumentalism), etc.-will qualify as such in virtue of
being adopted as a consequence of taking one side or the other on
the basic issue. Notice that this analysis of the situation will imme
diately account for the divergence of opinion over whether certain
specific theories, such as behaviourism and intuitionism, are anti
realist or not. As in the case of a disease, where a given condition
may in some cases be a symptom and in another case not, a given
philosophical thesis will be an anti-realist move in the case of those
philosophers who embrace it for certain reasons, but not for those
whose motives are quite different. Thus we might distinguish, say,
between a 'realist behaviourist' and an 'anti-realist behaviourist' :
both have exactly the same view about the reducibility of mental
facts to behaviour, but they diverge in their reasons for holding it.
What, then, is the essence of realism? The answer is very simple.
There is a question about how it is possible for us to know of the
existence of certain facts given our ordinary conception of their
nature. This is because there can seem to be a tension in ordinary
thinking between the metaphysical autonomy of the world (its inde
pendence of us) and its epistemological accessibility (our capacity
to find out about it) . The difference between a realist and an anti
realist, in a nutshell, is that the realist decides on reflection that there
is actually no difficulty here-so our ordinary ideas about what we
know can stand; whereas the anti-realist decides, on the contrary,
that the alleged conflict is genuine and that it has certain
ramifications for what we can take ourselves to know. The alterna
tive symptoms of anti-realism are alternative views about what these
ramifications are-alternative modifications of our naive view of the
world and our capacity to comprehend it. Thus it is not unusual for
an anti-realist with respect to some domain to embrace one of the
following strategies:
( 1 ) Deny that there are any facts of the sort at issue (e.g.
formalism, instrumentalism, emotivism, relativism, non
factualism);
(2) Deny that we have the capacity to know such facts (e.g.
scepticism, constructive empiricism, error theory);
(3) Reduce the facts in question to other facts whose epistemo
logical status seems unproblematic (e. g. phenomenalism,
behaviourism, logicism).
The central point, once again, is that none of these doctrines, nor
any collection of them, is either necessary or sufficient for anti
realism. Rather, anti-realism is the view that our common-sense
conception of what we know is incoherent: the supposed character
of facts of a certain type cannot be reconciled with our capacity to
discover them. It is in response to this view that one or another of
the above doctrines may well be espoused and in such case the adop
tion of the doctrine qualifies as an anti-realist move.
According to this way of thinking, any position whatsoever might
count as an anti-realist move for some philosopher, providing that
this philosopher regards the position as the proper solution to the
anti-realist dilemma. Thus the class of 'possible anti-realist pos
itions' (like the class of 'possible symptoms of diabetes') is com
pletely uninteresting. What is of interest, however, is something we
might call the class of 'natural anti-realist positions': that is, pos
itions which really would remove the alleged tension in our naive
world view. In other words, it is worth distinguishing, from amongst
the more or less irrelevant things an anti-realist might be inclined to
say, those positions that really would, if they could be adequately
sustained, address the alleged dilemma.
The three strategies just mentioned clearly have this character:
they are natural anti-realist positions. However, what is glaringly
absent from this group is any particular thesis about the nature of
truth. This is not of course to deny that someone might, as a mat
ter of biographical fact, feel forced into some account of truth by
the anti-realist dilemma. My claim, rather, is that any such response
would be irrational. If the dilemma is real, then no theory of truth
could help to resolve it.
1 7. But this concl usion is extremely counterintuitive. It

seems obvious that the natu re of truth bears d i rectly on
the structure of reality and the conditions for com p rehend
ing it. S u rely, 'truth' and 'reality' are semantically
i n extricable from one another; so how cou l d one's
position in the real ism debate be d ivorced from
one's conception of truth?
The term 'realism' is an over-used, under-constrained piece of philo

sophical jargon, and one can no doubt invent senses of it such that
the minimalist approach qualifies either as 'realist' or 'anti-realist' .
However, the substantial question here, as we have just seen, con-
cerns the relation, if any, between our conception of truth and the
justifiability of believing in facts that exist independently of thought
or experience. And there is no relation-or so I shall argue. On the
contrary, a significant source of confusion in the debates about
scientific realism is the tendency to assume that the problem of truth
is fundamentally involved. 2
Anti-realism, as we have seen, consists in a perceived tension
between the realist's twin, common-sense commitments to credibil
ity and autonomy. Some anti-realist philosophers have supposed
that since the facts are independent of experience they are non
existent, or at least epistemologically inaccessible to us. Thus we
arrive at the sort of instrumentalism or theoretical scepticism advo
cated, for example, by Duhem ( 1 954), Popper ( 1 962), and van
Fraassen ( 1 980). Other philosophers have supposed, conversely, that
since the facts are verifiable they must reduce to observation. Thus
we get the sort of reductive empiricism characteristic of phenome
nalism and the Vienna Circle. But none of these natural anti-realist
positions is in any way affiliated with the minimalist conception of
truth. In the first place, minimalism gives no reason to think that
theories are constructions out of data, and is quite at home with the
holistic considerations that have led most philosophers to reject that
aspect of logical positivism. Secondly, it is perfectly consistent with
the minimal account of truth to suppose that the scientific method
provides us with theories that we should believe to be true and not
merely observationally adequate. According to the deflationary pic
ture, believing that a theory is true is a trivial step beyond believing
the theory; and the justifiability of this attitude is certainly not pre
cluded by minimalism. 3
Not only is the minimalist conception of truth quite neutral with
respect to the two central aspects of realism (namely, the questions
of justified belief and empirical reducibility), but the same can be
said of alternative conceptions. As we shall now see, the choice of a
theory of truth is orthogonal to the issues surrounding realism. The
2 The independence of questions about truth from the traditional issues of real
ism was urged by Tarski ( 1 943/4), and has been emphasized by Michael Devitt ( 1 984).
For further discussion see Horwich, 1 996.
3 In support of the anti-sceptical component of realism I have argued elsewhere
( 1 99 1 ) that there is really no difference between believing a total theory and the appar
ently less-committed attitude of instrumental acceptance.
theory of truth can have no definite implications for either the epis
temological or the semantic component of the problem.
Consider, for example, the constructivist account, which identifies
truth with a kind of demonstrability or verifiability. There is an incli
nation to suppose that this conception of truth immediately entails
the falsity of certain forms of scepticism, and that it thereby sup
ports the epistemological aspect of realism. At the same time it is
also thought that the meanings of sentences would be given by their
'truth conditions' (in the constructivist sense); and so it seems that
the content of a claim such as
(4) There are electrons
could then be nothing more than
(5) 'There are electrons' is demonstrable.
Thus the autonomy of theoretical facts would be lost, and we would
have semantic anti-realism.
But both of these arguments are fallacious. In order to combat
scepticism regarding some theory, T, we must be able to argue (in
the face of the underdetermination of theory by all possible data)
that we are justified in believing T. Constructivism (according to
Peirce ( 1 932/3)) tells us that we can infer
(6) T is true
from the premise
(7) T will eventually be demonstrated (i.e. verified) in the limit
of scientific investigation.
In Putnam's ( 1 9 8 1 ) version of the doctrine, the premise should be
(8) T would be demonstrated in the course of an ideal inquiry.
However, it is not at all obvious, from the sceptic's viewpoint, that
we are entitled to believe either of these premises. Moreover, even if
that source of scepticism were removed, it would still be unclear why
a justified belief that T is true should carry with it the justification
to believe T. No doubt a minimalist can assume this; for he regards
the equivalence of such beliefs is as fundamental . But a construc
tivist, on the contrary, defines truth in terms of demonstrability. For
him, the equivalence schema is a substantive claim that must be sup
ported on the basis of this fundamental assumption about truth.
And the possibility of such an argument is precisely what the scep
tic will deny. Thus constructivism gives no easy proof of the ordi
nary claims to knowledge constitutive of scientific realism.
Nor does it amount to a form of anti-realism, as many writers,
following Dummett ( 1 977), have assumed. There is a tendency to
confuse the following three theses:
(9) The meaning of 'p' consists in the fact that 'p' is regarded
as demonstrated in such and such circumstances;
( 1 0) ' "p" is true' means ' "p" is demonstrable';
and
(1 1) 'p' means "'p" is demonstrable' .
The final thesis evidently conflicts with realism; fo r i t reduces facts
about external reality to facts about our thought and experience.
However, constructivism gives us the right to nothing more than
( 1 0), and arguably (9). And these premises provide no basis for deny
ing that scientific theories describe a mind-independent reality.4
A second account of truth with merely apparent implications for
realism is the view that truth is a primitive, non-epistemic property
that is grasped independently of the equivalence schema. This view
seems to wear on its face the radical autonomy of theoretical facts.
And, as a consequence, scepticism can appear to be unavoidable. For
if the property of truth is primitive and wholly unexplainable, then
we can surely have no reason to suppose that the propositions we
regard as confirmed tend to have this property. But, once again,
these attempts to link the theory of truth with realist and anti-realist
theses are misconceived. For the meaning of the word 'true' is one
thing, and the meanings of theoretical terms like 'electron' and
' super-ego' are quite separate. Whatever we say about 'true' cannot
determine our view on the question of whether our theoretical
terminology is reducible to observational terminology. Similarly,
however mysterious and inaccessible we think the property of
truth-however hard we suppose it is to assess ' T is true '-we will
not necessarily be saddled with scepticism; for we need not also
4 Putnam ( 1 98 3 : 280) appears to go wrong in this way when he argues that any
one who adopts the combination of a redundancy theory of truth and an assertibility
condition conception of meaning will is 'perilously close to being a solipsist of the
present instant' .
suppose that the property of (say) being an electron is mysterious
and inaccessible; there need be no scepticism about T itself. As
before, the essential point is that any theory that explains truth
independently of the equivalence schema, loses the right to assume
without further ado that the schema holds. Therefore, relative to the
conception of truth in question, problems regarding the justification
of ' T is true' are not automatically linked with problems regarding
the justification of T.
1 8. If, as the m i n imal theory implies, 'truth' is not defined

as the prod uct of ideal i n q u i ry, why should we believe
that an ideal i n q u i ry would provide the truth?
To regard a certain inquiry as ideal is to suppose that one should

not question its outcome. So if an inquiry into whether or not there
is life on Mars yields the result that there isn't, and if the inquiry is
taken to be ideal, then we should be absolutely confident that there
is no life on Mars. Moreover, given the equivalence schema, one
should also be confident in the truth of the proposition that there is
no life on Mars. Similarly, one should believe of any ideal inquiry
that it provides the truth. 5 This is a trivial consequence of the min
imal theory and the meaning of 'ideal inquiry'. Therefore construc
tivism is unmotivated; for what it feels the need to guarantee by
definition may in fact be derived from the minimal theory.
Notice that there is no presumption here that every hypothesis is
susceptible to some idealized inquiry. Therefore, although we have
grounds for the schematic thesis
( 1 2) If (p) is the product of an ideal inquiry, then (p) is true,
the converse claim-and, a fortiori, the identification of truth with
idealized verification-has not been supported. Indeed, this
identification-the constructivist theory of truth-greatly overesti-
5 In the terminology of Putnam ( 1 9 8 1 ) , we are here rebutting the 'metaphysical

realist' thesis that 'truth is radically non-epistemic'. Fine; but this does not mean that
we agree to incorporate epistemic ideas into the very notion of truth. Peter van
Inwagen ( 1 988) points out in a similar vein that Putnam's argument that 'a fair
amount of what we believe must be true' also does not imply that truth is an epis
temic concept.
mates our epistemological power. For there are truths beyond the
reach of even an ideal investigation. Consider the phenomenon of
vagueness (e.g. 'is bald' applied to a borderline case); or underde
termination of theory by data; or sentences with assertibility con
ditions that don't allow for conclusive verification (e. g. 'The
probability that drug X will cure disease Y is 0. 3'). Any of these phe
nomena might involve a proposition which is such that no ideal
investigation would engender either its assertion or its denial. In that
case we have
( 1 3) It is not the case that (p) is verifiable and it is not the

case that (-p) is verifiable.
But then, according to the constructivist's definition of truth, we can

infer
( 1 4) It is not the case that (p) is true and it is not the case
that (-p) is true.
And by the equivalence principle,
( 1 5) p H (p) is true,
we get
( 1 6) -p and - -p,
a contradiction! Thus not only is constructivism unmotivated (i.e.

not needed to account for whatever correlation exists between truth
and verification), it is extensionally false since it cannot acknowledge
the existence of truths that are not conclusively demonstrable. One
might try to argue, in response, that given a sufficiently pumped-up
construal of 'ideal inquiry', there is really no need to acknowledge
unverifiable truths. For it could be supposed that for every proposi
tion, including the problematic cases just mentioned, a sufficiently
ideal investigation would decide its truth. But although the con
structivist principle may be protected against counterexample by this
manamvre, it will then be even less appropriate than before to regard
it as our basic theory of truth. For the notion of 'sufficiently ideal
inquiry' will now be one that is most naturally explicated in terms
of the concept of truth. Thus the constructivist principle, once it is
cast into a plausible form, clearly becomes something to be derived
from the minimal theory of truth, and not to be taken as explana
torily basic.
1 9. How is it possible, g iven the m i n i mal theory, for truth

to be someth ing of i ntrinsic val ue, desirable i ndependently
of its practical util ity?
To value truth is, very roughly speaking, to wish for satisfaction of

the schema 'p iff I believe p ', and therefore to be committed to the
techniques of investigation that will apparently achieve this result. 6
To value truth for its own sake is to desire it to some extent regard
less of its compatibility with other goals. Such a value may be eth
ical or aesthetic, or it may be something one simply wants. Which
of these alternative possibilities is correct is left open by the mini
mal theory, as is the possibility that truth not be valued for its own
sake.
It might be thought that if truth is intrinsically valuable, then
minimalism is in trouble, since it surely lacks the resources to explain
that value. 7 But this criticism is unjust. For the difficulty that
attaches to explaining why true belief is intrinsically good is no more
or less than the difficulty of explaining, for any other particular
thing (e. g. kindness, happiness, etc.), why it is intrinsically good. The
problem stems from our failure to understand the concept of intrin
sic goodness, rather than from our adoption of the minimalist con
ception of truth. More specifically, in so far as we don't know what
it is for something to possess the quality of intrinsic goodness, no
explanation of why truth possesses it will suggest itself, regardless
of which theory of truth is adopted. On the other hand, if some
6 This way of articulating what it is to value truth involves a couple of oversim

plifications. First, it suggests that we wish to believe every true proposition; whereas,
in fact, the truth values of most of them are of no interest to us. And second, it does
not allow for degrees of belief. Really, what we want is to 'minimize', in a certain
sense, the error in our plausibility judgements. If these are represented by numbers
between 0 and 1 , then what we want is to minimize the square difference between the
probability assigned to each proposition (p), and either 1 (if (p) is true) or 0 (if (p)
is false). See Horwich, 1 982b for discussion of this matter, together with an argument
that these desires are accommodated by the acquisition of new data.
7 Bernard Williams ( 1 996) expresses scepticism about minimalism on these
grounds.
account of intrinsic goodness is assumed, then it is far from obvi
ous that a minimalistically acceptable explanation of why truth has
that quality could not be based upon the assumed account.
Consider, for example, an analysis of 'being intrinsically good'
roughly along the lines of 'being normally conducive to human wel
fare'. On the basis of this sort of theory the minimalistic account of
truth's pragmatic value (given in response to Question 1 1) might well
be developed into an account of its intrinsic value. Note that such
an account would not undermine the intrinsicality of the value. For
to recognize that truth is valuable 'for its own sake'-i.e. to suppose
that truth is good even in those cases where it will have no practical
benefits-is not to suppose that the explanation of this value is inde
pendent of the normal advantages of truth.
20. How can m i n i malism accom modate the idea of

science progressing towards the truth?
It suffices to imagine a temporal sequence of total theories T( l),

T(2), . . . , T(k), . . . , T(final), becoming gradually (but not neces
sarily monotonically) more similar to T(final)-where T(final) is
true. T(final) is a conjunction of unknown and presently inexpress
ible propositions. However, as we saw in the answer to Question 2,
this is no obstacle to applying the truth predicate. The notion of
'theoretical similarity' remains to be explained; but there is no rea
son to expect that this can or should be done with a high degree of
precision. We can get by with our ordinary crude intuitions of the
extent to which two bodies of claims are similar to one another. Of
course, in order to make these comparisons it is necessary that the
theory formulations be to a fair degree intertranslatable (or 'com
mensurable' in Kuhn's terminology (Kuhn, 1 962)). I have not tried
to show that this would be so. However, I am not arguing here that
there actually exists progress in science; but only that a minimalist
conception of truth does not stand in the way of such a thesis. 8
8 Even if we give up the idea that there exists a 'final true theory' we could still
make sense of a weaker version of the view that science progresses with respect to
truth. We might suppose (a) that a proposition, (p), is roughly true just in case the
proposition (Roughly, p) is true (for example, it is roughly true that John is six feet
tall when it is true that John is roughly six feet tall); (b) that later members of the
21 . From the perspective of the m i n i malist conception of

truth it is i m possible to p roduce an adequate justification
of scientific methods. (Friedman, 1 979)
An argument that our ways of acquiring beliefs take us in the direc

tion of truth might proceed on the basis of assumptions that are
themselves the results of methods we are trying to justify. That is to
say, an explanation of the fact that a certain scientific method M is
reliable might proceed from premises of which some are believed as
a consequence of employing M itself. According to Michael
Friedman, this sort of derivation will sometimes constitute a justi
fication of the method M. He argues that-given a sufficiently sub
stantive conception of truth-the circularity that is evidently
involved need not be vicious. It need not render the derivation so
easy to provide that it isn't worth having. After all, he says, there can
be no prior guarantee that the products of M will not suggest a
theory that implies M's unreliability. Therefore, it can be a pleasantly
surprising and epistemologically significant fact regarding M if it
turns out to be demonstrably reliable---even if the demonstration
employs the results of M itself
Friedman's objection to deflationism is that, from the perspective
of such an insubstantive conception of truth, it would be a trivial
matter (available regardless of what one's beliefs and methods actu
ally are) to produce demonstrations of reliability. As a consequence,
these demonstrations would be quite devoid of explanatory or epis
temological importance, and we would be left with no grounds for
confidence in the reliability of scientific inquiry.
He reasons as follows. Suppose
( 1 7) Method M has engendered beliefs
P 1 , P 2, . . . , and Pn ·
In order to derive the reliability of M it suffices to combine premise
( 1 7) and the additional premises, p 1 , p2, , and Pm that were
• • .
obtained from M. From these additional premises (and the equiva

lence schema) we infer that
possibly endless sequence of total theories tend to contain a greater number of

roughly true, basic theoretical claims than earlier members; and (c) that most mem
bers of the sequence are such that the earlier members tend to be increasingly theo
retically similar to it.
' '
( 1 8) p 1 is true,
' '
p2 is true,
. . .,
and 'pn ' is true.
And then, from premise ( 1 7), we get the result that M engenders true
beliefs.
This pseudo-explanation is indeed worthless and provides no sup
port whatsoever for method M. But before casting aspersions on
minimalism we should consider a series of further questions:
( 1 9) Does minimalism play any role in the above pseudo

explanation?
(20) Could there not be-consistent with minimalism-a
more substantive demonstration of M's reliability?
(2 1 ) Even from the perspective o f a non-minimalist theory of
truth, could there be a telling demonstration of M's relia
bility?
If the answers to these questions were, respectively, Yes, No, and Yes,
then we might indeed have to concede that minimalism has unwel
come epistemological consequences. But in fact I believe that the
answer to all three questions is No. In the first place, the above
pseudo-justification relies merely on the equivalences that are com
mon to all reasonable accounts of truth. Minimalism-the thesis
that such biconditionals exhaust the theory of truth-plays no role
and cannot be blamed. Secondly, the reliability of certain methods
way well be demonstrable on the assumption of facts discovered by
other methods-but their reliability would then be at issue.
Eventually the question would arise as to whether some method (or
collection of methods) is capable of justifying itself in the manner
that Friedman proposes. And it seems to me that, regardless of one's
theory of truth, no such justification can be given.
Here I am questioning whether it really is possible for science to
undermine itself in a thoroughgoing way. I would suggest that the
circular procedure envisaged by Friedman could not go badly wrong;
and that this has nothing to do with the account of truth that is
employed. I am not denying that the theories resulting from
method M might fail to provide an explanation of M's reliability.
For perhaps further theories are needed; or perhaps the reliability
of M is extremely hard to explain. What I question is that the the
ories resulting from M might imply that M is not reliable.
Consequently, since I agree that something is supported by its suc
cessful predictions only to the extent that they might have been mis
taken, I also question whether the success of Friedman's circular
explanatory procedure could constitute any sort of justification for
relying on M. 9
In order to motivate this scepticism, let me consider a couple of
examples. Suppose that, on the basis of observational beliefs O i ,
02, . . . , and Om we were t o postulate a theory T that entails that
our methods of observation were very unreliable. In that case we
would have a theory T that, on the one hand, is confirmed by the
fact that it entails O i , 02, , and On; yet, on the other hand, entails
• • •
that a high proportion of O i , 02, , and On are false. Therefore T

• . .
would have to be internally inconsistent, and we shouldn't have pos

tulated it in the first place. Similarly, suppose that our theory T*,
justified by induction, entails that inductive inference is unreliable.
In other words:
(22) T* � Although data have conformed to T* in the past,
they will not conform to T* in the future.
Again, such a theory is internally inconsistent and should not have
been taken seriously in the first place.
Thus there are limits to the extent that science, rationally pur
sued, can invalidate itself. But this has nothing to do with the nature
of truth and, in particular, is not a consequence of minimalism.
Friedman suggests, on the contrary, that a naturalistic (causal)
theory of truth would leave more room for scientific self-criticism
and self-validation, for he supposes that it would open up the pos
sibility that most of the beliefs we regard as verified will turn out
not to have the naturalistic property of truth. But this is an illusion.
If there were such a property-a naturalistic reduction of truth-it
9 Perhaps Friedman's concern is merely with the explanation of M's success, and
not with providing reasons for confidence in it. But this interpretation is hard to rec
oncile with various facts: ( 1 ) that no reason is given for thinking that a non-trivial,
minimalistically acceptable explanation of M's reliability cannot be found; (2) that it
is thought necessary to maintain that M might undermine itself; and (3) that it is
thought necessary to have a conception of truth according to which it is conceivable
that most of our beliefs are false.
could be recognized as such only by means of the assumption that
it does tend to be present in circumstances that we regard as
instances of verification. If we don't impose this constraint then we
violate not merely the minimal theory of truth, but any theory that
respects the equivalence of 'p' and '(p) is true' . Thus no remotely
plausible account of truth could make it conceivable that our beliefs
are predominantly false and our methods of arriving at them unre
liable. The limited applicability of Friedman's epistemological strat
egy will not be expanded by rejecting the minimal theory. 10
Having attempted in earlier chapters to make clear and plausible the

minimalist point of view, I have here begun to explore its philo
sophical implications, specifically with regard to the debate over
realism. Not surprisingly, what I have tried to show is that our
notion of truth does not occupy the central theoretical position that
philosophers often assume it must occupy. Indeed, many problems
are exacerbated by the conviction that truth is essentially involved
and that their solutions depend on finding out more about its under
lying nature. We have seen on the contrary that in so far as the notion
of truth is properly employed in the philosophy of science it displays
no more than its minimalistic function. And in the next chapter I
argue the same point with respect to a broad range of semantic ques
tions. I shall take up (a) the nature of understanding, (b) the basis
of logic, (c) empty names, (d) vagueness, and (e) the status of ethi
cal assertion; and I shall indicate in each case how the problems
becomes much simpler once it is acknowledged that the concept of
truth should not be relied on to solve them.
10 For some further criticisms of Friedman's line of thought see Williams, 1 986.
5
Meaning and Logic
22. As Davidson ( 1 967) has argued , understanding a sen

tence, say, 'Tachyons can travel back i n time', is a matter of
appreciating what must be the case for the sentence to be
true-knowing its truth condition. That is to say, one must
be aware that 'Tachyons can travel back i n time' is true iff
tachyons can travel back i n time. Therefore it is not possible
to agree with the m i n i malist claim that this knowledge also
helps to constitute our grasp of 'is true' . For in that case we
wou l d be faced with something l i ke a single eq uation and
two u nknowns. Rather, if knowledge of the truth condition of
'Tachyons can travel back in time' is to constitute our u nder
standing of that sentence then this knowledge would
p resuppose some p re-existing conception of truth .
( D u m mett, 1 959; Davidson , 1 984)
What is right in this point is that knowledge of the truth condition of

a sentence cannot simultaneously constitute both our knowledge of
its meaning and our grasp of truth for the sentence. What is wrong
about it is its choice of the first of these options. For, on such a view,
how might we come to know that
(M) 'Tachyons can travel back in time' is true iff tachyons can
travel back in time?
How could this feat be accomplished? The picture that comes to
mind is that we deliberately associate the sentence 'Tachyons can
travel back in time' with a possible state of affairs, where the form of
association is our decision to count the sentence 'Tachyons can travel
back in time' as true if and only if the state of affairs obtains. But
there are considerable difficulties in this position. It is surely impos-
Meaning and Logic 69
sible for an individual to conceive of such an explicit association
unless he employs some sort of mental event---call it 'R'-to repre
sent the possible state of affairs. And in that case two problems
emerge. In the first place, it is more straightforward to represent the
association of our sentence with the possible state of affairs by means
of the definition
(M*) 'Tachyons can travel back in time' is to have the same
meaning as R
rather than by means of (M); so the notion of truth need not be
involved. And, in the second place, we must raise the question of
what provides representation R with its meaning. In order to avoid an
infinite regress it must be conceded that certain representational enti
ties obtain their content by means other than having been explicitly
associated with possible states of affairs. Yet the truth-conditional
approach provides no place for such alternative means. Moreover, if
some cases of meaning do not arise from explicit associations with
truth conditions, then why should we assume that understanding
'Tachyons can travel back in time' must have that character?
The way to avoid this mess is to recognize that while understand
ing a sentence does indeed usually coincide with an explicit knowl
edge of its truth condition, understanding does not consist in such
knowledge. It consists, rather, in appreciating the sentence's syntactic
structure and understanding its constituent words, which, in turn,
consists in knowing the basic regularities in their use-those regular
ities that will explain the overall use (including the assertibility con
ditions) of all the sentences in which those words occur. Once
'Tachyons can travel back in time' is understood in this way by some
one with a conception of truth, then the minimalist account entails
that he knows that 'Tachyons can travel back in time' is true iff
tachyons can travel back in time-i.e. it entails that he knows the
truth condition of 'Tachyons can travel back in time'. Moreover, such
knowledge can usually be attributed only to those who understand
'Tachyons can travel back in time'. 1 Thus anyone with a conception
of truth who understands 'Tachyons can travel back in time' will
1 But not always. Someone who does not understand German and who is told that
'Schnee ist wei13' is true iff frozen H 2 0 is white, does not understand the German sen
tence, even though he knows its truth condition, because he does not know whether
'Schnee' means 'snow' or 'frozen H 2 0'.
70 Meaning and Logic
indeed come to know its truth condition. However, contrary to what
is assumed in the objection, the understanding does not derive from
this knowledge. 2
Let me stress some aspects of this position. First, the advantage of
relying on use in giving a naturaiistic characterization of under
standing is that it is obvious how knowledge of the use of a word is
manifested (namely, by accepting certain sentences containing it in
certain conditions), whereas it is not at all clear how knowledge of
truth conditions is manifested; that is, unless such knowledge is con
strued, in the way that I have suggested, as the product of a knowl
edge of meaning (which is in tum explained in terms of use) and a
grasp on the concept of truth. Secondly, it is no objection to either
the coherence, or the preferability, of the notion of the use of a word
that it might sometimes include accepting certain sentences when and
only when their truth conditions are satisfied. For example, it might
be that the use of 'blue' involves accepting 'That is blue' when and
only when the designated object is blue. This does not alter the fact
that we can see how knowledge of the word's use may be manifested,
whereas we cannot see how knowledge of its contribution to the
truth condition may be manifested. Thirdly, an analysis of meaning
(and truth condition) in terms of use does not imply that a sentence
cannot be true without being assertible. In the first place, we are not
simply identifying the meaning of a sentence with its assertibility
conditions. And in the second place, even if we did, the assertibility
of a sentence would not follow from its truth. 3 Finally, it is sometimes
claimed to be a special advantage of the 'truth-conditional' analysis
of meaning that it enables us to see how the meanings of composite
expressions depend on the meanings of their parts, and to see there
fore how it is possible for us, with our finite minds, to understand a
potential infinity of compound expressions. But this alleged advan
tage is an illusion; for the compositionality of meaning can equally
well be accommodated within the use conception of meaning. In so
far as the fact that a word has the meaning it has consists in a certain
fact about its use, then the meaning of a complex expression will con
sist in the fact that it has a certain structure together with the facts
2 See Harman, 1 974, 1 982, for a statement of this position. There is more on the
use theory of meaning in the answer to Question 32. See also my Meaning ( 1 998).
3 For an elaboration of this point, see the discussion of scientific realism in sec
tions 17 and 1 8 .
underlying the meanings o f its constituents. The main idea behind
this approach to compositionality is that understanding a complex
expression is, by definition, nothing over and above understanding its
constituents and appreciating how they have been combined with
one another. If this is so, then compositionality can put no constraint
on how the meanings of words are constituted. For whatever are the
underlying characteristics that do this, complexes will mean what
they do in virtue of being composed as they are from words with
specific characteristics of that sort. In particular, if a word's meaning
derives from its use, then a complex's meaning consists in its being the
result of combining, in a certain way, words with certain uses.4
23. What about falsity and negation?
The simplest deflationary strategy is to define falsity as the absence of

truth, as follows:
(1) (x)(x is false H x is a proposition & x is not true),
or, in other words,
(2) (p) is false H (p) is not true.
As for the word 'not', it is traditionally supposed that the best way to
explain both it and the other logical constants is by means of truth
tables. In the case of negation, this is generally taken to be
(3) p not p
T F
F T
However, in the present context such an account is unacceptably cir
cular-we have defined falsity in terms of negation, and would now
be defining negation in terms of falsity. Moreover, the logician's term
'not' (underlined), which is supposedly defined by the truth table,
does not mean the English 'not', but rather 'It is not the case that', or
in other words 'It is not true that' -which isn't what we needed to
define.
4 For elaboration of this view of compositionality see my 1 997 a, reprinted as ch. 7
of my Meaning ( 1 998).
In order to deal with the second of these problems, and to bring it
about that an account of the logician's 'not' will engage with a
minimalistic account of falsity, we must adopt a slight modification
of that account-we must replace (2) with the definition
(2*) (p) is false H not [(p) is true],
or equivalently
(2* *) (p) is false H not p.
And in order to deal with the first problem, and remove the circular
ity involved in combining (2*) with the truth table definition of 'not',
we must eliminate the notions of truth and falsity which occur in the
truth table by explicating them in accordance with the minimalist
proposals. In that case the lines of the truth table, which initially say
(N) (i) (p) is true � (not p) is false
(ii) (p) is false � (not p) is true,
and the implicit claim that one of the two lines applies, i.e.
(iii) (p) is true or (p) is false,
are transformed into the following theses:
(N*) (i*) p � not not p
(ii*) not p � not p
(iii*) p or not p.
In taking 'not' to be implicitly defined by (N*), rather than by (3)
or (N), we are no longer attempting to explain it in terms of falsity;
so the threat of circularity has been defused. However these three
theses, though to some extent constraining the meaning of 'not', are
not enough to fix it completely. A complete account of the meaning
of 'not' must contain those fundamental facts about its use that
suffice to explain our entire employment of the term. Such basic reg
ularities of use might well include acceptance of the theorems of
deductive logic-which include the laws implicit in (N*). But a fur
ther pattern of usage, not implied by (N*), must be recognized:
namely, that which is characterized by the principle
(K) 'not p' is acceptable to the degree that 'p' is unacceptable.
Perhaps the combination of (N*) and (K), when conjoined with facts
about the use of other terms, will be capable of explaining all our
ways of deploying 'not' . If so, then its meaning will be fixed and we
can proceed to define falsity in terms of it by means of definition (2*)
and without fear of circularity.
24. As Frege ( 1 9 1 8) said, logic is the science of truth ; so

s u rely o u r accounts of truth and logic should be, if not
identical , at least bound u p with one another. Yet the
m i n i mal theory does not even enable one to prove that
the principle of non-contradiction is true.
The concept of truth is involved in stating laws of logic and meta

logic; for example,
(4) Every proposition of the form 'p � p ' is true.
(5) If a conditional and its antecedent are true, then so is its
consequent.
(6) The laws of logical inference preserve truth.
Thus one easily gains the impression that truth and logic bear a pecu
liarly intimate relationship to one another. Indeed Frege maintained
that, just as biology is the science of living things and astronomy the
science of stars, so logic is the science of truth. And it is not uncom
mon (e.g. Dummett, 1 977; Putnam, 1 978) to see the competition
between alternative logics described as a choice between conceptions
of truth.
From the perspective of minimalism, this way of thinking is incor
rect. The reason that the notion of truth is heavily involved in logic is
not that logic is about truth, but simply that logic makes precisely the
sort of generalization that the truth predicate enables us to formulate.
Moreover, just as we should, if possible, prise apart our theories of
truth and reference, so, for the same reason, we should distinguish
logic from the theory of truth. One and the same theory of truth can
be combined with classical logic to demonstrate, for example, that the
distributive law is true, or combined with quantum logic to show that
it isn't. And just as our belief that the Principle of Relativity is true
requires for its derivation assumptions from physics as well as from
the theory of truth, similarly it should be quite unsurprising that in
order to prove the truth of the principle of non-contradiction we
need to invoke logic, and not simply the theory of truth. This means
that the issue between, for example, classical and intuitionistic logic
has as little to do with truth as has the issue between Newtonian and
Einsteinian physics. It would be absurd to describe the latter as a
conflict between Newtonian and Einsteinian conceptions of truth;
and for the same reason it is wrong to think of the logical dispute as
a competition between 'the classical conception of truth' and 'the
intuitionistic conception of truth' .
Although I use classical logic throughout this book, none o f my
claims about truth presupposes it. Indeed, as I have just emphasized,
a central tenet of the point of view advanced here is that the theory
of truth and the theory of logic have nothing to do with one another.
Thus minimalism is the proper conception of truth even in the con
text of deviant logics such as intuitionism or quantum logic, and
would not be undermined by any arguments demonstrating the
preferability of non-classical rules of inference.
25. Perhaps m i n i malism can be squared with alternative

logics; but it cannot be squared with the role that truth
must play i n the foundations of logic-i n justifyi ng one
logic over another.
But this suspicion is also mistaken; for the concept of truth plays no
substantial role in the justification of logic. To see this consider how
a system of basic rules of inference, L, could conceivably be justified.
One strategy, it might be thought, is to specify the meanings of the
logical constants by principles in which truth is the primary semantic
notion-by means of truth tables-and then to show on the basis of
these principles that the rules of L preserve truth. The trouble with
this strategy is that it is blatantly circular; for the principles specifying
the meanings of the constants are just trivial reformulations of the
very rules we want to justify. For example, the classical truth table for
'and' is
(7) p q p&q
T T T
T F F
F T F
F F F
which is tantamount to the rules of inference
(7*) p,q -p,q p,-q -p,-q
:.p & q :. -(p & q) :. -(p & q) :. -(p & q)
:. (p & q) v (-p & q) v (p & -q) v (-p & -q)
for the difference between them is nothing more than the equivalence
principles of truth and falsity. When these rules are supplemented
with the rules underlying the other constants we obtain the classical
propositional calculus. Thus the truth tables are not substantially dif
ferent from the rules of inference that we are trying to support, and
so do not form a suitable starting-point.
A second strategy for the justification of logic L would begin with
the specification of the meanings of the logical constants by means of
principles whose main semantic notion is something other than
truth-e.g. assertibility or proof. Consider for example the intuition
istic account of disjunction:
(8) Something is a proof of 'A v B' just in case we can see that
it is (or will yield) a proof of 'A' or a proof of 'B' .
One might then hope to justify logic L by showing that only proofs in
L would accord with these principles. But we should appreciate two
facts about this strategy.
In the first place, no account of truth is involved. So there is cer
tainly no reason here to think that something beyond the minimal
theory is required in the foundations of logic. It might be thought
that assumptions about truth must play a role at some stage; for isn't
the whole point to show that the rules of L preserve truth? But even
given that characterization of our goal, minimalism will do. If we can
show that L provides the only system of rules of inference that accord
with the meanings of the logical constants, then we have thereby
justified the rules of L. Then, if we want, we can introduce the equiv
alence schema to derive the conclusion that if the premises of an
argument of L are true then so is the conclusion. But this final step is
not needed and does not invoke anything beyond the minimal
theory.
The second point about this strategy is that it suffers from the
same flaw as the first strategy, in that it does not really offer a non
circular justification of logic. The principles that, for each logical
constant, constrain the assertibility conditions of sentences contain
ing them, have no epistemological priority over the rules of inference
one might derive from them. Both the assertibility conditions and the
basic rules of inference each determine regularities in the use of the
logical constants. So, in so far as meaning is constituted by use, they
could equally well be regarded as specifying the meanings of the con
stants. And if it is legitimate to accept the principles that specify
assertibility conditions without further justification, then it should
be legitimate to accept the basic rules of inference in the first place.
My inclination is to think that the principles of deductive logic are
neither susceptible of, nor in need of, demonstrative justification.
Instead, I would follow Quine ( 1 953) in supposing that the logical
principles relied upon in scientific theorizing are justified in so far as
they are members of the simplest way of accommodating experience.
I would say, moreover, that the logical principles deployed outside
science in ordinary life-which may or may not differ from those
used in science-are not subject to revision in light of experience.
Hence these commitments are a priori; their justification is prag
matic; purely practical goals are furthered by adopting them. 5 I have
not argued for these conclusions here. All I have tried to show is that
alleged justifications of logic from assumptions about the meanings
of the logical constants are not substantive, and anyway do not
threaten the minimalist conception of truth.
26. How can truth-value gaps be admitted?

(Dummett, 1 959)
They can't be. Begin with the minimalist view of falsity formulated
either as
(2*) (p > is false � not [( p > is true]
or as
(2**) (p > is false � not p.
Either of these definitions, in conjunction with the equivalence
schema
5 For further discussion see my 1 997e, reprinted as ch. 6 of Meaning (1 998).
(E) (p ) is true H p
yields the conclusion
(9) (p ) is not true and not false � not p & not not p.
Thus we cannot claim of some proposition that it has no truth value,

for that would imply a contradiction. Moreover, given classical logic
(in particular, the law of excluded middle, ' (x)(Fx v -Fx)'), we can go
even further, not merely refraining from the claim that some proposi
tions are neither true nor not true, but asserting positively that every
proposition is true or not-i.e. true or false.
Admittedly these results do not derive solely from the minimal
theory of truth, but depend also on our having defined falsity as the
absence of truth. So one might conceivably make room for proposi
tions that are neither true nor 'false' by constructing a narrower con
ception of 'falsity' than the one I have endorsed; one might hope to
distinguish various different ways for a proposition to be untrue
only one of which is 'being false'. The reasons for supposing, on the
contrary, that the minimalist characterization of falsity is the right
one are as follows:
(a) The account reflects our pre-theoretical intuition that if a

proposition is not true then it is false, and that if something is
not the case then the claim that it is the case would be false.
(b) No reasonably plausible alternative characterization of falsity
is able to accommodate these features of the concept.
(c) The minimalist picture of truth encourages a parallel account
of falsity-such as (2**)-according to which its attributions
are similarly equivalent to non-semantic propositions.
( d) The spirit of minimalism precludes accounts of truth and fal
sity which would equip them for theoretical work in seman
tics.
Thus we have ample reason to accept the minimalist view of falsity

embodied in principles (2*) or (2**)-and to embrace the conclusion
that no proposition can be neither true nor false.
27. But doesn't p h ilosophy requ i re truth-value gaps i n

order t o accom modate s u c h phenomena a s non-referring
names, vagueness, the emotivist conception of .
ethics, etc.?
These are all cases in which the misimpression that truth is a funda
mental ingredient of reality has fostered the idea that it should play a
substantive role in philosophical theorizing. The minimalist position,
on the other hand, is that truth can play no such role, and that when
ever we are tempted to give it one, the philosophical issues will be
clarified and resolved by recognizing that this is a mistake.
In the first place, as Russell ( 1 905) argued quite convincingly
against Frege ( 1 891), an atomic proposition entails that the referents
of its singular terms exist: a is F entails that a exists; and in that case,
it is natural to allow that if it is false that a exists then it is false that a
is F. Therefore atomic propositions containing vacuous singular
terms may very plausibly be regarded as false and don't call for truth
value gaps. One way of ensuring this result is to combine Russell's
theory of definite descriptions (Russell, 1 905) and Quine's strategy
for predicatizing names (Quine, 1 953), to produce logical forms that
are free of singular terms. For example,
(I 0) Everyone has an ancestor from Atlantis,
which contains the empty name '.Atlantis', becomes
( 1 0*) There is a place with the property of being-Atlantis, and
everyone has an ancestor from there,
which is uncontroversially false.
28. It is obvious that many p redicates-for example, 'blue',

'small ' , 'bald', 'heap'-do not have defin ite extensions;
and when such p redicates are applied to certain objects
the resu lt wil l s u rely be p ropositions with no truth val ue.
We can grant that there are indeed propositions-notably those

attributing vague properties to borderline cases-whose truth values
can never be determined and which might well be described as
'having no determinate truth value' or 'being objectively neither true
nor false'. But we are not forced, on that account, to give up the clas
sical law of excluded middle: (x)(Fx v -Fx). And this is just as well.
After all, classical logic is attractively simple and familiar. Further
more, it is the logic that has been extracted from ordinary linguistic
practice, and ordinary language is predominantly vague; so it would
be remarkable if the phenomenon of vagueness were to dictate any
departure from classical logic.6 Moreover, once we have steeled our
selves to apply the law of excluded middle come what may, it is not
additionally counterintuitive to retain also the principle of bivalence:
that every proposition is true or false. On the contrary, its rejection
leads quickly to contradiction, as we saw in the answer to Question
26. Thus we should retain the classical principles if at all possible. But
the urge to back away from them will not be completely relieved until
we have a satisfactory treatment of vagueness that will plainly allow
them to be preserved.
In order to obtain such an account, a useful step is to recognize a
distinction between ordinary truth and determinate truth (Field,
1 986; Wright, 1 987), enabling us to assuage our intuitions by saying
that a proposition in which a vague predicate is applied to a border
line case is not determinately true but might nonetheless be true. A
natural explication of the needed notion of determinacy may be
given by reference to the way in which the meaning of a vague predi
cate prevents us from finding out whether its application in certain
cases would be correct or incorrect: we can say that an object with the
property F-ness is determinately F when there is no such semantic
obstacle to discovering that it is F, and that it is not determinately F
when there is such an obstacle-i.e. when the meaning of the predi
cate 'F' precludes the prospect of our arriving at a stable conclusion
about whether or not it applies. Consider, for example, the ascription
of 'heap' to an unclear case-a little collection of sand particles.
6 It would not be appropriate to restrict the application of classical logic to propo

sitions with a determinate truth value. In the first place, such a strategy would not gen
erally be workable. For sometimes the use of logic is required to see what entails a given
proposition and thereby to discover whether or not it has a determinate truth value.
And in that case, the proposition could be tested for its conformity with logic only by
already presupposing that it will pass the test. And secondly, any restriction of logic to
propositions that are definitely true or definitely false would conflict with the a priori
character of logical laws, since their applicability would become contingent on the
favourable outcome of empirical investigation-the determination of whether we are,
or are not, dealing with a borderline case.
Given the meaning of 'heap', and given the number, size, and
arrangement of the grains, we cannot become inclined to attribute
either 'heap' or 'not heap' to this collection. Therefore
(1 1) I t i s neither determinately a heap nor determinately not a
heap,
even though, assuming classical logic,
( 1 2) It is, or is not, a heap.
Therefore, in light of the minimalist accounts of truth and falsity, we
may conclude that
( 1 3) (It's a heap) is not determinately true and not determi
nately false,
although
( 1 4) (It's a heap) is true or false.
Thus we are able to accommodate the phenomenon of vagueness
without questioning the law of excluded middle, the principle of
bivalence, or the minimalist conception of truth.
It remains, however, to say what it is about a predicate that makes
it vague. What character does the meaning-constituting property of a
vague predicate have to have in order to explain the indeterminacy
that is characteristic of vagueness-the semantically induced impos
sibility of coming to know, in borderline cases, whether or not it
applies? The answer, I would suggest, is that the vagueness of predi
cate, 'F' , consists in the distinctively 'gappy' character of the basic
regularity governing its use. Very roughly speaking such regularities
have the form
(V) 'F ' is applied to things which have underlying property # to
a degree greater than y, and 'not F ' is applied to things that
have # to a degree less that n, and neither is applied to
things that have # to an intermediate degree,
where y is greater than n.7 For in so far as 'F ' is deployed according
7 The basic use regularities of vague predicates are not quite as simple as (V), since
(a) they usually involve several underlying parameters whose values fix whether the
predicate applies, and (b) they engender higher-order indeterminacy (i.e. the bound
aries of the 'clearly F' and the 'clearly clearly F' , etc. are also unclear).
The present account is set within a use theory of meaning (sketched in Chapter 6)
to such a basic regularity, we will indeed be unwilling to apply either
it or its negation to the things that have # to some degree between y
and n, and we will indeed be confident that no further discovery
could help matters. This account of vagueness explains its main
symptom-namely, our irremediable inability to decide in certain
cases whether or not a term applies-without suggesting any depar
ture from excluded middle or bivalence.
Any decent theory of vagueness must also have something to say
about the notorious sorites paradox:
(Sor) 0 grains cannot make a heap;
For any n, if n grains cannot make a heap, then n + 1
grains cannot make a heap either;
Therefore: For any n, n grains cannot make a heap,
which seems to show, on the basis of incontrovertible premises and
impeccable reasoning, that there are no heaps. In so far as we want to
give up neither classical logic nor the view that some things are heaps
and others are not, then the only remaining option-which I think
we must learn to happily embrace-is to deny the second premise.
That is, we must allow that there is some unknown (indeed unknow
able) number, h, such that h grains cannot make a heap but h + 1
grains can. Thus we are allowing that the predicate 'is a heap' has an
extension-albeit an indeterminate one. True, we could not, even in
principle, discover the extension. In particular, we could never know
the fact of the matter as to whether our little pile is a heap. For, as we
have seen, such knowledge is precluded by the very meaning of the
word-by its being vague. But why should this be thought odd or
implausible? It is surely only the lingering seductiveness of
verificationism-an inclination to hold that the existence of a fact
requires the conceivability of knowing it-that gives rise to discom
fort with this situation. 8
according to which the overall use of each word is explained by means of some basic
regularity in its use, and the word means what it does in virtue of being governed by
that basic regularity. For a more thorough treatment of vagueness from this point of
view see Horwich, l 997b. See also Tim Williamson ( 1 994), who also retains excluded
middle and bivalence in the context of vagueness, but who offers a very different
account from mine of why knowledge is impossible in borderline cases.
8 A further source of discomfort with the idea that a vague predicate might apply
(albeit indeterminately) in a borderline case is that there could be no explanation, in
terms of its use, of why it applies. However (as I argue in Meaning), it is a mistake to
expect any such explanation. Even though the meaning of each word is constituted by
I should stress that the above treatment of vagueness and of the
sorites paradox has substantial merits, and should not be viewed
merely as a necessary, but rather unwelcome, outgrowth of minimal
ism. In the first place, it is not the case that minimalism dictates our
solution. The situation, rather, is that minimalism primes us to
expect-what could well be recognized independently-that the
problems of vagueness are simply not addressed by an account of
truth. Once this is appreciated, we can see that the real choice is
between the abandonment of classical logic and the abandonment of
verificationism. My sentiments in favour of the latter option are inde
pendent of minimalism. That response succeeds in preserving the
simple and well-entrenched principles of excluded middle and biva
lence, offers a positive analysis of vagueness, proposes a way out of
the paradox, and gives an explanation of why that way out might
have seemed so counterintuitive. Thus our approach to indetermi
nacy is not just a byproduct of minimalism, and has a great deal of
plausibility in its own right.
The rejection of minimalism would be of no help whatsoever in
dealing with these matters. For, as we have seen, the only genuine
alternative to what is proposed here involves the rejection of classical
logic-which is a move not made any less difficult by the adoption of
a non-minimalist theory of truth.9 Moreover, the semantic principle
its use, there need be no explanation of why a given use should provide a predicate with
the particular meaning (and hence extension) that it has.
Frank Jackson has expressed a related concern. He thinks it would be a violation
of our sense of symmetry (i.e. the principle of sufficient reason) to suppose that one of
the pair of propositions 'It is a heap' and 'It is not a heap' is a fact and the other isn't,
when what we have is clearly a borderline case. However, although I don't doubt that
symmetry is appealing, it seems to me that we would rather accept that the world is dis
appointingly asymmetrical, than abandon classical logic.
9 It is far from clear what the new logic would be. Putnam (1 983) suggests intu
itionistic logic. However, a 'vagueness logic', in so far as it is motivated by the idea that
'indeterminate truth' is impossible, would need to renounce the inference from the
conditional
k is F -7 k is determinately F
to its contrapositive,
k is not determinately F -7 k is not F.
Otherwise, in a case where k is evidently a borderline case, i.e.
k is not determinately F and not determinately not F,
it could be inferred that
(E) (p ) is true iff p,
which binds our logic and metalogic, and which prevents us from
combining classical logic with the thesis that neither (A) nor (not A)
is true, is by no means peculiar to minimalism. Just about all
accounts of truth subscribe to it. So it certainly cannot be suggested
that by adopting minimalism we are depriving ourselves of an easy
solution to the problems of indeterminacy.
Finally, it is perhaps worth noting how certain superficially differ
ent approaches to vagueness might be partially accommodated to the
present point of view. Consider, for instance, an idea of Kit Fine's
( 1 975), endorsed by Dummett ( 1 978), that a sentence containing
vague expressions is absolutely TRUE just in case it is true relative to
every admissible way of making the vague terms precise. If we take
'TRUE' to mean 'Determinately true' then this idea becomes a per
fectly consistent and illuminating elaboration of the minimalist view
sketched above. But if it is assumed that nothing can be true unless
TRUE, as Fine and Dummett are inclined to say, then the result is
disaster. For we can infer that neither (A) nor (not A) is true, which
implies a contradiction. So the former interpretation is clearly prefer
able. Another strategy for dealing with vagueness has been to invoke
infinitely many TRUTH VALUES : all the real numbers from 0 to 1 .
The idea is that when a vague predicate, 'F', is applied to a borderline
case, n, the resulting TRUTH VALUE is somewhere in the middle
between 0 and 1 ; that the conditional 'F(n) -7 F(n + 1)' has a TRUTH
VALUE just less than 1 ; and that the TRUTH VALUE of the con
junction of many such propositions gradually decreases as the num
ber of conjuncts increases-reaching 0 when we get to '(n)[(F(n) -7
F(n + l )] ' . As before, these ideas can be reconciled with the minimal
ist viewpoint. It suffices to identify 'TRUTH VALUE = 1 ' with
'Determinate truth', 'TRUTH VALUE = O' with 'Determinate fal
sity', and (perhaps) 'TRUTH VALUE of (p ) = x ' with something like
k is not F and k is not not F

-a contradiction. Thus 'vagueness logic' must renounce contraposition-a principle
which is perfectly acceptable from the intuitionistic point of view. Moreover, there is
no reason for a 'vagueness logic' to quarrel with the principle of double negation elim
ination,
k is not not F ---+ k is F,
which is intuitionistically unacceptable. Thus intuitionism is not the logic of vagueness.
'The probability of (p ) is x ' . So far, so good. Difficulties arise only if
the new TRUTH VALUES, 1 and 0, are identified with truth and fal
sity. But this step can and should be resisted. In order to eliminate the
paradoxes of vagueness it suffices to appreciate, first, that a notion of
determinacy should be invoked to articulate the nature of vagueness;
and second, that this will make it unnecessary to tamper with any of
the entrenched laws of logic and semantics, and will enable us to
accept with equanimity the falsity of the sorites' main premise.
29. There is a su bstantive issue in meta-ethics as to

whether evaluative utterances pu rport to assert
truths or whether they are merely expressions of
feeling; but this question would be trivialized by
m i n imalism.
There has indeed been a tendency for ethical emotivists (also known
as 'non-cognitivists' and 'expressivists'), to want to use the notion of
truth value to distinguish 'genuine descriptions' from syntactically
similar sentences whose linguistic role is arguably non-descriptive.
And this practice certainly is at odds with a minimalist perspective
from which ( 1 ) ethical pronouncements express genuine propositions
(as we shall see in the next section), for they form 'that-clauses' (e.g.
'that honesty is good') which function in the normal way to designate
objects of assertion, belief, etc.; and (2) ethical propositions provide
perfectly good and useful instances of the equivalence schema
instances which are needed to formulate generalizations (e. g. in logic)
that cover such propositions.
However, the moral here is not that minimalism and emotivism
are incompatible, but that emotivism should be reformulated. For a
minimalist might happen to accept the emotivists' central insights:
namely, that the function and assertibility conditions of certain ethi
cal claims are fundamentally different from those of empirical
descriptions, and that an appreciation of the difference will help to
resolve philosophical problems surrounding the notion of an ethical
fact. My point is neither that this position is entailed by minimalism;
nor that it is correct; but that it need not, and should not, be formu
lated in such a way as to preclude the minimalist conception of truth.
It should be articulated, rather, as a view about the unique character
o f ethical propositions. More specifically, the emotivist might
attempt to characterize the nature of ethical propositions by main
taining, very roughly speaking, that the meaning of 'X is good' con
sists in the fact that it is asserted by someone when and only when he
wants X (which is not to say that 'X is good' means 'I want X'). This
sort of account (suitably refined) has some claim to being able to dis
solve some of the epistemological problems of ethics and explain
why certain ethical beliefs have motivational force. Thus the essential
character of emotivism might be captured without having to ques
tion the existence of ethical propositions, beliefs, assertions, etc., and
without having to deny that they satisfy the usual logical and meta
logical principles. I o
10
The bearing of minimalism on the formulation of emotivism is debated in a
series of papers by myself ( 1 993, 1 994), Michael Smith ( 1 994a, 1 994b), and Jackson,
Oppy, and Smith ( 1 994).
6
Propositions and Utterances
30. P ropositions are highly d ubious entities. It is unclear

what they are supposed to be, and thei r very existence is
controversial . Would it not be better, therefore, to develop
a theory of truth that does not p resuppose them-by
assuming, for example, that utterances are the
pri mary bearers of truth?
According to the advocate of propositions, whenever anyone has a

belief, a desire, a hope, or any of the so-called propositional atti
tudes, then his state of mind consists in there being some relation
between him and a special kind of entity: namely, the thing that is
believed, desired to be the case, hoped for, etc. Thus if Oscar believes
that dogs bite, this is alleged to be so in virtue of the obtaining of a
relation, believing, between Oscar and a certain proposition: namely,
that dogs bite.
The considerable merit of this theory is that it appears to provide
an adequate account of the logical properties of belief attributions
and the like. We are inclined to infer from
(1) Oscar believes that dogs bite
that
(2) There is something Oscar believes,
and from
(3) Oscar doubts that it will rain
and
(4) Barnaby is saying that it will rain

Propositions and Utterances 87
that
(5) Oscar doubts what Barnaby is saying.
Such inferences may be subsumed under familiar logical rules pro
vided that apparently singular terms such as
( 6) what Barnaby is saying
and
(7) that it will rain
are taken at face value to refer to a species of object-to be called
'propositions ' .
I t might well b e countered, however, that such evidence should
not be taken as conclusive. After all, syntax is not an infallible guide
to semantic structure. Indeed, one of the central concerns of this
book-whether the truth predicate does or does not stand for some
sort of property---d erives from this well-known fact. And there are
many less controversial examples: e.g. 'Jones's sake', 'The average
man', 'Smith's appearance' . The last of these examples is particu
larly close to the issue at hand. There is a relation of resemblance
between certain things, but this relation is not transitive: it may be
that A looks like B and B looks like C, but A does not look like C.
Given this failure of transitivity, we cannot suppose that there is a
realm of entities, appearances, such that every object possesses
exactly one of them, which it shares with all the objects it resembles;
for that supposition would imply transitivity. Thus although we do
express resemblance claims by saying that certain things have the
same appearance, it turns out on closer scrutiny of the logical prop
erties of such claims that their structure cannot be what would seem
most natural given their syntactic form: namely,
(8) (3x)(x is the appearance of A and x is the appearance of B).
Similarly, it should not be taken for granted that our ordinary talk of
'what Oscar believes' is to be construed at face value as referring to a
special kind of entity. 1
1 Particular grounds for doubt lie in the possibility that translation is affiliated
with the notion of resemblance (roughly, via the identification of intertranslatability
with resemblance in use), and is infected with its intransitivity. If this were so, then we
could not suppose there to be a realm of entities such that every declarative utterance
88 Propositions and Utterances
In order to clarify this issue, let us step back for a moment from
the case of propositions and consider more generally why it is that
we impute logical form in the way that we do. The logical forms of
the sentences in a language are those aspects of their meanings that
determine the relations of deductive entailment holding amongst
them. Let us imagine a body of sentences characterized by their con
cern with a certain range of phenomena; and suppose that we have
mounted an investigation into the relations of deductive entailment
that hold amongst these sentences. Suppose that the results of our
investigation suggest an attribution of logical forms having the
implication that some of the sentences will clearly entail the exis
tence of entities of a certain type-call them 'Ks'. Suppose, finally,
that we believe that some of those sentences express truths. Taken
together, these considerations would provide a basis for thinking
that things of type K exist. But how powerful are these reasons?
Under what circumstances, if any, would it be right to resist them
to maintain that despite the aforementioned evidence there are really
no such entities as Ks?
Of course, one thing that would justify resistance would be the
discovery that our assignment of logical forms is unsatisfactory.
Suppose we found another way of doing it which gave a more com
plete representation of our inferential practice, and which involved
no commitment to Ks. In that case our prior ground for believing in
them would be completely undermined.
Another possible source of resistance would be the discovery of
non-philosophical arguments for the conclusion that Ks don't exist.
In calling these arguments 'non-philosophical' I have in mind that
they would come from inside the field to which the statements in
question belong. Consider, for example, arguments within physics
that the aether does not exist, or arguments within zoology against
the Loch Ness Monster. These are evaluated with respect to the same
canons of justification that govern the original body of statements.
If they are deemed acceptable then the result is a revision in our sci
entific beliefs; but no change is called for in the logical forms that we
attribute to them.
is associated with exactly one of these entities, sharing it with all the other utterances
with which it is intertranslatable. Thus the existence of propositions is put in question.
This problem is addressed directly in the answer to Question 32.
Finally, the most philosophically interesting case is that in which
general philosophical considerations motivate a disinclination to
postulate Ks. It might be argued, for example, that only material
objects exist and Ks would not be material; or that sheer ontological
parsimony requires us to do without Ks if at all possible; or that,
given the nature of Ks, knowledge about them could not be squared
with otherwise attractive epistemological theories; or that we would
be unable to answer basic questions, such as 'What are Ks?', 'When
is K1 the same as K2?', 'Where are Ks?', to which we would have every
right to expect answers if Ks really existed.
There are three types of response to any such argument: (a) we
may regard it as fallacious, and proceed to explain how this is so; (b)
we may find it persuasive, accept that Ks don't exist, and conclude
that a certain body of what we used to believe is mistaken; or (c) in
order to preserve our earlier scientific beliefs and still be able to
accept the conclusion that Ks don't exist, we might abandon the
account of logical forms that involves commitment to Ks and
replace it with one that doesn't.
Of these alternatives it seems clear that option (a) is always best;
for the arguments that it asks us to reject are extremely weak in the
first place. Often they involve barefaced overgeneralization of the
following sort. First, material objects are taken to be paradigm
examples of what exists; secondly, certain prominent properties of
such objects are identified; thirdly, it is inferred that only entities
with these properties could exist; fourthly, it is noticed that Ks would
not have them; and finally, the conclusion is drawn that Ks cannot
exist. Evidently no great conceptual strain is involved in rejecting
such arguments, which beg the whole question in their first premise.
So option (a) is quite acceptable. The other alternatives, however,
exhibit some highly undesirable features. Option (b) implies that we
must start denying certain things that we presently regard as cer
tainly true. Option (c) involves the idea that the correct logical forms
are not those that provide a perfectly adequate account of inferential
practice. We are to reject a certain way of articulating our beliefs
solely because it has consequences that are irrationally regarded as
unwelcome, and even though it accords precisely with our
entrenched criteria for being a good logical articulation.
It is easy to see how these general conclusions will apply to the
case of propositions. In the first place, we can suppose that an ade
quate account of the logical forms of belief attribution (and other
so-called propositional attitude statements) involves the supposition
that 'that p' is a singular term. Thus from 'Oscar believes that dogs
bite' we can infer 'Oscar believes something' , and this is assimilated
to the rule of existential generalization. In the second place, we may
assume that some propositional attitude statements are certainly
true. For example, it is certainly true that Einstein claimed that mat
ter is a form of energy. Thirdly, we should take these assumptions to
entail that there is an entity, what Einstein claimed, or in other words,
the proposition that matter is a form of energy. And fourthly, we
should not be troubled by an inability to say where this entity is, or
what it is made of For such questions are not appropriately asked of
things like propositions. Similarly, it should not trouble us that
propositions will not enter into causal relations with us, so that our
knowing anything about them will violate the 'causal theory of
knowledge' . For this theory derives its entire prima facie plausibility
from the sort of blatantly question-begging overgeneralization men
tioned above-in this case, from the unjustified presupposition that
everything is known in the same way that physical objects are
known.
I conclude that a compelling argument for the existence of propo
sitions may be built on the premise that they participate in an ade
quate account of the logical forms of belief attributions and similar
constructions. Moreover, the required premise appears to be correct.
Therefore, despite their peculiarities, we should not balk at proposi
tions and should not object to their use in a theory of truth.
31 . The case for propositions assu mes the adequacy of a

certa i n logical analysis of belief-one that construes the
state of belief as a relation between a person and a kind
of entity, the content of the belief. But this assumption
is plagued with fam i liar d ifficu lties and appears to
be mistake n .
The main alleged difficulties have t o d o with a dispute about the

nature of propositions-a dispute which goes back to Frege and
Russell. Russell ( 1 903) claimed that a proposition consists of the
very objects it is about. For example,
(9) the proposition that Hesperus is visible
would be made up of the object, Hesperus, and the property of being
visible. If Russell was right, then, since 'Hesperus' and 'Phosphorus'
have turned out to be two names for the same planet, then
( 1 0) the proposition that Phosphorus is visible
is the very same proposition as (9). But how, in that case, do we
account for the fact that someone may be aware-as we say, de
dicto-that Hesperus is visible and not that Phosphorus is? Frege
( 1 892) had already posed this question, and answered it by main
taining that the proposition expressed by a sentence (what he called
a 'thought') is composed of the senses, rather than the referents, of
its constituent words. So (9) and ( 1 0) would be different from one
another. But if he was right then so-called de re beliefs become prob
lematic. For if, in that sense, someone believes, regarding
Phosphorus, that it is visible, it does follow, given the identity of
Phosphorus and Hesperus, that he has that belief about Hesperus
too.
One way of dealing with these problems is to acknowledge the
existence of both Fregean and Russellian propositions-the first
being the objects of de dicto belief, and the second of de re belief.
The idea is that all propositions have some compositional structure
(e.g. $(x), @(x,y), (x)(3y)($(x) � @(x,y)), etc.), and that each singu
lar position in such a structure may be filled either by a Fregean
sense or by the referent of such a sense. Thus there are pure,
Fregean, abstract propositions, in which a compositional structure is
filled only with senses; there are pure, Russellian, concrete proposi
tions, in which each singular position in a compositional structure is
filled only with referents; and there are mixed propositions, in which
some of the singular locations are occupied by senses and the others
by objects. For example, corresponding to the sentence
(1 1) Hesperus is identical to Phosphorus
there is an abstract proposition which consists of the senses of 'is
identical to', 'Hesperus', and 'Phosphorus' embedded, in that order,
in the structure '@(x,y)'; the concrete proposition consisting of the
referents of 'Hesperus' and of 'Phosphorus', and the property of
identity, embedded in that structure, and two different mixed propo
sitions. Therefore,
( 1 2) Raphael believes that Hesperus is Phosphorus
has four readings 2-one for each of the propositions to which one
might be saying that Raphael is related. Disambiguation is achieved
(in English) by use of the qualifier 'of X' for any of the propositional
constituents that are intended to be referents rather than senses.
Thus, if one says
( 1 3) Raphael believes of Hesperus that it is Phosphorus,
the object of belief is the mixed proposition in which the 'x'-position
in '@(x,y)' is filled by the object, Hesperus, and the other positions
are filled with senses. And if one attributes the belief without any
such qualification, what is conveyed is a de dicto belief-a belief
whose object is the purely abstract proposition.
Moreover, it seems plausible to suppose that in order to have a de
re belief of something, it is necessary to have in mind a special way of
conceiving of the thing-a way that is sufficiently intimate to consti
tute 'knowing what the thing is' . 3 Therefore a certain relationship
exists between de dicto and de re belief. In order for someone to
believe a proposition that is not entirely abstract (i.e. in order for him
to have a de re belief), two things are required. First, he must believe
some wholly abstract proposition from which the other may be
derived by replacing some of the senses with their referents. And,
secondly, the believer's conception of these referents must be such as
to qualify him as 'knowing what they are'. Admittedly the second
condition is vague and context-dependent. But this is entirely appro
priate, since the set of circumstances in which de re belief may prop
erly be attributed is also vague and context-dependent.
It seems, therefore, that prima facie difficulties in what would
otherwise appear to be the most straightforward construal of belief
attributions may be overcome. Beliefs should indeed be analysed as
relations between persons and propositions. Thus our practice of
belief attribution commits us to the existence of propositions.
2 I am assuming (a) that the referents of only the singular terms can be what de re
beliefs are about, and (b) that the Russellian property of F-ness is the same thing as
the Fregeau sense of 'F'. For elaboration of this approach-including discussion of
de se attitudes and complications arising from context-sensitivity-see my Meaning
( 1 998), ch. 3 .
3 See Hintikka, 1 962, Kaplan, 1 969, and Quine, 1 977, fo r discussion o f this sort of
requirement for de re belief.
32. 'The proposition that p is true iff p' may be thought to

captu re our conception of truth only if truth is not al ready
presupposed in the very idea of a proposition. But this
req u i rement may wel l be violated . For a central component
of the notion of a proposition is lodged i n the statement of
identity conditions for propositions-the conditions for two
utterances to express the same proposition. But this is an
idea one might plausi bly explain in terms of the inter
translatability of the utterances, which , i n turn, m ust be
construed as thei r havi ng the same truth conditions.
And if the concept of truth is needed to say what
propositions are, then a theory of truth can not take
propositions for granted .
A complete treatment of this issue would go beyond the scope of

this essay. However, I agree that our having the concept of proposi
tion presupposes that we can make sense of intertranslatability
(interpretation) claims of the form
( 1 4) u expresses the same proposition as my present utterance
of 'p',
which, taking for granted that
( 1 5) My present utterance of 'p' expresses the proposition that p,
will yield
( 1 6) u expresses the proposition that p.
Therefore, it is incumbent on me to say something about the notion

of translation on which I am relying-at least enough to quell the
suspicion of circularity.
The general conception of meaning and translation to which I
wish to appeal is the so-called 'use theory' articulated in various
forms by Wittgenstein ( 1 953), Sellars ( 1 954), Quine ( 1 960), Harman
( 1 982, 1 987), Peacocke ( 1 986, 1 992), and many others. Roughly
speaking, the right translation between the words of two languages
is the mapping that preserves basic patterns of usage-where usage
is characterized non-semantically, in terms of circumstances of
application, contribution to the assertibility conditions and inferen
tial role of containing sentences, etc. What counts as the 'basic
patterns' are those regularities that (in conjunction with other fac
tors) best explain overall usage. For example, the basic use of the
word 'red' in English is arguably that it is applied to observably red
things and unapplied to observably not-red things; the basic use reg
ularity governing the word 'and' is arguably that we are disposed to
infer 'p and q ' from 'p' and ' q ' , and vice versa; and the basic use of
'true', as we have seen, is that we are inclined to accept instances of
'The proposition that p is true iff p'.4
Thus the concept of translation may be explicated, as required by
minimalism, in entirely non-truth-theoretic terms. However, this will
not quite suffice for an account of propositional identity. For in
order to determine which of my present utterances expresses the
same proposition as an utterance of u, we must not only translate u
into English but also take into account any differences between the
contexts in which u and my utterance are produced. For example
'J'ai faim' uttered by Pierre at noon yesterday expresses the same
proposition as my now saying 'Pierre was hungry at noon yesterday'.
This is so in virtue of the fact that 'J'ai faim' translates into 'I am
hungry', which is then adjusted for contextual variation by recourse
to a fixed set of rules:
'I' becomes a term 'a', such that a is the speaker;
'now' becomes a term ' t ' such that t is the time of the utterance;
'here' becomes some term 'l ' such that l is the place of the utter-
ance;
4 For a full articulation and defence of the use theory of meaning see my Meaning
( 1 998). This account bears some resemblance to the 'conceptual role semantics' artic
ulated by Field ( 1 977) and Block ( 1 986). A vital difference between them, however, is
that according to their view the meaning of a word is identified with its internal role,
whereas according to the use theory, as I understand it, the meaning of a· word also
includes its use in relation to the external world.
For certain terms it may be objectively unclear precisely which regularities for
their use are explanatorily fundamental. However, this is by no means fatal as far as
propositions are concerned. For if there is no determinate fact of the matter as to
whether two utterances are intertranslatable, then there is no determinate fact of the
matter as to whether they express the same proposition. But, as we saw in the answer
to Question 28, such facts may be indeterminate but none the less exist. Indetermin
acy is found throughout our conceptual scheme-whether we are talking about
material objects, persons, societies, species, or whatever-and so its presence in the
domain of propositions too should be no cause for sceptical concern.
'that F' becomes a term 'd' such that d is the F to which the
speaker was attending;
. . . etc.
Note that the rules of context adjustment do not invoke the concept
of truth. Moreover, as we have seen, the primary determinant of
propositional identity-namely, translation-is explicable in terms
of 'identity of use', where 'use' is characterized in terms that do not
cite truth. Consequently, propositions are not grasped via the notion
of truth; and so we are free to suppose that this notion is entirely
captured by the equivalence principle. 5
33. But the 'use theory' of meaning impl ies that p roposi
tions don't exist. For if translation is a matter of resem
blance i n use, then it is not a transitive relation , and so
there can be no such things as 'what i ntertranslatable
utterances have i n common'. (Harman , 1 973)
Formulated more carefully, this argument against propositions pro

ceeds by comparing the following three principles:
(A) Two utterances express the same propositions just in case
they are intertranslatable.
(B) Utterances are intertranslatable just in case they have corre
sponding constituents with a similar use.
(C) There are words, x , y , and z, such that the use of x resem
bles the use of y, which resembles the use of z; but the uses
of x and z do not resemble one another.
These principles are inconsistent with one another. For the intransi
tivity of resemblance expressed in (C), combined with the use theory
of translation expressed in (B), together entail the intransitivity of
translation. But, combining this with (A), which says that inter
translatable utterances express the same propositions, we can infer
5 For further discussion of this point-specifically the relationship between under

standing a sentence and knowing its truth conditions-see the answer to Question 2 1 .
the possibility that u and v express the same propositions, v and w
express the same propositions, but u and w do not express the same
propositions-which is absurd. Thus, given that relations of resem
blance clearly are intransitive, the use theory of translation (as
articulated in (B)) entails that the identity conditions for proposi
tions are incoherent, and therefore that propositions don't exist.
My response is to modify premise (B). The use theory of meaning
is indeed correct. But the argument shows that premise (B) gives a
misleading rendering of it. What we should say instead is that two
words are intertranslatable only when they have the same basic use
and not merely similar uses. But we must acknowledge that the
observable facts of linguistic behaviour, although providing the only
possible guide to the existence of this relation of sameness, may not
always be sufficient to establish it objectively. For there may be alter
native but equally good ways of explaining the overall deployment
of a word in terms of alleged basic regularities in its use. In that case
it will be indeterminate which of these patterns of use is really basic.
Consequently it may be indeterminate whether or not two words
conform to the same basic patterns of use. Thus claims of inter
translatability based on sameness of use are not always determinate.
Two people could have equally good grounds for divergent claims
and there be no way to settle the matter. So the relation between
translation (or 'same use') and observed linguistic behaviour is like
the relation between baldness and amount of hair, or between
poverty and net assets. In each case attribution of the former notion
can be based only on information about the latter; however, there is
often a degree of indeterminacy-a semantically caused, irremedi
able ignorance regarding the precise conditions in which the former
concept applies. None the less, there does exist a fact of the matter.
Either two words are properly intertranslatable, or they are not
even though it may be impossible to say which is so. 6
I have been trying to rebut an argument designed to show that
translation is intransitive-a conclusion that would undermine the
existence of propositions. However this rebuttal does not prove that
translation is transitive. Of course, one way of showing this is to
derive it from the existence of propositions. But if one is nervous
about propositions precisely on the grounds that translation might
6 See the answer to Question 28 for a discussion of how vagueness and

indeterminacy are treated from the minimalist point of view.
be intransitive, then this strategy is no good. An alternative line of
thought offering some reassurance goes as follows.
It is implicit in the existence of propositional attitude construc
tions that a person may adopt a given attitude no matter what the
natural language is in which he would express it. We do not regard it
as a necessary condition for the truth of
(17) Florence believes that snow is white
that Florence has our language available to her. Moreover, we do not
suppose that this fact about what Florence believes, if it is a fact,
depends on our having characterized it, or on our being able to char
acterize it in the particular way that we did-i.e. in English. One and
the same propositional attitude of Florence's may be equally well
described in other languages: for example, with
( 1 7-f) Florence croit que la neige est blanche.
But this sort of translatability of propositional attitude attributions
implies that translation is transitive. For suppose Florence asserted
in a third language some words whose translation into English is
( 1 8) Snow is white,
and suppose that on that basis we English speakers assert ( 1 7). If
translation were intransitive then there would be no guarantee that
( 1 8-f) La neige est blanche
-which is the translation into French of the translation into English
of Florence's words-is the translation into French of Florence's
words. So there would be no guarantee that ( 1 7-f) manages to char
acterize the same propositional attitude as ( 1 7) does. Thus the
assumption that our attributions of propositional attitudes are
translatable in the normal way presupposes that translation is transi
tive.
More formally, suppose u and A are particular utterances in dif
ferent languages, L(u) and L(A), and that A attributes a certain con
tent to u by means of the sentence 'v'. Utterance A says in effect that
' v' expresses in the language L(A) what u expresses in L(u) . Or, in
other words:
(I) A is true iff the translation of u into L(A) is 'v'.
Suppose, moreover, that we would translate the sentence 'v'-as it is
used within L(A)-as the sentence 'p' . Then, from our point of view,
A expresses the statement
(B) u expresses the statement that p.
Therefore, A is true iff u expresses the statement that p. In other

words,
(II) A is true iff our translation of u is the same as our transla-
tion of ' v ' from L(A).
Combining (I) and (II)-the two statements of A 's truth condi
tions-we can infer that u and ' v ' -in-L(A) are intertranslatable just in
case their translations into our language are the same. Therefore,
translation is transitive. It can indeed define the equivalence classes
with which propositions may be correlated.
34. Many philosophers would agree that if p ropositions

exist then propositional truth would be covered by
something l i ke the equ ivalence schema. But they m ight
stil l maintain that the truth of an utterance consists in its
'correspondence with reality' , or some other substantive
thing. Th us, it is for utterances that the deflationary
account is controversial , and this position has received
no elaboration or defence. (Field, 1 986)
Fair enough. So let me now indicate how minimalism deals with the
truth of utterances. The initial deflationary impulse is to say that
(D?) Any utterance of the sentence-type 'p', is true iff p.
But the trouble with this idea is that different utterances of the same
type (e.g. 'I am hungry', 'Banks have money', 'Mary's book is on
that table') have different truth values depending on the circum
stances in which they are produced and on what is meant by them.
Consequently, it is simply false that
( 1 9) A ll utterances of 'I am hungry now' are true iff I am hun
gry now,
and similarly for the other examples.
To avoid this difficulty we need a restricted form of the disquota
tional principle-one in which the way that 'p' is construed when it
is mentioned on the left-hand side is the same as the way it is con
strued when it is used on the right-hand side. In other words, we can
endorse instances of the schema
(D ') (u E 'p') � (u is true H p),
provided that they-specifically the right-hand sides of the bicondi
tionals-are construed with respect to the same pertinent contextual
variables as u itself In other words, an instance of the disquotational
schema holds if it is asserted in a context that is not relevantly dif
ferent from the context of the utterance whose truth is in question.
Implicitly attempting to ensure that this condition is satisfied,
Quine ( 1 970) suggests that we restrict instantiation of the disquota
tional schema to so-called eternal sentences-sentences whose
tokens exhibit no context-sensitivity because they don't contain ele
ments like indexicals, demonstratives, and ambiguous words. As for
utterances that are context-sensitive, Quine proposes to derive each
of their truth values from the truth values of 'equivalent' eternal sen
tences. For example, a particular instance of
(20) I am hungry
may be true because of its 'equivalence' to the eternal sentence,
(2 1 ) Albert Einstein is hungry at noon on 1 January 1 947.
However, there are a couple of difficulties with this strategy. In the
first place, the problem of context-sensitivity is not really solved
because the same string of words can appear in several languages
and for that reason have different truth values on different occasions.
Thus, strictly speaking, there are no eternal sentences. Moreover, as
soon as one tries to patch things up by inserting reference to a lan
guage in the disquotation schema, i.e.
(D+) 'p' is true in L iff p,
then the theory is undermined. For unless something is said about
what it is for an utterance to be 'in language L', the schema will
become a definition of that notion rather than a definition of truth.
But if we do fill out the modified schema with an account of what it
is for an utterance to be 'in language L', then, in effect, the theory
will be explaining the truth of utterances in terms of the truth of the
1 00 Propositions and Utterances
propositions they express, and so its distinctively disquotational
character will be lost. 7 In the second place, even if we allow Quine's
language-relative notion of 'eternal sentence' , it is unclear that there
would exist enough of them to provide a sufficient basis for the truth
values of all instances of non-eternal sentences. Ambiguity and
context-sensitivity are by no means restricted to indexicals and
demonstratives. Most names, predicates, and quantifiers can also be
construed in various alternative ways. So much so that it is not easy
to find English sentences whose instances must all have the same
truth value. 8
For these reasons I do not follow Quine's approach, but adopt an
alternative way of making sure that instances of the disquotation
schema are construed in the same way as the utterances whose truth
condition they. specify. My strategy is to introduce a kind of quote
name designed to pick out an expression-type on the basis, not only
of syntactic form (i.e. physical character), but also on the basis of
meaning (or, more specifically, the propositional constituent
expressed). Thus '*dog* ' picks out the word-type with a certain
shape and a certain meaning; '*bank*' is ambiguous, since there are
two different types (with the same shape but different meanings) that
it can designate; similarly, '*I am hungry*' is multiply ambiguous,
designating a separate expression-type for each of the limitless num
ber of different propositions it may be used to express (depending on
the speaker and time of utterance) . Thus we can accept
(22) (u E *I am hungry*) --7 (u is true iff I am hungry),
provided the quote-name is understood appropriately. This will be so
if the instance of the syntactic form 'I am hungry' in the consequent
of (22) is a member of the type designated in the antecedent. And
this is achieved by understanding the two tokens in the same way.
Generalizing, we have an inclination to accept any instance of
(D) (u E *p *) --7 (u is true iff p),
when what is put in place of 'p' is given a uniform interpretation.
However, this is not quite enough to capture our concept of truth
for utterances; for it leaves unspecified how we attribute truth to
7 See Putnam, 1988, for a forceful presentation of this point.

8 This point was impressed on me by Dan Sperber, Pierre Jacob, and Frarn;:ois
Recanati. See Searle, 1 978, and Barwise and Perry, 1 98 3 .
utterances i n foreign languages and, in general, t o utterances whose
types are not exemplified in our own current discourse (e. g. John's
utterance yesterday of 'I am hungry'). In order to cover these cases,
we must add a further principle enabling us to 'project' attributions
of truth from our own current utterances onto other, 'equivalent'
ones-where u and our 'p' are 'equivalent' when 'p' is the correct
interpretation of u (i. e. Int(u) E 'p'), and where interpretation is a
matter of translation plus context adjustment, as discussed in the
answer to Question 32. Thus we suppose that
(23) (Int(u) = v) � (u is true iff v is true),
which, together with the disquotation schema (D), yields the more
general schema
(DT) (Int(u) E *p*) � (u is true iff p).
Our grasp of truth for utterances is constituted by our inclination to
accept instances of this schema. Moreover, these instances are the
axioms of the minimal theory of that property.
The minimal theory of truth for utterances is equivalent-mod
ulo two further principles-to the minimal theory for propositions.
The auxiliary assumptions needed to derive either theory from the
other are, first, a specification of the conditions for an arbitrary
utterance, u, to express a given proposition,
(24) u expresses the proposition that p � Int(u) E *p* ,
and, second, a specification o f the relationship between truth for
propositions and truth for utterances,
(25) u expresses the proposition that p � (u is true � the
proposition that p is true).
These two schematic assumptions entail
(26) Int(u) E *p* � (u is true � the proposition that p is true),
which allows us to derive the equivalence schema for propositions,
(E) The proposition that p is true iff p,
from the disquotation schema (DT) for utterances, and also to derive
(DT) from the equivalence schema. Thus, the accounts of truth for
propositions and for utterances are unified and equally deflationary.
Notice; moreover, that a slight modification of the preceding
1 02 Propositions and Utterances
argument allows us to derive a minimal theory of truth for beliefs,
claims, insinuations, etc. Instead of the preceding auxiliary princi
ples governing the relationship between utterances and propositions,
we invoke analogous principles relating utterances and propositional
attitudes, namely,
(24*) u expresses the statement that p H Int(u) E *p*
and
(25 *) u expresses the statement that p � (the belief (claim, sug
gestion, . . . ) that p is true H u is true).
These principles imply:
(26*) Int(u) E *p* � (u is true H the belief (etc.) that p is
true) .
And, as before, given the interpretation of u we can now go easily
either from the disquotation principle (DT) to
(EB) The belief (etc.) that p is true iff p,
or the other way round.
Ordinary language suggests that truth is a property of propo
sitions, and that utterances, beliefs, assertions, etc., inherit their
truth-like character from their relationship to propositions.
However, the above derivations show that this way of seeing things
has no particular explanatory merit. The truth-like conception for
each type of entity is equally minimalistic. And by assuming any one
of them we can easily derive the others.
In so far as our aim is merely to understand our conception of truth,

and not to promote some allegedly better one, then I think we have
no choice but to acknowledge that truth is primarily attributed to
what we believe, question, suppose, etc. That is to say, truth is
applied to the objects of our so-called propositional attitudes.
However, there are a couple of influential sources of scepticism
about this article of common sense-an article wholly embraced by
minimalism-and it has been my purpose in this chapter to respond
to them. First, many philosophers feel a certain queasiness about
recognizing that propositions exist. Therefore I sketched the case in
favour of them (based on the logical form of belief attributions) and
tried to undermine the usual counterarguments (which derive from
Propositions and Utterances 1 03
their supposed weirdness and the alleged intransitivity of transla
tion). In the second place, there is a tendency to think that our con
ception of 'proposition' presupposes the notion of truth, so that the
minimalist order of explanation would seem to be the wrong way
round. And to this I replied that we might employ an account of
meaning-hence 'proposition'-in terms of use, which would not
require a prior grasp of truth. This idea was already mentioned in
the answer to Question 22. Finally, for those who are not convinced
by these arguments, I showed that minimalist theories of 'truth' for
utterances and belief-states can be given without making a commit
ment to propositions.
7
The ' Correspondence ' Int uition
35. Is it not patently obvious that the truth or falsity of

a statement is something that grows out of its relations
to external aspects of reality?
This sentiment is what lies behind so-called 'correspondence theories

of truth' . According to one such theory (Wittgenstein, 1 922) the
alleged correspondence is between representations and facts: the
truth of a sentence or proposition is said to derive from the existence
of whatever fact it 'depicts' . Alternatively, there are accounts (e.g.
Tarski, 1 958; Davidson, 1 969) in which the ontological category of
fact is eschewed, and the truth of a whole sentence or proposition is
built directly out of the relations of reference and satisfaction
between its parts and various external objects. Here, the truth of a
thing is engendered by the objects to which its constituents corres
pond.
Admittedly minimalism does not explain what truth is in any such
way. But it does not deny that truths do correspond-in some
sense-to the facts; it acknowledges that statements owe their truth
to the nature of reality; and it does not dispute the existence of rela
tionships between truth, reference, and predicate satisfaction. Thus
we might hope to accommodate much of what the correspondence
theorist wishes to say without retreating an inch from our deflation
ary position. Let us see how this can be done.
It is indeed undeniable that whenever a proposition or an utter
ance is true, it is true because something in the world is a certain
way-something typically external to the proposition or utterance.
For example,
(1) (Snow i s white)'s being true is explained by snow's being
white.
The 'Correspondence' Intuition 1 05
That is to say,
(2) (Snow is white) is true because snow is white.
But these intuitions are perfectly consistent with minimalism. In
mapping out the relations of explanatory dependence between phe
nomena, we naturally and properly grant ultimate explanatory pri
ority to such things as the basic laws of nature and the initial
conditions of the universe. From these facts we attempt to deduce,
and thereby explain, why, for example,
(3) Snow is white.
And only then, invoking the minimal theory, do we deduce, and
thereby explain, why
(4) (Snow is white) is true.
Therefore, from the minimalistic point of view, (3) is indeed explana
torily prior to (4), and so ( 1 ) and (2) are fine. Thus we can be per
fectly comfortable with the idea that truths are made true by
elements of reality. Since this follows from the minimal theory (given
certain further facts), it need not be an explicitly stated part of it.
36. Is it not equally clear that, contrary to m i n i malism,

statements are made true by the existence of facts
to which they correspond?
From the just-acknowledged explanatory relation

(2) (Snow is white) is true because snow is white.
it is not at all obvious that we can infer
(5) (Snow is white) is true because there exists the fact that
snow is white.
After all, the two suggested explanations,
(3) Snow is white
and
(6) The fact that snow is white exists,
1 06 The ' Correspondence' Intuition
report different (though intimately related) states of affairs-and so
something might well be due to one but not the other.
Moreover, there are positive grounds for resisting (5) . For con
sider, to begin with, a Russellian proposition: namely, an arrange
ment of objects and properties. 1 It is hard to see how there could be
any difference at all between such a proposition and the fact corre
sponding to it, since each of these entities would involve precisely the
same arrangement of precisely the same objects and properties.
Thus Russellian facts are just those Russellian propositions that are
true. In other words, Russellian propositions, when true, are identical
to facts, and not merely 'similar' to them; there is no difference
between the truth of the proposition that snow is white and the exis
tence of the fact that snow is white. Consequently there can be no
explanatory relationship between these things, and they can corre
spond to one another only in so far as 'correspondence' is construed
as identity. 2
Turning to Fregeau propositions (which are arrangements of the
senses of words), there are two ways to go. One option is to maintain
that they also, when true, are identical to facts. This idea may well
come from the feeling that, for example, the fact that Phosphorus is
Hesperus is not identical to the fact that Phosphorus is Phosphorus.
Thus we might countenance both Fregeau and Russellian facts
and it will be as hard to distinguish Fregeau facts from Fregeau
propositions as it was to distinguish Russellian facts from Russellian
propositions. Once again we would be excluding the possibility of a
substantive correspondence theory.
Alternatively one might suppose that any Fregeau proposition
and a sentence would presumably be treated in the same way-is
made true by the existence of the corresponding Russellian fact.
This, in effect, is Wittgenstein's ( 1 922) 'picture theory' . 3 The rough
idea here is that a sentence-token (or a Fregeau proposition) consists
of elements arranged in a certain logical form, and the fact that this
is so depicts that there is in reality a Russellian fact consisting of the
1 See the answer to Question 31 for a brief discussion of the distinction between
Russellian and Fregeau propositions.
2 The identity of facts and true propositions was urged by Moore ( 1 927) and has
been recently defended by Jennifer Hornsby ( 1 997).
3 What follows is just one way of reading Wittgenstein's cryptic remarks.
Arguably, the Tarskian approach, discussed in the answer to Question 3 8 , is closer to
what he had in mind.
referents of the elements arranged in the same logical form as in the
sentence. The sentence is true if there is such a fact and false if there
isn't. Formulated more precisely, the theory might look something
like this:
(PT) I. A sequence of terms (or senses), S, refers to a sequence
of entities, 0, when S and 0 have the same length and
the nth member of S refers to the nth member of 0.
II. A sentence (or Fregean proposition), x, corresponds to
a Russellian proposition, y, � there exists an S and an
0 such that
-x is composed of the members of S;
-y is composed of the members of O;
-x and y have the same logical form;
-S refers to 0.
III. y = the Russellian proposition that p -7
(y is a fact � p).
IV x is true � there exists a Russellian proposition, y,
such that
-x corresponds to y;
-y is a fact.
But although these principles are quite plausible and certainly
worth noting, they do not make a good alternative to the minimal
theory. The single respect in which the minimal theory can seem
unattractive is its infinite, list-like character. Sellars ( 1 962) compares
it to a telephone directory. And precisely this feature is preserved in
the picture theory's principle III, which specifies, for each Russellian
proposition, the circumstances in which it will qualify as a fact. As
we saw in the answer to Question 5, there can be no hope of encap
sulating this schema within a finite body of axioms. In addition, a
theory of reference is required; and, as we also saw in the answer to
Question 5, there is reason to think that this too would require infin
itely many axioms. And we have not yet mentioned the need for
theories of what it is for a sentence and a proposition to have the
same logical form, and for a proposition to be composed of a certain
sequence of entities. Thus if we were to trade minimalism for the pic
ture theory we would sacrifice purity and simplicity for absolutely no
benefit. This implies that, although its principles may well be correct,
1 08 The 'Correspondence' Intuition
the picture theory of truth should not be thought to qualify as our
basic theory of truth.
Thus we have failed to find any correspondence account of truth
possessing methodological virtues lacked by the minimal theory. No
doubt one may formulate some interesting, plausible schemata that
relate the concepts of truth, fact, and correspondence. But the con
junction of such schematic principles is best viewed as yielding a
legitimate extension of our theory of truth; it does not provide a
tempting alternative.
37. Certai n cases of representation (e. g . by maps) clearly

i nvolve a correspondence-a structu ral resemblance
to what is rep resented. So is it not reasonable to expect
some such relation in lingu istic representation also?
Perhaps so. However it remains to be shown, for either type of rep

resentation, that such structural similarities are in any way consti
tutive of truth. The minimalist view is that they are not.
The difference between a map (or any so-called 'realistic' repre
sentation) and a sentence of a natural language is that the interpre
tation of maps is more 'natural' -or in other words, less
'conventional' -than the interpretation of sentences. Let me explain
what I mean by this. In the case of certain maps, a representation
that consists in some set of objects (symbols) standing in some rela
tion to one another is supposed to be interpreted as saying that the
referents of those objects stand in that very same relation. For
example,
(7) The fact that point y is on a straight line between points x
and z
expresses
(8) The fact that the place represented by y is on a straight line
between the places represented by x and z.
Thus what is special about 'realistic', 'natural' , 'non-conventional'
systems of representation is that they have syntactic features (e.g.
one thing being on a line between two other things) that refer to
themselves. In other words, facts (7) and (8) share, not merely logical
form, but also a certain pictorial form: namely,
(9) That . . . is on a straight line between . . . and . . . .
This suggests the following account of pictorial truth:
(P) x is a true picture H the referents of x's pictorial elements
exemplify the very relation exemplified by those elements.
But the then needed accounts of exemplification and reference will
be far from unproblematic. For the former would involve the schema
(Ex*) The sequence of objects, (ar . a2 , , an), exemplifies the
. • •
relation R-ness iff R(a i , a2 , , an).

• • •
And, in so far as there is no limit to the number of possible pictorial

elements, the latter would also call for infinitely many axioms. Thus
the proposed theory of pictorial truth is not especially attractive in
comparison with minimalism.
An alternative strategy would be to construct a theory of truth for
pictorial representations that would parallel the theory of truth for
sentences, (PT), given above-the only difference being that in prin
ciple PT-II, we would speak of pictorial forms instead of logical
forms. However, although this is indeed possible, the result would be
just as cumbersome as (PT) turned out to be, and just as unsatisfac
tory compared to the minimalist theory.
So, even for pictorial representations, the best theory of truth is
the minimal theory. This assumes, of course, that the minimal theory
can be applied to pictures; and let us quickly confirm that this is so.
As we saw in the answer to Question 24, the minimalist axioms of
the theory of truth for expressions (as opposed to propositions) are
instances of
(DT) Int(u) E *p* � (u is true iff p).
Therefore the axiom-schema concerning the map
(M) A B C
is
( 1 0) Int(M) E *p* � M is true iff p.
But in fact the English interpretation of M happens to be the sen
tence
(1 1) B is between A and C.
Consequently
( 1 2) M is true iff B is between A and C.
Thus we are able to articulate the difference and similarity between
linguistic and pictorial representation, and show that in neither case
is there any reason to go beyond the minimal theory of truth.
38. The m i n imal theory fails to show how the truth of a

sentence depends on the referential p roperties of
its parts.
Another substantial account of truth that is sometimes regarded as

an elaboration of the 'correspondence theory', is the Tarski-inspired
compositional approach.4 Tarski's account resembles the picture
theory in that the truth of a sentence is explained in terms of the
semantic properties of its constituents. But it differs from the picture
theory in not explicitly referring to facts or to logical structures.
Most importantly, it does not call for a schema, such as PT-III, to
specify for each proposition when it is a fact; instead it offers a non
schematic, finite theory by deducing the truth condition of each
statement from the referents of its parts. Thus, roughly speaking and
simplifying enormously, we get:
(TT) u is true in language L iff
(i) u has the form '$(x)' and there is an object, x, such
that the predicate of u is true-in-L of x and the subject
of u refers-in-L to x;
'
or (ii) u has the form 'p & q and each conjunct of u is true-
in-L;
4 It is unclear that Tarski himself would endorse this construal of his work.
Davidson ( 1 969), however, has argued that it should be seen as vindicating the corre
spondence intuition. Note also that Tarski's own initial theory of truth (Tarski, 1 958)
was not given for languages containing names, and does not reduce truth to reference
but solely to predicate-satisfaction. Nevertheless, the following points apply to it.
The 'Correspondence' Intuition 111
or (iii) u has the form '-p' and what is negated in u is not true;
or (iv) u = . . . ;
which can ultimately be turned into an explicit definition of truth.
Tarski himself was content to show that for certain simple lan
guages, L, he could define a predicate, 'true-in-L', whose extension
would be the truths of L. But someone (e.g. Davidson, 1 969) might
hope to go further and arrive at a general theory of truth by show
ing, even for complex, natural language such as English, that 'true
in-L' could be defined in terms of 'refers-in-L' and 'satisfies-in-L'.
However, there are two basic objections to any such approach. In
the first place, as Davidson concedes, the Tarskian strategy applies
only to those sentences whose logical forms may be represented in
first order logic. Or, in other words, it applies only to those sentences
whose truth values are determined by the truth values of atomic sen
tences. But there is no reason to assume that every conceivable truth
has such a structure. Consider, for example, counterfactual condi
tionals, probability claims, laws of nature, and modal assertions of
various kinds. All of these constructions resist formalization in the
language of predicate logic, and so it is not clear that Tarski's theory
can be made to cover them.
In the second place, as we saw in the answer to Question 5, such a
theory of truth will be adequate only if it is supplemented with
theories of reference and satisfaction; and these require either sub
stitutional quantification, or an unformulable collection of axioms.
So no simplification is achieved by reducing truth to reference and
satisfaction. On the contrary, the extreme conceptual purity and
simplicity of minimalism is thrown away for nothing. The corre
spondence approach promised to rectify a certain perceived defect of
minimalism: namely its infinite, list-like character�its failure to say
what all truths have in common. But it turns out that some of the
terms to which truth is to be reduced are similarly intractable; and so
the initial promise of correspondence is not fulfilled.
In rejecting the idea that the correspondence theory describes the
basic nature of truth, we are denying that truth, reference, and satis
faction are constitutively interdependent on one another. But we are
certainly not trying to suggest that the theoretical principles relating
those properties are incorrect. We are supposing, rather, that such
principles should not be treated as explanatorily basic, but should
each be explained in terms of simple, separate, minimal theories of
truth, reference, and satisfaction. And in fact this can be done very
easily. For example, it is not a part of our basic theory of truth that
we should accept instances of
( 1 3) 'p and q' is true iff 'p' is true and 'q' is true.
However, this is trivially deducible from
(i) 'p and q' is true iff p and q,
(ii) 'p' is true iff p,
and
(iii) 'q' is true iff q.
Similarly, we do not define truth in terms of reference and satis
faction using such principles as
(T/R) 'Fa' is true iff (:Jx)('a' refers to x & 'F' is satisfied by x).
However, this schema is easily explained. For we have
(S ') (x)(x satisfies 'F' iff Fx)
and
(R ') (x)('a' refers to x iff a = x) .
Therefore the right-hand side of (T/R) is equivalent to
( 1 4) Fa,
which, given the disquotational principle, is equivalent to the left
hand side of (T/R),
( 1 5) 'Fa' is true.
Thus, the minimal theory does not preclude the possibility of show
ing how the truth value of a sentence is related to the referential
properties of its parts.
The best theory of truth will be the smallest, simplest collection
of statements that, in conjunction with theories of other matters
such as reference, will enable us to derive everything we believe about
truth. This desideratum points us towards minimalism and away
from theories, such as Tarski's, which are unnecessarily complex,
explanatorily misguided, and foster the misleading impression that
truth, reference, and satisfaction are inextricably intertwined with
one another.
The 'Correspondence ' Intuition 1 13
39. The g reat vi rtue of defi n i n g truth i n terms of reference

is that the account may be supplemented with a natu ralistic
(causal) theory of the reference relation to yield, i n the end,
a naturalistic and scientifically respectable theory of truth .
(Field, 1 972 ; Devitt, 1 984)
The usual line on singular reference goes roughly as follows. Once

upon a time we believed the description theory of Frege ( 1 892) and
Russell ( 1 905). It was held that every time a name is used it is
intended to be synonymous with some definite description. Thus
'Socrates' may, in some idiolect, mean 'the teacher of Plato' . This
idea was refined by Wittgenstein (1 953) and Searle ( 1 958), who
argued that a name, though perhaps not synonymous with any par
ticular definite description, is none the less associated with a certain
cluster of definite descriptions which jointly determine what it
stands for. But, so the story goes, this whole picture of reference has
been decisively refuted by Kripke (1 972) . Suppose, to give one of his
examples, an obscure mathematician, Schmidt, was really the person
who discovered the incompleteness of arithmetic, although Godel
has been unfairly given the credit. In that case, even if someone's
only belief expressed by means of the name 'Godel' were that Godel
discovered the incompleteness of arithmetic, we would none the less
say that his uses of the name do not refer to Schmidt, but to GOdel.
Kripke used such counterexamples to the description theory in order
to motivate an alternative 'causal' picture of the reference relation.
The idea here is that there is a certain type of causal chain connect
ing the uses of a name and the thing it stands for. What remains,
according to this line of thought, is to formulate a definite theory of
reference that would manage to capture this causal picture-but
unfortunately no-one has been able to do it yet.
From the minimalist perspective this failure comes as no surprise.
For truth and reference are so intimately related that the rationale
for a minimal account of truth will equally well motivate a minimal
account of reference. Reference, on this view, is not a complex rela
tion; a naturalistic or conceptual reduction is not needed and should
not be expected.
Notice, moreover, that Kripke's own model contains nothing to
preclude some role for representation in the fixing of a name's refer
ent; and the introduction of names by means of initial modes of
presentation (demonstrative or descriptive) is not given a non
semantic analysis. His central claim is merely that there need be no
reference-fixing characterization in the minds of users of the name.
For as use of the name spreads 'causally' through a linguistic com
munity, the initial modes of presentation might well be lost. But this
is merely an instance of a quite general point about meaning, apply
ing to all words, and not just to names. The point is that the meaning
of an expression is not intrinsic to the minds of individual language
users, but resides in the practices of the linguistic community as a
whole and is properly attributed to an individual on the basis of his
interactive relations with the community (see Burge, 1 979) . Thus
what is 'causal' in Kripke's theory has nothing specifically to do with
names, and the term 'causal theory of names' is quite misleading in
connection with his discussion. One shouldn't think that he gives
even so much as a crude version of a causal theory of reference.
Such a conception becomes even less plausible when it is com
bined, as it often is, with the thesis that names have no meaning (i.e.
no Fregean sense) and that their only possible contribution to the
propositions they help to express is their referent. As we saw in the
answer to Question 3 1 , this position makes it impossible to give a
satisfactory account of de dicto propositional attitudes (which at
least the description theory was able to do) . Moreover, it appears to
be the product of an overreaction to Kripke's arguments. Granted
they show that names do not have the same meanings as definite
descriptions; but this provides no reason to conclude that names
have no meanings at all. We might suppose, rather, that they are
primitive expressions. And there would be no mystery about this.
After all, most predicates are primitive; so why shouldn't names be
as well?
As in the case of singular terms, the search for a causal theory of
predicate satisfaction (for example, by Stampe ( 1 977), Dretske
( 1 98 1 ), and Fodor ( 1 987)) is conducted as if it were perfectly obvious
that there is such a thing and all that remains is to work out the
details. But this obviousness, this sense that from a naturalistic,
scientifically respectable viewpoint such an account must be right,
has its origin in the same linguistic illusion that motivates substan
tive theories of truth. Therefore the continual failure of its advocates
to produce an adequate theory is only to be expected. 5
5 See T. Blackburn ( 1 988) for a discussion of the failed attempts. As he says, 'it is
Paralleling our approach to truth, minimalist accounts of satis
faction and reference would begin with a story about the function of
those notions. Roughly and briefly, one might suppose that just as
truth helps us to talk about propositions that are wholly unarticu
lated, so satisfaction and reference help us to talk about propositions
of which certain parts those corresponding, respectively, to predi
-
cates and singular terms-are unarticulated. Just as truth performs

its function by providing a sentence of our language,
( 1 6) 'X' is true,
which is equivalent to the utterance
( 1 7) X,
which might be in any language, so satisfaction and reference do
their jobs because
( 1 8) satisfies 'F'
and
( 1 9) the referent of 'D'
provide home language equivalents of any predicate, 'F', or singular
term, 'D', whatever their language, and whether or not their transla
tions are available. Moreover, just as truth is a convenient alternative
to substitutional quantification into sentence positions, so satisfac
tion and reference enable us to do without substitutional
quantification into predicate and singular term positions. Thus
instead of
(20) { # } (x)(#x v -#x)
one can say
(20*) Given any predicate and any object, either the object
satisfies the predicate or it doesn't,
and instead of
remarkable that the "new approach" is still so much discussed and cited, when so lit
tle has been done to redeem those enthusiastically penned promissory notes which
marked its inception'.
(2 1 ) {:Jd} (Raphael's belief contains (d) and d = the moon)
one can say
(2 1 * ) The referent of a constituent of Raphael's thought is the
moon,
or
(2 1 * * ) Raphael is thinking about the moon.
Having characterized the non-descriptive role of 'satisfies' and
'refers' , the second stage of a minimalist account of satisfaction and
reference is to specify the theories that would suffice for the perfor
mance of these functions. Paralleling the theory of truth, these
would be, respectively, every proposition of the form
(S) (x)(x satisfies (F) H Fx),
and every proposition of the form
(R) (x)((d) refers to x H d = x).
Given such an explanation of the behaviour and purpose of 'refers'
and ' satisfies', it would be a very surprising coincidence if there were
any unified conceptual or naturalistic reduction of the reference
relation.6
The correspondence conception of truth involves two claims:

( 1 ) that truths correspond to reality; and (2) that such correspon
dence is what truth essentially is. And the minimalist response, urged
in this chapter, is to concede the first of these theses (properly under
stood) but to deny the second. The rationale for this response is that
the minimalistic equivalence biconditionals can easily be supple
mented with characterizations of correspondence and fact to show
that, indeed, for any true proposition or sentence, there is a corre
sponding fact. However, we have also seen that there are no advan
tages-and substantial disadvantages-in supposing that this entire
construction constitutes the basic theory of truth. One merely imag-
6 Arguments in favour of the conclusion that these minimalist axioms of reference

and satisfaction are indeed explanatorily basic may be constructed in parallel to the
arguments, given in the answer to Question 14, that MT is explanatorily basic. For a
more extensive deflationary treatment of reference and names see my Meaning ( 1 998),
ch. 5 .
inary benefit is that a correspondence account, by reducing truth to
reference and reference to causation, would leave us with a finite,
naturalistic model-and thereby make the concept of truth sci
entifically respectable. I argued, however, that reference, satisfaction,
correspondence, and fact are just as non-naturalistic, and in need of
infinite, deflationary theories, as truth is. Moreover, such theories are
already perfectly legitimate from a scientific point of view, without
any additional naturalistic reduction. Therefore the extra notions
employed in the correspondence theory do not earn their keep, and
merely introduce unnecessary complexity. Ordinary canons of
explanatory priority dictate that the equivalence biconditionals be
taken as fundamental, and that further characteristics of truth,
including its correspondence properties, be accounted for on the
basis of that theory.
Concl usion
Although the minimalist point of view has been subjected to many

objections, we have seen that it is by no means incapable of dealing
with them. Indeed, from this trial it emerges even stronger than
before. For we now have a clear formulation of the doctrine, a coher
ent set of replies to a very broad range of criticisms, a picture of the
alternative accounts and what is wrong with them, and a sense of the
theory's significant philosophical implications.
Let me end by summarizing the line of thought that has been pur
sued here. We began with just about the only uncontroversial fact to
be found in this domain: namely, that the proposition that snow is
white is true iff snow is white, the proposition that tachyons exist is
true iff tachyons exist, and so on. We then posed, and attempted to
answer, two questions about this general fact. First, what precisely is
it? Can we provide a clear and logically respectable characterization
of it-one which does not rely on improper locutions, like 'and so
on'? Second, is there any further, deeper, non-trivial theory of
truth-some account going beneath or beyond instances of the
equivalence schema? With respect to the descriptive question I
argued that a theory containing all the equivalence theses cannot be
formulated-unless notions are employed that themselves require
unformulable theories. This is because there are too many such the
ses, and also because some of them concern propositions that we can
not yet express. On the question of whether any further theory of
truth remains to be found my answer was a categorical No. Hence the
name 'minimalism'. The justification for this answer fell into a couple
of parts. I argued, in the first place, that all uses of the truth predicate
are explained by the hypothesis that its entire raison d'etre is to help
us say things about unarticulated propositions, and in particular to
express generalizations about them. It transpired in the course of the
book that our apparently deep uses of truth in logic, semantics, and
Conclusion 1 19
the philosophy of science are simply displays of this role. Secondly, I
showed that the performance of this function requires nothing more
or less than the truth of the equivalence axioms. Thus minimalism
perfectly explains all the pertinent facts; and that is its justification.
As for the philosophical import of minimalism, this should no longer
be in doubt. We have seen that many controversies-regarding, for
example, scientific realism, meaning, vagueness, normative emo
tivism, and the foundations of logic-are standardly assumed to
interact essentially with the nature of truth. To the extent that the
notion of truth is clarified and its independence of these problems
established, they can be certain to receive clearer formulation and be
more amenable to resolution.
Postscript
In the years since the first edition of this book was published 1 several
critical reactions to it have appeared. As far as I can tell, none of
these discussions succeeds in undermining the minimalist perspec
tive. Some of them show, however, that various points were poorly
formulated and stand in need of clarification, and some raise new
objections which should be addressed. I would like to take this
opportunity to deal with the most serious of these issues. I will begin
by repeating what I regard as the essence of the minimalist concep
tion of truth, and then go on to discuss some of the recent objections
to it. I shall focus particularly on questions raised by Anil Gupta,
Hartry Field, Mark Richard, Donald Davidson, Crispin Wright,
Scott Soames, Michael Dummett, Paul Boghossian, and Michael
Devitt.
1. What is m i n i malism?
The deflationary attitude toward truth-and the particular variant

of it that I call 'minimalism'-are a reaction against the natural and
widespread idea that the property of truth has some sort of underly
ing nature and that our problem as philosophers is to say what that
nature is, to analyse truth either conceptually or substantively, to
specify, at least roughly, the conditions necessary and sufficient for
something to be true. Amongst the products of this traditional point
of view there is the correspondence theory (x is true iff x corre
sponds to a fact), the coherence theory (x is true iff x is a member of
a coherent set of beliefs), the verificationist theory (x is true iff x is
1 Truth (Oxford: Blackwell, 1 990).

Postscript 121
provable, o r verifiable in ideal conditions), and the pragmatist theory
(x is true iff x is useful to believe). But nothing of this sort has ever
survived serious scrutiny-which comes as no surprise to the
defiationist, who denies that there is any prospect of an explicit
definition or reductive analysis of truth, even a very approximate
one. 2
Defiationism begins by emphasizing the fact that no matter what
theory of truth we might espouse professionally, we are all prepared
to infer
The belief that snow is white is true
from
Snow is white
and vice versa. And, more generally, we all accept instances of the
'truth schemata'
The belief (conjecture, assertion, supposition, . . . ) that p is
true iff p.
But instead of taking the traditional view that an analysis of truth
still needs to be given-a reductive account, deeper than the truth
schemata, which will explain why we accept their instances-the
defiationist maintains that, since our commitment to these schemata
accounts for everything we do with the truth predicate, we can sup
pose that they implicitly define it. Our brute acceptance of their
instances constitutes our grasp of the notion of truth. No concep
tual analysis is called for-no definition of the form
'true' means 'F',
where 'F is some expression composed of terms that are more basic
than the truth predicate. Moreover there is going to be no non
definitional analysis of truth either, however rough and ready-no
substantive discovery of the form
x's being true consists in x's having property F.
2 For a discussion of what is right and wrong about these traditional analyses of
truth, see Chapter I , and Kirkham, 1 992. It may be felt that since almost no concepts
are susceptible to exact reductive analysis (not even 'table' or 'house'), it cannot be of
much interest to deny that truth is. The alleged peculiarity of truth, however, is that
there is nothing to be said-not even very roughly speaking-about what it consists in.
1 22 Postscript
Hence the term 'deflationism' .
This, o f course, i s highly reminiscent o f the old 'redundancy
theory' of Frege ( 1 8 9 1 ) , Ramsey ( 1 927, 1 99 1 ) , Ayer ( 1 935), and
Strawson ( 1 964) : the idea that
The proposition that p is true
means no more and no less than simply
p.
And the redundancy theory is indeed an early form of deflationism.

However, there are various respects in which we have been able to
improve upon this original articulation of the view-leading to the
formulation of deflationism presented and defended in the preceding
chapters, namely, minimalism.
In the first place, as the name given to their doctrine suggests, the
redundancy theorists had nothing much to say about the function of
our concept of truth. But if it really is redundant, why on earth do
we have such a notion? A virtue of minimalism is that it contains a
satisfying response to this question-one that was first proposed by
Quine ( 1 970) 3-namely, that the truth predicate plays a vital role in
enabling us to capture certain generalizations. For we can generalize
The moon is subject to gravity
by saying
Every physical object is subject to gravity.
And similarly we can always obtain a generalization from a state
ment about a particular object by first selecting some kind or type G
(e.g. 'physical object'), and then replacing the term referring to the
object with the universal quantifier 'Every G' . However there is an
important class of generalization that cannot be constructed in this
way: for example, the one whose instances include
If Florence is smiling, then Florence is smiling.
How can we extract the law of logic it instantiates? Or consider
3 The redundancy theorists appreciated the need for some account of the function
of the truth predicate. They tended to maintain that it serves various merely pragmatic
purposes (e.g. to emphasize, as in 'It's true that I didn't see anything'; or to concede,
as in 'It's true that they are a lousy team; but they will win anyway'). But the Quinean
account is clearly superior.
Postscript 1 23
Physicists would like to believe that there are black holes only
if there are black holes.
How are we to generalize this to obtain the epistemological policy it
exemplifies? The solution provided by our concept of truth is to
convert each such proposition into an obviously equivalent one
but one that can be generalized in the normal way. Thus, given the
equivalence of
p
and
The statement that p is true
and
The belief that p is true,
we get, respectively
The statement that if Florence is smiling then Florence is
smiling is true
and
Physicists would like to believe that there are black holes only
if the belief that there are black holes is true,
each of which generalizes in the standard fashion, yielding
Every statement of the form 'If p, then p' is true4
and
Physicists would like to believe only what is true.
From these we can derive (given the truth schemata) all the state
ments we initially wished to generalize-and nothing logically
weaker would suffice. They may therefore be regarded as generaliza
tions of the initial statements. Thus it is indeed useful to have a term
that is governed by the truth schemata-despite their triviality.
There is a clear raison d'etre for a concept having precisely the char
acteristics that the minimalist attributes to truth.
4 'Every statement of the form " . . . " ' can be read as 'Every statement expressed
by a sentence of the form " . . . " '
.
1 24 Postscript
A second and related respect in which minimalism improves on
the original redundancy theory is in not claiming that 'p' and 'The
statement (belief, . . . ) that p is true' have exactly the same meaning.
This claim is implausibly strong; for after all, the words 'true' and
'statement' do have meanings, and those meanings would appear to
be, in some sense, 'components' of the meaning of 'The statement
that p is true' but not of 'p' . Moreover, there is no need to take such
an extreme position. The generalizing function of truth is perfectly
well fulfilled as long as instances of the truth schemata, understood
as merely material biconditionals, are accepted. Relative to that
body of assumptions, any generalization of the form 'All instances
of schema S are true' will entail all the instances of S-which are
precisely the statements we needed to generalize.
Anil Gupta ( 1 993 a) rightly notes that the instances of the gener
alizations that we use the concept of truth to formulate (e.g.
instances such as 'The statement that if Florence is smiling then
Florence is smiling is true') will not say exactly the same thing as what
we wished to generalize ('If Florence is smiling then Florence is smil
ing') unless corresponding instances of 'The statement that p is true'
and 'p' express the very same proposition-which, as we have just
conceded, is not especially plausible. But this point does not under
mine the minimalist story about the function of truth; for, as just
mentioned, that function requires merely that the generalizations
permit us to derive the statements to be generalized-which requires
merely that the truth schemata provide material equivalences. This
isn't to deny that the instances so understood are not only true but
necessarily true (and a priori). The point is that their mere truth is
enough to account for the generalizing function of truth.
It was perhaps an exaggeration to have suggested that the concept
of truth is needed for this generalizing purpose. An alternative strat
egy would be to introduce some form of non-standard (e.g. substitu
tional) quantification, by means of which we could say
(p)(If p then p)
and
Physicists would like it that
(p)(they believe that p, only if p) .
But in that case there would be required a battery of extra syntactic
and semantic rules to govern the new type of quantifier. Therefore,
Postscript 1 25
we might consider the value of our concept of truth to be that it pro
vides, not the only way, but a relatively 'cheap' way of obtaining the
problematic generalizations-the way actually chosen in natural lan
guage.
It is emphasized by some philosophers that the truth predicate is
a device for forming prosentences: just as one might use the pronoun
'he' instead of repeating a name (as in 'John said he was happy'), so
one might say 'That's true' instead of repeating the sentence just
asserted. Evidently this is a perfectly correct observation as far as it
goes. However, there remains the issue of how best to explain it; and
here is where minimalists such as myself part company with self
styled 'prosententialists' such as Dorothy Grover, Joseph Camp, and
Nuel Belnap ( 1 975), and Robert Brandom (1 994). The latter tend to
suppose that 'is true' should be analysed using substitutional
quantification-roughly speaking, that 'x is true' means '(p)(x = ( p)
-7 p )'. In contrast, the former contend that the overall use of the
truth predicate (including its use as a prosentence forming device) is
best explained by supposing that 'is true' is an unanalysable predi
cate governed by the equivalence schema; and that the virtue of such
a predicate is that it allows us to avoid the complexities and obscuri
ties of substitutional quantification.
A third defect of the original redundancy theory lies in its impli
cation that truth is not a property-or, in other words, that attribu
tions of truth (as in 'The proposition that dogs bark is true') do not
have the logical form, 'x is F', that is characteristic of attributions of
properties to objects. A better form of deflationism will reject this
thesis as inconsistent with the logical role of the truth predicate-in
particular with the inferences we make from 'x = that p' and 'x is
true' to ' that p is true', and hence to 'p' . No doubt truth is very dif
ferent from most properties in so far as it has no underlying nature;
but, in light of the inferential role of 'true' as a logical predicate, it is
nonetheless a 'property', at least in some sense of the term. This is a
point to which I return in section 8 .
M y defence o f the minimalist perspective will proceed a s follows.
In the next section I consider various criticisms of the idea that the
truth schemata provide an implicit definition of the truth predicate.
Then, in sections 3 and 4, I take up the question of what sort
of entity we apply it to, and restate the case in favour of pro
positions. Turning, in section 5, from the question of how the word
'true' means what it does to the question of what explains the
1 26 Postscript
characteristics of truth, I reinforce the claim that truth has no under
lying nature and that the basic theory of that property consists in
instances of the equivalence schema, 'The proposition that p is true
iff p'. In sections 6 and 7 I address the objection that the normative
status of truth implies that there must be something more to it than
is appreciated by minimalists. And I conclude by looking back at the
issue of whether or not truth is, in some sense, a property.
2. Does m i n i malism give an adequate account of our

concept of truth?
A central component of the minimalistic picture is its claim about

the particular non-semantic fact that provides the word 'true' with
its peculiar meaning. This is said to be the fact that a certain basic
regularity governs the overall use of the truth predicate-that a cer
tain pattern of behaviour provides the best explanation of our total
linguistic practice with the term. Now, as we have just seen, the reg
ularity which plays that explanatory role is the regularity in virtue of
which the term is useful: namely, our disposition to accept instances
of the truth schemata. Therefore, this is the fact about the truth
predicate that constitutes our meaning by it what we do. 5 More
specifically, the meaning-constituting fact is this: that the explanato
rily basic fact about our use of the truth predicate is our tendency to
infer instances of 'The proposition that p is true' from corresponding
instances of 'p' , and vice versa, whenever (a) each 'p' is replaced with
tokens of an English sentence, (b) these tokens are given the same
interpretation as one another, (c) under that interpretation they
express the content of a statement (a proposition), and (d) the terms
'that' and 'proposition' are given their English meanings.
It might be objected that this condition is not strong enough to
distinguish truth from various other concepts, since there are innu-
5 This claim about the meaning of 'true' is the product of two distinct ideas: first,
the fundamental thesis of minimalism with respect to truth (namely, that the truth
schemata are the explanatory basis of our overall use of the truth predicate); and sec
ond, the use theory of meaning (nameiy, that a word means what it does in virtue of
the basic regularity governing its deployment). Many of the radical consequences of
minimalism depend only on its basic tenet (that the schemata are conceptually basic)
and do not depend upon the further idea that they constitute our concept of truth.
Postscript 1 27
merable predicates, 'F', other than 'true', for which we would accept
instantiations of (F*):
(F*) The proposition that p is F iff p.
Imagine, for example, that 'F' is the predicate 'is true and
1 + 1 = 2'. However, this objection overlooks the fact that, although
there are many predicative expressions that obviously satisfy schema
(F*), it is not generally true, for each such expression, that this is the
full basic regularity governing its use. For example, our knowledge
that 'is true and 1 + 1 = 2' satisfies schema (F*) does not explain our
inclination to deduce, from the premise that this predicate applies to
an object, the conclusion that 1 + 1 = 2. And, in general, if a com
plex predicate satisfies schema (F*), then it will exhibit various use
relations to its constituents; and these facts will not be explained by
our reliance on the schema. Thus, in the case of the truth predicate,
the schema provides the complete basic regularity for its use. In con
trast, the predicates other than 'true' that satisfy the schema are gov
erned by additional regularities. And this is what distinguishes their
meanings from that of 'true'. 6
A separate objection, made by Hartry Field ( 1 992),7 is that the
minimalist account could capture the meaning of 'true' only in its
application to statements that we are able to formulate; for only in
those cases can we supply the relevant instance of the truth
schemata. The idea is that if we take a statement, s, to say, for exam
ple, that snow is white, we might then suppose that the content of
s is true
is more or less
snow is white,
whereas, if we don't know what s means, then we don't know the
content of 's is true' . Now this would indeed count against a theory
that aimed, by means of the truth schemata, to provide a reductive
analysis of each utterance containing the word 'true'. However one
need not, and should not, promise any such reductive analysis;
6 This potential difficulty was brought to my attention by Anil Gupta in October

1 992.
7 Although Field is critical of my particular brand of deflationism, he is one of
the strongest advocates of the deflationary point of view. See, for example, Field,
1 994.
1 28 Postscript
indeed it is a central tenet of minimalism that there is no such thing.
On the minimalist view, we aim to define the truth predicate, not by
providing another expression with the same meaning (nor even by
providing a rule transforming every sentence containing the word
'true' into a content-equivalent sentence without it), but rather by
specifying which property of the truth predicate constitutes its hav
ing the meaning it has; and to that end we must identify the property
that best explains our overall use of the term. In particular, the min
imalist thesis is that the meaning of 'true' is constituted by our dis
position to accept those instances of the truth schemata that we can
formulate. In that way, the word is provided with a constant meaning
wherever it appears-even when ascribed to untranslatable state
ments. And the justification for this thesis is that such a pattern of
acceptance accounts for our entire use of the term-including its
application to untranslatable utterances. Thus we might attribute
truth to an untranslatable statement on the basis, for example, of a
belief in the reliability of the person who made it; and we do, as
Field suggests, need an account of truth that will explain this sort of
attribution; but the minimalist proposal would do so perfectly well.
Mark Richard ( 1 997) has noted that some instances of the truth
schemata will be found, by some people, to be less obvious than
other instances. In particular, propositions involving empty names
('Atlantis is in the Atlantic'), vague attributions ('Smith is bald'),
ethical pronouncements ('Killing is wrong'), and future contingents
('The sea-battle will occur tomorrow') may be affirmed, yet none the
less be regarded as 'non-factual' hence incapable of truth or falsity.
In which case 'p' is not taken to imply the corresponding 'The state
ment (belief, conjecture, . . . ) that p is true' . Consequently, argues
Richard, minimalism is mistaken in maintaining that the concept of
truth is fixed by our disposition to accept instances of the truth
schemata.
However, the essence of minimalism is that certain uncontrover
sial schemata suffice to fix our concept of truth and that there is no
deeper definition taking the traditional, explicit form. Therefore
even if, in light of the above examples, we were to reformulate mini
malism using the qualified schemata
'p' has 'factual content' --7
(The proposition that p is true iff p)
(where the notion of 'factual content' is explicated non-truth-
Postscript 1 29
theoretically, in terms of 'use' or 'conceptual role'), we would leave
intact the core of the minimalistic perspective.
Perhaps some such qualified schemata do give a better account of
certain philosophers' concept of truth; but for the rest of us the
unqualified schemata are accurate. Thus one might be inclined to
suppose that there are two concepts of truth: that the original truth
schemata provide a minimalist account of one of them, and the
qualified schemata a minimalist account of the other. However, the
truth predicate is needed as a device of generalization in all domains
of discourse, including those in which certain philosophers are wary
of attributing truth. Thus the original, unqualified schemata are
what capture the concept we need. It seems reasonable, therefore, to
blame any anxiety regarding those schemata on philosophical con
fusion rather than on deployment of a more restricted concept.
Thus, despite the intuitions cited by Richard, I believe we can stand
by the view that the meaning-constituting use of 'true'-the 'expert'
use-is the explanatory role of our disposition to accept instances of
the unqualified truth schemata.
3. What are the bearers of truth?
In ordinary language what are said to be true are the things that we
believe and that our utterances express-so-called propositions.
Thus, on the face of it, propositions exist, some of them (presum
ably, half) are true, and the correlated truth-like attributes of utter
ances and of acts of believing, asserting, etc. are the complex,
derivative properties, ' u expresses a true proposition' and 'the object
of act x is a true proposition' .
However, it remains to be seen ( 1 ) whether this first impression is
correct-specifically, whether there really are such things as proposi
tions; (2) whether, if correct, this naive picture is consistent with
minimalism-specifically, with minimalism's requirement that
propositions be conceptually prior to truth; and (3) which of the
three truth concepts is really fundamental-specifically, whether it is
not explanatorily preferable to go against the current of ordinary
language and to explicate propositional truth and act truth in terms
of a previously developed account of utterance truth.
On the basis of the inferential behaviour of that-clause
1 30 Postscript
(propositional attitude) constructions (which are sentences like
'John believes that God exists' , 'Mary said that lying is wrong', and
'Charles imagines that he is Napoleon'), a good case can be made for
concluding that they articulate relations (such as believing, asserting,
and hoping) between people and whatever are designated by the con
stituent that-clauses-i.e. propositions. 8 Assuming, as I shall, that
this conclusion is right, there none the less might be no nominalistic
answer to the question 'But what are propositions?'-no theory
specifying, for example, some species of physical or mental entity
with which propositions are to be identified. An inability to give such
an answer does not justify scepticism about the existence of proposi
tions, for the nominalistic presupposition of any such sceptical argu
ment can easily be resisted. It may be supposed, rather, that the
conception we have of any given proposition lies in the circum
stances in which we would take it to be expressed. In other words, it
is by making explicit the grounds on which we maintain such things
as 'John's utterance, u, expresses the proposition that dogs bark' that
we will articulate our concept of proposition.
The minimalist view of truth has two important consequences
regarding the grounds for such interpretive claims. One is that they
not involve truth-theoretic notions (i.e. 'true', 'is true of' or 'refers').
For minimalism is primarily the view that the equivalence schema is
conceptually basic vis-a-vis the truth predicate-and, analogously,
that parallel schemata are conceptually basic vis-a-vis 'is true of' and
'refers' . This implies that the various truth-theoretic concepts be
posterior to the concept of proposition, and therefore that it be pos
sible to possess the concept of proposition without possessing the
concepts of truth, being true of, and reference. And this implies that
that-clause attributions not rest on grounds involving truth-theoretic
notions. For example, minimalism implies that nothing along the
lines of
u expresses the proposition that p = u is true iff p
could be what articulates our conception of proposition.

8 I have in mind the need to accommodate such inferences as: 'Oscar believes that
dogs bark and Barny denies that dogs bark; therefore there is something which Oscar
believes and Barny denies.' This sort of argument in favour of propositions is devel
oped in Chapter 6 and in Schiffer, 1 994. For the sake of simplicity of exposition I am
ignoring the various other ways of referring to propositions (e.g. 'John wonders
whether May loves him') and am writing as though propositions were invariably des
ignated by that-clauses.
Postscript 131
The second important consequence of minimalism is that the
grounds for that-clause attributions need not even involve anything
that might constitute the truth-theoretic properties. In other words, it
is not merely that the evidence for that-clause attributions does not
explicitly advert to truth and reference; it need not even advert to
anything non-semantic that might underly truth and reference. For
example, it need not be that 'dog' expresses the propositional con
stituent that it does in virtue of standing in some non-semantic rela
tion R to dogs-where R would constitute the relation 'x is true of
'
y . For if truth is fully explained by the schema 'The proposition that
'
p is true iff p , then the capacity of propositions and utterances to be
true is guaranteed regardless of how it comes about that an utter
ance expresses the particular proposition it does.
Both of these consequences of the minimalist perspective are vin
dicated by the use theory of meaning. To see this, let us, for simplic
ity, restrict our attention to context-insensitive sentences (such as
'dogs bark' and 'snow is white')-sentences not involving indexical
or demonstrative terms. In such cases it is plausible to identify the
proposition expressed by an utterance with the meaning of the
sentence-type to which the utterance belongs;9 and it is plausible to
suppose that the meaning of a sentence-type is determined by the
meanings of its component words and by the procedure by which
those words were combined. Thus the pair of minimalist implica
tions about propositions amounts to the claim that word meanings
are constituted neither in terms of truth-theoretic properties nor by
anything to which such properties might reduce. Now according to
the use theory of meaning, each word-type means what it does in
virtue of a certain non-semantic regularity regarding its tokens:
more specifically, that the occurrences of its tokens derive from the
acceptance of certain specified sentences (or inference rules) con
taining the word. 10 On this view there is no reason to expect that the
property in virtue of which a predicate 'f' means what it does will
consist in that predicate's standing in some non-semantic relation to
fs. Consequently, there is no reason to think that 'f's' meaning is con
stituted in terms of a non-semantic reduction of the 'true-of'
9 For simplicity, I am not only ignoring context-sensitivity but also the distinction
between de dicto, de re, and de se propositional attitudes. For a discussion of these
complexities, see my Meaning ( 1 998), ch. 3, Objection 1 2 .
1 0 For
elaboration and defence of this position see Meaning, ch. 3 .
1 32 Postscript
relation. I I Thus, if the use theory of meaning is correct, as I believe
it is, then the two predictions of minimalism are verified.
We are now in a position briefly to address Donald Davidson's
dissatisfaction with the minimalist point of view (Davidson, 1 996) .
His critique i s founded o n two objections t o the thesis that truth is
implicitly defined by the equivalence schema. The first of them is
that, contrary to minimalism, truth is conceptually prior to meaning
and proposition. This claim is an immediate consequence of
Davidson's own theory of meaning, according to which the meaning
of a declarative sentence is constituted by its having a certain truth
condition. However, Davidson's theory, although widely accepted, is
far from unproblematic. Amongst the reasons for scepticism about it
are:
( I ) After thirty years there still remains the notorious unsolved
problem of how to articulate a conception of truth condi
tion ( 'u is true iff p') that is strong enough to constitute the
corresponding attribution of meaning ( 'u means that p').
(2) Davidson's truth-conditional theory of meaning is largely
motivated by the desire to explain how the meanings of sen
tences depend upon the meanings of their constituent words.
The proposed explanations piggy-back on Tarski-style
derivations of the truth conditions of sentences from the ref
erence conditions of words. But (as we saw in the answer to
Question 38) there are long-standing questions as to whether
all the constructions of a natural language can be forced into
the mould of a Tarski-style truth theory.
(3) Moreover, Davidson's truth-theoretic approach to the expla
nation of the compositionality of meaning is not at all com
pulsory. One can 'deflate' the issue by supposing that
understanding a complex expression is nothing over and
above understanding its component words and appreciating
how they are combined. I 2
1 1 For a detailed discussion of this point see Horwich, 1 997 c, reprinted (in a
revised form) as ch. 4 of Meaning.
1 2 This idea is sketched above in the answer to Question 22. For a full discussion
of it, see Horwich, 1 997a, reprinted as ch. 7 of Meaning ( 1 998).
Postscript 133
(4) We have just seen how it is possible to give a use-theoretic
account of proposition that does not presuppose the
concept of truth.
It i s therefore far from evident that meaning and proposition should
be explicated in terms of truth.
Davidson's second objection to the brand of de:flationism pre
sented here is that expressions like 'The proposition that dogs bark',
construed as singular terms, are unintelligible. However, this rather
counterintuitive claim is entirely theory-driven: it is derived from
Davidson's inability to find any account (of the sort required by his
truth-theoretic paradigm) of how the referents of such expressions
could be determined by the referents of their parts. Therefore, given
the above-mentioned doubts already shrouding that paradigm, this
particular inability to accommodate it would merely constitute one
more reason to give it up. 1 3
4. Which i s the basic truth concept-propositional truth ,

utterance truth , or act truth?
Ordinary language suggests that propositional truth is fundamental

and that the notion of an utterance 'expressing a true proposition'
and the notion of a belief 'being directed at a true proposition' are
understood in terms of it. From this point of view the order of
explanation will go as follows:
First: begin to characterize propositions by means of the
principle
*p* expresses the proposition that p,
where the sentence-type, *p* , is individuated semantically as well
as physically, and where the two tokens of 'p' are understood in
the same way.
Second: explicate propositional truth via the equivalence schema
The proposition that p is true iff p.
1 3 For a more detailed response t o Davidson's critique o f minimalism, see

Horwich, 1 999.
1 34 Postscript
Third: complete the account of propositions (and propositional
truth) by adding
u and v express the same proposition iff u and v have the same
use-theoretic construction property.
Fourth: introduce utterance truth (' truth') and act truth ('truth*')
by means of the linking principles
Utterance u expresses x -7 (u is true iff x is true);
Act a is directed at x -7 (a is true* iff x is true) .
Fifth: prove the adequacy of these notions by deducing the corre

sponding equivalence schemata
u expresses the proposition that p -7 (u is true iff p)
(which in turn implies the disquotation schema: *p* is true iff p)
and
Act a, of believing (conjecturing, etc.) that p, is true* iff p.
However, despite its conformity with ordinary language, this

order of explanation is not at all compulsory: we can equally well
base everything on one of the other truth concepts. For example, we
might start by characterizing utterance truth (for our own utter
ances) by means of the disquotation schema
*p* is true iff p;
then add a projection principle to cover truth for other utterances:
u and *p* have the same use-theoretic construction property
-7 (u is true iff *p* is true);
then bring in the above account of propositions:

*p* expresses the proposition that p;
u and v express the same proposition -7
u and v have the same use-theoretic construction property,
together with the above linking principle
u expresses x -7 (u is true iff x is true);
from which we can easily derive the equivalence schema
Postscript 135
The proposition that p is true iff p.
The fact that the ordinary language truth predicate typically
expresses propositional truth makes the first approach more natural.
And this is my rationale for taking truth, rather than truth or truth* ,
t o b e the conceptually basic truth concept and fo r supposing that its
bearers are propositions; and in the same vein I take it that our con
cept of truth is engendered by our disposition to accept instances of
the equivalence schema and that this is what explains our confidence
in the truth schemata and in the disquotation schema. However, we
have seen that this conceptual ordering is not especially profound or
important. Our language certainly could have had a word expressing
a primitive concept of utterance truth or act truth*; and in that case
it would have been a simple matter to construct the notion of propo
sitional truth.
5. What does m i n i malism have to say about the natu re

of truth itself, as opposed to the nature of o u r concept
of truth?
We do not hesitate to distinguish a theory about the concept of

water (or the meaning of the word 'water') from a theory of the phe
nomenon of water itself. The first would be supplied by a semanticist
and the second by a physicist; the first would be designed to account
for our thoughts and statements about water and the second for the
properties of water itself; and the first need not say anything about
H2 0 whereas the second would have to. Similarly, it is prima facie
reasonable to distinguish between an account of our concept of
truth and an account of truth itself The former purports to specify
the conditions in which someone uses the word 'true' with a certain
meaning; the latter purports to specify the fundamental facts about
what that word stands for-about the phenomenon, truth.
Consequently, the former-the meaning of 'true'-is specified by a
generalization about the word 'true', a generalization that will
explain all our uses of it; whereas the latter-the theory of truth
itself-consists in principles about the property of truth on the basis
of which all the facts about truth are to be explained. And parallel
distinctions should be drawn between our concepts of truth expres
sion and act truth, and the phenomena themselves.
1 36 Postscript
What, then, from a minimalist point of view, is the basic theory of
truth itself? Which body of fundamental facts about truth provides
the best explanation of all the further facts about it? A natural con
jecture is that, although there is always in principle a difference
between the theory of the concept of X and the theory of X-ness
itself, perhaps, in the case of truth, these theories more or less coin
cide. Perhaps the axioms of the theory of truth are instances of the
equivalence schema
The proposition that p is true iff p,
comprising what I have been calling the Minimal Theory of truth.
An immediate problem for this proposal is created by the liar
paradoxes: the existence of contradiction-implying instances of the
schema. Notice that this is not an overwhelming difficulty for the
supposition that what we mean by 'true' is captured by the equiva
lence schema. For although certain instances yield contradictions, it
might be argued that anyone who means what we do by 'true' has a
certain inclination to accept even those instances-an inclination
that is overridden by the discovery that they lead to contradiction.
Indeed, one might suppose that it is only because we have such an
inclination that the 'liar' sentences present us with a paradox! But no
such manamvre is available to protect the schema if it is proposed as
a theory of truth itself. All that can be said (as we saw in the answer
to Question 1 0) is that the theory may contain only a restricted set of
the instances of the schema. But further theory will have to be
deployed in order to specify that restriction.
A second deficiency of the proposed Minimal Theory, one might
think, is that it has infinitely many axioms. Leaving aside the
banned, contradiction-inducing instances of the equivalence
schema, there is a separate axiom for each proposition. So would it
not be better to pursue a Tarski-like strategy of explaining the truth
of the infinitely many propositions in terms of the referents of their
finitely many constituents? The answer, it seems to me, is No. The
Tarski-style approach offers false hope ( 1 ) because, as is well known,
there are many kinds of proposition (e.g. statements of probability,
counterfactual conditionals, etc.) whose truth we have no reason to
believe can be explained on the basis of the referents of their parts;
and, more importantly, (2) because such a strategy would miss those
propositions that are constructed from the primitive concepts that
are not expressed in our language. If all propositions are to be cov-
Postscript 1 37
ered, then there would have to be axioms specifying the referents of
all the infinitely many possible primitives. So the Tarskian approach
would turn out to need no fewer axioms than the Minimal Theory.
A third objection to the Minimal Theory (emphasized by Anil
Gupta ( 1 993a) and Scott Soames ( 1 997)) is that it is too weak to
explain any general facts about truth. Consider, for example, the fact
that
Every proposition of the form 'p � p' is true.
We might try to explain this by assembling particular explanatory
inferences that invoke nothing more than axioms of the Minimal
Theory. For example, given that if snow is white, then snow is white,
and given the axiom
The proposition that if snow is white then snow is white is true
iff if snow is white then snow is white,
we may explain why it is that
The proposition that if snow is white then snow is white is
true.
But in order to arrive at the general fact to be explained we need to
collect all these conclusions together; and there is no logically valid
rule that would enable us to do so. Clearly, a set of premises attribut
ing some property to each object of a certain kind does not entail
that everything of that kind has the property. We would need a fur
ther premise specifying that we have a premise for every object of
that kind-and this would be tantamount to our conclusion.
However, it seems to me that in the present case, where the topic
is propositions, we can find a solution to this problem. For it is plaus
ible to suppose that there is a truth-preserving rule of inference that
will take us from a set of premises attributing to each proposition
some property, F, to the conclusion that all propositions have F. No
doubt this rule is not logically valid, for its reliability hinges not
merely on the meanings of the logical constants, but also on the
nature of propositions. But it is a principle we do find plausible. We
commit ourselves to it, implicitly, in moving from the disposition to
accept any proposition of the form 'x is F' (where x is a proposition)
to the conclusion 'All propositions are F'. So we can suppose that
this rule is what sustains the explanations of the generalizations
about truth with which we are concerned. Thus we can, after all,
1 38 Postscript
defend the thesis that the basic theory of truth consists in some sub-
·
set of the instances of the equivalence schema.

But there remains one further question. Even if all the familiar
characteristics of truth do indeed stem from the equivalence facts,
might not these facts in turn be explained by some more fundamen
tal theory-perhaps a theory that attributes to truth some specific
underlying nature? The answer, it seems clear, is No. For the equiva
lence axioms are conceptually basic and a priori. In these respects
they are on a par with instances of fundamental logical laws, such as
the law of excluded middle, and with the basic principles of arith
metic. And in no such case should we anticipate theoretical reduc
tion, as we should in the empirical domain. For in the area of a
posteriori facts, the behaviour of systems can be expected to be
caused by the properties of their parts and the way those parts are
combined with one another; therefore one might well anticipate
some reductive explanation of this behaviour. In the a priori
domain, however, there is no such reason to anticipate a reduction.
And in the particular case of the equivalence axioms thei;e are par
ticular grounds for scepticism. For an explanation of them would
have to provide a simple body of facts about truth and other matters,
from which they could all be deduced. But that body of facts would
itself need to be infinite-since the behaviour of infinitely many dif
ferent propositional constituents must be accounted for. Con
sequently, the unification and gain of simplicity required of a decent
explanation could be achieved only if the new theory were to explain
not merely the equivalence axioms but various other phenomena as
well. But we have absolutely no reason at all to think that there exist
any other phenomena whose properties might be explained by the
same theory that explains the equivalence axioms. It would be very
surprising if there were. Thus it is indeed reasonable to regard the
equivalence axioms as explanatorily basic; hence to suppose that
truth has no underlying nature.
Thus minimalism is not merely the view that truth has no concep
tual analysis. It involves the stronger thesis that no sort of reduction
can be expected. And it bases that claim on a certain line of thought:
namely, that the utility of the concept is non-descriptive and is
explained by our acceptance of the truth schemata, that such accep
tance constitutes our grasp of truth, and that the property so char
acterized will not be susceptible to reductive analysis. This is why we
should regard neither Moore ( 1 899) nor Davidson ( 1 990) as
Postscript 1 39
minimalists. For despite their insistence that truth is an indefinable
primitive, they don't acknowledge the non-descriptive function of
the concept, they don't take it to be implicitly defined by the truth
schemata, and they do not oppose the prospect of a substantive
reduction.
6. What does m i n i malism have to say about the normative

significance of truth?
In Michael Dummett's famous paper of 1 959 he argues against the

redundancy theory that it fails to capture the desirability of truth. 14
The redundancy theory tells us when our beliefs are true: it says that
our belief that p will be true if and only if p. But it leaves out the
important fact that we want our beliefs to be true; we aim for the
.
truth. And the same point could be made against minimalism.

In response, however, it can be said that in order for an account of
truth to be adequate it suffices that it be able to explain the desirabil
ity of truth-it is not required that the desirability of truth be an
integral part of the account. Moreover, such an explanation (or, at
least, a sketch of one) can indeed be given.
Roughly speaking, it is easily seen why I should want it to be the
case, for example, that I believe that if I run I will escape, only if I will
escape if I run. I want this because, given a desire to escape, that
belief would lead to a certain action (running), and that action
would satisfy my desire if indeed it implies escape. This is why I
would like it to be that I believe that I will escape if I run, only if I
will indeed escape if I run.
Clearly, one has parallel reasons for wishing that
If I believe that D, then D
whenever 'D' expresses what we might call 'a directly action-guiding
proposition'-namely, a conditional of the form 'If I do A, then I
will get G', whose antecedent describes a possible action under one's
1 4 Dummett's view is elaborated by Crispin Wright (1 992) (to whom I respond in

section 8 of this Postscript) and by Bernard Williams ( 1 996) (to whom I responded in
the answer to Question 1 9).
1 40 Postscript
control, and whose consequent describes a state of affairs that one
might wish to be realized.
Now recall the use of our concept of truth to capture such gener
alizations. We can reformulate the schema as
If I believe that D, then that D is true,
which has the logical form
If I believe X, then X is true,
enabling it to be generalized in the normal way as
All my directly action-guiding beliefs are true.
This is what we have good practical reason to want. But, in addition,
bear in mind that directly action-guiding beliefs are derived from
other beliefs by means of inferential rules that tend to preserve truth.
And any one of our beliefs might be a premise in some such infer
ence. It thus becomes pretty clear why we might like all our beliefs to
be true. 1 5
7.But doesn't the existence of empi rical general izations

about truth-i ncl uding the fact that true beliefs faci l itate
successful action-suggest that truth is, after a l l , a
property with an u nderlyi ng natu re?
The idea behind this objection is that an empirical generalization is

typically explained by the underlying natures of the properties that it
relates to one another. Therefore one might expect that if there are
any general, empirical facts about truth, their existence would
depend, contrary to minimalism, on truth's having some underlying
nature. And, as we have seen, it does appear to be a general fact
about truth that, as Hartry Field puts it, 'agents tend to be more suc
cessful when the truth conditions of their beliefs are realized' . 16
It is a mistake, however, to suppose that generalizations involving
1 5 See the response to Question 1 1 and Barry Loewer (1 993) for refinements of
this idea.
1 6 Both Field ( 1 992) and Michael Devitt ( 1 99 1 ) endorse the suspicion, already
articulated in Question 1 1 , that minimalism may well be undermined by the role of
truth in explaining successful action.
Postscript 141
truth may b e explained by analysing that property. Given the func
tion of our concept of truth, we can see that these generalizations
are not focused on truth, not really about truth. Rather they belong
to that class of special schematic generalizations that rely on the
equivalence schema for their formulation. It was shown in section 5
that such facts must be explained in terms of yet more basic
schemata and in terms of the equivalence axioms; nothing can be
gained by attempting to identify truth itself with some deeper
property.
As we have just seen, it is more or less possible to explain the gen
eralization in question in terms of the truth of the instances of the
schemata
I. If S believes that if p then q, and if S wants that q, then S
wants that p;
II. If S wants that he does A , then S does A;
III. If S believes that q follows from p i , p2 , , and P m then
• • •
(probably) q does follow from p i , p 2 ,. , and Pn·

• .
And, by the same token, each of these schematic facts 'about truth'
will be explained by reference to yet more basic schemata and to
instances of the equivalence schema. No analysis of truth is called
for.
8. Is truth a property?
Minimalism does not involve, in itself, any particular answer to this

question. For it may be combined with a variety of different concep
tions of property, some of which wilJ yield the conclusion that the
truth predicate does stand for a property, and some that it doesn't.
Consider, for example, the liberal conception according to which
every term that functions logically as a predicate stands for a prop
erty. Now it is a vital feature of truth that we can argue as follows:
x is true;
x = the proposition that p;
 The proposition that p is true;
 p.
1 42 Postscript
Therefore the truth predicate must indeed be rendered in logic as
a predicate. Thus there is a perfectly legitimate, weak conception of
property according to which minimalism implies that truth certainly
is one. However, this does not preclude the construction of various
stricter, more 'robust' conceptions of property, by adding further
conditions. One might say, for example, that a predicate expresses a
'substantive' property if and only if there is no a priori obstacle to its
being reducible to non-semantic terms. Minimalism, in so far as it
maintains, on the basis of a priori considerations, that truth is not
naturalistically reducible, will imply that it is not, in that sense, 'sub
stantive' . Thus the only reasonable way to approach the question of
whether, from a minimalistic point of view, truth is a property, is to
distinguish various conceptions of property, and to answer on a case
by case basis.
Paul Boghossian ( 1 990) has argued that minimalism is incoherent
('unstable') on the grounds that it implies both that truth is, and that
it is not, a property. For, according to Boghossian, the essence of
minimalism-that which distinguishes it from traditional accounts
like the correspondence and coherence theories-is its claim that the
truth predicate does not really stand for a property. However, he con
tinues, it is also constitutive of the doctrine that any grammatically
appropriate sentence be substitutable into the disquotation schema
' '
p is true iff p .
And this is tantamount to accepting the generality of the schema

' '
p expresses a fact iff p,
which implies
'Fx' expresses a fact iff Fx,
which, quite plausibly, implies
'F' stands for a property of x iff Fx.
Therefore, since, for example, the proposition that snow is white is
true, it follows that 'true' stands for a property.
There is, however, no reason for a minimalist to fall into this con
tradiction. The second leg of Boghossian's argument invokes the
weak, logical notion of property; and we can accept that truth is a
property of this sort. In that case we will say that what distinguishes
the minimalist conception from traditional accounts is its character-
Postscript 1 43
istic view of the meaning and function of the truth predicate,
together with what that implies about the nonexistence of any reduc
tive theory of truth. Alternatively, we can operate throughout with
some more substantive conception of property. And in that case we
might agree with the characterization of minimalism as implying
that truth is not a property. However, relative to this notion of prop
erty the minimalist should not accept the general schema
'F' stands for a property of x iff Fx,
and can well deny that this follows from the generality of the equiv
alence schema.
Crispin Wright ( 1 992, 1 997) has also claimed that the minimalist
view of truth is 'unstable'-but on quite different grounds from
those of Boghossian. Wright argues that since truth is a 'distinctive
norm'-i.e. since (a) it is desirable that our statements be true, and
(b) truth is distinct from justification-then one must recognize that
truth is a 'real' property.
But we must consider whether these claims really conflict with
minimalism. And to that end we must review what might be meant
by 'a real property' . There are various possibilities.
( 1 ) Were it to mean 'the sort of property that any logically normal

predicate stands for', then, as we have seen, the minimalist has no
disagreement that truth is indeed this sort of property.
(2) Perhaps, then, Wright has in mind what I have called a 'substan
tive' property: namely, the sort of property for which there might
well be a constitution theory of the form
x is true = x is F.
But it surely does not follow, from the normative character of truth,
that truth is 'real', in that sense. On the contrary, we saw in section 7
that no reductive theory of truth is likely to be correct because no
such account will explain the equivalence axioms.
(3) Another possibility is that we count the following infinitely dis

junctive property as 'real':
x = the proposition that snow is white and snow is white;
or x = the proposition that God exists and God exists;
or . . . .
1 44 Postscript
But again this is unlikely to be what Wright is supposing since, apart
from terminology, it is more or less exactly the minimalist account of
truth.
(4) Of course one might introduce a special sense of 'real property'

for which it is stipulated that any 'distinctively normative' property
(as defined above) is to count as a 'real' property. But again the min
imalist could have no quarrel-except concerning the point of such a
stipulation.
(5) The only other possibility I can think of is that Wright's notion
of 'real' property resembles my notion of 'substantive' property
except for not requiring that there be any prospect of reduction.
It may be, in other words, that his conception of 'property with a
real nature' appeals to an intuitive picture whereby it is possible
for a property to have a 'real nature' even though we can see
a priori that no such nature is likely to be specifiable. If this is
indeed the right interpretation of Wright's thesis then there are two
replies. First, in the absence of any explicit articulation of the
criterion for being such a property, and in the absence of any
argument for Wright's contention that a distinctively normative
property would have to satisfy the criterion, we have absolutely
no reason to accept that contention. Second, there is intuitive
evidence for the opposite conclusion. For suppose that a concept of
'truth' (perhaps not identical to our own) is introduced by means
of the stipulation that it will apply to the proposition that snow is
white if and only if snow is white, to the proposition that E = mc2
if and only if E = mc2 , and so on. Then it would seem to be consis
tent with our intuitive conception of 'real nature' and of 'property
constitution' that the 'truth' of the proposition that snow is white
simply consists in snow being white, that the 'truth' of the proposi
tion that E = mc2 simply consists in E being equal to mc2 , etc.
which will imply that 'truth' as such has no real nature. And this
despite the fact that it would nonetheless follow from the equiva
lence schema that 'truth' is a 'distinctive norm' (in Wright's sense).
Consequently, it does not follow from the fact that our actual con
cept of truth is 'distinctively normative' that it stands for a property
with a real nature.
Postscript 145
Conclusion
The minimalist picture of truth has three principal components:
first, an account of the utility of truth (namely, to enable the explicit
formulation of schematic generalizations); second, an account of
the concept of truth (namely, that 'true' is implicitly defined by the
equivalence schema); and third, an account of the nature of truth
(namely, that truth has no underlying nature, and that the explana
torily basic facts about it are instances of the equivalence schema).
These ideas are supported as follows. The thesis about utility is
justified by a couple of considerations. First, there appear to be no
other convenient ways of expressing the generalizations that can be
captured with the concept of truth; the alternative would be to sup
plement our language with the relatively cumbersome apparatus of
sentential quantification. And second, there appear to be no other
advantages of having the concept.
The thesis about the meaning of 'true' is based on two assump
tions, both of which are themselves justifiable. First, the facts in
virtue of which we mean what we do by 'true' are those that best
account for our use of the term. And second, our use of the term is
best explained by our acceptance of the equivalence schema. The
first of these assumptions comes out of a general theory of meaning
sketched in Chapter 6 and elaborated in my Meaning ( 1 998). The
second is justified by the difficulty of finding any uses of the truth
predicate that cannot be explained in terms of the equivalence
schema. Or to put it another way: it is justified by showing that the
generalizing function of the truth predicate is explained by our
acceptance of that schema.
The thesis that truth has no underlying nature derives from the
foregoing. Given the function of truth, we may infer that the general
facts which we need the concept of truth to articulate are not really
about truth; therefore their explanation would not be facilitated by
an account of truth's underlying nature. Rather, the facts about truth
that will enter into the explanation of those generalizations will be
instances of the equivalence schema. These instances are conceptu
ally basic and a priori; hence very likely to be unsusceptible to reduc
tive explanation. It is reasonable to conclude, therefore, that truth is
not constituted by some more fundamental property.
This is not the place to expand further on the philosophical
1 46 Postscript
ramifications of minimalism-but they are evidently considerable.
For the concept of truth is deployed throughout philosophy, often in
ways that are in tension with the minimalist picture. As we have seen,
one example is the formulation of emotivism as the theory that eth
ical pronouncements have no truth value. Another is the view that
the realism/anti-realism debate concerns the nature of truth.
Another is the doctrine that meaning should be explained in terms of
truth. Another is the idea that truth is precluded by vagueness and
other forms of indeterminacy. In each of these cases, progress
towards clearer formulations, and a better sense of where the prob
lems really lie, is achieved by appreciating that truth is metaphysi
cally trivial-nothing more than a device of generalization.
Bibliography
AUSTIN, J. L. ( 1 950). 'Truth' , Proceedings of the Aristo telian Society, supp.

vol. 24: 1 1 1 -28.
-- ( 1 9 6 1 ) . 'Unfair to Facts', Philosophical Papers. Oxford: Oxford
University Press.
AYER, A. J. ( 1 935). 'The Criterion of Truth' , Analysis, 3 .
-- ( 1 936). Language, Truth and Logic. London: Dover.
-- ( 1 963). 'Truth' , The Concept of a Person and Other Essays. London:
Macmillan.
BALDWIN, T. ( 1 989). 'Can There Be a Substantive Theory of Truth?' ,
Recherche sur la Philosophie et le Language, 1 0 . Grenoble: Universite des
Sciences Sociales de Grenoble.
BARWISE, J., and PERRY, J. ( 1 983). Situations and A ttitudes. Cambridge,
Mass. : MIT Press.
BIACK, M. ( 1 948) . 'The Semantic Definition of Truth' , Analysis, 8: 49-63 .
BIACKBURN, S. ( 1 984). Spreading the Word. Oxford : Oxford University Press.
BIACKBURN, T. ( 1 9 8 8 ) . 'The Elusiveness of Reference', in P. French, T.
Uehling, and H. Wettstein, eds. , Midwest Studies in Philosophy, 1 2 .
Minneapolis, Minn . : University o f Minnesota Press.
BIANSHARD, B. ( 1 939). The Nature of Thought, vol . ii. London: Allen &
Unwin.
BLOCK, N. ( 1 986). 'Advertisement for a Semantics for Psychology', in P.
French, T. Uehling, and H. Wettstein, eds. , Midwest Studies in
Philosophy, 1 0 . Minneapolis, Minn . : University of Minnesota Press.
BOGHOSSIAN, P. ( 1 990) . 'The Status of Content', Philosophical Review, 9 9 :
1 57-84.
BRADLEY, F. H. ( 1 9 1 4) . Essays on Truth and Reality. Oxford: Oxford
University Press.
BRANDOM, R. ( 1 9 84) . 'Reference Explained Away', Journal of Philosophy, 8 1 :
469-92.
-- ( 1 988). 'Pragmatism, ·Phenomenalism and Truth Talk' , in P. French, T.
Uehling, and H. Wettstein, eds. , Midwest Studies in Philosophy, 1 2 .
148 Bibliography
BRANDOM, R . ( 1 994) . Making it Explicit. Cambridge, Mass. : Harvard
University Press.
BROAD, C. D. ( 1 9 1 4) . 'Mr. Bradley on Truth and Reality' , Mind, 2 3 : 349-70.
BURGE, T. ( 1 979). 'Individualism and the Mental', in P. French, T. Uehling,
and H. Wettstein, eds. , Midwest Studies in Philosophy, 4. Minneapolis,
Minn . : University of Minnesota Press.
CARNAP, R. ( 1 942) . Introduction to Semantics. Cambridge, Mass. : Harvard
University Press.
-- ( 1 947) . Meaning and Necessity. Chicago, Ill. : University of Chicago
Press.
CARTWRIGHT, R. ( 1 962) . 'Propositions' , repr. ( 1 987) in his Philosophical
Essays. Cambridge, Mass. : MIT Press.
-- ( 1 987) . 'A Neglected Theory of Truth', Philosophical Essays.
Cambridge, Mass. : MIT Press.
COHEN, J. ( 1 9 50) . 'Mr. Strawson's Analysis of Truth' , Analysis, I O : 1 3 6--44.
COUSIN, D. R . ( 1 959/60) . 'Carnap's Theories of Truth' , Mind, 59: 1 -22.
DAVIDSON, D. ( 1 967) . 'Truth and Meaning' , Syn these, 1 7 : 304-2 3 . Repr. in
Davidson ( 1 984) .
-- ( 1 969) . 'True to the Facts ' , Journal of Philosophy, 66: 748-64. Repr. in
Davidson ( 1 984) .
-- ( 1 9 84) . Inquiries into Truth and Interpretation. Oxford : Clarendon
Press.
-- ( 1 990) . 'The Structure and Content of Truth' , Journal of Philosophy,
8 7 : 279-3 2 8 .
-- ( 1 996). 'The Folly of Trying t o Define Truth', Journal of Philosophy, 9 3 :
263-7 8 .
DEVITT, M . ( 1 984) . Realism and Truth . Princeton, NJ: Princeton University
Press.
-- ( 1 9 9 1 ) . ' Deflationist Truth : A Critical Notice of Paul Horwich's Truth',
Mind and Language, 6 : 273-8 3 .
-- and STERELNY, K. ( 1 989). Language and Reality. Cambridge, Mass. :
MIT Press.
DEWEY, J. ( 1 9 1 6) . Essays in Experimental Logic. Chicago, Ill . : Dover.
-- ( 1 938). Logic: The Theory of Inquiry. New York : Irvington.
DRETSKE, F. I . ( 1 9 8 1 ) . Knowledge and the Flow of Information. Cambridge,
Mass. : MIT Press.
DUHEM, P. ( 1 9 54). The A im and Structure of Physical Theory. Princeton, NJ:
Princeton University Press. First published in 1 906.
DUMMETT, M . ( 1 959). 'Truth', Proceedings of the Aristotelian Society, 59:
1 4 1 -62; repr. in Dummett ( 1 978).
-- ( 1 975). 'What is a Theory of Meaning? Part I', in S. Guttenplan, ed . ,
Mind and Language) . Oxford : Oxford University Press.
-- ( 1 976) . 'What is a Theory of Meaning? Part II' , in G. Evans and J.
Bibliography 1 49
McDowell, eds. , Truth and Meaning: Essays in Semantics. Oxford :
Oxford University Press.
-- ( 1 977). Elements of Intuitionism. Oxford : Oxford University Press.
-- ( 1 978). Truth and Other Enigmas. Cambridge, Mass. : Harvard
University Press.
EVANS, G. ( 1 982). The Varieties of Reference. Oxford : Oxford University
Press.
-- ( 1 985). Collected Papers. Oxford : Oxford University Press.
EzoRSKY, G. ( 1 963). 'Truth in Context' , Journal of Philosophy, 60:
1 1 3-3 5 .
FIELD, H . ( 1 972) . 'Tarski's Theory of Truth', Journal of Philosophy, 69:
347-7 5 .
-- ( 1 977) . 'Logic, Meaning and Conceptual Role' , Journal of Philosophy,
74: 379-409.
-- ( 1 986) . 'The Deflationary Conception of Truth', in G. MacDonald and
C. Wright, eds. , Fact, Science and Morality. Oxford : Blackwell.
-- ( 1 992) . 'Critical Notice : Paul Horwich's Truth' , Philosophy of Science,
5 9 : 3 2 1 -3 0 .
-- ( 1 994) . 'Deflationist Views of Meaning and Content', Mind, 9 4 :
249-8 5 .
FINE, A. ( 1 9 84) . 'The Natural Ontological Attitude' , i n J. Leplin, ed.,
Scientific Realism . Berkeley, Calif. : University of California Press.
-- ( 1 984) . 'And Not Anti-Realism Either' , Nous, 1 8 : 5 1 -65 .
FINE, K. ( 1 975). 'Vagueness, Truth and Logic', Synthese, 3 0 : 265-300.
FODOR, J. ( 1 984) . ' Semantics, Wisconsin Style' , Synthese, 5 9 : 2 3 1 -50.
-- ( 1 987). Psychosemantics. Cambridge, Mass. : MIT Press.
FORBES, G. ( 1 986) . 'Truth, Correspondence and Redundancy' , in G.
MacDonald and C. Wright, eds. , Fact, Science and Morality. Oxford:
Blackwell.
FREGE, G. ( 1 89 1 ) . 'On Function and Concept' , in Translations from the
Philosophical Writings of G Frege, ed. and trans. M. Black and P. Geach.
Oxford : Blackwell, 1 960.
-- ( 1 892). 'On Sense and Reference' , in Translationsfrom the Philosophical
Writings of G Frege, ed. and trans. M. Black and P. Geach. Oxford:
Blackwell, 1 960.
-- ( 1 9 1 8) . 'The Thought' , trans. A. Quinton and M. Quinton, Mind, 65
( 1 956): 289-3 1 1 .
FRIEDMAN, M . ( 1 979). 'Truth and Confirmation' , Journal of Philosophy, 76:
3 6 1 -82.
GEACH , P. T. ( 1 960) . 'Ascriptivism' , Philosophical Review, 69 : 221-5.
-- ( 1 965). 'Assertion', Philosophical Review, 14: 449-65 .
GROVER, D., Camp, J. , and Belnap, N . ( 1 975). 'A Prosentential Theory o f
Truth' , Philosophical Studies, 2 1 : 73- 1 2 5 .
1 50 Bibliography
GUPTA, A. ( 1 982). 'Truth and Paradox' , Journal of Philosophical Logic, 1 1 :
1 -60; repr. in R . L. Martin, ed . , Recent Essays on Truth and the Liar
Paradox. Oxford: Oxford University Press, 1 984.
-- ( 1 993a) . 'A Critique of Deflationism', Philosophical Topics, 2 1 : 57-8 1 .
-- ( 1 993b) . 'Minimalism' , in James Tomberlin, ed . , Philosophical
Perspectives, 7: 3 59-69.
-- and BELNAP, N. ( 1 993). The Revision Theory of Truth . Cambridge,
Mass. : MIT Press.
HARMAN , G. ( 1 973) . Thought. Princeton, NJ: Princeton University Press.
-- ( 1 974) . 'Meaning and Semantics' , in M. K. Munitz and P. Unger, eds. ,
Semantics and Philosophy. New York: New York University Press.
-- ( 1 982). 'Conceptual Role Semantics' , Notre Dame Journal of Formal
Logic, 2 3 : 242-56 .
-- ( 1 987). ' (Nonsolipsistic) Conceptual Role Semantics' , i n E. LePore, ed . ,
New Directions in Semantics. London : Academic Press.
HEMPEL, C. ( 1 935). 'On the Logical Positivist's Theory of Truth', Analysis,
2: 49-5 9 .
HINTIKKA, J. ( 1 962) . Knowledge and Belief Ithaca, NY: Cornell University
Press.
HORNSBY, J. ( 1 997) . 'Truth: The Identity Theory', Proceedings of the
Aristo tellan Society, 97: 1 -24.
HORWICH, P. G. ( 1 982a) . 'Three Forms of Realism' , Synthese, 5 1 : 1 8 1 -20 1 .
-- ( 1 9 82b) . Probability and Evidence. Cambridge: Cambridge University
Press.
-- ( 1 987). Asymmetries in Time. Cambridge, Mass. : MIT Press.
-- ( 1 990) . Truth, 1 st edn . Oxford : Blackwell.
-- ( 1 9 9 1 ) . 'On the Nature and Norms of Theoretical Commitment' ,
Philosophy of Science, 5 8 : 1 - 1 4 .
-- ( 1 993). 'Gibbard's Theory of Norms' , Philosophy and Public Affairs,
22: 67-7 8 .
-- ( 1 994) . ' The Essence of Expressivism', Analysis, 4 5 : 1 9-20 .
-- ( 1 995). ' Meaning, Use and Truth', Mind, 1 04: 3 5 5-68 .
-- ( 1 996) . 'Realism and Truth', i n J. Tomberlin, ed . , Philosophical Per-
spectives, 1 0 .
-- ( 1 997a) . 'The Composition o f Meanings' , Philosophical Review, 1 06 :
503-3 1 ; repr. i n Horwich ( 1 998).
-- ( 1 997b) . 'The Nature of Vagueness', Philosophy and Phenomenological
Research, 5 7 : 929-3 6 .
-- ( 1 997c) . 'Deflationary Truth and the Problem of Aboutness', i n E.
Villanueva, ed . , Philosophical Issues, 8 : 95- 1 06 . Atascadero, Calif. :
Ridgeview Publishing Company.
-- ( 1 997d) . 'Realism Minus Truth', Philosophy and Phenomenological
Research, 56.
Bibliography 151
-- ( 1 997e) . 'Implicit Definition, Analytic Truth, and Apriori Knowledge' ,
Nous, 3 1 : 423--40 .
-- ( 1 998). Meaning. Oxford : Oxford University Press.
-- ( 1 999). 'Davidson on Deflationism', in U. Zeglen, ed . , Discussions with
Donald Davidson: On Truth, Meaning and Knowledge. London:
Routledge.
JACKSON, F. , OPPY, G. , and SMITH, M. ( 1 994). 'Minimalism and Truth
Aptness', Mind 1 0 3 : 287-302.
JACOB, P. ( 1 997) . What Minds Can Do . Cambridge: Cambridge University
Press.
JAMES, W ( 1 909) . The Meaning of Truth . Cambridge, Mass. : Harvard
University Press.
KAPLAN, D. ( 1 969) . 'Quantifying In' , in D. Davidson and J. Hintikka, eds. ,
Words and Objections. Dordrecht: Reidel.
KATZ, J. J. ( 1 972) . Semantic Theory. New York : Harper & Row.
KAUFMAN, F. ( 1 948) . 'Rudolf Carnap's Analysis of Truth' , Philosophy and
Phenomenological Research, 9: 294-9 .
KINCADE, J. ( 1 958). 'On the Performatory Theory of Truth', Mind, 67:
394-9 .
KIRKHAM, R . L . (1 992) . Theories of Truth . Cambridge, Mass. : MIT Press.
KRIPKE, S. ( 1 972) . 'Naming and Necessity' , in D. Davidson and G. Harman,
eds. , Semantics of Natural Language. Dordrecht: Reidel.
-- ( 1 975). 'Outline of a Theory of Truth', Journal of Philosophy, 72:
690-7 1 6 .
KUHN, T. S . ( 1 962) . The Structure of Scientific Revolutions. Chicago, Ill . :
University o f Chicago Press.
LEEDS, S. ( 1 978). 'Theories of Reference and Truth' , Erkenntnis, 1 3 : 1 1 1 -29 .
LEMMON, E. J. ( 1 966) . 'Sentences, Statements, and Propositions', in B. A. 0.
Williams and A. C. Montefiore, eds. , British Analytic Philosophy .
London: Routledge & Kegan Paul.
LEWIS, D. ( 1 972) . 'General Semantics', in D. Davidson and G. Harman, eds. ,
Semantics of Natural Language. Dordrecht: Reidel.
LEVISON, A. B. ( 1 924) . 'Logic, Language and Consistency in Tarski's Theory
of Truth' , Philosophy and Phenomenological Research, 2 5 : 3 84-92.
LOAR, B. ( 1 9 8 1 ) . Mind and Meaning. Cambridge: Cambridge University
Press.
-- ( 1 987) . 'Truth Beyond All Verification' , in B. Taylor, ed . , Michael
Dummett. Dordrecht: Nijhoff.
LOEWER, B. ( 1 993). 'The Value of Truth', in E. Villanueva, ed . , Philosophical
Issues, 4. Atascadero, Calif. : Ridgeview Publishing Company.
MAYO, B. ( 1 959). 'Truth as Appraisal', Mind, 6 8 : 80-6 .
McGEE, V ( 1 9 9 1 ) . Truth, Vagueness, and Paradox. Indianapolis, Ind . :
Hackett.
1 52 Bibliography
MILLIKAN, R . ( 1 984) . Language, Thought and Other Biological Categories.
Cambridge, Mass. : MIT Press.
MOORE, G. E. ( 1 899). 'The Nature of Judgment', Mind, 8: 1 76-9 3 .
-- ( 1 9 1 0/ 1 1 ) . Lectures, i n Some Main Problems of Philosophy. London:
Allen & Unwin, 1 9 5 3 .
-- ( 1 927) . 'Facts and Propositions', Proceedings of the Aristo telian
Society, supp. vol . 7: 1 7 1 -206.
PAP, A . ( 1 9 52) . 'A Note on the " Semantic" and the "Absolute" Concept of
Truth', Philosophical Studies, 3: 1-8.
-- ( 1 9 54) . 'Propositions, Sentences, and the Semantic Definition of
Truth', Theoria, 20: 23-3 5 .
PAPINEAU, D. ( 1 987). Reality and Represen tation. Oxford : Blackwell.
-- ( 1 99 1 ) . 'Truth and Teleology', in D. Knowles, ed., Explanation and its
Limits. Cambridge: Cambridge University Press.
PEACOCKE, C. ( 1 986) . Thoughts. Oxford: Blackwell.
-- ( 1 992) . A Study of Concep ts. Cambridge, Mass. : MIT Press.
PEIRCE, C. S. ( 1 905). 'What Pragmatism Is', The Monist, 1 5 : 1 6 1 -8 1 .
-- ( 1 955). 'How to Make Our Ideas Clear', in Philosophical Writings of
Peirce, ed. J. Buchler. New York : Dover.
-- ( 1 9 32-3) . Collected Papers, vols. ii-iv. Cambridge, Mass. : Harvard
University Press.
PITCHER, G. , ed. ( 1 964) . Truth . Englewood Cliffs, NJ: Prentice-Hall.
POPPER, K. ( 1 955). 'A Note on Tarski's Definition of Truth' , Mind, 64: 388-9 1 .
-- ( 1 9 62) . Conjectures and Refutations. New York : Basic Books.
PUTNAM, H. ( 1 975). 'The Meaning of " Meaning" ' , in K. Gunderson, ed. ,
Language, Mind and Knowledge : Minnesota Studies in the Philosophy of
Science, 7 . Minneapolis, Minn . : University of Minnesota Press.
-- ( 1 978). Meaning and the Moral Sciences. London: Routledge & Kegan
Paul.
-- ( 1 9 8 1 ) . Reason, Truth and History. Cambridge: Cambridge University
Press.
-- ( 1 9 8 3 ) . 'Vagueness and Alternative Logic', Philosophical Papers, vol . iii:
Realism and Reason. Cambridge: Cambridge University Press.
-- ( 1 9 8 8 ) . Representation and Reality. Cambridge, Mass. : MIT Press.
QUINE, W. V ( 1 9 5 3 ) . From a Logical Point of View . Cambridge, Mass. :
Harvard University Press.
-- ( 1 960) . Word and Object. Cambridge, Mass. : MIT Press.
-- ( 1 970) . Philosophy of Logic. Englewood Cliffs, NJ: Prentice-Hall.
-- ( 1 977). 'Intensions Revisited', in Midwest Studies in Philosophy, 2.
-- ( 1 990) . Pursuit of Truth . Cambridge, Mass. : Harvard University Press.
RAMSEY, F. ( 1 927) . 'Facts and Propositions', Proceedings of the Aristo telian
Society, supp. vol . 7: 1 5 3-70 .
-- ( 1 9 9 1 ) . 'The Nature of Truth' , O n Truth: Original Manuscript
Bibliography 1 53
Materials (192 7-1929) from the Ramsey Collection at the University of
Pittsburgh, edited by N. Rescher and U. Majer, Episteme, 1 6 : 6- 1 6 .
RICHARD, M. ( 1 997). 'Deflating Truth' , in E. Villanueva, ed. , Philosophical
Issues, 8. Atascadero, Calif. : Ridgeview Publishing Company.
RoRTY, R. ( 1 982) . Consequences of Pragmatism. Minneapolis, Minn . :
Minnesota University Press.
RUSSELL, B. ( 1 903). The Principles of Mathematics. New York: Norton .
-- ( 1 904) . 'Meinong's Theory of Complexes and Assumptions', Mind, 1 3 :
204-52. Repr. in D . Lackey, ed. , Essays in Analysis. New York : Braziller,
1 97 3 .
-- ( 1 905). ' On Denoting', Mind, 1 4 .
-- ( 1 906) . 'On the Nature o f Truth and Falsehood', repr. i n his
Philosophical Essays ( 1 966) . New York: Simon & Schuster.
-- ( 1 9 1 9) . 'On Propositions : What They Are and How They Mean' ,
Proceedings of the Aristo telian Society, supp. vol. 2: 1 -4 3 .
RYLE, G . ( 1 929) . 'Are There Propositions?' , Proceedings of the Aristo telian
Society, 3 0 : 9 1 - 1 26 .
SCHIFFER, S. ( 1 994) . 'A Paradox o f Meaning', Nous, 2 8 : 279-34.
SCHLICK, M . ( 1 935). 'Facts and Propositions', Analysis, 2 : 65-70.
SEARLE, J. R . ( 1 958). 'Proper Names', Mind, 63: 1 66-7 3 .
-- ( 1 962) . 'Meaning o f Speech Acts' , Philosophical Review, 71 : 423-32.
-- ( 1 978). 'Literal Meaning', Erkenntnis, 1 3 : 207-24.
SELLARS, W ( 1 954) . ' S ome Reflections on Language Games', Philosophy of
Science, 2 1 : 204-2 8 .
-- ( 1 962) . 'Truth and "Correspondence" ' , Journal of Philosophy, 5 9 :
29-56 .
SMITH, M. ( 1 994a) . 'Why Expressivists about Value Should Love
Minimalism about Truth' , Analysis, 54: 1 - 1 2 .
-- ( 1 994b) . 'Minimalism, Truth-Aptitude, and Belief', Analysis, 5 4 : 2 1 -6 .
SOAMES, S. ( 1 984). 'What is a Theory of Truth', Journal of Philosophy, 8 1 :
4 1 1 -29.
-- ( 1 997) . 'The Truth about Deflationism' , in E. Villanueva, ed . ,
Philosophical Issues, 8. Atascadero, Calif. : Ridgeview Publishing
Company.
STALNAKER, R. ( 1 984). Inquiry. Cambridge, Mass. : MIT Press.
STAMPE, D. W ( 1 977) . 'Toward a Causal Theory of Linguistic
Representation', Midwest Studies in Philosophy, 2: 42-6 3 .
STERELNY, K. and DEVITT, M . ( 1 989). Language and Reality. Cambridge,
Mass. : MIT Press.
STRAWSON, P. ( 1 950). 'Truth' , Proceedings of the Aristotelian Society, supp.
vol. 24: 1 29-56.
-- ( 1 964) . 'A Problem About Truth-A Reply to Warnock', in Pitcher
( 1 964) .
1 54 Bibliography
TARSKI, A . ( 1 943/4) . 'The Semantic Conception of Truth', Philosophy and
Phenomenological Research, 4: 34 1 -75.
-- ( 1 958). 'The Concept of Truth in Formalized Languages', in his Logic,
Semantics, Metamathematics: Papers from 1 923 to 1 938. Oxford: Oxford
'
University Press.
THOMSON, J. F. ( 1 948) . 'A Note On Truth', Analysis, 8 : 67-72.
UsHENKO, A . ( 1 944) . 'A Note on the Semantic Conception of Truth',
Philosophy and Phenomenological Research, 4 : 1 04-7.
VAN FRAASSEN, B. ( 1 980). The Scientific Image. Oxford : Oxford University
Press.
VAN INWAGEN, P. ( 1 988). 'On Always Being Wrong', in P. French, T. Uehling,
and H. Wettstein, eds. , Midwest Studies in Philosophy, 1 2 . Minneapolis,
Minn. : University of Minnesota Press.
WALSH, W. H. ( 1 952). 'A Note on Truth', Mind, 6 1 : 72-4.
WARNOCK, G. ( 1 964) 'A Problem about Truth' , in Pitcher ( 1 964) .
WmTE, A. R. ( 1 957). 'Truth as Appraisal', Mind, 66: 3 1 8-30.
WILLIAMS, B. ( 1 996). 'Truth in Ethics' , Ratio, 8 : special issue, Truth in Ethics,
ed. Brad Hooker, 227-42.
WILLIAMS, M . J. ( 1 9 8 6) . 'Do We (Epistemologists) Need A Theory of
Truth?' , Philosophical Topics, 1 4 : 223-42.
WILLIAMSON, T. ( 1 994) . Vagueness. London: Routledge.
WITTGENSTEIN, L. ( 1 922) . Tractatus Logico-Philosophicus. London: Rout
ledge & Kegan Paul.
-- ( 1 9 5 3 ) . Philosophical Investigations, trans. G. E. M. Anscombe. Oxford :
Blackwell.
WRIGHT, C. ( 1 987). 'Further Reflections on the Sorites Paradox' ,
Philosophical Topics, 1 5 : 227-90.
-- ( 1 9 8 8 ) . 'Realism, Anti-realism, Irrealism, Quasi-realism' , in P. French,
T. Uehling, and H. Wettstein, eds. , Midwest Studies in Philosophy, 1 2 .
-- ( 1 992) . Truth and Objectivity. Cambridge, Mass. : Harvard University
Press.
-- ( 1 996) . 'Responses to Commentators', Philosophy and Phenomeno
logical Research, 56: 863-8 , 9 1 1 -4 1 .
ZIFF, P. ( 1 962). Semantic Analysis. Ithaca, NY: Cornell University Press.
Index
absolute truth 29-30 Davidson, D. vii, 8, 68-7 1 , 1 04, 1 1 0 n,

action 9, 44-6, 1 39-4 1 1 1 1 , 1 20, 1 32-3, 1 3 8
actuality 29-30 decision, see action
analysis 33-5 deduction, see logic
anti-realism, see realism definition 33-6
approximate truth 63 n determinate truth 78-83; see also inde-
Austin, J. L. 8 terminacy
Ayer, A. J. 1 1 , 5 n, 38, 1 22 Devitt, M. 8, 44, 57 n 1 1 3 , 1 20, 1 40 n
defiationism, see truth conceptions
Baldwin, T. 5 n, 26 n Dewey, J. 9
Barwise, J. 1 00 n disquotation 98- 1 02
behaviourism 54-5 Dretske, F. 1 1 4
belief Duhem, P. 57
analysis of 90-2 Dummett, M . 9, 46, 59, 68, 73, 76, 83,
degrees of 62 1 20, 1 39
truth of 1 6-17, 1 02, 1 3 3-5
Belnap, N. 5 n, 26 n, 42 emotivism 55, 84-5, 1 28
bivalence 78-83 empty names 78, 128
Black, M. 27 n excluded middle 78-8 1
Blackburn, T. 1 1 4 n existence 2, 38
Blanshard, B. 9 Ezorsky, G. 39 n
Block, N. 94 n
Boghossian, P. A. vii, 1 20, 1 42-3 fact 1 04
Bradley, F. H. 9 falsity 7 1-3
Brandom, R. 5 n, 26 n Field, H. vii, 5 n, 8, 29 n, 37, 44, 50, 79,
Burge, T. 1 1 4 94 n, 98, 1 1 3 , 1 20, 1 27, 1 40 n
Fine, A. 5 n
Camp, J. 5 n, 26 n Fine, K. 83
Cartwright, R. 1 0 n Fodor, J. 1 1 4
causal theory of knowledge 90 Forbes, G. 26 n
causal theory of reference 1 1 3-14 formalism 55
Cohen, J. 39 n Fraassen, B. van 57
coherence, see truth conceptions Frege, G. ix, 5 n, 29, 38, 73, 78, 90-2,
conceptual role semantics 94 n 1 06, 1 1 3, 1 22
conjunction 74-5 Friedman, M. 64-7
constructive empiricism 55 future contingents 1 28
constructivism, see truth conceptions
correspondence, see truth concep- Grim, P. 20 n
tions Grover, D. 5 n, 26 n
1 56 Index
Gupta, A. vii, 23 n, 42, 42 n, 43, 1 20, metaphysical realism 60 n
1 24, 1 27 n minimalism 5-8, 1 0-12, 23-5, 3 1-3,
1 20-6, 1 45-6
Harman, G. 70 n, 93, 95 wrt meaning of 'true' 3 5-7, 1 26-9,
Hempel, C. G. 9 1 35
Hintikka, J. 92 n wrt propositional attitudes 1 02,
holism 57 1 3 3-5
Horwich, P. G. 5 n, 35 n, 57 n, 62 n, wrt reference and satisfaction 1 1 5- 1 6
70 n, 7 1 n, 8 1 n, 85 n, 92 n, 94 n, wrt theory o f truth 5, 1 7-20, 2 3 n,
1 1 6 n, 1 3 1 n, 1 32 n, 1 3 3 n, 145 n 25-3 1 , 36-7, 50- 1 , 1 3 5-9
wrt utility of truth predicate 22-3,
indeterminacy 78-83, 94 n, 96 48-9, 1 22-3, 1 40
instrumentalism 54--5 wrt utterance truth 98-1 02, 1 33-5
interpretation 1 00; see also translation Moore, G. E. 9, 1 0 n, 1 06 n, 1 3 8
intuitionism 53-5, 74, 75, 82 n
Inwagen, P. van 60 n names, see empty names and reference
necessity 2 1 n
Jackson, F. 82 n, 85 n negation 7 1-3
Jacob, P. 1 00 n non-cognitivism, see emotivism
James, W 9 non-factualism 55
justification
of logic 74-- 6 observation 46-8
of scientific methods 64--7 ontological commitment 86-7
Oppy, G. 85 n
Kaplan, D. 92 n
knowledge, see scepticism Papineau, D. 9, 50
Kripke, S. 42, 1 1 3-14 paradoxes, see liar paradoxes
Kuhn, T. 63 Peacocke, C. 93
Peirce, C. S. 58
Leeds, S. 5 n, 8 , 50 n Perry, J. 1 00 n
liar paradoxes 40-2, 1 3 6 phenomenalism 54, 55, 57
Loar, B . 5 n Popper, K. 57
Loewer, B. 140 n pragmatism, see truth conceptions
logic progress in science 63
formulation of 3-4, 7 1 -3 propositions 1 6-17, 86-103, 1 29-35
foundations of 74--6 evidence in favour of them 86-90,
of vague predicates 82 n 1 29-30
logical form 87-90 and facts 1 06
logical positivism 57 Fregean and Russellian 90-2
logicism 53, 5 5 identity conditions of 93-5
and utterances 98-1 02, 1 3 3-5
McGee, V. 42 prosentences 4 n, 125
maps 1 08-1 0 Putnam, H. 9, 44, 46, 48, 53 n, 58, 59 n,
materialism 5 3 60 n, 73, 82 n, 1 00 n
meaning
compositionality of 70- 1 , 1 32 quantum logic 73-4
of names 1 1 4 Quine, W V. ix, 5 n, 76, 78, 92 n, 93,
of 'not' 7 1 -3 99-1 00, 1 22
specifications of 33-6
and truth conditions 68-7 1 Ramsey, F. ix, 5 n, 38, 1 22
of the truth predicate 35-7, 1 26-9 realism 7-8, 53-60
use theory of 69-7 1 , 93-4 Recanati, F. 1 00 n
meta-ethics, see emotivism reductionism 53-60
Index 1 57
redundancy theory, see truth concep- irreducibility of 50- 1 , 1 3 8
tions operator 1 6 n
reference 27-8, 1 1 3- 1 6 , 57 n property of 37--40, 125, 1 4 1 --4
relativism 52-3 , 55 and reference 1 0, 23-5, 1 1 2
reliability 46-8, 64--7 traditional problem of 1-2
Richard, M. vii, 1 20, 128-9 utility of concepet of 2-5, 3 1-3,
Rorty, R. 9 1 22-5
Russell, B. 9, 1 0 n, 29, 78, 90-2, 1 06, see also minimalism
1 13 truth conceptions
coherence 8-9, 1 20
satisfaction 28, 5 1 n, 1 1 5-1 6 constructivism/verificationism 9, 58,
scepticism 53-60 60, 1 20
Schiffer, S. 1 30 n correspondence 8, 1 04, 1 1 7, 1 20
scientific realism, see realism defiationism vii, 1-6, 1 0-1 1 , 1 20--6;
Searle, J. 1 00 n, 1 1 3 see also minimalism
Sellars, W. 93, 1 07 pragmatism 9, 46, 1 20
Smith, M. 85 n primitivism 9-1 0
Soames, S. 5 n, 1 20 redundancy/performative theory ix,
sorites paradox 8 1-3 3 1 , 38--40, 1 22
speech acts 39--40 truth conditions 68-7 1 , 1 32
Sperber, D. 1 00 n truth-value gaps 76--7
Stampe, D. W. 1 1 4
Sterelny, K. 44 understanding 68-7 1
Strawson, P. ix, 5 n, 3 8 , 1 22 utterances 1 3 , 1 6-- 1 7 ; 98-1 02
substitutional quantification 4 n, 25-6,
32-3, 39 n, 1 24--5 vagueness 78-83, 1 28
success, see action validity, see logic
verficationism 8 1
Tarski, A. 8, 27, 28, 39 n, 4 1 , 42, 57 n, Vienna Circle 57
1 04, 1 1 0-12, 1 36--7
Thomson, J. F. 39 n Warnock, G. 40
translation 87 n, 95-8; see also interpre Williams, B. vii, 62 n, 1 39 n
tation Williams, M. J. 5 n, 53 n, 67 n
truth Williamson, T. 8 1 n
bearers of 1 6-17, 1 29 Wittgenstein, L. ix, 5 n, 8, 93, 1 04,
beneficial consequences of 22-3, 1 06--7, 1 1 3
44--6 , 1 39--4 1 Wright, C. vii, 446, 79, 1 20, 1 39 n, 143--4
and implication 2 1 -2
intrinsic value of 62-3 Ziff, P. 39 n

Paul Horwich - Truth-Clarendon Press (1999)

Uploaded by

Copyright:

Available Formats

Paul Horwich - Truth-Clarendon Press (1999)

Uploaded by

Document Information

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

Paul Horwich - Truth-Clarendon Press (1999)

Uploaded by

Copyright:

Available Formats

TRUTH

CLARENDON PRESS · OXF O R D

Oxford is a registered trade mark of Oxford University Press

Published in the United States

©Paul Horwich 1990 and 1998

First edition published by Basil Blackwell Ltd. 1990

The moral rights of the author have been asserted

All rights reserved. No part of this publication may be reproduced,

British Library Cataloguing in Publication Data

Library of Congress Cataloging in Publication Data

Typeset by Invisible Ink

1 The Minimal Theory 1

The following is a list of the questions and problems regarding

2 The Proper Formulation 15

3 The Explanatory Role of the Concept of Truth 44

4 Methodology and Scientific Realism 52

22. 'As Davidson has argued, understanding a sen­

6 Propositions and Utterances 86

7 The 'Correspondence' Intuition 1 04

The Minimal Theory

A Sketch of the Minimalist Conception

'What is truth?' , we sometimes ask-but the question tends to be

And it is from precisely this inferential property that propositions

we are provided with an equivalent sentence

The Space of Alternative Theories

It will help us to focus on what is at stake in accepting the minimalist

The second most popular view of truth is known as the coherence

In the third place we have the so-called pragmatic theory of truth

Fourthly-perhaps the least attractive conclusion-there is the one­

These traditional approaches do not typically impugn the correct­

1 . An account may or may not be compositional-it may or may not

2. One may or may not insist on a conceptual analysis of truth, a

4. One may or may not attempt to formulate a non-trivial, finite the­

5. One may or may not propose an account which inextricably links

Summary of Alleged Difficulties

Objections to deflationary approaches have concerned six related

The Proper Formulation. It has been no easy matter to provide even a

The Explanatory Role of the Concept of Truth. The concept of truth

Methodology and Scientific Realism. A natural (realist) view of sci­

Meaning and Logic. A further body of objections concerns the role of

Propositions and Utterances. Propositions can appear to be such

(E) The proposition (p) is true iff p

which presupposes them. At the same time, the natural deflationary

(D) Any utterance of the sentence 'p ' is true iff p,

has difficulty with indexical expressions (try 'I am hungry'), foreign

The 'Correspondence ' Intuition. The idea that a representation is

Each of these topics receives separate treatment in the following

The Proper Formulation

The conception of truth to be defended in this essay is similar

1. Of what kinds are the entities to which truth may be

The list of candidates includes: (a) utterances-individual sounds

2. What are the fundamental principles of the minimal

The axioms of the theory are propositions like

2 This claim will be modified slightly in the answer to Question 1 0 in order to

3 I am employing the convention that surrounding any expression, e, with angled

3. I t seems unlikely that instances of the equivalence

6 See sect. 5 of the Postscript for further discussion of this premise.

According to the minimalist thesis, all of the facts whose expression

4. The m i n i mal theory must be i ncomplete, for it says

A theory of any phenomenon, X, is a collection of principles (i.e.

22. 'As Davidson has argued, understanding a sen

Fourthly-perhaps the least attractive conclusion-there is the one

These traditional approaches do not typically impugn the correct

4. One may or may not attempt to formulate a non-trivial, finite the

Methodology and Scientific Realism. A natural (realist) view of sci

which, as we have just seen, must be regarded as an infinite collec

( 1 * * *) (Snow is white) is true iff the actual world is a mem

Quite right. And it is not part of the minimalist conception to main

10. If the equivalence schema is rel ied on indiscrimin

(9) The electron microscope works so well because the theo

The term 'realism' is an over-used, under-constrained piece of philo