Fact and Meaning

P H IL O S O P H IC A L T H E O R Y

SERIES EDITORS
John McDowell, Philip Pettit and Crispin W right

For T ru th in Semantics
Anthony Appiah
The Dynamics o f Belief: A N orm ative Logic
Peter Forrest
A bstract O bjects
Bob Hate
Fact and M eaning
Jane Heal
Conditionals
Frank Jackson
Reality and Representation
David Papineau
Facts and the Function of T ru th
Huw Price
M oral Dilemmas
Walter Sinnott-Armstrong
 Fact and M eaning
Q uine and W ittgenstein on
Philosophy o f Language

JANE HEAL

Basil Blackwell
 Copyright © J a n e Heal 1989
First published 1989
Basil Blackwell Ltd
108 Cowley Road, Oxford, 0 X 4 1JF, UK.
Basil Blackwell Inc.
432 Park Avenue South, Suite 1503
New- York, NY 10016, USA
All rights reserved. Except for the quotation of short passages for the purposes of
criticism and review, no part of this publication may be reproduced, stored in a retrieval
system, or transmitted, in any form or by any means, electronic, mechanical,
photocopying, recording or otherwise, without the prior permission of the publisher.
Except in the United States of America, 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

A C Il’ catalogue record for this book is available
from the British Library.

Library o f Congress Cataloging in Publication Data

Heal, Jane.
Fact and meaning: Quine and Wittgenstein on. philosophy of
language / Jane Heal.
p. cm. — (Philosophical theory)
Bibliography: p.
Includes index.
ISBN 0-631-14591-5
1. Languages—Philosophy. 2. Quine, W. V. (Willard Van O rm an)-
-Gontributions in philosophy of language. 3. Wittgenstein, Ludwig,
1889-1951—Contributions in philosophy of language. I. Title.
II. Scries.
PI06.H355 1989
!49'.94—del 9 88-7890
CIP

Typeset in 11 on 13 p t Baskerville
by Vera-Reyes, Inc.
Primed in Great Britain at
The Camclot Press Ltd, Southampton
 Contents

Preface vii

1 Introduction I

2 V arieties of Realism 11
2.1 M inim al Realism 11
2.2 Reflecting on M inim al Realism 21
2.3 Realism, Idealism and Empiricism 25
2.4 Realism, Relativism and Intcrsubjectivity 28

3 Instrum entalism and M eaning Scepticism 35

3.1 Q uine’s A rgum ents for the Indeterm inacy of
T ranslation 35
3.2 Setting up a Predictive Sentence M achine 40
3.3 Indeterm inacy o f Function and M eaning 45
3.4 Instrum entalism and the Revisability of Logic 55

4 Q uine’s N aturalized Em piricism 60

4.1 Q uine’s Epistemology 60
4.2 Ontological Relativity and D isquotation 70
4.3 Q uine’s Version of Realism 75
4.4 Should Em piricism be Naturalized? 82

5 T he M ona Lisa M osaic 86

5.1 Sem antic Holism 86
5.2 Holism and Indeterm inacy: the Mosaic Analogy 92
5.3 H olism , Indeterm inacy and Language 98
5.4 C an W e Restore Dctcrminacy? 102
vi Contents
6 T he Slide into the Abyss 112
6.1 T he Incom patibility o f Realism and M eaning
Scepticism 112
6.2 Thoroughgoing Pragm atism 121
6.3 W ittgenstein and Pragm atism 129
6.4 Empiricism and Platonism 134

7 T he Dissolving M irror 143

7.1 W ittgenstein’s Hostility to M irroring Realism 143
7.2 Explanation and the Absolute Conception 149
7.3 Kripkc and N orm s 160

8 Interpretations and M isinterpretations 167

8.1 Speech Acts and Language Gam es 167
8.2 W ittgenstein and Anti-Realism 178

9 Interests, Activities and M eanings 192

9.1 Proof and New Concepts 192
9.2 Conceptual C hange and D eterm inacy o f Sense 196
9.3 O ther Forms of Life? 206
9.4 M odal Realism 210
9.5 Facts about M eaning . . . 216

Notes 229
Bibliography 240
Index 245
 Preface

T he aim of this hook is to form ulate a view of some very wide-

ranging themes about which an enorm ous am ount has been writ­
ten, and it would be absurd to claim any kind of comprehensiveness
or finality for this treatm ent of them. Philosophical discussion
forms a kaleidoscopically changing scene. New, or seemingly new,
pieces are constantly emerging into view and new juxtapositions of
old pieces reveal themselves. 1 hope that T have detected and
described the salient features of some recurrent and significant
patterns - but it may be th at I have only provided a muddled
account of some tem porary configuration. In any case it is certain
th at other interesting arrangem ents are possible, other connections
can be made and other them es justifiably brought to prominence. I
hope that this does not provide an excuse for sloppiness or for
failing to notice m atters which need discussion. But it goes some
way to explain why a great am ount of excellent work goes undis­
cussed here.
I have tried to acknowledge the m ain sources o f and influences
on my views; but I am far from sure th a t 1 have identified and
rem em bered all of them , since philosophical views develop in such
devious ways and over such a length of time. I have several times
had the disconcerting experience of w riting down some thought I
took to be new and then discovering the very sam e idea on looking
through old notes o f my own. I f in this way I can mislay and
seemingly re-invent my own ideas, it is surely possible to do it to
those of others as well and I am certain th at many cases of such
borrowing exist in the following pages. T o all those whose ideas I
have thus appropriated w ithout acknowledgement, I apologize.
vi Contents
6 T h e Slide into the Abyss 112
6.1 T h e Incom patibility o f Realism and M eaning
Scepticism 112
6.2 Thoroughgoing Pragm atism 121
6.3 W ittgenstein and Pragm atism 129
6.4 Em piricism and Platonism 134

7 T he Dissolving M irror 143

7.1 W ittgenstein’s H ostility to M irroring Realism 143
7.2 Explanation and the Absolute Conception 149
7.3 K ripke and Norms 160

8 Interpretations and M isinterpretations 167

8.1 Speech Acts and Language Games 167
8.2 W ittgenstein and Anti-Realism 178

9 Interests, Activities and M eanings 192

9.1 Proof and New Concepts 192
9.2 C onceptual Change and D eterm inacy of Sense 196
9.3 O ther Forms of Life? 206
9.4 M odal Realism 210
9.5 Facts about M eaning 216

Notes 229
Bibliography 240
Index 245
 Preface

T he aim o f this book is to form ulate a view of some very wide-

ranging themes about which an enormous am ount has been w rit­
ten, and it would be absurd to claim any kind of comprehensiveness
or finality for this treatm ent o f them. Philosophical discussion
forms a kaleidoscopically changing scene. New, or seemingly new,
pieces are constantly emerging into view and new juxtapositions of
old pieces reveal themselves. I hope that I have detected and
described the salient features of some recurrent and significant
patterns - but it may be th at I have only provided a muddled
account of some tem porary configuration. In any case it is certain
th at other interesting arrangem ents arc possible, other connections
can be made and other themes justifiably brought to prominence. I
hope th at this does not provide an excuse for sloppiness or for
failing to notice m atters which need discussion. But it goes some
way to explain why a great am ount o f excellent work goes undis-
cusscd here.
I have tried to acknowledge the m ain sources of and influences
on my views; but I am far from sure that I have identified and
rem em bered all of them, since philosophical views develop in such
devious ways and over such a length o f time. I have several times
had the disconcerting experience of writing down some thought 1
took to be new and then discovering the very same idea on looking
through old notes of my own. I f in this way I can mislay and
seemingly re-invent my own ideas, it is surely possible to do it to
those o f others as well and I am certain that many cases of such
borrowing exist in the following pages. T o all those whose ideas I
have thus appropriated w ithout acknowledgement, I apologize.
 T he references in the Notes section are given in an abbreviated
and informal fashion. In connection with W ittgenstein’s works,
numbers refer to sections, except for Philosophical Investigations P art
11, where numbers refer to pages. Full details of the works cited are
to be found in the bibliography.
1 am conscious of a debt to m any friends and colleagues who are
not explicitly mentioned in the book and with whom I have
discussed these m atters. I would like to record in particular my
gratitude to Geoffrey Midgley, conversations with whom over
many years have much enlivened and enriched my understanding
of Wittgenstein. I would also like to thank Crispin W right who
encouraged me to undertake this project in the first place and
whose acute comments on an earlier draft saved me from many
blunders.

St Jo h n ’s College
Cam bridge
 1
Introduction

When we: make remarks which attribute content or significance to

things - lor example when we say what a sentence means or what a
person’s thought is — should the claims we make he treated as
factual? Tlu: aim of this hook is to clarify what this question means
and to examine some possible answers to it.
It proceeds primarily by consideration o f what Q uine and W itt­
genstein say, or have been thought to say, about the m atter. Some
have supposed that their views are im portantly similar. Quine
himself triggers this com parison when lie quotes with approval
W ittgenstein’s dictum ‘U nderstanding a sentence means under­
standing a language’, and comments in a footnote ‘Perhaps the
doctrine o f indeterm inacy of translation will have little air of
paradox for readers familiar with W ittgenstein’s latter-day remarks
on m eaning.’1 O thers also have claimed to lincl significant
resemblances.'*' But are the Q uinean and W iitgenstcinian outlooks
really alike?
It is certainly very plausible to hold that there are. negative views
which are shared by Q uine and W ittgenstein. Both are hostile to
w hat we may label ‘Platonism ’ - that is, to the idea that there are
concepts or meanings existing fully formed and determ inate in
some crystalline realm. They are hostile also to a conception of
philosophy which is the natural com panion of this - namely the-
view that philosophers should spend their time analysing concepts
and laying out the connections existing in the Platonic realm.
T here is also another negative thesis which it seems highly plaus­
ible to hold that they share, although discussion explicitly directed
2 Introduction
to the issue docs not figure prom inently in the writings o f either.
T his is scepticism about that strategy in philosophy of mind and
semantics which we may call reductive materialism , o f which
functionalism in its m any varieties is the current favoured represen­
tative. It is characteristic o f this strategy to take it th at psychologi­
cal and sem antic concepts arc p articu lar cases of the kind of
concepts used in the natural sciences. Given this framework, func­
tionalism proposes to elucidate psychological notions using cen­
trally such ideas as causation, m echanism and law. Both Q uine
and W ittgenstein seem to think th at this sort of approach under­
estimates the differences between the concepts of the natural sci­
ences and those o f psychology and sem antics - differences in the
kinds o f explanations we give by their use and the methodology of
establishing claims couched in their terms.
T his leads us to rem ark upon another and more positive point of
resemblance - the one Q uine draw s attention to in the quotation
he cites. Both Q uine and W ittgenstein lay stress on w hat we m ight
call ‘holistic considerations’, that is on the idea that items are
m eaningful only if they play a role in some complex structure and
th at their meaning is bound up with the whole operation o f the
structure.
A nd, to note one more apparent sim ilarity, it is easy enough to
read some of the cpistemological rem arks in On Certainty (for
example, the image of the river and its shifting banks)3 as express­
ing a fallibilism and pragm atism sim ilar to those famously es­
poused by Q uine in ‘Tw o Dogmas o f Empiricism*.4
Pursuit of these lines o f thought suggests an interpretation of
both authors in which they arc revealed as sceptics about meaning,
on grounds having som ething to do with inextricable links between
holism, pragm atism and the idea of m eaning.
B ut before we rush to elaborate this picture we should set against
the striking similarities (or seeming similarities) some equally
striking differences. We find in Q uine explicit statem ents to the
effect that there is no fact of the m atter where m eaning is con­
cerned. W e have the use by him of disparaging epithets about
m eaning and the intensional and, in contrast, laudatory rem arks
about the cxtensional and scientific; philosophy is presented as
study of the methodology of science and m oreover as continuous
with science itself.5
 Introduction 3
. All this is totally absent from W ittgenstein. He may attack
Platonism but he does not make any overtly sceptical remarks
ab o u t meaning. And, for him , science does not play any central
standard-setting role for intellectual enterprises. W ittgenstein’s
concerns with meaning, if we take th at notion narrowly, link up
w ith concerns about m eaning as we might understand it more
broadly - as significance, im portance or value. A recent study
proposes that we read the Philosophical Investigations as prim arily a
w ork on ethics.6 This suggestion may not be totally persuasive but
it is by no means absurd - as it surely would be if offered about
Word and Object.
I t may be th at these arc diiTcrcnces merely of tone and context.
Perhaps the views of both Q uine and W ittgenstein about meaning
arc sceptical for sim ilar reasons and they dill'cr only in their
response to this insight. But on the other hand it may be that the
diiTcrcnces we have noted connect with differences about the nature
o f the ‘holism’, which each can be said to endorse, and indeed
about the conception each has of w hat it is to be ‘factual’. T he view
I shall endeavour to defend is this second one. I shall argue that
although there are some genuine resemblances, these occur in the
contexts of radically different fundam ental pictures of w hat it
would be like to understand meaning. And the most im portant
outcom e of these different approaches is that W ittgenstein is not a
sceptic or non-realist about m eaning whereas Q uine is.
W h at has been said so far may suggest that the book is primarily
cxcgctical. T his would however be a misleading impression. W hat
I actually have to offer is som ewhat uneasily balanced between
exegesis and freestanding discussion of the issues. I do not say
enough to defend fully the claims I advance about how to read
Q uine or W ittgenstein. T h a t would require more textual citation
an d canvassing of rival readings than is included here. But perhaps
I do not say enough cither about the opposed theories I discuss to
m ake their viability or otherwise apparent. I can only hope that I
have said enough to suggest th at both interpretations and evalua­
tions have some interest and plausibility.
T h e shape of the discussion and its main themes are as follows.
C h ap ter 2 considers various senses we may give to the word
‘realism ’. We start with w hat I label ‘minimal realism ’, which is
presented as an interlocked structure of comparatively uncontentious
4 Introduction
claims, about non-contradiction and epistemological indepen­
dence, which are involved in being a realist. I then introduce the
idea o f trying to understand or explain this minimal realism.
Eventually I shall recommend a quietist response to the challenge,
conforming to the W ittgensteinian injunction to be content with
description and with assembling rem inders. But to see why this is
attractive we need first to allow free rein to the impulse to seek for
more positive seeming accounts. O ne such attem pt at justification
may proceed pragmatically, by rem arking on the convenience,
serviceability etc., of minim al realism. A nother will call upon the
doctrine of mirroring realism. T his is a metaphysically am bitious
view which may be sum m arized as follows: there is a way th at the
world is, independent o f any observer; and in some, but only some,
of our judgem ents we succeed in describing the world as it thus is,
in itself. C hapter 2 suggests further that some variety o f empiricism
will be a natural outcom e of accepting m irroring realism, given
commitment to a naturalistic account of concept acquisition. If,
moreover, we dem and foundational validation of at least some
concepts and exercises of them , we shall end up in sense-datum
empiricism.
C hapters 3 and 4 take such an empiricism as a starting point for
understanding Q uine. C hapter 3 argues th a t instrum entalism will
be the natural construal of those sentences which do not record or
predict our sense d ata and th at the idea that there are no facts
about the meanings o f instrum entally construed utterances can be
made plausible on very Q uinean-sounding grounds. C h ap ter 4
then attem pts to m eet the objection th a t this instrum entalist talk
grossly misrepresents Q uine, who professes to be a physicalist and
scientific realist. I suggest that we can make a great deal o f sense of
Q uine by supposing him to be, in truth, fundam entally an in­
strum entalist (albeit a ‘naturalized’ one) and that both his realism
and his physicalism are, when considered in the light of his disquo-
tational theory of truth and his doctrine of ontological relativity, a
good deal less robust than some have supposed.
M irroring realism in its sim plest empiricist form is, on this
interpretation, at the root o f the Q uinean view. This is not to say
th at Q uine ends up with any straightforw ard realism, let alone a
sense-datum view. H e does not. And when, in section 4.3, I unpack
 Introduction 5
his account of the activities we might label ‘describing the world as
it is in itself, it is a strange one, which no more commonsensical
m irroring realist would be happy to accept. But empiricist m irror­
ing realism nevertheless has the crucial role o f providing Q uine
with his conception of w hat meaning is like, if there is any.
T h e upshot o f this early phase of the argum ent might well be that
is not particularly surprising that we get sceptical conclusions out
when we feed in em piricist premises about limitations on the
concepts atid knowledge that people can acquire. Moreover it
seems far from m andatory to accept the premises in the Q uincan
form. So realism about meaning, if assailed only on the basis
allowed in chapters 3 and 4, will not be seen as under serious
threat. But stronger attacks are in prospect.
T h e theme of ‘holism’ surfaces briefly, in chapter 3, bu t only in
the form o f rem arking on the impossibility o f understanding in
isolation the function o f the elements of the instrum entalist’s theor­
ies. In chapter 5 another sort of holism comes to prominence, in
considering which 1 abandon (or seem to abandon) the whole
em piricist and m irroring realist line of thought and move to con­
sider a less controversially based, and consequently more powerful,
attack on realism about meaning.
T his attack starts from the claim of sem antic holism, namely the
view that items with representational roles, w hether they are
sentences in a language or elements in a picture, have the signifi­
cance they do in virtue of their placem ent in a suitable assemblage
of sim ilar representational items. Sem antic holism is a different
thesis from that o f epistemological holism and needs to be clearly
distinguished from it (as it is not by Q uine). They are not, however,
incom patible and chapter 5 argues that sem antic holism, in tandem
with a com paratively mild and uncontroversial form of epistemo­
logical holism, delivers the Q uinean-sounding conclusion of inde­
term inacy of meaning. Hence (following my account o f what is
involved in minimal realism) I have an argum ent against a realistic
construal of our talk about meaning.
T he picture presented in this discussion might be thought to be
latent in much of w hat Quine says bu t prevented from full
emergence by his residual strong attachm ent to simple empiricism.
It has, 1 would suggest, a good deal in common with Davidson’s
6 Introduction
outlook — although Davidson’s anti-factualist conclusions about
m eaning arc not as strong as the ones I endeavour to extract from
the premises.
T h e question now is w hether the resulting theory is acceptable;
and in chapter 6 we discover th a t trouble lurks. T he position of
chapter 5 purported to be realist about things other than meanings,
for example, material objects. But in chapter 6 non-realism about
m eaning is argued to be incom patible w ith realism about anything
at all. How, then, arc we to account for the appearances o f realism
a t various points in our linguistic practices? We arc here led to
consider a t last that other strategy, m entioned in chapter 2, for
justifying our minimal realist practices —namely a thoroughgoing
pragm atism . T his seems to be the direction in which our considera­
tions are tending. (M any have thought th at Q uine, when we have
discounted in his views the rem nants o f the em piricist notion o f the
‘given’, suggests such an outlook.) But w hat we find on closer
inspection is that pragm atism docs not, after all, olfcr a stable
resting place since it is itself incoherent.
T h e final section of chapter 6 suggests th at m irroring realism ,
although not explicitly invoked in the argum ents o f ch ap ter 5,
nevertheless has an im portant p art in making them seem plausible,
It does this by presenting as natural and inevitable certain modal
and epistemological assum ptions which do play an im portant role
in the argum ents.
T h e claim then, at this stage of the discussion, is th a t a certain
com bination of views will land us in a morass. O ne elem ent is
m irroring realism and another is the idea that sem antic holism
com m its us to the view th a t there is an im portant difference
between the workings o f sem antic concepts and those o f natural
science. T here are then a t least two lines o f enquiry we may pursue
in endeavouring to extract ourselves from the difficulties. Indeed
there are m any more, since the conclusion th at we are in a morass
has been derived through an intricate collection of moves, concern­
ing w hat I claim to be the im plications o f certain leading ideas and
the independent plausibility o f some extra premises which arc
requited. But the lines to be m entioned are those which a good
num ber o f philosophers have in fact chosen to follow.
T h e first is to challenge the second o f the two elem ents men-
 Introduction 7
tioncd above (which I remarked earlier was a view plausibly held
in common by Quine and Wittgenstein) namely accepting the
existence of a radical difference between the concepts o f natural
science and those of psychology and semantics. This challenge may
come to seem very plausible: Suppose otic thinks that the world
revealed by the natural sciences is certainly real and also that
C artesian dualism will not do; the material world, then, is all there
is. W hat will one do, confronted with the argument of chapter (> —
th at realism about the material is incompatible with non-realism
about meaning? Clearly one will make every attem pt to arrive at a
view 011 which statem ents about meaning can be realistically
construed. But given the imagined assumptions about what is real,
this involves trying to assert that statem ents about meanings
describe, or arc true in virtue of, facts discovered in the non-human
sciences. (M any other factors push us in the same direction, for
example, a desire to assimilate all explanation to the same broadly
causal pattern.) O ne’s effort then will be directed to defending
some broadly m aterialist theory of m ind (most plausibly this will
be a form o f functionalism) and finding some flaw in the considera­
tions, about semantic holism and the like, which seemed to lead to
indeterm inacy and non-realism about meaning.
T here arc m any difficulties in this m aterialist and functionalist
program m e in the philosophy of mind. Equally, we m ust note,
there arc many ingenious theories which aim to overcome these
difficulties. But the dcfcnsibility o f these theories is a m atter which
I shall not discuss in detail. Instead I shall pursue the second line of
enquiry to be described below. T here is thus at this point a great
gap in my argum ent, if one were to regard it as attem pting some
kind of dem onstration of the likely correctness of a W ittgenstcinian
outlook. Philosophers sym pathetic to the functionalist view of mind
and m eaning may nevertheless find something o f interest in the
later stages o f the discussion, if only o f an excgetical character.
T h e first line, ju st discussed, leaves largely unexamined the
issues of w hat ‘realism ’ am ounts to (although it probably has an
inexplicit com m itm ent to some version o f the mirroring view) and
tries by ingenious manoeuvres to persuade itself that psychological
and sem antic notions fit into some (as it seems to any exponent)
quite commonsensical notion o f ‘reality’. The second line of
8 Introduction
thought, however, explores the idea that the problems exposed by
chapter (i cannot be handled this way but require a more drastic
move - namely a reappraisal of the m irroring realist idea. This is
the thread which is pursued through chapters 7, 8 and 9. T he
interest of proceeding in this way is that it oilers the best chance of
finding the interesting points of contrast between Q uine and W itt­
genstein. If W ittgenstein’s views are im portantly different from
Q uine’s, in that W ittgenstein is a realist about m eaning where
Quine is not, the difference certainly docs not lie in W ittgenstein’s
greater sympathy to functionalism. My suggestion is rather that we
should conceive W ittgenstein’s outlook as even more absolutely
holistic than Q uine’s and as involving a rejection of the whole
mirroring realist (and relatcdly empiricist) m anner of conceiving of
meaning.
C hapter 7, then, starts by trying to make plausible the idea that
W ittgenstein is hostile to m irroring realism. T he grounds for this
may be sketched as follows. Semantic holism insists that in under­
standing a person’s claims about the world we have to place them
iu an extensive web of other judgem ents. But w hat if this web is not
the whole story? W hat if to understand a person’s thoughts and
concepts we require also to place them in the context of a life, with
its characteristic interests and activities? W hat if the idea of a
purely cognitive response to a prc-sliced world is unintelligible?
T he outcome of these suspicions is the idea that we cannot explain
or justify our concepts in the way m irroring realism desires, by
pointing to the world whose nature forces these concepts on us for
its correct description. W hat we can do instead is to describe how
possession of our concepts interlocks w ith our interests and charac­
teristic activities to constitute our form of life. C h ap ter 7 attem pts
also to link these thoughts with the so-called ‘rule following con­
siderations’.
The later parts of chapter 7 and all o f chapter 8 are concerned
with exploring some implications of this reading of W ittgenstein -
in part by contrasting the interpretation I would like to recommend
with those offered by others, notably Kripke, D um m ett and
Vvright, t consider also a counter-attack on behalf of a sophisti­
cated version of m irroring realism which is m ounted by Bernard
Williams.
 Introduction 9
C hapter 9 explores some elements in W ittgenstein’s views on the
philosophy of mathematics and uses this example to try to clarify
his views on necessity and possibility and on other ’forms of life’.
T he last section of chapter 9 tries to assemble the various clues we
have been olfered to show finally how and in what sense realism
about meaning can be defended.
In the terms of the original three-fold distinction of chapter 2 —
pragm atism , mirroring realism and quietism — the outcome is a
justification of the quietisl approach to these m atters. T he dillicully
that many have had in understanding W ittgenstein is that of seeing
how rejection o f mirroring realism can avoid bringing with it
something in the relativist or pragm atist, and so non-realist, line.
T he constant tem ptation is to construe W ittgenstein’s remarks
against the metaphysical pretensions of the mirroring conception as
some form of scepticism. 1 hope to have suggested why this is a
misconstruction and thus to have vindicated the claim that his
outlook on meaning is significantly different from that of Quine.
W hy, one might ask, is this whole question about the ’reality’ of
meaning or the ‘factuality’ of m eaning statem ents, o f importance?
At the extreme risk of vagueness and portentousness, something
needs to be said about this. In part the answer is that consideration
of w hat the question could m ean, and how one could defend any
particular view, is part of the modern phase of a long-running
attem pt by hum an beings to arrive at some satisfactory conception
of themselves and their relations to their world. T he reflective
self-consciousness which leads to this philosophical project is an
im portant hum an characteristic. But one can add also that the
question is im portant because the notion o f ‘reality’ is itself a
weighted notion. There are strong links between w hat is real, what
is w orth attending to and w hat deserves respect. T he claim that
som ething we had thought real is not real (however we hedge about
that claim or attem pt to soften its impact) will almost inevitably be
seen as an attem pt at dow ngrading or debunking. T h e suspicion
that it might be true that there are no facts about m eaning can
hardly do other than give us a curious-sinking or dwindling feeling.
However little we understand the denial o f facts about meaning,
the threat implicit in it seems to be that of our futility.
O ne thing which is at issue, then, in the question of realism
10 Introduction
about m eaning (as with the closely related topic of ethical realism)
is how much confidence and what kind o f confidence we can have
in our various hum an enterprises and modes of thought. W hatever
the details of the fully articulated picture, it is clear that a W ittgen-
stcinian will stand opposed to the view in which the enterprise of
natural science acquires some kind o f pre-eminence. He or she will
w ant to rem ark that the construction o f scientific theories is only
one hum an project am ong many and that the (undoubted) value
and validity it has should not be seen as underm ining (indeed
probably cannot be m ade intelligible without) the equal, if differ­
ent, value and validity of other projects.
Som ething like this would be common ground among many (if
not all) com m entators on W ittgenstein. But it seems to me im port­
ant also to emphasize that, on the view I would like to recommend,
no easy endorsements or condem nations are to be had. We do not
arrive a t an immovably conservative viewpoint where we see that
there could be no basis for criticism of or alteration to any of our
‘language gam es’. We do not discover, for example, that the idea of
enriching, altering or extending the ‘psychological language gam e’
in the light of investigations in cognitive, psychology, formal sem an­
tics or artificial intelligence m ust be rejected as completely mis­
guided. W ittgenstein remarks that philosophy leaves everything as
it is. And one of the things which is left the sam e is the practice of
applying ideas from one area of thought to another, of seeking
parallels, of perceiving tensions and trying to resolve them. It may
be that, after working through the W ittgcnstcinian lines of thought,
we shall have a subtly altered sense of w hat kinds o f endeavours on
these lines will be fruitful or of how to describe some projects. But
the proof of the pudding is in the eating, and W ittgenstein is not in
the business of t rying to prevent the cooking of new puddings or of
predicting what, all possible future puddings will taste like.
 2
Varieties of Realism

2.1 M IN IM A L REALISM

There arc a variety of ways in which we talk about the meaning or

content of items. We make remarks like ‘.John said that the train
was late’, or ‘M ary has learned that there is a planet beyond Pluto’,
or ‘ “ Es rcgnct” means in G erm an that it is raining’. Here we
ascribe meaning or content to something, an utterance, psychologi­
cal state or sentence type. O n other occasions we speak of people
knowing or not knowing the meaning o f certain words or utter­
ances. Wc say that wc ourselves had not understood something and
then suddenly saw w hat it m eant. Wc report that two things
(words, sentences, remarks) mean the same.
T he question I w ant to address is whether in doing all or any of
these things wc arc making statem ents. In other words (or words
which I shall take to be equivalent) I shall ask whether there arc
facts about meanings, and w hether we should be realists about
meaning.
A num ber of im portant questions can be posed about the rela­
tions between thought and language and thus about the connec­
tions between ascriptions of content to linguistic and to
non-linguistic items. But th a t there arc close links seems evident.
And it would be strange if content ascriptions for linguistic ex­
pressions could be defended as factual while those concerning
thoughts could not, or vice versa. Hence I shall, in what follows,
talk indifferently o f thoughts, remarks, judgem ents, sentences etc.,
except in those contexts where attention to type/token or vehicle/
meaning distinctions is dem anded.
Once we have posed our initial question, further questions
12 Varieties o f Realism
immediately present themselves. W hat is it to be a realist, either in
general or about some specific subject m atter? And w hat is it to be
right in being a realist? T he exploration o f various possible answers
to these questions, their motivations and ramifications, will occupy
much of this book. Let us start obliquely by making some observa­
tions, the acknowledgement of which would, I think, be common
ground to those debating these issues.1 In accordance with the
policy announced in the last paragraph, I shall present the m atter
via discussion of language, although it could equally be presented
in terms of the psychological states which are expressed by utter­
ances (except for some complications which I shall mention later).
If we consider a particular, possible, move in language we find it
very natural to say that in a large num ber of cases there are one or
more other individually possible moves which are incompatible
with it. Hy this I do not mean that it is impossible (e.g. physically)
to do both moves but rather that we can make no sense of someone
who makes the given move and also one o f the set o f incompatibles.
The doing of the latter stultifies or makes nonsense of the former
and the whole linguistic episode is one to which the hearer will be
at a loss to respond. For example if I say ‘Go aw ay’ and then (in the
same breath and to the sam e person) ‘Don’t go away’, the hearer
will be bemused. O r if I say ‘My new coat is bright red all over.
And also it is deep black’ my audience will not know what to make
of it.
It is not clear that all classes o f utterances have incompatibles of
this kind. For example optatives (‘Would th at so and so’ or ‘How I
wish . . .’) may be imagined not to have incompatibles. C ontradic­
tory wishes may not be self-stultifying bu t merely unfortunate for
their subject. But with intentions and beliefs (and their linguistic
relatives, indicatives and imperatives) the notion is very much at
home. This contrast between desires and these other intentional
psychological states is perhaps connected with the fact that our
desires are (on some conceptions at least) merely supplied to us as
raw material, while for our intentions and beliefs we take some kind
of responsibility; we regard their form ation as subject to rational
control and scrutiny which it is our business to apply.
The notion of incompatibility here invoked is not entirely per­
spicuous. It depends upon that of stultifying oneself and is only as
 Varieties o f Realism 13
clear as that is. Let us consider .a few more examples to help get the
idea in focus. If someone says ‘T his is a lemon’ and also ‘This is
sweet’ we may lind it bizarre. Ilut he does not stultify himself. We
may easily suppose that lie has strange beliefs about lemons (while
still understanding ‘lem on’ quite adequately) or we may suppose
that what is in question is an unusual lemon. We certainly do not
have to entertain the suspicion that something has gone radically
wrong with his thinking. W hat if someone says ‘This is red - but it
is sort of orange too’? This may again cause bewilderment for an
instant, lint a way o f removing it is not far to seek - what she is
looking at is a borderline case. Tim s ‘This is red' and ‘This is
orange’ are not incoinpatibles. ’
Merc are two cases where we have, by various moves, explained
prim a facie strange remarks. And one might wonder whether these
sorts of moves, or analogous ones, might not lead us to withdraw
other judgem ents of .self-stultification. There might be suspicion
that the notion is not a particularly robust or fundam ental one but
merely reflects some superficial feature of our imaginations, of the
familiarity of various situations or some other contingency of our
actual experience.
Now it is certainly true that we need not be totally nonplussed by
incidents such as those 1 mentioned earlier, e.g. someone saying
both ‘Go’ and ‘Don’t go’. We have a variety of labels and tech­
niques lor assim ilating and dealing with them. We may, prosaic­
ally, diagnose a slip of the tongue or look lor indexical features of
the utterances which would assign them to different types from
those which we lirst assumed; more excitingly, we' may discern
deep psychological ambivalences or metaphorical significance.
But these moves are unlike the ones wc made in response to the
lemon or red/orange examples. They do not leave the full force of
the original rem ark unaffected. In various ways they lead us to
distance the speaker from the utterance produced. The doubter
about incom patibility is seeking something stronger than this. His
view will be that the notion of incom patibility, at least as 1 have
introduced it, smacks strongly of some suspect conceptual/non-
conceptual or analytic/synthetic, distinction. His suggestion will be
th at there is no com bination of linguistic moves which is to be ruled
out as senseless. T he point is not ju st that we m ight get muddled
14 Varieties o f Realism
and suppose something m ade sense when it did not, or vice versa.
T he suggestion is rather th at the distinction itself involves the
imposition of a two-fold division on w hat is really a continuum ;
moreover, the suggestion continues, this continuum is not one with
respect to something in the same conceptual area as the suspect
distinction (e.g. making more or less sense, where the extremes
would be close relatives of the original concepts) but with respect to
something of a different kind (e.g. familiarity or unfamiliarity).
T his Q uincan style o f challenge introduces something of im port­
ance to which we shall need to return. T he issues it raises will be
discussed further in section 6.2. But for the m om ent we need note
only two things, both of which the doubter should be happy to
concede. T he first is that, however fundam ental we suppose it and
whatever its underpinnings, some such notion as that o f incom pati­
bility is operable in connection with our thoughts and utterances.
T he second is th a t it is bound up with realism in the way to be
sketched in the rem ainder of this section. O u r construal of incom­
patibility will thus feed back into our construal of realism.
Suppose we accept then that some moves arc incom patible,
however this is to be elucidated. Wc may move then to rem ark
another fact: it is not always the case th at of two incom patible
linguistic moves one m ust be preferable to or more defensible than
the other. For example, if my child asks me ‘W hat colour shall I
paint this dinosaur?’ 1 may answ er cither ‘Paint it green’ or ‘Paint
it orange’, and it is surely possible that no Taull should exist in
cither answer. 'Phis docs not mean th at I can with im punity do
both. If I did then then each would have a fault, namely th a t it is
stultified by the other. It means only th at if I do cither one, and
taking into account all circum stances (except w hether or not I have
done the other) there is no fault to be found in that move. It is, of
course possible that there should be some reason, in the paint
available or the existing aesthetic state of the picture, for giving one
answer rather than the other. Choice between incom patible alter­
natives is far from generally arbitrary. But it is no part o f our
understanding of what it is to answer such a question th at there
must be some ground, even if unknown to us, for preferring one to
the other. Wc are not speaking here merely of the two moves being
equally rationally defensible from some limited state o f knowledge,
 Varieties o f Realism 15
but rather of there being no shadow o f criticism to be levelled at
cither from any being or hypothetical being however knowledge­
able.
O n the other hand, in some areas of language use we take it that
of two incom patible moves, if either is in order, one m ust be
preferable to the other; we do not, in the way that we treat these
moves, allow that both could be totally and in all respects without
fault. For example, it cannot be that making the linguistic move of
saying ‘W e’ve ju st missed the bus’ is completely in order in all
respects and also that the incompatible ‘We have not missed the
bus’ is also fully in order. It may be that, given a lew dillicult-to-
intcrpret glimpses of traffic passing and the unreliable state of our
watches, both remarks arc rationally defensible. But they cannot
both be true. O ne of them at least m ust have something serious to
be laid against it. (Perhaps both have something unfortunate about
them in th at there is no bus on that route.)
Let me, for the sake of having a label, call this idea that two
incompatibles cannot both be fully acceptable ‘the principle of
non-contradiction’. There is a risk in this label, since it might be
taken to mean simply the idea that one should not utter incom pat­
ibles. But I hope that, with this explicit warning, we can remember
that it has a stronger content.
T he relevance of this to realism is as follows. It would, I think, be
agreed on all hands that if some sentences arc treated by us in the
first way and not the second, i.e. as not subject to the principle of
non-contradiction, then those sentences can have no claim to a
fact-stating role. W hatever useful function they have it is not that of
describing the real. Im peratives would he a clear example of
something in this category.3 So for some utterances to be rightly
construed as realistically intended statem ents about some subject
m atter it is required th at those utterances be made in a way which
acknowledges the principle o f non-contradiction. (Let me stress
that we arc not at the moment talking about the correctness of
realism but only about what it is to intend some utterance realisti­
cally.)
‘Acknowledging the principle of non-contradiction’ in the rel­
evant sense does not mean merely having a poor opinion of an
utterance which is incom patible with that which one chooses to
16 Varieties o f Realism
make oneself. For example, it is characteristic of moral dispute
(perhaps essential to its being moral) that one has a poor opinion of
someone who differs from oneself. Hence, it) a derived sense, one
will have a poor opinion of those o f his or her utterances which
embody the opposed moral view; one thinks them expressions of
wickedness, bad character or whatever. Iiut rem arking on this fact
is not enough to show that we are all, in virtue of exhibiting this
pattern of reaction, moral realists in the minimal sense. T o be a
moral realist a person m ust in addition take it that one or other of
the opposed judgements is at fault; and it will be at fault by
standards grasp of which is involved in grasp of proper use of the
term. T h at is to say that the moral realist thinks that there is such a
thing as a shared grasp of a standard for correct use of the moral
term and that one or other o f the disputants will be, even if
unwittingly, violating a rule he or she aims to follow. (This is not to
say that the dispute m ust be taken to be resoluble. See the remarks
below on convergence.) T he moral non-realist supposes, by con­
trast, that whatever grasping the rule for employment of a moral
term amounts to, it cannot have this shape.
Let us move on to consider a second observation about realism
which would, I tlunk, be common ground. It concerns the existence
of a connection between the ideas o f ‘realism ’ and ‘mind indepen­
dence’. The absolutely minimal version of this is the claim that if 1
am a realist about, (or example, the existence of marigolds in my
(lower bed, then 1 take it that the existence of those marigolds is not
constituted hv my thinking or sincerely saying that they are there.
When I think about the marigolds my thought is about some state
of atfairs other than itself. And the mere existence and nature of my
thought does not constitute the existence o f w hat it is a thought
about, i.e. does not make the thought correct.
To say this is very far from saying that if I am a realist I must
also accept various sceptical theses about my realistically intended
thoughts, it does not imply that I m ust acknowledge some doubt
about the truth of eveiy belief o f mine, let alone th at I suppose it
possible that they could all be false together, it does not show that
we must be able to make sense of the idea of us agreeing on some
complete and ideally well established theory which is nevertheless
false. Also it does not rule out (at least not w ithout a few more
 Varieties o f Realism 17
moves) the possibility o f C artesian direct and infallible access to
certain items. It insists only that if one is a realist one adm its that
‘W hat do you think?’ and ‘Are things like that?’ are diíTerent
questions.4
I have presented these two, as I would claim, generally agreed
observations about realism, namely its link with non-contradiction
and its link with mind independence, as though they had no
connection. But in fact the second is a consequence of the first.
Suppose I have two remarks that I can make, e.g. ‘T his is a melon’
and ‘This is not a melon’. They are incompatible and I treat them
realistically, that is as subject to the principle of non-contradiction.
Remember that the incom patibility o f which we speak is not a
m atter of the impossibility of my in fact thinking both ‘This is a
melon’ and ‘This is not a m elon’ or of my sincerely uttering both
sentences. It is adm itted that this might by mischance or m uddle
come about. But now let us suppose that it has come about and let
us suppose also that my thinking or sincerely asserting ipso facto
constitutes the thinking or asserting as correct. W hat we have
described is a case where each thought or remark is, barring the
presence of the other, fully correct. We have thus contradicted the
initial assumption th at the incompatibles were o f the type where
both could not be fully defensible.
There are some difficult points that I have skated over in the
remarks above. The physical possibility ofm y uttering both ‘This is
a melon’ and ‘This is not a melon’ is much clearer than is the
possibility of my thinking both that something is a melon and that
it is not a melon. T he vehicle/content distinction does need some
attention at this point. There are problems about attribution of
contradictory thoughts - problems which connect with how we
stultify ourselves and what it is that we prevent happening, when
we make incompatible moves. O ne thing that we prevent is the
smooth development of action and conversation, because others
(and we ourselves) will be at a loss to ‘see w hat we m ean’. And we
may stultify ourselves so severely that people begin to think that we
do not mean or think anything at all. But I take it to be obvious
th a t all of us do have at least some contradictory thoughts -
w hatever this am ounts to - and, hence, that it absurd to try to
fault the claim th at the first uncontentious observation about
18 Varieties o f Realism
realism implies the second by m aking out th at we are always
perfectly consistent.
T he considerations above indicated a connection between mini­
mal cpistcmological independence and the principle o f non­
contradiction. But a version o f the ‘m ind independence’ idea is
already embedded in the very notion that certain moves arc risky,
even w ithout advancing as far as the distinction between ‘realisti­
cally’ construed utterances or thoughts and others. I f we consider
im peratives, for example (or the stance towards the actions and
intentions of others which they manifest), then it is clear th at we do
not suppose the merely sitieercly and wholeheartedly issuing a
com m and or request makes it im m une from criticism. As the
uttcrcr, I may myself, in the light of inform ation gathered later,
come to think of it as a stupid com m and or request. And if, on the
contrary, my uttering it constituted it im m une from all criticism
then it becomes quite opaque how it could be incom patible with
any other. It becomes dillicult also to sec how it could have any
significance beyond itself, and how issuing it or not issuing it could
be a m atter o f importance.
So the idea of an item which docs not carry with it the guarantee
of its own correctness, the idea o f an item which lias incompatibles
and which is such that it m ight have been better to have produced
one o f the incompatibles, all this is not as such the idea of a
realistically m eant statem ent or judgem ent. R ather the externally
underpinned correctness is truth and the items in question arc
realistically meant judgements when the principle of non-contra­
diction is added. W hen the moves in question are seen as (always
and o f their nature, rather than occasionally because of contingent
circum stances) such that full correctness of one rules out that of
another, then the idea o f their being m ade correct by the stale of an
independent reality, and the idea o f their being aimed at a descrip­
tion of th at reality, will be available.
A third observation about realism - which is again a conse­
quence o f the first —is th at taking a realistic stance towards some
subject m atter involves supposing that when people differ about
which one of a set of incom patible moves should be made, further
investigation will produce either agreem ent on the question or
 Varieties o f Realism 19
recognition by both parties th at no firm judgem ent should be
made. A realist cannot, in other words, accept of some realistically
construed area of language th at people can justifiably persist in
confident but incom patible moves w ithin it. This is not to say that
the realist may always hope for convergence of judgem ent. And he
may tolerate continued disagreem ent where neither party is confi­
dent and recognizes the strength o f the other’s case, But he cannot
make sense of a continued, clash of justificdly confident claims
which is irresoluble in principle.
A reductio may make this clearer. Im agine that you and I both
take ourselves to have a realistic understanding of the remarks
‘T here arc marigolds in that (lower bed’ and ‘There are no m ari­
golds in that (lower bed’; imagine also that we accept the same
linguistic system — we operate according to the same standards
and accept the same constraints in assessing the remarks. (This
sameness of language is a trivial condition of the convergence
question arising at all in an interesting form.) Suppose further that
we disagree on w hether there arc m arigolds in that llowcr bed. Clan
we in these circumstances make sense o f the idea that we also take
it as a perfectly acceptable outcome th at this disagreement should
persist through all possible future investigation, even though each
of us fully acknowledges the legitimacy of the grounds that the
other has for his or her continued confident claim? This seems
unintelligible (in the realist framework) because it am ounts either
to the acknowledgement th at standards for assessing these linguis­
tic moves provide contradictory and so incoherent guidance, or to
adm itting that the standards are really o f the non-realist form.
So the realist thinks th a t there is defensible hope of convergence
- in the somewhat limited sense that if a verdict on the m atter is
reached at all it will, with enough open mindedness and in favour­
able conditions o f investigation, be the sam e verdict. But this does
not commit him to thinking that every question which is realisti­
cally construed, and rightly so construed, is in fact resoluble, even
• ‘in principle’.
Let us return now to the starting point of these reflections,
namely the idea that two elements are essential to the minimal form
of realism, the recognition of incompatibles within our linguistic or
20 Varieties o f Realism
conceptual system and, if that is granted, the acceptance th a t some
at least of these incom patibilities are subject to the principle of
non-contradiction.
A possible strategy for underm ining realism - a powerful one
because it invokes such an uncontentious framework - becomes
apparent at this point. It is to try to show th a t (contrary to w hat
one might a t first have assum ed) some set o f linguistic moves is
performed by us in a way which does not and could not acknowl­
edge the principle of non-contradiction. Consider for example the
familiar case of secondary qualities. O ne could argue thus: the
sentences ‘This is hot’ and ‘This is cold’ are incompatible; but the
standards and rules for their use are also such th at it could, in the
familiar case of the warm w ater which feels differently to the two
hands, be fully in order for me to say ‘This is hot’ and equally fully
in order for me to say ‘This is cold’; hence realism here is inappro­
priate.
The realist can o f course fight back. H e m ay deny that ‘This is
hot* is a complete sentence; it needs supplem entation by an extra
relational term like ‘to my right han d ’. O r perhaps some mention of
‘norm al conditions’ should be built in. These moves the realist will
defend as being not merely arbitrary stipulations, designed ad hoc to
preserve the practice o f non-contradiction, but as being required
for the more accurate representation o f the real item which we were
talking about all along.
There is thus an im portant difference between cases where we
acquiesce easily in the failure of non-contradiction when it is
pointed out to us (imperatives, for example) and cases where
discovery of such apparent failure makes us, at least initially,
uneasy (tem perature, for example). T o m ake a denial of realism
plausible in the latter sort o f case one m ust show th at there are no
repairs (introduction of extra- terms or the like) which are non-
arbitrarily motivated and which restore obedience to the principle
of non-contradiction. O ne needs to show also that this failure does
not m atter, that no fundam ental interest or practice of ours is
threatened. In showing this one will allay the uneasiness and
dem onstrate that the realist images or impulses which motivated it
were the upshot of illusion or muddle. T his is, in a nutshell, the
strategy which we may see some meaning-sceptics (e.g. Quine)
 Varieties o f Realism 21
pursuing. We shall see how the argum ents m ight go in chapters 3 ,4
and 5.
It may seem to some th at the invocation o f notions like ‘linguistic
rule’, ‘standards of appraisal’ and the like which occurs in the
above discussion is question-begging. Some sceptical views about
m eaning and realism make play with the idea th at there is no sense
to be (bund in the idea th a t anyone follows any particular linguistic
rule or acknowledges any standard in w hat he or she does. So we
might seem to be setting o(T in the wrong style. B ut this would be
unfair. T he argum ents o f the m eaning sceptics, whom we shall
consider, themselves implicitly invoke considerations about rules,
standards and so forth. They ask us to reflect how we would
establish claims about m eaning and when it is correct to make
them . It may be that the argum ents are intended to have the form
of a reduction But reflection on the m ethod of" proceeding strongly
suggests that, if we are to establish anything about the different
statuses or logical shapes of the concepts we use (physical, psycho­
logical, semantic,- etc.) wc have nowhere else to start except with
some ideas about our thoughts o r remarks and w hat wc suppose to
m ake some more correct or defensible than others.

2.2 R E FL E C T IN G O N M IN IM A L REALISM

We have, then, certain areas of language in which we (at least

seem to) recognize the existence of incom patibles and the impossi­
bility of their being equally and fully correct, where we hope for
ultim ate agreem ent in verdict and where we recognize a potential
gap between our taking some rem ark to be correct and its being so.
These are descriptive claims. Let us for convenience call w hat they
describe ‘our realist practices’. I believe there would be general
agreem ent that going along with such practices about a particular
set of remarks (i.e. not supposing the practices open to serious
criticism, not abandoning them) is a necessary condition for being
a realist about the subject m atter o f those remarks.
But can we say anything further? W hat are we to make of the fact
th a t this is how we carry on, that we have these realist practices?
C an we explain why wc think and speak in ways which fit that
22 Varieties o f Realism
description? And is that explanation such as to justify our so doing?
If so, docs it justify it in every case or m ight it give grounds for
criticism or reappraisal in connection with some sorts o f subject
m atter?
O ne line o f thought, which I shall call pragm atism , supposes
that it is possible to explain and thereby justify our realistic
practices, in fact every aspect o f them (e.g. the facts th at we
recognize incompatibles at all, that we operate the law of non­
contradiction, th a t we do so in certain particular areas, etc.) by
pointing out th at the policy of adopting these practices is advan­
tageous to us; it helps us to organize and anticipate experience, live
satisfactory lives and so forth. O u r way of carrying on is seen as the
outcome of some decision which can be defended in the way
appropriate to decisions. T here is on this view no more to being a
realist than recognizing the usefulness o f realist practices.
It is im portant however to realize that this is not the only form
which an explanatory or justificatory enterprise could take. T he
way we find it natural to speak o f ‘rules’ or ‘practices’ in these
contexts, together with the fact th a t I have talked of language
rath er than thought, may encourage us into an uncxamined as­
sum ption th at any explanation and justification there was had to be
o f the kind suitable lor actions.
But reflection shows that this is in tension with some other
natural ways of taking the m atter. Realism may also, it seems, be
explained and justified in the way appropriate to an opinion. T he
question here would not be w hether it is practically advantageous
to think or talk in a certain way but w hether we think truly in
taking it that the world has such and such features.
T here may seem to be a smack of circularity or question-begging
in this approach. I t is the status of the idea o f ‘tru th ’ which is under
scrutiny and we are told to settle the question by asking whether
certain things are true. We thus continue to use our to-be-
elucidated notion in our enquiry. But a realist of a non-pragm atist
character will not be bounced into supposing th a t every notion
needs underpinning by some reductive or foundationalist m a­
noeuvres. M oreover he may counter-attack by rem arking that the
pragm atist cannot himself do w ithout the non-pragm atist notion o f
tru th he attem pts to discredit.
 Varieties o f Realism 23
For a realist of this stam p, a realist-seeming practice which is
underpinned only by practical considerations will be merely
pseudo-realist. Someone adopting it for these practical reasons has
no right to call him self a realist about the subject m atter of which
he professes to speak. M oreover the sham nature of his so-called
‘realism* is liable to show up in certain extreme circumstances,
even if this is not ap p aren t on the surface of everyday discourse.
T he pragm atist m ay respond either by suggesting that a t a certain
fundam ental level the distinction between opinion and decision
becomes blurred, or by attem pting to discredit the non­
pragm atist’s proposed notion of truth. For the moment we need
note only that there arc two stances to our realist practices, one of
which takes it th a t they need pragm atic elucidation, the other of
which rejects that strategy as inappropriate.
O ne version of the second stance, an extremely powerful and
attractive line of thought, starts with the idea of an independent
world with a determ inate character or nature, a nature which is
fixed quite indepcdcntly of us. We, on this view, are (at least part of
the time) trying to get an accurate, unbiased and uncontam inatcd
view o f w hat this independent world is like in itself. T he general
character of o u r intellectual practices, e.g. the im portant role of the
principle of non-contradiction, is bound up with the fact that it is a
determinate world we have to deal with. The particular features of
our practices - e.g. realist ways of carrying on about material
objects, the past, values or w hatever —are to be explained (at least
in some cases) by saying th a t our intellectual life contains them
because we believe th at possession of those concepts and use of
them in that way is the response dem anded by the actual nature of
the world; these features of our thought are as they arc because we
suppose they need to be that way for us to think about and describe
the world as it is in itself.
I f this is the kind of description we give of our practices, then the
im portant questions to ask in probing our realism will be ‘Is there
indeed a world independent o f us, with a nature of its own?’ And
‘In which o f our judgem ents do we succeed in representing that
nature accurately?’ To answ er affirmatively to the first question is
to be w hat I shall call a ‘m irroring realist’ (borrowing a useful term
from Rorty).6 And for such a realist the second question will seem
24 Varieties o f Realism
pressing. He supposes th at some o f o ur thinking and speaking does
get to grips with the world as it actually is while other elements of it
may fail to do so because influenced and partially determ ined by
our interests, sensory faculties, affective natures and so forth. A
crucial m atter is then how we are to sort out our concepts into those
which do represent the world as it is in itself and those which do
not.
We may say, then, that this m irroring perspective on our realistic
practices adds to the epistemological independence emphasized in
minimal realism the extremely im portant extra element o f conceptual
independence. It is not only our individual judgements which are
answerable to something other than themselves for their truth or
falsity; the very concepts in terms of which they are couched m ust
also (if the judgem ents are to be o f the real) answer to something
‘out there’ and independent of us.
Both pragmatism and m irroring realism are agreed in supposing
that explanation and justification o f our realist practices are called
for, although they differ in the resources they call upon for the
explanation and justification. But a third im portant view rejects
this assumption. I shall call this position ‘quietist’ realism. O n this
view we can make nothing of the question ‘Is so and so really the
case?’, beyond a request to examine again our particular grounds
for affirming so and so, to see w hether we still wish to do so in a
wholehearted fashion. If we find on reflection that we can make
nothing of not continuing to operate our existing realist style of
linguistic practice with respect to ‘so and so’-type utterances, and
th at we see good reason to affirm this particular sentence, then that
is enough to say ‘Yes, so and so is really the case’.
Q uietist realism finds both pragm atist and m irroring realist
attem pts at justification pf linguistic practice misguided. It denies
that we can make any sense of the choice that the pragm atist
supposes us to make and denies also th a t the idea of a, so to speak,
pre-sliced world makes sense. It invites us instead to become aware
of the interlocking complexities of o ur thought and action and to
become aware also of how little sense or use we can make of the
idea of (certain sorts of) things being otherwise.
The discussions of this book have to do with the virtues and
failings of these approaches to w hat ‘realism ’ is and with the
 Varieties o f Realism 25
implications of each lor the factual or non-factual status of our
remarks about meaning.
Given that there are (at least) two subjects o f prim a facie realist
discourse - namely meanings (people, their thoughts, actions,
speech etc.) and the rest (paradigm atically space, time, stulf etc., as
dealt with in the natural sciences) - we have at least four different
possible positions, '[’hey are as follows:

1 We should be realist about both meanings and the rest.

2 We should be realist about the rest but not about meanings.
3 We should be realist about meanings but not about the rest.
4 Wc should be realist about neither.

Each of these positions will, of course, be interpreted differently,

depending on w hat variety of ‘realism ’ we have in mind.
T he two possibilities we shall be mainly concerned with arc (1)
and (2), although (4) will also make an appearance. 15i.it there are
complications in the form of the existence o f some hybrids and it is
to the tracing o f the developm ent of one o f these that we. will turn in
the next chapter. T his is partly because it seems to be one way into
an intelligible reading of Q uine’s diilicult pronouncem ents on these
m atters and partly because it allows the introduction of some
notions, such as ‘holism’, the later development of which will prove
im portant. But before we turn to that we will, in the final two
sections of this chapter, consider first the connection between
realism in the sense in which I have been speaking o f it and other
outlooks to which the label is applied, and, secondly, some of the
ramifying connections between the notion of realism and the ideas
of cultural relativism.

2.3 R EA LISM , ID E A L ISM AND E M P IR IC IS M

As I shall use the term ‘realism ’ it will m ean, first, the commit­
m ents embodied in minim al realism and, second, w hatever further
moves (e.g. of a pragm atist, m irroring or quietist kind) are at issue
in some particular attem pt to understand m inim al realism. But the
term ‘realism ’ is used in a variety of other ways too.
26 Varieties o f Realism
T he question of w hether wc should be realists about some class
o f entities is sometimes taken as the question of w hether or not
statem ents about them can be reduced to sets of statem ents about
some other class of entity. I f the statem ents can be reduced then the
supposed entities arc not real; if they cannot then the entities are
real. This debate is one which, as sketched, proceeds on the agreed
assum ption that both types o f statem ent in question arc factual. If,
o f course, th at itself is in dispute —i.e. if there is a debate over the
‘realistic’ construal of one o f the sorts o f statem ent — then the
dem onstration of a reduction may be a way of preserving rather
than underm ining the ‘realistic’ status of the suspect statem ent and
saving it (rom relegation to the realm of the instrum ental or
emotive.
A second sense in which ‘realism ’ is used, which is not directly
our concern, is as a label for com m itm ent to the idea o f ‘verification
transcendent tru th ’. T here may indeed be connections between
views o f this shape and some o f the views I label ‘realism ’. But
w hether this is so is som ething for further discussion and its-truth is
not built into our meanings for the term ‘realism ’. (T here is a brief
discussion of some o f these m atters in section 8.2.)
In yet a third usage, ‘realism ’ is opposed to ‘idealism ’, in the
sense of the view that nothing exists other than minds and their
states. B ut rejection o f this ‘idealist’ metaphysics is not required (or
a t least not w ithout further argum ent) by realism in the senses that
interest us. It might, for example, be possible to be an idealist, in
the sense of affirming that all that fundam entally exists is states of
consciousness, while being a thoroughgoing realist of the m irroring
kind. This may seem an odd claim. T h e view that w hat exists are
states of consciousness might seem to com mit us to a denial of any
of our sorts of realism because it is definitive of consciousness that
one cannot be mistaken about w hat is going on in it. So we m ight
seem to have a case where thinking that something was so consti­
tuted its being so and thus have a case violating the mind indepen­
dence clem ent of minimal realism. But this is too hasty. T h a t I
have infallible knowledge of my states of consciousness does not
entail that my thinking th a t I am, for example, in pain, makes it the
case th at I am in pain or constitutes my being in pain. We m ight
explain the infallibility in terms of the ‘closeness’ o f the item to me
 Varieties o f Realism 27
and the consequent ‘clarity’ and ‘directness’ of my awareness o f it.
Indeed this is exactly the form th at m irroring realism takes, when
combined with certain further assum ptions. Let us‘ consider how
this comes about.
Suppose th at one is convinced o f the mirroring realist general
conception and also one is convinced th a t certain of our concepts
are genuinely reality-representing. W h at account might one give of
how we acquired the approved concepts? How can we certify them
as ones fit for accurate representation o f the world as it is in itself?
And how do we justify the claim th at any particular exercise of
them is right? O ne m ight appeal to a benevolent and truthful God
in answ er to all three questions. O ne might take it that he had
endowed us with m ental capacities suitable for judging the world
he has put us in, that reflection on his nature shows that he has
done this and that those exercises of concepts which seem natural,
indeed compelling, to us m ust consequently yield truth.7
B ut w hat if a theologically based epistcmology docs not appeal?
A plausible move is surely to say th at we acquired the right
concepts forjudging the world by interaction with the world itself,
that the world reveals its nature to, or enforces awareness of its
nature upon, the enquiring mind.
A philosopher with this naturalistic and empiricist outlook might
approach the other two problems with the assumption that the
questions have to be dealt with on an individual, judgem ent by
judgem ent, basis - or at least th at it is im portant to be able to
answer them in such a way in a t least some cases. This assumption
m ight well be grounded in the im m ensely attractive idea that
epistcmology needs some certainties on which to found enquiry. So
the idea is that we are to be able to get assurance that in thinking as
we do on some particular occasion wc are both exercising a reality
representing concept and also doing so correctly. W hat would the
world have to be like if this sort o f assurance were to be forth­
coming? T he answer is (very plausibly) that it would have to
contain some ‘ideas’ or ‘sense d a ta ’ - i.e. some things which
present themselves to a sentient being in such a way, so clearly and
directly, that he or she cannot but grasp both their nature and their
existence.
It is this view, this radically em piricist and idealist version of
28 Varieties o f Realism
mirroring realism, which we shall consider in the next chapter.
(There are other, non-atom ist and non-sensory versions o f m irror­
ing realism. Discussion of them will be postponed until section 7.2.)
But before turning to th a t we need to rem ark on some potentially
confusing variant uses o f ‘idealism ’.
As I have ju st sketched it, ‘idealism ’ is the thesis that nothing
exists other than states o f consciousness. But the label may also be
used for the view th a t the epistemological independence element of
minimal realism fails, i.e. for the view th at (in general or for some
particular sort o f judgem ent) thinking it so constitutes its being so.
This second sort of idealism is distinct from the first. O ne might
plausibly argue that if we reject minimal realism everywhere then
we must accept that all th at exists are stales of mind. T he idea
would be that everything required for any judgem ent’s being
correct is given in the fact th a t it occurs. (W hether this is a coherent
view is not at all clear. There is extreme difficulty in saying what
the content of any judgem ent is. B ut these difficulties need not
concern us.) But the converse im plication, from acceptance of
metaphysical idealism to denial of m inim al realism, does not
obtain, as I remarked above. H ence the distinctness o f the theses.
There is also a third sense o f ‘idealism ’. T his is th at in which it is
used as a label for any position not of the mirroring realist kind, i.e.
for any position which sees our concepts as somehow dependent
upon us.8 The use o f the term in this way hints at the idea that we
cannot allow our concepts to be dependent upon us w ithout also
allowing acceptance o f the view th a t the tru th or falsity o f judge­
ments exercising those concepts is dependent upon us. I f this were
so then there would be no separating m inim al realism from m irror­
ing realism. This is an extremely im portant issue and we shall
return to it, especially in section 7.2, and section 8.1.

2.4 R EA LISM , R E L A T IV IS M AND

IN T E R S U B JE C T IV IT Y

I have, in the first section o f the chapter, presented those features o f

realism which I wished to stress —for example, acknowledgement
of the centrality of non-contradiction - mainly from the standpoint
 Varieties o f Realism 29
of one imagined judger. I stressed cases (that of im peratives or the
two hands in w arm water) where one person m ight (so it seems)
find him or herself in a position defensibly to make either of two
incompatible linguistic moves. B ut it would have been equally
natural, and perhaps equally persuasive, to have m ade the point by
using the idea that two speakers o f the one language might recog­
nize that they conllict in the linguistic moves they m ade (I say
‘G o’, you say ‘Don’t go’, or I say ‘I t ’s hot’ anil you say ‘It’s cold’)
without either needing to acknowledge that one o r other m ust be in
error. O ne would say that where this could be m ade out to be the
correct description o f the situation, there the speakers would ac­
knowledge that their utterances were not realistically intended.
(W c have already seen considerations like these at work in the
reflections on the notion of convergence.)
T he assum ption which underlies this multi-person way of pre­
senting the issue is that, for the moves in question, the identity of
the speaker does not aflect the question of w hether the moves are
‘incom patible’ in the relevant sense. If i make a certain move (for
example saying ‘G o’) then the fact that it is another person who
produces the ‘D on’t go’ does not ipso facto turn that token into one
o f a type compatible with my original utterance. (Note that w hat it
is for utterances to be incom patible is still to be explained in terms
o f how one speaker would stultify him or herself by producing
both.) But the recognition th at I am not the only person capable of
making a given linguistic move —that other can do the same as me
in the relevant sense of ‘the sam e’ - m ay make it easier for
anti-realist argum ents to get olT the ground, because the circum ­
stances which justify the conflicting utterances can now be distri­
buted around among different persons.
W e have so far imagined one language spoken by one o r more
persons, and the question has been how certain intra-linguistic
features (the structure of the rules which we recognize as governing
the various aspects of correctness of utterances) connect with how
parts of the language should or should not be realistically under­
stood. But the question o f ‘realism ’ is often raised in the context of
comparison between languages and cultures. Is this inter-linguistic
issue quite a different one? T h e suggestion 1 wish to make is th at it
is not.
30 Varieties o f Realism
‘C ultural relativists’ - to put the ease crudely - can be seen as
draw ing our attention to the existence of very widely differing
‘conceptual schemes’ (which arc often said to be ‘incom m ensur­
able’) and using their existence to unsettle our previous realist
confidence. We are invited to acknowledge th a t each culture has its
acceptable collection o f utterances (or practices of utterance-
making) which it is entitled U) go on using. Hut we should (it is
said) recognize that the making o f these utterances is not to be
construed as the description o f how things are independent of
people; rather they arc moves in complex social rituals, constitutive
of various distinctive ways of life.
Hut how are we to be motivated to make this move? T he
relativist supposes that we find ourselves in a situation where
several conditions are simultaneously fulfilled. T he first is that we
have come to understand what it is th at some persons in another
culture say. T he second is that we see th at saying what they say is
incom patible with saying w hat we say. (This is clearly im portant.
T h a t w hat they say is different from w hat we have hitherto said and
so cannot be translated into our fam iliar language is itself no bar to
simply adding their remarks to o ur stock o f truths. T he supposed
incom patibility is what blocks this move.) T he third condition is
that we sec that there is no way of rem oving the incompatibility by
adding in extra relational terms o r the like, and thus no way of
revealing their and our utterances as, at bottom , complementary
parts o f a unified account of w hat there is. T he fourth and final
condition is the most crucial. It is th a t we adm it that there are and
could be no grounds for saying that one set o f remarks is right and
the other wrong. It is at least p a rt of the role of the idea of
incom m ensurability in these discussions to provide support for this
last claim. T he fact th at the other way o f looking at things is
incom m ensurable with ours is supposed to show th at our familiar
standards of assessment get no grip on their remarks and practices.
H ence wc arc (it is said) in no position to condemn them. Indeed,
in understanding the remarks (as wc have been adm itted to do) we
presum ably grasp w hat would make them correct. And wc sec that
their justifying conditions are indeed fulfilled.
W hat arc we to make of the idea o f such a situation? T he
question now is not w hether it is coherent or w hether some parti­
 Varieties o f Realism 31
cular case ought to be seen as satisfying the conditions. T he
questions are w hat the logical structure of such a case would be if it
were dem onstrated to exist, and w hat it would have to do with
realism. My contention is that the structure is the same as that
found in our previous intra-linguistic examples, such as that o f‘hot’
and ‘cold’, and th a t the threat to realism comes from the same
source. By imagining ourselves as understanding but unable to
fault the alien but incom patible rem arks, we have, in effect, ima­
gined ourselves in possession of an extended language, built from
the am algam ation o f ours and theirs, in which the intra-linguistic,
realism -threatening, pattern is repeated. T he idea of ‘incommen­
surability* here plays another o f its roles, namely that of providing
a parallel to the logical relation of incompatible dcterminablcs of a
determ inate in the more domestic, intra-linguistic case. Wc have in
both cases prim a facie characterizations which arc in some logical
relation in as m uch as they exclude one another (they are incom­
patible if both are affirmed); but the logical relation cannot be
explained through the content of one characterization being de­
rived from th at of the other by rearrangem ent of elements and/or
the addition of negation.
But w hat of a potential threat to realism arising from the idea of
m utually inaccessible ‘conceptual schemes’? Does this also have the
same structure? T o say so might seem strange since ex hypothesi it
seems it will be im proper, even by extension, to regard this as a
case o f an intra-linguistic situation. Yet nevertheless it is illuminat­
ing to regard the possibility as a kind of limiting case of the same
structure.
Let us first try to be clear w hat we arc to consider. Imagine that
we can make sense of there being a set of concepts in terms of which
judgem ents can be m ade which are, of their nature, incomprehensi­
ble to any thinker able to employ our concepts and such also that
the inaccessibility o f our concepts is guaranteed to any thinker
capable of using the others. Now the question is: if there were such
another conceptual scheme, would that undermine the realistic
status of our judgem ents?
O ne might urge that it did not. Perhaps wc could imagine that
there is, interpenetrating with our real world, another equally real
but quite independent world and th a t users of the two schemes are
32 Varieties o f Realism
thus responding to the two separate aspects of w hat there is. I do
not wish to say that this is something which is instantly to be ruled
out. O ne can have all kinds of philosophical (or science fiction) fun
with such a hypothesis. But it is not clear that it is in fact com pati­
ble with the imagined conditions. It seems plausible that the idea of
the two accounts’ being com plem entary would need to be ex­
plained in terms o f a G od’s eye viewpoint from which their dove­
tailing is apparent. But the idea that there is such a viewpoint
contradicts the claim that capacity to use the concepts of one
scheme excludes capacity to use the concepts of the other. If God is
to be allowed to violate this restriction it becomes opaque why we
should be supposed to be subject to it.
However this may be, there is a version of the ‘totally different
conceptual schemes’ idea 011 which the two schemes are taken to be
incompatible in the same way as the different but mutually com­
prehensible schemes were in the earlier example. I f two ju d g e­
ments, taken from the two schemes, arc psychologically incom­
patible, in the sense that entertaining one excludes so much as
comprehending the other, they will be a fortiori also logically
incompatible; were someone {per impossibile) to make both ju d g e­
ments, the result would be hopeless confusion, fragm entation of
mind and, so, scll-stultification.
If this is adm itted to he an intelligible way of conceiving the
m atter then wc do have a skeletal version of our original intra-
linguistic argum ent against realism . T o be sure, the alternative
judgem ents appear merely as w hat is imagined to occupy a box
labelled ‘the other way of thinking’. B ut their appearance within
our language, even in this sketchy fashion, is enough to prevent our
judgem ents from satisfying the law of non-contradiction. O u r
judgem ents have been supplied with incom patibles which, ex hypo­
thesis arc as acceptable as themselves.
We see then that some ways o f taking seriously the idea of
different conceptual schemes - ‘taking seriously’, for example, in
the sense of actually affirming their existence or possible existence
- is liable to result in the rejection of minim al realism. It may seem
from this that m irroring realism has acquired a powerful argum ent
in its support. It may seem to be the only metaphysical outlook
which, with its stress on the one right set o f concepts, can save us
 Varieties o f Realism 33
from falling into conceptual relativism and the abandonm ent of any
realism at all. But I hope to suggest later (especially in section 9.3)
that this is not so. In order to avoid affirming that there are two
conceptual schemes wc do not have to insist baldly that there is
only one, nor yet do we have to underpin the. concepts exercised in
the judgements we wish to interpret realistically by appealing to a
classification scheme built into N ature.”
A nother line of thought leading to the claim o f a link between
minimal and m irroring realism starts from reflections on the idea of
intersubjectivity. If I suppose that (some ol) my judgements are of
their nature intelligible only to me, I credit myself to that extent
with a private conceptual scheme. I think o f myself as confronted
with some aspect of w hat is real which is nevertheless in principle
not even thinkable by (let alone cpistemologieally accessible to)
any other person. But as we have just seen, this view is o f doubtful
coherence. In making the imagined subject m atter of my judge­
ments in principle unintelligible to anyone else, while at the same
time rem aining non-solipsistic, I supply my thoughts with incom­
patible rivals (namely those accounts of the real offered by other
people in which my private objects do not figure) with which I can
find no fault. I fence 1 seem to lose the right to be a minimal realist
about my private objects.1”
Perhaps, then, in thinking of the subject m atter of a realistically
intended judgem ent as independent of my thought aliout it I
necessarily think o f it as something intelligible to other thinkers?
(‘Intelligibility’ and its relation to contingent lim itations of intellect
and also its relations to epistemological access need clarification,
but wc need not trouble with this lor our present purposes.) If this
is so then a tem pting extension is to argue that I need to think of the
real .as that which is intelligible to (the idea of which is accessible
to) any thinker whatsoever. And what sense can we make of this idea
except by adopting the mirroring realist notion of the world?
Again I wish to suggest that this is too hasty. T h a t any realisti­
cally intended thought m ust be imagined to be som ething which is
capable of being shared by another person is one thing. T o grasp
this idea we need only the conception o f there being other people, a
concept with which it is hardly extravagant to credit ourselves. T he
other move, however, requires that we make sense of the idea o f ‘all
34- Varieties o f Realism
other possible thinkers’, and it is m uch less clear that this is a
notion we can claim to he in firm control of.
Some o f these issues will surface again in later chapters, espe­
cially section 7.2. But to start, we shall explore w hat happens when
we adopt m irroring realism and also, persuaded by the argum ents
sketched above in section 2.3, we adopt it in the form of bold and
simple sense datum empiricism.
 3
Instrumentalism and Meaning
Scepticism

3.1 Q U IN E ’S A R G U M EN TS FO R T H E IN D ETER M IN A CY
O F T R A N SL A T IO N

W hat is the root of Q uine’s scepticism about meaning? Why docs

he claim that there arc no facts of the m atter about what our
sentences (other than observation sentences) mean? In this and the
following chapter I shall defend the idea that the appeal of m irror­
ing realism plays an im portant role in his outlook. It does so via its
tendency to generate a ‘direct confrontation’ empiricist view and
via 'the tendency of that view in turn to lead to instrumental
construals of judgem ents which cannot be fitted into the direct
confrontation account.
However, in trying to understand Q uine’s highly paradoxical
views on meaning, we should surely start with something else
which Q uine stresses again and again, namely his naturalism . He
rejects the idea of a ‘first philosophy’; he docs not think that we arc
supplied a priori with definite principles about the nature of the
world, ourselves or m eaning.1 Philosophers cannot claim access to
some special body o f truths about knowledge (how it is to be
acquired, how certified etc.) in the light of which they are entitled
to lay down the law to other investigators. Philosophy may have its
own subject m atter - the most general features o f the world and of
our knowledge of it. But its concerns arc continuous with those of
other disciplines such as physics or psychology and it does not have
any special methodology. Any thoughts that philosophy comes up
with are to be judged in the same way as other claims, for example
by how well they fit in with other beliefs we hold firmly and by
w hether they link m atters in an illum inating way. And philosophy,
36 Instrumentalism and Meaning Scepticism
in setting out to offer ideas about m eaning, knowledge or w hat not,
does not put aside other kinds o f knowledge, seeking firm and
independent foundations quite ap art from commonsense and sci­
ence. O n the contrary, w hat it does is accept our world picture as a
going concern. It is by reflecting on th at picture, taken as mainly
correct, by noticing general features discernible in it and implica­
tions latent in it, that the philosphcr makes his contribution.
Q uine’s thoughts about m eaning arc therefore to be seen as
underpinned in the way appropriate to this conception of philo­
sophy. They are supposed to arise from commonsense truths, or
from scientific claims which are so well attested as to have the same
status.
Now Quine also claims to be a physicalist.2 And some have seen
this as playing a very im portant role in his discussion of meaning.
By Q uine’s own pronouncem ents, this in some sense m ust surely be
right. Yet it is not entirely clear w hat this physicalism am ounts to.
Wc may take it in a somewhat uninteresting formal way and also as
a more substantive thesis.
T h e former reading arises from Q uine’s view th a t it is built into
the aims of physics ’to find a m inim um catalogue o f states —
elementary states let us call them —such that there is no change
without change in respect o f them ’.3 And he also says ‘If the
physicist suspected there was any event th at did not consist in a
redistribution o f the elem entary states allowed lor by his physical
theory, he would seek a way of supplem enting his theory. Full
coverage is the very business of physics and only o f physics.’4 Being
a physicalist in the sense indicated here is, for all we have yet seen,
consistent with believing that having m eaning is an intrinsic and
fundam ental property of some items and th at its possession appears
am ong the elementary states.
Physicalism in the more substantial sense involves com m itm ent
to some particular view about w hat these elem entary states are like.
We might take it, rather crudely, to be the view th a t only material
(solid, extended) things existed and th at their only properties are
mass, velocity, electric charge and the like. M ore subtly, we could
characterize physicalism as starting from the idea th a t medium-
size material objects are the paradigm or central case o f the real;
 Instrumentalism and Meaning Scepticism 37
then the physicalist’s com m itm ent is to the idea that whatever is
spoken of in the best current theory needed to explain the behav­
iour o f such objects is all that is needed for the elem entary states. It
will be taken for granted (and rightly so given the current state of
physics) that such a theory would not invoke the m ental or the
intentional in its fundam ental vocabulary. T his sort o f physicalism
does not directly rule out the notion that some items have meaning.
But it requires that minds and meanings, if they exist, are to be
seen as in some way derived from or constructed out o f assemblages
of inam inate and noil-intentional items.
Q uine in his com m itm ent to pliysicalism endorses both the
formal version and the subtler of the substantive versions.5 And
given his views about the analytic and synthetic, it would be silly to
charge him with equivocation. T he view is that if we reflect
philosophically we can, early and easily, see the appropriateness of
some such thesis as the first, thin, physicalist one, but th a t further
exploration of our world view will soon lead us to flesh this out in
the more substantive way.
T he interesting question however is w hether in understanding
Q uine’s views alwmt m eaning we need to take the substantive
physicalism as a premise or w hether something weaker will do.
Let us leave this tricky question for the moment and look at his
central argum ent for the indeterm inacy o f translation (and hence
for the lack o f any fact of the m atter about meaning) as it is
presented in ‘O n the Reasons for the Indeterm inacy of T ranslation’
and other places in Q uine’s later writings. (It seems best to
concentrate on this paper, since Q uine explicitly remarks th at it is
more fundam ental to his thought than the argum ent presented in
chapter 2 o f Word and Object.)*'
T he line of thought may be presented as follows:

1 O bservation sentences have determ inate meanings; there is no

difficulty in establishing which sets of stim ulations in the world
they are responses to; and in virtue of this detectable correla­
tion each one will have its own distinct, identifiable (stimulus)
m eaning.7 But:
2 O bservations underdeterm ine theory; even given the truth of
38 Instrumentalism and Meaning Scepticism
all possible observation sentences, there may well be two or
more theories which are at odds with each other but consistent
w ith all the data. And:
3 In considering w hat m eaning to assign to theoretical sen­
tences, the facts about the m eanings o f observation sentences
are the only thing that could be relevant. So:
4 A scientist speaking another language, whose observations
have been translated and entirely agree with ours, can be held
to accept any of the incom patible theories which systematize
those data, consistently with all his or her patterns of linguistic
behaviour. Therefore:
5 As far as this scientist’s theoretical sentences are concerned
there is no fact o f the m atter as to which of the opposed
theories they express and hence as to w hat meaning they have.

It is unclear, looking a t this, w hether or not a substantive

physicalist assumption is doing im portant work. It may seem that
it is, and that there is a gap in the argum ent between (4) and (5)
which it is required to bridge. O n this view we would read ‘rel­
evant’ in (3) as solely an cpistemological notion. T hen we will
suppose that it is essential to an explicit account of the Q uinean
strategy to spell out the thought th at we arc not allowed to
postulate special sui generis intrinsically intentional states to provide
ontologically for facts of the m atter about theoretical meanings, left
cpistemologically unlixcd by the d ata. Such things, we say, violate
our com m itm ent to physicalisin —nam ely ‘physicalism’ taken in a
substantive sense.
M any have thought that this is the right way to read the
argum ent. But it is worth rem arking that substantive physicalism
does not appear as a premise in ‘O n the Reasons for Indeterm inacy
o f T ranslation’, nor docs Q uine examine anti-dualist argum ents in
detail anywhere in his writings. It may be th a t he takes them as so
obviously correct as to be hardly w orth setting out.
B ut there is another way of understanding m atters which does
not require us to saddle him with assum ption. It would stem from
taking ‘relevant’ in (3) in a constitutive and not merely epistcmo-
logical sense. O n this view Q uine would be inviting us to agree that
the only way theoretical sentences could acquire and m aintain a
 Instrumentalism and Meaning Scepticism 39
meaning is via determ inate linkages with observational sentences.
And indeed he docs say things like this. For example, ‘It is evident
th a t these further linguistic structures [i.e. theoretical sentences]
are based, however precariously, on the observational vocabulary
that was learned by direct confrontation and simple conditioning.’8
And ‘The sort of m eaning that is basic to translation and to the
learning of ones own language is necessarily empirical meaning and
nothing more . . . Language is socially inculcated and controlled;
the inculcation and control turn strictly on the keying of sentences
to shared stim ulations . . . Surely one has no choice but to be an
empiricist so far as one’s theory of meaning is concerned.’9 T he
notions of being ‘based’, ‘inculcated’, ‘controlled’ and ‘keyed’
which arc here invoked seem to have a constitutive rather than an
epistemological flavour.
In offering his account of our coming to have our theoretical,
sentences, Q uine is indeed relying on parts of our general picture of
the world, parts which he takes to be solidly established. Me is
thinking of very generally accepted facts about the physiology of
perception, about natural selection and the like. (Wc shall discuss
this at greater length in section 4.1.) So this view about meaning is
not (he thinks) a reprehensible deliverance of some first philo­
sophy. But w hat is interesting is th at the required parts of the world
view can be spelt out w ithout explicit commitment to a fully
fledged, anti-intcntionalist version of physicalism.
T he outcome of these reflections is as follows. It is a hypothesis
worth consideration that it is the constitutive connection between
the meanings o f theoretical sentences and the observation sentences
through which they are taught, which is central to Q uine’s argu­
ment. T he claim th at there is such a connection springs from
premises Q uine believes himself entitled to because of his natural­
ism. If this is so, then the increasingly more substantive version of
physicalism (which involves for example repudiation of the intrinsi­
cally intcnsional) would indeed be a Q uinean view, but it would be
in p art a consequence of his hostility to the notion of meaning
rather than a foundation for it.
To see how such a reading o f Q uine might work in more detail I
shall turn to examine the implications for meaning of an apparently
entirely different view, namely traditional sense datum empiricism,
40 Instrumentalism and Meaning Scepticism
when developed in a certain way. Tins m ay seem an exceedingly
bizarre proceeding, since we are here very m uch in that territory of
first philosophy, and foundations outside ordinary commonscnse
and science, which Q uine so vigorously repudiates. Nevertheless
there is interestingly, something, in common between such empiri­
cism and Q uine’s position (as I suggest we understand it), namely
the idea that theoretical meaning (in so far as we can speak of it at
all) is founded solely on the relation o f theoretical with observa­
tional statements. My conjecture is that it is exactly this similarity
which Quine means to m ark by saying that we have no choice but
to be empiricists so far as our theory of linguistic meaning is
concerned.10
W hat follows in the rest of this chapter is not, for the most part,
directly about Quine, although I shall take the opportunity at
certain points to indicate some resemblances between the theses
and arguments under discussion and some well-known Q uinean
claims. I shall in the next chapter return to more explicit discussion
of his views.
O ne point of interest in what follows will be the emergence of one
sort o f ‘holism’, which will play a pivotal role in determ ining what
is to be said about meaning. This intriguing but vague notion will
re-appear, im portantly transform ed, at later stages in the book.

3.2 SETTIN G U P A PR E D IC T IV E SEN TEN C E M A C H IN E

imagine, then, that we hold two theses which are arguably defini­
tive of a central strand o f empiricism. We believe first that there are
events in which we have unproblem atic confrontation with reality:
we exist and are conscious subjects, each one a percipient tabula
rasa, and there arc also real things in the world; the latter act upon
or get presented to the former and the upshot is an experience in
the subject in which he or she is aware, in an undistorted way, of
what is actually there. O u r second em piricist assum ption is that the
kind of item (object, fact) with which we are confronted in these
experiences is the only kind of thing that we can intelligibly claim
to exist; wc can have no conception of, and a fortiori no belief about
 Instrumentalism and Meaning Scepticism 41
or knowledge of, any other kind o f item; the concepts wc have are
and can only he of w hal we directly experience.
Let ns suppose also that we have a language in which we can
report upon the items wc are confronted with — a ‘fancifully
fancyless medium of unvarnished news’. Let us call this the ‘basic
language’ and abbreviate the sentences in it to ‘S I’, ‘S2’ etc. These
will be type sentences o f an indexical character describing the
nature of our experiences. IT we w ant some examples, we may
imagine them to have contents such as ‘hot now’, ‘hall'red and half
green then’ and so forth.
Suppose that, equipped with our basic language, wc wish not
only to be able to comment on w hal presents itself to us but also to
predict what will happen. Perhaps we merely wish to predict for the
satisfaction of seeing if we can get it right or so that wc can prepare
ourselves better to enjoy or w ithstand what occurs. But, more
plausibly, on the assum ption that we can ourselves intervene
actively to determ ine the future, we wish to be able to predict in
order to be able to control events. In either case we will need to
establish inductively the existence of certain regularities.
Perhaps the observed sequences will be simple and we shall be
able to express them neatly in our basic language with sentences
like ‘W henever SI and S2 then S3’. But perhaps w hat regularities
there arc turn out to be exceedingly complex and so cannot be
handled in such a brief way. How then might we proceed?
It might become convenient to introduce a new form of notation.
Let the items in the new notation be abbreviated as ‘'I T , ‘T 2 ’ etc.
Armed with these, our general shape of procedure in arriving at a
prediction is to assemble as many observation statem ents o f the
basic language as seem relevant, to move thence to some more
easily handled formulae in the new notation, to m anipulate these
according to prescribed rules so as to arrive at a different formula of
the new notation and finally to move back from there to a sentence
in the basic language which expresses our prediction.
It is im portant to stress at this point that the items in the new
notation, together with the rules for introducing, m anipulating and
discarding them, are invented solely to facilitate the derivation of
predictions. This is the instrum entalist stance which comes built
42 Instrumentalism and Meaning Scepticism
into our form o f empiricism. Let us call an assemblage of such
items, together with suitable rules, a ‘predictive sentence m achine’
or PSM . To have a PSM is ju s t to have a device into which one can
put some sentences of the basic language and which will, after
certain crankiugs o f the handle, deliver some other basic language
sentences at the other end. W e may, if we like, say that the items
‘T T , ‘T 2 ’ etc. arc ‘sentences’. But we m ust beware of assuming
w ithout further justification that they function like the sentences of
the basic language - that they, for example, report or describe
anything. Perhaps they do. But th at will depend upon the exact
workings of the PSM in which they occur. And it is to the varieties
of different possible PSMs that we now turn.
W hat will an effective and useful PSM be like? There are various
different cases here. In one it will be possible to devise a helpful
PSM in which ‘T T , ‘T 2 ’ etc. arc tied up one by one with (in most
eases extremely long ami cumbersome) truth functional packages
of basic language sentences. So (to take an artificially simplified
ease) ‘T l ’ might abbreviate ‘SI and (S2 or S3)’, and ‘T 2 ’ might
stand in for *S4 and S5 ami St>’. Here the ‘sentences’ of the PSM
can properly be said to describe the world. We have moved up in
level of complexity but not in conceptual terms beyond the situa­
tion where we did not need a PSM at all.
It is part of w hat we imagine in imagining this that the rules of
the PSM do not go beyond the rules we have for transforming and
inferring in the basic language, i.e. do not go beyond the resources
of classical prepositional logic. If they do then we have a somewhat
different situation. It is im portant to rem em ber that the rules of
transform ation are as im portant a component o f a PSM as its
formulae. If we have PSM formulae introduced according to strict
equivalences with basic language packages but m anipulablc in
ways which have 110 intelligible analogue in the basic language -
e.g. by perm uting the letters, writing the sentences back to front, or
otherwise applying a ‘logic’ which would be invalid for basic
sentences - then it is not so clear th at we can regard these PSM
formulae as abbreviations for basic language sentences. We would
have moved to a stance which cannot but be taken merely instru-
m entally, where we take it that we can do w hat we like with the
formulae we devise, provided the result is all right in the end. And
 Instrumentalism and Meaning Scepticism 43
we would thus not regard ourselves as answerable, in our m anipu­
lations, to some descriptive meaning which the formulae possess.
Yet another sort of PSM would move us further again from the
idea that PSM sentences had descriptive meaning. T o see how this
m ight be so let us consider a PSM which helps us - believers in
sense d ata who wish to predict future sense data - in something
like our actual world. Here the possibility of simple bundle correla­
tion, for input to the PSM, is most unlikely to be realized.
T o sec this more clearly consider an example. Suppose that our
basic language is a visual one in which we describe two-
dimensional arrays o f shapes and colours. O ur predictive problem
is akin to th a t of trying to anticipate the shapes and colours which
appear on a screen in front of us - as in the cinema. If the patterns
resemble in complexity and structure those that we sec in actual
cinemas then it is difficult to sec how a PSM could make any useful
contributions unless it introduced something which - taking up for
a m om ent a non-empiricist, non-sense datum perspective - looked
like the concept of a third dimension. In such a PSM we would
move from a set of basic language statem ents describing a sequence
of two-dimensional arrays to a sentence like ‘A small red cube is
circling a large blue sphere’. T he internal manipulation rules of the
PSM might then allow us to derive ‘T he small red cube will soon be
behind the large blue sphere’, from which we would in turn derive
some basic language statem ent about the disappearance of a red
square from our visual held.
Now it is clear that there is no way of simply pairing off sets of
basic language descriptions and ‘sentences’ of the PSM. Speaking
again from outside our empiricist assumptions, a given two-
dimensional array can be produced by projection from indefinitely
m any different three-dimensional situations and, conversely, a
given three-dimensional situation can project, by variations of
angle and distance, to indefinitely m any two-dimensional arrays.
Introduction of temporal sequences of two-dimensional arrays will
not help to eliminate this flexibility of linkage when we remember
th at the three-dimensional vocabulary will contain such terms as
‘growing’ and ‘shrinking’. A further feature which the PSM will
need to incorporate, if the world it copes with is to be anything like
ours, is something corresponding to the (perhaps undetected)
44 Instrumentalism and Meaning Scepticism
existence o f mirrors, distorting lenses and the like. Building all
these kinds of complications into the PSM makes it clear that
reductive equations are not to be had. H ence when I spoke earlier
of moving from basic language sentences to PSM sentences, that
moving cannot be a m atter of simple, m echanical replacement.
How then do we work the PSM? Given some set o f basic
observations we will have to make a choice of PSM sentence or
sentences and similarly at the other end o f the process when
deriving our final prediction. T he existence of this need for choice
docs not make the PSM useless. It only shows th a t operating it
cannot be an entirely deterministic mechanical procedure and that
we cannot expect it to deliver a unique set o f predictions. Consider
again the earlier example of the sequence o f observations which
leads me to introduce the PSM sentence ‘A small red cube is
circling a large blue sphere’. My prediction about the disappear­
ance of a red square from my visual field may turn out wrong;
perhaps the red square becomes superim posed on the blue circle
but continues to dim inish in size. It looks then as if I would have
done better, in this particular case, to have m ade the move to
‘shrinking’ rather than ‘circling’ as my response to the sequence of
observations of the changing size of the red square.
W e can see additionally from this case that slack between basic
observations and licensed PSM sentences may in practice be les­
sened by my adopting some policy with regard to w hat sort of PSM
sentences to work with when I have a choice. Perhaps, for example,
I have a policy o f using if possible sentences which talk only of
‘circling’ rather than those which speak o f ‘shrinking’ as well as
‘moving’. We may represent this as my taking it th a t things remain
the same size and move more often than they shrink and move. T he
policy thus corresponds to a further formula in the PSM language.
T his can be treated as another input, alongside the observation
statements, which helps here to determ ine my ultim ate prediction.
But of course it is not an independent input; it is only one o i the
rules of the PSM and as such subject to modification in the same
way as the others.
A PSM which is, in the way discussed, non-dcterm inistic in
operation is thus risky to use. But it may still be a good deal better
than nothing, if the choices or probabilistically determ ined moves
 Instrumentalism and Meaning Scepticism 45
turn out right in a good num ber of'cases. And I can always keep on
the look out for ways to refine and improve the rules for PSM
m anipulation.

3.3 IN D ETER M IN A C Y OK FU N C T IO N AND M EA N IN G

Im agine now that we have a given body of data specified in the

basic language and we are setting about devising a PSM. Is there
any reason to suppose that our data will yield for us a unique, best
PSM? Let us define the ‘goodness’ of a PSM purely in terms of its
yielding accurate predictions. (If wc allow such features as whether
it is easy or fun to work with then the case for the existence of
alternatives is strengthened further, since it seems likely that the
various desirable features need not all be manifested together.) It is
surely overwhelmingly plausible to suppose that the answer is ‘no’.
Even when wc do not have to move beyond truth functions and the
basic language to state the laws o f the world there may well be
notationally different but truth functionally equivalent ways of
packaging the materials. And as soon as we move to PSM s whose
sentences cannot be reduced then it is yet more compelling to
suppose that alternative PSMs arc possible, even if we in fact
succeed in thinking up only one. T he striking feature of such PSMs
is that their notation moves beyond that available in the basic
language, it breaks new ground. So the supposition that there is
only one PSM turns into the implausible view that there is only one
novel extension of the basic language which could be of use.
At this point a worry about the coherence of our empiricism in
the face of the non-reducibility point may surface. W e take
ourselves to be empiricists, and so allowed only a limited repertoire
of basic notions, But have we not breached these constraints in
supposing that wc have thought up these novel notations? T he
answer is that wc have not. But to prevent the charge sticking we
must hang on firmly to the formalist claim that to devise a PSM is
merely to hit upon the idea o f some pattern or shape which is
m anipulahle by us according to certain rules. T he shape and the
rules are both specifiable in the vocabulary of the basic language
and such merely syntactic items present no problems for the
46 Instrumentalism and Meaning Scepticism
em piricist — at least 110 problems that we have not overcome
already in imagining ourselves equipped with the basic language.
T he ‘novelty’ in question in a PSM is, on this view, merely the
novelty one gets by assembling familiar syntactic types of item in
new instances o f lam iliar syntactic relations.
T o get the llavour of the instrum entalist claim here we could
consider as an analogy increasing the complexity o f the m athem ati­
cal symbolism we use to deal with spatial position by introducing
elements standing for fourth, fifth, sixth etc. ‘spatial dimensions’. It
is a t least not obvious th at usefulness of the talk o f ‘the curvature of
space’ has to commit us to non-instrum entalist construal of such
talk. T o say, as the instrum entalist would, th at we have here
something which is merely formally analogous to the bits of notation
which describe ‘real position’ is a move w ith some attractions. And
p art of its attractiveness is precisely that some of us do not think
that we understand or can make any sense o f more than three
spatial dimensions.
A nother feature of PSMs is worth remarking. As with any
m achine designed to perform a certain function, if one extracts one
elem ent of th at machine and asks w hether it is ‘right’, w hether it is
one whose presence we endorse, the enquiry will only make sense if
we presuppose the existence and determ inate working of the rest o f
the m achine. T h e endorsability or otherwise o f a com ponent has to
do with how it interacts with the other components in delivering
the goods. T here is no such thing as inspecting a piece in isolation
and deciding th at it is a good p art for our machine. So we can
assess a particular piece of notation which is a potential component
of a PSM only if we know w hat other sentences the PSM contains
and w hat the rules for m anipulating them arc.
T h e notion of a unified machine which is to be assessed by its
perform ance o f one function thus im ports the first sort of ‘holism’
we have come across. It is holism about w hat is relevant to
assessment of parts of a machine and it stem s from the fact that it is
only the m achine as a whole which produces the output which
interests us.
If we were to change our terminology and talk not o f ‘PSM s’ but
o f ‘theories’ then these last two points are the familiar underdcter-
m ination of theory by data and D uhem ’s thesis that the various
 Instrumentalism and Meaning Scepticism 47
claims o f a scientific theory cannot be tested in isolation. These are
views about ‘theories’ which an instrum entalist will find obvious.
And let us remember th a t Q uine is a strong proponent of both
claim s."
W hat, then, are the implications of all this for meaning? C an we
speak of the meaning of a PSM or o f its component elements? The
em piricist/instrum entalist view of meaning is that an item has
m eaning if it is a (potentially true) description of what exists -
namely sense data. Note that the nieaningfulncss of elements of the
basic language - construed in the ‘realistic’ or ‘m irroring’ way 1
alluded to briefly at the start of this chapter - is not in question at
all. T his is taken entirely for granted. T he question is whether a
PSM or its parts have m eaning — i.e. (on the given construal)
w hether they describe sense data.
A PSM as a whole may be said to have meaning, in the following
way. T o utter the whole set of sentences constituting the PSM,
against the background o f some known rule lor entering, m anipu­
lating and leaving the system, can be taken as tantam ount to saying
T h e PSM constituted by these sentences and rules is a useful one’.
T h a t claim is, in effect, a description of sense d ata because it
commits the claim ant to a complex conjunction of conditionals of
basic language statem ents — namely the one that is defined by
conjoining all the conditionals we get by feeding basic language
statem ents into the PSM, seeing w hat others come out at the end
and linking input and output as antecedent and consequent. So
here we can sec why an instrum entalist might say such things as
th at a theory as a whole has empirical content. (Let us note in
passing that this is another characteristic Quinean view.)12
But what of individual components of the PSM —sentences and
words? It is clear at once th at on the empiricist conception of
m eaning (and given non-rcducibility) the individual components
can have no meaning. T o u tter a sentence in a PSM is not to make
some claim about w hat sense d ata have occurred or will occur; it is
merely a necessary stage in the working of the machine.
It does not follow from this that the meaning of a PSM compo­
nent is indeterm inate, i.e. th at there arc various ways of assigning it
meaning but one is as good as another.13 W hat follows is that it is
totally inappropriate to speak o f m eaning here at all, as inappropri­
48 Instrumentalism and Meaning Scepticism
ate as it is for the non-m orphem ic syllables of a word. T he syllables
have a role in the language in enabling us to distinguish one word
from another, but they do not mean anything in the way that a
word does.
The mention of words raises another possibility. W ords, after all,
do not have meaning in the sense in which sentences do. T hey do
not make true or false claims about reality. R ather they have their
meaning in virtue of some systematic contribution which they make
to the sentences in which they occur. So could we not say some­
thing analogous for PSM components?
Although the whole PSM is, in a sense, an enormous sentence, it
will not do to press the w ord/sentence analogy; we cannot think of
components of a PSM, components with the superficial appearance
of sentences, as being words, with roles like referring or predicat­
ing. T hat philosophically standard way of talking about word
meaning (and seeing sentence m eaning as the resultant of it) goes
together with commitments about existence —with a willingness to
adm it that there are objects which exhibit various characters and
that words connect with these objects and characters. So, although
word meaning is of a very different sort from sentence meaning, it
still involves ‘connection with the w orld’. And hence talk o f ‘m ean­
ing’ as belonging both to sentences and to words involves little
sense of strain.
But the syntactic structures which reflect this thing/property
type of metaphysics are clearly lacking in the formal relations
between a PSM and its components. So the most that wc can derive
from the proposed w ord/sentence analogy is the idea of system atic
contribution to meaning, i.e. the idea of a characteristic sort o f role
that a PSM component could play in the overall functioning of the
machine.
How can we specify the roles o f individual constituents of a
PSM? We could describe the workings of the whole m achine,
indicating at a certain point that we were talking about the consti­
tuent which was of particular interest. But this approach does not
give us the idea o f ‘the’ role o f one com ponent, something in which
it may differ from other com ponents, because every com ponent gets
the same specification — namely via a description o f the whole
 Instrumentalism and Meaning Scepticism 49
machine. So, on this way of treating the m atter all components are
seen as having the sam e role, namely that o f jointly constituting a
machine which works thus and so.
In order to isolate one role and specify it in a way which does not
bring in all the others we need some more economical vocabulary.
W ith ordinary machines we do have such a vocabulary. We can
talk o f the engine of a car and of the handle o f a mincing machine etc.
T he existence of such a standard vocabulary is m ade possible by
the com parability o f different mincing m achines or cars. We can
make sense of the question ‘W hat item in that machine plays the
role th at this item does in this machine?’ C an we do anything
sim ilar for PS Ms?
It may sometimes be the case that a person wishes to enquire
about the role in a PSM o f some piece of notation which he has not
come across before. If he is already familiar with some other PSM,
which seems to be of closely com parable structure, then it may be
informative to tell hint that the unfam iliar sentence plays the same
role as ‘T l ’ with which he is already acquainted. T he word ‘m eans’
m ight seem appropriate in such a report - e.g. ‘T l ’ means the
same as ‘T 2 ’ - because we here link two ‘sentences’ and this will be
something like the form of report we are already used to employing
for explaining the (genuine) meaning of basic language sentences.
In the basic language unfam iliar notation may need to be intro­
duced and will be explained either ostensively or by production of
another sentence with the same meaning.
Recent discussions in philosophy .of language have m ade us
familiar with the point that if the type o f rem ark we have ju st
mentioned —e.g. ‘S I’ means that S2 —is really to do its work of
telling the enquirer how ‘S I’ relates to the world then ‘S2’ must
occur in it unquoted. T he semantic report is not a mere rem ark
about synonym y.14 (O ne could report on the existence of synonymy
in the basic language — ‘S I’ means the sam e as ‘S2’ - but this is
different thing from telling someone how the new notation ‘S I’
actually relates to the world.) ‘T l ’ and ‘T 2 ’, our PSM components,
do not describe the world. So in their case it can make no difference
w hether we choose to p u t in the quotation marks or not. It is
interesting to note that Q uine (as opposed to Lewis or Davidson)
50 Instrumentalism and Meaning Scepticism
seems to see no im portant difference between studying translation
and doing semantics, and has m ade no move to alter or reformulate
his position in the light of the distinction they draw.
So possession of meaning for PSM com ponents (or the best wc
can do for an ersatz) am ounts, on this line of thought, to the
possibility of finding determ inate ways of equating or pairing off
elem ents in one PSM with those in another. If there is such a
pairing then wc can speak with confidence o f 'the* meaning (i.e.
roughly, function) o f a PSM com ponent and usefully specify such
m eanings by citing an appropriate sentence. Such ‘specification of
m eaning’ although it docs not report a connection between sen­
tences and the world will nevertheless report a real fact —by our
em piricist lights - about the sentences in question. T he elements of
the PSM are really in the em piricist’s world, as are also the facts
th a t they have such and such perceptible features and occur in such
and such patterns. T he imagined determ inate one-to-one pairing
will be of this order, if it exists. If on the other hand there arc
reasons in principle for supposing th a t pairing sometimes ojr always
cannot be dclcrm inately enforced then our ersatz ‘m eaning’ for
PSM components will lose its factual status for the empiricist.
Let us consider machines in general — floor cleaners, sausage-
m aking machines, lawnmowers and the like. Is it guaranteed that if
we take two that perform a sim ilar function and deliver the sam e
o u tp u t wc shall be able, on inspecting the innards, lo pair elements
in the one dctcrm inatcly with elem ents in the other? It may seem
that the obvious answer is ‘no’ - there can be no certainty of such
pairing. Al ter all, even if the two machines deliver the same output
they may lie designed to cope with very different sorts of input. A
m achine for making sausages out of ordinary m eat would work very
differently from one designed to cope with an input of completely
formed sausages encased in shells. T he output ‘clean floors’ m ight
be produced very differently if the input were dirty carpets rather
than dirty tiles. So it would hardly be surprising if two PSMs
designed to produce the same outp u t (under the description ‘accu­
rate predictions’) had very different structures, if different inputs,
i.e. different bodies of experience to systematize, had been sup­
plied.
M y aim in pursuing this line of thought is to throw light on
 Instrumentalism and Meaning Scepticism 51
Q uine’s indeterm inacy of translation thesis. But wc cannot suppose
that the above unrem arkable result has helped much. Q uine’s
claim is that given two languages however alike there is no certainty
o f finding a unique best translation scheme from one to the OLher.
O n the contrary there will be various alternative schemes of trans­
lation which preserve all the com m unity’s dispositions to linguistic
behaviour. This applies even when wc arc translating our own
language into itself. T he hom ophonic scheme may be the easiest
option and the only one wc can in practice think of. But there will
be non-homophonic ones which pair off sentences having no intuit­
ive claim to equivalence with each other at all.15 I f we regard
preserving dispositions to linguistic behaviour in the face of the
same experience as an instance o f preserving input and output (of
our PSM or w hatever m achine we arc considering) then the Qui-
nean claim generalized comes out as something like (but as wc
shall see not exactly like) this: T h a t given two apparently identical
machines which not only perform indistinguishably but also look
alike when taken apart there is more than one way of pairing
elements 011 the basis of identity of function. Now this is a prim a
facie startling claim, very different from the earlier truism. Can wc
throw any light 011 it by pursuing the instrum entalist line?
Quine links indeterm inacy of translation with the undcrdctermi-
nation o f theory by data. H e claims that they arc im portantly
different theses but that the latter underpins the former. Now in the
instrum entalist reading of ‘theory’ the undcrdetermination of
theory by d ata comes out as a particular version of the more
general thesis th a t two machines which perform identically (pro­
cessing identical input into identical output) may nevertheless have
- strikingly different internal organizations. This seems unexception­
able. But som ething im portant th at it brings to our attention is that
to grasp Q uine’s view properly wc need to be willing to deploy,
initially a t least, the distinction between language ancl theory, or, in
m achine term s, the distinction between an array of equipment
(nuts, bolts, cogs, belts, words, rules) and the machines that one
might build from it. T he underdeterminalion thesis more accurately
expressed is th at from some extensive, imaginable set of equipment
two identically performing machines can be built, which are never­
theless significantly different in internal working. T he indeterminacy
52 Instrumentalism and Meaning Scepticism
o f translation, extended to the case o f machines in general, comes out
like this: if we consider two equipm ent sets, not from the point o f
view mere shape (= pure syntax) but from the point of view o f
potential use in machines (= m eaning), then there will be more
than one way of pairing them. For example, even in two quali­
tatively indistinguishable equipm ent sets, it is not m andatory to
pair off each cogwheel with its qualitatively indistinguishable counter­
part in the other set; it could defensibly be paired with an axle,
pulley or bolt, i.e. the cogwheel could be thought o f as contributing
the same function to potential machines as the axle, pulley or bolt.
C an we make any sense o f this idea?
Let us start by asking w hat underlies the undcrdcterm ination,
what makes it possible that there should be two machines? T he
precondition is that the defining differences between input and
output can be broken down into collections o f subdifferences in a
variety of ways. For example consider the contrast between chunks
of meat and sausages. They differ in two im portant respects, texture
and shape. So if one could first change the texture and then change
the shape one could transform the one into the other. But equally if
one could half transform both texture and shape and then complete
lx»th transformations one could produce the same result. O u r
sausage machines work on the first principle; we mince the meat
and then stuff it into the skins. But we could imagine a machine
which first produces sausage-weigh ted balls o f partially re-textured
m eat and then puts these balls through a process which sim ulta­
neously modifies the texture and encloses the balls in properly
shaped skins. Let us put this schem atically. We have in the one
case an X -er and a Y-er operating sequentially and in the other a
W -er and a Z-er operating sequentially.
T h a t these two machines are different is not in dispute - ju st as
it is not disputed by Q uine that two languages or theories may be
different in the sense o f being distinguishable collections o f marks,
contrasted in shape and m anipulation rules. T he question however
is about pairing of elements on the basis of function performed. So
let us ask ‘W hat is involved in som ething being an X -er, e.g. a
mincer?’ This is not so clear cut as one m ight suppose. Transform ­
ing meat into sausages necessarily involves changing the texture of
the meat. T hat is ju st a trivial consequence of the specification of
 Instrumentalism and Meaning Scepticism 53
the overall functioning o f a sausage-m aking m achine. But then
there m ust be something which does this in any m achine which
actually delivers the goods. So that thing, w hat ever it is, can be
called the mincer. As it turns out, the mincing function in our
second machine is going on continuously throughout the course of
sausage m anufacture. B ut that could ju st lead us to say th at the
second machine consisted o f an X-er and a Y-er (m incer and
shaper) working in parallel, rather than in the fam iliar sequential
way.
Suppose now th at I have an equipm ent set sufficient to build
either type of m achine and th at you do also. Suppose further that
we have in fact constructed identical machines. It follows from
w hat we have said above that even so it is not obligatory to pair ofr
pieces of the equipm ent sets, functionally speaking, in the obvious
‘homophonic’ way. I may do that, of course, but 1 can also equate
the pieces which it is most natural to regard as my W -cr with those
pieces of your equipm ent set from which you could have made an
X -er had you chosen to construct the other type of m achine. So for
example my W -er involves a cogwheel; but I pair th a t cogwheel not
with your cogwheel (which is actually doing exactly the same thing
in your machine as it is in mine) but with the pulley from which
you could have m ade the X-cr. And conversely 1 pair my pulley
w ith your cogwheel. How can I justify this? My grounds are that it
is possible to regard an X -er as p art of a W -er. It all depends on
how I choose to segment the machines anti w hether the idea of, say,
finding ‘the m incer1 seems to me im portant. T he language and
translation version of this is that 1 can translate you - who
seemingly speak English and affirm ju st the sam e sentences as I do
- as expressing by your utterances an entirely different view o f the
world, namely the view you would have had had you adopted the
alternative PSM which I recognize to be possible.
T here is som ething very odd about this in the case o f physical
machines and one may well w ant to resist these conclusions. There
are, after all, physical constraints in ease of handling and repairing
machines, which make certain kinds of segm entation overwhelm­
ingly useful to us and rule out others as mere freewheeling fantasy.
But these constraints will not help us to resist the Q uinean-
sounding conclusions in the case of the PSMs because in their
54 Instrumentalism and Meaning Scepticism
realm technological limitations are at a minimum ; the sentences
and rules provide no resistance to any segm entation or m anipula­
tion.
W e can sec then how, following the instrum entalist line, the
central Q uincan thesis of indeterm inacy of translation emerges
w ith some plausibility. M oreover, this interpretation has the ad­
vantage of showing how the indeterm inacy thesis links with the
undcrdetcrm ination one while still being a distinct claim .16
W h at then of the attack on the distinction between analytic and
synthetic truths? O n one way o f looking at the m atter, the non­
existence of analytic truths (i.e. ones guaranteed by meaning) is a
direct consequence of the non-existence of determ inate assignment
o f function to PSM parts. Supposed discovery of analytic equiv­
alences is ju s t discovery that one bit within a language is a transla­
tion of another.
13ut there is another way o f looking at the m atter which is
perhaps more illuminating. In ‘Tw o Dogmas o f Em piricism ’ Q uine
argues lor the non-existence of a linn class of analytic truths (which
I shall treat, as Q uine docs, as the sam e as the non-existence of
necessary truths) on the grounds that, given bizarre enough d ata to
systematize, any general statem ents or rules could in principle be
discarded. He is sometimes criticized for confusing the question of
w hether one could be mistaken about necessity and come to recog­
nize that mistake with the question of w hether there are any
necessary truths. H e is said to move from a perhaps acceptable
fallibilism to an unw arranted scepticism about necessity. But on
the instrum entalist reading the criticism is unjustified. T h e best
th a t the instrum entalist could do lor the notion o f ‘necessary tru th ’
in a PSM is to find elements, rules or functions, which have to be
present in any PSM whatsoever. If we can find features which no
usable PSM could lack then they could be said to be indispensable
because of their function i.e. necessary on account of their meaning.
But given that the test of a PSM is entirely in its working it would
seem rash to take such a dogm atic stand th a t ‘whatever is altered in
my PSM this shall stand firm.’ An easy-going and pragm atist
attitude to PSM features should be the instrum entalist one. And
this is ju st w hat wc find in Quine.
 Instrumentalism and Meaning Scepticism 55
3.4 IN ST R U M E N T A L IS M AND T H E REV ISA BILITY O F
L O G IC

It has frequently been rem arked th at Q uine seems to be am bivalent

about the status of classical logic. O n the one hand he claims that it
is révisable, up for holistic appraisal like any other element in our
intellectual system (even if it is extremely unlikely that wc will
substantially revise it); on the other hand he insists that change of
logic is change o f subject.171 shall suggest that the instrum entalist
ideas explored in the previous sections provide a framework for
seeing how this clash m ight arise.
Let us make explicit the assum ption th at we have operated with
so far, namely that the sentential operators ‘not’, ‘and’, ‘i f . . .
then . . .’ and so forth arc part o f the basic, language and that when
we speak the basic language wc reason in classical style. This is no
more than a reflection of the robust m irroring realist outlook we
have at this level. But w hat o f such operators and the principles
governing their use in a PSM?
A problem immediately presents itself when wc pursue this
thought. Suppose we try to raise the question ‘Is it conceivable that
we should wish to modify the PSM so as not to include in it the law
of non-contradiction or modus lollensV Wc discover that wc have not
succeeded in asking anything coherent. Finding ‘the law of non­
contradiction’ or the like in the PSM depends upon making sense of
the idea that some of the PSM com ponents arc to be identified as
m eaning ‘not’, ‘i f . . . then . . .’ and the like. O ur difficulty is not
ju st th at we cannot identify any items as negation, conditional etc.
unless they obey classical laws. It is the much more primitive
difficulty that we have given no sense to the idea that PSM
components could have this sort of m eaning at all. Remember PSM
components are merely shapes (or noises or whatever) which we
m anipulate. T heir production by us is not assertion, they cannot be
denied or inferred. So it looks as if the appearance o f an ‘and’ or a
‘not’ in the PSM could be no more than a pun. And then the whole
question of the status of logic in the PSM collapses.
B ut this is prem ature. We can, on reflection, see how logical
operators might, in a not wholly punning way, be incorporated into
56 Instrumentalism and Meaning Scepticism
a PSM and thus give rise to the question about the revisability of
logic.
Let us note first th at there is likely to be a tendency to abbreviate
remarks about the PSM to utterances which, in a different context,
would constitute actual operations of the PSM. If a large part of
our lives consists in thinking and debating oti how to improve our
PSM wc may fall into the shorthand of uttering merely ‘T l ’ when
we m ean ‘Let’s pul “T l ” into the m achine’. O r we say ‘I f T l then
T 2 ’ meaning ‘If we put in “T l ” we should put in “T 2 ” as well’.
And we say ‘Not T 2 ’ to express disapproval o f the inclusion o f ‘T 2 \
This does not give us logical connectives as PSM elements but
one further move puts the operators inside the machine. This move
is taken if we develop the idea of a self-modifying m achine, i.e. we
recognize not only rules about how PSM elements should be
m anipulated but rules about how those rules should be modified.
It seems likely that in any case o f modifying a machine, sensible
modification will often come in groups. Suppose that I have a
mangle which every now and then gets into an unfortunate con­
figuration in which a part sticks out and trips me up. I partially
dism antle the m achine and discover that the obstructive item is
linked to something else which occasionally moves in such a way as
to push the offending part out further than usual. I can now
relocate this further thing, alter the linkage, or pursue yet deeper
into the works. Cut a very probable scenario is th at I shall have to
make a sequence o f adjustm ents. For exam ple merely undoing the
linkage may produce some serious malfunction elsewhere, preven­
tion of which requires further manoeuvres.
W ith a PSM one m ight analogously discover th at every now and
then it produces ‘T 2 ’ and that use o f th at ‘T 2 ’ leads to false
predictions. ‘T 2’ is produced, we discover, because the PSM con­
tains ‘T l ’ and in the operating instructions we have the rule ‘Put in
“ T 2 ” when “T l ” is present’. Suppose I express this rule to myself
in the abbreviated form mentioned above; I say ‘If T l then T 2 ’. O n
discovering that operation o i the m achine is producing trouble in
the way outlined and taking the view th a t the machine has got to be
prevented from producing ‘T 2 ’, I realize th at I have certain
options. Among these are, first, removal o f ‘T l ’ and, secondly,
modification of the rule. But perhaps I am determ ined to hang on
 Instrumentalism and Meaning Scepticism 57
to the latter. My best policy then seems to be to remove ‘T P .
Supposing this sort o f modification to be needed fairly frequently
Ï m ight devise for myself a higher level instruction which (using ‘P’,
‘Q ’ to capture the generality I need) I lind it natural to express as
follows: ‘If, if P then Q and not Q , then not P \ O r a related
instruction might be ‘If not Q then either not P or not (if P then Q )\
I have, in doing these things, taken up a stance in which the
PSM together with the rules for operating it (which all along I
expressed using logical operators in im peratival sentences) be­
comes itself an assemblage subject to further rules of modification.
I have rules about w hat rules 1 shall adopt for the m anipulation of
the basic elements, lh it it is im portant to rem em ber that these
higher level rules are not purported statem ents of truths. If we
unpack the one given above in a particular case it comes out like
this:

If you hang on the rule ‘Put in “ T 2 ” if you have put in “ T P ” and

you do not want to put in ‘T 2 ’ then do not put in ‘'PP.

O r we might express it in a kind of interm ediate notation thus:

If you include ‘If T1 then T 2 ’ and also include ‘not T 2 ’ then

include ‘not T P .

What does ‘including “not T 2’” amount to? It is, in cflcct, writing
myself a note to remind myself to take ‘T2* out of the machine. But
(he writing o f such notes has now become part o f operating the
machine. T he notes themselves arc, in a way, parts o f the machine.
T he upshot, then, of allowing the conception of the self­
modifying machine is to make natural the use of items like ‘not T 2 ’
which are, regarded from one perspective, parts o f the m achine but
regarded from another are instructions on how to work the m a­
chine. ‘Not* thus occurs in the machine and its occurrence is not a
pun; its occurrence is an intelligible extension of its use in the basic
language, namely in the description of and giving of instructions
about the sense data and sentences.
Having in this way put logical operators into the m achine we can
ask a question about classical logic, namely "Must we accept as an
58 Instrumentalism and Meaning Scepticism
overriding principle o f m achine construction the need to keep
logical operators distributed am ong the sentences of the PSM
according to only classically acceptable patterns?’
Suppose for example th a t I have a PSM and, as a result of
various operations, it comes to include ‘T l ’, ‘if T 1 then T 2 ’ and
also ‘not T 2 \ I notice this configuration. M ust I now say to myself
‘T here is something terribly wrong with this machine which I am
obliged to attend to’? Do 1 become subject to necessities in my
handling of my machine in virtue of having adopted this self­
modifying policy and the notation which naturally goes with it?
T he symbols ‘not’, ‘i f . . . then . . etc., could not appear in the
PSM w ithout being mere puns unless there was some presum ption
th at I would at some stage of the operation of the machine take
notice o f the instructions that 1 give myself by including them. I am
ex hypolhesi in the business of modifying my PSM in a systematic
m anner and the general patterns of modification arc w hat justify
the use of these particular mnemonic marks. But this observation
alone docs not show that 1 shall find myself subject to necessities.
And there arc two further considerations which show that it
would be inappropriate to insist th at classical logic be preserved
within the PSM. T he first is that PSM construction is only one
among various projects which 1 may have. W ith the inconvenient
mangle I may decide that I cannot be bothered to dism antle it. I
will instead keep a good lookout for the obtruding piece and ju st
shove it back in when it threatens to be a nuisance. Similarly with
my PSM 1 may merely keep a good lookout for ‘T 2 ’ when it turns
up and remove it before it generates any misleading predictions,
without bothering to readjust the innards so that it docs not appear
again. To do this is to adopt a PSM which says ‘T 2 ’ and at the
same time ‘not T 2 \ But if life goes on all right why should 1 worry?
Well, it may be said,' clearly you have not got an entirely
trouble-free PSM; it is imaginable that you could devise a better
one; you should stop lazing about enjoying your sense d a ta and get
on with improving your PSM. Perhaps I take this advice to heart.
O r perhaps I like constructing PSMs. O r perhaps I think that I
might get even better sense d ata if I got rid of this occasional
awkwardness of the current PSM. So I have now as my goal the
devising of the maximally smooth-working, economical, reliable
 Instrumentalism and Meaning Scepticism 59
etc. PSM. Docs this restructuring o f my objectives give us w hat we
want, namely the idea th at I should recognize myseiras subject to a
necessity o f preserving classical patterns for the connectives? It
does not.
T h e problem is that we do not have and cannot have a guarantee
that there is a PSM which preserves the classical patterns and
works better overall than the imagined non-classical one. Until I
have found the alternative I am rationally justified in continuing to
use the non-classical PSM and need have no qualms at all about
doing so. T he ad hoc policy adopted earlier on grounds of laziness
may be the best policy even when laziness has been discarded.
A nd, more elaborately, we can see how such things as quantum
logic m ay turn out to be highly useful for incorporation into certain
parts o f the PSM.
My suggestion, then, is that this sort o f approach might make
some sense of a Q uinean style o f ambivalence about classical
propositional logic. O n the one hand wc need it as an clement in
the basic language but on the other hand when it, or its extension,
gets incorporated into a many layered PSM we want it to be
malleable, as other PSM elements arc.
 4
Quine’s Naturalized Empiricism

4.1 Q U IN E ’S E P IST E M O L O G Y

T he contention of this chapter is that we can best understand

Q uine by supposing that he rejects sense d ata and the ‘first philo­
sophy’ underpinnings of the instrum entalism explored in the last
chapter, while carrying many structural features of the position
into a naturalized version o f empiricism. T he retained features
include the existence of a sharp distinction between observational
and theoretical sentences and the constitutive thesis about theoreti­
cal meaning, namely that what makes theoretical sentences have
whatever meaning they do have is their relations with observa­
tional sentences. T h e upshot is that the sceptical consequences for
theoretical meaning implicit in instrum entalism also arc retained.
T o make this interpretation plausible I shall first examine the
evidence for resemblances between Q uine’s later epistemology and
instrumentalist empiricism. Then 1 shall, in later sections, consider
the ramifications o f the position sketched. T he theme here will be
that a non-instrum entalist interpretation lands Q uine fairly di­
rectly into gross and obvious difficulties. T h e naturalized version of
instrumentalist empiricism, on the other hand, does indeed require
one to do some very delicate balancing acts, concerning notions
such as ‘real’, ‘tru e’ and the like (balancing acts of a type which we
find Quine attem pting) but can be m ade to yield a whole picture
which is less obviously incoherent, if exceedingly strange. In the
final section I shall ask w hether we need to accept this unfamiliar
vision of ourselves.
In 1951 Q uine wrote

As an empiricist I continue to think o f the conceptual scheme of

science as a tool, ultimately, for predicting future experience in
 Quine’s Naturalized Empiricism 61
the light o f past experience . . . T he m yth of physical objects is
epistemologically superior to most in that it has proved more
efficacious than other myths as a device for making a m anage­
able structure of the flux of experience.

In the famous paper from which that comes he also introduced his
image o f the web o f belief: ‘T h e totality o f o u r so-called knowledge
or beliefs, from the most casual m atters of geography and history to
the profoundest laws o f atom ic physics or even of pure m athem atics
and logic, is a m an-m ade fabric which impinges on experience only
along the edges.’1 T hirty years later in 1081 his com ment on his
earlier formulation is this: ‘It was an interim indication of an
attitude and an altitude that I still hold.’ But he adds ‘My non­
committal term “ experience” awaited a theory.’'’ His later theory of
‘experience’ is revealed in this:

O u r talk o f external things, our very notion of things, is ju st a

conceptual apparatus that helps us to foresee and control the
triggering of our sensory receptors in the light of previous
triggerings of our sensory receptors. T h e triggering, first and
last, is all we have to go on.3

T he striking feature of*these quotations is the persistence in them

of the ideas o f something (sensory experience or triggerings of
sensory receptors) which is to he predicted, foreseen, m anaged or
controlled, and of som ething else (tool, m an-made fabric or appar­
atus) which we devise to do the job. Let us look at the two elements
in turn, starting with Q uine’s notion of experience, his rejection of
sense data and his later idea of an observation sentence.
Q uine speaks mockingly in Word and Object of ‘a fancifully fancy-
less medium of unvarnished news’, a ‘protocal language’, in which
we could offer an ‘account of the passing show’ or of the ‘unsullied
stream of sense experience’. 1 It is clear that he thinks that we do not
have such a language and that we cannot attain one by rational
reconstruction. But if we look in more detail at his reasons for
denying the ‘fancyless m edium ’ various interesting facts emerge.
He remarks that it is references to physical things which hold
together the domain o f im m ediate experience which would other­
wise not cohere and that reference to physical objects gives us ‘our
62 (¿nine’s Naturalized Empiricism
main continuing access to past sense d ata themselves; for past sense
d ata arc mostly gone for good except as comm emorated in physical
posits’.3 T he thought here is that if I set out to think about w hat my
sense experiences have been like in the past lew minutes w hat I
actually do is to consider the objects around me, how my eyes have
been oriented and the like, and to reconstruct from this w hat I m ust
have seen and felt. And surely this is right. T he point sccnis to be
th at ‘actual memories mostly arc traces not o f past sensations but of
past conceptualization or verbalization’.6 Later he says ‘We cannot
rest content with a running conceptualization of the unsullied
stream of experience; w hat we need is a sullying of the stream .
Association of sentences is wanted not ju st with non-verbal stim u­
lation but with other sentences if wc arc to exploit finished concep­
tualizations and not ju s t repeat them .’7 T he emphasis here is on the
limitations of reports o f the unsullied stream ; by themselves they do
not supply any predictive tools.
Consider also the following quotation.

Even a strictly sensory idea is elusive unless it is reinforced by

language. T he jxjint was m ade by W ittgenstein. U naided by
language wc might treat a great lot of sensory events as recur­
rences of one and the same sensation, simply because of a
similarity between each and the uext, and yet there can have
been a serious cum ulative slippage of similarity between the
latest of these events and the earliest of thcm .H

T he view here seems to be that wc can recognize similarity between

one sensation and another but, as things are, not accurately enough
to ensure that only ones exactly or really closely resembling the first
get called by the sam e name; wc lose grip on the standard of
sim ilarity set by the first, but only because o f some confusing scries
o f interm ediate events.
O ne striking thing that emerges from these passages is that the
rejection of sense data seems curiously half-hearted; the general
impression given is th at they occur but are elusive and tend to be
remembered not a t all or inaccurately. A sense datum instrum en­
talist with a bad memory would be happy with w hat Q uine says.
But perhaps this is ju s t a rhetorical device on Q uine’s part; perhaps
 Q jnm ’s Naturalized Empiricism 63
he is ju st trying to loosen the grip o f a simplc-mindccl sense datum
view with an eye to more radical alterations later.
M uch more im portant is the fact that none of Q uine’s objections
arc objections in principle to the very idea of an event with (many
of) the essential characteristics of the supposed confrontations with
sense data. In the latter part of Ward and Object, and in subsequent
writings, the idea of the sense datum report docs indeed fade away
—after its last ghostly appearance in the remarks al>ovc. But in its
place we arc olfercd the idea of an observation sentence, which, as we
shall see, has a very good claim to be ju st another sort o f ‘fancyless
m edium ', whose main dilfcrcnce from the sense datum report is its
subject m atter rather than its cpistemological characteristics.
T he notion of an observation sentence is introduced by Q uine in
Word and Object and the account is slightly modified and clarified in
certain respects in other places.51 W hat is interesting is that it is
apparent that the guiding principle in constructing the account
given is to arrive at a notion o f a kind o f a remark which is merely a
response to w hat a person is currently confronted with, in the case
o f sense datum reports the ju d g cr can tell by introspection whether
a rem ark has this character. But if we have rejected that way of
approaching things and have adopted the third person perspective
within which Q uine works, we cannot find out ju st by looking at
the linguistic behaviour of one subject whether his reports arc of
this kind. Ilis responses will necessarily depend upon past experi­
ence as well as w hat he is currently confronted with, because he
m ust have had some experiences in learning the language. And we
cannot separate out, by looking at him alone, which of those
experiences determ ine the meanings ofscntenc.es and which contri­
bute to possession of collateral information which may influence
the judgem ents made.
At this point in Q uine’s exposition one might think that the
notion of a linguistic response which is merely a non-committal
labelling of the current stim ulus is doomed and that it is Q uine’s
own rejection of the analytic/synthetic distinction (which is another
aspect of the m eaning/collateral information distinction) which has
brought about this fate. B ut Q uine is determined to restore the idea
and docs so by the ingenious move of bringing in other speakers.
‘An observation sentence is one on which all speakers of the
64 Quine's Naturalized Empiricism
language give the same verdict when given the same concurrent
stimulation. To put the point negatively, an observation sentence is
one that is not sensitive to differences in past experience within the
speech community.’10 An interesting feature of this definition is
that obscrvationalitv is clearly a m atter of degree. It will depend
upon how wide a speech community we consider and how much
agreement in verdict we dem and. Q uine explicitly acknowledges
this; and the moral should be (I think) that the notion of a sharp
observation/theory distinction is untenable and that other related
theses (like thoroughgoing cpisternological holism) follow. It is
striking that this is not the path Q uine takes; rather he chooses to
ignore those concessions anti to treat the notion of observation
sentence as one with a sharp boundary and to suppose, in conse­
quence, that an observation sentence can have a well-defined class
of stimulations which elicit it and which it can be said to mean.
T he observation sentence which emerges from this definition is
notably unlike the sense datum report in not presupposing con­
sciousness and in not being about som ething which is private to the
subject. But in other respects it rem arkably resembles it. T he
observation report is som ething which is a direct, non-infercntial
response to certain patterns of stim ulations of nerve endings. It ‘has
its particular range o f admissible stim ulating occasions indepen­
dently of wider context’.11 And ‘observation sentences peal nicely;
their meanings, stim ulus meanings, emerge absolute and free of
residual verbal tain t’.12 'Thus the observation sentence is free o f the
Duhcmian web and can be verified or falsified in isolation. ‘In my
view observation sentences can indeed be cognized in isolation,
independently of scientific theory as a whole, by the closeness of
their association with stim ulation.MS A corollary of this is th a t they
are (pretty nearly?) infallible. A nd, as Quine him self remarks,
‘speech thus confined would be strikingly like bare reporting of
sense d ata.’14
There is an im portant qualification and clarification of this,
concerning isolation and infallibility, which is worth noting. It is
only when regarded holophrastically, with the meanings appropri­
ate tu sentences lacking significant internal or referential structure,
that observation sentences can be regarded as being conceptually
and cpisternologically isolated. It is possible for us, however, to pay
 Quine’s Naturalized Empiricism 65
attention to the words th a t we discern in some observation sentence
and think o f the m eaning o ftb c sentence as something only to be
characterized in terms of the notions that are brought in when
analytical hypotheses u I k h i I these wortls have been accepted. If we
do this then we shall (ind that the observation statem ents look less
like sense datum ones. T heir meanings become enmeshed with
those of theoretical statem ents and, because of this enmeshing of
m eaning, theoretical sentences come to have some bearing on the
episternological acceptability of these observation sentences. This
may lead to the overturning of an observation statem ent by theo­
retical considerations. At this point it may look as though Quine
has recognized the points mentioned earlier and is urging an
observation/theory distinction which is merely a m atter of degree.
But things arc not so simple. lie also says that what is then re­
jected is not the original observation report. Consider the following.

These occasion sentences qualify holophrastically as observa­

tional, but their parts recur in terms in theoretical science. T h at
is how observation sentences link up with scientific theory,
bearing evidence. If a recorded observation wildly at odds with
theory is ultimately repudiated, sis /nay happen, still what is
repudiated is a dated statement o f history, itself theoretical. and not the
occasion sentence at its utterance.l5

Q uine here says that w hat is repudiated from theoretical considera­

tions is theoretical —a dated, and hence eternal, sentence concern­
ing the existence of some theoretical item at the specified time.
T hus it need not be taken to be the same as what was originally
asserted, which was (so Q uine seems to suggest) some indcxical
and holophrastically construed claim about the nature of current
stim ulation. 1'his distinction (if I have interpreted the passage
rightly) seems to be designed to preserve in full force the infallibil­
ity of the observational.
We have thus far looked at Q uine’s continuing com m itm ent to
the idea of observation reports which are risk-free labelling re­
sponses to stim ulation. Let us now turn to the other element of the
instrum entalist picture — namely the idea o f the tool or apparatus
which we devise to handle o r predict our observations.
66 Quine’s Naturalized Empiricism
Q uine writes, as we saw above, ‘O u r talk o f external things, our
very notion of things, is just a conceptual apparatus that helps us to
foresee and control the triggerings o f our sensory receptors.’ He
continues this passage as follows: ‘T he triggerings, first and last,
arc all we have to go 011 . . . W hat I ain saying is not m eant to be
sceptical . . . But there remains the fact - a fact o f science its e lf-
th at science is a conceptual bridge of o ur own making linking
sensory stim ulation to sensory stim ulation.” 6 And a t another place
he says something related:

Science itself tells us that our inform ation about the world is
limited to irritations of our surfaces and thus the cpistctnological
question is in turn a question within science: the question of how
we hum an anim ats can have m anaged to arrive at science from
such limited information. O u r scientific cpistcmologist pursues
this inquiry and comes out with an account that has a good deal
to do with the learning of language and with the neurology of
perception. He talks of how men posit bodies and hypothetical
particles, but he docs not mean to suggest that the things thus
posited do not exist. Evolution and natural selection will doubt­
less figure in this account.17

O ne thing to note here is th at Q uine explicitly distances himself

from possible anti-realist im plications o f this instrum entalist
sounding talk. And we shall need to return to the question o f w hat
Q uine’s realism am ounts to. B ut for the mom ent we need to gel
clearer about what he means by the fact o f science, namely that
science is a bridge of our own making linking sensory stim ulation to
sensory stim ulation. W hich parts of our generally accepted picture
of the world is he here calling 011?
M ost probably he has in mind the following kinds of considera­
tion. '1'he behaviour o f our bodies, including our movements,
production of noises etc., is determ ined by a combination of im­
pinging physical stimuli (light rays, pressure, heat etc.) and the
given internal state of the body. The current internal state is the
result of previous impinging stimuli and previous internal states.
M atters can be traced back ultimately to our genetic constitution
and the environm ent in the womb. T h e theory of natural selection
tells us th a t we have the kind of genetic constitution we do because
 Quine’s Naturalized. Empiricism 67
its possession contributed to those who had it in the past behaving
in ways which resulted in more of their olTspring surviving than
those of creatures with different genetic constitutions. Surviving
and reproducing are, in large part, a m atter of being so constituted
as to produce, on a certain stim ulus, behaviour which will result in
certain future patterns o f stim uli rath er than others. T he causal
effects o f the environm ent on the development o f species and
individual animals is all (given the non-existence of action at a
distance) m ediated via causal effects on the exterior surface of the
anim al. N atural selection mimics teleology. So one may say that
our internal constitution (i.e. genetic structures, the development of
the nervous system which it largely controls, and the. consequent
disposition of the nervous system to change under the im pact of
stimuli) is an apparatus, the biological function of which is to
ensure that the organism anticipates and controls the triggering of
sensory receptors in the light of previous triggerings. But different
genetic constitutions m ight, for all we know, be equally effective in
producing the same external behaviour, i.e. the same stimulux-l)e-
haviour-stim ulus sequences. Equally, for the individual animal it
is highly plausible th a t different ncurophysiological structures
could have developed in response to given stimuli and have pro­
duced indistinguishable pattern s of future stim ulus—behaviour
correlations.
It is this set of facts, about physiology, evolution and so forth
which, I conjecture, Q uine sees as justifying his remarks about
w hat science is and his related claim that we have no choice but to
be empiricists as far as our theory of linguistic meaning is con­
cerned. There are, as I shall suggest in the last section of this
chapter, reasons for grave dissatisfaction with the reading Quine
offers of the scientific facts. But these wc shall set aside for the
moment. T he question we need to consider at the moment is
w hether the view of science, and in particular of the nature of the
meaning of scientific sentences, is instrumentalist.
H ere I would like to m ention three considerations. O ne is the,
already noted, constitutive flavour of the remarks about meaning
which I quoted above, in section 3.1. T he second is the distinctly
- instrum entalist tone of those Q uincan remarks about science which
wc have ju st set out. (‘O u r talk of external things . . . is just a
68 Quine's Naturalized Empiricism
conceptual apparatus’'. ‘Science is a conceptual bridge of our own mak­
ing’.) T he third is his treatm ent o f the postulation of physically
respectable inner items which m ight be thought to supply some
facts for translation m anuals to describe. This third topic deserves
a little more discussion.
In a discussion of the idea o f differentiating in point of correct­
ness between two translation schemes which fit all observable
behaviour, Quine remarks

Since translators do not supplem ent their behavioural criteria

with neurological criteria, much less with telepathy, w hat excuse
could there be for supposing th a t the one m anual conformed to
any distribution of elem entary physical states better than the
other manual? W hat excuse, in short, lor supposing there to be a
fact of the m atter?18

O ne thought Quine seems to be pursuing here is that it would be

wrong to think that translators are using behavioural d ata to make
guesses about neurological states. I f they were then they would use
more direct information about, neurological states if they could get
it. But they show no disposition to do this; the neurological d a ta are
of no interest to them. And so two translation m anuals cannot be
seen as differentiated in truth value by appeal to hypothetical
differences in neurology. B ut the same holds, Q uine also suggests,
for items which might be the object of telepathy. Presum ably he is
here thinking of some supposed intrinsically intentional states of
mind-stuff.
T he supposed conclusion is th a t imagined inner items which arc
epistemologically underdeterm ined by observable behaviour -
whether these items are neurological or m ental - are not w hat
translation is about. T he argum ent o f the passage quoted is not
that we know that some sort o f item — e.g. non-physical ones —
would do the job, but unfortunately our physicalism rules out
postulating them. As the m ention o f neurology shows, it is not the
non-physical nature of the postulated items which is the sticking
poim, despite the rhetoric about ‘distribution of elem entary physi­
cal states’. Q uine’s point seems to be rath er that the enterprise of
translation, as we actually conduct it, ju st is not a theoretical
 Quine’s Naturalized Empiricism 69
enquiry after unobservables, w hether meanings or neural states.
It is interesting to think o f K ripke’s interpretation o f W ittgen­
stein at this point.15* Both Q uine and K ripke’s W ittgenstein main­
tain th at the kind o f fact that we arc tem pted to pu t in, by way of
(leshing out w hat makes m eaning statem ents true, turns out not to
be w hat we were after. But, on the interpretation I wish to recom­
mend, the parallel breaks down beyond this point. K ripke’s W itt­
genstein (whom we shall discuss further in section 7.3) thinks that
wc cannot find such facts about meaning because o f a link between
the notions of meaning and the normative. Also he thinks that the
point is quite general and sutlic.es to underm ine the idea of a fact
about meaning for all kinds o f sentences. Q uine, on the other hand,
thinks that observation statem ents have clear stim ulus meanings.
And his reasons for denying facts about theoretical meaning do not
have to do w ith norms. M oreover K ripke’s W ittgenstein thinks that
we cannot, on reflection, make sense of the idea o f there being a fact
about meaning; we cannot get our m inds round any coherent
conception of w hat such a fact could be like. Q uine does not seem
to be hesitant on these grounds: if translation procedures could be
shown to deliver a unique m anual, Q uine implies, then we need
have no qualm s in talking of facts about meaning: such facts would
simply be w hat was stated by the relevant parts of the uniquely
correct m anual.
But if it is not the non-physical nature of meanings which is the
trouble, nor yet the fundam ental unintclligibility o f facts about
such things a t all, what is Q uine’s root thought here? An in­
strum entalist assum ption about the nature o f theories provides at
least one answ er to this question.
It may be thought however that the three points I have ju st
mentioned are but flimsy evidence on which to base an instrum en­
talist reading of Q uine, in the face of his denials o f sceptical intent
and repealed professions o f realism. So I shall turn in the next
sections to a fourth and stronger kind of consideration, namely that
of internal consistency. I shall argue th at, contrary to w hat one
might at first suppose, Q uine will have much more trouble main­
taining coherence if we read him non-instrum entally than if we
read him instrum cnially.
70 Quine’s Naturalized Empiricism
4.2 O N T O L O G IC A L R E L A T IV IT Y AND
D IS Q U O T A T IO N

T h a t Q uine’s views about m eaning rest upon the constitutive thesis

about the relation of observational and theoretical meanings, which
is in turn bound up with som ething of significantly instrum entalist
style about the nature of theory, has seemed to many implausible
because, as we have seen, Q uine again and again denies that his
stance is sceptical and claims to be a realist about the subject
m atter o f theories.
So the crucial question, for the spelling out of Q uine’s doctrine
w hich 1 wish to defend, is w hether we can make sense of Q uine’s
realism in a way which preserves at the same time both a signifi­
cant elem ent of instrum entalism and also something which Q uine
would call ‘realism’. I believe th at we can, but what emerges under
the nam e o f ‘realism’ is som ething which docs not come under any
of the three types discussed in chapter 1 and which many might
have difficulty in recognizing as a form of realism at all.
T he alternative and more usual way of spelling out Q uine’s
thought has the attract ion o f starting with something much more
robust and obviously ‘realistic’ — namely by taking Q uine to be
som ething like a m irroring realist. But T shall suggest that this
(together with its related reading o f the argum ent for indeterm i­
nacy of translation) is untenable. If we pursue it we shall find
ourselves attributing to Q uine some fairly obvious muddles and
incohercncics. M oreover it is inconsistent with much that he him­
self explicitly says about the notions ‘tru th ’, ‘reality’, ‘empirically
equivalent’ and the like.
I shall first try to make this claim plausible by examining the
m atter of ‘ontological relativity’. This will provide an example of
the kind of difficulties we get Q uine into if we take his realism too
naively. I t will also introduce the im portant m atter of his treatm ent
of truth. In the next section we shall move on to further considera­
tion of his realism and its relation to indeterm inacy of translation,
and his naturalism .
M irroring realism has as a n atu ral corollary a correspondence
theory o f truth and a belief in the obtaining o f referential relations
between names and objects and between predicates and the classes
 Quine’s Naturalized Empiricism 71
of things of which they arc true. Q uine’s professions of realism,
together witli his well-known approval o f the cxtensional semantic
idioms, may combine to suggest that he is sympathetic to these
views.20
Suppose th at this were so, w hat sense could we then make of his
doctrine of ‘ontological relativity’, which he states as follows: ‘It
makes no sense to say w hat the objects of a theory arc, beyond
saying how to interpret or reinterpret that theory in another.’21
T he claim o f ontological relativity, or ‘inscrutability of reference’
as Q uine sometimes calls it, is a close relative of the indeterminacy
o f translation. In Word and Object it is not sharply distinguished
from it and is presented as the m ain line of support for indetermi­
nacy of translation. It is only with ‘O n the Reasons for Indeterm i­
nacy of T ranslation’ th at we iind the shift to emphasis on the
central part played by underdctcrm ination of theory by data. It is
this later line o f thought which we have discussed up to now. Rut in
‘O n the Reasons for Indeterm inacy o f Translation’ Quine still
assigns an im portant role to the inscrutability of reference. It can
provide support for indeterm inacy via ‘pressure from below’ —as
opposed to the ‘argum ent from above’ o f underdctcrmination.
T he idea of inscrutability of reference is this: if I am translating a
foreign language I will start by finding sentences which, as wholes,
are asserted or denied in conditions the same as those in which
certain sentences of my own arc asserted or denied. But I cannot
make substantial progress, especially in translating highly theoreti­
cal sentences, unless 1 move beyond this to detect sub-sentential
units in the foreign sentence which I can pair with my own words. I
m ust, in Q uine’s terms, adopt some ‘analytical hypotheses’.
Q uine’s claim is that this can be done in a variety of competing
ways. T o take the familiar example, I could equate their term
‘gavagai’ with my term ‘rab b it’ and some other term of theirs with
our ‘is identical w ith’. This might give me the right assent and
dissent conditions for some utterance containing both terms of
theirs. But, says Q uine, if I paired ‘gavagai’ with ‘undctachcd
rabbit p a rt’ I could adjust the translation of the other term -
taking it as equivalent to ‘is p art of the same animal as’ - thus
preserving the overall assent and dissent conditions of the
sentence.22
72 Quine’s Naturalized Empiricism
This is not itself a very exciting form of alternative translation.
Some one who says ‘This rabbit is identical to th a t’ and someone
who says ‘This rabbit p a rt belongs to the sam e animal as th a t’
have, in some sense, ‘said the sam e thing’. And if indeterminacy
went no further than this we would hardly be worried. Q uine’s
claim however is that in other cases o f less observational sentences
the two elements of the original analytical hypotheses will cease to
cancel each other out and we shall be led to offer two competing
translations for the one native sentence which in no plausible sense
‘say the same thing’. In these cases moreover, because the sen­
tences are highly theoretical, we shall find that we have no evidence
to support one translation rather than another. T his is how inscru­
tability of reference is supposed to support indeterm inacy of trans­
lation.
Wc need not concern ourselves with Q uine’s argum ents for the
existence of different analytical hypotheses or why they should be
supposed to cease to cancel each other out. Let us look simply at
how Q uine puts w hat he claims. H e says that it makes no sense to
press the question ‘W hat does “ gavagai” refer to, rabbits or unde­
tached rabbit parts?’ There is no fact of the m atter about what it
refers to. O n the other hand it does make sense to enquire w hat it
refers to relative to some theory, i.e. relative to some translation
manual for linking terms in different languages. He explicitly likens
this to Leibniz’ replacem ent of the idea o f absolute space with that
of relative space.23
H artry Field launches an interesting attack on this position.24 He
claims that the notion of ‘reference relativized to a translation
m anual’ makes no sense. C onsider ‘“ G avagai” refers to rabbits
relative to manual M l and to undetached rabbit parts relative to
m anual M 2.’ T he remarks seems to set up some genuine link
between the word ‘gavagai’ and rabbits - b u t only relative to a
manual. But what will the m anual do? It will only tell of a relation
between the word ‘gavagai’ and the word ‘rab b it’. A relation be­
tween ‘gavagai’ and actual rabbits is only thereby set up if we
assume the existence of an unrelativized relation between our word
‘rabbit’ and rabbits.
Can it really be that Q uine has fallen into this m uddle of trying
to set up a genuine, if relativized, reference relation, oblivious of his
 Quine's Naturalized Empiricism 73
claim th at the same inscrutability infects our own utterances and
hence oblivious of the fact (according to him) that my lingking
‘gavagai’ and ‘rabbit’ in a m anual cannot secure any connection
between ‘gavagai’ and rabbits that it does not equally secure
between ‘gavagai’ and undetached rabbit parts? Field’s view is that
he has and that he needs to be rescued from it by an ingenious
doctrine of partial denotation, which Field explores in the rest of his
article.
lin t we can get a lead to another reading which has Quine
looking less m uddled (if more bizarre) by considering something
Field himself says in a footnote.
W hat we do in translation manuals is set up relations of co­
extensiveness or co-denotationality. But Quine thinks that we are
not simply limited to locutions like ‘ "G avagai” is co-cxtcnsive with
“ rabbit” ’, because we can say also that ‘rab b it’ occurs in a ‘back­
ground language’ which we acquiesce in ‘taking at face value’.'25
But w hat is this ‘taking at face value’? It may look simply like
sweeping Field’s problem under the carpel. But Q uine might have
more to say in defence of the idea, by spelling out the analogy with
Leibniz’ view of spatial location talk. O n that view wc cannot
sensibly ask for the absolute position of some object x, bu t we can
say what spatial relation x has to some other object y; and we can
avoid an infinite regress in fixing position by pointing to y. This is a
move which, according to Q uine, removes need lor further discus­
sion, ju s t as ‘acquiescing in our m other tongue’ is the move which
ends discussion o f reference. But, says Field, ‘there is a crucial
disanalogy here: on Leibniz’ theory we can understand relativized
claims about the relation o f physical objects to places only because
places are understood as constituted by the relations o f physical objects;
whereas no one holds th at physical objects are constituted by the
relations of words.’20
But is it so clear that Q uine rejects the idea that ‘physical objects
are constituted by the relations of words’? O f course in the natural
first reading of ‘constituted’ it is absurd. Physical objects are
constituted by atoms, fields o f force or w hat have you, as specified
in physical theory. Suppose however that one were an instrum en­
talist?
O n the relational theory o f space one denies the existence of
74 Quine’s Naturalized. Empiricism
absolute space but adm its the reality o f objects and relations
between them; talk o f ‘place* is ju s t a convenient abstraction from,
or shorthand for, talk o f objects and the relations between them.
O n an instrum entalist construal o f theoretical sentences, one denies
the existence of extra-sentential entities but adm its the reality of the
sentences and the relations (of shape, role etc.) discernible between
them. T alk of ‘objects’ is thus for the instrum entalist merely a
convenient way of rem arking on certain high level features (of
organization, structure etc.) of those patterns of useful sentences.
T hus the doctrine that Field thinks no one, even Q uine, could be so
crazy as to hold, is ju st w hat an instrum entalist actually thinks.27
It was the introduction of the idea o f a real relation o f reference
which precipitated Field’s Q uine into the m uddle. But it is clear, if
we look further, that Q uine’s approval o f the extcnsional sem antic
idioms is not to be understood in this way. For him ‘true’, ‘refers’
etc. are mere devices for disquotation, useful tools for ‘sem antic
ascent’.aB Let us put this view at its simplest. W hen I u tter ‘T he
moon is round’ in the normal run of things, I use that sentence to
do w hatever is its proper jo b in language. I may also form a name
of the sentence and lienee a device for talking about it by putting
quotation marks around it. So insulated it docs not perform its
norm al task. But if I now add ‘is true’ I have used a linguistic
device the central purpose of which is to undo the effect o f the
quotation marks — i.e. to produce a more complicated looking
syntactic sequence which has the same potential role in discourse
as the original, quotation free, sentence. Similarly the phrases ‘the
referent o f “ . . .” ’ ami ‘is in the extension o f “ . . .” ’ are devices for
undoing the effects of quotation m arks on singular term s and
predicates.
O n this approach, w hat should we say of contexts (as for exam­
ple in translation m anuals) where we say such things as “ ‘Snow is
w hite” is true ill" snow is w hite’? Here we do not simply disquotc
but seem to use the devices to speak o f a link between a quoted item
and some thing or condition in the world. T he disquotationalist has
an account of this too. He says that we give a hearer some grasp of
how to use a foreign, quoted, linguistic item by exhibiting to him a
sentence o f our own which he already knows how to use and which
has the sam e use for us as the foreign sentence for the foreigner.
 Quine’s Naturalized Empiricism 75
(Rem em ber Q uine’s comparison of acquiescing in our mother
tongue and pointing.) O n this approach it is not of any particular
significance th at the clause on the right-hand side of the ‘iff should
occur unquoted. Its unquotcdncss signals merely that the speaker
thinks that he is now using a clause already understood by both
him self and the hearer he is trying to help - i.e. that he has reached
a background language.
A proponent of this view would reject the idea that unquotedncss
is im portant on the right-hand side because in such remarks we arc
‘doing sem antics’ as opposed to merely reporting synonymy rela­
tions. Recall again w hat we rem arked in chapter 3, that Quine
seems quite unmoved by those who have attacked his translational
approach to the study o f m eaning on the grounds that it confuses
these things.

4.3 Q IJIN K ’S V E R SIO N O F REALISM

Let us return to the question o f w hat we are to make of Q uine’s

insistence that lie is not a sceptic.29 I shall start by raising another
difficulty, sim ilar to the one Field discusses, which appears in
Q uine’s position if we suppose th at Q uine’s realism is of any
substantial, i.e. non-instrum entalist, variety (whether it is m irror­
ing, quictist or whatever). O n this view we take it that Q uine’s
‘conceptual apparatus’ talk about science is not instrum entalist but
is rather ju s t an expression of the familiar idea that in science we
make hypotheses about items we cannot perceive and that one test
we apply to such hypotheses is w hether they enable us to make
accurate predictions about w hat we can observe.
How, in this framework, is the argum ent from undcrdetcrmina-
tion o f theory by d ata to indeterm inacy of translation supposed to
work? T he instrum entalist/constitutive considerations I have been
exploring arc out. Instead it seems probable that we will have to
invoke a line of thought relying on a substantive version of physi-
calism, as we sketched in section 3.1. Let us recapitulate that
argum ent.
W e have two theories, IT and T 2, which are incompatible but
each consistent with all the observation sentences. We accept T L
76 Quine’s Naturalized Empiricism
W e imagine all the "physical facts to be fixed, i.e. all the facts about
positions of fundam ental particles, movements o f bodies, utter­
ances of noises, etc. We will doubtless use the terms supplied us in
T1 to characterize these facts and the characterizations will not
include any mention of meanings. Next we recognize that it is
consistent with the statem ent o f all these facts to say o f some alien
scientist both that his theoretical utterances are to be translated as
expressing T l and also th at they are to be translated as expressing
T2. So given the data available to us (which is all physically
specifiable) about his utterances and w hat stim ulate them (and
given all other physical facts as well for good measure) we still lind
th at hypotheses about theoretical meanings are underdeterm ined.
But our physicalism insists that there are no further facts which
could make one of the translation hypotheses true and other false,
for there can be no factual difference w ithout a physical difference.
'Therefore there is no fact o f the m atter about which theory the
foreigner accepts and no fact about what his theoretical utterances
mean.
The tension in this story emerges as follows. O n the one hand we
refuse to identify the ‘truth* of a theory with its coping adequately
with all the data. 'This is the identification which our non-
instrum entalist realism rules out. W e say th at two theories may
cope equally well with all the d ata but that, if they clash, one o f
them will in addition have some further virtue which the other
lacks, namely that of describing how things really are. In saying
this we clearly allow for the possibility that there should be two
syntactically distinguishable systems each of which has the syntac­
tic property of generating all the true observation sentences, but
one o f which has, whilst the other lacks, an im portant semantic
property - namely that of saying something true. O n the other
hand when we consider the foreign scientist’s utterances, we must,
on Q uine’s view, deny that there can be any fact o f the m atter
about whether his set of sentences has or does not have this
sem antic property. It can be translated equally correctly by that set
of our sentences which has it and by the set which lacks it.
But what could justify the discrim ination implicit in this between
w hat we are prepared to say o f our sets o f sentences and w hat we
say of his? It seems unm otivated and Q uine is indeed anxious to
 Quine’s Naturalized. Empiricism 77
emphasize that the indeterm inacy o f translation can be applied in
our own case. T hus we may ourselves accept T1 and reject T 2, in
the sense of choosing one set o f theoretical sentences to u tter rather
than the other. But we m ust say o f ourselves th at in uttering the
sentences of T1 we can be translated either homophonically as
holding T1 or non-homophonically as holding T2.
And now w hat are we to make of the ‘realism ’ we earlier
professed? T h at realism offered us a view on which an im portant
difference existed between T l and T 2, while the current view
makes a nonsense of that claim. If my uttering the sentences o f T l
is acceptable, whilst it is also the case that in uttering them I can
defensibly be taken to have ‘m eant’ T2, then I might as well have
uttered T2 instead; and if I had it is hard to see how it would not
have been other than equally defensible.
This seems such an unpalatable and muddled position that we
should, if at all possible, avoid attributing it to Quine. And there is
am ple textual evidence that it is not his view. T he first point to note
is his theory of truth, which we have already discussed. Acceptance
o f a disquotational theory is not a position with clear metaphysical
com m itm ents of its own. It is com patible with mirroring realism,
with ‘quictist’ realism (of the sort mentioned in chapter 1) and also
with instrum entalism . But as deployed by Q uine it is clearly
designed to oiler us a view o f ‘true’ which frees us from the
tem ptation to suppose that we m ust hypothesize some correspon­
dence relation, of the type implicit in m irroring realism. And Quine
is hostile to the idea of such a relation. He speaks of the correspon­
dence theory as ‘a hollow mockery’10 and points out with relish the
consequences of his views about m eaning for the notion o f ‘state of
affairs’ which would be required at the world end of such a
relation.31 It is, on this evidence, surely possible to suppose th at he
means to deploy the disquotational view in such a way as to make
instrum entalism seem more palatable than it otherwise would;
instrum entalism docs not debar us from talking earnestly of truth.
T h e second relevant point from Q uine’s writings which bears on
this m atter is his handling o f the hypothetical situations where we
are confronted with two empirically adequate theories. His moves
here are subtle. In the version of ‘Em pirical C ontent’, originally
published in 1981 in Theories and Things, Q uine wrote ‘Still, let us
78 Quine’s Naturalized Empiricism
suppose that the two formulations arc in fact empirically equivalent
though not known to be; and let us suppose further th at all of the
implied observation calcgoricals arc in fact true, although again
not known to be. N othing more, surely, can be required for the
truth of cither theory formulation. Are they both truc? I say yes.’3'2
But Roger F. Gibson pointed out a disagreem ent between this and
a passage in Theories and Things where Q uine also talks of alternative
empirically equivalent theories.33 (H e is here considering theories
th at may be derived by the use of proxy functions and of considera­
tions having to do with ontological relativity. But the issues which
concern us arc the same.) Here Q uine says ‘But it is a confusion to
suppose th at wc can stand aloof and recognize all the alternative
ontologies as true in their several ways, all the envisaged worlds as
real, i t is a confusion of truth with evidential support. T ru th is
im m anent and there is no higher. Wc m ust speak from w ithin a
theory, albeit any of various.’34
Q uine’s response to G ibson’s rem arks was to modify the earlier
cited passage (which as it stands is clearly congenial to the in­
strum entalist interpretation I wish to defend) to bring it more in
line with the other. But let us rem ark that there surely must be
some quasi-instrum entalist strand which led Q uine to the first
formulation; someone who was a straightforw ard m irroring realist
would never have dream t of w riting it. Also Q uine’s inclination
when faced with the idea of incom patible theories is to try to
minimize the need for choice. He points to the possibility of
reinterpreting terms in one theory so as to present it as a notalional
variant of the other or to the possibility o f simply conjoining them.
(If he were able to say that different but empirically adequate
theories arc impossible lie could avoid the m uddle wc examined at
the beginning of the section - but equally he would have lost his
argum ent for the indeterminacy o f translation.) And in fact Quine
does recognize that there may, hypothetically, be occasions where
choice cannot be avoided: O n this, his linal and subtle position
hardly am ounts to a robust version of realism. R ather he insists
truth is immanent, we must speak from within a theory?* And it is this
thought which provide the crucial clue to the nature of his realism.
I f we follow this through wc can credit him with a coherent, if
strange, vision.
 Quine’s Naturalized Empiricism 79
T o see how this may be so, let us return to some remarks in Word
and Object.

Have we now so far lowered our sights as to settle for a relativis-

tic doctrine of truth - rating the statem ents of each theory as
true for th at theory, and brooking no higher criticism? Not so.
T he saving consideration is th a t we continue to take seriously
our own particular aggregate science, our own particular world
theory or loose fabric of quasi-thcorics, whatever it may be.
U nlike Descartes, we own and use our beliefs of the moment
even in the midst o f philosophising, until by what is vaguely
called scientific m ethod we change them here and there for the
better. W ithin our own total evolving doctrine, we can judge
truth as earnestly and absolutely as can be; subject to correction,
but that goes w ithout saying.36

T h e emphasis here comes back to one aspect of Q uine’s natural­

ism, namely the idea th at we cannot do without actually accepting
some theory of the world. (It is worth remarking on the slightly
unusual phraseology: we are said to ‘use’ our beliefs). There is no
entirely non-committal standpoint, no possibility of Cartesian jetti­
soning of all substantive and risky commitments, no finding of
some certain starting points for a first philosophy.
O ne might enquire a t this point why talk about meaning was not
finally vindicated - as far as the possibility of realistic construal is
concerned - by parallel moves. Do I not always and unavoidably
‘take seriously’ some view about w hat utterances, my own or
others’, mean? Q uine’s answ er, at least for the case of other people,
is ‘no’. He says ‘T he param eters of truth stay conveniently fixed
most of the time.’ (We may take him here to be talking of truth in
connection with sentences o f physical theory and ‘the param eters
staying fixed’ is our continuing to take seriously some background
picture.) But he continues ‘Not so the analytical hypotheses that
constitute the param eter of translation. We are always ready to
wonder about the m eaning o f a foreigner’s rem ark without refer­
ence to any one set of analytical hypotheses, indeed even in the
absence of any.*37 Q uine’s view then is that I can carry on theoriz­
ing and talking about physics w ithout accepting any one theory
about w hat other people mean.
80 (¿nine’s Naturalized Empiricism
But again what holds in my relations with the foreigner holds in
my relations with myself. So the upshot is that 1 can operate
w ithout ‘taking seriously’ (in the relevant se»se)any assum ptions
about what J mean by my own remarks. And now the shape of the
whole Quincan construction can be seen more clearly. T he claim is
that a bit of my world picture that I do and cannot but take
seriously - namely the bit th at starts with talk o f material objects
and proceeds to theories about them - will, when reflectively
examined, lead me to see th at the other bit of the picture - namely
the bit about meanings - is one that I do not and should not take
seriously.
But now wc w ant to know w hat it is to ‘take an utterance
seriously’. And the price to be paid for avoiding the obvious
muddles, the ones that we saw in the discussion o f Field earlier in
this section, becomes apparent at this point. T he price is a radical
impoverishment o f the picture we have of ourselves, or rath er an
inability to endorse fully certain apparently central elements o f th at
picture. This occurs because Q uine cannot allow anything but the
thinnest account of what ‘taking seriously’ am ounts to.
T he natural account of what it is to take something seriously
calls upon those notions which Q uine wishes to see as secondary.
For example we might well say that to take an utterance seriously is
to believe what it says, or at least to suppose th at there might be good
reason to believe th at it says som ething which m ight be true. But all
the italicized phrases (except ‘true’) are ones which Q uine tells us
do not occur in descriptions o f w hat really goes on. (‘Reason’,
although we have not discussed it explicitly, clearly goes along with
meaning, because it is in virtue o f their meanings th a t sentences or
beliefs enter into relations of epistemological relevance to each
other.) And the account he gives of ‘true’ merely points us back to
the idea o f ‘taking seriously’, since to say something is true is, on his
view, merely a disquotational m anifestation of my taking it se­
riously.
VVe can return here to a reflection on Q uine’s realism and w hat
differentiates it from instrum entalism . Earlier I offered a formula­
tion of the anti-instrum entalist thought in these terms: empirically
equivalent theories may differ in th a t one has a virtue th at the other
lacks. Q uine’s remarks suggest that he wants to respect this
 Qitine’s Naturalized Empiricism 81
thought. But it turns out th at the only fundam ental difference he
allows for, the difference which takes the whole burden of distin­
guishing the ‘true’ from the ‘false’ theory, is the fact that I actually
use (‘take seriously’) one and not the other. Q uine’s non-
instrum entalism , on my reading, consists only in this: he thinks
that we are always actually employing one o f o ur instruments and
so, in some respects, cannot stand back from it; but we can stand
back from it enough (he supposes) to see its instrum ental nature
and hence to derive, sceptical conclusions about meaning.
We may indeed adm it that being used or not used is a difference
between the theories; and it is dilference which, given disquotation
and so forth, allows me to make realist-sounding remarks. But it
can hardly be said that 1 have revealed a virtue which one theory
has and the other lacks or that I have made sense of my adopting one
rather than the other in terms of my taking it to have that virtue. I
am returned simply to the blank fact that this is the theory that I da
work with.
W e are, of course, allowed by Q uine to say those intcnsional
things about ourselves —about reasons, beliefs, meaning and truth
- in which we express our convictions that our utterances have
virtues. He is not in favour o f outlaw ing such idioms. But in saying
these things we do not, according to him, capture anything that is
really going on. W hat is really occurring, he claims, what I can say
‘seriously’, is that we are caused to make certain noises in ways
which then cause us to make other noises and so forth.
But even this remark (‘W hat is really occurring is . . .’) comes to
have an em pty ring when we realize that it is itself ju st a noise we
are caused to make. Even for it, I am not allowed to explain what
iny taking it seriously am ounts to in any way which goes beyond
the fact that I recognize that I am disposed to utter it. We thus end
up, it seems, not only not taking seriously our remarks about
m eaning but, in some sense, unable to take seriously anything we
do.
This is only a brief outline of the reflections that gazing at the
Q uinean picture leads us into. We shall return to the topic at the
start o f chapter 6 and explore the m atter further. It is, however,
already apparent that Q uine’s vision is exceedingly strange and
uncongenial. Should we, nevertheless, endorse it?
82 Quine’s Naturalized Empiricism,
4.4 S H O U L D E M P IR IC IS M BE N A TU R A LIZED ?

T here arc Iwo questions I would like to address in this section.

First, assum ing that the sketch I have given of Q uine’s position is
correct, should we adopt that position? And second, can 1 substan­
tiate my claim that the lure of m irroring realism (in its empiricist
form) has a great deal to do with Q uine’s view? T he second
question m ay seem especially pressing, since the Q uine who has
em erged could hardly he said to he a central ease o f a m irroring
realist. So we will start with it.
M irroring realism is, I contend, at the heart of the m atter in this
sense: th a t it is a commitment to the idea th a t i f m eaning is to be
respectable at all then it will fit the m irroring realist model, which
drives the whole construction. How docs this come about?
I would like to rem ark (irst, and rath er baldly, that it is not,
straightforw ardly, ‘a fact o f science itself that science is a concep­
tual bridge o f our own making linking sensory stim ulation to
sensory stim ulation’. Such a view o f science .is not a mere com­
m onplace of scientifically educated commonscnsc. T he sciences
Q uine cites, physiology, theory o f n atu ral selection and so forth, tell
us nothing directly about meaning, the nature of science or any
such things. They do indeed tell us a great many fascinating things
about light waves, retinas, nerves, neurones, genes, phenotypes,
reproduction, survival etc. But we cannot superimpose upon the
picture built up in these terms some story about meaning, beliefs,
reasons, theories and so forth w ithout making further assumptions
about those notions, assum ptions which arc not supplied by physi­
ology or natural selection theory. Q uine may sincerely deny com­
m itm ent to a lirst philosophy — but it would be impossible to get
insight into meaning on the basis o f physiology etc. w ithout some
ideas about meaning got from elsewhere, because physiology etc.,
in their central and uncontroversial parts, ju st do not deal with
such m atters.
A nd it is indeed evident that Q uine does make some assumptions
about meaning, about what a clear ease o f one item m eaning
som ething would be like. His paradigm is that of a situation in
which items m eant (stimuli) produce w ithout significant mediation
other things (responses), in such a way th a t a system atic and
 Quine’s Naturalized Empiricism 83
absolutely reliable correlation between their natures exists. And
where has he got the paradigm from? H e has got it from mirroring
realism in its em piricist form.
I offer this as a historical claim and a brief look at Q uine’s
intellectual developm ent, and a t w hat he himself says about it, will
substantiate it. T h e influence of C arnap upon Quine is very great
and openly acknowledged. T h e m ajor intellectual enthusiasms and
insights of the young Q uine, in the areas of epistcmology, meta­
physics and theory of meaning, sprang from his contact with the
Vienna Circle, and particularly with C arnap.38 Consideration of
C arnap’s system at this time reveals it as providing an interesting
precursor to Q uine’s.39 C arnap held that wc have a body of
immediate experiences, described 111 ‘protocol sentences’, which wc
need to systematize. In order to do so we choose pragmatically
among various different linguistic frameworks. Each framework
involves its own particular set of terms, the meaning of which is
fixed by analytic principles, linking sentences using the terms to
protocol sentences and to sentences containing other terms. So,
once a framework is chosen —the so-called ‘external question’ has
been decided - then questions posed in the chosen vocabulary -
the ‘internal questions’ —will have definite answers dictated by the
course of experience. Q uine’s response to this, which wc can sec
developed in a num ber of his early papers (most notably ‘Two
Dogmas of Em piricism ’)'*0 is to deny the tcnability of a sharp
distinction between internal and external questions and conse­
quently the tcnability of the analytic/synthetic distinction. To deny
this distinction is to pass w hat Q uine calls the fourth milestone of
empiricism (the earlier ones being the shift of attention, when
considering vehicles o f m eaning, from ideas to words, the shift from
words to sentences and the holistic shift from sentences to theories).41
At this point, having passed the fourth milestone, and if wc still
accept the idea o f protocol sentences reporting on immediate
experience, we have reached the instrum entalist position as elabor­
ated in chapter 3.
T he next developm ent, required according to Q uine to move on
past the fifth milestone, is to repudiate first philosophy and espouse
naturalism . And now we can tu rn back to the first question posed
in this section and ask w hether we should adopt the final Quinean
84 Quine’s Naturalized Empiricism
position. Two sorts of consideration are relevant. O ne concerns the
acceptability of the m irroring realist paradigm and the other con­
cerns Q uine’s correctness in supposing th at the scientific facts, even
if they do not supply those ideas about m eaning, do indeed dovetail
with them and give them support.
We cannot deal with the first sort o f consideration properly until
we come to discuss W ittgenstein’s worries about mirroring realism
and his rejection of the passive, im printing, conception of meaning.
Here I shall only put down a m arker to the effect th at there is
something seriously wrong with the paradigm and hence that we
should not desire to find naturalized structures into which we can
read it.
But considerations of the second kind alone suggest serious
worries about the internal cohesion o f the Q uinean system. Q uine’s
claim is that we find reproduced w ithin physiology etc. the struc­
tures that a sophisticated sort o f sense datum empiricism claimed
to find embodied in instrospectible experience and our deliberated
response to it. B ut to read the scientific facts this way involves a
good many procrustean m anoeuvres. Someone who was not al­
ready thoroughly committed to the em piricist picture m ight well be
more struck by the difficulties in the program m e than Q uine seems
to be.
We have already seen that he has to skate over the difficulties o f
the fact that observationality is a m atter o f degree. Further recent
work makes it look a murky and complex m atter w hether many or
indeed any utterances tire related to patterns of stim ulation in the
way Quine imagines.42 And it is even less clear that utterances
which fulfil this condition (if there are any) form a sharply distin­
guishable observational base for scientific theorizing. Related wor­
ries arise from the way Q uine mingles the personal, intentional
level of description with the subpersonal and causal. Scientific
theories, he says, are o f our own devising. T his is unexceptionable.
And equally it is unexceptionable th a t neural changes are caused in
us by the triggering of sensory receptors. But the attem pt to
superimpose the one scheme o f description upon the other produces
various obvious tensions. O ne has to do w ith the content which, at
the personal level, we would naturally take our evidential state­
ments to have. We suppose them to be about tables,.trees and so
 Quine’s Naturalized Empiricism 85
forlh and are not aw are o f any ‘holophrastic’ level at which another
sort of claim is m ade. But Q uine tells us that it is these holophrastic
claims upon which we base our theorizing. A nother difficulty
concerns the idea o f w hether judgem ent and choice are involved
and, if so, how. At the personal level we suppose th at it may
sometimes be up to me which hypothesis I adopt, since the evi­
dence does not point unam biguously one way rath er than another.
Indeed it is part of Q uine’s underdetennination thesis that this
should be so. But at the subpersonal level there will be no underde­
termination; the fact that different neural structures could underlie
sim ilar external behaviour does nothing to show that there is some
causal indeterm inacy about which structure I actually coruc to
contain.
T he point o f the above sketchy rem arks is not to show that
Q uine’s supcrim posilion is obviously untenable. It is only to make
clear that it is a supcrim posilion and that it is fraught with
difficulties which Q uine docs not resolve. T h e conclusion is that his
view is far from commonsensical or m andatory and hence that we
may justifiably regard the extreme bizarreness of the position it
leads us to as grounds for suspicion o f the premises.
 5
The Mona Lisa Mosaic

5.1 SE M A N T IC H O L ISM

In this chapter I shall set out another argum ent concluding that
there are no facts about m eaning and th at ascriptions of m eaning
arc subject to indeterm inacy. It may be th at I have been obtuse
and unjust in my earlier interpretation of Q uine and th at the (it
seems to me considerably better and more interesting) argum ent
which I am now about to sketch is the one he is really putting
forward- A nother possibility is that the argum ent of this chapter
represents one strand in Q uine which he never gets properly
disentangled but which has been extracted and presented in pu rer
form by a sym pathetic developer of Q uine’s ideas, namely
D avidson.1
T h e new line of thought depends crucially upon ‘holistic’ con­
sideration - but it invokes holism interpreted in a significantly
different sense from th at presented in chapter 3. There we started
from the idea that the property we were interested in - m eaning -
•has a clear prim ary sense in which it belongs only to certain wholes,
namely observation sentences and entire theories. These things,
said our earlier theory, perform some function, namely th a t of
labelling an individual stim ulus or encapsulating a whole course of
experience; and we can see th a t they do so from contem plating the
connection which they as wholes have with the things m eant, quite
ap art from any consideration o f their internal structure or o f the
m eaning of other items; sm aller units, words or individual theoret­
ical sentences, may be said to ‘have so and so m eaning’, but only in
a derivative sense (as the parts o f a sausage machine may be said to
 The Mona Lisa Mosaic 87
‘make sausages’); and this sense m ust be explained in terms of an
idcntiliablc contribution m ade to the functioning of the whole.
T he kind of holism we shall now consider, however, starts from a
different idea, namely th a t possession of meanings is quite properly,
and in no secondary sense, ascribed to small units such as words or
sentences of a theory. But, it adds, possession of this meaning
dem ands the presence of other sim ilar m eaning-bearing items of a
suitable sort, which together with the individual item considered,
form some assemblage; and, it says further, the mcaningfulncss also
requires the absence from th at assemblage of certain items of an
unsuitable sort. I f these conditions arc fulfilled then there will, it
seems likely, be a sense in which the assemblage as a whole has
meaning. And, given w hat we have ju s t said, its existence and
mcaningfulncss arc required for the m eaning of the elements. But
the dependence is, on this view, m utual. T he mcaningfulncss of the
whole is not prior to th at of the parts, in any significant way,
conceptually, causally or metaphysically. We cannot make sense of
the idea that the whole should have the m eaning it docs without the
individual parts bearing their appropriate meanings. So the slogan
‘the fundam ental unit o f m eaning is the whole’ would be highly
misleading as a sum m ary of this sort of holism, although accurate
enough for the earlier variety. T he central idea now is rather that
m eaning necessarily involves complexity.
T here are, plausibly, at least three levels to this complexity -
that o f word, sentence and (for w ant o f a better word) discourse.
But I shall not, in w hat follows, say much about words, concen­
trating rath er on some features of the required connections between
sentences and other sentences in discourse. (It may he that the
existence of words is a corollary o f the existence of further features
of the sentential patterns needed for mcaningfulness, but I shall not
pursue this thought).
T o help get a grip on the kind of meaning holism now under
consideration it may help to look at a ease where a very similar sort
of reciprocal dependence is found, namely pictures and their ele­
ments. I have in mind here th a t kind of pictorial representation
which may be found both in artefacts and in such natural objects as
clouds or stains on the wall.2 W e shall be concerned, initially, only
with the sort of manifest content which can be seen in a picture in
88 The Mona Lisa Mosaic
virtue of its visual qualities, without special knowledge of its
m aker’s intentions or the identity of any model it may have had.
In a picture of a face, it may be clear th at a certain circle
represents the right eye, a line represents the nose and so forth. But
the circle standing entirely by itself could not represent the eye. It
requires the presence of (at least some) other suitable items which
can be seen as representing (at least some of) nose, m outh, other
eye, etc. The assemblage as a whole is a picture of a face and the
circle represents an eye only in the context o f the other items which
together make up that picture. So one could be tempted to say that
the whole picture was the prim ary vehicle of representation. But
this would be to ignore the fact that this whole only represents a
face because the individual parts have their representational roles.
1 do not wish to say that anything representing a face m ust have
such internal complexity. A simple item, such a white blob, may
indeed do so and also lack significant relevant structure within
itself (as the simple circle may represent an eye without having
internal complexity connected with pupil, eyelid etc.). But for the
blob to represent a face in itself requires placement in a suitable
setting, say a picture of an immense crowd.
So the general conclusion is that one cannot come to grasp the
picturing role o f some item, w hether simple of complex, w ithout
attending to the picturing role of other items, w hether simples
juxtaposed to the first, if it is itself simple, or elements of it, if it is
complex. W here there is pictorial representation there will always
be complexity and hence possibility of talking both o f w hat the
whole picture represents and also of w hat the elements represent.
I have talked so far of the required presence of other items
needed to make something simple, like a circle, represent an eye.
But equally we can see that the absence o f certain disruptive
influences is required. T he very same lines and circles that rep­
resent a face might, occurring in the context o f some further lines
and circles, clearly not be a picture o f a face or not suggest a face at
all, but be instead a straightforw ard picture o f a salad on a plate,
the wheels of a m achine, or almost anything else one’s ingenuity
might devise.
When I speak of sem antic holism in w hat follows it is a thesis
suggested by this example that I shall mean. M ore specifically it is
 The Mona Lisa Mosaic 89
the claim that an individual sentence is both correctly said to be the
unit of meaning (in some sense) and is also such that it can have the
meaning it has for a person only if it occurs in a certain setting,
which both contains other sentences used by that person to which
suitable related meanings can be assigned and also lacks sentences
inappropriate to that m eaning.
We may leave the notions of what is ‘suitable’, ‘inappropriate’,
etc. in a vague state for the purposes o f the overall argum entative
strategy, although I shall have a little more to say about them in
section 5.3. Hut it is w orth noting that an epistemological flavour is
discernible in our sem antic rem arks at this point. W hat is suitable
and what inappropriate has to do with what is involved in a
person’s being in a psychological state with a given content, for
example, with how fragm ented, partial, contradictory and so forth
thoughts can be while still being thoughts. T he questions will be
ones like ‘Could someone think this while also thinking that, or
while failing to think such and such?’
Epistemology, narrowly construed, tells us about w hat people
ought to judge or not to judge in certain circumstances, about how
confident they should be and about whether their thoughts consti­
tute-knowledge. It is thus norm ative and might not seem to have
much to do with w hat people actually do think or might possibly
think. But the kind of outlook being discussed in this chapter makes
the assum ption th at ‘thinks’ is, in Rylcan terms, a success or
achievement word.3 O ne docs not think at all unless one, to some
extent, succeeds in doing w hat, qua thinker, one ought to do -
namely, satisfying the norm s which epistemology (among other
enquiries) aims to lay out and clarify. T he claim here is extremely
weak. It is only that completely unsuccessful and chaotic thought,
thought in which no shred o f tru th or rational connectedness is
discernible, is an incoherent notion. It does not deny the possibility
o f extremely bizarre beliefs, great degrees of muddle, contradiction
and so forth.
W h at of the notion of a ‘setting’ which has been invoked? In the
case of a pictorial representation there will (usually) be some clear
boundary to the area w ithin which relevant items could occur, a
boundary such as the edge o f a sheet of paper or a frame. But in the
case of a person it is difficult to see how one could find any such
90 The Mona Lisa Mosaic
boundary at a place short o f the limits of the whole assemblage of
w hat the person says. (In as m uch as this includes responses to
things said by others, relevant items may well be found even further
afield; but we shall not pursue this complication.)
T h e earlier claims made, th at m eaning requires suitable pre­
sences and absences, and rclatcdly that thinking is a success notion,
have considerable plausibility. B ut this last move may seem more
startling. O n reflection however it will become less outrageous. T he
first impulse to deny it may come from the (correct) perception that
the requirem ent lor the presence o f relevant items may well be
satisfied in a small and temporally limited set of utterances. Yet it is
not the presence condition but the absence condition which re­
quires us to look wider. And w hat principled reason could one give
for ruling out any stretch of w hat a person says from consideration?
All stretches ex hypolhesi include things said and done by the very
same person and hence subject to the constraints (however m ini­
mal) which arc required for them to belong to the unified psycho­
logical life of the one subject. In as much as we weaken any such
rationality and coherence conditions - for example, allowing the
thoughts to divide into two groups the union of which violates the
constraints - we have started to think in terms of sequential or
parallel occupation of the one body by two persons. (It may well,
for example, become necessary to know, when interacting with the
person, which group of thoughts is operative.)
It is worth explicitly distinguishing sem antic holism in the sense
sketched above from the thesis which Fodor discusses under that
nam e and which he summarizes as follows: it is ‘the idea that the
identity - specifically the intentional content - o f a propositional
attitude is determ ined by the totality epistcmic liaisons’.4 If one
embraces sem antic holism in Fodor’s sense then one is forced to the
conclusion lh a t two people who dilicr at all in their views about
evidential relevance cannot share any propositional attitude and
hence cannot mean the sam e by any utterances they produce.
Fodor, understandably recoiling from this outcome, suggests that
the best hope for setting up a non-holistic account of m eaning is to
pul denotation ~ determ ined on a one-by-one causal basis for
meaningful items - in the central place.
 The Mona Lisa Mosaic 91
B ut the result o f doing this, at least in the way Fodor proposes to,
is th at, as he allows, we have to contem plate the possibilities that
‘people can believe things th at are arbitrarily m ad’ and that ‘enter­
taining the thought th at three is a prim e num ber could constitute
an entire m ental life.’5 Fodor professes himself happy with these
views but they are arguably a t least as intuitively unacceptable as
the untoward consequence of Fodor’s version of holism remarked on
above, namely that people with different views cannot mean the same.
So, were Fodor’s kind o f sem antic holism and his causal-deno-
tational theory the only two options, we would be in a bad position.
Fortunately there is a t least one more possibility, namely the more
m oderate holism to be examined here. T h e crucial difference is that
our holism claims only th at for a certain meaning to be expressed,
the whole constituted by the person’s utterances must be suitable,
in terms of presences and absences. B ut we have not said that there
will be only one suitable setting in which a given meaning can
occur, so wc arc not committed to the view that any difference
between two wholes m ust make every meaning expressed in the one
dilTer from every meaning expressed in the other. And to say (as we
will later) that every statem ent in a whole collection is relevant to
(has a potential bearing on) the meaning of any other given
statem ent is not to say th at a change in the former m ust necessarily
correlate with change in the m eaning assigned to the latter.
Nevertheless there is, perhaps, a highly paradoxical implication
o f the view lurking in the wings, namely a revival of the indetermi­
nacy claims about m eaning. And it is to consideration of this that
wc now turn.
Wc have so far been talking metaphysically or constitutively
about w hat is required for an utterance to have a certain meaning.
However, there arc, it seems, implications in this for the methodol­
ogy of the enterprise of ascribing meaning to the sentences of
someone elsc’s language. T he methodology will be subject to
holistic constraints - in a sense which I shall try to lay out more
fully in the next section. But, it will then be argued, when applica­
tion of a concept is governed (only) by this sort of methodological
constraint then it is implicit in the rules for its use that each of two
incom patible meaning ascriptions may be equally and fully in
92 The Mona Lisa Mosaic
order. Hence the concept does not obey the law of non­
contradiction anti, in accordance with the connection outlined in
chapter 1, cannot be regarded as describing the real.
T he interest o f the argum ent to be explored is that it invokes only
the comparatively innocuous seeming premises we have discussed
and is not cumbered with baggage o f a controversial empiricist
character. We start oiT the argum ent allowing ourselves to be
realists about both the observable and the unobservable and un­
burdened by any views on how to spell out that contrast or whether
it is im portant. We also carry no commitments about how concepts
may be acquired or of what they can be concepts. All we have in the
way of substantive premises are the outlined views about meaning,
which might well be thought plausible and to reveal something
interesting and distinctive about the notions of semantics and
psychology as opposed to those o f physics.

5.2 H O L ISM AND IN D ET ER M IN A C Y : T H E M O SA IC

ANALOGY

In this section 1 shall try to make the argum ent from holism to
indeterminacy vivid and plausible by examining a particular case
of the picture analogy already mentioned. T he advantage is that in
the example to be discussed the essential logical features are clearly
exhibited and, moreover, the conclusion is not seriously paradoxi­
cal. The case is th a t o f constructing a mosaic. In the next section I
shall move back to the case o f language and meaning, showing how
it can plausibly be construed as an instance of the same conceptual
pattern.
Imagine that 1 am set down before a certain scene - a woman,
Lisa, sitting in front of a landscape. I am supplied with a large flat
tray and a box of little chunks o f glass, ceramic, stone etc. o f various
sizes and colours. T h e task 1 set myself is to produce in mosaic a
recognizable picture o f the scene. O ne constraint on the enterprise
is that I must use, if at all possible, all of the pieces I am supplied
wii.li; I must not throw out a large proportion in order to leave
myself with a handy set. A nother constraint is th at I m ust not
myself m anufacture pieces to fill in inconvenient gaps. I am given
 The Alona Lisa Mosaic 93
no guarantee that the pieces are peculiarly suitable for the jo b —
my task is not like that of doing a jigsaw. I do not lay down in
advance w hat size the final picture is to be, nor do I dem and that I
should depict the scene from the viewpoint I currently occupy or
th at I should use any particular style of representation. M oreover
the business is open-ended, ■since from time to time I may be
supplied with new batches o f pieces which m ust also be used.
flow would 1 set about the task? Perhaps this way: by sorting the
pieces for colour and size, noting proportions of various colours in
the actual scene and m atching as well as I can the elements o f the
real scene to the m aterials supplied. 1 may plunge in by m atching
some pieces to her hair, some to her hands and some to the distant
hills. Having pul these in place I am limited in my later choices. I
may be lucky and discover th a t it all works out; but if I have not
been careful enough I may find myself without enough left to model
her face satisfactorily, so I m ust start again.
M any kintls of developm ents and outcomes are possible. Perhaps
I complete one acceptable picture and, bored with nothing to do, I
tip the pieces out and start again, producing another equally good
picture from a dilfercnt viewpoint and in a different idiom. Perhaps
a promising picture has to be scrapped because a new batch of
pieces arrives and cannot be fitted in. But perhaps again the new lot
can be fitted in just by expanding the old arrangement. Arid so on.
In the context of this task and the various possible outcomes,
consider the questions ‘W here does this piece go? Is it a left
eyebrow piece?’
There are various ways in which these questions can be con­
strued, but on only one interpretation do they have a clear-cut
answer, namely when they are taken as asking alter a decision
already made. O n other interpretations there is no guarantee that
there will be only one answ er to where the piece goes and hence no
guarantee that one could not answ er ‘Is it a left eyebrow piece?’
w ith both ‘Yes’ and ‘No’.
O n one interpretation the questions may be taken tim dcssly -
suggesting that there is some one role th at any piece should play in
all defensible arrangem ents or suggesting that there is one final best
arrangem ent in which it can play only one role. But the assum ption
th a t all defensible arrangem ents m ust assign the same role to a
94 The Mona Lisa Mosaic
given piece is wrong. New pieces may always arrive and disrupt a
given satisfactory arrangem ent, necessitating drastic displace­
ments. And, much more im portantly, with a given set of pieces —
even the final complete set supposing we know when that has
arrived - there may be alternative ways of constructing an accept­
able picture.
It may be th a t with the final set o f pieces one arrangem ent stands
out as vastly preferable to all others - e.g. the pieces m ay form a
jigsaw. But our impression that there is only one arrangem ent may
be due to lack of imagination. T he terms of the task show that we
cannot insist a priori that there shall be one and only one best
arrangem ent of any given set of pieces.
I f I cannot insist that there is no one best role o f a particular
piece, can I at least rule out certain roles for it? Again the answer
seems to be that a priori 1 cannot, although (as with the assignment
o f a given role and for the same sort of reason) a particular given
collection, e.g. a set of jigsaw pieces, may in practice rule out a
particular move. It is even clearer that if I take a piece or small
collection of pieces in isolation then I cannot, by inspection of them
alone, dismiss certain roles as ones impossible for them to play. For
example one might suppose th at a particular bright red piece could
at least be ruled out as an eyebrow piece. But if the pieces supplied
turned out, on inspection, all to be in various shades of red, then
the eyebrow is ju st where it m ight end up.
A nother interpretation o f the questions ‘Is this a right eyebrow
piece?’ and ‘W here docs this go?’ secs them as asked of me when I am
half-way through completing the task with a given set of pieces. So
it enquires, in e ject, ‘Given th at you have already placed a good
num ber o f pieces and determ ined a certain style, colour scheme
and degree of detail for your picture, where will this piece go?’ Here
there is more likely to be determ inacy in the answer; the greater the
num ber of pieces already fitted and the clearer the style, the less
choice I have on where the rem aining pieces shall go. But even so
determ inacy cannot be guaranteed.
These indctermirxacics do not mean th a t we would have no use
for the predicate ‘left eyebrow piece’. Context will often serve well
enough to distinguish the vaxious questions being asked, or claims
being made, by its use; and by employing it conversational moves
 The Mona Lisa Mosaic 95
can be made in a less cumbersome way than would otherwise be
required. But the indcterminacies do show that anyone who
thought that this predicate operated like more straightforward
descriptive ones, such as ‘red’ or ‘square’, would be mistaken. In
particular he would be mistaken in supposing that there was some
‘fact of the m atter’ in virtue o f which lie could insist on getting a
clear cut answer to the timclessly intended questions.
W hat are the general structural features of the situation that
explain the indeterm inacy and by discovery of which we might
hope to throw light on the case o f meaning?
We have a group of objects, the mosaic pieces, to which two sets
of predicates arc applicable. O ne set is relatively unproblematic —
‘red’, ‘square’, ‘mem ber o f a large group of like-coloured pieces’
etc. Let us call them the U-prcdicatcs or (relatively) unproblem atic
predicates. These predicates are the ones about which we arc
realistic. We take it that there arc facts about whether the pieces
are correctly described by them and that investigation can put us in
a good position to make largely reliable, if not infallible, judge­
ments about what properties of this kind the pieces have. The other
sort of predicate - let us call them P-predicatcs to remind ourselves
of their problem atic status - are applied on the basis of the
U-prcdicates. There arc such things as ‘being a left eyebrow piece’
and so forth.
T he P-prcdicatcs come as a set. As was stressed in section 5.1
above, for one to be applicable at least some others from the set
m ust be applicable. N othing can be an eyebrow piece unless there
arc at least some pieces which can be taken to represent nose,
forehead, eye, etc. T h e items to which P-prcdicatcs apply must
have certain relations. But these relations arc defined at a high level
of abstraction; they concern such things as contrasts between
U -propcrties and contrasts between contrasts between properties.
T h a t some things stand in such relations docs not presuppose much
about the actual nature of the items related. Let us say that
P-prcdicates generate an abstract pattern of demands. For example, an
eyebrow piece and a forehead piece must exhibit some visual
contrast, a contrast which would be possible to use to represent the
diflerence between hair and skin. Being certain shades of brown
and pink respectively would do the job, but so would many other
96 The Mona Lisa Mosaic
differences such as being spotted and plain, black and white, small
and large, etc. A hair piece and a cheek piece m ust also exhibit such
a contrast. M oreover within one picture the two contrasts should
be similar. How sim ilar will depend on features o f the general style
of the picture. But 1 m ight well find myself in trouble if I failed to
co-ordinate the two contrasts I intend to employ.
The crucial points in generating the indeterm inacy are now
thrown into relief. T h e only constraint on a defensible application
of P-prcdicates is that the properties described by the U-predicates
of the items to which they are applied provide a realization of the
abstract pattern. So to apply P-predicatcs we try to m atch the
abstract pattern of dem ands with some actual set o f contrasts,
relations etc., constructed from the U-predicates o f our given set of
objects. But the U-predicates of the objects contrast and relate in
many different ways. Also the set of P-predicalcs, together with the
abstract set of dem ands it generates, is open to extension or revision
as we perceive more of the scene or decide to represent it in greater
detail, and the dem ands so generated arc not clearly ordered. An
abstract pattern, in other words, is not a fixed m atter dictated once
and for all by the scene we are to depict. All that is required of it is
that it be rich enough to enable us to recognize a picture embody­
ing it as of the desired scene.
In constructing a picture we have then two points at which we
may be required to make a choice. O ne point is that a t which we fix
on how much of the scene is to be represented, in w hat style, level o f
detail and so forth. T he other is that at which we fix on how the
U-propcrtics o f the pieces shall be deployed to fulfil the pattern of
demands generated by the earlier decision. At neither point is there
any guarantee that there will be one unique best decision; hence the
indeterminacy. It is worth stressing, however, that the second
factor mentioned is, by itself, quite sufficient to generate indeterm i­
nacy. Even if we had something fairly firm in the way o f instruc­
tions about what features o f the scene are to be represented in the
picture we would still have no guarantee th a t we could not deploy
our pieces in two different ways to carry out the task.
Why should it seem ap t to describe application o f the P-
predicates as ‘subject to holistic constraints? The point is this.
Given an item o r group of items I cannot, ju st by exam ining them
 The Mona Lisa Mosaic 97
in isolation, decide w hat P-predicates are applicable to them.
However well they fit together to m ake an eye, I must not say
straight oil' that an eye is w hat they represent. I m ust first look to
the whole set of elements to make sure th at the other P-prcdicatcs
have suitable resting places and hence to make sure that I am
working witli the most convenient sub-set of P-predicates and
abstract dem ands. W hether I am operating a convenient set, and
have m ade workable initial moves in applying it, depend upon the
structure of the total set of elements. T hus every element is relevant
to the placement of every other clement. It is the need to keep an
eye on the whole which thus operates as a distinctive constraint on
this kind of enterprise. And it is the fact that the P-predicates come
as a set which in turn underlies that methodological constraint.
T his sort of holism is to be sharply contrasted with some general
policy, applicable to any judgement, o f keeping an eye on all the
evidence. O u r current holism is rooted in something peculiar to the
predicates involved and consists in a necessity to perform certain
specific kinds of cross-check 011 a certain particular group of items.
T he general policy of w atching the whole of the evidence on the
other hand is rooted in views about the unreliable nature o f our
.capacity to acquire knowledge and views about the ramifying and
interconnected nature o f the world. In the routine application of
many predicates this general policy does not give rise to the need to
check against the rest o f our beliefs; and when larger scale re­
appraisal does seem in order the policy docs not enjoin that any
specific sort o f check be undertaken. So, as I have stressed, the
indeterm inacy generated by the holism in the mosaic case is of a
different character from the admission th at future evidence may
shake some of our current firm beliefs. As we shall see in section
5.4, the epistcmological m atter of needing to keep an eye on the
whole o f our current view, and the consequent coherence element
in justification, does play an im portant role in the indeterm inacy
argum ent. But we m ust nevertheless keep cpistemological holism
sharply distinguished from the indeterm inacy being discussed.
W hat the mosaic case illustrates is that a predicate like ‘left
eyebrow piece’ (as f have imagined it used) is a different sort o f
linguistic item from a predicate like ‘red’ or ‘square’. T he point of
having this P-prcdicate is, we might say, different from the point of
98 The Alona Lisa Mosaic
having these U-prcdicatcs. T he point o f a P-prcdicate is, roughly,
to be able to report on or ask after certain projects or decisions; the
point is not to ask about items independent of our choices. And
the dcfensibility of those decisions involves a different sort o f
consideration (namely about the picture as a whole) from the
considerations involved in the defence of particular judgements
involving U-prcdicates. T he upshot o f all this is that it is not
difficult to persuade oneself that there is ‘no fact of the m atter’ to be
got at here.

5.3 H O L ISM , IN D E T E R M IN A C Y AND LANGUAGE

I low might all this apply to the case o f m eaning, i.e. to the problem
o f understanding the utterances and identifying the thoughts of
another person? This is what I shall now try to spell out.
In a radical translation project as usually conceived we have two
problem s, that o f translating the language of o u r subject and that of
finding out what he believes. T he tasks arc different because a
speaker may not say everything he believes and he ntay not believe
everything lie says. J want to avoid all the puzzles generated by
these facts and will adopt instead the fantasy of a simplified
situation - that in which we are confronted by someone I shall call
‘the candid babbler’. T he candid babbler is a person who conducts
all his mental life, his uoticings, musings, reasonings, concludings
etc. overtly and in words and who never knowingly speaks falsely.
W ith the candid babbler the problem o f translation and th at of
ascribing thoughts arc collapsed into the one problem of under­
standing his utterances.
T h e fantasy of the candid babbler is useful not only as a concep­
tual device for isolating the interpretation/translation problem
from the conceptual soup we w ant to clarify, but also for other
reasons. Thinking of our problem as that of understanding a candid
babbler brings out a plausible presupposition, common to the
approaches of both Q uine and Davidson, namely that in under­
standing another person w hat we m ust go on is his or her utter­
ances and behaviour in response to confrontations with the
 The Mona Lisa Mosaic 99
environment. T here is some level of description which is available
prior to (detailed) identification of meaning and which is the only
thing on which we can call to supply material, conceptually or
epistemologically, to ground specific ascriptions of content.6 If the
facts here are not enough to yield determ inate interpretations, says
this methodology, there is nowhere else to go; in particular we must
not appeal to further inner events which arc intrinsically inten­
tional, i.e. to C artesian ‘thoughts’.
O u r starting point for interpretation will he the utterances
themselves. These play a role parallel to that of the mosaic pieces.
T he features corresponding to the size, colour etc. of the pieces arc
the phonetic character of the utterances and what we might
vaguely call the circum stances - such things as the environment of
the speaker, the movements he and his hearer’s go through before
or after the utterance, what other utterances precede or follow etc.
These are described by our U-prcdicatcs. T he P-prcdicales arc
ones by which wc assign meaning — ‘said that it was raining’,
‘concluded that the train was late’, ‘reminded her that she must
take her um brella’, and the like.
T o interpret is to make intelligible, and this in two senses. First
there is the minim al sense where it simply means ‘such as to have a
meaning assigned’. But secondly there is a richer one, where
‘intelligibility’ also involves our seeing how the person could have
come to have those thoughts; this intelligibility requires us to see
some point, sense or dcfcnsibility in the other’s thoughts. The first
sort o f intelligibility docs not directly involve the second. T hus one
may, on authority from others, come to use some translation
scheme which enables one to provide a reading for various marks,
noises or gestures, while it still seems to one that in another sense
the interpreted marks, noises etc. ‘make no sense’. But the two
ways of using ‘intelligible’ do hang together in the long run,
because use of interpretation in the first sense m ust ultimately be
rooted in the second. It is understanding creatures in the sense of
seeing them as beings with a point of view of their own, a point of
view of which one can gel some inkling, that justifies postulating
translation schemes. I f this idea fails us in some particular case,
then we lose the idea that there is a person, a point of view, at the
100 The Mona Lisa Mosaic
origin o f some putatively meaningful items. But if it becomes clear,
so to speak, that there is no one there, then the meaningfulness
drains away.
T he condition o f ‘intelligibility’ th at we here impose on interpret­
ation is an obscure m atter. I t is not clear how much sym pathy with
or insight into another point of view (set of opinions, project, way
o f thinking) [ have to have in order to find it intelligible. It is not
clear that it is an all-or-nothing m atter.
But we may suggest at least two constraints which the condition
places upon interpretation. First our babbler, or candidate for
interpretation, will only be intelligible if we can see him to be
thinking, in part at least, about the world we share with him. This
is a fairly elastic constraint. It does not rule out our making sense of
someone who turns out to be unaw are o f his immediate surround­
ings and tuned in to broadcasts from the A ndrom eda Galaxy; nor
does it rule out our making sense of someone who claims to be
aware, some of the time, of things we cannot detect. T he point is
only that if his thoughts were solely about things which were
entirely inaccessible to us then we would never make any sense of
him.
T h e second constraint is that our candid babbler cannot be seen
as thinking about anything a t all unless he can be m ade out to be at
least minimally rational, i.e. to exhibit some sensible pattern in his
thoughts. It is im portant to be clear about the logical shape of this
claim. T he point is not that there is some particular pattern (modus
ponens, universal instantiation, avoidance of the gam bler’s fallacy or
whatever) which I can now specify and which he m ust exhibit. T he
constraint is only that he m ust exhibit some pattern or other which
allows us to see him as reasoning, however misguidedly.
The upshot of these two requirem ents is that the P-predicates we
wish to apply to him come as- a set. Both the nature o f rationality
(as we conceive it) and the nature o f the world (again as we
conceive it) help to suggest w hat the m em bers o f the set will be.
And, as in the mosaic case, the predicates generate a structure of
abstract demands, for example, that utterances assigned different
meanings should, in general exhibit phonetic differences and that
utterances assigned related meanings (where one would eviden­
tially support or be supported by the other) should occur some­
times in certain sorts of temporal sequence.
 The Mona Lisa Mosaic 101
From this it follows that holistic constraints operate in the
application of these P-predicates. We m ust not use one o f them
unless we are sure th a t the others have resting places and that
radically disruptive elements are not present. T he dcfcnsibility of
an assignment of meaning to an item depends on how things work
out across the whole set o f utterances. Indeed the task o f interpreta­
tion is now surprisingly close to that of mosaic construction. It is to
take some gabble o f noises and to tie them up to facts in the world
in such a way th at their utterer is shown to be someone who
succeeds in representing that world in a recognizable manner.
A nother person is a living picture of the world. But (if the analogy
holds up) it is our business as interpreters to arrange the assemblage
as a whole into an actual picture. A person may also do it for
himself, but he has no special authority such as m ight be given by
access to C artesian items.
T he method will be to plunge in with some likely looking initial
hypotheses and then to hope that the rest of the assignments can be
fitted in. Deciding to adopt a certain translation scheme is like
adopting a certain scale and style in the construction o f the picture.
I t commits us to the existence of certain other U -characterizable
items to fulfil the rest o f the abstract pattern. And if these are not to
be found, o r if in addition some further radically different and
unforeseen items appear, we shall be in trouble.
But as with the mosaic there is no guarantee that two or more
ways of fulfilling the abstract dem ands cannot be constructed. Part
of the flexibility comes from the fact that we have some discretion in
fixing the set of P-predicates we work with, i.e. we can attribute
bizarre beliefs to our subject (if we can find enough of a setting to
make them intelligible) and we can make him more or less observ­
ant. But even if we imagine this degree of freedom to be not very
extensive, even if we take it that there are a good many opinions,
rules of inference etc. th at m ust be found in any recognizable
thought about the world, the possibility still remains that we might
be able to find more than one way of tying up those obligatory
opinions with the utterances we take to be their vehicles.
So, in sum m ary, the methodology o f interpretation is holistic. It
can only consist in m atching some chosen set of abstract dem ands
generated by the P-predicates with the actual patterns detectable in
the U-predicates. But given that we have a choice of subset of
102 The Mona Lisa Mosaic
P-prcdicatcs and, more im portantly, that there may be different
ways of arranging U-prcdicatcs to satisfy the dem ands, there can
be no guarantee that there is some unique best interpretation o f a
particular utterance. Two entirely defensible translations schemes
m ight assign it radically different roles.
Hence the conclusion is th a t investigation o f the nature of
m eaning ascriptions themselves - the distinctive methodology of
applying them - shows th at insistence on a ‘fact of the m atter’ is
out of place, because even minimal realism is not defensible here.

5.4 CAN WK RESTO RK DETKRM INACY?

We move next to consider a counter-argum ent and some responses

to it. Someone might point out that the existence of holistic con­
straints on placement of mosaic pieces is not by itself suflicicnt to
ensure that there is no fact of the m atter about whether a given bit
is an eyebrow piece. We have implicitly used another im portant
assum ption, the alteration o f which allows us to construct a ease in
which the appl¡ration o f such P-prcdicatcs is both subject to
holistic constraints and also determ inate.
T h e extra assumption is this: th a t the pieces from which we are
to construct the picture have no natural suitability for the task.
T hey have not been produced with a view to constructing a picture,
so their num ber and colour are in no way controlled by the
elem ents in the scene which they arc used to depict. R ather the
construction o f the picture is entirely my project and the con­
straints on methodology result from my decision on w hat shall
count as a successful completion of the project. Knowledge of the
causal origin o f a piece (e.g. as p art of the rejects from a crockery
factory) is thus of no relevance to my decision on where to place it.
But let us alter this assum ption. Im agine that my task is as
before except that I see my raw m aterials actually being produced by
some device and I operate on the hypothesis that it is a device
which has been built precisely in order to produce m aterials to
construct a mosaic of the scene with which it is confronted. T he
device operates, I surmise, by directing some light sensitive mech­
anism at an area of the scene and stam ping out a mosaic piece in
 The Mona Lisa Mosaic 103
accordance with the intensity and nature of the light arriving. But
(let us imagine) I do not know as yet w hat optical resolution the
device incorporates, w hat sort of scanning pattern it has, whether it
introduces some systematic distortion of perspective or colour in its
stam ping process, w hether it spits the pieces out as soon as it
produces them or randomizes them inside first. Given this ignor­
ance, my methodology in constructing the mosaic must, in practice,
be the same as before, namely, entirely holistic. Consequently
something like indeterm inacy may arise. I cannot conclude from
the fact that some group of pieces arrives together and work
excellently to represent an eye that they must remain in that
position in my finished picture; I may well be able to make two
equally good pictures from the same set of pieces, and so on.
Nevertheless the indeterm inacy now looks more like the undcrde-
tcrm ination o f theory by data, because the predicate ‘right eyebrow
piece* will now have a new meaning, one upon which there is a
determ inate answer to the question ‘Is this a right eyebrow piece?’
- namely it will mean 'piece produced by the machine when
scanning her right eyebrow’.
O ne m ight object to the claim th at the methodology still incor­
porates a central holistic element on the grounds that, in theory at
least and were my ignorance less, there should be more direct ways,
which would supersede the holistic ones, of deciding what V-
prcdicate to apply to a piece. Might 1 not actually see a piece being
produced as a result o f scanning her right eyebrow? Might I not
have enough insight into the workings o f the machine to assign a
causal origin and hence a pictorial role to a piece, without the
consideration of any other factors?
T he answ er is that 1 cannot do this unless 1 have reason to
suppose th a t the holistic constraints are also obeyed — that the
other P-prcdicatcs will find suitable resting places and that some
set of abstract dem ands are fulfilled. I may have this assurance on
inductive grounds - previous success with this machine on other
scenes - or because I know in detail how the machine works. So I
may, given these sorts of evidence, be prepared to override holistic
considerations in the short term. But I can only properly do this
because 1 take it that in the long ru n the constraints will be
observed. I t is exactly this th a t induction or the design evidence
104 The Mona Lisa Mosaic
makes me think. I f it begins to look as i f l am wrong in anticipating
this, i.e. my reliance on the non-holistic indicators for placem ent of
pieces continues over a long period o f tim e to produce a non-
picturc, then ipso facto I lose my confidence in the device, think I
have misdescribed it or that it has broken down. And at this point
‘left eyebrow piece’ will cease to have determ inate application on
the basis of causal factors and will revert to the role or collection of
roles we originally assigned to it.
My well-supported conjecture about the nature of the device
thus docs not enable me to ignore holistic constraints in applying
P-predieatcs. Belief that the constraints are satisfied is part of what
1 endorse in describing the device as I do. But, and this is the
im portant point in rendering the predicates determ inate, if my
conjecture alxjut the device is correct it provides a standard by
which I can sensibly choose between alternative picture construc­
tions both of which satisfy the holistic constraints - a standard by
which 1 can say ‘T his is the picture the machine was designed to
produce’ or ‘This is the m achine’s picture’. T h u s we come to an
outlook in which a concept, ‘right eyebrow piece’, is subject both to
holistic constraints and also to causal conditions. And if we get an
insight into the workings of the machine wc can employ the causal
information to fix on placements for pieces. It is when holistic
constraints arc the only ones operative th at indeterm inacy ensues.
T he more complex concept now available is free of that feature and
hence fit to describe an aspect o f reality.
How could we apply a move of this sort to the case of meaning?
T he obvious suggestion will be that some non-holistic, e.g. causal,
constraint be incorporated, as an extra elem ent, into the meaning
of talk about meaning. U tterances will be linked with particular
stales of the world which causally explain their occurrence. T hus
alternative schemes of translation which satisfy the holistic con­
straints can be evaluated differently and m eaning rendered deter­
minate.
Something of this sort seems prim a facie very plausible since no
collection of bits of behaviour can be utterances or expressions of
beiicf unicss they arc under the control of, formed in response to,
the events and states they are about. So, one m ight say, if a person
has a belief as a result of causal interaction with a certain state o f
 The Mona Lisa Mosaic 105
affairs in the world (with many other provisos o f course), then it is a
belief that that state o f affairs obtains.
Rut although this general program m e is easy to sketch in outline,
it is difficult to carry through in detail. It encounters many diffi­
culties, some of which may be deep ones of principle. 1'or a start it
is clear that in applying it to determ ine content in the case of some
particular utterances we shall need a way of sieving out utterances
which are not assertive at all and also sieving out ones which arc
assertive judgem ents produced by ‘im proper’ causes, such as H u­
mean associative mechanisms or self-deception. If we attended to
causes to fix content in these eases, we would go badly wrong.
It cannot but improve the chances of the strategy we are con­
sidering to imagine all these problems done away with. So let us
modify our specification of the candid babbler to dem and that he
be an extremely literal minded fellow, all of whose utterances are
concerned to express truths about the world and all of whose
judgem ents are arrived at by epistemologically respectable routes.
This latter cannot be taken to mean that the babbler never makes
mistakes in reasoning, otherwise he becomes too unlike us to be
interesting. So we will allow that he may be in error in reasoning, in
the sense of being com m itted to far-reaching and principled mis­
takes. But he docs not make trivial slips and is consistently rational
by his own lights.
In the case of such a subject can the causal antecedents of belief
fix meaning? It is here that some difficulties of principle appear and
that Q uinean considerations about epislcmological holism become
relevant. In the case of the mosaic-making machine we were able to
home in on those particular causal antecedents which were of
im portance because we were able to distinguish sharply between
the standing causal conditions for the production of pieces (namely
the perm anent design structure of the machine) and the varying
conditions of input which explain, in particular cases, the varying
output. But we arc not like the machine. C ertainly we are like it in
th at facts in the world impinge on us and causally contribute to our
acquiring beliefs about them ; moreover causal connections of these
kinds may well be necessary for some sorts of meaning. It would be
absurd to deny this. But w hen facts impinge upon a person, they for
the most p art encounter a subject already possessing a complex
106 The Mona Lisa Mosaic
world view, itself the outcome of innum erable previous encounters.
T he ‘o u tp u t’ of judgem ent on a given occasion thus depends not
only upon current ‘input’ (if it depends on it at all, which, in cases
of highly theoretical reflection seems problematic) but also upon
previous inputs, llo w then do we choose among the causal ante­
cedents o f a judgem ent those which we take to be relevant to its
content? T he strategy under consideration requires that we be able
to do so dctcrm inatcly, e.g. in the light o f some other determ inate
(causal?) facts. But w hat if this cannot be done and the choice has
rath er to be made on holistic grounds?
W e may approach the problem from another angle by intro­
ducing in more detail the (heme o f epistcmological holism. T he
central claim here is that the correct epistcmological status of a
judgem ent cannot be determ ined by looking a t the judgem ent in
isolation. A judgem ent cannot be certified as something one ought
to have a great deal of confidence in, or something one ought to
throw out, except by considering how it relates to the whole body of
beliefs.
Acceptance o f this sort of holism docs not mean that we have
bought the idea that whenever we consider some question we ought
to survey the whole body of our beliefs before answering. This
would be absurd. Clearly a great deal is taken for granted and docs
its supportive epistcmological work unnoticed in the background.
T he need to acknowledge this should not disconcert the holist. But
all the same, his picture sceins much more plausible for some kinds
o f beliefs and their relations than for others.
Let us consider some examples. Suppose that the question has
arisen ‘Is this flower a fritillcry?’ T he other beliefs which I am likely
to bring consciously into play concern the colour, habitat etc. of
this flower and the colour, habitat, etc. of fritillerics. Beliefs such as
th at I can read, that a certain word is ‘purple’ or that the sun is
shining arc not going to figure explicitly in my deliberations. Yet it
is surely plausible to say th a t my assum ptions on these m atters play
an im portant background role in the m atter, in that were I to
change my views on them 1 might well change my verdict on this
flowci. But arc all my other beliefs playing a similar behind-the-
scenes supportive role? W hat about the belief that men have flown
to the moon or that W illiam II had red hair? It is certainly not clear
 The Mona Lisa Mosaic 107
that change in these would alter my judgem ent about the ilowcr.
T h e holist however will reply to this that these other beliefs might
support or disrupt my views about the flower because one cannot
rule out as impossible circum stances in which they would be
relevant. It could be the case th a t my confidence that 1 can
correctly interpret w hat it says in this foreign flower book rests in
p a rt upon the fact th at I can interpret some other sentence in that
language as saying that W illiam II had red hair. O r it might be the
case th at the fact th at men have flown to the moon is p a rt of the
support for a view of gravity which connects with a view about
geology which dovetails with a view about habitats which . . . and
so on. We arc not in a position to rule these out as intelligible
accounts of w hat someone m ight, given certain other evidence,
come to believe. And if someone did believe such tilings then the
beliefs about the moon or W illiam would have supportive and also
potentially disruptive roles in connection with the judgement about
the flower. T h e upshot is that my belief that William II had red
hair is doing something im portant vis-à-vis my judgement about the
flower merely by not exercising the disruptive potential it would
have if the right m ediating body o f beliefs came to be accepted. In
other words, the fact tfiat my belief about the flower is consistent
with my belief about W illiam is not as trivial as it looks. And it is
only given the structure of my whole set of beliefs —given the lack
of interm ediating items of certain kinds —that they arc consistent.7
W hat are the relations of this kind of cpislcniological holism to
sem antic holism? T he former may serve as a premise from which
the latter may be deduced, provided th at suitable extra premises of
a vcrificationist character are to hand. But if we reject such extra
premises then it seems much less likely that there should be a link.8
T his line o f thought would take us back in the direction of the
themes of chapters 3 and 4 and it is not how semantic holism has
been defended in this chapter.
W hat about a converse connection? If one accepts semantic
holism (as introduced in section 5.1 rather than in Fodor’s sense)
does epistemological holism follow? It. does not. To consider the
failure of epistemological holism one must imagine a schematic
situation where, for some p and q, in establishing whether or not p
it ju s t could not, ever o r for anyone, be relevant whether or not q.
108 The Mona Lisa Mosaic
This is prim a facie plausible for various judgem ents, for example
whether this is a fritillery and w hether nine times nine makes
eighty-one. Let us, for the sake o f argum ent, take it th at the
plausible supposition is correct. Does the denial of sem antic holism
follow? T o say that it did would mean th a t I was now in a position
to establish that a certain rem ark of another person expresses the
judgem ent about the square of nine w ithout considering everything
else that person says. Hut why should this follow from the epis-
temological irrelevance o f the fritillery judgem ent? T h e point o f the
inspection of everything else the person says is to establish the
absence condition required by w hatever dem ands we have for
minimal coherence and rationality. T h e epislcmological irrel­
evance of the fritillery judgem ent helps us to establish this; the
fritillery judgem ent is not going to do any disruptive work in
making our subject turn out to be incoherent. But that does nothing
to undermine the idea that meaning requires a discourse-wide
absence condition, and hence does nothing to underm ine sem antic
holism.
T he upshot then seems to be that sem antic and epistemological
holism are independent theses. W hat then underlies the latter? The
claim in the earlier discussion was that intelligible links between
any two judgem ents could be set up; one could always imagine a
case in which the one bore upon the other. (O n closer inspection we
shall see this to be so, contrary to our earlier concession, for
arithm etic and botany. For example, botanical views about fritil-
leries might bear upon the question o f w hether I was or was not
drugged and so upon the question of w hether I should accept the
conclusion of this seemingly valid proof about the square o f nine.)
But why is this so?
T he answer seems to have to do with the fact th a t we are not in a
position to insist that the world exists in isolated w atertight com­
partm ents. To establish epistemological non-holism we would have
to be certain that some aspects o f the world were independent of
each other in an extremely strong sense. It would be required th at
no theory could possibly unify the two sorts of phenom ena in such a
way that considerations from th at theory could bear cpistemologi-
cally upon the acceptance of individual statem ents about the two
realms. If such a theory was imaginable (even if we did not possess
 The M o m Lisa Mosaic 109
it) then we could im agine evidence from the two realms supporting
the theory which in tu rn bore upon an individual judgem ent about
one realm. And this imagined structure has brought the individual
judgem ents from the two areas to bear upon each other. T he
impossibility o f such a theory could come about in either of two
ways. It might be th at we knew ourselves to be infallible with
respect to the two subject m atters. In this case a unifying theory
m ight exist in the sense o f a compendious or illum inating overview.
But such a theory (in effect a mere conjunction) could not, because
o f the infallibility, play any cpistemological role. O n the other hand
wc might have confidence in the impossibility of any linking theory
on the basis o f some scientific or metaphysical com partm cntaliza-
tion of things. But it seems extremely implausible to suppose that
cither the infallibility or the com partm enlali/.ation condition is
fulfilled.
T he im portant upshot, as far as the attem pt to restore determi-
nacy of meaning is concerned, is that we should accept some
version of the Q uincan thought that our beliefs face the world en
bloc. Let us now go back to our rational candid babbler. We may
establish that some particular fact which impinges on him is
causally relevant to the production of some particular utterance.
But if the babbler is, as im agined, rationally conscientious then the
explanation o f his making th at judgem ent will involve his back­
ground view of the world. He judges as he does in part because, for
example, there are no elements in his view which lead him to reject
the judgem ent which the current confrontation prom pts him to.
A nd w hat explains the absence of such elements? Both the nature of
all the actual encounters which led to the beliefs he does have and
also the absence of some encounters which, given his cognitive
nature, would have led to the formation of disruptive beliefs had
they occurred.
T h e non-holist supposes that the description given in the last
sentence, fixes some totality o f causal information which, in theory
if not in practice, one could imagine assembling and which, if
possessed, would single out some causal factors from others as the
ones proper to supply intentional content. But it is far from clear
that these suppositions are correct.
O ne sort of worry arises from the idea that we can find a
110 The Mona Lisa Mosaic
principled limit to the num ber o f standing negative conditions to be
taken into account. Identification o f an utterance as expressing a
given thought involves the idea th a t the m aker of the judgem ent
would respond appropriately to a variety of new juxtapositions,
challenges, and so forth. But how arc we to list in advance w hat
they all arc or w hat the appropriate reaction is, given that we arc
ourselves enmeshed in the middle of the development?9
A second difficulty lias to do with the identification of falsehood.
N ot every utterance can mean even one of its causes, unless our
subject makes no mistakes. So, given that we arc aw are of the
possibility of error, we cannot ever move from causal information to
assignm ent o f m eaning w ithout some extra premise about the
tolcrabiiity of im puting error. W here is this to come from? Will it be
non-holistic? T he answer is far from clear.
A third difficulty is marginally more concrete than the highly
abstract considerations m arshalled so far and gives more sense of
the problems encountered in moving from causal to intentional. It
poses the question of whether we could even be sure of distinguish'
ing indcxical statem ents from non-indcxical theoretical ones on the
basis o f causal inform ation, not itself employed under the con­
straint of a holistic story.
Suppose that we have securely identified sentence types and
tokens, and also their negations. An indcxical type will then, it
sccrns, be one, different tokens o f which arc happily accepted as
having different truth values. So it is supposed we can observe a
group of persons oscillating between asserting a sentence and its
negation and decide that the sentence m ust be indcxical. Having
got this clear we then move to identify causes and so content. This
is the usual story. But how do we know that we have got even the
first stage right? Perhaps the persons arc changing their minds
frequently about some theoretical m atter. Suppose they arc gazing
at a screen on which two slowly growing lines arc exhibited. They
u tter affirmative-and negative tokens of a type according as the
lines converge or diverge and we establish th a t the positions of the
lines are causally relevant to their utterances. Docs it follow that
they are talking about the position of the lines? M ight they not be
expressing changing theoretical views about w hether or not the
m oon has a m agnetic core, according as the results o f some analysis
gradually unfold?
 The Mona Lisa Mosaic 111
O ne might object here th a t wc have forgotten the condition that
our candid babblers m ust voice all their thoughts. I f things were as
I have suggested, would there not have to be utterances other than
those I have mentioned to which wc could assign an observational
and indexical role? C an we make sense of the idea of someone
‘seeing’ the composition o f the moon in a screen display without
th at observer being aw are o f and voicing cpistcmologically inter­
m ediate links of an observational kind? Wc may well be able to
make sense of ‘seeing’ things other than what is (in some sense)
present and causally effective; this would be a consequence of
allowing (at least some) observations to be ‘theory laden’.10 Hut the
currently imagined case seems to go even further than this.
Concession of the force of this objection however leaves us able to
construct an analogous problem . W hat of somebody who utters
pairs of apparent indcxicals, one o f which supports the other? Can
we be sure that both arc really indexical? M ight not one be an
indexical and the other a theoretical view whose plausibility
changes in step with the acceptability of the indexical? It seems
that it m ust be features of how things pan out, from the point of
view of coherence etc., in the rest of w hat he says, which would
enable us to decide between the two views. And what entitles us to
think th at these features will be non-holistic?
T he thrust of all these considerations taken together is to suggest
that causal considerations will indeed be highly relevant in the
assignment of meaning but that they do not play a role alongside
and independently of holistic constraints. R ather they arc them­
selves among the U -predicatcs wc use to support our attributions of
P-prcdicatcs. As such they serve as grist to the holistic mill and not
as the desifed independent check on its results.
I am far from claiming th a t the arguments of the past few pages
show that this is so, only th at it is a hypothesis which wc cannot
dismiss. And if it were correct then wc seem to have no defence
against the conclusion that m eaning ascriptions of their nature
cannot be guaranteed to obey the law o f non-contradiction and
hence to the conclusion th at meanings are not real and that there
are no facts about them.
 6
The Slide into the Abyss

6.] T H E IN C O M P A T IB IL IT Y O F REALISM AND

M EA N IN G S C E P T IC IS M

How should wc respond to the conclusion of chapter 5? It ofl’c rs us

scepticism about meaning - not in the sense of claiming that there
is something In know which we do not know hut in the sense of
claiming that there are no facts to be known. This position is the
same as that recommended by Q uine, but generalized to all
sentences and reached by a different route. Psychology, history,
literary criticism and everyday talk of thoughts and feelings are
thus revealed as (mere) pattern construction rather than fact
discovery. But other investigations are, for anything wc have so far
said, still to be credited with finding how things really are. T o make
this contrast is not necessarily to think psychology etc. unim port­
ant, still less to suppose them dispensable. But it is certainly to view
them as something sharply different from investigations in the
natural sciences. And given the honorific status of words like ‘fact’
and ‘real5 it is difficult to avoid the implication o f some kind of
downgrading.
Tw o questions now arise. Should we accept this view? And does
W ittgenstein follow something like the route mapped out in chapter
5 to reach it? I wish to suggest th at we should answer ‘no’ to both.
O n the second question, as I rem arked in the Introduction, we
certainly do not find in W ittgenstein any explicit draw ing of
sceptical conclusions about m eaning such as we find in Q uine, nor
do we find claims that only the natural sciences describe the real.
But on the other hand wc do find attacks on a certain conception of
w hat facts about m eaning could be like (the negative argum ents so
forcefully expounded by K ripke), And the materials for some
 The Slide into the Abyss 113
argum ent like that of chapter 5 may seem to be provided. T here are
many passages in W ittgenstein where he stresses that when we see
a gesture, words or facial expression as ‘full of m eaning’ this is
bound up with its occurrence in a context.1 T hus ‘holism’ of some
kind there certainly seems to be in W ittgenstein.
But (as 1 shall try to suggest in the next chapter) on examination
this holism proves to be even more all em bracing than the version
we have discussed so far. T hai version emphasized what a person
could coherently be said to believe, given that he or she believed
such and such else. It thus had to do with the existence of suitable
patterns among things which are to be taken as representing the
world. And its central theme is that the things a person takes to be
the case must hang together in a certain way. But on W ittgenstein’s
view to see what someone means by some utterance I may need to
see not ju st w hat else he or she takes to be the case but also what he
or she is trying to do, w hat he or she takes to be im portant and why;
it is only in grasping these things th a t T can understand even his or
her most austerely descriptive words.
T his introduces a very im portant theme — the interpenetration
and interdependence of description and project, of theoretical and
practical, to which wc shall return. T his broadening of the scope of
possible considerations needed for understanding does not, I shall
suggest, leave the indeterm inacy of meaning even more (irmly
rooted. R ather it moves us on to a ditlerent conception o f ‘realism’
from either the m irroring or the pragm atist one —and to a different
conception of episteinology - which ultimately allows us to escape
from the sceptical conclusion of chapter 5. But before we. can begin
to examine this alternative vision we need to turn to the first
question 1'mentioned above, namely the acceptability of the posi­
tion now arrived at. In the rem ainder of this section I shall argue
that no kind of realism can live happily with the sceptical view of
meaning we have arrived at; in the next section I shall contend that
realism is not easily dispensed with.
In the mosaic-construction story, and equally in the story about
interpretation of language, we took it for granted that we, the
picture builders or interpreters, were seeing Lisa in her landscape
or thinking about the world around us. W hat we do is tell a story
about how we might go about assigning meanings to items other
114 The Slide into the Abyss
than our own utterances or states. We give an epistemological and
methodological account, from which certain ontological con­
clusions arc supposed to follow.
B ut w hat happens if we try to think through the application of
these conclusions to our own case? H ere we return to a theme
touched on briefly in section 4.3. Suppose for example 1 say ‘There
is a tree over there’. Application of the conclusion of chapter 5 to
this utterance tells me that 1 have made a noise which someone else
(or I myself if I had enough ingenuity) could perhaps with perfect
defensibility take to mean not that there is a tree over there but
som ething quite different — that the moon is spherical.
1 am now driven to exclaim ‘But 1 don’t mean that the moon is
spherical. I mean that there is a tree over there.’ O u r theory
however tells us that this is itself just another piece of noise-making
to which, in turn, various different meanings m ight be assigned.
‘Assignment o f m eanings’ cannot itself be taken to set up some
connection between noises «and som ething other than noises, be­
cause the noises th at arc used to do the assigning cannot themselves
be allowed to have connections with non-noises which the original
subjects o f interpretation were denied. All that ‘assignment of
meanings* can do is to set up correlations between noises.
W e arrive thus at a (tem porary) resting place in which we say to
ourselves: ‘T he notion o f meaning has been revealed as incoherent;
there arc really nothing but noisc-makings, which in turn provoke
other noisc-makings. But none o f these noises have any properties
beyond their non-reprcseniational ones. T he truth is that the idea
of there being items (or performances or states or activities) which
point beyond themselves, things in virtue o f the occurrence of
which we are in touch with or open to the nature o f something other
than ourselves, is an illusion.’
But this conclusion cannot- provide a very satisfactory stopping
point. In enunciating it we seem to ourselves for an instant to have
got the real truth about how things arc. W e have plumbed the
depths and there is some melancholy satisfaction, even a sort of
thrill, in having got to grips with the hum an condition, however
illusion-ridden and unfortunate it turns out to be. However a
m om ent’s reflection upon w hat our conclusion says (or at least
seems to say) puts any such feeling of satisfaction or stability in
 The Slide into the Abyss 115
jeopardy. O u r understanding of the conclusion (whatever ‘under­
standing’ is allowed to be, even if it is ju s t a disposition to move on
to some other utterances under the stimulus of this one) drives us
on to say that if the conclusion is ‘right’ (whatever that now comes
to) then its enunciation cannot have the status of ‘being the truth
about how things are’ which we hoped for it. T he more dcfensibility
the conclusion has, the less any such defcnsibility (‘tru th ’) can
am ount to. In particular it cannot am ount to that representation of
w hat is independent o f us, belief in which is the hallmark of the
realist of whatever variety.
A central thread of these rem arks is that there is something
extremely strange about assigning a high semantic status (‘really
true’, ‘limning the true and ultim ate nature of reality’ or what not)
to some class of utterances, and a diJfcrcm and lower status (‘no
fact of the m atter’) to utterances which report, or would prima facie
be taken to report, the features o f the first sort of utterance - their
meanings —in virtue of which they were suitable recipients of the
honorific-seeming label. T he suspicion arises that the disparage­
ment of certain supposed properties of the vehicles (‘they do not
really have one m eaning rather than another’) must carry over to
disparagem ent of w hat one m ight attem pt to do (‘truly describing
the real world’) by use o f the vehicles.2
I w ant to argue now th at we can make this charge stick a little
more firmly by showing that an adherent of indeterminacy about
meaning cannot with confidence defend the law of non-contra­
diction, even for those statem ents about which he says that there is
a fact of the m atter as to their truth or falsity. Me thus cannot
deliver even the minim al dem ands of realism.
Let us remind ourselves of what the principle of non-contra­
diction says. It is the claim th at when I know that each of two
incompatible utterances could be equally and fully defensible then
I cannot take it th at those utterances arc to be realistically con­
strued. T his is not to say th at I take it that I can always know
without any difficulty o r reflection w hether any given incompatible
sentences could both be fully defensible and so whether realism is
the right construal o f them . T h a t would rule out the possibility
of intelligible philosophical dispute about controversial cases. W hat
it is to say is th a t reasons for th in k in g th a t the principle of
116 The Slide into the Abyss
non-contradiction does not hold in a particular case are ipso facto
reasons against realism in th a t case. T h e greater the likelihood th a t
non-contradiction fails, the greater the im plausibility of realism.
T o get the argum ent under way we now need to note an im port­
ant feature of interpretation as described in chapter 5. It is this:
Although I must make my subject come out as having a set of
opinions which is recognizably a view of the world 1 share with
him, there is no b ar to attributing to him w hat I take to be false
opinions. Putting things in the translation idiom, I do not always
have to map tilings to which he assents onto things to which I
assent, or vice-versa. M ore particularly, there is no bar, in the
methodology as sketched, for supposing that in some cases I might
not find two acceptable translation schemes, which translate a
given sentence of the other person into two sentences of my
language which 1 take to be incompatible and to be subject to the principle o f
non-contradiction. For example, there might be a sentence of the
other’s language which I can translate defensibly cither as ‘T he
moon is spherical’ or as ‘T he moon is a flat disc’.
Remembering th at the indeterm inacy claim is applicable to my
own case also, let us imagine th at my own set o f utterances is an
example of what has ju st been described. So when I say ‘T he moon
is spherical’ I can either translate myself as speaking familiar
English (let us call it English 1), and take it that I mean th at the
moon is spherical or 1 can take m yself to be speaking another
language, call it English 2, in w hich w hat I mean is that the moon
is a flat disc. (You may object that this is an implausible claim
about our actual utterances. Yet even if this is conceded, and it is
not clear th at it should be, the objection will turn out not to help as
much as might be imagined. W e shall return to it below.)
Now what the indeterm inacy o f m eaning thesis tells us is that ‘is
a speaker o f English 1 who' says “ T he moon is spherical” afid
believes that the moon is spherical’ is one predicate applicable to
me; it also says that the predicate ‘is a speaker of English 2 who
says “ T he moon is spherical” and believes that the moon is a flat
disc’ is also applicable to me; and finally it asserts that there is no
iact of the m atter as to which description is really true of me; both
are equally and fully defensible.
At this point we shall make a further assum ption, which m ust be
 The Slide into the Abyss 117
allowable, namely that when I utter ‘T he moon is spherical’ w hat I
do is ‘right’ and fully defensible. T h e assum ption m ust be allowable
because all parties to this dispute are agreed that some utterances
are ‘right* and ‘T he moon is spherical’ seems an uncontrovcrsiai
candidate. At the m om ent we are all claiming to understand
rightness in a realistic way, and, on my contention - which the
believer in indeterm inacy claims to be able to endorse - this
implies th at the utterance is subject to the principle of non­
contradiction.
W hat does this ‘lightness’ have to do, if anything, with the
language I am taken to be speaking? T he advocate of indeterm i­
nacy is in trouble w hatever he says, lie could m aintain that
w hether I am speaker o f English 1 or of English 2 boars directly on
this assessment. I might, for example, be right qua speaker of
English I but wrong qua speaker o f English 2. But if this is the line
we take then one could hardly say that the two descriptions o f me
are merely different notations for the same facts. O r to put the point
another way, if we say that 1 am right qua speaker of English 1 and
wrong qua speaker of English 2 and, at the same time, say that there
is no fact of the m atter as to which language 1 speak, then we have
already thrown in the realist sponge as far as the moon is con­
cerned. For what we have said implies th a t there is no fact of the
m atter as to w hether I am right or wrong and hence no fact about
how it is with the moon.
So to avoid this and to m aintain, as he thinks, his right to be a
realist, the advocate of indeterm inacy ought to hang on to the idea
that my rightness is not bound up with the (non-factual) question
of which language 1 speak but is independent of it. Then the
difficulty with being a realist comes out another way. 'I'he claim
now is that I can properly be taken as speaking English 2 and
saying in it that the moon is a flat disc and that my utterance is
correct. Now, by anybody’s lights it must be agreed that if it is ‘all
right’ to do w hat we properly describe as ‘saying that the moon is a
flat disc in English 2’ then it is ‘all right’ to do what we would
describe as ‘saying that the moon is a flat disc in English 1’.
Changing the language in which I express my views (whatever
account we give o f w hat th at am ounts to) cannot change the views
from right to wrong.
118 The Slide into the Abyss
But the effect of this admission is disastrous. T he advocate of
indeterm inacy now has no grounds lor finding any fault with a
person who utters those sentences of English 1 which wc would take
as expressing the bizarre beliefs which arc attributable to me qua
English 2 speaker but urtercr of all the sentences I now utter. In
short, the advocate of indeterm inacy has got into a position where
acknowledging the full dcfcnsibility o f ‘T h e moon is spherical’ as
said in English 1 is no bar against acknowledging the full dcfensi-
bility of saying ‘The moon is a flat disc’ in the very same language. And
now we have lost grip on our supposed com m itm ent to the law of
non-contradiction with respect to these utterances about the moon
in th a t language, and with it wc have lost the right to claim to be
even minimal realists about the shape of the moon.
But w hat of the objection th at this is all irrelevant because the
imagined alternative translations arc not plausibly forthcoming?
T his may seem a forceful riposte. T he claim o f indeterm inacy is not
th a t I can take an arbitrary utterance, assign it any meaning I like,
and then be certain of finding a translation scheme which will
vindicate the assignment. In the case o f the mosaic wc adm itted
th at certain particular sets o f elem ents m ight in fact determ ine or
rule out given roles for a particular piece. So the sam e m ay be the
case for interpretation o f sets of utterances. And if our actual sets of
utterances are not such as to allow for interpretation o f ‘T he moon
is spherical’ as ‘T he moon is a fiat disc’, then my argum ent seems
to lapse.
O ne reply would be to rem ark th a t access to this fact about
lim itations on interpretation m ay be pretty difficult to obtain.
Indeed it is far from clear that we could tell a story about how we
could be reasonably certain of it. For the mosaic, we have a finite
and small set o f elements to play with; there is some sense in the
idea o f a full and systematic survey o f all possibilities or arrange­
m ent. B ut how would wc do anything sim ilar for utterances (es­
pecially if wc c an n o t even be c e rta in th a t wc have rightly
distinguished indexicals from non-indcxicals, as suggested at the
end o f chapter 5)? But let us waive this point and allow th a t we can
have reasoned confidence th a t our actual utterances about the
moon, together with the other things we say, form a set which
resists the kind o f translation imagined above.
 The Slide into the Abyss 119
Does this help to restore a respectable sense of commitment to
realism? It does not. W hat is required to undermine the argum ent
is to make plausible the m uch stronger claim that there could not
be any set o f utterances such that both (1) there is in it a sentence
which is suitable to be interpreted as meaning that the moon is
spherical and also (2) it is as a whole so structured that it is possible
to set up an equally good rival scheme in which that sentence is
interpreted as saying th a t the moon is a Ilal disc. And what grounds
have we been supplied with so far on the basis of which we could
rule out the possibility o f such a set of utterances? T he cpistemo-
logical task is far m ore dem anding than the adm ittedly tough one
we contem plated in connection with our actual sets of utterances. It
is difficult to see how one could so much as begin to go about trying
to show the non-existence of the specified sort of set. O ne might be
tem pted by a line of argum ent starting in a priori fashion from
something we take ourselves to know about the meaning of ‘T he
moon is spherical’, for example th at it is incompatible with ‘T he
moon is a flat disc’. Such an argum ent could proceed via contra­
position o f the considerations I have urged to the denial of the
existence of the relevant sort of set of utterances. But this approach
is emphatically not available to the proponent of indeterminacy.
W hat lie can allow him self to know about meaning must be
established through exam ination of patterns in sets of utterances
and not the other way around.
But if the existence of a set of utterances of the type described is
an cpistcmic possibility, indeed even a probability (given only the
kinds o f data wc can imagine ourselves to have on the holistic
methodology of chapter 5) then our confidence in the propriety of a
realistic construal o f rem arks about the moon m ust be correspond­
ingly weak. And an exactly sim ilar argum ent can be run for any set
of incompatible sentences we please, rendering realism an un­
attractive position right across the board. In as much as wc have
not proved the existence o f sets o f sentences of the kind described,
minimal realism rem ains an option. But, given that minimal re­
alism looks so epistemologically precarious, it would surely be
desirable to give an account of w hat we are up to in our judge­
ments, and w hat success in making them amounts to, which docs
not require the tru th of realism . From the perspective of the
120 The Slide into the Abyss
position sketched, it looks much more certain th at our judgements
are well-founded and that some of them are ‘right’ than it does that
any of them conform to the conditions of minimal realism. So there
is now strong pressure to find an account of ‘being right’ and
related notions which cuts them free from the marginal looking
concern of whether a ‘realist’ construal is defensible. T his is also
something which is strongly indicated by the earlier move which
separated the question o f w hether some utterances is right or not
from the question o f what it is to be taken to mean. Realism
remains on the books, perhaps, but concern with its truth, in
general or in the particular case, appears a profitless enterprise.
W hat might an alternative account of judgem ent and its con­
cerns be like? It would very plausibly take the form of a general
version of pragmatism, and it is unclear that there is any other
option. The dillicultics that have been raised for the com bination of
indeterminacy claims about m eaning with realism are closely re­
lated to the arguments advanced in section 4.3 against the coher­
ence of a possible Q uinean position com bining realism about the
subject m atter o f underdetennincd theories with claims of indeter­
minacy of meaning lor sentences o f those theories. In both cases the
trouble arises from attem pted com bination or the position about
approval and condem nation of sentences implicit in the realism
with the position on approval and condem nation which the indeter­
minacy claims involve. T h e way out for Q uine was, I suggested, to
embrace an explicit instrum entalism , and it is the analogous move
that we now need. Since Q uine left observations and logic un­
touched by the indeterm inacy, their realism is similarly left un­
touched by his instrum entalism . But our current claim about
indeterminacy, being that m uch m ore general, brings with it a
correspondingly generalized form o f instrum entalism and denial of
realism. Thus the claim will be that w hat we do in thinking is aim
at the formation of some workable, coherent and useful body of
judgem ents. We do not have a body of given data to systematize;
each potential new judgem ent is rath er to be assessed by its
coherence with w hat is already there. And any already accepted
judgem ent is assessed, if we need to, in sim ilar fashion.
It may be, one might add, that we get tem porary configurations
of judgem ents in which some of them appear as incompatible anil
 The Slide into Ike Abyss 121
as subject to the principle of non-contradiction. If such a configura­
tion occurs then it will give its users grounds for claiming to be,
tem porarily and in a modified sense, ‘minim al realists’ about the
subject m atter of those judgem ents. It may also be that these
speakers are, through lack o f imagination, under the impression
that certain of their sentences arc such as to be always subject to
the principle of non-contradiction. Perhaps this gives rise in them
to the idea o f ‘describing the real nature o f a determ inate indepen­
dent world’. But, says this view, such a metaphysical reading of any
o f our judgem ents or utterances is always a gratuitous and epistc-
mologically ill-founded extra, from which it is better to refrain; in
any case it has nothing to do with any feature of the judgem ents
which really m atters to us.
In sum m ary then, the thrust of the argum ents o f this section has
been to suggest th at denial o f realism about meanings (on the
grounds given in chapter 5) leads on to dow ngrading of the im port­
ance of realism about any subject m atter whatsoever. (Another line
of thought to the sam e conclusion can be constructed from a
generalized version of the epistemological holism we exam ined in
section 5.4. W ithout pursuing this in detail, we can see th at the idea
that every judgem ent gets its epistemological status from its coher­
ence with all rem aining judgem ents, is likely to lead to the idea of
generalized fallibilism, which in turn leads to the idea that commit­
m ent to non-contradiction in the case o f any particular judgem ents
can never he more than a m atter of provisional convenience.
W e contem plated at the start of this section a stable position
which combined realism about the subject m atter of the natural
sciences with non-realism about the subject m atter of the (broadly
speaking) hum an sciences. But such a position has proved unten­
able. This does not yet provide reasons for supposing th at there is
something amiss with meaning scepticism. T he question now is
w hether thoroughgoing pragm atism is som ething we can live with.
I wish to suggest in the next section that it is not.

6.2 T H O R O U G H G O IN G PR A G M A T ISM

By ‘pragm atism ’ I understand the view th at we choose what

122 The Slide into the Abyss
concepts to have and w hat judgem ents to make. T he choices are
m ade in the light o f the expected usefulness of the judgem ents,
given the rest of the scheme o f thought; and any current feature of
the scheme is conceived of as open to modification.
T h e appearance of the notion of choice may seem odd in this
context, particularly if we have arrived at this position by the
cpistcmological route. T hat route asked us to accept some thesis
about w hat form evidence always had to take. But how do we get
from the thought that justification is always a m atter of coherence
to the idea that judgements and concepts are chosen? There is a
link here which is worth examining.
T o sec it wc need to consider w hat account a pragm atist is to give
of the pressures which lead us to modify and improve our scheme of
thought. W hat is it for a scheme to be useful, to succeed or fail? O n
the earlier explicitly instrum entalist view discussed in chapters 3
and 4 wc had ‘recalcitrant experiences’ in the form of the subject
m atter of infallibly known datum statem ents. These provided
pressure to modify the scheme by which wc systematized and
predicted them. But lacking experiential d ata, what account arc wc
to give now?
W right olTcrs the only plausible answer.'* We have ‘recalcitrant
experience’ in a revised sense appropriate to pragm atism , he sug­
gests, if wc find ourselves spontaneously inclined to accept an
inconsistent body ol judgements. T o sec how this might work let us
draw a rough and ready distinction between observational and
theoretical judgements, not on grounds o f the supposed complete
certainty o f the former or of their conceptual independence of the
latter, but simply on the grounds of w hat is and w hat is not
explicitly arrived at inferentially. O bservation statem ents are those
which present themselves non-infcrentially to nearly all of us as
acceptable and which w c a r e agreed in w anting to preserve -
although o f course we can reject some o f them on holistic grounds.
Tension in a system of thought then appears in the following
schem atic form: wc find ourselves inclined to assent to two observa­
tion statem ents, say P and not R. But our theory T , together with
its underlying logic L, entails th a t if P then R.
T his now is where the choice comes in. Wc have to decide which
elem ent to alter. No particular adjustm ent is forced upon us. I f it
 The Slide into the Abyss 123
were we would be back in the realm of the given. All that is
required is removal of the tension, but this can be achieved in
various different ways. How do wc choose? According to the
pragm atist wc do so in the light of expectations about which move
would minimize future tensions, i.e. future occurrences of recalci­
trant experiences. (O r a t least this is one im portant consideration
guiding choice. Some kinds of pragm atism would allow in other
factors, e.g. the pleasantness of having in sonic particular element.
But wc shall not pursue these complications.)
O ne may still find the talk of choice odd. Why speak of choice
and action when judgem ent of probabilities is al! wc need? Why not
ju st recognize that possibly not P and possibly R but most prob­
ably not T? In effect the pragm atist can concede this. He builds in
recognition o f these possibilities as proper and continuing re­
sponses to the tension, in as much as he knowingly holds himself
ready to discard die statem ent that P or not R il'al some future time
that seems the advantageous thing to do. But the choice he insists
on is that of putting down some one thing rather than another as an
assum ption which (for the time being a t least) wc arc going to work
with. ‘W orking w ith’ an assumption has two faces. It is a m atter of
exploring a theoretical formulation but also of a policy of fixing
non-theorctical courses of action. (These two things, cxploriug a
theory and acting upon it, do not necessarily go together. [ can
explore one theory while acting on another. But it would be odd
were I to act upon a theory which 1 did not explore.)
So the pragm atist’s only difference from the more purely cpis-
temological holist (who wants to talk only of probability) is that the
pragm atist (plausibly) adds in the idea that, confronted with
differing world views, all of which are defensible in differing de­
grees, we will have a non-cpistemic decision to make about which
of them to take seriously - i.e. about which of them to use as a
basis for action, both practical and more purely investigative.
For the thoroughgoing pragm atist then, everything, including
the very idea of there being incom patible judgem ents obeying tfic
law of non-contradiction, is up for grabs. There is no feature of
current practice which he is not prepared to contemplate abandon­
ing; and his taking seriously w hat lie docs now take seriously is, he
says, the outcome of choice.
124 The Slide into the Abyss
B ut is everything up for grabs simultaneously? Even the hardiest
pragm atist might boggle at this since it asks us to contem plate a
choice made by a conceptless (and hence goalless and information-
less) creature. So the N eurath boat picture is surely a better image.
Everything is up for grabs but piecemeal. T he pragm atist thus
adm its that in order to conduct any intellectual operations of
assessement and choice we m ust make some assumptions and
adopt some procedures. But any given assum ption or procedure
can be asked to justify itself and it will do so by appealing to its
advantages, compared to its rivals, in contributing to the effective
functioning of the whole. So the ‘facts’, in the light of which we
make choice of further ‘facts’, are there only because we have
chosen to put them there.
There is, however, something incoherent in this view when we
follow it through. A central point in seeing why this is so is the
following. It is taken for granted (and rightly so) that we arc not
able to choose w hat to believe in the sense o f m aking up some world
entirely to suit ourselves. Pragm atism is not the fantasist’s or
self-deceiver’s charter, telling us each to build a world to fit our
heart’s desire. It is in the business of giving an account o f recogniz­
able hum an intellectual endeavour. H ere, all too clearly, things
sometimes go badly wrong. Things turn out other than we would
wish them and we recognize that they have done so. C onstraining
forces thus operate upon our belief system; there is something,
independent of us and not subject to our choice, to which we are
endeavouring to accommodate ourselves. O u r pragm atic choice
cannot then be seen as freedom to ignore E x tern al’ pressure en­
tirely; it must rather be read as freedom over which adjustm ent to
make in response to that pressure.
To see the incoherence in thoroughgoing pragm atism , let us now
return to the notion of recalcitrance as W right sketches it and
follow through the ingenious line of argum ent he develops.5 T he
situation was that we found ourselves inclined to assent to both P
and not R while already being com m itted to a theory T together
with a logic L which entails th at if P then R.
initially one m ight suppose that the choice was between T , P and
not R. But the pragm atist will deny that these are the only options.
‘W hat about adjusting L?’ he will say. So we have a list of four
 The Slide into the Abyss 125
things to choose from in making our adjustment - T , P, not R and L.
‘But is this really all?’ enquires W right. W e seem to be taking it
as absolutely unquestioned that. L in application to T yields that if
P then R. Let W be the claim th a t T entails th at if P then R when
we use logic L. M ight not acceptance of that be o ur mistake?
Perhaps we are w rong in thinking that we are in a situation where
experience is recalcitrant! So now we have five things to choose
from, T , P, not R, L and W.
At this point W right perceives a problem for the pragm atist
which he sets out as follows. According to the pragm atist, choices
are to be m ade in the light of the expected am ount o f recalcitrance
they will produce. B ut w hether or not some experience is recalci­
tran t - i.e. prom pts us to tension-inducing observation statem ents
—has now itself become a m atter of hypothesis and so o f choice. In
accepting or rejecting W we render P and R recalcitrant or not
recalcitrant for T with L. And the advice on how to choose any
hypothesis is to see how m uch recalcitrance it would produce. So
the injunction to choose a system which minimizes recalcitrance
becomes an injunction to choose a system which minimizes what
the best system says is recalcitrance. And, W right rem arks, this is
‘hopelessly im predicative’ and can yield no methodological advice
a t all.6
How should we respond to this? It is clear that (at least some of)
the kinds of judgem ents we are interested in at. the m om ent are,
broadly speaking, judgem ents o f logic, necessity or conceptual
connection. They are judgem ents in which we lay out fundamental
links between linguistic moves, in virtue o f which the truth or
falsity of some can be brought to bear on the tru th or falsity of
others. Before we continue by discussing W right’s view o f the
difficulty we may note that we are in a position to rule out one style
of theory about these problem atic judgem ents, nam ely a crudely
construed projectivism. This sort of theory says th at our judge­
ments of necessity represent a reading onto the world o f some
psychological facts about the limitations of our im aginations.7
W hat seems attractive here is the replacem ent of metaphysically
and epistemologically problem atic ideas with the unproblem atic
notion of an ordinary, unmysterious and merely psychological
fact. But the more the theory emphasizes how commonplace and
126 The Slide into the Abyss
everyday it is to find that some people are stymied when asked to
imagine this or that, the less it can provide a satisfactory explana­
tion of how there can he pressure on a system of thought, pressure
to which it ought to respond. T h e projcctivist’s view is th a t when
we arc forced to abandon one o f our judgem ents, all that is going on
is that, because o f our im aginative limitations, we find ourselves
psychologically unable to m aintain a certain conjunction o f beliefs.
But if ibis is conceived as a mere psychological fact, it am ounts to
saying that there are no real constraints on the system, only
supposed constraints which we cannot imagine not to be there. And
the implication of this is that, if we could unshackle our im agina­
tions, we might have our heart’s desire.
Sometimes, of course, this is the right picture. Suppose I am
confronted with a green fruit which 1 unhesitatingly, through
limitation of im agination, take to he sour; 1 am ju s t unable to
conceive how a green fruit could be other than sour, but, all the
same, it is in fact sweet. My limitation here may well prevent me
getting w hat 1 want, namely a sweet fruit. But this structure can
hardly be extended to all cases. Suppose I desire th at this bare iloor
should he carpeted. C an wc make sense o f the idea that it is ju st
lim itation o f imagination which prevents me recognizing that this
bare door is, perhaps, at this very moment, carpeted? Is it merely a
limitation on my imagination that I take it that bareness and
carpctcdncss arc incompatible? T o put the point more generally, is
it ju st limitation on our im aginations th at when wc think that p we
find that wc cannot at the sam e time take it that not p? I f this were
so then indeed, for all wc know, we have our heart’s desire but
have, unfortunately, failed to realize it.
Let us return to the difficulty for the pragm atist posed by the
need to give an account of the nature and justification of judge­
m ents like W. W right’s response to the difficulty is to say that the
consistent pragm atist cannot allow that W is up for pragm atic
judgem ent. ‘Such statem ents adm it of totally convincing proof. We
m ust take seriously the idea o f proof as a theoretically uncontam i­
nated source of rational belief.’8 It is not entirely clear w hat W right
means by this. It is not anything as strong as th at wc m ust credit
ourselves with infallible access to a priori truths. T he thought
seems to be more that a proof supplies some prim a facie support to
 The Slide into the Abyss 127
a statem ent like W, support which although defeasible has its
strength independently of the empirical considerations which might
oppose it. W hatever the details, W right’s wording strongly suggests
that he thinks that the difficulty uncovered for the pragm atist is a
methodological or epistemological one, which requires correspond­
ingly some epistemological or methodological solution, a solution
which tells us about the kinds o f w arrant various statem ents may
have and when we arc entitled to rely on them.
Now there may be som ething defensible in the suggestion of
epistemological privilege, and we shall return to the m atter below.
But it seems misapplied as an answer to die current difficulty. In
the case outlined, W has been framed precisely so as to be the
judgem ent o f whether or not recalcitrance is now being manifested
in the system. According to the pragm atist programme that press­
ure which is to guide us in shaping our system (i.e. the appearance
or non-appearance of recalcitrance), is exerted or not exerted
precisely as we accept or do not accept W. Hence if we try to regard
the choice of W itself as to he guided in this way, we discover that
we have no independent source of pressure. T he problem then lor
the pragm atist is to give an account of w hat lie is doing or trying to
do in debating about W. W hat would it be for W to be right or
wrong? W hat arc the standards o f success here? These are the
conundrum s that face the pragm atist. He cannot give his usual
answer, that the standard of success is minimizing recalcitrance,
because to do so would lead to his losing touch with the idea of
constraint and hence to his moving off into the realm of wish
fulfilment.
T his problem of making sense of w hat counts as right or wrong in
accepting W is not solved by providing reason to regard W as
methodologically or cpistemologically privileged, even if such were
forthcoming. A methodological privilege might resolve the circu­
larity question of how actually to proceed in assessment, by by­
passing the question of w hether or not to assess W; but it leaves the
question o f w hat W ’s rightness consists in as obscure as ever.
A move th a t would resolve the difficulty here would be to
recognize th at the question ‘Is my system in trouble or not?’ has to
be taken in a realist spirit. And it is quite unclear to me that there is
any other move th a t would restore coherence to the story. A
128 The Slide into the Abyss
realistic treatm ent of the question seems to be w hat is implicit in
the idea of constraint. A constraint the operation o f which one can
choose at one’s convenience to recognize or disregard is no con­
straint at all. T o put it simply, there may be something right in the
picture of an indefinitely malleable system. But even given that, the
pragm atist has presented it to us as a system for doing something,
for coping with w hat impinges on us. (And if we extend the range of
things which the system is supposed to do for us beyond that of
enabling us to predict, the fundam ental idea of the system having
to deliver something remains.) C an we then abandon the idea that
there is a source of pressure on the system independent of our
choice? Only, I think, at the cost o f abandoning the attem pt to give
an account of something, however attenuated, which is still recog­
nizable as rationality and intellectual integrity. These virtues have
to do with facing up to the possibility of failure in thought. So they
can only have as much substance as we can give to the idea of there
being such success or failure.
Tf we feed this thought back into reflection on how one would
decide on some judgem ent like W we see th at it docs not enjoin us
to give any particular epistemological privilege to apparent proofs.
It certainly leaves open the possibility th a t a person should ration­
ally abandon a proof th at he or she finds psychologically compel­
ling upon non a priori grounds. But the thought does enjoin us to
hang on firmly to the idea that a proof is either valid or not. Wc will
thus need a distinction between a posteriori considerations which
are clues to the invalidity of a proof (even when the flaw has not yet
been detected) and considerations which rationally, but in fact
mistakenly, lead us to doubt the validity o f w hat is actually
correct.3 It may well be th a t it is something along these lines th a t
W right has in m ind w ith his notion o f a ‘theoretically uncontam i­
nated source of rational belief. T o hear his rem ark this way is to
take the lack of contam ination (of the a priori by the theoretical a
posteriori) as having to do with the content of the belief in question
rather than with any epistemological mode of investigating it. ‘This
proof is valid’ rem ains the sam e thought, has the same correctness
or incorrectness conditions, through all the varying epistemological
contexts in which it m ight occur. This is something we cannot say,
on the pragm atist story, about other judgem ents, because their
 The Slide into the Abyss 129
correctness consists only in forming part of the most convenient
total system.
The conclusion that I would like to stress at this point is that the
difficulty we have found with thoroughgoing pragm atism consti­
tutes a problem for the indeterm inacy of meaning. T he lirst section
of this chapter argued that indeterm inacy of meaning seriously
underm ined the defensihility of even minimal realism. At th at point
it seemed that the position reached m ight he acceptable if we could
make sense of thoroughgoing pragm atism - because pragm atism
seemed designed precisely as a theory which jettisons the ncccl for
realism. But the suggestion now is that even pragm atism requires
at its foundation some realist element. And it is quite unclear that,
if we accept pragm atism as springing from the indeterminacy of
meaning, we could possibly be entitled to make the repair to
pragm atism which would incorporate this essential strand. M od­
ified pragm atism , m ade coherent by inclusion of a realist treatm ent
of the question of w hether the system as a whole is coping or not,
presupposes that we are able to identify those thoughts of ours in
which we address the question of how the system as a whole is
faring; and it presupposes also that we are justified in treating them
in realist fashion. Given indeterm inacy of meaning, we have reason
to suppose that these conditions are not fulfilled. How are we to
identify the sentences which m ean ‘My system is in trouble* and
‘My system is not in trouble’? And why are we entitled to assume
th a t they will not succumb to the same argum ent which showed us,
in the first section of this chapter, that indeterminacy of meaning
left us unable to defend minimal realism with respect to ‘T he moon
is a flat disc’ and ‘T h e moon is spherical’?

6.3 W IT T G E N S T E IN AND PRA G M A TISM

In this section I would like to indicate a further difficulty for

thoroughgoing pragm atism which has to do with the idea o f ‘taking
for granted’. I shall argue th a t the category of the ‘taken for
granted’ needs to be divided into two and th at the items in one of
the subdivisions are ‘taken for granted’ in a more profound sense
130 The Slide into the Abyss
than those in the other. On this more profound sense a certain sort
o f justification for w hat is thus taken for granted is neither available
nor called for. W ittgenstein’s ideas in On Certainty have sometimes
been seen as im portantly sim ilar to the thoroughgoing pragm atism
discussed above. O ne suggestion of this section is that they differ
precisely in recognizing this contrast between kinds of 'taking for
granted’.
•The usual pragm atist story stresses that it is not committed to
the absurdity o f claiming that we can think w ithout using some
concepts and assumptions; but it emphasizes that any elements on
which we rely in one deliberation can be questioned and holistically
justified in another deliberation, provided that we take some other
appropriate things its (temporarily) fixed. Any p art of our boat can
be inspected and, if need be, reconstructed provided there is
enough o f it left to keep us afloat while we work. So far we have seen
no reason to question the viability of this story. T h e point em pha­
sized above was th a t in our assessment of W we had to think of the
question in a realist way; but th at says nothing directly about how
in practice we handle the assessment o f such a claim as W. And in a
particular ease it seems enormously plausible that w hat we would
actually do is w hat the pragm atist recommends, namely take for
granted various other m atters, both logical and non-logical, and
balance them as best we can to cotnc up with some verdict.
T he strength of this view is its openness to the idea of our making
radical changes in the way we look at things, changes which, from
the point o f view of many earlier users of the. system, arc extremely
hard to assimilate and may seem quite unintelligible. It is clear
that, psychologically speaking, any thinkers will have some very
deeply entrenched elements in their thought system. These will
manifest themselves in part in the form of unquestioning reliance
on certain principles of inference. It may seem to the thinkers that
the identity o f their concepts is bound up with recognizing these
principles. And they may also suppose that these concepts are
indispensable elements in any system. They will take it th at those
principles of inference that arc bound up with their concepts are
used in the assessment of other judgem ents; these principles are the
tools employed in gauging the degree of support th at other judge­
ments have; but they are not themselves supported in th at fashion.
 The Slide into the Abyss 131
I f they are questioned, their defenders are liable to fall back on
saying that they arc obvious, th a t they arc grasped intuitively or
something o f the kind.
T he pragm atist would like to unsettle this, and to deny that there
is any significant difference in the kind o f support different elements
have. It is a m atter of degree, of depth of entrcnchcdness, not of'
kind. He will point to cases (replacem ent of the ‘square of opposi­
tion1 view o f quanlificational inference by the Frcgcait one, the
laying out of coherent non-Euclidcan geometries, Einsteinian in­
sights into the notion o f sim ultaneity, etc.) where such unquestion­
ing reliance has been shown mistaken. When a particularly deeply
entrenched principle is overthrow n, we may, il'wc like, say that a
concept has been abandoned and a new otic adopted, rather than
say that an opinion about an existing subject has been changed.
But nothing of this im portance hangs on this distinction, says the
pragm atist.
It is possible to view such fundam ental changes as upheavals
which arc perhaps causally but not rationally explicable. T he
pragm atist however will not lie sym pathetic to this. His view is
that, with hindsight at least, wc can sec justification for change in
the increasing am ounts of recalcitrance in the old system and in the
way the new system eliminates or drastically decreases it.
It is at this point that the question 1 wish to focus upon appears.
Suppose wc agree with the pragm atist in seeing even the most
fundamental changes as instances of rational adjustm ent of the
system in view of the pressures on it. The abandonm ent of old
certainties is justified by the range of difficulties that have been
encountered in continuing to try to incorporate them. It may seem
to follow from this th a t we could, as things now stand, provide a
parallel justification for hanging on to fundamental features which
arc not now diagnosed as sources of difficulty. If certain occurrences
would provide justification for adjustm ent of some fundamental
principles (and so of the concepts th a t are thereby defined), does
not the fact of their non-occurrence provide our justification for
currently keeping the principles entrenched? It may be that wc find
difficulty in form ulating descriptions o f these non-occurrcnccs. But
it looks as if this is merely lack o f imagination on our part. Thus it
may seem that these negative facts provide positive support for the
132 The Slide into the Abyss
fundamental principles of the system. And if this were so then the
pragm atist would be right in seeing every element of the system as
justified in the same way, namely in the light o f certain considera­
tions which favour it rather than its rivals when we engage in the
enterprise of testing that particular plank o f the ship.
'Phis is a mistake. It is true that within the everyday parts o f our
scheme positive and negative facts stand on a p ar for justificatory
purposes. It is the fact, for example, that there are no fingerprints
on the glass which leads me to suppose that the burglar wore
gloves. But we cannot extrapolate from this case to the limiting one
of conceptual stability. It we are to imagine providing sim ilar
justification for fundam ental principles in terms of the hypothe­
sized non-occurrences we m ust suppose ourselves equipped with
some vocabulary for describing those 11011-occurrcnces. T h a t vo­
cabulary will in turn be individuated by its own associated funda­
mental inferential principles. And what is the pragm atist to say
about our (temporary) acceptance of these while we specify the
facts? Presumably the same, namely that it is justified by the
non-occurrence of the happenings which would show them to be
inappropriate elements o f a conceptual scheme. And so on, down
an infinite regress of vocabularies and schemes.
We might avoid the regress by supposing that the questioning of
schemes takes us round in a finite circle, or that we arrive at some
one ultimate vocabulary in which all basic evidential facts are to be
described. But to suppose either o f these is in efTect to return to
instrumentalism and thus to betray the most central tenet of
thoroughgoing pragm atism . But it is equally unacceptable (and not
only to the pragm atist) to suppose that any chinker, however
imaginative, is capable of working through the infinite series of
negative scenarios which would be required to provide positive
justification for staying with our current conceptual scheme.
T he conclusion is th a t it is not a mere superficial psychological
fact, for example about the limits o f people’s im aginations, that we
arc brought to a standstill by the request to justify certain of our
concepts and their attendant principles. T o suppose th at we were
never at such a standstill is to suppose that our minds responded
easily to every one of an infinite num ber of requests to skip on from
one conceptual scheme to another. This indefinite flexibility of
 The Slide into the Abyss 133
mind (which wc here imagine to be exercisable without even the
stim ulus o f actual puzzles and adverse circumstances which re­
quire to he dealt with) is as incompatible with recognition of our
finitudc as is the idea that we have already traversed the infinite
series.
T o say this is not to say that there is some one definite part o f our
conceptual scheme for which it is impossible for anyone to perform
the imaginative feat o f conceiving o f occurrences which would lead
us to w ant to change it. It is to say, rather, that we cannot make
sense oflhinkiug except as proceeding in accordance with at least
some principles and assum ptions which the thinker ‘takes for
granted’ in the sense of being nonplussed at the request to justify
them and being unable to m uster any positive evidence in their
support.
W c may, o f course, on identifying such principles in our thought,
point to such things as that they are fundam ental for us and that we
would hardly know how to carry on thinking at all if we were
denied the use of them. But this is not the sam e kind ofjustification
as th at which we are able to supply for other elements in the
system, elements for which we can think of alternatives and can
form ulate descriptions oí' the circumstances in which we would
adopt the alternatives. 'Falk of ‘coherence’ and ‘holism’ may dis­
guise this difference. Hut what we do on becoming aw are o f the
status of fundam ental assumptions is not to decide to keep them
fixed because they seem useful or to judge that are correct because
all the evidence speaks in favour o f them. It is not a move of the
same shape as choosing some basic elem ents o f a theory to hold
steady or take for granted for the time being. T h at kind o f taking for
granted, which goes with justification, requires ability to envisage
alternatives. But in this case there are none envisageable. The
conventionalist moves on necessities and fundamental principles,
moves which are congenial to the pragm atist, are in consequence
empty posturing. R ather naturalism (in one sense of that term) is
the outcome. W hat honest appraisal brings us to do is recognize a
fact about how we conduct our thought, and recognize also th at we
cannot but acquiesce in it.10
O ne final im portant point which needs to be made here is that
when we look at our own patterns o f thought, among the principles
134 The Slide into the Abyss
we find ensconced arc many which underpin minimal realist styles
of thought about all sorts of subject m atters. We find, for example,
that we arc equipped with the ideas of ranges of different and
incom patible properties (position, shape, colour, tem perature, m a­
terial composition . . .). These arc some of the basic tools in terms
of which we articulate our experience and their use brings minimal
realist practices along with it.
T he kinds of ideas I have been discussing in the past lew
paragraphs arc closely related to those which Wittgenstein raises
in On Certainly.11 Some have seen W ittgenstein as a thoroughgoing
pragm atist and there arc indeed im portant negative themes which
he pursues in common with the pragm atist, most notably opposi­
tion to m irroring realism. (The grounds of the criticism however
arc different, and we shall sec more of this in the next chapter.)
Both W ittgenstein and the pragm atist reject the idea that the world
dem ands certain concepts for its correct representation. But W itt­
genstein differs from the pragm atist in his view th at we come in the
end to recognition of facts about our practices and not to decisions.
He is also at pains to stress that this is not a blemish in our
thinking, the intrusion into it of unw arranted and dogmatic asser­
tion. For that criticism to be in order it would have to be possible
for us to do better, to cease to be so dogm atic. But there is no such
possibility.
T here may still however be worries about (so to speak) the
metaphysical status o f our fundam ental principles. Are they, for
example, the recognition of truths? If so can we use them to
underpin minimal realist practices not merely psychologically but
also metaphysically? W e shall return to this in section 9.4.

6.4 E M P IR IC IS M AND PL A T O N ISM

O ne m ight be struck by the fact that, in practice and by and large,

we have no difficulty in understanding each other, even in cases of
radical interprctaion. We ju st arc not faced with com peting inter­
pretations in the way in which the argum ents of chapter 5 sug­
gested th at we m ight be. There are indeed cases, for example those
explored in the novels of Henry Jam es, where we become aware
 The Slide into the Abyss 135
that our everyday psychological labels arc crude and that insistence
on getting a yes/no answ er in their terms (‘Does he love her or
not?’) would be misleading. But these situations are ones which
spur us on to develop subtler forms of description and in so doing
we have the sense th a t we represent more accurately w hat is
dctcrm inatcly there. F u rth er interesting questions may then arise
about exactly w hat the propriety is o f describing people’s beliefs,
feelings and aspirations in terms which they themselves might not
(or not w ithout a good deal of prom pting) acknowledge, and about
w hat further complications o f hum an interaction become possible
with the developing complexity of our self-descriptions. But again
none of the subtle reservations raised by these lines of thought have
anything to do witli the indeterm inacy worries generated by the
mosaic comparison. A nd in m any everyday cases we would find it
. absurd if such subtle reservations were entertained.
W hat use, however, can one make of the fact about the practical
dcterm inacy o f meaning ascription in arguing against the indeter­
minacy thesis? W hat I w ant to suggest now is that if one is a
m irroring realist it is highly unlikely th a t this thought can be of any
help in re-establishing our right to talk about the facts of what
people mean.
This is because the m irroring realist will, in all probability, he a
modal realist of a Platonist sort as well as a realist about the sense
data, physical objects or w hatever else he lakes Lhe world to consist
of. T h at is to say th at he is committed to the view that states of
affairs arc possible or necessary independently of our views about
w hether they arc possible or necessary, and independently also of
w hether we have so much as thought of them. So the mirroring
realist takes it that, independent of him and his thoughts, either
there is or there is not an alternative equally defensible translation,
i.e. a possible way o f m atching sentences which observes all the
appropriate constraints. If there is such another scheme then its
mere existence, even if entirely unsuspected by us and beyond the
reach of our im aginations, will nevertheless do its malign work of
unravelling our entitlem ent to realism about meaning (and hence
about anything else).
T he claim th a t the m irroring realist is also likely to be a Platonist
modal realist may seem odd. T h a t it is coherent to combine
136 The Slide into the Abyss
mirroring realism with such m odal realism (as in a rationalist style
of system) seems plausible. But should it not also be feasible to
separate the two? And is this not w hat empiricism (the form of
mirroring realism we are prim arily considering) actually does? T he
em piricist is one who supposes th a t we acquire concepts and
knowledge from confrontation in sense perception w ith the facts. It
seems immensely plausible to suppose th at we cannot have that
kind of confrontation with supposed modal facts. Hence the em piri­
cist turns to conventionalism, or so it is supposed, to explain our
use of modal concepts.
'Phis story is, o f course, partly right from a historical point of
view. T he logical positivists did indeed turn to what they called
conventionalism, but their resort was half-hearted. T he conven­
tionalism advanced, for example by A. }. Ayer, is modified and not
radical conventionalism.1'2 O n the latter every individual necessary
truth is separately adopted. T here are not logical discoveries, no
logical compulsions, revealed to us on reflection. But on the former
wc make certain decisions, e.g. on w hat rules we shall follow in
using the words ‘all’, ‘not’ etc., and in consequence find ourselves
committed to certain other uses of these words which we had not
explicitly anticipated; we discover, in other words, that doing what
the initial rules prescribe necessarily involves doing w hat certain
other, derivative, rules dem and.
But, as Dummett remarks, ‘this account is entirely superficial
and throws away all the advantages o f conventionalism’.13 His
point is that the modified conventionalism leaves entirely unex­
plained the nature of the truth: ‘Necessarily, if you follow this rule,
you will follow that other rule also.’ And it leaves unexplained
(from an empiricist point of view) our access to this fact. These
matters are conceived, by the modified conventionalist, in a modal
realist manner.
But is this just an accident of w hat these particular empiricists
said? Gan we make sense of an em piricist style m irroring realism
which in a thoroughgoing way repudiates any taint o f modal
realism, but allows talk of necessity as an upshot of conventions? I
do not think that we can.1-1 Let us consider how such a theory
would have to be articulated.
Such a view will say that there are no facts about how things
 The Slide into the Abyss 137
m ust be but there are facts about how they are. W hat then are we
doing when we use modal terms? We are expressing decisions. The
decisions will be, for example, to hang on to any statem ent of a
certain form come w hat may, or to accept a statem ent of one type
as unimpeachable w arrant for making a statem ent o f some other
type.
Hut if there are no modal facts to underpin these decisions, how
are they to be defended against the charge that carrying them out
might lead us astray? If the realist thinks, as he does, that there are
facts and we have epistcmological access to them apart from these
decisions, then the decisions can do nothing except introduce some
short cuts and rules o f thum b into the process o f acquiring judge­
ments. There may be room for tightening up vagueness by conven­
tion. But leaving th a t on one side, c ith e r the n a tu re o f the
ju d g e m e n ts alread y g u a ra n tee s c ertain tru lh -v a ln e links - in
which case modal realism is conceded —or it does not - in which
case deciding to treat some proposed links as absolutely lirm is
pointless foolhardiness. W hat this suggests is that radical conven­
tionalism is only fully at home in a pragm atist selling, where all of
our judgements arc open to choice and so can be jerrym andered to
lit the chosen necessities. Radical conventionalism thus makes an
uncomfortable ally for empiricism, conceived in a robustly realist
way.
W hat then are we to make of the, at first sight plausible, claim
that experience teaches us only what is the case and cannot yield
the concept of w hat must be the case? T h e conclusion to which the
argum ent points is that this does not stand up to serious pressure,
when empiricism is allied with the m irroring realist metaphysics
and when perceptual confrontations are conceived of as those events
in which the nature o f the world impresses itself upon us. There are
indeed certain sorts of necessity (e.g. the synthetic a priori, or
physical necessity) for which the empiricist claim can be defended.
B ut the idea that no notion whatsoever of any necessity is derived
from experience is in tension with the em pricist idea that it is facts
which are revealed to us in confrontations. T he idea of a fact
involves the idea of a property being instantiated. B ut how are we,
in any realist and non-pragm atist spirit, to conceive of a property
which does not, of its nature, enter into relations of exclusion or
138 The Slide into the Abyss
inclusion with other properties? Such relations between properties
reveal themselves as necessities in the world of instantiations.
M oreover we see that the empiricist docs have an account to oiTer of
how we come to know them: awareness of them is impressed upon
us as a component of that awareness o f fact which experience gives
us. Experience shows us the univcrsals as much as the particulars.
C an the upright empiricist l)r Jckyll reject this unwelcome
conclusion th at he has all along been a Platonist M r Hyde as well?
O ne way of doing so is to play down the conceptual and to play up
the sensory element in empiricism. But on this tack empiricism loses
plausibility as an account of thinking and loses also its attractive
role in the m irroring realist story o f how we come to know the true
nature of the world. So is there perhaps a coherent view which
combines the empiricist and the realist elements but which is
neither conventionalist nor modal realist? Could such a view not
simply repudiate the idea of necessity altogether?
A move to sketch such a theory would be to say that all positive
features revealed to us in experience arc simple and m utually
compatible. Hence all relations of exclusion or involvement among
them arc discovered a posteriori. O ne might enquire w hat these
features are. ft is rem arkably difficulty to specify any whicfi fit the
bill. But waiving this objection we may allow lliat the move does
indeed do away with a good deal of Platonist m achinery. However,
it docs not quite satisfy the requirem ents o f disposing of all modal
notions. W hat is to be said, for example, about negation?
T here arc two lines the empiricist could take here. lie could say
that ‘not F’ and *F’ arc independent simple characters; the presence
o f the phonem e lF’ in ‘not F’ is just a coincidence. The lack of
co-occurrence of F and not F is a very well supported empirical
generalization and not a logical tru th a t all. I f the empiricist says
this then, for him, no rem ark -is incom patible with any other. But
w hat sense can we now make o f the idea th at what is going on is
description o f how things arc in an independent and determ inate
world? C ertainly this empiricist position no longer fits our require­
ments even for minimal realism.
O n tlic other hand the empiricist could say, much more plau­
sibly, th at in grasping w hat it is lor som ething to be F (and even if
one docs not grasp anything about any other property G or about
 The Slide into the Abyss 139
relations of exclusion or inclusion with G) one docs grasp that there
is such a thing as not being F; and one also grasps that being F and
not being F are incompatible. But this thought supplies us with the
modal notion o f incom patibility, th a t which is necessarily not so.
From this, together with the view that no simple property is
incompatible with any other simple property or its negation (which
wc conceded above in talking o f compatibility) wc can generate a
rich and determ inate modal structure. (It is, or course, closely akin
to that of the Tractatus).
T he upshot then is th a t the link between mirroring realism and
modal realism seems to stand. And in consequence the mirroring
realist is in no position to resist the indeterminacy of meaning
argum ents presented in chapter 5. From our current perspective we
need to emphasize how two factors work together to generate the
indeterm inacy conclusion. T he first is the one we have spent most
time expounding, namely th a t meaning ascription is justified by
holistic considerations. T he second is equally im portant. It is the
assumption th at dctcrm inatcly either there is or there is not a
holistically acceptable alternative interpretation of the utterances
of a given person. T his assum ption is the one. to which the current
version of mirroring and modal realism commits us. If we make this
assumption then the question of w hether we arc entitled to think of
our meanings as determ inate, turns into the question of whether wc
can prove that there are no acceptable alternative interpretations.
But this, as I emphasized in section 6.1, is a daunting task winch
we have no idea how to carry out. Hence our seeming loss of right
to be confident that we do dctcrm inatcly mean something in our
remarks and th at we can know w hat we mean. The actual unavail­
ability of alternative translations and the practical workability
(indeed indispcnsability) of a realist practice of meaning ascription
is now represented as a mere psychological or sociological observa­
tion. Given the idea o f a totality of metaphysical possibilities,
cpistemologically independent of us, and the envisaged methodol­
ogy o f meaning ascription, the actual success and importance of
meaning talk all go for nothing, metaphysically speaking. The next
step (as 1 argued in section 6.1) is the loss of a right to take a
coherent realist stance about anything and the emergence of thor­
oughgoing pragm atism as the only option. But (as 1 argued in
14-0 The Slide into the Abyss
section 6.2) this in turn proves unattractive. We need an account of
our thinking which allows us at least those repairs to pragm atism
which make it coherent. B ut we cannot effect these repairs when we
still recognize the force of the argum ents for indeterm inacy of
meaning.
C an we extract ourselves from this imbroglio? T here are many
routes out from this uncomfortable position, as many as there are
disputable assumptions that I have m ade on the way into it —and
they are not a few. One route would be to try to defend thorough­
going pragmatism against the charge o f incoherence, or to deny
that importation of realist elements is the only way of making it
coherent. Another would be to try to uncouple the linkage I have
endeavoured to set up between realism about the objects o f our
thoughts and realism about the content o f those thoughts. T he idea
I offered (at the start of this chapter) was that there is bound to be
tension in a philosophical position which endeavours to m aintain
any robust form of realism (i.e. som ething more than mere Qui-
ncan disquotationalism) about the natural world while denying
correspondingly robust realism about the sem antic properties of
our thoughts. I do not claim to have proved this in full generality
however, only to have m ade it plausible by examining some cases.
So one might find a way out by showing the combination to be,
after all, defensible.
Even supposing that these moves are not attractive, many paths
remain open. If we start with a m irroring realist view of the subject
m atter of natural science and wish to preserve it, then our concern
must be to give an account of m ind and meaning on which they
have reality of this kind. O ne could achieve this by denying the
claims about holistic constraints on meaning ascription. And
another move, even more attractive perhaps, would be to adm it
them but to reject the argum ents offered in section 5.4 for saying
that these holistic considerations could not be supplem ented by
causal, determinacy restoring, facts. T o pursue these program mes
is to endeavour (by talk of functionalism in philosophy o f mind,
teleology and natural selection, etc.) to produce a naturalized
theory of mind. The upshot of success in such an enterprise would
be the defensibility of m irroring realism all across the board, for
natural and semantic facts alike. And given the extremely sketchy
 The Slide into the Abyss 141
and inconclusive nature of the argum ents advanced in section 5.4 1
am in no position to deny the viability of these philosophical
programmes.
But w hat if wc are, all the same, dubious o f the chances of
success by this road? W hat if it seems likely to us that mind and
m atter, subject and object, rationality and causality, will not lie
down together in the required way? And w hat if we combine this
scepticism with a com m itm ent to the reality of the natural world
and also commitment to the linkage o f realism about objects of
thought with realism about content? Arc we now stuck in an
impasse? I wish to suggest th at there is a crack through which we
can slip - namely the denial of m irroring realism. My suggestion
above was th at m irroring realism, via its generation of commitment
to the totality of epistemologically independent metaphysical possi­
bilities, is an im portant underpinning clem ent in the move from
remarks about the holistic nature of meaning to indeterminacy
claims. If wc could somehow rid ourselves o f the need to take
seriously these shadowy, unimaginable, but (it seems) theoretically
possible other interpretations, then there m ight be no bar to
combining the holistic story about m eaning with the view that our
remarks, by and large, detcrm inately mean what we take them to
mean.
I have concentrated here on the modal and metaphysical el­
ements in the m irroring realist position and our need to be rid of
them. It is worth rem arking that were we to see how to satisfy that
need, certain epistcmological consequences would follow. If we
can, so to speak, ignore the possibility o f an arrangem ent or
interpretation other than the one that we normally take for granted,
that is tantam ount to allowing us to be confronted epistemologi­
cally by an utterance-i/Kfl-element-of-such-and-sueh-a-pattern rather
than by mere uninterpreted noise.11 'I'o put this the other way round,
the idea of the need to take the other arrangement seriously brings
with it the idea th at wc m ust separate the secure datum that so and
so uninterpreted utterance occurs from the less secure hypothesis
that it can be placed in such and such a pattern.
This epistemological presupposition of the indeterminacy argu­
ments (one which explicitly acknowledged by D avidson)16 suggests
another respect in which m irroring realism , in its naturalist
14-2 The Slide into the Abyss
version, is powerfully influencing the developm ent of the ideas.
Such naturalized m irroring realism brings with it em piricist sym­
pathies, and preconceptions about w hat kind ofl'act could impress
itself upon us or reveal itself to us. M eanings, when regarded as
residing in patterns, clearly arc not the kind of thing which could
do the required impressing. Hence they have to be seen as inferred
or constructed.
A nother corollary of the shift in outlook sketched above would be
a need to rcfmc and recast all the remarks wc have made about the
methodology o f meaning ascription. We would have to distinguish
views on the kinds of consideration th a t judgements of meaning arc
answ erable to if doubted, and reflections on how, in their holistic
com m itm ents they dilfcr from those o f natural science (about which
the discussions of chapter 5 are largely right) from the idea that wc
go about interpreting Our lirst language by assembling d ata on
uninterprctcd utterances (on which chapter 5 is wrong). All of
these themes could well have been m ore fully represented in an
alternative approach to the outlook wc arc about to explore.
Let us return to our main theme: How are wc to be realists
w ithout mirroring? How could ignoring these possibilities be other
than cpistcmological irresponsibility? It is to the W ittgcnsteinian
treatm ent of these questions that we now turn.
 7
The Dissolving Mirror

7.1 W IT T G E N S T E IN ’S H O S T IL IT Y T O M IR R O R IN G
R EA LISM

M irroring realism is now the focus of our discussion. I suggested in

chapters 3 and 4 how attraction to it, in its empiricist form, might
issue in a Q uincan version o f m eaning scepticism. It was perhaps
not surprising to find sceptical theses about theoretical meaning
issuing from premises which themselves explicitly concern our
limited ability to acquire concepts and knowledge. T he argum ent
of chapter 5 is of interest in p art because it starts from something so
much less controversial. It docs not involve, explicitly, mirroring
•realism and even less does it involve a particular empiricist form of
it. It seems to invoke only some prc-theorctical, commonscnsical,
idea of realism about the ordinary material world, together with
some ideas about the contrast between concepts used for describing
it and concepts used for describing the psychological and semantic.
But the idea floated at the end of the last chapter was that
m irroring realism, appearances notwithstanding, is one of the
crucial assum ptions in generating the indeterminacy conclusion of
chapter 5 and hence the difficulty which that precipitated us into in
chapter 6. U ndoubtedly other assumptions are also im portant in
enabling us to derive the conclusions. Hence there are many
possible routes out o f that difficulty to be explored. Philosophical
reflection can resemble a giant maze with decisions about promis­
ing paths needing to be m ade all the time. The interest of the path
we shall enter is, I shall try to suggest, that it resembles the one
followed by W ittgenstein.
W hat evidence have we th a t W ittgenstein was hostile to m irror­
ing realism? First there are some rem arks which are most naturally
144 The Dissolving Mirror
read as explicit statements to that effect: secondly, we can detect in
his writing some powerful argum ents directed against the intelligi­
bility of the idea; and thirdly, if we take rejection of this form of
realism as the end of a thread and follow it through, we lind the
whole tenor of W ittgenstein’s thought becoming more com prehen­
sible and many difficult itleas fitting into place.
T he first kind of consideration will not get us far w ithout the
others. But let us rem em ber the following:

I f anyone believes that certain concepts are absolutely the cor­

rect ones and that having different ones would mean not realiz­
ing something that we realize - then let him imagine certain
very general facts of nature to be different from w hat we are used
to and the formation o f concepts different from the usual ones
will become intelligible to him .1
O ne is tempted to justify rules o f gram m ar by sentences like ‘But
there really are four prim ary colours’. And the saying that the
rules of gram m ar are arbitrary is directed against the possibility
of this justification, which is constructed on the model o f justify­
ing a sentence by pointing to w hat verifies it.2
Do I want to say, then, that certain facts are favourable to the
formation of certain concepts; or again unfavourable? And does
experience teach us this? It is a fact of experience that human
beings alter their concepts, exchange them for others when they
learn new facts, when in this way w hat was formerly im portant
to them becomes unim portant and vice versa.3
We have a colour system as we have a num ber system. Do the
systems reside in our nature or in the n atu re of things? How are
we to put it? Not in the nature o f num bers or colours.
T hen there is something arbitrary about the system? Yes and
No. It is akin both to what is arbitrary and to w hat is not
arbitrary.'1
Yes, but has nature nothing to say here? Indeed she has - but
she makes herself audible in another way.
‘You’ll surely run up against existence and non-existence
somewhere.’ But that means against facts, not concepts.5
 The Dissolving Mirror 145
I want to say: an education quite different from ours might also
be the foundation for quite different concepts.
For here life would run on differently. - W hat interests us
would not interest them. Here different.concepts would no longer
be unimaginable. In fact this is the only way in which essentially
different concepts are imaginable.0

W hat is the vision o f people, thought and language iron» which

these remarks spring? Let us consider some of the themes treated in
the Philosophical Investigations up to section 242. T he later elements
of this sequence of sections, from about 138 onwards, are some­
times known as the ‘rule following considerations’ and have been
much discussed. A certain consensus (which I do not at all desire to
undermine) has emerged upon at least one of the things which is
brought to our attention here. T he point in question is a negative
one and has been expounded very clearly and elegantly by K ripke.7
Let me sum m arize these ideas briefly, for the ground is fairly well
trodden.
Suppose we consider the questions ‘W hat does this person
think?’, or suppose we are faced with the challenge of defending the
attribution of some particular thought to him or her. In either case
we arc inclined to look for something th at having a certain thought
consists in. We try to find some item the occurrence of which in or
to the person, the existing o f which in some relation to the person,
fixes it that the person m ust be having a certain thought (express­
ing a certain m eaning). T he kinds of items wc look for, although
they may be very various - sequences of marks, noises, pictures
(which are external to the person), images, formulae before the
m ind’s eye and feelings (which are ‘internal’) —all have something
in common. This is that we can conceive of their existence as
consisting in the exhibition of certain determ inate, intrinsic, non-
representational properties over some area of space-tim e (or men­
tal quasi space-tim e). T here is, so to speak, clearly a ‘w hat it is
like’ for these things to occur and it is this which makes us confident
that there is such a thing as their occurring (or not occurring)
which we could, in epistemologically favourable circumstances,
establish. It is also the case that in virtue of their possession of such
properties the things in question are fitted to enter into causal or
146 The Dissolving Mirror
quasi causal explanatory relations. O u r looking for these kinds o f
item s in answer to the original questions about thought is bound up
with the dom inance of scientific modes of understanding and our
disposition to assume th at we can find an account of thinking
closely analogous to the kinds o f account which science has given us
o f the constitution and workings of the natural world.
O u r search, however, is doomed to failure. T he sort o f thing
which wc arc determ ined to find always turns out on closer inspec­
tion to be a vehicle of meaning rath er than m eaning itself. M ore­
over in these cases it is not a vehicle suited to carry only one
m eaning but one capable o f varying interpretations. W hatever kind
of item wc consider - a diagram , image, formula, picture, feeling,
etc. — wc always find that we can suppose that it occurs in
w hatever relation to a person is appropriate to its nature but that
the person carries on using it, responding to it or whatever, in a
way which shows that he or she did not mean by it th at thought
which wc had intended it to pin down. Something other is required
for m eaning than the mere presence to or in the subject of the item
itself. T his is most obviously true o f the external items, bu t W itt­
genstein takes us through a variety o f inner items and shows us that
it is true of them too.
But it would be wrong, 1 suggest, to sec the implications of these
ideas as bearing only on the question of how or how not to conceive
o f ‘m eaning’ ‘understanding’ or ‘thinking’ as states of persons. T he
m oral is not merely, to put it crudely, th at ‘understanding is not
having an image before the m ind’s eye' or something of th a t shape.
T h e remarks about formulae, images and the like follow' on from
discussion in the earlier sections ot Investigations (1—138 roughly) o f
os tensive teaching, of the role o f examples and o f the phrase ‘and so
o n ’ in explanation of meaning. T hey thus follow on from discussion
o f the need for some response from the pupil which the teacher
should not suppose to be m ade inevitable by the circumstances o f
teaching and the examples he presents. T his phase of W ittgen-
stcins’s investigation bears upon m irroring realism and wc can,
with a backwards look, see it as a version o f the later thought.
W hat W ittgenstein’s considerations show is that we cannot make
sense of the idea that being confronted with some set of objects or
chunk of the world should force possession o f a certain concept
 The Dissolving Mirror 147
upon a person. W c clo indeed develop our concepts and make our
judgem ents as the outcome o f such confrontations. But sets of items
we can be confronted with will clearly, in many cases, fall under
many different concepts. A set of objects cannot of itself determine
that one rather than another similarity shall be noticed and so
cannot force extrapolation o f the set one way rather than another.
This may look at first sight a very different type of consideration
from the earlier ones about the insulliciency of formulae, mental
imagery and the like to constitute meaning, but it is the same
fundam ental thought at work. If the items themselves were capable
of forcing an extrapolation then, in the extreme case, I could bring
them along to be my formula or image, which of itself carried my
meaning.
T here is som ething initially strange about the idea of the world,
or part o f the world or a set o f objects in the world, (other than the
small, easily recognized subset of items which make natural lan­
guages) constituting some sign or formula which carries meaning.
But in an only slightly extended usage this clearly is a defensible
way of looking a t things. T he not-obviously-linguistic world (or
some part o f it) can be the sort of thing wc bring attention to when
we try to determ ine or convey to one person w hat another person
thinks. ‘W hat is she thinking?’ someone asks. 1 might in response
ju st nod in the direction of some happening in the environment. O r
I m ight say T ic thinks this one is like those’, pointing to soine
objects. Ostensión cannot but play an im portant part in concept
acquisition and hence it can continue to serve the role of concept
identification w ithin the practice o f a community which has already
learnt the language. T h e points that W ittgenstein stresses about
the indispensable role o f the nature of the learner in acquisition
reappear in the other context as points like the ones about the
possibility of alternative employments o f formulae, images and the
like.
Now m irroring realism is, in effect, the claim th at the world, if
properly attended to, docs determ ine one and only one (set of)
thought(s). I f there were a way that things arc in themselves then
there would be such a thing as thinking that things were that way.
T h a t thought would be, so to speak, the correct reading of the item
which is the world. We can see W ittgenstein, then, as saying that
148 The Dissolving Mirror
the world does not in itself carry any message. It does not as it were
say ‘I ’m like this’. It cannot carry a message, any more than
formulae, images or pictures can.
T he view put this baldly m ight be thought to have sceptical
implications. It might be thought, for example, to be committed to
the idea that the world is an indeterm inate mush upon which we
impose form. It might be thought th a t even minimal realism could
not be preserved. This is a m atter to which we shall return at
several points in the chapters that follow.
Relatedly, but more subtly. it might be thought that the view I
am ascribing to W ittgenstein involved clear cut com m itm ent to the
idea that there are alternative ways o f ‘reading’ the world and so
alternative conceptual schemes. W e have, after all, both in the
considerations about formulae and in the considerations about the
role of samples in learning, stressed the seemingly real possibility o f
alternative responses in the user or learner. T h u s some version o f
the realism -threatening relativism which we m entioned in section
2.4 appears to be looming.
It is im portant however to make a distinction here and to attend
more exactly to w hat the consideration of formulae and samples
shows. We may in particular cases draw attention to the fact that
formulae or samples can actually carry diiferent meanings. But
what this makes vivid is that the formula, or set o f samples, is itself
inert; it has no life as a sign unless it is read, i.e. used in a certain
way. Reflection on these cases leads to the thought that this is a
quite general m atter. W hat someone is thinking is never a question
of the existence ofsom e item that he or she stands in some relation
to, but always a m atter of w hat life that person is living, and hence
of a role that the item plays in the life. Concepts are explicable and
defensible only as elements of a whole which m ust also include
projects, actions etc. This goes for the placem ent of a creature in
the item which is the. world ju st as m uch as for the placem ent o f the
creature vis-à-vis a formula or set o f samples. Stress on the alterna­
tive readings is merely a heuristic device, a step towards this
realization, not something which itself is to be unthinkingly extra­
polated to ail cases.
So, given this interlocking of concepts, lives and interest as a
starting point, the question ‘Is there a way th at the world is in
 The Dissolving Mirror 149
itself?’ comes out as the question ‘Is there only one way th at a
person could live an intelligible life in which he or she dealt
rationally with this world?’ O r, better, it comes out as ‘Is there a
large common clement in all such lives? And m ust fully rational
lives converge upon one account of the nature of the world?’ To
claim that m irroring realism is not a good m etaphysical picture is
not to claim that we see that the answers to these questions is ‘N o’ —
i.e. the claim is not that of relativism. R ather it is to point out first
that one is entitled to m irroring realism only if one can make out a
clear case for a ‘Yes’ answer, and it is to suggest, secondly, that we
cannot make out such a case because the question docs not have the
kind o f sense which would enable us to give a clear cut answer. We
may come to think that we cannot make much of the idea o f a ‘form
oflife’ dilferent from our own without thereby coming to think that
we see that ours is the only possible one. We shall return to these
m atters in chapter 9.

7.2 E X PL A N A T IO N AND T H E A B SO LU TE
C O N C E P T IO N

T he line of thought concerning the interdependence of concepts

and interests which is opened up in the preceding section requires
much more exploration. But before we turn to that directly we can
get more light upon some aspects of the problem by discussing a
recent defence of the central idea of mirroring realism. T his defence
is offered by B. A. O. W illiams in his attem pt to clarify a notion he
calls ‘the absolute conception’. W illiams’ claim is th a t we can make
sense of the idea that some o f our beliefs represent the world ‘in a
way to the m aximum degree independent of o ur perspective and its
peculiarities’.8 And he makes another claim also, namely th at this
conception is bound up w ith our idea of knowledge in such a way
that commitment to the possibility o f knowledge is com m itm ent to
the intelligibility of the absolute conception? This is an interesting
claim from our point o f view because w hat it proposes is, in effect, a
move directly from minimal realism to mirroring realism; it insists
that the latter is implicit in the former.
W e started out by considering m irroring realism prim arily in the
150 The Dissolving Mirror
form of sense datum empiricism and I suggested (in section 2.3)
th at this is the form such realism is likely to take when we dem and
some onc-by-onc certifications of concepts anti judgem ents. But the
form of realism proposed by W illiams cuts free from all the sensory
and atom ist limitations o f th a t scheme by w ithdrawing any de­
m and for onc-by-onc endorsem ent. For Williams, to certify a
concept as rcalily-rcprcscnling, we look to the explanatory role o f
judgem ents employing it. Here arc some quotations which sum ­
m arize the view:

Wc can select among our beliefs and features of o u r world

picture some that wc can reasonably claim to represent the
world in a way to the m aximum degree independent o f our
perspective and its peculiarities. T h e resultant picture o f things,
if wc can carry through this task, can be called the ‘absolute
conception* of the world. In terms o f th a t conception wc may
hope to explain the possibility of the conception itself and also
the possibility of other, pcrspectival representation.10

And at another place wc find this:

T h a i possibility, i.e. of forming the absolute conception, as I

have explained it, depends heavily on the notion of explanation.
T he substance of the absolute conception . . . lies in the idea that
it could uou-vacuously explain how it itself and the various
perspcctival views of the world are possible. It is an im portant
feature of modern science th at it contributes to explaining how
creatures with our origins and characteristics can understand a
world with properties th at this same science ascribes to the
world. T he achievements of evolutionary biology and the neuro­
logical sciences arc substantive in these respects."

T h e advantages of this view over the more crudely empiricist

versions of m irroring realism are considerable. T he ‘absolute con­
ception’ is seen as something that we strive towards, not som ething
handed to us on a plate. T he striving towards it, and the m easure of
w hether we have attained it, take the form of reflecting on what
explains w hat within our view o f ourselves and the world.
An objectionable feature of the older empiricism is the concep­
 The Dissolving Mirror 151
tion o f thought as something passive, to which the knowing subject
makes no contribution. T he them e o f some VVittgensteinian critic­
ism o f this (as we have seen in the first section of this chapter) is
firstly th at the nature of the subject m ust make some contribution
to the content of the thoughts th a t spring from a given series of
confrontations, and, secondly, that the subject’s contribution needs
to be thought of as, in some sense, bound up with activity.
It m ay seem that W illiams’ version o f realism can accommodate
w hat is correct in these responses. His view is that we start out with
some set of representations o f the world which arc (or may well be)
‘pcrspectival’. T h a t is to say that they involve concepts which we
exercise only because o f our peculiar interests, sensory equipment
etc. W illiams does not thus suppose th a t we can make sense of our
m ind being a tabula rasa which from the start forms uncontam inatcd
judgem ents. How thought gets under way, so to speak, is only
comprehensible by bringing in a large dose of activity, interest and
the like. B ut we also, W illiams implies, come to develop an interest
in representing things ‘in a way to the maximal degree indepen­
dent o f our perspective and its peculiarities’. And in pursuit of this
interest we reilect upon our representations, perhaps come to have
some new ones, and so arrive at som ething we can dcfensibly take
to be (at least nearer to) the absolute conception.
I t looks as though we do not a t any stage in description of this
developm ent have to make play with the suspect notions o f a
self-extrapolating set o f objects, passive thinkers or signs which can
be correctly read only one way. So we may seem to have driven a
wedge between the ideas about m eaning which W ittgenstein has
shown to be nonsensical and the defensible idea of there being a
way th a t things are in themselves.
But on closer inspection it is far from clear that W illiams’ notion
should satisfy the W itlgcnsteinian. It is an assumption of W illiams’
approach that there is such a thing as an interest in how things arc
in themselves which makes intelligible the formation of one set of
concepts, those of the absolute conception, rather than another and
merely ‘perspectivaP set. But the VVittgensteinian may well re­
spond that the supposed ‘interest’ is purely formal and cannot
contribute to commending one extrapolation rather than another.
M ere interest in ‘the tru th ’ w ithout any specification of subject
152 The Dissolving Mirror
m atter is of’doubtful intelligibility as a value.12 And as soon as we
put in some specific subject m atter - w hether it be die origin of the
universe or the growth of my neighbour’s m arrow —which makes
the curiosity comprehensible, we shall find that we are employing
particular concepts, the role and im portance of which in our
thought the W ittgcnsteinian will wish to relate back to our inter­
ests, practices etc. T h a t the concepts of fundam ental physics have
some interesting differences from those of everyday hum an and
social life is undoubted. But that the interest which drives us to
construct the concepts of physics is that in ‘pure tru th ’, rather than
an interest itself rooted in that fife, is not an unquestionable datum .
A further worry could be put this way. If a certain idea makes no
sense, namely the idea of being a passive tabula rasa upon which the
unambiguous sign im prints a judgem ent, then it makes no sense
either to suppose that each o f us should strive to become a tabula
rasa by selecting am ong our beliefs so as to remove the contam ina­
tion or limitation of our perspectives and interests.
In order to get any further we need to look more closely a t w hat
Williams means by ‘explanation’. C an be supply an account of a
sort of explanation which would validate the claim o f a notion to be
p art of the ‘absolute conception’ and lay to rest the W ittgensteinian
worries?
His references to evolutionary biology and the neurological
sciences suggest on first reading that it is some substantive account
of what it is to think this or th at (something o f the kind which
functionalist accounts in the philosophy of mind aim to supply)
which Williams has in m ind.13 T he idea could be developed like
this. We find that, as science progresses, we are able, at last, to give
an account of what it is to have thoughts, o f w hat thinking consists
in. We find that the explanation calls upon only some of the
concepts which previously we unreflectively used; the others turn
out unnecessary and the supposed ‘thoughts’ in which they were
exercised are revealed by science to be not central cases of intellec­
tual actions or states but to be characterizable in some other way.
For example, one m ight claim that thinking th a t something is
square is explained as a state caused by confrontation with square
things and apt itself to lead to behaviour appropriate for dealing
with square things. ‘Square’ is thus vindicated as p a rt of the
 The Dissolving Mirror 153
absolute conception, since m ention of squareness is invoked in the
account of w hat it is to think about it. O n the other hand thinking
that something is wicked is explained as having a certain feeling
and producing certain behaviour on confrontation with certain
things - where all of the items in question are characicrizablc
w ithout mention of wickedness. O u r propensity to have the reac­
tion to the items is itself explained (perhaps) as a result of natural
selection an d /o r social conditioning. ‘W ickedness’ is thus thrown
out o f the absolute conception.
But this bold and simple strategy is not what Williams has in
mind, as is clear from a variety of considerations. O ne is that he
explicitly expresses sym pathy lor Q uinean and Davidsonian
themes (akin to those discussed in chapter 5) which he takes to be
incompatible with any ‘reductive’ and functionalist style of account
such as that proposed above.14
A nother reason for thinking that the above misrepresents Wil­
liams’ strategy is that he acknowledges the force of the ‘shapeless­
ness’ objection (advanced, for example, by McDowell) against
projectivist accounts of ethical notions.15 Williams thus concedes
th at the items grouped together under one of the concepts he wishes
to label ‘pcrspeclival’ may not have any neat set o f ‘absolute’
features in common. W hat is common, what makes it intelligible to
group them together, can, he adm its, only be grasped by someone
who sees the world from the relevant point of view (or at least has
sym pathetic insight into w hat that would be like). So what is called
for (and all that is available) to attain understanding o f such
perspectival concepts is insight into the possibility of these various
points o f view. But this insight, although according to W illiams
facilitated by possession of the relevant absolute concepts, will not
necessarily be the kind o f thing which enables us to carry through
the imagined kind of reductive manoeuvres.
And a third consideration, of a more structural but still weighty
kind, is that W illiams ju s t does not attem pt to provide himself, or to
cite others as providing, the sort of substantive philosophy of mind
which would be required to m ade the above strategy work. This
would be a serious oversight were his ‘explanations’ designed to
have the form suggested. But they are not. We can get a much more
coherent and interesting reading by pursuing another line.
154 The Dissolving M inor
O n this second strategy, the focus is not on the act of thinking
but rather upon the contents of thoughts and how they are seen as
knowledge advances. The idea is th at such advances reveal certain
notions to be ‘merely pcrspcctival’ and hcncc as not to be taken as
representing how things are in themselves. But these same ad­
vances presuppose, and have to be seen by us as presupposing, that
we have at least the idea o f otiicr notions which resist any such
relegation. Let me quote the crucial and exceedingly condensed
argum ent in which the vital moves arc made. (I have inserted
num bers at certain points, for case o f later reference.)

(1) Knowledge is of what is there anyway . . . (2) Suppose A and

11 each claims to have some knowledge of the world . . . Now
with respect to their supposed pieces o f knowledge A ’s and B’s
representations may well differ. (3) If w hat they both have is
knowledge, then it seems to follow that there m ust be some
coherent way of understanding why these representations differ,
and hence how they arc related to one another . . . (4) A story
can be told which explains how A’s and li’s can each be a
perspective on the same reality. T o understand this story, one
needs to form a conception of the world which contains A and II
and their representations . . . (5) But this process, it seems, can
be continued. For if A or 11 or some other party comes in this way
to understand these representations and their relation to the
world, this will be because he has given them a place in some
more inclusive representation . . . If this is knowledge, then we
must be able to form the conception once more o f how this would
be related to some other representation which m ight equally
claim to Ire knowledge; indeed we must be able to form that
conception with regard to every other representation which might
make that claim, (fi) If we cannot form that conception, then it
seems that we do not have any adequate conception of the reality
which is there ‘anyw ay’ . . . But th at conception appeared at the
beginning as basic to the notion of knowledge itself. 'That concep­
tion we might call the absolute conception o i reality. (7) If
knowledge is possible at all, it now seems, the absolute concep­
tion must be possible too."*

It will be noted that Williams does not in this passage give the
characterization of the ‘absolute conception’ as th a t which rcpre-
 The Dissolving Mirror 155
sents how the world is independent of the peculiarities of observers.
But wc may justifiably take that to be the right interpretation o f the
phrase here. This is both because he elsewhere uses the same words
explicitly with that sense and equally explicitly with reference back
to this argum ent,17 and also because (as wc shall sec) it is the
interpretation required by the internal coherence of w hat is
claimed.
W hat is going on in this argument? It starts at (I) from a claim
which is very plausibly a version of something Jikc our minimal
realism — namely that certain of our thoughts we take to be of
things independent of those thoughts. T o claim to have knowledge
is to claim (in part a t least) that such a thought is correct and in
claiming knowledge we do not suppose that this correctness is
derived solely from the nature o f the thought. O n the contrary, wc
think of it as dependent on how it is with things which are there
‘anyway’, i.e. things independent of the thought. But, as stressed at
(2), I must recognize that my thoughts may dillcr from those of
others who also claim knowledge of this independent reality. Now,
i f l sec some representation o f mine as knowledge then I will take it
that any seemingly incompatible representation can cither be shown
up as simply erroneous or shown not to be really incompatible.
This is w hat is asserted at (3) and again our earlier thoughts in
chapter 2 chime in with it. T o adm it that my representation and
the incompatible one could both be right is ju st to adm it that
neither is a thought about w hat is there ‘anyway’, i.e. neither is
deserving of realistic, interpretation. Williams docs not stress the
notion of incompatibility. B ut its importance is implicit in his
notion o f mere difference. I f l suppose that what I think is a correct
representation of w hat is there anyway, then any differing represen­
tation presents a problem, namely ‘W'hy is it not the same as mine,
since mine is of w hat is?’ If the question docs not worry me it will be
because I already have w hat Williams next speaks of, at (4),
namely the wider representation in which both of these differing
ones are accommodated, for example a view o f ‘what is’ in which it
has various different parts, only some of which I have detailed
information about.
Let us now pause in our examination of t he argument to consider
w hat happens at the point where representations seem to clash and
156 The Dissolving Mirror
then we arrive at some resolution. It is essential to W illiams’
position that sometimes at this point one o f the representations is
revealed as being merely perspectival. T h a t is, we say to ourselves
‘Earlier we might have taken it to represent w hat was there anyway
but now we see that it does not; it merely represents how things
seem to us because of some peculiarity o f our situation.’ But what
form does this insight take? How do we deploy the resources of the
wider view of things we have now arrived at to enforce this
judgem ent of the status of our earlier concepts? Rem ember that this
is not to take the form o f the functionalist or projectivist style of
story which we earlier contem plated.
Williams himself offers the realization that A and B are in
different places as a very primitive example of how one m ight come
to understand how representations could differ.18 And it is presu­
mably from this prim itive case th at he derives the word ‘perspecti-
val’. So meditating on it should give us the clues we need.
Suppose, then, that A and B speak the same language (operate
the same system of representations) and A claims 'I t is raining’
while B says ‘It is not raining’. T heir representations are different
and incompatible so they cannot rest content with the view that
both, ju st as they stand and w ithout any further setting o r explana­
tion, embody knowledge. A and B then reflect and come to enlarge
their (explicit) representation of the world by the idea that there
are different places in it. Perhaps they add place-marking elements,
whether indexical or of some other character, to their vocabulary.
T he representations using these notions are now accepted as abso­
lute (or nearer the absolute) and the earlier ones are thought of as
perspectival. T he ‘explanation’ that A and B offer for their previous
way of talking might go something like this: ‘We used to talk in a
way which took no account of the possibility that we m ight be in
different places; we, as it were, “ took it for granted” that we were at
the same place.’ Let us note that this explanation deploys the new
conceptual apparatus to illum inate the content of w hat was pre­
viously said. It operates in an interpretive or rationalizing mode. It
does not purport to offer any reductive account of w hat it is to
think.
T he ‘perspectivalness’ of the previous representations, on the
view sketched, consists in their ‘recognisably and diagnosably
 The Dissolving Mirror 157
coming from a point of view’19 in the sense that their use has built
into it a certain assum ption. ‘U se’ m ust be read here as invoking
the role of the representation in the whole system. It is the system,
rather than the particular elem ent in it, which, in its limitations,
embodies the assum ption or point o f view. The structure of the
range o f remarks which the language or system provides for has no
place in it for marking the possibility which has now thrust itself
upon the speakers.2"
Now that we have in hand this picture o f what pcrspectivalness is
and how it manifests itself in unexpected disagreement, let us
return to W illiams’ argum ent. Wc have seen how, if I make a
knowledge claim, I take it that a prim a facie incompatible know­
ledge claim can be shown to be either erroneous or not incompati­
ble. This can come about in various ways. The other’s knowledge
claim, if not simply discredited as false, may turn out to be
perspectival and mine will already have the resources to explain
why it is different; or perhaps the reverse is the case; or perhaps
both claims will be seen as perspectival from the wider view we
shall then command. But such a com mitment to the possibility of
vindication o f knowledge is open-ended - as Williams remarks as
(5). He expands on this by claiming that we m ust be able to make
sense of every possible challenge from rival knowledge claims
having been successfully negotiated. We m ust thus, he concludes,
be able to make sense of the idea of a way of conceiving of things
which cannot itself be regarded as perspectival from any more
inclusive standpoint.
This is W illiams’ way of puttings things, but we may add the
following gloss. The suggested ‘absolute’ way of conceiving of things
would have to be one which could not be seen as taking anything
for granted. T he nature of the subject who judges (e.g. the fact that
he or she seems not to countenance certain possibilities) does not
have to be grasped by anyone who is to understand that person’s
representation. T he representations can thus be taken as providing
an account of how the world is in itself.
W illiam s’ argum ent then concludes at (7) by drawing attention
to the conditional claim: in making sense of knowledge at all we
commit ourselves to the intelligibility of the incorporation of every
worthy (i.e. not simply false) knowledge claim into one, unique,
158 The Dissolving Mirror
unified and pcrspcctivelcss picture. So if we can make sense of
knowledge a t all, we must be able to make sense of the absolute
conception.
Is the argum ent persuasive? There seem to be two points at
which one might have qualms. The first is the move from the
open-ended ness of the com m itm ent in the making of a knowledge
claim to the supposed intelligibility of an endstatc of having incor­
porated all knowledge claims in one unified picture. T h e idea of
knowledge may well involve the idea of ability to negotiate success­
fully with every challenge that arises. IJut why need we suppose
th at there is some given totality o f challenges with which it is
intelligible to suppose oneself having negotiated? W illiams, as we
have seen, claims that if we cannot make sense of this then we ‘do
not have an adequate conception o f the reality which is there
“ anyway” ’. But then perhaps in this sense o f ‘adequacy’ we neither
need nor can have adequate conceptions.
This leads to consideration of a second questionable assumption,
more deeply buried in the argum ent but even more fundam ental.
This is that a knowledge claim which negotiates a challenge by
adm itting itself to be ‘pcrspectivaP has thereby been shown to be in
some sense at fault. T he idea is that it has been shown to be less
than we were striving for, less than we implicitly claimed to have in
having knowledge; perhaps the fault is merely that of some kind of
incompleteness, but a falling short of some sort there is. This is the
idea that fuels the first questionable move. T he assumption is that
in having the idea of knowledge we have a certain ideal, namely
th a t of the ‘faultless’ representation, where the possibility that there
is another representation which differs from this and is not ex­
plained by it is taken to be a fault. A representation can only fulfil
the ideal we have in having the conception of knowledge if there arc
no further worthy rivals for it to negotiate with. So in having the
idea of something faultless we m ust have the idea of the non­
existence of rivals. And knowing ourselves to have knowledge
would involve knowing th at there are no more rivals to be defeated
or negotiated with.
But the idea that there is a fault here is part of, not separately
m otivated from, the ‘absolute conception’ idea itself. If one sup­
poses th a t there is a ‘how things are in themselves’ and th a t every
 The Dissolving Mirror 159
knowledge claim is a move in an intelligible project of getting a
picture of it, then indeed revelation of pcrspectivalncss will be
discovery of a fault. But if one thinks that making a claim about
what is so is always an aspect of (is only intelligible in the context
of) some life-involving project - a project not rcvcalingly charac­
terized merely as ‘discovery of the tru th ’ — then what counts as
fault in a representation will be bound up with what the life and its
projects are.
W ittgenstein draws attention to the idea that, an activity may be
fully defensible although it is not everywhere bounded by rules.
This is a prom inent them e in the early sections of the Investigations.'*'
It may seem that W illiams can acknowledge whatever truth there is
in this. He docs not w ant to condemn pcvspectival representations
in the sense of urging us not to have them or even in the sense of
denying that they can ever rightly be classed as knowledge in some
sense. So, for W illiams, languages arc not everywhere bounded by
rules; there arc, quite properly, ones where the participants are not
equipped to deal with every possibility.
But W illiams’ conception docs clash with the Wittgcnsteinian
one in this, that Williams thinks that the full ideal of knowledge is
bound up with the idea of a language which is everywhere bounded
by rules — i.e. a language which is equipped to describe every
possibility. If there were a possibility which it could not describe -
which its speakers would be at a loss to deal with - then the views
expressed in th at p a rt of the language are, for Williams, merely
pcrspcctival.
B ut W ittgenstein is not merely concerned to point, out that many
of our perfectly acceptable linguistic practices arc not everywhere
bounded by rules (e.g. the case of the disappearing chair).22 N or is
he saying that there is a given totality of possibilities which our
language will always, but excusably, fail to articulate in ever)'
detail,23 He is, I suggest, trying to get us to see that the idea of the
given totality of possibilities makes no sense. Any linguistic move at
all has presuppositions because it is bound up with, and its sense is
dependent on, a context o f interests and activities - interests and
acitivities which, in turn, only make sense given other empirical
facts. We do not make sense of the idea of a totality of possibilities
because th at would require us to get some overview of the totality of
160 The Dissolving Mirror
ways life could intelligibly be lived. And although we may m outh
this phrase, it does not have for us the kind of sense which would
enable us to do substantive philosophical work with it.
I t may sound as if an unacceptable form of pragm atism was here
being recommended. For example it may sound as if it were being
suggested that the making of a statem ent was always a move in
some limited pratical project and that its ‘tru th ’ could am ount to
no more than its turning out effective in securing the desired goal. I
shall try to suggest, in section 8.1 th at this is not so.

7.3 K R IP K E AND NORM S

In order to understand m eaning more fully we need to bring in

projects, interests, values or the like. T h a t was the theme of the first
section o f this chapter. But how exactly do they enter an account of
meaning?
O ne theory is th at which is offered by Kripke.24 O r if not offered
by him it is at least a plausible reading of w hat he says which has
been adopted by some.25 This theory constitutes a move which we
might make in response to the negative claims, about the insuffic­
iency of formulae, images etc. to constitute meaning, although it is
also bound up with the exact form th a t the diagnosis of insuffic­
iency takes. The view is th at those negative conclusions arc essen­
tially linked to the fact th a t m eaning is a normative notion. W hen we
realize this, we can devise a sceptical solution to the sceptical
paradox, a solution which shows w hat the role of attributions of
meaning is in the context of a language-using community.
I shall argue in this section th at the proposed theory is unattrac­
tive, first because meaning attributions are not ‘norm ative’ in the
sense required and, secondly, because if we accepted it we would be
back in a version of the unfortunate position examined in the first
section of chapter 6, nam ely o f endorsing a form o f meaning
scepticism which undercuts the realism which we had thought to
preserve about things other than meanings.
Suppose we are considering some speaker and we say of her that
she means addition by ‘plus’. W hen we do this are we describing
 The Dissolving Mirror 161
something? Is there some fact which makes our remarks true or
false? This is the question that Kripke asks us to consider.'20 We
then examine various candidates Tor what her m eaning addition by
‘plus’ might consist in. We may find items, (formulae, brain states
or whatever) which cause or causally contribute to utterances of
‘125’ in answer to the question ‘W hat is 68 plus 57?’ Now some­
one’s meaning something by ‘plus’ surely (helps to) explain why
she speaks as she does. So it seems that we have in these brain
states, formulae or whatever plausible candidates for what it is to
mean.
But, says Kripke, ‘the relation of m eaning and intention to future
action is normative, not descriptive'’.27 O ne central point here seems to
be this. Given that she docs mean addition by ‘plus’ (and also of
course 68 by ‘68’ and 57 by ‘57’) and that she is asked ‘W hat is 57
plus 68?’ then she ought to reply ‘125’. T his is the correct, the true,
thing to say.
I t is worth stressing that the strategy here (as I am interpreting
it) requires us to deny the link between m eaning and utterance in
the form in which the alternative, causalist, theory assumes it. The
fact that someone means something by a certain noise does not
(according to Kripke) imply th at she will or is likely to say a given
particular thing in answ er to some question. M eaning is no guaran­
tee against mistake. T h at is w hat 'seems to be implied by the
remark th a t the relation of meaning to future action is ‘not descrip­
tive’. I f the relation to action were both descriptive and normative
then it would be open to someone who thought that meaning
consisted, for example, in the occurrence o f some causally effica­
cious item to argue that all his theory needed was supplem entation
by some further element which captured the desired normative
features. Kripke’s contention is that items which contribute to causal
explanation of noise-making are not merely an incomplete analysis of
what it is to mean but could be no part o f the analysis at all.
K ripke’s emphasis on future action in his initial exposition has
misled some com m entators into thinking that cross-temporal link­
ages of some sort are the crucial issue.28 Bur he could ju st as well
have said th at the relation o f meaning to current action is norm a­
tive and not descriptive. T h e point has to do not only with language
162 The Dissolving Mirror
b u t with thought more generally. A ttributions of content of any
kind can be said to be norm ative and not descriptive. I f I say of
someone that she is thinking that the tree is about to fall, I have not
said anything descriptive (about the imagery going on in her mind
or w hat not) but 1 have m ade a com m itm ent about the conditions
under which her judgem ent will be true, right and what she ought
to have judged.
Let me emphasize once again th at there is much more to be said
on this issue of w hether a causalist, or broadly functionalist,
account of mind of the kind Kripke attacks can give an acceptable
underpinning to whatever notion o f the norm ative is required. But
we arc proceeding here upon the assum ption (to which I am very
sym pathetic) that Kripke is right in rejecting such approaches.
W e have, as we have seen, a trio of terms for appraising u tter­
ances and judgem ents, when they arc regarded as mcaning-
bcarcrs. We can say that they are what one ought to say or judge,
that they are correct or that they are true. Each of these terms can be
used to express the key ‘norm ative1 idea we are trying to capture.
’Fhe naturalness o f ‘ought1 in this context sets us oil’ on a line of
thought in which we hear ‘norm ative’ as meaning ‘evaluative’ and
where we take ‘evaluative* in the sense ol the traditional fact/value
distinction.
By the ‘traditional’ distinction I mean the one grounded in
H um e’s style of consideration.29 T h e key elements are these: M oti­
vation to act always involves desire (pro-attitude, feeling) as well as
belief. A clear line can be draw n between belief on the one hand
and desire (attitude, feeling) on the other. T h at is to say that given
any mere belief it is always possible, i.e. intelligible, i.e. coherent
and not irrational, to take up various different attitudes to the state
o f alfairs the belief is about. T o apprehend that things arc thus and
so (or might be thus and so) is never of itself a motive to action.
T here m ust also be the (at least in theory) separable m atter of
w hether one cares about things being thus and so. No caring about
anything is logically required by the fact apprehended. Conse­
quently any intentional state, however prim a facie beliel-like,
which docs necessarily provide some motive for action m ust con­
tain a desire (feeling, attitude) com ponent as well as the belief
com ponent. So any prim a facie descriptive rem ark which is such
 The Dissolving Mirror 163
that in accepting it one necessarily allows that one has some reason
to perform an action cannot be entirely fact-stating, it cannot be
expressive merely of belief but will also express some desire, feeling
or attitude.
As Blackburn has argued, wc arc not committed by this outlook
to some crude ‘B oo-H urrah’ theory about evaluative utterances.30
We do not need, on acceptance o f this view, to reform our linguistic
practice of treating these utterances almost exactly as wc treat
fact-stat ing utterances. W e may talk o f them as ‘true’ or ‘false’ and
operate as though the laws o i non-contradiction and excluded
middle held for them, we may include them as premises in argu­
ments and so iorlh. But acceptance of the view docs involve
supposing th a t there is some highly reflective philosophical stand­
point from which wc can distinguish the real facts from the secon­
dary or projected ones, the seeming presence of which in the world
is the result of the interaction of the value-free world with our
affective natures.
Suppose we hear K ripke’s remarks about what someone ‘ought
to say’ (given w hat she means by ‘plus’ and so forth) in this light,
how do things then unfold? Perhaps like this: In saying ‘she ought
to say “ 125” ’ I am not describing any real fact about her. Rather I
am expressing my approval o f her now uttering ‘125’. Why do I
have a pro-attitude towards her uttering this noise? Because it is
what I myself feel impelled to say and 1 like other people to keep in
step with me. I like that because it makes me feel at home in the
world and enables me to do various of the other things I like doing.
T he rem ark ‘she ought to say “ 125” ’ is entailed by ‘she means
addition by “ plus’” together with some other statements. W hat
entails an evaluative statem ent m ust be in part evaluative itself. So
unless the extra premise not concerning meaning, i.e. that 68 plus
57 is 125, is evaluative (which seems implausible) then the ‘means’
statem ent itself m ust be evaluative. W hat attitude or feeling docs it
express? It expresses my willingness to dignify this other person as
a m athem atician of sorts. It shows that I regard her as a creature
whose utterances in context like this - where someone has asked
‘W hat is 68 plus 57?’ —m atter to me. I have the attitude of taking
seriously w hether she is in step with me or not. I welcome her into
the linguistic com m unity.31
164 The Dissolving Mirror
But will all this do as a coherent story about meaning? (We are
not here concerned directly with its accuracy as a representation of
W ittgenstein. But many com m entators have made telling points
against supposing Kripke’s argum ent, w hether or not this is quite
the right reading o f it, it to be acceptable in that role.)32 T he
problems with accepting it are first th at the premise is false and,
secondly that the conclusion self-underinining.
For the first of these claims I have argued elsewhere.33 T he
central point is that, w hether or not we accept the H um ean account
of evaluative (i.e. action-guiding and practical reason giving) vo­
cabulary, it would be wrong to put ‘true’, ‘false’ and ‘means’ in the
same bag with ‘am using’, ‘noble’, ‘kindly’, ‘cruel’, ‘betrayal’, ‘bor­
ing’ and so forth. T o say that a rem ark is true, or that some course
of action would lead to acquisition o f a true belief is not an
action-guiding or reason-giving com m ent. I do not come to under­
stand why someone made a rem ark or pursued some course of
action simply by learning that the rem ark was true or that the
action led to acquisition of a true belief.
We are often led to think the opposite by concentrating on cases
where a context of interest in some particular subject m atter, a
context requiring remarks o r beliefs about it, is already set up. This
is w hat Kripke’s imagined question does. B ut if we imagine cases
where there is no such context then it becomes plain that truth as
such has no evaluative pull. T here are m any true sentences th at I
could insert into this book. I could for example write next what I
ate for breakfast this morning. But you are hardly likely to find my
doing so intelligible if I tell you th at I ju s t could not resist enun­
ciating this truth. There are m any facts that I could learn by
observing the colours o f the book jackets on my shelves. B ut I am
not all the time struggling with some tem ptation, albeit very tiny,
to go and learn those facts. W hy we find it intelligible to make
certain remarks and pursue certain enquiries has to do with the
subject m atter involved and With values bound up w ith talking and
knowing about that subject m atter, not with the evaluative status of
truth in itself.
It is a good thing that we can thus resist em barking on the path
suggested because the conclusion th at m eaning is itself a projective
notion directly undermines the idea th at there are any facts (or at
 The Dissolving Mirror 165
least any of which we can have knowledge) which can intelligibly
be contrasted with the pseudo and merely projected facts about
values, truth and meanings. T he dilliculty the position encounters
here is closely akin to th a t which we explored in the first section of
chapter 6. We are told th a t the ‘real’ truth about meaningful speech
is th at it is social noise-making accompanied by various patterns of
feelings of satisfaction and dissatisfaction in speakers and hearers.
T he noise-makings, and w hether or not the noises are the same as
the ones I feel impelled lo produce, are the asscrtibility conditions,
the skeleton, upon which my projection of feelings o f like or dislike
builds the phenomenal body of apparently meaningful language
use. This is the story. But how can I (or you) apprehend a fact in
the m anner required by the attcm ptedly realist part o f this picture,
if the states such as thinking and speaking which at first glance we
took to be that apprehension turn out to be mere noise-making
accompanied by feelings o f .s atisfaction? 1 have cut off the branch
on which I hoped to sit.
Suppose that we push on, accepting tlte consequence of thor­
oughgoing projcctivfsm about meanings, we then arrive at projec-
tivism about the facts which that (apparent) meaningfulness
(seemingly) allows us to apprehend. So the end point is a thorough­
going transcendental idealism in which an interaction between an
unknowable subject and an unknowable something else results in
states in which we (seemingly) apprehend a world of things and
persons. And we cannot even say this with a confident sense that,
however mysterious, this is how things really are. T he content we
(seem to) discern in th a t very rem ark is, by its own lights, only a
figment.
Let me here reiterate that I am not sure that the line o f thought I
have traced here accurately represents the one followed by Kripke.
It could be said that I have imported talk of ‘liking’and ‘approval’
where Kripke speaks only o f primitive dispositions to give answers
and to respond to training. Hence, it may be argued, there is little
firm evidence that K ripke’s ‘norm ative’ is to be heard as a Hum ean
‘evaluative’. I would also like to acknowledge that Kripke is
im portantly right in his initial attack on the dispositionalist^ or
causalist account of m eaning and in his invocation of ‘norms at
this point. T he two crucial claims here would be these: fust, that
166 The Dissolving Mirror
talk of m eaning is appropriate only for items which play a role in
the life o f creatures for whom some things, including 011 occasion
their own utterances, arc in various ways im portant, valuable anil
cared about; and secondly th at the idea o f something being im port­
an t to a person, m attering to him or her, cannot be explained in
terms of causes and dispositions. T his is one of the points where
battle might be joined, were wc to pursue the issue raised a t the end
o f chapter 6 concerning the feasibility o i naturalized account of
mind.
However right Kripke is in the general thrust of his claim here,
and however unfair to him the more detailed interpretation o f the
course of the argum ent which I have offered, the im portant point
(or the line of argum ent in this book is th at the notions of m eaning
and value should not be lied together in the way set out in that
detailed interpretation. And a further im portant point is that, on
the view 1 recommend, there will be something wrong about
K ripke’s argum ent, whatever its details, namely th at it issues in a
sceptical thesis about meaning. T h e strategy I am pursuing is th at of
bringing interests, values, forms of life and so forth into the discus­
sion of meaning as a way o f loosening the grip of m irroring realism
and hence finding a stance from which wc can happily acknowledge
the reality o f meaning (and very probably of value too - although
th a t is not a theme I shall pursue).
 8
Interpretations and
Misinterpretations

8.1 SPEECH ACTS AND LANGUAGE GAMES

T he last chapter introduced the idea that Wittgenstein is hostile to

m irroring realism and attem pted to sketch the grounds for that
hostility. I t also endeavoured to disarm a counter-argum ent on
behalf of m irroring realism and to forestall one misinterpretation of
the proposed m caning-intcrest link. But many questions remain.
Some concern the nature of the interconnection between concepts
and interests on the W ittgcnstcinian view; others are about
w hether any non-m irroring view can rightly claim to be realistic at
all. T he aim of this chapter is to explore these issues and to clarify
the W ittgcnstcinian position further, in part by distinguishing it
from tem pting but ultim ately misleading views with which it may
be confused.
T h e m irroring idea oilers us the thought that Lhere is one way
that the world is in itself and which (some ol) our thoughts strive to
capture. A proposal to abandon this way of looking at things is
liable to be seen as recom m endation o f the idea that ‘there is
nothing determ inate out there’ —th a t the world is a kind of mist,
upon which we project shadows which we then mistake for pre­
existing figures. Recoiling from this picture, while still striving to
respect the insight that the world cannot enforce any classificatory
scheme upon us, we might hit upon the idea of crediting the world
with immense richness of characteristics, rather than with none at
all. W hat was wrong with the m irroring picture, we say, is that it
assumed ju s t one nature for the world; the truth is that it has
indefinitely m any.1
168 Interpretations and Misinterpretations
Given this idea we seem to have no difficulty in saying w hat the
role of our interests (values, projects) is in the business: they
determine what facts, or what sorts of fact, are o f concern to us,
what conceptual net wc shall use. O u r fishing w'ith a net of a certain
mesh fixes that there are some fishes we shall not catch, but it does
not create the fish that we do catch. To put it less m etaphorically,
our interests make certain aspects o f the world of im portance to us
and others of no relevance. But the aspects that we do notice arc
there all right and we need have no qualm s about realistic con-
sirual of our thoughts about them. T he sort o f exam ple one might
use to illustrate the idea would be the fam iliar one of the Eskimos
and their recognition of many types of snow. T heir interests require
them to use snow as a building material; consequently they mark
all kinds of differences between types of snow o f which we are
oblivious.
I do not want to suggest that this kind of explanation can never
be given or that it is not illuminating. W hat it provides, however, is
a causal or quasi-causal account of why certain concepts are
possessed. We may agree that having certain interests (in building
shelters) is likely to lead to acquisition of certain concepts (of
different kinds of snow and their properties) if the agents and their
environment are of certain kinds. But the relation between interest
and concept remains contingent and external. We have not been
given any reason to suppose that the same features o f snow could
not be noticed by someone with very different interests. So posses­
sion of the interest is not in any sense constitutive of possession of
the concept; the interest explains possession of the concept but not
in a way which gives us deeper insight into w hat it is to have th at
concept at all.
W e should note also th at there is nothing in this story to prevent
us supposing that we can set aside the practical interests which
make our familiar concepts useful to us and ask, disinterestedly,
w hat further features things have which we have so far failed to
notice. We might hope to get some sort of schematic view o f a
greater range of properties and to locate our own conceptual
scheme among various others which we see would be appropriate
to creatures with different interests. An enterprise o f this shape may
be intelligible. But if we ask w hat metaphysical picture goes with it,
 Interpretations and Misinterpretations 169
it is clear that m irroring realism is not ruled out. 'The theory oilers
us the facts that creatures have interests, the fact that they have
concepts and the fact th at the world includes such and such a range
o f properties, and it knits them together in an explanatory way. But
there is nothing to prevent us from conceiving of these things
themselves on the m irroring model; that is how things really arc
(we think) and we have become aw are of it through our ability to
grasp how things are in themselves.
T o understand W ittgenstein’s vision better we need, 1 suggest, to
move to something more radical than this. Perhaps one of his own
starting points, the com parison o f a judgement to a move in a
game, will help us.
Let us consider something like making a bid in bridge. In order
to understand what is going on one needs to grasp that making a
bid is a particular sort of element in a whole complex pattern of
manoeuvres. W hat would be a good or bad bid (and what counts as
making a bid at all) is seen only when one sees the structure and
objectives of the whole set up. T he bid may be assessed in various
dimensions, as sensible, exciting, successful and so forth. T he
possibility of such assessments is implicit in the structure of the
activity. But it would be odd to assess the bid as true or false or to
think of it as a factual claim.
Com pare with this my m easuring a sofa, writing down ‘4ft Gins’
on a piece of paper, carrying out a similar operation on a space of
wall, comparing the written marks and finally moving the sofa into
space. One might say analogously that ‘m easuring’, ‘noting the
result’ and the like can only be understood as elements in the
complex whole. T he objective is to fit the sofa somewhere in the
room; judging that it is 4ft Gins long is a move which is assessable in
various dimensions (‘justified by the rules of m easurem ent’, ‘actu­
ally leading to a successful placing o f the sofa’ etc). And even if one
of these dimensions is labelled ‘being true’ (says this line of
thought) that should not blind us to the fact that what is really at
issue is success in the enterprise.
Contem plating an example such as this, or W ittgenstein’s ‘live
red apples’ case,2 it m ay seem th a t w hat he is recommending is a
‘speech act’ account of the ways of talking under discussion. This
‘speech act’ manoeuvre is one th a t wc are familiar with from the
170 Interpretations and Misinterpretations
discussion of such things as the emotive and prescriptive theories of
ethics or the suggestions of A ustin.3 W hat is distinctive in such
theories is the idea of taking some philosophically controversial
word (‘ought’, ‘true’, ‘responsible’ etc.) and denying that it has a
descriptive role. Philosophers, it is said, have spent many years
trying to understand w hat facts one could be describing by the use
o f such words and bow we could provide a respectable epistem-
ology for them; but wc now see th at these attem pts were all
misguided and doomed to failure, because the role o f the words is to
be conceived diU'crcntly; the essential thing to realize is th at in
uttering them we are making moves in some complex social activity
and the question of whether such moves are right or wrong should
not be construed as the question of w hether they arc true or false.
Perhaps, then, W ittgenstein is proposing a wholesale extension
of this policy to every expression o f our language, not just those that
have proved philosophically controversial? N um ber words, colour
words, tem poral and modal expression, psychological vocabulary
- to understand any of these wc m ust grasp w hat language games
their use is em bedded in, we m ust see w hat moves wc make in
uttering sentences containing them .4
There arc serious things wrong with this proposal, as wc shall sec,
namely that it yields an unacccptably non-realist story and, worse,
that it is incoherent. But before considering these objections we should
note that there is something very importantly right in the idea. This is
the prominence it gives to the idea that the meaningful employment of
any linguistic item requires a great deal of stage setting, in terms of
circumstances and kinds of objectives, a stage setting which, in that
employment of the item, is just taken for granted.
It is worth m editating on this further, for it is one of the
im portant corollaries of'the thought in section 7.1 and one which
we again and again shy away from or fail to get into focus. Consider
a person who, as the outcome of some confrontation, makes a
judgem ent of the form ‘This is F’. I f we are not meaning sceptics
then wc take it that this judgem ent invokes a standard of w hat is to
count as the sam e which places the confronted item as a m em ber of
one extended grouping rather than another. T h e judgem ent is in
order as a judgem ent, not in the sense o f being true but in the sense
of having a content and being assessable as true or false, provided
that there is such a standard.
 Interpretations and Misinterpretations 171
B ut w hat is requited for such a standard to exist? O n the
m irroring view all that is needed is th at ‘F ’ be a label marking one
of the ways things can be, i.e. picking out an independent, built into
the nature of things, property. W hether or not a classificalory rule
has this status is not a contingent m atter. So there arc no contin­
gent facts the existence of which is required to underpin the
respectability of ‘F ’ as a predicate. T he use of ‘F’ runs no risk of
turning out to be radically flawed because of the non-existence of
such facts. O n the W ittgcnsteinian view o f section 7.1 by contrast,
every concept is bound up with a way of living and with the
possesion of certain interests which inform that way of living. The
viabilility of a concept will thus be dependent on whatever is
required for that way o f life to be pursued and for its interests to be
continuingly defensible. And it is difficult to sec how this could fail
to involve a ramifying set of contingent facts, which then constitute
a needed stage setting for the use of the concept *F’.
Suppose then that wc recognize some unavoidable entanglement
of concepts with stage setting, docs this have the consequence that
any linguistic move is to be seen as (really) an action, as opposed to
a judgem ent of the truth? T his is the way that the ‘speech act’ idea
chooses to spell out the idea. T w ant to argue now that such a view
is incoherent. For it to work it is required that an account of various
assessment conditions for a given utterance should be available in
terms not invoking the utterance to be explained. For suppose this
were not so; suppose for example that in talking of the success
conditions of the utterance ‘T he sofa is 4ft bins long* one had to
invoke the condition th at the sofa is 4ft 6 ins long. Clearly here we
have lost any leverage for saying th at the utterance is to be given a
‘speech act’ construal. Let us imagine that this pitfall has been
avoided and th at wc do have some independent specification of
w hat counts as success o f various kinds. W hat story are we now to
tell about the linguistic moves involved in giving this account of
success? Im agine, for example th a t I say a bid in bridge is success­
ful provided that such and such a num ber of tricks is secured by the
bidding player and his or her partner. This certainly avoids (and
plausibly in this case) any risk of precipitation back into some idea
that making a bid is producing some sort of statement. But what
am I doing in saying of some bridge players ‘They made such and
such a num ber of tricks’? T h e natural thing is to construe it as a
172 Interpretations and Misinterpretations
description. But then the theory is not fully general. And if we
endeavour to make it fully general by finding some further
language game in which th at ‘description’ is a move, we are
embarked on an infinite regress. Let us spell this out in more detail.
The original speech act theories, as proposed by the emotivists,
prescriptivists and Austinians, had built into them a contrast
between content-giving, sense-giving or ‘phrastic’ elements on the
one hand and mood, force-indicating or ‘neustic’ ones on the other;
they also had built in a contrast between assertive and non-
assertive force. Particular ‘speech a ct’ theories were offered within
this framework and involved trying to re-locate suspect items from
the ‘sense-indicating’ box into the ‘force-indicating’ one. But w hat
we seem to be asked to do under the present proposal is different
from these piecemeal manoeuvres. We are asked to pu t everything
into the ‘non-assertive force indicator’ box.
O ne might suppose that the incoherence here lies in the fact that
notions like ‘force’ require the intelligibility of ‘sense’. But this
objection is not a serious one. 11 is true th at some force notions do
require sense as a contrast —for example the idea o f assertive force
requires the idea of the content of w hat is asserted. But not all force
notions have this feature. (Consider the force o f ‘H ello’.) W hat we
are being asked to do (on this construal of W ittgenstein) is to
construct a gram m ar in which every sentence is built up from
force-determining elements and has, as a consequence, a complex
force, i.e. suitability for perform ance of some appropriately com­
plex move.
So far so good. But the incoherence emerges, the infinite regress
appears as vicious, when we ask ourselves how we are to conceive of
w hat we do when we set out the conditions under which these
linguistic items are to be used and the conditions under which the
moves made by them succeed. In giving these conditions we seem
to ourselves to state circum stances which, among other things,
provide reasons for making the linguistic moves we arc trying to
explain. But giving the conditions is, o f course, ju st itself making
more linguistic moves. These moves themselves m ust, if the theory
is right, be regarded as actions which we choose to do in the light of
circumstances which make them appropriate or not. So we now
 Interpretations and Misinterpretations 173
have a picture of ourselves choosing w hether or not there will be
justifying reasons for some other action. And our current action of
so doing will turn out to have reasons in favour o f it only in the light
o f certain other moves which again wc m ust choose to make. And
so on.
T h e dilliculty we have arrived at here is a close relative of
(indeed perhaps another incarnation of) that which wc encoun­
tered with thoroughgoing pragm atism in section G.2. There the
insurm ountable problem was that of com bining the idea that every
feature o f our world is put into place by o u r choice with the idea
that the choice could be thought o f as constrained. T he upshot was
that the idea of constraint on choice (and the linked idea of rational
motivation in choice) requires the notion o f fact. It requires it both
to give substance to the idea of that in the light of which the choice
is to be made and to give substance to the idea that something
hangs on the choice - th a t it could turn out to be wrong as well as
right. T he earlier debate o f section 6.2 took place in a rarefied
atm osphere. We talked of the advantages of adopting theoretical
positions, of recalcitrant experience and the like. Now wc are
talking about more m undane activities like shopping, arranging
furniture and playing bridge, but (he fundamenetal point carries
over. We are invited in both cases to see ‘making a judgem ent’ as
an action, something done by choice, the success or failure of which
is to be assessed accordingly. But the inadequacy o f this pragm atist
stance is shown in the fact th at we can think of certain of the things
we do (in a broad sense o f ‘do’) as motivated and constrained
choices only if we see certain others in a quite different way, namely
as recognition of facts.
W e should mention one more consideration against the ‘general­
ized speech act’ proposal. As applied, for example, to the case of
measuring the sofa, it hopelessly falsifies w hat is going on, from a
phenomenological point o f view. It ju st is not the case that I
suppose that w hether or not the sofa fits at the end of the day
(given, o f course, certain results of the other measurem ent and the
comparison process) is of crucial im portance in determ ining
w hether or not I judged correctly when I wrote down ‘4ft Gins’. I do
not conceive of myself as doing som ething (like making a bid in
174- Interpretations and Misinterpretations
bridge) which has its sole significance from its role in this sharply
bounded, if complex, set of transactions. O n the contrary 1 take it
th a t I am finding out how things arc in themselves, th at I am
discovering which is ‘there anyw ay5. O f course (one will add) this
inform ation may be useful and having it guides m e to the satisfac­
tion o f my desires. But w hat it is to have the information and
w hether it is correct is independent of these particular desires and
w hether they get satisfied. T he phenomenology, in other words, is
such as to make the mirroring conception eminently satisfying.
If, then, we arc to find any distinctive and defensible Wittgenstci-
nian view on the interdependence o f concepts and interests we need
to find a way of rejecting a m irroring or ‘absolute’ conception
which docs not precipitate us into the absurdities of the thorough­
going pragm atist or speech act view.
Some commonscnsc observations m ay help us here. O u r lives do
not segment into self-contained episodes o f action. T hat is not to
say that we all have a ‘life plan’ to which every particular action is
tightly subordinated; all kinds o f inter-relations, both strong and
weak, may hold between all kinds, o f projects conceived at various
levels of specificity. But typically we do have various long-term
aims and persistent interests, even if they arc only vaguely charac­
terized; particular actions are often seen as means to, or elements
of, or instances of, some more general enterprise. A nother obvious
feature of hum an experience is memory and the cum ulative nature
(up to a point) of our stock of abilities and elements of our picture
of the world. We suppose ourselves to learn something on one
occasion, a fact or a skill, which we may be able to apply on
another.
It is the combination of these two features which makes the
‘speech a ct’ reduction o f ‘this sofa is 411 6ins long’ so implausible. It
is not w hether it now fits the space (given the other results) which is
the standard of ‘tru th ’, because I m ay well remember the dim en­
sions and use the information on other occasions in the course of
other projects. (Suppose, for example, that the space and the sofa
m easure the same, b u t the sofa fails to fit the space because the
other pieces of furniture bordering the space expand between the
m easurem ent and the attem pted fitting. Even though the fitting
w ent wrong, 1 get the dimensions right, and may be able to use the
 Interpretations and Misinterpretations 175
information elsewhere.) We can make the same point more gener­
ally about the possession of concepts exercised in the judgem ent.
My grasp on the concept ‘four’ docs not consist simply in my ability
to go through some circum scribed set of manoeuvres of measuring
and fitting sofas. O nce 1 have acquired numerical notions I am in a
position (if other things are propitious) to embark on all sorts of
projects, counting apples, m easuring the distance to the moon and
so on.
W hat can I now say about this ‘and so on’, about the nature of
those other occasions on which I shall use the inlormation about
the length of the sofa or exercise my grasp of t he concept ‘four’? On
the one hand there is som ething that tempts us in the mirroring
direction. It is th at there seems to be nothing T can say of a general
character except th at those situations will lit the given description,
namely will be ones where the information cornes in useful or the
concept is exercised. I certainly cannot give any reductive ‘speech
act’ account because that would be to imagine, absurdly, that I can
foresee in detail the future course of my life and that it will consist
ju s t of the repetition of some limited number of self-contained
activities; I cannot now lay down limits to the way in which my
information and concepts will interact with other similar items, nor
can I foresee the nature of the projects in which the resulting ideas
Will play a part.
O n the other hand there is something else which can be seen as
pointing another way. I find on reflection that there is at least one
more thing that I can say about these future activities, namely that
they will be like this one, in which I measured and attem pted to fit
the sofa. To have the concept ‘four’ or the knowledge that the sofa
is 4ft 6ins long is to have a flexible and openendcd capacity Tor new
kinds o f undertaking. But it is precisely kinds o f undertaking like this
th a t it is a capacity for.
T h e m irroring conception imagines that judgem ents can, so to
speak, ‘float free’ from their historical and psychological origins in
training rituals and particular practical contexts. And in one way
they can. T he child learns to say ‘all gone’ as a move in a
parent-satisfying and milk-drinking ritual. But he or she docs not
understand w hat ‘all gone’ means until the words lose their tight
connection to this particular context and thé child acquires the
176 Interpretations and Misinterpretations
ability to recognize of all kinds of things on all kinds of evidence
that they are ‘all gone’. But judgem ents and concepts cannot ‘float
free’ from their roots in the sense th a t we can fully justify and
explain them by pointing solely to the world which they are (on this
view) designed to m irror. I f we ask ‘why do we have these con­
cepts?’ and attem pt an answer by trying to show that the world of
itself requires that we do then we run into the difficulties outlined in
seciion 7.1, namely those o f supposing that the world is a gigantic
sign with only one interpretation and that we can make sense of
having or aiming to have the content of our thoughts fixed by that
inert item.
So we come finally to saying som ething like this: ‘Possession of
such and such a concept is exercised paradigm atically like this, at
such and such a point in so and so kind o r enterprise; we have the
concept, and it is the right one for us to have, because it is part of
our being able to do such things; if you doubt this, try to imagine
w hat hum an life would be like if we did not undertake this sort of
enterprise in this sort of way; you will find that you cannot make
much of it.’
T o say this is to invite ourselves to ‘assemble rem inders’, to do
'the natural history of hum an beings’, and it is to suggest th a t it will
be a help to use certain simple ‘language gam es’ as starting points
for reflection.5 The upshot of pursuing the line, however, is not th at
wc get some vantage point from which the overall layout of hum an
thought can he surveyed and on which it turns out to be pragm ati­
cally rather chan mirroringly describable. T he central thing we
need to hang on to is that the interdependence of concepts and
interests is not to be spelt out as the priority of interests over
concepts. T hat way lies incoherence (as in the ‘speech act’ view) or
re-imporlation of another version o f the m irroring view (as in the
‘Eskimo and snow’ view first canvassed in this section). T he
interdependence is genuinely that. And it manifests itself in the fact
that ‘assembling rem inders’, getting a sense of our language games
and their not now foreseen ramifications, is the form that ground­
ing and justifying our concepts will take. B ut none of this will
supply us with materials for constructing reductive accounts of the
content of any judgem ent or of w hat its correctness will consist in.
Suppose that we wish to find out w hether the sofa is more than four
 Interpretations and Misinterpretations 177
feet long. O u r W ittgcnsteinian reflections on length, num ber,
m easurem ent, furniture and so forth do not tell us th at w hat we are
really trying to find out is such and such else. It oilers us no
substitute focus for our thoughts when we ask questions about or
investigate the state of the sofa. And, relatedly, if we enquire what
is dem anded for it to be correct to claim th at the sofa is more than
four feet long, the reflections tell us that all we can say is that it is
necessary that the sofa be more than four feet long.
Despite these reassurances about rcductivism and die continuing
defensibility of certain truisms about truth, is there nevertheless a
threat to ‘realism ’ in the view proposed? This is an outstanding
question about which something needs to be said. T here are two
points to stress. T he first is that there is nothing in w hat has been
said to commit us to the idea that our judgem ents are uncon­
strained; in other words wc have not been precipitated back into
the situation threatened by thoroughgoing pragm atism where we
lost grip on the idea o f pointfulness in utterance, of succcess or
failure to be faced. T h e second closely related point is that there is
nothing in w hat has been said which requires us to deny the idea
that there are certain elements of our thought th at we take for
granted (in the m anner outlined in section 6.3), th at these involve
operation with the law of non-contradiction in m any areas, and
that our practices, in consequence, satisfy the dem ands of minimal
realism.
O u r situation then is this. We are not in a position to determ ine
the success of our enterprises. We propose things to the world, but
the world disposes. T h e central point th at we have em phasized, in
rejecting the m irroring view, is that the proposals (the concepts in
terms of which all our judgem ents and actions are to be character­
ized) come from us and that their nature cannot be understood or
our possession of them defended by pointing to the world. But this
fact, that the proposals are ours, does nothing to soften or limit the
fact of our non-omnipotence. This observation seems to protect one
very im portant elem ent which distinguishes realist from idealist
stances. And we need to note also that the proposals or questions
we put to the world are highly specific because they arise in the
course of projects, thoughts and actions which have already a detailed
and taken-for-granted structure. Moreover they are frequently
178 Interpretations and Misinterpretations
posed on the unquestioned assum ption that the answers arc subject
to the principle of non-contradiction. It is answers to these kinds of
questions, success or failure of projects conceived in these highly
specific term s, which the world, by and large, delivers back to us. It
may be th a t wc need to revise our rnininal realist practices in
particular cases. But the view under consideration offers no encour­
agem ent to the idea that wholesale revision is desirable or even
intelligible. It may be felt that all this is not enough as a defence of
a claim to be realistic. But the W ittgcnsteinian will then w ant to
pose a challenge to the opponent, nam ely to articulate more clearly
and intelligibly what is missing. T h e suggestion will be that any
such attem pt at articulation will fall back into the diiiicultics o f the
m irroring view outlined in section 7.1.

8.2 W IT T G E N ST E IN AND A N TI-REA LISM

T h e object of this section is to consider w hether wc should sec

W ittgenstein as an ‘anti-realist’ in the sense explored by D um m ctt,
K ripke and W right and to say som ething about the argum ents for
anti-realism .'1 There is no agreed view offered by all those sym­
pathetic to anti-realism , but the central idea is hostility to the view
th at our understanding can outstrip what wc arc capable o f getting
evidence about. We must not say that understanding a sentence
involves grasping that it describes some state of affairs (has the
obtaining of th at state as its truth condition) when this state of
affairs is one whose obtaining or not obtaining is, wc acknowledge,
something upon which wc can get no epistemic handle.
W here we go (or suppose W ittgenstein to have gone), having got
this starting point, is far from clear. Arc there some base state­
ments, which are ‘dccidable’*and can quite unproblcm atically be
allowed to have truth conditions, using which we can give some
system atic asscrtibility conditions theory of meaning? Did W itt­
genstein favour such a thing? O r did he (mistakenly) get his
anti-realism mixed up in an abandonm ent of the sense/force dis­
tinction and so lose the chance of a systematic theory he should
have pursued? (This last is D um m ett’s view.)7 Does the idea of a
base class require that its members be indefeasible? I f so then
 Interpretations and Misinterpretations 179
perhaps there are no such statem ents. Did W ittgenstein recognize
this and so move on from anti-realism to some form of idealism, to
the view that our thinking it makes it so? (This is a possibility
W right canvasses, although he is by no means totally hostile to the
base class idea.)8
AH these exceedingly difficult questions are, fortunately, ones
with which we need not concern ourselves. We need only ask what
the connection is between the core anti-realist thought outlined
above and the hostility to m irroring realism sketched in section 7.1,
and elaborated in section 8.1. Is the former a corollary or natural
ally o f the latter? Are they independent? O r arc they (as I shall try
to show) opposed? M y suggestion will be that wc have got W itt­
genstein seriously wrong if we see ‘anti-realism ’ as any central
strand in his thought. His concerns arc, it seems to me, quite other.
And the rejection of m irroring realism , far from entailing anti-
realism , provides a perspective from which the fundamental anti-
realist argum ents are seen to be unpersuasivc. To see W ittgenstein
as an anti-realist is to sec him as, at some level, in sympathy with
empiricism, whereas, if section 7.1 is right, he invites us to make a
radical break with th at way of approaching things.
Before we em bark on the discussion let us be quite clear that the
issue now to be considered is quite different from that of the
dcfensibility of what I have called ‘m inim al realism’. T h at position
insists upon the lack of a guaranteed move from what wc think to
be the case to how things arc. But the ‘realism’ we arc now
considering consists in the denial of a guaranteed move from how
things are to be the possibility of our knowing that they are so. Both
views insist upon separation o f the ontological from the cpistcmo-
logical or psychological. (This is perhaps the hallmark of any
‘realism ’.) But the epistcmological end is different in the two cases.
In one it is thinking or being convinced that so and so, in the other
it is possibly knowing that so and so. And the direction of depen­
dence denied in the two positions is different. It may be that the two
forms o f realism arc im portantly connected. But to show this would
require a radically different set of argum ents from any considered
in detail here. For the rest of this section ‘realism’ and ‘anti-
realism ’ will be used in the D um m cttian senses unless there is
specific indication to the contrary.
180 Interpretations and Misinterpretations
T he anti-realist position is supported by a variety o f argum ents.
As elegantly assembled by W right in a recent discussion, there are
at least three main strands: considerations having to do with
acquisition, with m anifestation and with norm ativity.9 T he first
line of thought presses the question ‘How is a person supposed to
arrive at a conception of a verification transcendent or unevidenced
state of allairs when (obviously) all he or she actually experiences
are detectable or evidenced states?’ T he second poses the challenge
of making intelligible how one person could know, or even reason­
ably conjecture, o f another that he or she associated verification
transcendent truth conditions with some sentence of the language.
T he claim of the manifestation argum ent is th at we cannot give any
defensible account of this. And consequently, even if the acquisition
problem can be dealt with, the m anifestation considerations show
that such grasp of verification transcendent tru th conditions could
not be invoked as a central element in any account o f the workings
of a public language. T h e norm ativity argum ent claims th a t m ean­
ing is normative, in th a t grasp of it involves grasping a set of
constraints to which use can intelligibly be designed to conform
and, it adds, grasp o f supposed verification transcendent truth
conditions could provide no such constraints. In all o f these, but
most particularly in the second and third, we are supposed to find
some distinctively W ittgensteinian elements — emphasis on p u b ­
licity and on a link between m eaning and use. And it is here th at we
arc supposed to see W ittgenstein as probing deeper than traditional
empiricists, although arriving at conclusions which would not be
uncongenial to them and on the basis of argum ents no t totally
different. O ne slogan in which these thoughts are presented is
‘understanding is a practical ability’. W right formulates m atters
this way and takes it be roughly equivalent to the D um m ettian
‘meaning cannot transcend use’. Both slogans clearly have some
W ittgensteinian resonances.
How should we respond to the three arguments? I would suggest
that they are not equally powerful. I f the manifestation challenge
could be met then the acquisition problem , although it m ight still
dem and treatment, would hardly be enough 011 its own to under­
mine claims to a ‘realist’ understanding. T he norm ativity argu­
ment also seems less than compelling and I shall discuss briefly
why this is so.
 Interpretations and Misinterpretations 181
W right starts this argum ent by m aking some plausible claims
about intention - namely th at the notion gets no grip where the
agent knows that he or she can have no rational beliefs about how
to achieve the supposed end or about whether it has been attained
or n ot.10 Let us agree th a t these things are so. W hat then follows
about verification transcendent sentences which are adm itted to be
such? All that follows is th a t a rational agent will not intend to
assert one. W right takes this to lie in tension with the idea that
meaning is norm ative and places some constraint on correct use.
But this dictum can be taken in two ways. U ncontentiously it
claims (among other things) th a t if S m eans that p then an
assertion of S will be correct, if and only if p. This truistic con­
ditional connection (If I were to assert S then it would he true
iff. . . ) is not unsettled by the discovery that rationally I have no
way of finding out w hether S and of so putting myself in a position
to assert it. We have a tension only if we take the dictum about
norm ativity in a stronger sense and tie it in with views about
understanding being a practical ability - namely as saying that
grasp of the meaning o f a sentence m ust involve grasp of an
intelligible project of asserting that sentence. But why should we
say this? T o deny it is not to deny that understanding is practical in
the sense of having some actions in which it can manifest itself;
there arc plenty of things that I can do in virtue of my understand­
ing of a verification transcendent sentence - e.g. lament that we
will never be in a position to assert it, judge that another foolishly
has asserted it, employ it in hypothetical reasoning, and so forth.
If I am right in saying that the acquisition and normativity
argum ents are, for various reasons, m arginal then it appears that
the manifestation argum ent is the central one. T he challenge is to
set out some account of w hat it could be to grasp verilication
transcendent truth conditions, given that we accept that grasp
cannot transcend use and must be a practical capacity manifestable
in action. Let us suppose th at someone has no grasp of verification
transcendent truth conditions for a sentence but has instead grasp
of some suitable and complex set o f rules for varying responses to
detectable conditions. If the set of rules has been properly chosen it
seems th at there could be no difference between what this person
does and w hat someone with the verification transcendent grasp
does; each reacts the same way to all the detectable differences in
182 Interpretations and Misinterpretations
the world. So, practically speaking, they do ju st the same things.
And if m eaning cannot transcend use, m ust not the use, and so the
m eaning, be the same?
T he first move to make in questioning this persuasive line of
thought is to ask how we should understand ‘meaning cannot
transcend use’ or ‘understanding is a practical ability’, if we regard
them as sum m arizing some W ittgcnsteinian line of thought. There
are two ways in which the idea can be interpreted, only one o f
which points in the anti-realist direction.
T here is a way of hearing the slogans in which they express
something which need not be controversial and may seem unexcit­
ing. Suppose someone understands that a sentence S means that p.
Then lie or she will, ipso facto, have the ability to use S to assert that
p and to use suitable relations of S to ask w hether p, express the
speculation that p, claim that if p then q, and so forth. M oreover
someone who has these abilities to assert that p, ask w hether p etc.
by S and its relations understands that S incans that p. ‘U nder­
standing’ is an ability word; there is no gap between possession of
the ability to do the actions described and understanding the
sentence. Pointing out this linkage — although it indicates some­
thing which may seem obvious to us now - is by no means useless.
W hat it stands opposed to is the idea that understanding is some­
thing (like a catcgorial base) from which an ability flows, some­
thing which explains or underlies the ability. T he point now being
stressed is th at understanding is not something o f this character,
but the ability itself.
I do not want to suggest th at these last remarks are entirely clear.
W hat it is to ‘explain’ an ability, and how we arc to conceive of
abilities are im portant questions. But it is surely plausible to
suppose that the line of thought opened u p here docs not necessarily
tend in the direction the anti-realist wants. We arc invited to ask
why we have ability concepts, w hether they are different from
dispositional ones, how abilities might relate to categorical bases
and the like. It is not obvious that all plausible answers to these
questions constrain us to views about limitations on the intentional
content th at an item, w hether it be an ability, like understanding,
or an act, like speaking, can be said to have. In particular it is far
from clear th a t verification transcendent tru th conditions will have
 Interpretations and Misinterpretations 183
to be expelled from am ong the admissible intentional contents. My
suggestion would be th a t this unflam boyant interpretation is the
correct reading o f W ittgenstein. Philosophically exciting conse­
quences follow, in virtue o ffu rth cr links the slogan may have to the
themes pursued in section 7 .1 and in the first section of this chapter
— namely to ideas about the failure of mirroring realism and the
interdependence of concepts and interests. But it has yet to be
shown that any o f these move in an anti-realist direction.
This interpretation, however, will not do for the anti-realist. He
wishes to get more mileage out of the idea o f ‘use’ or ‘practical
ability’ than is here allowed for. Let us consider the passage in
which W right poses the difficulty.

T he argum ent is directly that the realist misdescribes what

understanding a statem ent of the relevant sort, consist in, or
better th at he overdescribes it, finding it to involve more than there
is any w arrant to suppose . . . According to the realist, under­
standing a statem ent of one of the relevant kinds consists in
knowing that a certain sort of potentially evidence transcendent
state of affairs both suffices and is necessary for its truth. Ilow
can that account be viewed as a description of any practical, ability
of use? No doubt some one who understands such a statem ent
can be expected to have m any relevant practical abilities. He will
be able to appraise evidence Tor or against it, should any be
available or to recognize that no information in his possession
bears on it. He will be able to recognize at least some of its logical
consequences and to identify beliefs from which comm itment to
it would follow. And he will, presumably, show himself sensitive
to conditions under which it is appropriate to ascribe prep­
ositional attitudes em bedding the statem ent to himself and
others . . . But the headings under which his practical abilities
fall so far make no m ention of grasp of evidence transcendent
truth conditions. Is it the realist’s view th at such mention is
somehow there implicitly? I f so, let the implication be brought
o u t.11

T he challenge is to set out the practical ability involved in under­

standing some S to m ean that p, with imagined evidence transcen­
dent p. But the thrust of the earlier remarks is that we answer that
184 Interpretations mid Misinterpretations
challenge by citing exactly the sam e things as we would lor non­
evidence transcendent content. W e say, that is, that understanding
that S means that p is to have the ability to remark on others’
assertions that p, debate on w hat would follow if p, and so forth.
Why are these not enough? Are these not practical efforts we have
described? They arc intentional activities, carried out purposely
under certain descriptions. Is this not enough to make them ‘practi­
cal’? From W right’s point o f view, something vital has clearly been
left out: ‘the headings under which his practical abilities fall so far
involve no mention o f grasp o f evidence transcendent truth condi­
tions’, he says. But they do, says the realist. W hether or not the
proposition that p is evidence transcendent, exercising the practical
ability to recognize its expression by others, reason with it, etc.
must involve grasp of the proposition that p - i.e. of the evidence
transcendent stale of affairs, if th a t is w hat the proposition specifies.
Clearly the ships are passing each other in the night here. W hat
is the worry which the realist’s apparently bland rejoinder ignores?
Note the recurrence in W right’s exposition of evidential notions;
one who understands is ‘able to appraise evidence’ and ‘sensitive to
conditions under which it is appropriate to ascribe propositional
attitudes’. W hat this strongly suggests is that, for W right, some­
thing counts as a ‘practical’ ability only if it issues in actions which
are done under descriptions showing them to be responses to, or
productive ofj observable aspects of the world. W right’s ‘practical’
is not just ‘intentional’ or ‘purposive’ but something closer to
‘makes a detectable difference’. W right’s worry is how one could
justifiably get from a description o f w hat is done which calls only
upon responses to and changes in detectable states of affairs to
another description which involves im puting grasp of w hat it would
be for a state o f affairs to obtain undetectably. But the realist does
not accept this starting point. For him, the practical activities are
from the start recognized and described in terms which im pute such
grasp.
How should we respond to this dispute? Before considering this
question, I shall pause to say som ething about W ittgenstein’s use of
the term 'criterion’. T he m atter is relevant because it is clear th at
W ittgenstein’s linkage of m eaning with use is im portantly bound
up with his taking ‘use’ to be the criterion o f meaning. His rem arks
to this effect form some of the evidence for an anti-realist interpreta­
 Interpretations and Misinterpretations 185
tion. But. is there any clear indication in the way W ittgenstein uses
the ‘criterion’ idea that such a reading is correct? 1 shall suggest
that there is not.
Let us look at the uses o f ‘criterion’ in the text of the Philosophical
Investigations. We may first extract all those cases where he speaks of
use (or application or the like) as the criterion of meaning, since
they are the ones whose interpretation is under scrutiny. We. may
also set aside cases where he asks what the criteria for such and
such arc (w ithout supplying an answer) or remarks that the criteria
are more difficult to lix than one would imagine or the like. We are
left witli a handful of cases in which he gives examples of criteria
not directly concerned with meaning. I quote them here:

344 O u r criterion lor someone’s saying som ething to himself is

w hat lie tells us and the rest of his behaviour.
377 W hat is the criterion for the sameness of two images? . . .
For me, when it is someone elsc’s image: what he says and does.
542 ‘But the point is, they felt to him like the words o f a
language he knew well.’ - Yes: a criterion for that is that he later
said ju st that.
625 ‘How do you know that you have raised your arm ?’ — ‘I
feel it’. So what you recognize is the feeling? And are you certain
that you recognize it right? - You are certain that you have
raised your arm: isn’t this the criterion, the measure, of the
recognition?
633 ‘You were interrupted a while ago; do you still know what
you were going to say?’ — If I do know now, and say it —does
that mean that I had already thought it before, only not said it?
No, Unless you take the certainty with which I continue the
interrupted sentence as a criterion o f the thought’s already
having been completed at that time.
W hat does it mean for me to look at a draw ing in descriptive
geometry and say: ‘I know that the line appears again here, but 1
can’t see it like th a t’? Does it simply mean a lack of familiarity
with operating with the drawing; that I don’t ‘know my way about’
too well? —This familiarity is certainly one of our criteria. W hat
186 Interpretations and Misinterpretations
tells us that someone is seeing the drawing thrcc-dimcnsionally
is a certain kind of-‘knowing one’s way about’. C ertain gestures,
for instance, which indicate the three-dimensional relations: fine
shades of behaviour. (P art II, p. 203.)
Let us assume there was a inan who always guessed right w hat I
was saying to myself in my thoughts . . . But what is the criterion
for his guessing right? W ell, I am a truthful person and I confess
th at he has guessed right. (P art II, p. 222.)
O ne thing which is striking about all of these is that there is not a
single one of them in which the item cited as a criterion is a mere
piece of behaviour; all of the criteria arc actions, w hat people say or
do, or (in two cases) their being certain. So a behaviourist view of
W ittgenstein, with ‘behaviour’ somehow neutral between bodily
m ovement and action, receives no support at all.
B ut what is really im portant, for our current purposes, is th at
none of the items cited as a criterion has a clearly less problem atic
or less complex intentional content (from the anti-realist point of
view) than the item for which it is said to be a criterion. In all cases,
the criterion is a contentful item whose content is the same as, o r at
least out of the same conceptual basket as, the content o f the item
for which it is criterial. 'l'hus the criterion for the m ind-rcader’s
getting my thoughts right is th a t I say that he did so; the criterion
for seeing a drawing in a certain w ay is that I make gestures which
indicate three-dimensional relations.
I f we transfer the structure of these models to the m eaning and
use case we will have the thought th at the criterion for someone
m eaning that p by some sentence is that lie (for example) uses it to
say that p. In other words we have a version of the (from W right’s
point o f view) bland m eaning/use link and a correspondingly
non-committal sense of ‘practical’.
W hat this interpretation will seem to leave completely opaque is
w hat epistcinological use ‘criteria’ can have. C riteria are things by
which we judge of other things. W hatever kind of items they arc
taken to be, and whatever account we give of knowledge, this seems
to be clear. They m ust therefore be cpistcmologically available
when the things for which they are criteria arc not. T his is the
thought which pushes one in the direction of giving prom inence to
 Interpretations and Misinterpretations 187
the detectable and provokes suspicion of the evidence transcendent.
B ut w hat is involved in being ‘epistemologically available’? As with
the question about ‘practical ability’ we can cither give this ques­
tion an unexciting seeming answer or offer something more con­
troversial.
O n the unexciting front we may note that some things arc not
(directly) available epistemologically because they belong to the
wrong category of item to be observed. Dispositions and abilities fit
this description. H aving them is a ‘state’ in some broad sense, but
not a state being in which involves the exhibition of some observ­
able appearance. T hey can thus be epistemologically accessed only
via some exercise or realization, which will take the form of an
occurrence or episode. Such occurrences or episodes are of the right
category to be observed. But nothing wc have said so far requires
that the occurrences or episodes be describablc in terms of concepts
independent of th at of the state for which they arc criteria.12
T he anti-realist would like to suppose that W ittgenstein goes
beyond this ‘gram m atical’ style of remark. T he idea would be that
what is observed should be available to inspection. And in terms of
his notion o f w hat sort of things could be available for inspection
some characterizations o f events (e.g. those which invoke the
notion of the disposition of ability it manifests) will seem qucstion-
begging. B ut worries of this kind, for example about how I could
possibly tell that a certain gesture indicates a three-dimensional
relationship perceived in a flat picture, do not seem to pre-occupy
W ittgenstein at all. Wc have, then, no evidence that he thinks of the
cpistemological availability of criteria in anything other than the
first and extremely non-conunittal sense.
T he upshot of the reflections o f the past few pages is that, as far
as interpretation of W ittgenstein is concerned, it is far from clear
that we can pin anti-realist thoughts 011 him. But of course this
leaves the plausibility o f the anti-realist position on its own grounds
uncxam incd. And it is to some brief remarks on this that I now
turn.
Let us go hack to the thought that understanding is possession of
a practical ability and to the centrality of the manifestation chal­
lenge. The realist thinks that one finds out whether someone under­
stands S to mean that p (where it may be verification transcendent
188 Interpretations and Misinterpretations
whether p) by seeing w hat actions he does with the sentence S. I f he
uses it to speculate w hether p, express the view th a t someone else
would like it to be the case th at p and so forth, then we have the
relevant evidence to say th at he does so understand it. T he anti­
realist objects that one cannot just observe th at someone expresses
the speculation w hether p and so on, where it is evidence transcen­
dent whether p. All one can do is see th a t S is used in so and so
detectable circum stances.13
Could the underm ining of m irroring realism provide any support
for the realist? O ne might argue th at the idea o f w hat one can
‘observe’, what can be ‘manifested’, to a person is crucially unclear.
W hat is observable to a given person will depend upon the concep­
tual apparatus which that person brings to the business of observa­
tion. T he beauty of the music or the brilliance o f a chess move
might be apparent to the expert b u t unobservable by the beginner.
So, one might say, why should not grip of verification transcendent
truth conditions be apparent in someone’s use of language, when
the interpreter is him self or herself equipped with the relevant
‘realist’ concepts?14 Is this not the nub of the anti-m irroring view -
that the thinker’s nature and equipm ent contribute vitally to the
content of any thought arising from a given confrontation?
Put thus baldly, however, the thought will not move the anti­
realist at all. From his point o f view the im portant thing is that we
all agree that the sentence whose m eaning is in question is such that
no evidence bears on its tru th or falsity; we agree that it has
verification transcendent truth conditions (if that is the right way of
speaking at all). IIow, then, can all these rem arks about different
things being verifiable or observable by different people provide
any grip on what ability someone has in virtue of understanding a
problematic sentence - a sentence which is on all hands adm itted
to be unfitted for the expression of the outcome o f some verification
process? Surely, the anti-realist will urge, anything th at a person
does with such a sentence could be explained w ithout postulating
grasp of verification transcendent tru th conditions, namely by
saying that its user has grasped some (complex) set of rules for its
use under evidenced conditions. To suppose th a t there is something
else going 011 is to move unw arrantably beyond w hat we can
 Interpretations and Misinterpretations 189
properly defend on the d ata available, because (to stress the
truistic) it m ust be adm itted by everyone that, as far as the
circumstances o f utterance are concerned, we can only detect what
is detectable; and so, it seems, all we know about the utterance is
that it occurs in such and such detectable circumstances. Thus it
may seem that the rejection o f mirroring realism - read ju st as the
idea that approaching the same situation with different concepts
will yield different judgem ents - is quite powerless to ward ofT the
manifestation challenge.,s
But this underestim ates the disruptive potential of the rejection
of m irroring realism, while at the same time slightly misrepresent­
ing its central theme. In the current context its im portance lies not
ju st in the thought th a t w hat you see in a situation depends upon
the conceptual equipm ent you bring to it. T h e im portance is rather
this: O nce we have rejected m irroring realism then we have
nothing pushing us in the direction of empiricism, no reason to be
attracted to the idea th a t we have concepts only of things which can
reveal themselves to us or impress themselves upon us. In conse­
quence we need have no predisposition to sym pathy with the root
idea of anti-realism , namely that there m ust be cpistemic con­
straints on ideas of m eaning and truth. T he acquisition challenge
entirely loses its force. W e do not get our concepts from anywhere,
in the sense of extracting them from, or having them supplied by,
presented instances. T here is thus no difficulty in principle (m ir­
roring view and attendant empiricism having been jettisoned) in
supposing that we have concepts of which we see that we can never
apply them categorically in rational assertive judgem ents.16 Posses­
sion of such concepts is a m atter o f having the interests and living
the life in which they have their role. And now we see th at the
manifestation challenge loses its force as well. K nowing that some­
one, oneself or another, has such concepts is a m atter of knowing
th at the person has the relevant interests and lives the relevant life.
And why should this be supposed insuperably difficult?
T he anti-realist may not be much impressed by this last rhetori­
cal question. He will dem and to know w hat it could possibly be
about how people talk and act which, if someone know of it, would
show him that the subject grasped evidence transcendent truth
190 Interpretations and Misinterpretations
conditions. He will reiterate that all that is there to be seen arc
responses to detectable conditions, responses which could be ex­
plained more economically by grasp of assertibility conditions.
W hat he shows by this protest is that he takes it as obvious th at
claims about meaning, if cpistcmologically respectable, m ust be
based on, explicable as rationally derived from, d a ta about uninter-
prctcd utterances observed when occurring as responses to observ­
able circumstances. I f one were a mirroring realist (particularly of
an empiricist kind) this view about how judgem ents o f m eaning are
to be established would be natural, indeed perhaps m andatory. O n
that view one supposes that all our notions are derived from
situations in which the relevant properties or relations are instan­
tiated and impress themselves, directly or indirectly, upon us.
M eaning is such a notion and is a m atter of some kind of
corrcspondcnce-stylc relation between labels and features of the
world. So how could one establish that it obtained except by
observing both rclata in the required configuration? At the thinker’s
end o f the relation all one can establish is th a t an assertively
intended response occurs. The burden o f determ ining the content
o f the response falls entirely onto the w orld.17
But if this whole ‘trying to label the independently given facts’
picture o f meaning is the wrong one then the conception o f evidence
about meaning to which it gives rise will be the wrong one too. My
suggestion is that it is implicit attachm ent to this picture which
provides much of the motive lorcc Tor anti-realism . O nce we have
got free o f the tem ptations o f this picture, we see that nothing has
yet been found in the rival non-m irroring account of w hat it is to
have a concept (including the account of w hat it m ight be to have
the concept of meaning) to give us any reason to suppose th at there
is some special difficulty in crediting people with grasp o f verifica­
tion transcendent truth conditions.
Let me adm it at once that the above considerations are entirely
negative. They have been aimed at underm ining a possible source
of support for anti-realism and not at establishing a positive case
for realism. T he anti-realist is still perfectly entitled to ask for an
account of a form o f life in winch the notion o f the verification
transcendent plays an im portant part, and no such account has
been supplied. M y own hunch is that there are two ideas which are
 Interpretations and Misinterpretations 191
fundam ental to our way of looking at things which do indeed give
D um m ettian realism an im portant place in our view of the world.
These are, first, the idea of the world as separate from us and,
second, the idea of the dependence of our awareness of any item in
it upon the, only contingently related, existence of other things of
the same kind. (For example my awareness of the tree over there is
dependent upon the existence of light rays and the non-existence of
intervening objects. My awareness of the existence o f Napoleon is
dependent on the existence of certain pieces of paper with ink on
them , etc.) It is difficult to see how these two ideas could be
prevented from combining to deliver the idea of the undetectable
event, undetectable because of the non-cxistcnce of the requisite
finks between it and us. It may be that in this area there is an
argum ent from the ‘separateness’ of the world invoked in minimal
realism to the claims of D um m ettian realism. But I shall not pursue
these difficult issues here.'8
T here is however one extreme version of Dummctt-stylc realism
which a W ittgcnsteinian will certainly oppose. This is the idea
(called by Putnam ‘metaphysical realism’) that f/w c had got a
complete and, from inside, perfect world view, which satisfied all
dem ands on world views (experimental adequacy, elegance, sim­
plicity etc.) then it would nevertheless be intelligible to suppose that
it was false.19 T he problem with this ‘realist’ thought is not that, on
reflection, we sec that there can be no sense in the idea that such a
perfect theory was false. If that were so then we might, at this limit,
have to abandon even our minimal realism and adm it the coinci­
dence of the ontological fact of w hat is and the cpistcmological fact
of the perfection of the theory. T he problem with the idea is
however much simpler; it is that we can make no real sense of the
antecedent. T he idea of a theory which meets and could be known
to meet all possible dem ands presupposes the idea of a totality of
dem ands on a theory - both of kinds of demand and of members of
each kind. B ut it is entirely mysterious how one could give sub­
stance to any such notion.
 9
Interests, Activities and
Meanings

{). I PR O O F AN 1) N EW C O N C E PTS

In the earlier part of this chapter J shall make some suggestions

about W ittgenstein’s views on the philosophy o f m athem atics. 'Flic
point is nol to provide a full account of these m atters - which
would require a much more extended treatm ent - but to give a
little more substance to the schematic ideas about concepts, interests
and their interdependence and thus to put us in a belter position to
appreciate W ittgenstein’s views about necessity, possibility and
"other forms of life’. T his in turn will stand us in good stead when
we return, in the final section, to the question o f realism about
meaning.
Let us start by considering one of the puzzling things that
Wittgenstein says in the Remarks on the Foundations o f Mathematics
(hereafter RFM ), nam ely that a proof creates new concepts.1This
is a thought to which he returns throughout the period 1937-44
spanned by the writings in RFM and so seems to represent some­
thing important for him .2 He also uses the notions of decision and
invention in this context, speaking of the m athem atician as an
inventor and introducing the idea o f the discovery of a proof as
winning through to a decision. A nother well-known strand in his
thought on m athematics is hostility to Platonism - the idea th a t in
doing mathematics wc are doing a sort of n atural science where,
using some special faculty, we peer into a crystalline realm of
abstract objects forever fixed in necessary relations to each other.
T he conjunction of this anti-Platonism with the talk of new
concepts and decisions has led many com m entators, very naturally,
 Interests, Activities and Meanings 193
to the idea that W ittgenstein is some kind ol' conventionalist,
probably a radical conventionalist, about necessary tru th .9
T he direct textual evidence lor this is far from-conclusive. If we
consider the way W ittgenstein wishes to invoke the decision idea,
we find that his wording is, predominantly, oblique and suggestive
rather than directly assertive. For example Mi would almost be more
correct to say, not that an intuition was needed at every stage, but
that a new decision was needed at every stage.’1( )r ‘ Why should I not
say: in the proof I have won through to a decision’ (my italics)’’ So
he docs not say, baldly and in propria persona that all or oven some
judgem ents are math: at will or are the outcome of decisions.
R ather he invites us to play with the notion, misleading though he
implies it is, as an antidote to some other even more unfortunate
picture of the m atter. It is im portant also to remember and set
against these passages other remarks where he is speaking baldly
and in propria persona and where he links the idea of compulsion and
rule-following, e.g. Ml'a rule does not compel you tlam you aren’t
following a rule’, or ‘W hen I say ” 1 deride spontaneously” nat­
urally that does not mean I consider which num ber would really be
the best one here and then plum p l o r . . . ”■This latter remark
shows clearly that it cannot be decision in any ordinary sense, a
decision where one is aware of alternatives, that he has in mind.
Let its note also that conventionalism can be seen either as an
empiricist response to the epislemological and metaphysical prob­
lems of necessity or (and as I suggested in section O.'l more
defensibly) as a component of a thoroughgoing pragm atism . But 1
have tried earlier to suggest that W ittgenstein is neither an empiri­
cist nor a pragm atist. Arid if these earlier interpretive moves were
in the right direction then wc have further reason not to take the
talk o f ‘decision’ in a direct conventionalist way.
W hat then is going on with these invocations of decision? Clearly
it is in line with our earlier proposed reading to see them as
expressions of hostility to m irroring realism. W ittgenstein is, in
them , emphasizing the clement ol’ activity (spontaneity, practice,
form oflife) in concept-use. Thus taken they lit in well with the line
of thought sketched in section 7.1. Passages where W ittgenstein
talks of the limits of empiricism and of the difficulty of combining
empiricism and realism also make good sense in this light.7
194 Interests, Activities and Meanings
But if conventionalism is not w hat W ittgenstein is after, w hat arc
wc to make of his hostility to Platonism and of his talk linking proof
and change of concept? O n one reading, that favoured by Crispin
W right, the hostility to Platonism is an implication of a general
scepticism about the notion of meaning. O n this view the rule-
following discussion (together, perhaps, with other considerations
of the kind mentioned in section 7.2) gives us a picture in which we
see th at there could be no such thing as the fact that a person m eant
one thing rather than another at a given time. So, in particular, it
makes no sense to suppose that there is such a thing as ‘grasp of a
concept’ which could commit one to recognizing necessary or
conceptual truths about a certain subject m atter.
But from the point of view 1 propose this is not at all an attractive
line of interpretation. My (heme is that W ittgenstein’s rejection of
the general m irroring realist outlook allows us to defend a (more
sensible and minimal) version o f realism about meaning. It is a
welcome fact, then, that the meaning-sceptical reading is (as
W right notes) in serious tension with the other element mentioned
above - namely the idea of proof as modifying concepts.” If all talk
of determ inate meaning is illegitimate how can we make sense of
the idea of change of meaning? O n the view of W ittgenstein which
sees him as a radical sceptic about meaning he is not entitled to the
idea that there is anything to be changed.
So let us look afresh at anti-Platonism and at the idea that proof
modifies concepts, starting with the latter. T here is, of course,
something paradoxical about it, however we read it. It seems to
deny th at we can understand some m athem atical conjecture per­
fectly well, look for a proof o f it, find one and thus prove exactly
that statem ent which wc set out to prove. If the proof changes the
concepts —i.e. the meanings of the words in which we express the
object o f our interest - then wc have not proved w hat we set out to
but some other statem ent instead. T he paradox is made particu­
larly acute —indeed so acute as to be intolerable - if we suppose
not only that proof of a sentence alters its m eaning but also that the
meaning it had before the proof was found remains available to be
expressed by some other sentence. For now it looks as if any
attem pted proof of a statem ent was bound to miss its m ark. We arc
going to prove a proposition different from the one which we set out
 Interests, Activities and. Meanings 195
to prove, even if, at the end o f the proof, wc express it by the sam e
words as wc used to express our original proposition. And the
originally questioned proposition rem ains unproven but still avail­
ab le to be wondered about. T h is story leaves it totally m ysterious
how the p roof is supposed to shift the m eaning o f the sentence -
given that w c can still entertain the original conjecture. H ow could
the p roof force us to abandon this meaning? And should w e not
have rejected as invalid any step o f the proof w hich tem pted us to
alter our understanding o f som e word or construction?9
T h e paradox how ever docs not becom e intolerably acute unless
w e m ake this extra assum ption - and wc need not m ake it. O ur
‘concept’ o f som ething m ay be altered, colloquially speaking, when
w e learn som ething new about that thing. And our earlier concept
o f the item , i.e. our body o f inform ation about it, no longer exists as
the sam e total body, because it has been superseded by the new
body. For Platonism , a m athem atical investigation tells us more
about the properties o f num bers, shapes and so forth, and so, in the
colloquial sense it m odifies our concepts. I f wc have a view o f
concepts which is bound up w ith som e sharp distinction between a
priori and a posteriori truths then acquisition o f em pirical informa­
tion about som ething will not count as ‘m odifying the concept’. But
even on this second and restricted sense the Platonist can still make
som e sense o f the conccpl-m odilication idea because the truths
revealed by proofs do precisely alter, by addition, our range o f a
priori know ledge and the set o f conditions recognized as necessary
or sufficient for the application o f a given word. So the Platonist
m ight adm it that m athem atical formulae do ‘m ean som ething
different’ to m e before and after the proof: my understanding after
grasping the proof is deeper and richer; I have seen m ore into the
nature o f things; I have unfolded w hat was im plicit in m y pre­
viously grasped concept. It w ould not, o f course, be congenial to
the Platonist to say that the concept, regarded as the abstract focus
o f m y som ew hat confused thoughts, has changed. But my concept,
in the sense o f im portant structural features o f my intellectual
relation w ith that object, has been altered. W right is struck by the
appearance o f incongruity in the idea that action in conformity
w ith or dictated by a concept - e.g. following a proof em ploying it
- should at the sam e tim e change the concept.10 But the Platonist
196 Interests, Activities and Meanings
conception of ‘unfolding’ w h at w as im plicit is adm irably designed
to allow us to rem ove the appearance.
T h is interpretation is, o f course, not available as a reading o f
W ittgenstein. H e is hostile to the ideas o f ‘unfolding’ and ‘the
im plicit’ which it em ploys. So how can w e im prove matters?

9.2 C O N C E P T U A L C H A N G E A N D D E T E R M IN A C Y O F
S E N SE

W e have just considered two m odels o f conceptual change — the

absurdity generating one, w here the original concept rem ained to
be used, and the Platonist one. W e are now in search o f a third,
w hich m ight provide som e insight into W ittgenstein’s view s. 1 shall
first consider a tem pting interpretation. It is deeply m istaken but
looking at it will provide som e useful clues.
Let us start by settin g in p lace an idea (w hich w e shall later
abandon as thoroughly u n W ittgcn stein ian ), nam ely that o f the
fixed totality o f m etaphysical possib ilities. A s w e have seen before
(in section 6.4) this idea is very naturally found as an elem en t in
the mirroring conception o f the w orld. O n one view about thought
(a view found in Frege and also in the early W ittgenstein) any
respectable thought m u st be fully determ inate. T h at is to say that,
for any possible w ay the w orld m ight be, it m ust be the case that
the thought would be true if things w ere that w ay or false if they
were. In grasping a m ean ing, on this conception, we em b ody a
function from possib le w ays the world m ight be to truth values.
T here is on this outlook on ly on e conceptual schem e - th e one
w hich recognizes those features the presence or absence o f w hich
differentiates and defines the various possibilities. But w e do not
have to suppose that w e now mark all those features explicitly w ith
separate words in our language. T h e concepts associated w ith our
words m ay represent com plexes o f sim pler elem ents. G iven this
view , there is plenty o f room for P latonistic unpacking o f w hat is
im plicit, but no room for conceptual change of an any m ore radical
kind.
H ow ever, one com paratively sm all m odification o f this story will
give us an account in w hich a notion of'conceptual change m akes
 Interests Activities and Meanings
, 197
sense. Suppose w e leave the set o f m etaphysical possibilities in
place but abandon (he dem and that possession o f an intelligible
thought m ust involve ability to classify every possible w ay the
world m ight be under one or other o f the headings ‘situation in
w hich the thought is true’ and ‘situation in w hich the thought is
false’. W ould-be thinkers are allow ed to be d oing som ething
respect-w orthy as long as they are equipped to respond to a lair
range o f situations o f the kinds they arc norm ally confronted w ith,
even if in other situations they w ould have to adm it that they did
not know w hether their thought w as true or false. H ere w e can
allow for conceptual change because we can allow different think­
ers, or the sam e thinker at different tim es, to em body different
incom plete classificatory dispositions.
Before proceeding further I shall sketch a possib le case o f con­
ceptual alteration w hich will give us som ething m ore concrete to
think about and m ay help us in d evelop ing a view o f the m ath­
em atical case." C onsider som eone w ho has been taught the use o f
colour vocabulary but only on the basis o f seeing objects in sun­
light. H e is now show n his fam iliar toy postbox, w hich he is used to
thinking o f as red, by the light o f a sodium lam p and he is asked
w h at colour it is. '[’here are two w ays he m ight respond. (There are
m any others as well, but I shall not discu ss them , since their
presence does not affect the argum ent.) H e m ight say that the
postbox w as no longer red but bail becom e w hite. O r he m ight say
that it w as still red but som ething w as m aking it look funny.
W e m ay reconstruct his puzzlem ent as arising from the fact that
he has been used to applying colour vocabulary in cases where two
features coincide. O n the one hand w e have the view that colour is
a stable property o f things and hence that how it was in the
im m ediate past (provided it has undergone no causal process
w hich could alter it) is adequate ground for attributing the sam e
colour now. O n the other hand w e have the view that one can
detect colour by looking. W hat the surprising experience show s is
that these w ays o f ascribing colours can conflict. So we may say
that colour ascription was earlier governed by two criteria, each
treated as both necessary and sufficient; the speaker w as thus not
equipped to deal with those possible situation s where the criteria
diverged.
198 Interests, Activities and Meanings
Faced w ith such situations, language learners m ay well be at first
n onplussed. But various m oves arc m ade to encourage them (as w e
w ould describe it with hindsight) to associate the word w ith one
criterion rather than another. And after a w hile the learners find
them selves proceeding in step w ith th e rest o f the linguistic com ­
m unity. G iven the picture o f the total set o f m etaphysical possi­
bilities, w hat has happened here is that the learners first had a
m ethod o f dealing with situations w here the two criteria were cither
both present or both absent but had no response ready for situ a­
tions w here they diverged. T h e teaching or training at the point o f
m u d dlcm cnt som ehow alters the function their understanding em ­
bodies so that they arc subsequently equipped to deal w ith a w ider
range o f possibilities.
It is a corollary o f the way we have described the m atter so far
that the training which resolves the m uddle is not a m atter o f
gettin g the child to think, in the sen se o f form ing and testing
h ypotheses, in terms o f concepts he already h as.12 E x hypotfmi he
docs not have the concepts to cope w ith the new situation and w e
have to supply him with them by som e appropriate non-cognitivc
procedure. I f this conclusion js found too olfcnsive w e can devise a
hybrid schem e w hich com bines elem en ts o f the original ‘total
detcrrninacy o f sen se’ model w ith our new ‘partial function’ m odel
by speaking o f tacit as opposed to explicit know ledge. O n this hybrid
view , w c say that the child did have separate concepts o f each o f the
tw o criteria but had also an inexplicit and deeply rooted theoretical
com m itm ent to their co-occurrence, so that the two concepts were
not differently marked at the surface o f the language. (T h e differ­
ences betw een the view s may appear m erely verbal. T h is outlook
finds it natural to think o f a concept as a disposition to produce a
classificatory response. T h e dispute is them only about w h at kind
o f disposition is required - 'e.g. w h ether a second-level disposition
to acquire a disposition is enough.)
It is worth noting that the sort o f ‘conceptual ch ange’ and
‘conceptu al difference’ that w e are trying to get a grip on in these
m odels is not that w hich is som etim es discussed w hen the question
o f the possibility o f different con cep tu al schem es is raised. T h e
possib ility canvassed in these other discu ssions is that o f a kind o f
 Interests, Activities and Meanings 199
thinking w hich is so different from ours that no m utual under­
standing at all is to be hoped for. V arious general argum ents have
been proposed (e.g. by D avid son ), the thrust o f w hich is that this is
a seriously defective notion; attem pts to defend relativism by its use
need not, it is said, worry u s.13 O n this line o f thought, if w e can
m ake som ething o f another creature’s thoughts and offer defensible
translations then that creature is operating with our conceptual
schem e.
O n the view o f w hat it is to have different concepts which we arc
now exploring, this last m entioned connection docs not hold. We
can m ake som ething o f the children’s thoughts and translate their
utterances. But the difference in their w ay o f looking at things, the
distinctness o f their conceptual schem e from ours, com es out, we
im agine, in the fact that none o f these translations is entirely
satisfactory.
U nderlying this contrast is a m ethodological difference. O ur
sense that none o f the translations is entirely satisfactory is bound
up with the fact that w e are prepared to recognize a greater variety
o f responses to circum stances than just whether those circum ­
stances provoke assent or d issent to a sentence. W e build in, for
exam ple, the idea that a circum stance m ight provoke an im pulse
both to assen l and to dissen t or, rclatcdly, that it m ight provoke a
particular kind o f surprise. T h e surprise is not ju st that o f discover­
ing that som ething one took to be the case is not the case but is
m ore akin to bew ilderm ent or ju st not knowing w hat to say. I f we
flatten out all this variety o f response to mere assent or dissent we
shall indeed end up w ith on ly two options - nam ely that those wc
seek to understand either have our conceptual schem e or have none
(that w c can m ake sense o f at le a st).14
Let us return to the m ain line o f thought and try the experim ent
o f view ing W ittgen stein ’s claim s about proof and new concepts in
m athem atics in the context o f our above m odel o f conceptual
change.
T h e idea w ill be that at certain points in learning to use m ath­
em atical vocabulary a person m ay find him self or herself in a
position where tw o criteria for the application o f som e description
conflict; he will be confused and bew ildered. T h e job o f the ‘p roof
200 Interests, Activities and Meanings
is to remove his uncertainty as to how to proceed by encouraging
the association o f the word with one o f the criteria rather than the
other.
H ere is an exam ple on w hich to try out this highly abstract
schem a. Suppose that I have learnt to use the num bers up to ‘20’
by being taught how to count the m em bers o f groups o f apples,
beaus etc., and to answ er questions o f the form ‘H ow m any so and
sos arc there here?’ Som etim es 1 count two groups and then put
them together and count the resulting larger group. W hen I put
together groups o f seven and four 1 usually get a group o f eleven.
T h en one day a disconcerting thing happens. I count a group
resulting from the putting together o f groups o f seven and four and
the result is not eleven but tw elve. W hat am 1 to say? Perhaps I say
‘H ere’s a strange thing; seven and four d o not make eleven in this
case. I wonder if that happens very often?’ T h is remark is not well
received and I am told ‘N onsense! You m ust have m iscou n ted .’
A nd, perhaps by exhibition o f this picture

{ hi } { m } i / {n r } // i

I am led to exclaim ‘O h I see. It m ust b e eleven .’

Som ething has changed in m e. Hut what? I f I assum e that I now
m ean exactly the sam e by ‘7’, ‘4 ’ etc. as I did before then I m ust
have discovered som ething expressible in term s o f those concepts.
But I have not discovered em pirically that, by and large, seven and
four m ake eleven. So (it seem s) 1 m ust have unearthed som ething
that was all along im plicit in m y concepts - a relation that
necessarily and antecedently held betw een the essences o f seven,
four and eleven.
But surely (says the old Platon ist A dam in all o f us) this is the
right response! H ow could seven and four not equal eleven? W hat
are the ‘two criteria’ w hich previously governed application o f the
numerals? W hy should we not su pp ose that there is only one, the
one we all understand, grasp o f w hich rem ains constant?
Let us go back to the slightly easier case o f colours to give
CUi'aCl'vCS a model to work from . Sup pose that I have applied colour
term s to som ething up to now on the basis o f a good look at it in
ordinary daylight. But I am already aw are, let us suppose, o f the
 Interests Activities and Meanings
, 201
possibility o f m aking m istakes; 1 know that people can be inatten­
tive ancl not look carefully enough; I know also that illness may
lead them lo perceive abnorm ally. So in reporting on the colour o ia
thing 1 am reporting on a public feature o f the world. (I em phasize
this in order to m ake clear that the question at issue here is not that
o f whether colour is only a feature o f private experience. T h e issue
concerns the possibility o f different w ays o f thinking Of som ething
like colour, given already (hat what w c think o f is a feature o f the
world, accessible to other perccivers.) It is a corollary o f this that
w hat counts as a ‘good look’ cannot be som ething fully definable by
reference to how things seem lo me, what procedures I seem lo
m yself to have carried out, etc. If I adm it, as in the im agined state 1
do, that features o f the world inter-relato in m any com p licated and
as yet unknown w ays (e.g. that I m ay suffer d elusions from as yet
unknown causes) then w hether 1 have had a good enough look to
pronounce definitively on the colour o f som ething will only be as
certain as w hether I can rightly discount all sources o f error.
N ow for the first tim e 1 encounter a red object under sodium
lighting and I have a problem . O n the one hand m y usual w ay o f
applying the ‘good look’ procedure yields one result — the post box
is white. O n the other hand I have, up to now, m et with changes in
the colours o f things only w hen they were produced by fairly drastic
or prolonged procedures (painting, boiling or ripening); colour has
been a stable feature o f objects. T h is current experience involves
colour change (if that is what it is) o f a very different sort from any I
have previously encountered.
T h ere are two w ays o f proceeding. O n e is to insist that the colour
has not changed and to search for som ething to add to the list o f
possible confusing circum stances. T h e other is to leave the ‘good
look’ conditions in practice m uch as before and to allow that the
object has inexplicably changed colour.
W hichever o f these courses I take I can still say ‘O n e tells w hat
colour a thing is by having a good look at it’. In that sense the
‘criterion’ for w h at colour a thing is has not changed. A nd neither
the w ay that I w ou ld go about teaching colour w ords to som eone
else nor my judgem ents about earlier cases need be affected. T h ese
points are worth rem arking on because som e have seen it as a
difficulty for the idea that our concepts get m odified (e.g. by a
202 Interests, Activities and Meanings
proof) that earlier verbal forms o f explanation ancl teaching by
exam p le arc not im pugned or replaced even after the concept is
su pp osedly m odified. T h e problem som e have detected w ith this is
that if a concept is m odified then w c w ould expect exactly these
thin gs, i.e. exam ples and forms o f explanation, to change as w e ll.15
B ut on the view wc arc now exploring this problem is an illusion
sin ce the new concepts arc acquired only via developm ent o f the
old. T h e m ethods o f initiation rem ain as before and also the verbal
clarifications. T h e concept-develop ing procedure is itself a continu­
ation o f w hat w as achieved by those m ethods; and in so far as it
alters understanding it docs so sim ultan eou sly lor all the concepts
linked together in the verbal statem ents o f criteria.
B oth o f the two possible courses outlined arc (as I have stressed
before) m oves in the public realm and involve treating colour as a
publicly accessible property. B ut they nevertheless operate 011
different assum ptions about w hat kind o f objective property colour
is. T o take the first view is to treat colour as a stable and persisting
feature o f objects but one w hich m ay not be as easily detectable as
w e m ight like. Colour reveals itself w hen one can get a good look,
but settin g up the circum stances for that look m ay present prob­
lem s. H ow ever, once one has establish ed a colour one can rely on it
stayin g the sam e in the short term , barring drastic treatm ent o f the
object. T o take the second view is to treat colour as (com paratively)
easily accessible. T ak ing a good look is som ething w c arc alm ost
alw ays in a position to do. But w c pay the price in that w hat w e
thus d etect cannot be guaranteed to persist.
Im agine that som eone finds the first view natural. H e m ay
express h im self in remarks such as these*. ‘H ow can a thing change
from red to w hite w ithout anything acting upon it?’ (N ote the
appearance o f m odal term s.) H e will find it difficult to hear the
remarks o f anyone w ho takes the second view in any other w ay than
as expressing em pirical m istakes or nonsense. Let us use ‘w hite (1 )’
to express the first person’s colour concept, and ‘w hite (2 )’ for the
secon d. T h en the second person will say o f the postbox in the
strange light ‘But it is w hite [m ean in g w hite (2)]; I sec that it is’
T h e first person will hear this as the false claim that it is w hite (1).
And if the second person remarks that lie docs not sec why colours
 Interests, Activities and Meanings 203
should n ot change w ithout cause the first will hear this as an
expression o f grotesque confusion.
T h e best w e can do in the first vocabulary to suggest the second
person’s idea to the first is to stress that use o f ‘looks w hite (1 )’ in
w hich it is used to report a pub lic fact about an item , the sam e sort
o f fact as that w e report about the two lines in the M ullcr-Lyer
illusion w hen w e say that the one looks longer than the other. T h is
explanation w ill not give exactly the right extension for the second
person’s use o f ‘w h ite’ and will m isrepresent his notion as a
com pound one, but it m ight push the ‘w hite (1 )’ user in the right
phenom enological direction.
Suppose then that w e do, by this or som e other m eans, give the
user o f the first conception som e inkling o f how the other is
em ploying colour term s, how w ill he respond? His view well m ay be
that, althou gh there is here a consistent policy o f application, one
w hich he could h im self follow , nevertheless there is no ‘real’ feature
o f the world b eing recognized. Part o f w hat he is saying in express­
ing this view is that he cannot see how anybody could find it
w orthw hile to bother w ith the classification; he him self uses colour
as a clue to the ripeness o f fruit, the heat o f the fire and so forth; the
other m an’s notion, he thinks, has no such useful links; as far as he
can sec it is hoked up and arbitrary. B ut perhaps he could be
persuaded to ch a n g e his m ind by b ein g encouraged to take up
painting or interior decorating. O r perhaps it could turn out that
application o f the other m an’s notion does, after all, provide useful
clues to ripeness or heat.
C an we find for the m athem atical case analogies to the things we
have said about colours? C onsider again my disconcerting experi­
ence with the groups o f seven, four and eleven (or twelve). W ith
reflective hin d sigh t w h at description can we give o f my puzzle­
ment? I have m astered the use o f num ber words where the criterion
o f application is careful counting o f the w hole group. C ounting is
w hat establishes num ber, ju st as looking is what establishes colour.
But I have also another governing assum ption about num bers,
w hich has never yet clashed w ith the counting criterion, nam ely
that num ber relations are perm anent. T h us if seven and four ever
m ake eleven then they alw ays do. H ow am I now to respond in the
204 Interests Activities and Meanings
,

disconcerting case? I could harden this perm anence assum ption

into a rule, insist that there m ust have been a m iscount and look
around for an explanation o f it.16 O r 1 could allow that seven and
four m ight som etim es com e to som eth in g other than eleven.
W c are not here to consider the easy w ays out, like supposing
that one o f the objects has divided into two or another has been
added between the countings that yielded seven and four and that
which yielded twelve. Let it be agreed on all hands that these things
have not happened - ju st as in the colours case w e did not seek to
make sense o f the strange experience or the divergent responses to
it by im agining that the two respondents had different view s about
whether or not som eone had nipped up and painted the post box.
But if I deny the easy w ay out how can I intelligibly say that I
had a groups o f seven and four, put them together and got a group
o f twelve? I f 1 say this do I not ju st reveal that I cannot have
properly understood ‘7’, ‘4’, ‘none have been ad d ed ’ and so forth?
It is essential to attend closely to w h at it is w c are being asked to
im agine here. Let us as before use ‘7 (1 )’, ‘7 (2 )’ etc. to express the
putatively different num erical concepts m anifested by the two
different responses. T h e question is not w hether w e can make sense
o f the idea that 7 (1) apples and 4 (1) apples do not m ake 11 (I)
apples. T h e answer to that question is that we cannot. B ut then
neither can the person w ho thinks in term s o f colour (1) concepts
make sense o f the idea that som eth in g should change its colour (1)
w ithout cause. T h is fact alone docs not sh ow that colour (2)
concepts are unintelligible or (m ore im portantly) arc not a possible
developm ent o f the colour concepts possessed before the strange
experience. So the real question about the m athem atical notions is
w hether there are num ber (2) concepts w h ich can be regarded as
possible extensions o f the original, and (it is suggested) not fully
developed, num ber com petence.
T h e best route to grasping ‘w hite (2 )’ w as via the idea ‘looks
w hite (1 )’ - where the latter w as understood as a public property
about w hich one m ay be m istaken. O n e cannot m ake sense o f
som ething being red (I) and w hite (1) at the sam e tim e, but one
can m ake sense o f the idea that som eth in g w h ich one know s to be
red (I) should nevertheless look w hite (1). T h a t is ju st w hat has
 Interests Activities and Meanings
, 205
happened in the case o f the post box in the sodium light. T h e
num erical analogue will be as follows: O n e cannot m ake sense o f a
group w hich contains exactly the sum o f seven and four and yet
also contains twelve. But w e can make sense o f the idea that group
w hich w e know to contain the sum o f seven and four (from previous
counting, together w ith the fact that nothing has been added etc.)
should nevertheless yield a count o f twelve. Perhaps it does so again
and again. W e make marks on the item s as w e count them to avoid
double counting; w e pin num bers on them; we cross-check on each
oth er.17 But they defeat all our attem pts to get a sensible count of
eleven.
H ow could this be? Perhaps, as we later discover, this particular
group o f objects m akes the air around it sh im m er so that it is
difficult to fixate properly; one item alw ays gets counted tw ice or
m arks and num bers on ob jects g el overlooked or m isread. O r
perhaps the group exerts a hypnotic influence on anyone attem pt­
ing to count it so that he or sh e alw ays om its the num eral *(>’. O r
p erh ap s. . .
W hat w e are searching for in trying to const ruct such stories is an
explanation o f why this group yields a count o f twelve. And this is the
property in w hich the user o f num ber (2) concepts is interested.
T h ose groups w hich from his point o f view go im portantly together
from our point o f view fall into two subgroups - nam ely those
w hich yield a count o f tw elve because they contain tw elve m em bers
and those w hich yield a count o f twelve because they are be­
w itched, shim m ering, h yp n otic or w hat not. W e are inclined to say
that the user o f num ber (2) concepts is m istaking shadow y and
parasitic features o f the w orld for what is really there. W e cannot
see w hat point there could be in so grouping item s, sin ce all the
things we are interested in d oing connect up (or at least so far
em pirically seem to connect up) with how m any things there are in
the group, specified in num ber (1) concepts, and not w ith how
m any things there seem to be. B ut we are invited to consider how
things m ight appear to us if w e had different interests or if the world
were em pirically different.
T h e suggestion about proof, then, is that w hat it does is provide a
stim ulus for us to develop the earlier num erical concepts (which
206 Interests, Activities and Meanings
have no provision lor responding to cases where the tw o criteria fall
apart) in the direction o f num ber (1) concepts rather than num ber
(2) concepts.

9.3 O T H E R F O R M S O F LIFE?

T h ere is m uch in the above account w hich m akes it attractive as an

interpretation o f W ittgenstein. It gives us an account o f how
‘proofs’ can lead to the acquisition o f new concepts and it seem s to
accom m odate thoughts about how different em pirical circum ­
stan ces and interests w ould m ake possession o f different concepts
intelligible. But there is nevertheless som ething fundam entally
w rong with it.
Let us rem em ber that w c have been proceeding all this tim e
against the background assum p tion that there exists a given total­
ity o f m etaphysically given possibilities. W c have been speaking o f
different w ays in w hich an ability to cope with som e sub set o f the
possibilities m ight be extended to bring a larger set w ithin its scope.
It is a clear im plication o f this that the sense o f inevitability about a
p r o o f - the idea that it sh ow s one w hat one had to say, - is illusion.
T h is is so w hether w c take the view that concept develop m en t is a
sort o flca rn iu g (and use the contrast betw een explicit and im plicit
know ledge) or w hether w e think that conceptual d evelop m ent is
m erely a sort o f non-cognitivc training w hich som ehow results in
cognitive abilities. Either way there is a real alternative d evelop ­
m ent to the one that actu ally occurred - even when that develop ­
m ent is triggered by a proof. Anti a clearsighted person is, at least
in principle and if he bothers to spend tim e in w orking it out,
capable o f com ing to see w hat the alternative w as. T h e p roof has
indeed induced him to adopt a certain way o f using num ber term s.
B ut it is clear that there is another way and his follow ing this one
rather than that other is a m atter o f choice and convenien ce, not o f
necessity.
A ll o f this m ight seem to lit well with w hat W ittgenstein says
about proof, especially if w c concentrate on those passages w hich
encourage a con ven tionalist interpretation. But there are serious
grounds for doub t w hether the w hole story w c have told docs give
 Interests Activities and Meanings
, 207
the W ittgensteinian view . W c saw earlier reason to doubt that he
w as a conventionalist. But even worse is the Pact that if one thing is
clear about W ittgenstein’s view s on m eaning it is that he is not
prepared to work w ith the idea o f a given totality o f m etaphysical
possibilities. The early section s o f the Philosophical Investigations
(w here h e discu sses sim plicity and com plexity and w hether rule
system s need to be com plete and the like) are not designed to
provide m aterials for rejecting total detcrm inacy o f sense in favour
o f the sort o f theory o f ‘perm issible vagueness’ which wc have been
discu ssing. Rather they provide m aterial for throwing out the
w h ole picture underlying the detcrm inacy o f sense view . T hese
passages form part o f the discu ssion o f ostensive definition and
related topics w hich, on the interpretation advanced earlier (in
section 6.1) arc in turn part o f the rule-follow ing considerations, the
upshot o f w hich is the rejection o f m irroring realism. But the ‘given
totality o f possib ilities’ w c have been considering is precisely that
defined by the combinations o f the supposed ‘real’ properties. And if
the conception o f such real properties is an illusion then so too is the
conception o f the given totality which they contribute to defining.
O n e w ay o f seein g the incongruity o f the picture o f W ittgenstein
w c have been trying to build up is to consider the following. As long
as the ‘total set o f possib ilities’ is in place wc m ay make room for
‘p r o o f as a tool o f som e con ven tionalist type o f choice for sharpen­
ing up not fully determ inate notions. But the possibility o f
Platonist-style p roof still rem ains in those eases where our concepts
arc adequately sharp. T h ere is n othing in the picture to rule out the
idea that som etim es wc do em body a fully determ inate function, for
exam p le to som e conjunction o f properties. I f w c did, then there
w ou ld be room for an activity o f reflective unpacking in which the
im plications o f this were brought out in thought experim ents and
hitherto unrealized necessary connections noticed. T h is is surely
not a W ittgensteinian position. T o put things at their sim plest, the
idea o f concept possession as the em bodying o f som e psychic
m echanism , in virtue o f w h ich a person is hooked up (perhaps in
determ inate w ays as yet unclear to him or her) to som e predefined
set o f possibilities is exactly the tem pting m isconception from
w h ich the W ittgensteinian reflections are designed to free us.
S o som ething has gone badly w rong. We need, if we can, to
208 Interests, Activities and Meanings
rescue what in the earlier discussion seem ed illum inatin g for the
understanding o f W ittgenstein and to disen tangle it from the
unsatisfactory m etaphysical setting.
O ne illum inating thing seem ed to he the reflections on the kinds
o f circum stances and interests w hich w ould m ake other concepts
appropriate. Wc should, I suggest, accept the cases roughly sketched
as instances o f w hat W ittgenstein w ants to bring to our attention.
H e does wish us to entertain the idea that som eone m ight, for
exam ple, respond to the carrying ou t o f a counting procedure or to
scanning through a p roof by regarding them as som e kind o f
experim ent, rather than as pointing to necessities. And lie w ishes us
to try to im agine how the world m ight look to such a person and
w hat features o f the situation he is noticing. But let us look again at
the counting case and ask ourselves w hether, or better, in what
w ay, we can make sen se o f the supposed alternative. Let us pose the
blunt question ‘Is it possib le to use num ber words in the alternative
w ay indicated?’ T h e idea o f the ‘given totality o f p ossib ilities’
enjoins us to give to this, as to any other m odal question, a clear-cut
‘yes’ or ‘no’ answer. E ither there is or there is not a possible
situation in which the facts and the interests com bine to deliver the
workability o f the im agined concepts. (In follow ing through this
train o f thought, w e seem to be con ceiving o f facts and interests
explaining concepts very m uch on the ‘Eskim o and sn o w ’ m odel
discussed towards the start o f chapter 8. And if w c were then right in
rejecting that as an account o f W ittgenstein’s view then its reappear­
ance here is another sign that w e have been on the wrong track.)
N either a bald ‘yes’ nor a bald ‘n o ’ gives a coherent interpreta­
tion o f W ittgenstein. It seem s right that he would like us to consider
such im agined situ ation s, other ‘forms o f life’ and so forth. But
w hat sort o f upshot is this consideration to lead to? T o su pp ose that
he im agines it leading to either a clear ‘y es’ or a clear ‘n o ’ answ er is
to com m it him to landing in a m uddle. T h u s if w e say ‘n o ’ w e are
back w ith Platonism ; the claim that there is only one possible
developm ent o f the num ber system stands vindicated; in develop ­
ing the system in the w ay w e did, w e discerned w hat w e were
already tacitly com m itted to. B ut if on the other hand w e say ‘y es’
then w e are back with the sort o f con ven tion alist/P laton ist hybrid
w c had above.
 Interests, Activities and Meanings 209
B ut w hat if w e were to say rather that w e just clo not know
w hether such an alternative is possible or not? T h is claim o f
ignorance is indeed on reflection entirely plausible. W hat w ould be
required to show the possibility or im possibility w ould be som e
rigorous p roof that attem pting to use such concepts w as either
hound to result in som e sort o f absurdity or contradiction or
alternatively could never so result. But how could w e have such a
proof? An attem pt to exam ine w hat it would be like to use these
other concepts takes the form o f trying to specify in non-question-
begging ways a set o f circumstances and interests which make them
appropriate. But this involves us in reflections which ramify to
em brace more and more o f our interests, assum ptions and practices,
in ways that arc less and less easy to handle formally. T h e hope o f
finding som e unconlroversially acceptable demonst ration o f the coher­
ence or incoherence o f these alternative concepts dwindles aw ay.18
But let us not ju st say that w c are invincibly ignorant on the
m atter, for that is not quite right either. It is not that we are
ignorant o f som ething about w hich there is an answer; it is not that
w e finite beings cannot handle the com plexities required to get an
answer; if that were all that w as am iss perhaps a com puter could
help us. T h e problem is rather that we d o not know in enough
detail w hat to d o w ith the questions, ‘W hat do you m ean “ Could
people go on like this?” ?’, ‘W hat do you m ean “like this” ?’ T h e
bew ilderm ent w hich poses these, counter-questions is not, how ever,
a straightforward and clcar-cut discovery that the question is
nonsense and so could never be given any intelligible answer. T h at
w ould precipitate us back into Platonism , since relegating the
remark ‘people carry on in this other w ay’ to the realm o f the
clearly nonsensical is a close relative o f relegating it to the realm o f
the im possible and w ould have th e sam e effect.
O u r bew ilderm ent is also not the ou tcom e o f know ing o f nothing
one could do to answ er any question posed in this form o f w ords. In
other contexts, e.g. w here w e knew w hat features o f our current
world view and practices w e are required to hold constant or w hat
sort o f evidence is required for talk o f ‘p ossib ility’, w e could arrive
at an answer. N eith er is the problem that, in this particular
philosop hical context w e have n o inkling at all o f how to go about
answ ering the question. T h e position is rather that we have enough
210 Interests Activities and Meanings
,

grip on the question to set us off dow n a path w hich w e soon realize
w e can not get to the end of.
Is asking the question pointless then? N o. Posing it is one w ay o f
articulating our aspiration towards self-understanding, towards
insight into our w ays o f thinking and w hether they are all right or
not. W ittgenstein d ocs not w an t to deny us this aspiration or to
prevent us pursuing the enquiries it leads us into. B ut m y sugges­
tion is that he supp oses that working through the lines o f reflection
suggested will show us that w h at w e end up w ith, and w ill find
satisfying, will be aw areness o f the interconnectedness o f our con­
cepts, practices and interests. W c shall becom e reflectively aware o f
the internal com plexity and m utual d ependence o f the elem ents in
our form o f life — not aw are o f its advantageous placem ent in som e
(independent and given) array o f ‘possible forms o f life’. Phil­
osophy therefore ‘assem bles rem inders’ w hich ‘enable us to com ­
m and a clear view ’; it describes, but in a way w hich satisfies
w hatever deserved satisfaction in the aspiration to explanation. 19
T h e considerations set out above arc the ones w c need to bear in
mind in considering w hether rejection o f m irroring realism com ­
m its us to a form o f relativism w hich, in turn, underm ines even
m inim al realism . T h e threat o f a breakdow n in m inim al realism ,
w hich wc saw (in section 2.4) to be im plicit in relativism , recedes
on the W iitgcnstcinian view (as I read it) because the putative
‘other w ays o f looking at th in gs’ never acquire in our thought m ore
than the shadow y and notional status w hich they have at the
beginning o f the process o f reflection. W c never answ er the ques­
tion ‘Arc there other conceptual schem es?’ affirm atively in a way
w h ich w ould precipitate us into relativism . O n the other hand, the
intelligibility o f the initial question w h ich sets off the reflection,
together with the non-availability o f a clear negative answ er, is
enough lo scupper m irroring realism . W e are thus enabled to walk
the tightrope between the two unsatisfactory options.

9.4 M O D A L R E A L IS M

So now , alter all this, w hat account arc we offer o f m athem atical
 Interests, Activities and Meanings 211
proof? H o w are w e to understand the id ea that proof alters con­
cepts? I w ould like to m ake som e (sketchy and unsatisfactory)
gestures in the direction o f w h at I take to be a W ittgenstcinian
account.
T h e connections will go som ething like this: W c m ust hang on
firmly to the idea the p ossession o f a concept is not being the
container for som e psychic m echanism linking one to the structure
o f the w orld. Rather a con cep t is an clem en t in a way o f living; it is
that w hich w c say som eon e has w hen w e see that he or she is able to
participate in that w ay o f living. So a change in a concept is not a
change in som e disposition to respond to stales o f allairs, a d isposi­
tion w hich is grounded in a state o f o n e ’s m ind. Rather it is a
change in on e’s w ay o f life. But, one m ight protest, any kind o f
change in things a person does - a change o f jo b or o f habits in
eating - is a change in w ay o f living. W hy should going through a
proof and accep ting its conclu sion (w ith whatever alterations in
behaviour are consequent on that) be singled out as the sort o f
change to be labelled ‘co n cep tu a l’? T h e answer will be that this sort
o f change is interestingly fundam ental from a particular point o f
view . It is a change w hich puts a new rule into circulation. T h is
rule is not som e abstract object (e.g. an aspect o f the supposed
psychic m achinery) but an actual formula which is used in a
certain w ay. People w h o have accepted the proof and the formula
w hich is its conclusion, have altered the range o f descriptions they
are prepared to give o f w h at they do, the range o f discrim inations
they see them selves as m aking and the techniques they have for
m aking them . T h u s the range o f things that m akes sense to them
has altered. But no such alteration is to be seen as an unfolding o f
w hat w as im plicit before — w hen this is envisaged as som e dis­
covery about the totality o f possib ilities and the m achinery linking
us to it. W e m ay alw ays d islod ge that picture by attem pting, as we
did in the seven, four and eleven case, to im agine alternative
developm ents o f the earlier practice and reflecting that w e cannot
say that such develop m ents are im possible.
C an w e com bine acknow ledgem en t o f this shadow y presence o f
other ‘possible d evelop m en ts’ w ith any kind of m odal realism? Can
w e thus have our cake and cat it? And what has happened to the
212 Interests, Activities and Meanings
distinction between valid and invalid proofs? T h ese questions now
becom e pressing. W e need, if w e can, to show that m odal realism o f
som e sort is an elem en t o f the W ittgensteinian position.
Let us remind ourselves o f how these questions are linked w ith
som e w e have already discu ssed. In section 6.3 I claim ed that
W ittgenstein’s position w as significantly different from that o f
thoroughgoing pragm atism . W ittgen stein does not think that w e
can m ake sense o f the idea o f every fundam ental principle being
chosen in the light o f supp ortin g reasons. T h ere are certain things
w hich w e just take for granted, in the sense o f relying on them
unhesitatingly w hile not know ing how to argue for them . W e m ay
becom e aware o f their pivotal status. B ut w e do not ju stify our
adherence to them (in the light o f this or any other considerations)
in the sense o f com paring them w ith som e alternative and seeing
their superiority.
At that earlier stage in the discu ssion w e had not considered
W ittgenstein’s rejection o f m irroring realism and all that flows from
this. So the conclusions there were supported on different grounds
from the ones (about the im pon derab ility o f questions ab ou t dif­
ferent forms o f life) w hich w e have ju st considered. But there is no
conflict in the upshot o f the tw o discussions. Both lines o f thought
draw our attention to the fact that certain principles are ones that
we rely on unhesitatingly and that the idea o f not so d oin g can be
allow ed no more than the m ost shadow y verbal status by us.
W e-also carry forward from earlier discussion in chapter 6, and
from the end of section 8.1, the id ea that m inim al realism is bound
up with these principles. M in im al realism has to do w ith incom ­
patibles. And the fundam ental principles w e are talking o f include
articulations of what w e find incom patible, o f those rules in think­
ing and speech w hich w e m ust n o t violate on pain o f stultifying
ourselves.
An im plication o f these earlier m oves is that if, at the end o f the
day, m odal realism is found to be an illusion w hich needs outright
rejection, then the w hole W ittgensteinian realist edifice begins to
totter again. Let us im agine that w e accept the remarks about the
centrality to us o f certain principles and w e accept also their role in
sustaining minim al realist practices. Suppose also that w e take the
view that it is definitely w rong to construe these principles as
 Interests Activities and Meanings
, 213
them selves records o f facts to be realistically construed - then w e
shall need an alternative characterization o f their status. W hat
could this be other than the annou ncem ent that they are projec­
tions o f the w ay our m inds work or o f w ays w e have (som ehow , at
som e level) decided to m ake them work? A nd if w e say this, then
the practices that arise from acceptance o f the principles and the
things in the world w hich those practices have us talking of, all
inherit this non-realist status.
So returning to the discu ssion o f W ittgenstein’s view s on proof,
validity and the like, the problem is to reconcile, if we can, the
em p hasis on p roof as introducing new concepts with a realist
treatm ent o f m odality. W e have to rem em ber here that allow ing the
notion o f valid proof to be defensible in a (m inim al) realist fashion
will bring some notions o f ‘unfolding’ and ‘sam e con cep t’ riding
piggyback w ith it.
T h e only w ay o f effecting the reconciliation is to see talk o f new
concepts as aim ed at underm ining a particular m etaphysical pic­
ture o f w h at valid proof is, rather than at underm ining our ordinary
(and rightly realist) use o f that notion. C an we m ake any sense o f
this? A n extrem ely im portant fact, to w hich w e have not hitherto
given its proper place, is that our w ay o f life is one in w hich
enunciation o f certain truths as necessary, together with formal
techniques o f m anip ulating and extend ing them , play an im portant
part. W e are trained to look at things in a certain w ay, by pictures
being exhibited, by people sayin g ‘But seven and four must m ake
eleven ’ and by being put through routines. But we are also trained
to construct further pictures and patterns for ourselves, to put
ourselves through routines and to say to ourselves ‘Ah! Such and
such must be the case’. For exam ple, som eone w ho has learnt how to
add has learnt how to put him or herself through a procedure which
w ill result in firm acceptance o f som e form ula like ‘25 + 47 = 72’
and in confident perform ance o f the kind o f actions that flow from
that, even w hen he or sh e has never been confronted with that sum
before.
W e take it for granted that m ost hum an beings will cotton on to
these procedures and w ill agree in the verdicts and actions they
produce. Even if we have com e, by philosophical reflection, to the
aw areness o f endless shadow y possibilities for alternative responses
214 Interests, Activities and Meanings
branching off at every stage o f every proof, this docs nothing
to w eaken the conviction w ith w h ich w e put our children through
the training or the confidence with w hich w e ourselves perform
calculations, seek proofs and use the results. T h ose alternative
responses, as w c have seen , arc not live options for us; w hereas the
enunciation o f these fam iliar necessities and their extension by the
fam iliar techniques arc not only actually livable by us but entirely
com pelling.
It is in this context, o f everyday training in and use o f formal
procedures that the concept o f valid p roof has its hom e. It is
because w c live the kind o f life in w hich w c calculate and train each
other in p roof procedures - and because it all works out satisfac­
torily as it docs — that w c have the concepts o f validity and
invalidity. So from one perfectly proper point o f view , it is absurd to
say that som eon e perform ing the usual sort o f calculation shifts to a
new way o f life or adopts new concepts; d oing this sort o f thing —
and ad ap tin g our practices in the w ay that goes with it — is a
norm al part o f our (one) w ay o f life; m oreover w c m ay perfectly
properly say, from this point o f view , that a person is, for exam ple,
gettin g a better understanding o f num bers. B ut the W ittgensteinian
remarks w hich seem to conflict with these com e from another point
o f view , or arc appropriate at another stage in discu ssion, when w e
arc trying to com bat a m istaken philosophical account o f w h at such
activity involves.
T h e tem ptation w c m ust resist is to try for som e kind o f further
account o f w hat m akes use o f concepts like validity the proper thing
in particular cases. T h ere arc, broadly speaking, two kinds o f
account available —som ething in the ‘realist' and Platonist style or
som eth in g in the prqjcctivist, speech act or pragm atist style. O n the
first w e try to explain talk o f validity in term s o f discerning m ore o f
the structure o f the totality o f possib ilities by unpacking w hat w as
hitherto in exp licit in our grasp o f it. O n the second w e say that
w hen som eon e says ‘seven and four m ust equal eleven’ or ‘I f this is
true then that m ust be true as w elt’, w hat he or she is really doing is
expressing som e state o f psychological conviction or perform ing a
speech act o f enjoining everyone to behave in a certain w ay.
W e asked above w hether it w as possib le to have our cake and eat
it — to be, in the m inim al sense, m odal realists w hile rejecting
 Interests, Activities and Meanings 215
Platonism . T h e answ er is that it is — but that there is a price to be
paid. And the price is that w e abandon as m isguided (and, at
bottom , unintelligible) the am bition to give a m etaphysical account
o f m odality in any such w ay as those m entioned above. I f som eone
w ants to know w hat m akes it right to say that seven and four make
eleven or that such and su ch a p roof is valid the only answer to be
given is that it is that seven and four do make eleven or that the
p roof is indeed valid. A nd if our questioner doubts these things w c
m ight attem pt to convince him in the usual ways - by show ing the
pictures, running through the proofs and so forth. T h is is w hat
q uictist realism , as opposed to the m irroring or pragm atist variety,
enjoins. A ll this should not com e as a surprise. T here should be,
from a strategic perspective, a place in a W ittgensteinian outlook
for m odal realism w ithout Platonism , if there is a place for realism
in general w ithout the m irroring view . Indeed the former will be
ju st a particular case o f the latter.
T h e natural objection iierc w ill be that we are going round in
circles. T h e project w c set ourselves w as that o f defending m odal
realism , in order to prevent its failure ram ifying through the system
and underm ining the defen ce o f m inim al realism that w c had
attem pted to put in place for other areas o f our thought. But all we
have been given is the sam e story over again! And it was ju st the
acceptability o f that story w hich w as at issue.
W hat this protest am ounts to, how ever, from the non-m irroring
point o f view , is yet another m anifestation o f the sam e im pulse
w hich all along w c have been trying to com bat. It is yet another
attem pt to get (as its proponents suppose) our concepts firmly and
properly grounded by link ing them to som ething quite outside our
practices and responses. B u t right across the board w c need to be
coaxed ou t o f the im pu lse to m etaphysical speculation, to be
persuaded that w c can and should be content w ith that under­
stan din g o f our concepts w h ich com es from seeing how judgem ents
using them are placed in a context o f actions, interests and other
ju d gem ents, so that together they constitute the only sort o f life that
w e have any idea how to live. ‘N ecessity ’, ‘possibility’, ‘validity’
and the like are ju st as m uch deserving o f this treatm ent as any
other o f our concepts. It is not the ease that they alone are
inseparable from the m irroring/P latonist picture and so m ust fall
216 Interests, Activities and Meanings
into disrepute with it. T h e challenge to any objector w ho w ishes for
som ething other than w h at has been offered, is to give som e
coherent account o f w hat it is that he w ants. T h e claim is that any
such attem pt will run into the problem s outlined in section 7.1 -
those o f dem anding the self-extrapolating series or the self­
interpreting sign.

9.5 FA C T S A B O U T M E A N IN G

In this final section w e turn at last to the question o f w hether there

are facts about m eaning, how w e are to understand the claim that
there are and how such claim s should be m ade defensible. T h e key
ideas w e need have already been introduced. It rem ains to as­
sem ble them in order and to in dicate how they enable us to avoid
the disastrous paradoxes explored in chapter 6.
In order that w e should be realists about m eaning, in the
m inim al sense, w e need to show tw o things: O n e is that there is a
practice o f m aking ju d gem ents about m eaning w hich fits the m ini­
m al realist pattern. T h e second is that the practice is proper or
legitim ate. T he second task is likely to he m ore difficult, but let us
start w ith som ething about the first.
T h e requirem ent is that w e should m ake statem ents about our
ow n and other p eop le’s m eanings w hich w e recognize to be subject
to the law o f non-contradiction. W e take it, in other words, that
there are incom patible m ean ing ascriptions and that incom patible
ones cannot both be fully defensible. T h e w ell-nigh intolerable
paradoxicality and bizarreness o f Q u in ean scepticism is strong
evidence that we do operate su ch a practice.
But w hat are the incom patibles in the case o f m eaning? It is
worth remarking that incom patibility o f m ean ing ascription, w here
‘m eans that X ’ and ‘m eans that Y ’ are an incom patible pair,
includes but also extends far beyond cases w here ‘X ’ and *Y’
them selves are incom patible m oves in the language. In K ripke’s
case o f the ‘plus’ and ‘q u u s’ functions, the two functions yield the
sam e answer in application to certain num bers and hence are not
incom patibles.20 N everth eless, on a realist view , the related m ean­
ing ascriptions do exhibit incom patibility. O n e im portant feature
 Interests Activities and Meanings
, 217
underlying the intuition o f incom patibility here is that the 'plus’
and ‘q u u s’ functions do or clearly could diverge, and that this is
linked with our grasp o f them as different concepts.
W e should note at this point that the realist about m eaning is not
com m itted to the idea that people arc never m uddled, never have
vague ideas, or never have ideas w hich fail to tally exactly with
ours. It m ay be, for exam p le, that w e possess two concepts which
w e distinguish because wc know they can diverge, w hile another
person operates som e notion w hich does not equip him or her to
deal with the cases w here, as w e see it, the notions com e apart. T h is
is not a case where ascription o f both o f our concepts to the other
person is justified, but rather a ease w here either m ight do (or som e
purposes but, in other contexts, neither w ould be entirely happy.
A nother thing the m eaning realist is not com m itted to is the idea
that there is som e one thing which is ‘the’ m eaning o f any given
remark. For different purposes, in different contexts and so forth,
different m ean ing ascriptions m ay be in order. T h e point is only
this: given a particular question in a particular context, there will
be a range o f answers; som e o f these answ ers will be incom patible;
and we properly take it that not m ore than one o f the incom patibles
can be fully defensible.
Som eone m ight object that w e conceded in chapter 2 that it was
possible for a person to m ake contradictory ju d gem en ts to think
that both p and not p. Indeed, in section 2.1, that claim played an
im portant role in linking together the various si rands characteristic
o f m inim al realism . But if this is so, the objection continues, how
can one claim that m eaning that p and m ean ing that not p (for
exam ple) are incom patible?
But let us m ake som e careful distinctions here. W e m ay have a
person w ho says that p and also says (in another, distinguishable,
remark) that not p and w e m ay take it that both remarks are
sincere. W e m ay even have a person w ho produces one utterance
w hich m eans that p and not p. But these are not cases w hich would
threaten the claim o f the existence o f incom patibilities in m eaning
ascription. For that w e need a case in w hich a person can rightly be
claim ed to say, in a given remark to a given audience, that p and
also rightly be claim ed to say, in that very sam e remark and to that
very sam e audience, that nor p, w here each o f these, that p and that
218 Interests, Activities and Meanings
not p, is represented as the full con ten t o f w hat is said and is
thought o f as conveyed by the very sam e features o f w hat is uttered.
T h is is the sort o f ease w hich the m ean in g realist will reject as
incoherent. T h e im portant difference from the other cases is that in
them w c had two remarks to interpret or som e relevant com plexity
w ithin one remark, so that the m ean in g that p and the m eaning
that not p got attached to different vehicles. In these circum stances
w hat leads to the interpretation and w hat w c expect in virtue o f it
can be intelligibly set out. But w hat w c arc now asked to im agine is
that o n e and the sam e vehicle he rightly given two opposed inter­
pretations. T h e problem for the realist is not that w e m ake the
person w e are interpreting out to he so confused that one can
hardly supp ose that anyone could be that m uddled. T h e problem
rather is that ive, the interpreters, can n ot give any sensible account
o f w h at w c think we arc up to. But the non-realist about m eaning, a
believer in the indeterm inacy o f interpretation for exam ple, su p ­
poses on the contrary that when w c reflect on how w e proceed in
m ean ing ascription we will see that this is exactly the kind o f thing
the possibility o f which (even if it d ocs not happen in practice) is
built into the notion o f m eaning. For the non-realist the realist
asp ects o f our practice o f m eaning ascription arc mere superficial
appearances.
L et us turn now to the second o f the two conditions which I
m entioned at the start o f the section. It is here that all the serious
problem s arise. W e need to show that our realist practice does not
need re-interpretation or rejection but is properly taken at face
value. Som e flaw m ust be found in those sceptical argum ents w hich
purport to show that our practice o f m ean ing ascription is, by its
ow n lights, not realistic. T h e scep tic undertakes to show us, by
exam in in g the working o f our notion o f m eaning, that the realistic
seem in g practices with m eaning arc at best tem porarily advan­
tageous ploys and at worst plain m istakes.
T h e sceptical m ove that w e have exam ined in chapter 5 pin­
p oints the crucial source for the dilHculties w ith realism about
m ean ing. T h ese difficulties arise from the acknow ledgem ent o f the
fact that possession o f m eaning by any item presupposes the
existence o f an im m ense and intricate pattern o f other m eaning-
bearing item s, stretching out over tim e, including the future, and
 Interests, Activities and Meanings 219
also o f the fact that the particular m eaning which an item has is
bound up w ith its role or place in the pattern. W hatever the exact
account w e give o f the kinds o f item in the pattern, or o f the nature
o f the organizing relations, the central problem in defending re­
alism about m ean in g rem ains. It is to sec how , if possesvsion o f a
particular m eaning is bound up with place in the pattern, anyone
could be justified in attributing a given m eaning to a particular
item unless he or she had som e independent access to the rest o f the
pattern in w hich the item occurs. But how could such independent
access be m ade sense oi? Large parts o f the pattern are in practice
or principle unobservable. A nd, worse, even if we suppose that
problem overcom e and ourselves supplied with all im aginable data
about other utterances and behaviour, it seem s that the evidence
m ay still be in adequate and w c m ay be faced with the indeterm i­
nacy problem explored in chapter 5. W c do indeed make confident
ju d gem en ts about m eaning, as though the limited part o f the array
o f utterances w c are confronted w ith allow ed us to be certain o f the
existence and nature o f a unique best pattern into w hich those
utterances, and others m ade by our subject, will lit. It is as though,
having picked up ju st a handful pieces from an im m ense and
difficult jig sa w w c im m ediately and confidently assigned roles to
them . But how can w c m aintain such confidence in the light o f our
inability to sh ow that an alternative interpretation o f our subject’s
utterances docs not exist? 'l'his is the crucial question to w hich w e
m ust now address ourselves.
W ittgenstein’s interlocutor at one point asks ‘But if you arc
certain, isn ’t it that you arc sh u ttin g your eyes in the face o f doubt?’
A nd W ittgenstein replies ‘T h ey arc sh u t’.21 M y contention, to put it
provocatively, is that, with W ittgenstein, wc should keep our eyes
shut. T o vin d icate our realism about m eaning wc do not have to
attack directly the conten tions o f chapter 5; we do not have to deny
that m eanings ascriptions arc answ erable to placem ent in a coher­
ent pattern or to sh ow the existence o f som e strategy for discerning
patterns w hich yields a guarantee that utterances fit together in
only one w ay. T h e m istake rather is in the sceptical conclusion that
is drawn from these observations about the holistic nature o f the
concept o f m ean ing. T o avoid the sceptical upshot all w e have to do
is keep our ey es sh u t, to ignore the possib ility o f alternative
220 Interests, Activities and Meanings
interpretation which the scep tic thrusts at us. And w e are not to be
criticized for so doing.
T h e sceptic, or any philosopher w h o w ishes to see ju stice done to
him , will find this q u ite outrageous. It will seem to him to be a
blatant exam ple o f the advantages o f theft over honest toil. H ow
can all that intricate and laborious argument, about holistic m eth­
odology and its im plications be sw ep t aw ay as irrelevant? H ow can
keeping one’s eyes shut be other than irresponsible?
'['here arc indeed circum stances in w hich a person m ight irres­
ponsibly shut his eyes and refuse to adm it a doubt w hich should
have been adm itted and should perhaps have influenced behav­
iour. Im agine that 1 am confronted w ith a row o f boxes in one o f
w hich, I am told, there is a diam ond. I am given som e clues as to
w hich box the diam ond is in - it is in an intricately carved box,
m ade o f oak and having com ers bound w ith brass. I see a box
w hich answers all elem ents o f the description and leap to the
conclusion that the diam ond is in it. It is then pointed ou t to m e
that another box also fits the requirem ents. B ut I rem ain entirely
convinced that the original box is the one containing the diam ond
and refuse to take out any insurance policies to cover error. I am ,
however, wrong. T h e diam ond is in the other box. C learly I have
been foolish and am in for a nasty shock. I should have hedged m y
bets. And even if we tell the story differently and im agine m e to
have been right, it is still clear that I w as foolish and my confidence
m isplaced; I have been saved from disaster only by luck.
T h e structure o f this situation closely parallels that in w hich
traditional sceptical problem s arise in the context o f m irroring
realism . I have em phasized at several earlier points how m irroring
realism goes naturally together w ith the id ea o f a given totality o f
m etaphysical possibilities or w ays that things m ight be. Let us
equate these with the boxes ‘and let us equate being actual with
containing the diam ond. T h en the scep tic im agines that w e are
supplied with som e inform ation about the actual world (by being in
it and having it im pinge on us) and w e also have som e inform ation
about the range o f possibilities (from som e o f the real properties o f
the universe having revealed their nature and fitness for com bina­
tion to us). O n this basis w e try to find out m ore about the actual
 Interests Activities and Meanings
, 221
world, i.e. to identify w hich o f the possibles is the on e w c actually
inhabit.
M aterial object scep ticism , for exam p le, em erges like this: T h e
sceptic adm its that w e know o f the actual world that it contains
sensory experience o f a certain character. But, he says, am ong the
range o f possible w ays things m ight be arc not on ly ones in w hich
such experiences are explained by m aterial objects but also ones in
w hich they are the ou tcom e o f m alicious dem ons, chance, scientists
m anipulating brains in vats and the like. H ow are w c to justify our
confidence that our situation answ ers to the first description? O ne
m ight suppose that goin g out to have a further look at our actual
w orld, interacting w ith it som e m ore, would supply relevant evi­
dence. B ut, o f course, the problem lias been set up precisely to
m ake this inefficacious, since all it could supply us w ith (the sceptic
urges) is m ore o f the sam e stuff, i.e. sensory experience, the expla­
nation o f w hich is w hat is in question.
M atters are com p arab le in the case o f m eaning. It is supposed
that we are aw are o f (supplied with inform ation about) uninter­
preted utterances. O u r problem is to show the legitim acy o f our
confidence in the linkage o f those utterances w ith particular fea­
tures o f the w orld, in the face o f the possibility that they could be
arranged in such a w ay as to display different linkages.
T here are, how ever, som e interesting differences betw een the two
cases. In the case o f m aterial objects the desired upshot o f the
sceptical reflection is not (at least in the case I have im agined) the
underm ining o f the concept o f material object. T h e notion remains
usable and we continu e to adm it that in som e possible situations
there is experience explained by such objects. W h at is supposed to
have evaporated is our confidence that there are any such objects in
our world. In the case o f m eaning how ever the upshot is more
drastic. W e are accep ting, at this stage in the argum ent, the
im portant negative thesis w hich is in com m on betw een Q uin e and
W ittgenstein, nam ely that to hold that an utterance m eans so and
so is not to take it that som e item lies behind or causally explains
that utterance — rather it is to take it that the ¡tern has a certain
place in a pattern o f item s. T h e effect then o f coun ten an cin g the
possibility o f another arrangem ent, o f discerning another pattern in
222 Interests, Activities and Meanings
w h ich the utterance m eans som eth in g quite different, is to under­
m ine the viability o f m eaning as a realistically intended descriptive
tool. W c d o not ju st say ‘W c cannot be sure that people m ean w hat
w c earlier took them to m ean — although they m ight m ean that,
since in som e possible situation they d o m ean it.’ W c say rather
that in 110 possible situation d ocs anyone really m ean anything -
that talk cl'm can in g is not any part o f sayin g how things really arc.
T h is now casts an odd light backw ards on the m oves by which
the conclusion w as reached. T h e claim w as that there is a possible
situation in w hich the utterances I am confronted with arc differ­
en tly arranged from the way in w hich I arrange them. T o talk in
this way seem s to adm it that I do indeed arrange them one w ay,
even if som eone else docs it another way. But m y arranging
utterances (i.e. my interpreting them ) consists, given the non­
existence. o f m eanings outside patterns (ff utterance, m erely in my
producing certain utterances o f m y ow n which in turn m ay be
formed into a pattern. It is as though all one could to towards
assem blin g a jigsaw w as to throw dow n another handful o f jigsaw
pieces — w hile allow ing that this second lot them selves could be
built into m ore than one picture o f the assem bling o f the first lot.
B ut then no single arrangem ent has been im posed 011 the first lot by
the appearance o f the second lot and the w hole notion o f an
arrangem ent begins to crum ble 011 us. 'I'his head-spinning outcom e
is not, how ever, a diiliculty that w c can w heel up against the
non-realist about m eaning. H is p oin t is that the notion o f m eaning
underm ines itself; his will therefore be happy to deploy it adhominem
at certain stages o f the argum ent.
L et us pause to assess m atters here. O n e m ight remark that the
sceptical argum ents presented above proceeded on the assum ption
that w e know that it is (m etaphysically) possible that sensory
experience like ours should be explained by som ething other than
m aterial objects, or that utterances like the ones w e hear should be
differently interpreted. But this is very arguably too strong, too
m uch o f a concession to the sceptic. T h e m eaning-sceptical argu­
m en ts considered by us have not claim ed to show that other
arrangem ents can for certain be produced or to suggest an algor­
ithm for producing them , only that w e do not know that they
can n ot.2“ Sim ilarly one m ight argue that it is not clear that sensory
 Interests, Activities and Meanings 223
experiences like ours could occur in the absence o f physical objects.
H ow confident w e arc in the existence o f either o f these possibilities
will depend on how confident w e are that wc now correctly grasp
the nature and relations o f the real characteristics which define the
world. Perhaps the supposed possibilities of other arrangem ents o f
utterances or other explanations o f sensory experience are m erely
cpistcm ic and not real?
U nfortunately this thought, in the context o f the current picture,
cannot help us. C onsider a slightly different situation with boxes
and a diam ond. 'Phis tim e I am told there is am ong these boxes
(and m y inform ant gestures vaguely at a group) just one box w hich
contains a diam on d and I ain given chics as before on what the box
is like. O n this occasion how ever I am not. so foolish as in the
previous case. I look carefully at all the boxes 1 can sec and I make
sure that I select the only one w h ich fits the description. But then
the thought strikes m e that there m ight be other boxes which I have
not seen but w h ich w ere included in the group. Are there som e
under the floorboards or behind the curtains? I realize that 1 do not
know w hether or not the ones i can see are the only ones I need
bother about and m y original inform ant is no longer at hand to
help. Should m y confidence that the selected box contains the
diam ond be underm ined? It seem s that it ought. (It m ight be
sensible, for exam p le, to take ou t an insurance policy on its not
containing the d iam on d , if a lot hangs on the question and the
policy is ch eap .) A n d this is so irrespective o f whether I am right or
w rong in m y supposition that there arc m ore boxes.
T h e an alogue in the possible situations case is clear enough. I
realize that I do not know w hether or not there are possible
situations in w h ich sensory experiences like ours is to be explained
by som ething other than m aterial objects or in w hich utterances are
intelligibly interpreted differently. So m y confidence in the exist­
ence o f m aterial objects or the coherence o f the notion o f m eaning
ought, it seem s, to be correspondingly underm ined. W hether I am
right or n ot in m y su p p osition about possibilities is irrelevant.
A s a m atter o f psychology, it seem s likely that in som e actual
versions o f the box and diam ond story the guesser’s confidence
m ight w ell be shaken by the reflection that he had no clear
inform ation on w hat boxes w ere included in the relevant group. By
224 Interests, Activities and Meanings
contrast it is notorious that sceptical argum ents d o not in fact shake
our beliefs about m aterial objects or m eanings (to say nothing o f
the past, the sensations o f others, the su n ’s rising tomorrow etc.).
W e m ay adjure ourselves ‘Face up to the (possible) facts! O pen
your mind to the thought that there is no such thing as determ inate
m eaning!’ But we find that, like G loucester trying to ju m p off the
cliir, we. com e dow n exactly w here w e w ere before. T h e sceptic,
however, may not be im pressed by this. H e m ay suppose it to be
merely force o f habit or im m ensely strong natural propensity.
T h e key question then is w hat we are to m ake o f propositions like
‘Sensory experience such as ours is explained by som ething other
than material objects’ and ‘U tterances like these are arranged in a
different pattern’. O n the m irroring realist view either they describe
possibilities or they do not. W e m ay adm it ignorance o f the nature
o f the full range o f possibilities, but in so far as w e have a grasp on
it, it m ay seem to us (epistem ieally) m ore likely than not that the
sentences do describe real possibilities. A nd even if w e are. not
prepared to attach a probability to the ju d gem en t that they d e­
scribe possibilities, the mere fact that, as we conceive things, w e do
not know that the situations arc im possible ought to m ake us adjust
our confidence in the existence o f m aterial objects or the dcfensi-
bilily o f m inim al realism about m eaning. T h u s it is irresponsible
and im proper to ignore that (epistcm ic) possibility that these
things are (m etaphysical) possibilities. But on w hat 1 suggest is the
W ittgenstcinian view w e arc entitled to ignore them , to shut our
eyes to them. H ow can this com e about?
It is clear that if w e reject m irroring realism then w e m ay also
reject the idea o f the given totality o f possib ilities defined by
com binations o f real properties. But this thought alone is not
enough to give us w hat w e need. W hy should w e not suppose that
any conceptual schem e brings with it its ow n set o f possibilities in
connection with w hich honest accounting m ust be done w hen
assessing the degree o f confidence to be reposed in various ju d g e­
ments? T h e assum ption behind the question is, in a sense, right.
For exam ple, in our conceptual schem e w e m ay describe situations
like the box and diam ond one, and such a situation brings with it
relevant ranges o f possibilities o f w hich participants should take
account. And som etim es participants fail, through inadequate
 Interests, Activities and Meanings 225
reflection, to notice som e o f the possibilities w hich arc relevant.
T h e possibilities are thus given and are independent o f the epis-
tem ological state o f the participants. W hy is it not the sam e for the
philosophically controversial ‘possib ilities’ which interest its?
T h e answ er is that in abandoning m irroring realism we cease to
be in a position w here we can insist on sayin g o f every (in som e
sense) intelligible sen ten ce ‘Kither it describes a possibility or it
does n o t’. W hat has been em phasized again and again is that our
concepts do not com e isolated. T o understand a concept is to see it
at work in its setting. As I tried to suggest in the last section,
possibility is a notion (or belter a family o f notions) o f w hich this is
true, ju st as m uch as o f any other. And the fact is that w e are not
trained in, we have in our lives no role for, assessing the possibility
o f every kind o f conceptual com bination. W e talk o f possibility in a
variety o f contexts - som e having to d o with derivations in formal
system s, som e with w hether men can sw im rivers, pegs lit into holes
or diam onds are to be found in boxes.23 In these contexts w e know
what to do with the notion o f possibility - how to go about
establishing w hether or not som ething is possible and w hat accept­
ing the ju d gem en t m ight lead to. Hut in connection w ith the
sentences the sceptic or non-realist thrusts at us w c are at a loss on
all these fronts. In particular w e are com pletely stym ied on, so to
speak, how to proceed w ith thinking (how to d o it, how to conceive
o f it, how to feel about it) if the indeterm inacy, and hence non-
reality, o f m eaning w ere to be accepted. I’he thesis that m eaning is
indeterm inate presents itse lf as im m ensely im portant, as concern­
ing the fundam ental m atter o f w hat we and the world arc really
like. Surely then even entertaining it, let alone accep ting it, ought
to produce som e kind o f upheaval - as with considering that the
Earth is not the centre o f the universe or that G od does not exist.
But in fact nothing o f the kind occurs. T h e m ost strenuous attem pts
to get to grips with the idea m ay produce a sense o f gloom , allied
with im ages o f being tied up alone in a cold dark place. But these
do not persist; and if they did, or if they cam e to d om in ate one’s
thoughts, it would be evid en ce not ol'dcep philosophical insight but
o f m adness.
From the point o f view o f mirroring realism , this all looks like a
paradoxical im passe. W e have, it seem s, excellent argum ents that
226 Interests, Activities and Meanings
m ean in g is or m ight be indeterm inate. T h e diificulty in doing
anything with the thought seem s a kind o flim ita tio n or blinkcrcd-
ness in us. B ut from a stance w h ich rejects mirroring realism , the
scep tic’s im passe takes on q uite another appearance; now it pre­
sents itse lf as an indication to us o f w hat our form o f life is.
L et us com e at this from another angle, linking it in to the
reflections about stage setting w hich were introduced in section 8.1.
W hat w c arc doing in attributing m eaning looks to the scep tic like
picking up jigsaw pieces and straightaw ay assigning them roles in a
picture, on the basis m erely o f the sh ape, colour etc., w ithout w hat
he conceives to be proper attention to the other utterances (for
exam p le future ones) or to the possibility o f quite different strate­
gics o f picture construction. W hen som eone utters a remark to us in
a fam iliar language wc hear it as a sayin g that so and so. W c do not
concern ourselves cither with the possibility that the future utter­
ances will go against our expectation s and force us to reinterpret,
nor yet with the possibility that alternative construals o f the w hole
lot m ight strike som eone. Instead w e carry on as though the
m eaning o f an utterance, its place in the pattern, could be directly
discerned in the individual item . T h e scep tic will adm it this as a
phenom enological description but insists that it m asks the under­
lying cpistcm ological situation, w h ich is one o f perceiving unintcr-
preted utterances and arriving at hypotheses about the nature o f
other such utterances and their possib le arrangem ents. H e w ill
protest vigorously at the idea that w c can perceive an utterance as
havin g such and sucli a m eaning, that its place in the pattern could
be for us present at an instant.
W hat fuels the protest is the idea that in hearing m eaning in an
utterance wc arc goin g beyond w hat we arc entitled to. T h e
burden, however, o f the earlier discu ssion, in section 8.1 and again
with the exam ple o f num bers earlier in this chapter, is that there is
a way of regarding the application o f any concept on w hich it
appears risky. T o use any concept is to make a m ove in on e sort o f
life rather than another. W hen w c reflect wc sec that the livability
o f that life, and hence the usefulness and dcfcnsibility o f that
concept, presupposes a w h ole ram ifying set o f circum stances - that
results o f various sets o f procedures will coincide, that projects w ill
d evelop and give rise to other projects and so forth. And the idea o f
 Interests Activities and Meanings
, 227
discharging all these presuppositions is quite fantastical. T here are
so m any o f them , each one spaw ns m ore, and w e soon lose even the
feeblest grip on how w e could set about investigating them . T h e
hope o f finding a set o f concepts free o f any such entanglem ent with
things w hich present them selves, w hen w e becom e aware o f them,
as contingencies, concepts the use o f w hich carries only the risk of
falsehood and not the risk o f having the concepts them selves turn
out m isguided, is a will o’ the w isp. It is the illusory am bition o f
discovering the real jo in ts at w hich the world is to be sliced.
T h e im plication o f the illusoriness o f that am bition is that the
application o f any concept involves a risk of ‘going beyond w hat is
su p p lied ’. B ut this is not quite the right way o i putting it. T here is
no such thing as ‘not going beyond w h at one is entitled to ’ ir this is
conceived as m aking a ju d gem en t the propriety o f which requires
nothing beyond the m om ent. T h is is not lo say that there are no
things w hich occur all in an instant - jabs o f pain, flashes o f
lightning. It is to say that their occurring all in an instant is bound
up w ith those other happenings at other tim es which m ake it
possible for talk o f lightning or pain to be appropriate.
So use o f any concept carries the risk brought by involvement,
without external guarantee o f success, in living the life o f which its
use is a part. T o point out that som eon e runs this risk is thus, in
itself, 110 criticism o f his concepts. It is sim ply to remark on an
aspect o f w hat it is to live and to think. W hether w e take ourselves
to be confronted with a flash o f lightning, a pen resting on a table or
a person sayin g that dinner is ready the com m itm ents and the risks
are there. T h ey are im portantly different in the different cases. T h is
is w hat wc have tried to mark w ith the talk o f holism and so forth.
But the use o f a concept is not to be criticized unless the critic can
indicate a livable, real alternative in which that concept is not
em ployed. A nd this, as w e have seen, is what the sceptic in the case
o f m ean ing is unable to do.
‘F ollow ing according to the rule is F U N D A M E N T A L to our
language gam e. It characterizes w hat w e call description.’24 W c act
spontaneously, in a w ay which is n ot to be explained by pointing to
som ething outside us w hich m akes us behave that w ay, and yet
quite unhesitatingly, w ith no sense o f choice or arbitrariness. W hat
it is that w e thus do and say, and the w ay that it all hangs together
228 Interests, Activities and. Meanings
w ith the things about w hich w c d o deliberate, m ake the notions
both o f m eaning and o f reality available to us.
 Notes

C ha pter 1 I n t r o d u c t io n

1 Quine, Word and Object (I960), p. 77.

2 S. Kripkc, Wittgenstein on Rules and Private Language (1982), pp. 55—8;
R. Rorty, Philosophy and the M inor o f Nature (1980), pp. 169, 265; C.
Hookway, (¿nine (1988), p. 47; C. Wright, Wittgenstein on the Foundations
o f Mathematics (1980), pp. 85811’.
3 Wittgenstein, On Certainty, 96-99.
4 Quine, ‘ Two Dogmas of Empiricism’, in his From a Logical Point o f View
(1953).
5 E.g. Word and Object, pp. 218-21; From a Logical Point o f View, pp. 138;
The Ways o f Paradox, pp. 193-4.
6 James C. Edwards, Ethics without Philosophy: Wittgenstein and the Moral
'Life (1982).

C ha pter 2 V a r ie t ie s o k R k a i . ism

1 The ideas pursued in this section, and also in section 2.4, have been
stimulated by the thought of B. A. O. Williams (in ‘Consistency and
Realism5 (1973) and ‘The Truth in Relativism5 (1981)) and David
Wiggins (in ‘What would be a Substantial Theory ol’Trulh?5 (1980)).
2 This is not to say that they do not have other things with which they
are incompatible, for example ‘This is pure white5 or ‘This is pale
blue5. The existence o f other incompatibles is enough to enable us to
defend minimal realism with respect to colour ascriptions, in terms of
the account of realism oiferccl below.
3 To say this is not to deny that imperatives have a dimension of ‘fit
with the world5, in some ways analogous to truth and falsity. They do;
it is fulfilment and non-fulfilment. Neither is it to deny that fulfilment,
or the likelihood of it, is important in the assessment of an imperative.
And it is not to say that the issuing of an imperative can carry no
230 Notes
implications about the Tacts. The claim is simply that when I say of
the dinosaur ‘Paint it green’ I am not thereby making a claim about
how it is with the world, a claim which would be contradicted by
‘Paint it orange’.
4 Compare here Crispin Wright’s ‘Docs Philosophical Investigations I
258-60 Suggest a Cogent Argument against Private Language?’
(1986), csp. pp. 227«'.
5 See JR. Kirk, Translation Determined (1986), pp. 6 Iff for such a proposal
in connection with Quine.
6 R. Rorty, Philosophy and the Mirror o f Nature (1980).
7 This is, of course, Descartes’ view.
8 This is how 13. A. O. Williams uses the word, for example in his
‘Wittgenstein and Idealism’ (1981).
9 Similar thoughts arc found in Davidson ‘On the Very Idea of a
Conceptual Scheme’, in Inquiries into Truth and Interpretation (1984).
10 There may be a link between this thought and the whole tangle of
considerations known as ‘the private language argument’. But I shall
not explore this issue here.

C ha pter 3 I n s t r u m e n t a l is m a n d M e a n in g S c e p t i c i s m

1 See for example Quine, Ontological Relativity and Other Essays, (1966),
pp. 26, 97-8, 126-7; Theories and Things (1981), pp. 21, 72, 85.
2 H. Field, ‘Quine and the Correspondence Theory’ (1974); C. Hook­
way Quine (1988).
3 Quine, ‘Facts oT the Matter’ (1979), p. 166.
4 Quine, Theories and Things, p. 98.
5 ‘Facts of the Matter’, csp. p. 165.
6 Quine, ‘On the Reasons for the Indeterminacy of Translation’ (1970),
p. 178. See also his ‘Iipistetnology Naturalised’, in Ontological Relativity.
7 For more about the notion o f ‘observation statement’ invoked here, see
Quine, Word and Object (1960), ch. 2 and also section 3.2 below.
8 ‘Facts of the Matter’, p. 158.
9 Ontological Relativity, p. 81.
10 Let us note also Quine’s approval of Follesdal’s remark, in ‘Indetermi­
nacy of Translation and Undcrdctermination of the Theory of Nature’
(1973), that indeterminacy of translation can be derived from holism
and the verification theory of meaning. See Hahn and Schilipp, pp.
155-6.
11 ‘On the Reasons for Indeterminacy of Translation’; Ontological Relativ­
ity, pp. 8 0 - 1; Theories and Things, p. 70.
 Noies 231
12 Theories and Things, pp. 24-30; Schilpp ancl Hahn, p. 427; Ontological
Relativity, pp. 79-80.
13 This seems to be Dancy’s view in An Introduction to Contemporary
Epistemology (1985), pp. 93-4, although at other points he expresses, a
view closer to the one I suggest.
14 D. K. Lewis, ‘General Semantics’, in Collected Papers I (1983), pp.
189-90; D. Davidson, Inquiries into Truth and Interpretation (1984), pp.
129-30, 175, 220-1; G. Evans, ‘Identity and Predication’, in Collected
Papers (1985), pp. 25-7.
15 Ward and Object, p. 78.
16 Many commentators have Ibund this dillicull to fathom: e.g. N .
Chomsky, ‘Quine’s Empirical Assumptions’ (1969); R. Rorty, ‘Inde­
terminacy of Translation and of Truth’ (1972). See also the useful
article by Roger F. Gibson Jr, ‘Translation, Physics and Facts of the
Matter’, in Hahn and Schilpp, Philosophy of W. V. Quine (1986).
17 Quine, ‘Two Dogmas of Empiricism’, in From a Logical Point of View
(1953), and Philosophy o f Logic (1970), ch. 6.

C haktkr 4 Q u in k ’s N a t u r a l i z e d E m p i r i c i s m

1 Quine, ‘Two Dogmas of Empiricism’, in From a logical Point o f View

(1953), pp. 44, 42.
2 Quine, Theories and Things (1981), pp. 39-40.
3 Ibid., p. 1.
4 Quine, Word and Object (1960), pp. 2-3.
5 Ibid.
6 Ibid., p. 3.
7 Ibid., p. 10.
8 Quine, ‘Facts of the Matter’ in Essays on the Philosophy o f W. V. Quine,
cd. R. W. Shahan and C. Swoyer (1979), p. 155.
9 Word and Object, pp. 401T. Ontological Relativity and other Essays (1969),
pp. 85-9.
10 Ontological Relativity, pp. 86-7.
11 Word and Object, p. 9.
12 Ibid., p. 76.
13 Hahn and Schilpp, The Philosophy o f W. V. Quine (1986), pp. 336, 364,
428.
14 Word and Object, p. 9.
15 Hahn and Schilpp, Philosophy o f W. V. Quine, p. 336. (The emphases
arc mine.) See also Theories and Things, pp. 29-30.
16 Theories and Things, pp. 1-2.
232 Notes
17 Ibid., p. 72.
18 ‘Facts of the Matter’, p. 167.
10 S. Kripke, Wittgenstein on Hides and Private Language (1982).
20 For example ‘Notes on the Theory of Reference’, in From a Logical Point
o f Vim, pp. 13011'.
21 Ontological Relativity, p. 50.
22 Word and Object, pp. 68If.
23 Ontological Relativity, p. 40.
24 II. Field, ‘Quine and the Correspondence Theory’ (1974).
25 Ontological Relativity, p. 49.
26 Field, ‘Quine and the Correspondence Theory’, p. 208n.
27 It is worth bearing in mind here Quine’s downplaying of the import­
ance of ontology; according to him, effectively .systctnatising the
holophrastically construed observation sentences is the important
matter; ontology is a mere by product of this. li.g. Halm and Schilpp,
Philosophy o f II . f\ (¿nine, p. 115; ‘Facts of the Matter’, pp. 164-5.
28 Word and Object, p. ‘24.
29 For the line of thought in this section I am much indebted to C.
Ilook way’s discussion in bis (¿nine (1988), csp. ch. 12.3.
30 Quine Philosophy oj Ltgic (1970), p. 1.
31 Ontological Relativity, p. 8 1.
32 Quoted by R. F. Gibson in Hahn and Schilpp, Philosophy o f IF. V.
Quine, p. 153. In the context o flh is discussion it is also worth noting
that Davidson clearly reads Quine as endorsing some thesis of the
quoted kind. See his ‘On the Very' Idea of a Conceptual Scheme’, in
Inquiries into Truth and Interpretation (1984). Quine rejects Davidson’s
interpretation (‘On the Very Idea of a Third Dogma’, in Theories and
Things) in the same spirit as he reformulates the passage quoted by
Gibson, but my suggestion is that the rejection, when closely exam­
ined, preserves the spirit if not the letter of the instrumentalism
charged.
33 Hahn and Schilpp, Philosophy o f IF. V. Quine, p. 153.
34 Theories and Things, pp. 21-2.
35 Hahn and Schilpp, Philosophy o f W. V. (¿nine, p. 157.
36 Word and Object, pp. 2-1—5.
37 Word and Object, p. 76.
38 Hahn and Schilpp, Philosophy o f W . V. Quine, p. 12.
39 For an illuminating discussion o f this, to which I am indebted, see
Monkway’s Quine, esp. chs 2 and 3.
40 See also ‘Truth by Convention’ (1935), ‘Carnap and Logical Truth’
 Notes 233
(1954), 'On Carnap’s Views on Ontology’ (1951), all in The IVays o f
Paradox and Other Essays (1966).
41 ‘Five Milestones of Kmpiricism’, in Theories and Things.
42 See, lor example, R. Gregory, Eye and ¡train (1972) and J. 1’odnr,
Modularity of Mind (19(13).

G iia pt k r 5 'I’m: M ona I . isa M o sa ic

1 T shall not here pursue excgetical questions about Davidson’s views.

The papers ol'particular relevance an* 'Radical Interpretation’, 'Relief
and the basis of Meaning’, 'On (he Very Idea of a Conceptual
Scheme’ (all in his Inquiries into Truth and Interpretation (19(14)), and
‘Mental Kvenls’ (in his Essays on Actions and Events (19(10)).
2 See the discussion of ‘seeing in’ in R. Wollheim, /lr/ and Its Objects
(1970).
3 G. Ryle, Concept o f Mind (1919), pp. I30-I.
4 J. Fodor, Psychosemantics (1975) p. 50.
5 Ibid., pp. («1, 119.
6 Quine and Davidson dill’er in what they take to be offered at this level.
Quine thinks that, the descriptions are purely behavioural, Davidson
that they will involve the minimally intentional notion of ‘holding
true’. (See Davidson’s ‘Radical Interpretation’ and also ‘The Inscru­
tability of Reference’, both in Inquiries into Truth and Interpretation). Rut
the difference is not relevant to the slrurture of thr particular argu­
ment wc are considering.
7 Acceptance of this sort of holism docs not rule out allowing lor some
useful distinction between observational and theoretical sentences,
couched for example in terms of things that we tend to agree on, that
we do not arrive at by explicit inference and the preservation of which
is agreed to be an important aim in a world view. Rut this rough and
ready itlca of the ‘observational’ is far from Quine’s. Neither does this
holism entail insistence on the idea that all ‘observational’ concepts
are ‘theory laden’. That a judgement may be rejected in the light of
theoretical considerations docs not (without a link of cpistemological
to semantic which I reject) imply that the. concepts in the first arc
theoretical. For some recent work on the idea o f ‘observations’ see C.
Wright, ‘Scientific Realism, Observation and the Verification Princi­
ple’, in Fact, Science and Morality (1986), ed. MacDonald and Wright.
8 As Fodor remarks, Psychosemantics, 62if.
9 Related worries are expressed in my 'Replication and Functionalism’,
234 Notes
in Ijmguage, Mind and Logic (1986), cd. Butterfield. See also J.
McDowell, ‘Functionalism and Anomalous Monism’, in The Philosophy
o f Donald Davidson (1986), cd. Lcpore and McLaughlin .
10 This constitutes yet another difficulty lor simplistic inference from
cause to intentional content. The rest of a background view will need
to be taken into account in determining what an item is seen as. And it
is not clear that there must be a determinate way o f so doing.

C h a pter 6 T iie S i . id k I n t o t h e A by ss

1 Zellel, sections 135, 170, 172, 173, 175, 176, 534; compare also Philo­
sophical Investigations, Part II, p. 174; lllue and Ilrown ¡looks, 5; Remarks on
the Foundations o f Mathematics, 334, 335, 345-7, 390.
2 Cl. Wright expresses similar views in ‘Kripkc’s Account of the Argu­
ment against Private Language’ (1984), pp. 769-70.
3 This seems to be Korty’s position, as expounded in his Philosophy and
the Mirror o f Nature (1980) and Consequences o f Pragmatism (1982) and
sonic have seen it as the logical outcome of the position Quine urges in
‘Two Dogmas o f Fmpiricism’, in From a logical Point o f View (1953).
4 (1. Wright, ‘Inventing l<ogical Necessity’, in Language, Mind and Ij)gic
(1986), cd. Butterfield, pp. 192-3.
5 Ibid., pp. 193-4.
6 Ibid., p. 194.
7 S. Blackburn proposes some view of this sort. See his Spreading the Word
(1984) section 6.5. Wright also makes remarks suggesting sympathy
with some such outlook in 'Inventing Logical Necessity’, pp. 204-5.
But it may well be that they lx>th intend something subtler than the
crude view I address. The suspicion however will lie that the less
crudely psychological the view, the less it will have any power to
demystify ontological and cpistcmological questions about necessity.
8 Wright, ‘Inventing logical necessity’, p. 194.
9 For more on this theme see Bob Hale’s discussion, in Abstract Objects
ch. 6 of Philip Kucher's The Nature o f Mathematical Knowledge (1984).
10 Wright is attracted to such a conventionalist view and thinks that
Wittgenstein held it. (Sec his ‘Facts and Certainty’ (1985), pp.
451-3) And he also thinks that Naturalism is more defensible in
conjunction with some kind of conventionalism (i.e. in the context of
the thought ‘I’ve got to choose some rule or other; but I can’t conceive
of any one but this; so I’ll choose this’) than simply in the context of
remarking that wc do not, in fact, doubt such and such things (sec
‘Facts and Certainty' pp. 467-8). At the same place he criticizes
 Notes 235
Strawson for inappropriately invoking Naturalism as a response to
scepticism (in Scepticism and Naturalism: Some Varieties (1985), pp.
10-29) in the form o f a merely descriptive remark that we do not
doubt. It will be clear that my sympathies arc with Strawson. We may
remark also that Wright’s interpretation of Wittgenstein docs not
make sense of those passages where Wittgenstein speaks of our search
for justification coming to an end in description, on our discovering
the propositions that stand lirm for us (e.g. On Certainly 152, 189;
Philosophical Investigations 12411).
11 On Certainly, sections 56, 83, 88, 89, 94, 95, 96, 98, 105, 109, 110, 117,
119, 120, 130-1, 144, 150, 152, 189, 210, 321, 337, 343, 375, 411,
421—2. Wittgenstein in some of these passages toys with the idea dial
‘everything speaks for' some of our framework or scailblding beliefs
(On Certainly, sections 89, 117, 119). But he clearly finds this an
unsatisfactory this view of things and emphasizes instead our aware­
ness of inability to do without them (On Certainly, sections 94, 105, 110,
117, 120, 144, 150, 152). Wittgenstein is, in On Certainly, discussing
most of the time certain empirical propositions which he takes to have
this interesting status, while we have been concerned here mainly with
conceptual ones. But we shall need to return to the fundamental
‘empirical' propositions at the end of ch. 9 below.
12 A. J. Ayer, Language Truth and Isgic (1971), pp. 113-15. For another
similar positivist view sec Hans Hahn's ‘Logic, Mathematics and the
Knowledge of Nature’, in logical Positivism (1959), ed. Ayer, pp.
156-7.
13 M. Dummctt, ‘Wittgenstein’s Philosophy of Mathematics’, in Witt­
genstein, Critical Essays (1968), etl. Pilcher, pp. 424 If
14 This argument is urged in ibid., and elaborated in C. Wright, H’tVf-
genstein on the Foundation o f Mathematics (1980), csp. ch. 5.
15 Sec J. McDowell, ‘Anti-Realism and the Epistcmology of Under­
standing’ in Meaning and Understanding (1981), ed. Parrcll and
Bouvcrcsse, and also ‘Wittgenstein on Following a Rule’ (1984) for
more on this theme.
16 D. Davidson, Inquiries into Truth and Interpretation (1984), p. 230.

C ha pter 7 T h e D is s o l v i n g M ir r o r

1 Philosophical Investigations, Part II, p. 230.

2 ZetUl, 331.
3 Ibid., 352.
4 Ibid., 357-8.
236 Notes
5 Ibid., 364.
6 Ibid., 387-8.
7 S. Kripkc, Wittgenstein on Rules and Private language (1982). See also the
useful discussion by C. McGinn in Wittgenstein on Meaning (1984).
8 Williams, Ethics and the Limits o f Philosophy (1985), pp. 138-9.
9 Williams, Descartes: The Project o f Pure Enquiry (1978), pp. 64—5.
10 Ethics and the Limits o f Philosophy, pp. 138-9.
11 Ibid., p p . 139-40.
12 For more on ibis theme sec my ‘The Disinterested Search lor Truth’
(1987-8).
13 His idea would, on this reading, (it in with the programme of giving
naturalized accounts o f mind and meaning, and so defending mirror­
ing realism, which was mentioned at the end o f ch. 6 above.
14 Williams, Descartes, pp. 297-9.
15 J. McDowell, ‘Non-Oognitivism and Rule Following’ in Wittgenstein: to
hollow a Rule (1981), ed. Holtzman and I .etch; Williams, Ethics and the
Limits of Philosophy, 141-2.
16 Williams, Descartes, pp. 64—5.
17 Williams, Ethics and the Limits o f Philosophy, pp. 111—12, 138—40.
18 Williams, Descartes, p. 64.
19 Ibid., p. 243.
20 It may be that Williams includes also other phenomena under the
heading of ‘pcrspectivalness’, for example indexicality or the use o f
special vocabulary which marks one out as having some particular
social position. And it may be that he is right to say that wc must be
able to make sense of a language from which these have been elimin­
ated. But my objection will have force as long as the kind of perspecti-
valncss I have indicated is one of the things Williams means.
21 Philosophical Investigations, sections 70, 71, 97-107.
22 Ibid., 80.
23 For further remarks to this effect sec G. P. Baker and P. M. S. Hacker,
Wittgenstein: Meaning and Understanding (1983), ch. 11.
24 In S. A. Kripkc, Wittgenstein on Rules and Private Language (1982).
25 li.g. S. Blackburn, ‘The Individual Strikes Back’ (1984), p. 287.
26 Kripkc, Wittgenstein on Rules and Private language, pp. 81F.
27 Ibid., p. 37.'
28 See. for example, McGinn, Wittgenstein on Meaning, chs 2 and 4. For
some further discussion see my critical notice of McGinn, in the
Philosophical (¿aarlerlv (1986), pp. 412-19.
29 David Hume, Treatise o f Human Nature, Pari III; Enquiry Concerning the
Principles o f Morals, Appendix 1.
 Notes 237
30 S. Blackburn, Spreading the Word (1984), ch. 6; ‘Truth, Realism and the
Regulation of Theory’, in Midwest Studies in Philosophy (1900), ed.
French, Uchling and Woltstem.
31 Kripkc, Wittgenstein on Rides and Private language, pp. 89-92.
32 J. McDowell, ‘Wittgenstein on Following a Rule’ (1984); McCJinn,
Wittgenstein on Meaning.
33 Heal, ‘The Disinterested Search for Truth’ (1987—8).

C ha pter 8 I n t e r p r e t a t io n s a n d M is in t e r p r e t a t i o n s

1 It is possible to see this idea in I). Wiggins ‘On Singling out an Object
Delerminately’ in Subject, Thought and Context (I98G), ed. Pettit and
McDowell, csp. pp. 179-80, and Sameness and Substance (1980), chs 4
and 5. Another possible expression of it is P. M. S. Hacker ‘Are
Secondary Qualities Relative?’ (198(5).
2 Philosophical Investigations, section 1.
3 J . I.. Austin, How to do 'I'hings with Words (I9G2).
4 ‘Can Analytic Philosophy be Systematic and Ought it to Be?’, pp.
452-4.
5 Philosophical Investigations, sections 127, 415.
6 The literature here is immense but some central items arc M. Bum-
melt, ‘Wittgenstein’s Philosophy of Mathematics’ in Wittgenstein, Criti­
cal Essays (19(58), ed. Pilcher, and ‘The Philosophical Basis o f
Intuilionistic Logic’, S. Kripkc, Wittgenstein on Rules and Private
Language (1982). Wright’s papers are collected and usefully intro­
duced by him in Realism, Meaning and Truth (198(5).
7 Dummetl, 'Can Analytic Philosophy be Systematic and Ought it to
Be?’
8 C. Wright, ‘Anti-realist Semantics; The Role of Criteria’ in Idealism
Past and Present (1982), ed. Vesey.
9 Introduction to Wright’s Realism, Meaning and Truth.
There is also a line o f argument mentioned by Wright centring on
the interpretation of Wittgenstein's rule-following considerations. On
Wright’s view these considerations arc designed to show the imten-
ability o f ‘objectivity’ about meaning (roughly, minimal realism) and
hence the untenahilily ofDmnmetlian realism. I shall not discuss this
line of thought explicitly. The objectivity of meaning is what the whole
discussion is about and, on the general view I wish to propose, the
rule-following considerations connect with Wittgenstein’s rejection of
mirroring realism (as sketched in section 7 .1 above) and so play an
important role in defusing anti-realist pressure and making objectivity
238 Notes
about meaning defensible. What I would wish to oppose to Wright’s
rule-following argument in favour o f anti-realism is thus implicit in
section 7.1, in the latter part of this section and in ch. 9.
10 Wright, Realism, Meaning and Truth, p. 26.
11 Ibid., pp. 16-17.
12 For more on the difficult topic o f criteria see S. G'avcll, The Claim o f
Reason (1979), csp. Part 1, and J. McDowell, ‘Criteria, Dcfcasibility
and Knowledge’ (1982).
13 The argument o f this section owes a great deal to J. McDowell
‘Anti-Realism and the Fpistcmology o f Understanding’ in Meaning and
Understanding (1981), ed. Parrctt and Bouvcrcssc.
14 This is a point pressed by K.J. Craig in ‘Meaning, Use and Privacy’,
and by S. Blackburn in Spreading the Word (1984), pp. 64-6.
15 Wright, Realism, Meaning and Truth, pp. 20-2.
16 Wittgenstein’s interlocutor asks ‘But how can human understanding
outstrip reality and itself think the uuvcriiiablc?’ lie answers himself,
‘Why should we not say the uuvcriiiablc? For we ourselves made it
unvcrifiablc’ (Zellel, section 259).
17 It is worth noting that Wright is attached to the mirroring idea and to
the reality rcprcscnting/non-rcaiity representing distinction to which
it gives rise. He explicitly remarks on the mirroring idea as an element
in ‘realism’ —but he docs not seem to think it likely to be one which
will cause difficulties. See, for example, Wright’s Realism, Meaning and
'Truth, pp. 6-7 and ‘ Inventing Logical Necessity’ in Language, Mind and
Logic (1986), ed. Butterfield, pp. 19511".
18 For more on this issue see D. Kdgington ‘The Paradox of Knowabiiity’
(1985) and ‘Verification and the Manifestation of Meaning’ (1985);T.
Williamson, ‘Intuilionism Disproved’ (1982); and Wright, Realism,
Meaning and Truth, pp. 309-16.
19 H. Putnam, Reason, Truth and History (1981), ch. 2; Meaning and the
Moral Sciences (1978), p. 125.

C h a pter 9 I n t e r e s t s , A c t i v i t i e s a n d M e a n in g s

1 See for example Remarks on the Foundations o f Mathematics, III 3 Iff, III
41, V 45, VII 45.
2 Ibid., I 168, III 27, IV 23, VI 8, 24.
3 M. Dummctt, ‘Wittgenstein’s Philosophy of Mathematics’ in Wit­
tgenstein, Critical Essays (1968), ed. Pitcher; C. Wright, Wittgenstein on the
Foundations of Alalhematics (1980).
4 Philosophical Investigations, 186, my italics.
 Notes 239
5 Remarks on the Foundations o f Mathematics, III 28.
6 Ibid., V I1 47, VI 24.
7 Ibid., IV 29, VI 23, VII 21.
8 C. Wright, Wittgenstein on the Foundations o f Mathematics (1980), pp. 451!’.
9 Cr. ibid., pp. 42-3.
10 Ibid., pp. 431!'.
11 For some other inlcrcstitig cases of conceptual development see T. S.
Kuhn ‘A Function for Thought Experiments’ in ScientificRevolutions
(1981), cd. Hacking.
12 Compare the possibilities discussed by J. Fodor in The Language o f
Thought (1975), pp. 79-84.
13 D. Davidson, *On the Very Idea of a Conceptual .Scheme’, in Ins
Inquiries into Truth and Interpretation (1984).
14 Wittgenstein is interested in these kinds o f differences. Sec Remarks on
the Foundations o f Mathematics, I 42—50, 58, 68, 69.
15 Wright, Wittgenstein on the Foundations o f Mathematics, pp. 43, 96, 44211’.
16 Compare Remarks on the Foundations o f Mathematics, VI 22.
17 As imagined by E. J. Craig in ‘The Problem of Necessary Truth* in
Meaning, Reference and Necessity (1975), cd. Blackburn.
18 Compare an illuminating attempt to do just this in connection with
Wittgenstein’s remarks about measuring in S. Cavcll, The Claim o f
Reason (1979), pp. 115-18.
19 This is a line of Wittgcnstcinian interpretation explored by B. A. ().
Williams in ‘Wittgenstein and Idealism’ (1981); sec also J. Lear
‘Leaving the World Alone’ (1982) and ‘The Disappearing “We” ’
(1984).
20 S. Kripkc, Wittgenstein on Rules and Private IMnguage (1982), pp. 81!’.
21 Philosophical Investigations, Part II, p. 224.
22 There arc arguments which do aim to suggest the existence of just
such algorithms. Quine’s discussion of rabbit and rabbit parts ( Word
and Object (1960), ch. 2) and in Ontological Relativity and Other Essays
(1969) may be regarded as such. Sec also J. Wallace ‘Only in the
Context of a Sentence do Words Have Any Meaning’ 1977; D.
Davidson ‘The Inscrutability of Reference’ in Inquiries into Truth and
Interpretation (1984); H. Putnam Reason, Truth and History (1981), ch. 2.
These arguments deserve consideration which I cannot give them
here.
23 See the Iirown Hook, Part I, 44-9, 60-9, for some of Wittgenstein's
examples.
24 Remarks on the Foundations o f Mathematics, V I, 28.
 Bibliography

Anscomhc, G. I1'.. M., 'The Question o f Linguistic Idealism’, in her From

Parmenides to Wittgenstein, Collected Philosophical Papers, vol. 1
(Oxford: Basil Blackwell, 1981).
Austin,,J. L., How to do Things with Words (Oxford: Clarendon Press, 1962).
Ayer, A. J., Language Truth and Ijrgic (IIannotulsworth, Middx: Penguin
Books, 1971).’
Baker, G. P., and Hacker, P. M. S., Wittgenstein, Meaning and Understanding
(Oxford: Basil Blackwell, 1983).
Blackburn, S., Spreading the Word (Oxford: Clarendon Press, 1984).
‘The Individual Strikes Back’, Synthese (1984), pp. 281-301.
‘Truth, Realism and the Regulation of Theory’, in Midwest Studies in
Philosophy, vol. V, cd. P. A. French, T . IS. Uchling and H. Wcttstcin
(Minnesota: University of Minnesota Press, 1980).
Gavcll, S., The Claim oj Reason (Oxford: Oxford University Press, 1979).
Chomsky, N., ‘Quine’s Empirical Assumptions’, in Words and Objections,
ed. I). Davidson and J. Hintikka, (Dordrecht, Holland: Rcidcl,
1969).
Craig, IS. J., ‘Meaning, Use and Privacy’, Mind (1982), pp. 541-64.
‘The Problem of Necessary Truth’, in Meaning, Reference and Necessity,
cd. S. Blackburn (Cambridge: Cambridge University Press, 1975),
pp. 1-31.
Dancy, J., An Introduction to Contemporary Epistemology (Oxford: Basil Black-
well, 1985).
Davidson, D., Essays on Actions and Events (Oxford: Clarendon Press, 1980).
Inquiries into Truth and Interpretation, (Oxford: Clarendon Press,
1984).
Dummcit, M., ‘Wittgenstein’s Philosophy of Mathematics’, in Wittgens­
tein, Critical Essays cd. G. Pitcher, (London: Macmillan, 1968).
‘Can Analytic Philosophy be Systematic and Ought it to be’ and ‘The
Philosophical Basis of Intuitionistic Logic’, in Truth and Other
Enigmas (London: Duckworth, 1978).
 Bibliography 241
Edgington, D., ‘The Paradox of Knowability’, Mind (1985), pp. 557-68.
‘Verification and the Manifestation of Meaning’, Aristotelian Society
Supplementary Volume (1985), pp. 33-52.
Edwards, James C., Ethics without Philosophy, Wittgenstein and the Moral Life,
(University Presses o f Florida, 1982).
Evans, G., 'Identity and Predication’, Journal o f Philosophy (1975) pp.
343-63. Also in his Collected Papers (Oxford: Clarendon Press, 1985).
Field, IT., ‘Quine and the Correspondence Theory", Philosophical Review
(1974), pp. 200-28.
Fodor, J., The Language o f Thought (New York: Thomas Y. Crowell, 1975).
Modularity oj M ind (Cambridge, Mass.: M IT Press, 1983).
Psychosemantics (Cambridge, Mass.: M IT Press, 1987).
Follcsdal, 1)., ‘Indeterminacy of Translation and Underdetcrmination of
the Theory of Nature’, Dialectica (1973), pp. 289-301.
Gibson Jr, Roger F., ‘ Translation, Physics and Facts of the Matter’, in The
Philosophy o f W. V. Quine, ed. H. Mans and S. Schilpp, (La Salic,
Illinois: Open Court, 1986).
Gregory, R. L., Eye and Brain (London: Weideufeld and Nicholson, 1972).
Hacker, P. M. S., ‘Arc Secondary Qualities Relative?’, Mind (1986), pp.
180-97.
i-Iale, 11., Abstract Objects (Oxford: Basil Blackwell, 1987).
Hahn, H., ‘Logic, Mathematics and the Knowledge o f Nature’ in logical
Positivism, ed. A. J. Ayer (London: Allen and Unwin, 1959).
Hahn, L. E., and Schilpp, P. A., (ed.), The Philosophy o f W. V. Quine (La
Salle, Illinois: Open Court, 1986).
Heal, J., ‘The Disinterested Search lor Truth’, Proceedings o f the Aristotelian
Society (1987-8), pp. 97-108.
‘Replication and Functionalism’, in iMnguage, Mind and Logic ed. J.
Butterfield (Cambridge: Cambridge University Press, 1986), pp.
135-50.
Hookway, C., Quine (Cambridge: Polity Press, 1988).
Hume, D., A Treatise o f Human Nature (Oxford: Clarendon Press, 1978).
Enquiries Concerning Human Understanding and Concerning the Principles o f
Morals (Oxford: Clarendon Press, 1975).
Kirk, R., Translation Determined (Oxford: Clarendon Press, 1986).
Kitcher, P., The Nature o f Mathematical Knowledge (Oxford: Oxford Univer­
sity Press, 1984).
Kripke, Saul A., Wittgenstein on Rules and Private Language (Oxford: Basil
Blackwell, 1982).
Kuhn, 1’. S., ‘A Function for 'Thought Experiments’, in Scientific Revolu­
tions, ed. I. Hacking (Oxford: Oxford University Press, 1981).
242 Bibliography
Lear, J ., ‘Leaving the World Alone’, Journal o f Philosophy (1982), pp.
382-403.
‘The Disappearing “ We” ', Aristotelian Society Supplementary Volume
(1984), pp. 219-42.
Lewis, D. K., ‘General Semantics’, Synthese (1970), pp. 18-67. Also in his
Collected Papers, vol. 1 (Oxford: Oxford University Press, 1983).
McGinn, C., Wittgenstein on Meaning (Oxford: Basil Blackwell, 1984).
McDowell, J., ‘Anti-Realism and the Epistcmology of Understanding’, in
Meaning and Understanding, cd. Herman Parrctt and Jacques
Bouvercsse (New York: Walter de Gruytcr, 1981), pp. 225-48.
‘Non Cognitivism and Rule Following’, in Wittgenstein: to Follow a
Rule, ed. S. I I. Iiolt/.man and C. M. Lcich (London: Routlcdgc and
Kogan Paul, 1981), pp. 141-62.
‘Criteria, Dcfcasibility and Knowledge’, Proceedings o f the British
Academy, vol. LXVIII (Oxford: Oxford University Press, 1982), pp.
455-79.
‘Wittgenstein on Following a Rule’, Synthese (1984), pp. 325-63.
'Functionalism and Anomalous Monism’ in The Philosophy o f Donald
Davidson: Perspectives on Actions and Events, cd. E. Lcpore and B.
McLaughlin (Oxford: Basil Blackwell, 1986), pp. 385-98.
Putnam, H., Meaning and the Moral Sciences (London: Routlcdgc and Kegan
Paul, 1978).
Reason, Truth and History (Cambridge: Cambridge University Press,
1981).
Quine, W. V., From a logical Point o f View (New York: Harper and Row,
1953).
Word and Object (Cambridge, Mass.: M IT Press, I960).
The Ways o f Paradox ami Other Essays (Cambridge, Mass.: Harvard
University Press, 1966).
Ontological Relativity and Other Essays (New York: Columbia University
Press, 1969).
Philosophy of Ij>gic (Englewood Cliffs, NJ: Prentice Hall, 1970).
Theories and Things (Cambridge, Mass.: Harvard University Press,
1981).
‘On the Reasons for Indeterminacy of Translation’, Jom w/ o f Philos­
ophy (1970), pp. 178-83.
‘Facts of the Matter’, in Essays on the Philosopliy of W. V. Quine, cd. R.
W. Shahan and C. Swoycr (Hassocks, Sussex: Harvester Press,
¡979).
R. Rorty, ‘Indeterminacy of Translation and of Truth’, Synthese (1972),
pp. 443-62.
 Bibliography 243
Philosophy and the Mirror o f Nature (Oxford: Basil Blackwell, 1980).
Consequences o f Pragmatism (Brighton, Sussex: Harvester Press, 1982).
Ryle, G., The Concept o f Mind (I^ondon: Hutchinson, 1949).
Strawson, P. F., Scepticism and Naturalism: Some Varieties (London: Met­
huen, 1985).
Wallace, J., ‘Only in the Context of a Sentence do Words Have Any
Meaning’, Midwest Studies in Philosophy, vol. II (University o f Minne­
sota Press, 1977), cd. P. A. French, T. li. Uchling and H. K. Wctlstciu.
Wiggins, D., On Singling out an Object Dclcrminatcly’, in Subject, Thought
and Context ed. P. Pettit and J. McDowell (Oxford: Clarendon Press,
1986).
Sameness and Substance (Oxford: Basil Blackwell, 1980).
‘What Would be a Substantial Theory of Truth?’ in Philosophical
Subjects: Essays Presented to P. F. Strawson, cd. Z. van Straatcti (Ox­
ford: Clarendon Press, 1980).
Williams, B. A. O ., ‘Consistency and Realism’ in Problems o f die Self
(Cambridge: Cambridge University Press, 1973), pp. 187-206.
‘The Truth in Relativism’ and ‘Wittgenstein and Idealism’, in Moral
Luck (Cambridge: Cambridge University Press, 1981).
Descartes: The Project o f Pure Enquiry, (Harniottds worth, Middx: Pen­
guin Books, 1978).
Ethics and the Limits o f Philosophy (London: Fontana Press/Collins,
1985).
Williamson, T., Tntuitionism Disproved’, Analysis (1982), pp. 203-7.
Wittgenstein, L., Philosophical Investigations, 2nd cdn (Oxford: Basil Black-
well, 1958).
The Blue and Brown Books (Oxford: Basil Blackwell, 1958).
Zettel (Oxford: Basil Blackwell, 1967).
On Certainty (Oxibrd: Basil Blackwell, 1969).
Remarks on the Foundations o f Mathematics, 3rd cdn, (Oxford: Basil
Blackwell, 1978).
Wollhcim, R., Art and Its Objects (Harmondsworth, Middx: Penguin
Books, 1970).
Wright, C., Wittgenstein on the Foundations o f Mathematics, (London: Duck­
worth, 1980).
‘Anti-realist Semantics: The Role o f Criteria’, in Idealism Past and
Present, ed. G. Vescy (Cambridge: Cambridge University Press,
1982), pp. 225-48.
‘Kripkc’s Account of the Argument against Private Language’, Jour­
nal o f Philosophy (1984), pp. 759-78.
Realism Meaning and Truth (Oxford: Basil Blackwell, 1986).
244 Bibliography
‘Facts and Certainty’, Proceedings o f the British Academy, vol. LX XI
(1985), Oxford University Press, pp. 429-72.
Inventing Logical Necessity’ in language, Mind and Logic, ed. J.
Butterfield (Cambridge: Cambridge University Press, 1986), pp.
187-210.
‘Does Philosophical Investigations I 258-60 Suggest a Cogent Argument
against Private Language?’ in Subject, Thought and Context, ed. P.
Pettit and J. McDowell (Oxford: Clarendon Press, 1986), pp.
209-66.
‘Scientific Realism, Observation and the Verification Principle’, in
Fact, Science and Morality, ed. G MacDonald and C. Wright (Oxford:
Basil Blackwell, 1986) pp. 247-74.
 Index

absolute conception 33-4, 149-60 fact, factual statement see realism

abstract pattern of demands 95-6, fact/value distinction 162—3
100-2 Field, H. 72-3
analytic truth 54, 03-4, (13 Fodor, J. 90-1
analytical hypotheses 65, 71-2, 79 Follesdal, D. 230
anti-realism 170-91 forms or life 9, 206-10, 213-15
‘assembling reminders’ 176—7, 210 Frege, G . 196
Austin, J. L. 170,172 functionalism 7, 8, 140-1, 152-3
Ayer, A. J. 13G
G ibson, R. F. 78
Blackburn, S. 1G3, 234
holism 2, 5, 46,
candid babbler 98, 105, 109 epistcmological 5, 97, 105-9, 123
Carnap, R. 83 functional 46-7,86-7
causal theory of meaning 90-1, semantic 5, 6, 8, 86-91, 99-100,
104-11, 140, 145-6, 161, 162, 107-8, 141, 218-19
165-6 holistic constraints 91,96-7,
colours 197-8,200-3 100-2, 103-4, 140, 219
conceptual change 192, 194—5, Hookway, C. 232
196-206,211,213 Hume, D. 162, 165
conceptual schemes 30, 31, 32,
132-3, 196-7, 198-9, 210, 224 idealism 26, 27, 28
conventionalism 133, 136-7, 193, imperatives 12, 14-15, 18, 20, 229-30
206, 208 incommensurability 30, 31
convergence in j udgemen t 18-19 incompatibility of linguistic
criteria 184-7, 197-8, 201-2, moves 12-14, 15, 18-20,
203-4 29-30, 32, 126, 138-9, 155, 212,
216-18
Davidson, D. 5-6, 49, 86, 98, 141, indeterminacy 20, 93-5
153, 199, 232, 233 of function 48,50,51,52-3
determinacy of sense 159, 196, 207 of meaning and translation 37, 47,
Duhem’s thesis 46-7, 64 51-2, 54, 68-9, 71-2, 75-6,
Dummett, M. A. E. 8, 136, 178, 77, 78, 91-2, 101-2, 115-21,
179, 180, 191 129, 135
see also meaning scepticism
empiricism 4, 5, 27, 35, 40-1, inscrutability of reference see
45-6, 136-9, 150, 179, 193 ontological relativity
246 Index
instrumentalism 4, 35, 41-2, 46, 47, ontological relativity 71-4, 78, 239
54, 55, 66-9, 120
intelligibility 99-100, 148-9, P-predicates 95,96-7,99,101,111
159-60 pcrspcctival representation 150, 151,
see also semantic holism 153-4, 156-7, 158, 159, 236
intention 12, 181 philosophy of mathematics 9,
interests as connected with 192-6, 210-16
meaning 8, 113, 144-5, 148-9, physicalism 36-7, 38, 39, 68-9,
151-2, 158-9, 160, 165-6, 75-6
168-9, 174-7, 203-6, 208-10, pictorial representation 88, 92-8
215-16, 226-8 Platonism I, 3, 135-9, 192-3, 194,
interpretation see translation 195-6, 207, 208, 209, 214-15
intcrsuhjectivity 33-4 pragmatism 2, 6, 22, 54, 120,
121-9, 130-2, 137, 140, 173, 177
James, H. 134 predictive sentence machine 42
function of parts in 46-54
Kripkc, S. 8,69,112,145, 160 6, self-modifying 56-9
178, 216 setting up 42-5
principle of non-contradiction 15,
language games 169-70 16, 17, 18, 20, 91-2, 111,
I^cibnia, G. W. 72, 73 115-18, 123, 177-8, 217-18
I^cwis, D. K. 49 private language argument 230
proof 192, 194-5, 199-200, 205-6,
McDowell, J. 153 206-7, 210-16
manifestation argument 180, prpjcctivism
181-90 about meaning 163-5
meaning and use 180, 182, 183-4 about necessity 125-6, 214-15
meaning scepticism 4, 5-6, 9-10, about value 162-3
20-1, 69, 111, 112-21, 194, Putnam, II. 191
218-19, 221-2, 224. 227
see also indeterminacy of meaning Quine, W. V. O. I, 2, 3, 4, 5, 7, 8,
aiul translation 20, 25, 35, 36-40, 47, 49-50, 51,
measuring 169, 171, 173-4 54, 55, 60-85, 87, 98, 105, 109,
metaphysical realism 16, 191 112, 120, 153, 221
mind independence 16-18, 24,
26-8 realism 3, 9, 12, 25, 26, 70, 77, 115,
modal realism 54, 133-4, 135-9, 177-8, 179
211-16 minimal 3, 12-20, 21—5, 32, 33,
moral realism 16, 162-3 115-21, 133-4, 139, 148, 149,
mosaic construction 92-8, 102—4 ' 212,214
mirroring 4, 6, 8, 9, 23-4, 27, 28,
naturalism 35-6, 66-7, 79, 82-5, 32-3, 35, 70, 82-4, 135, 140-2,
133, 234 143, 167, 174, 179, 188, 190, 210,
Neurath, O. 124 225-6
normativity of meaning 69, 161, 162, pragmatist 4, 6, 9, 22, 120-1,
HUM 123, 127-9
quietist 4, 9, 24, 176-7, 214-15
observation statement 35, 37, 39, 41, recalcitrant experience 122—3,
61, 63-5, 75, 84, 110-11, 122 124-9
 Index 247
relativism 9, 30-4, 148-9, 210 see also indeterminacy of translation
revisability of logic 55-9,124-9 and meaning
rule following 145-9, 193, 194, truth 18, 22, 151-2, 162, 164, 177
227-8, 237 disquotational theory of 74, 77,
Ryle, G. 89 80, 81, 140

secondary qualities 20 U-predicates 95,96,99, 101, 111

self-stultification 12, 13, 17, 29 undcrdetcrmination of theory by
sense data 27,40-1,47,60,61-3,64 data 46, 51, 52, 75, 77-8
speech acts 169-75, 178
stage setting for meaning 170-1, verification transcendence 26, 178,
226-8 180, 183-4, 188-90, 190-1
Strawson, I*. F. 235
Wiggins, I). 229
‘taking for granted’ 129-34,157, Williams, B. A. O. 8, 149-60, 229,
212 230, 236
‘taking seriously’ 79, 80-1 Wittgenstein, I,. I, 2, 3, 7, 8, 10, 62,
theoretical statements 38, 39, 41, 69, 84, 112-13, 130-4, 139,
110-11, 122 141-9, 159, 164, 167, 169, 170,
totality of possibilities 135, 139-40, 172, 178, 179, 180, 183, 185-7,
141, 159, 196-7,207,208,211, 196, 199, 206-10, 211-16, 221
214, 220-1, 223-5 Wright, C. 8, 122, 124, 125, 126-7,
translation 38,51,68-9,74,98-9, 128, 130, 178, 180, 181, 183, 184,
141-2 194, 195, 239