Context, Truth and Objectivity: Essays on Radical Contextualism 1138094080, 9781138094086

Citation preview

Context, Truth, and Objectivity

The claim according to which there is a categorial gap between m ­ eaning and saying – between what sentences mean and what we say by u ­ sing them on particular occasions – has come to be widely regarded as ­being exclusively a claim in the philosophy of language. The present essay collection takes a different approach to the issues connected to this claim. It seeks to explore the ways in which that claim – as defended first by ordinary language ­philosophy and, more recently, by various contextualist projects – is grounded in considerations that transcend the philosophy of language. More specifically, the volume seeks to explore how that claim is ­inextricably linked to considerations about the nature of truth, knowledge and representation. It is thus part of the objective of this volume to rethink the current way of framing the debates on these issues. By framing the debate in terms of an opposition between “ideal language theorists” and their semanticist heirs on the one hand and “communication theorists” and their contextualist heirs on the other, one brackets important controversies and risks obscuring the undoubtedly very substantial disagreements that exist between different currents of thought. Eduardo Marchesan is a Post-Doctoral Fellow at the Philosophy ­Department of the University of São Paulo. David Zapero is a Research Fellow at the Philosophy Department of the University of Bonn.

Routledge Studies in Contemporary Philosophy

106 New Directions in the Philosophy of Memory Edited by Kourken Michaelian, Dorothea Debus, and Denis Perrin 107 A Pragmatic Approach to Libertarian Free Will John Lemos 108 Consciousness and Physicalism A Defense of a Research Program Andreas Elpidorou and Guy Dove 109 The Value and Limits of Academic Speech Philosophical, Political, and Legal Perspectives Edited by Donald Alexander Downs and Chris W. Surprenant 110 The Significance of Interdeterminacy Perspectives from Asian and Continental Philosophy Edited by Robert H. Scott and Gregory S. Moss 111 Questions of Practice in Philosophy and Social Theory Edited by Anders Buch and Theodore R. Schatzki 112 The Phenomenology of Real and Virtual Places Edited by Erik Malcolm Champion 113 Context, Truth, and Objectivity Essays on Radical Contextualism Edited by Eduardo Marchesan and David Zapero

For more information about this series, please visit: https://www.routledge​.com

Context, Truth, and Objectivity Essays on Radical Contextualism

Edited by Eduardo Marchesan and David Zapero

First published 2019 by Routledge 711 Third Avenue, New York, NY 10017 and by Routledge 2 Park Square, Milton Park, Abingdon, Oxon OX14 4RN Routledge is an imprint of the Taylor & Francis Group, an informa business © 2019 Taylor & Francis The right of the editors to be identified as the authors of the  editorial material, and of the authors for their individual  chapters, has been asserted in accordance with sections 77 and 78 of the Copyright, Designs and Patents Act 1988. All rights reserved. No part of this book may be reprinted or reproduced or utilised in any form or by any electronic, mechanical, or other means, now known or hereafter invented, including photocopying and recording, or in any information storage or retrieval system, without permission in writing from the publishers. Trademark notice: Product or corporate names may be trademarks or registered trademarks, and are used only for identification and explanation without intent to infringe. Library of Congress Cataloging-in-Publication Data CIP data has been applied for. ISBN: 978-1-138-09408-6 (hbk) ISBN: 978-1-315-10625-0 (ebk) Typeset in Sabon by codeMantra

Contents

1 Introduction

1

Dav id Z apero

2 What Is a Statement?

20

J ocelyn B enoist

3 Ordinary Language Philosophy needs Situation Semantics (Or Why Grice needs Austin)

35

K rista L aw lor

4 Beyond Unnatural Doubts: Lessons from Wittgenstein

59

M ic h ael Williams

5 Meaning and Ostension: From Putnam’s Semantics to Contextualism

88

F ran ç ois R ecanati

6 The Role of Intention in Truth

100

E duardo M arc h esan

7 Is Seeing Judging? Radical Contextualism and the Problem of Perception

124

S ofia M i g uens

8 Externalism and Context-Sensitivity

159

Dav id Z apero

9 Contextualism and the Twilight of Representationalism Av ner B az

187

vi Contents 10 Their Work and Why They Do It

205

C h arles T rav is

List of Contributors Index

249 251

1 Introduction David Zapero

In his 1969 inaugural lecture as Waynflete Professor at Oxford, ­Strawson presented a debate – a “Homeric struggle”, he called it – b ­ etween two fundamentally different attitudes towards language and its formalization.1 On one side of this struggle he placed all those ­authors who grant the idea of communication and, more specifically, the idea of a communication-intention, a central role in the philosophical analysis of language. Affiliation to this side of the debate was conditional on an adherence to the following general thesis: the meaning of linguistic expressions is governed by rules and conventions the nature of which can only be adequately accounted for by drawing on the idea of a ­communication-intention. That is, according to the kind of philosophical attitude which the authors on this side of the struggle were meant to give expression to, the rules governing the meaning of linguistic expression cannot be understood independently of the use that speakers make of them for the purposes of communicating with one another. The very simple thought […] which underlies [this] type of ­analysis is that these rules are, precisely, rules for communicating, rules by the observance of which the utterer may achieve his purpose, fulfil his communication-intention; and that this is their essential character”. 2 Consequently, on such a conception of language, the prime function of language is to enable communication between members of a community. It is natural to construe the case in favor of such an outlook as an argument in two stages. First one fleshes out an adequate notion of communication, then one shows that the rules governing meaning are, essentially, rules enabling communication. It is of course essential that the first step of the argumentation does not rely on the second; that is, that one’s notion of communication does not rely on the rules which that very notion is meant to help elucidate. This means that the notion of communication cannot involve any reference to linguistic meaning: it must be a pre-linguistic notion. This does not entail ruling out the possibility that the rules of a language should enable the emergence of

2  David Zapero novel communication-intentions, unavailable to a speaker without the mastery of the relevant rules. Yet, if the nature of those rules is to be accounted for in terms of communication-intentions, the notion of communication that one works with must ultimately be explicable without making ­reference to those rules.3 On the other side of the Homeric Struggle Strawson placed authors whom he called the “theorists of formal semantics”. The authors who belong on this side of the debate do not deny that language is a powerful tool of communication; they do not deny that the rules governing meaning can be – and often are – put to such a use. But they deny that the nature of these rules is explicable in those terms. They claim, instead, that the general character of the rules governing the meaning of linguistic expressions must be understood with reference to another notion, namely the notion of truth – or, more specifically, the notion of a truth-condition. On this view, the rules that determine the meaning of sentences achieve their task by fixing the conditions under which those sentences are true. It is thus to the idea of a truth-condition that one must turn to understand the nature of those rules. Naturally, the theorists of formal semantics acknowledge that ­sentences cannot be said to be true or false in isolation. Due to context-­ dependent elements in the sentence, different utterances of the same sentence will have different truth-values. But this fact does not in itself pose any fundamental problem for the theorist of formal semantics: it merely obliges him to suitably relativize his account of meaning to the contextual conditions of utterance. A theory of the rules of a language will not simply be a theory of conditions under which the declarative sentences of that language are true; it will, instead, be a theory of the different types of conditions under which a given sentence will issue in different particular truths. This way, the theorist of formal semantics acknowledges the existence of context-dependent linguistic expressions while keeping in place his central claim that the rules governing meaning have to be understood with reference to the notion of truth. On this modified version of his claim, the meaning of a sentence is determined by the rules that determine which statement would be made by someone uttering the sentence on a particular occasion; and which statement is made on an occasion is in turn explicated in terms of truth-conditions. The rules governing meaning thus continue to be accounted for in terms of the notion of truth: the meaning of an expression is construed as the contribution that that expression can make to a truth-evaluable content. The Homeric Struggle that Strawson set out was thus, in the first ­instance, a conflict about the primary purpose or function of language: what united the authors on either side of the debate was a certain general conception of the prime function of language. On one view, the primary purpose of language is communication; on the other, it is the making of (true or false) statements, the expression of thought. But this

Introduction  3 disagreement about the purpose of language is closely tied to another disagreement. The schism that Strawson presented can also be understood as a conflict between two different views on how formal calculi or so-called formal languages relate to natural languages. For it is of course not with reference to just any conception of truth that Strawson’s theorist of formal semantics seeks to explicate the nature of the rules governing meaning. His claim is that a conception of truth of the kind developed by Tarski, one applicable to formal systems, can provide the key for a satisfactory theory of meaning. Therefore, in tying the theory of meaning to a (certain) theory of truth he is also establishing a close relation between natural languages and so-called formal languages. And in rejecting the claim that a semantic conception of truth can ultimately account for the nature of the rules governing meaning, his adversary, the theorist of communication-intention, also rejects the establishment of such a close link between natural and formal languages; he rejects the idea that the  “nature” of linguistic meaning can be captured by formal systems. The Homeric Struggle about whether the rules governing the meaning of linguistic expressions (of natural languages) can ultimately be accounted for in terms of a theory of truth is thus at the same time a Struggle about the depth of the gulf that separates formal languages from natural ones.4 In this respect, there is an important continuity between the debate that Strawson presented in his inaugural lecture and more recent debates about the distinction between semantics and pragmatics. 5 For while these more recent discussions are not – or at least not explicitly – about the nature or prime function of language, they do involve – and are explicitly motivated by – a reflection on the relation between natural languages and formal systems. Here too the fundamental issue is whether there can be a seamless transition from formal systems to a theory of meaning of natural languages or whether the two things are, rather, separated by an unbridgeable gulf. The issue is approached from a quite different angle, though. The point of departure of these recent debates is the question of just how pervasive and important context-dependent elements are in natural languages. On one position, the importance of such elements is minimal. Such a ­“minimalist” position acknowledges that natural language ­sentences do not, in general, express a proposition or a truth-evaluable content. The fact that sentences should contain a host of different kinds of ­context-dependent elements – of which indexicality is the paradigmatic type – means, for the minimalist, that one can only determine a truth-­ evaluable content by taking into consideration the circumstances in which a sentence is used. Yet, despite this, minimalists consider that that appeal to context is well delimited in the following sense: the ­(semantic and syntactic) rules of the language establish just what role the context plays in determining content. Thus, on this view, the truth-evaluable ­content of a declarative utterance departs only ­minimally from the l­inguistic meaning

4  David Zapero of the sentence type; it diverges only to the extent that the meaning of some element of the sentence involves the assignment of a contextual value. On the other hand, there is the view that the gap between the ­linguistic meaning of a sentence and the truth-evaluable content of an utterance – “what is said” by using a sentence on an occasion – goes much deeper and has more significant consequences. According to this view, which generally goes by the name of “contextualism”, the influence of context on the determination of content is not “linguistically controlled”: the rules of the language do not establish univocally how context will affect the task of determining the content of – “what is said” by m ­ aking – an utterance. The transition from the meaning of a sentence to a truth-­evaluable content involves not only “bottom-up processes” in which the rules of the language determine univocally what contribution the circumstances of an utterance can make to the content that is expressed by using a given sentence; the transition also involves, on this view, ­“top-down processes” in which the context of the utterance determines the content expressed in ways that aren’t fixed in advance by those rules. Therefore, the appeal to “pragmatic” considerations is indispensable to determine the content of an utterance: to arrive at a truth-­evaluable content it doesn’t suffice to apply the rules of the language to the context at hand by plugging values into the functions that serve to model the ­context-dependent elements of a sentence. For the ways in which context determines content cannot be accounted for solely by drawing on the rules of the language. The gulf between the meaning of a sentence and what is said by using the sentence on an occasion can only be bridged by relying on considerations that transcend the rules governing the meaning.6 In making this claim about an unbridgeable gulf between meaning and saying, the contextualist rehabilitates an idea that was central to ordinary language philosophy. In denying that truth-evaluable content can be ascribed to sentences in virtue of their meaning, the contextualist concurs with Austin when he writes: If you just take a bunch of sentences […] impeccably formulated in some language or other, there can be no question of sorting them out into those that are true and those that are false; for […] the question of truth and falsehood does not turn only on what a sentence is, nor yet on what it means, but on, speaking very broadly, the ­circumstances in which it is uttered.7 The same point also plays a central role in Strawson’s influential ­argument against Russell: [An] expression […] cannot be said to mention, or refer to, a­ nything any more than the sentence can be said to be true or false. The same expression can have different mentioning-uses, as the same

Introduction  5 sentence can be used to make statements with different truth-values. ­‘Mentioning’, or ‘referring’, is not something an expression does; it is something that someone can use an expression to do. Mentioning, or referring to, something is a characteristic of a use of an expression, just as ‘being about’ something, and truth-or-falsity, are ­characteristics of a use of a sentence.8 There are of course some differences between these two passages. ­Austin’s remarks contain an indication of what, in his view, truth and falsehood turn on (the circumstances of the utterance), whereas S­ trawson’s don’t. I will return to this issue shortly.9 Both, however, make a crucial negative claim, namely that sentences aren’t in the business of being true (or false); only the use that we make of a sentence – that is, only a particular statement – can, they both insist, be legitimately ­characterized as true or false. This claim – which I will, from now on, simply refer to as “Austin’s Claim” – can at first seem to be a merely grammatical one, and thus quite trivial. The insistence that statements, and not sentences, are true or false can sound like the insistence that only a person using a calculator, and not the calculator itself, can multiply two numbers. But Austin and Strawson are of course not primarily concerned with the dangers of metonymy. Their claim is a quite substantial one about the relation between meaning and truth. Even bracketing ambiguity and the systematic fixing of reference to context, what a sentence means does not yet determine what is said when one uses it on an occasion. One may, on the contrary, say an indefinite number of things on different occasions by using words alike in meaning. “What is said” or “what is stated” refers here to the kind of stance that is evaluable in terms of truth and falsity, and the point can thus be reformulated in the following way: for different speakings of words alike in meaning, there will be an indefinite number of different ways the world must be to be as it was, on the different speakings, stated to be. Even if one takes into account the systematic ways in which the meaning of a sentence depends on the context of its utterance, there is no way of determining, solely by relying on that meaning, how the world must be to be as it is stated to be on a given use of that sentence. There are, however, various, substantially different ways of understanding this claim. One way – the more prominent way – of ­cashing out the claim is the following. What is stated, on the different speakings, is not only determined by the meaning of the sentence and other linguistically well-delimited factors, but it also depends on pragmatic considerations which cannot be accounted for by relying solely on the rules governing the meaning of the sentence. This way of fleshing out the claim involves acknowledgement that what is stated, on different occasions, depends on what have been called “top-down” pragmatic

6  David Zapero processes – but, on this understanding, what one arrives at, when one takes into account those pragmatic considerations, is the kind of content which the critics of such a claim think can be reached solely by relying on the meaning of the sentence. Whether, under different circumstances, things would be as they have been stated to be on the present occasion can thus be established solely on the basis of what allowed us to establish whether things were as stated on this present occasion. Or, to put the matter in slightly different terms: what it would be to make, about another situation, the same statement as we made about the present situation is something which we can determine solely by relying on a grasp of the present statement and without attending to the particular ­circumstances of that other situation. In other words, one may defend the idea that the meaning of a ­sentence doesn’t determine what would be said by using it on a given occasion (even bracketing ambiguities and systematic fixing of reference to context), and consider that what is said, on a given occasion, is a proposition that determines, over possible worlds, when what was said would be true. If what is said is a proposition in that sense, then whether what one has stated to be so on a given occasion would also be true on another occasion is determined solely by two things: by what one has stated to be so on the present occasion (i.e. by the relevant proposition) and by how things are in the other situation. The circumstances that would be relevant to decide whether what has been said would be true on that other occasion would be those that were relevant to deciding whether, on the present occasion, what was said is true. Or, putting it again in a more cautious manner: what it would involve to make the same statement about another situation is not something that requires taking into account the particularities of the situation that one would make the statement about but would be wholly determined already by the present statement. Indeed, herein resides the purpose of the notion of a possible world: by partitioning possible worlds into those in which a statement would be true and those in which the statement would be false, one can determine independently of the particular circumstances of a situation what matters for establishing whether a statement made about another situation is true or false. Of course, for the authors that understand Austin’s Claim in this way, my presentation of their position will seem overly laborious and even misleading. The idea that one could establish, solely by drawing on what is said on a given occasion, whether what was then said would also be true on other occasions will be deemed a banality – something following simply from the definition of “proposition” or “what is said”. To have determined what is said on a given occasion just is to have established under what conditions it would be true. And one way of fleshing this out is to say that one has a partition of possible worlds: a partition that distinguishes between the situations in which the relevant conditions are

Introduction  7 satisfied (and the proposition is true) and the situations in which they are not satisfied (and the proposition is false). It is, on such an understanding of the Claim, simply part of “what is said” that it should determine what, in other circumstances, should be relevant towards deciding whether what was said is or is not true. What can make my analysis seem laborious is that, on my account of the matter, an attempt to bridge the gulf between meaning and truth involves not just one but two major steps. There is, firstly, the step from the meaning of a sentence to the proposition expressed on an occasion and, secondly, the step from the proposition expressed to the conditions under which that proposition is true. The understanding of Austin’s Claim that I’ve just outlined focuses on the first step: there are, it insists, distinctly “pragmatic” factors, irreducible to any “semantic” considerations, that contribute towards determining what proposition is expressed on an occasion. Authors that defend such a claim thus concede that that first step already requires an appeal to considerations that transcend “semantics”. But what I have taken to be a second possible gap between meaning and truth will not, for such authors, be a gap at all. And it will not be a gap for reasons that are (allegedly) wholly trivial: a proposition is true when things are as they are according to it. Searle, for instance, defends the claim that what would be said by ­using a sentence on an occasion can only be accounted for by taking into account a host of pragmatic considerations that are irreducible to semantic ones.10 In the debates about the distinction between pragmatics and semantics, he is thus a prime representative of what is generally labeled the contextualist (or even radical contextualist) position. Yet, while he holds that distinctly pragmatic considerations play an essential role in determining what is said on an occasion, he also holds that what is said on an occasion fixes, unambiguously, whether another situation would also (or not) be as thus stated. On this view, once one has taken into account the relevant “background assumptions”, one arrives at the kind of proposition that his adversaries (amongst which is Strawson’s “theorist of formal semantics”) seek to arrive at solely by relying on the syntactic and semantic rules of the language. Searle’s exposition has the merit of making this point fairly clear. For he does not, like Austin and Strawson, present his point in terms of a categorial distinction between the meaning of a sentence on the one hand and what is said on the other. Instead, he frames his claim as one about the assumptions required for a sentence to have “literal meaning” – where that is intended as a claim about the requirements on a sentence’s having “a clear set of truth conditions”.11 The terminology can at first seem misleading. After all, Searle’s claim is one conceivable way of cashing out the idea that only a speech act, made on an occasion, can be true or false. It can seem puzzling that Searle should frame that claim in terms of what is required for a sentence to have literal meaning. Usually,

8  David Zapero this last term is used to denote something that sentences possess solely in virtue of the syntactic and semantic rules of a language; it is thus a matter of definition that they should have such a meaning independently of the use that one makes of them. Yet, if one takes into account Searle’s conception of saying, the ­terminological choices are not as misleading as they may first seem. A sentence, Searle claims, can only have a “clear set of truth conditions” when a particular set of contextual background conditions ­obtain. Those background conditions are not themselves specified in the ­sentence and cannot even be specified by the construction of an equivalent sentence (“equivalent” in the sense that the sentence would have the same truth-conditions). Now, having “a clear set of truth conditions” involves the possibility of establishing, for a (presumably, indefinite) number of other scenarios, whether the sentence with the present literal meaning (i.e. the literal meaning it has under the present background conditions) would be true or false. The “clear set of truth conditions” refers, in other words, to the idea of a proposition ranging over possible worlds, and establishing clearly such conditions means being able to determine, without taking into account the different circumstances, whether a situation other than the one at hand would (or not) make the relevant proposition true. Of course, one could reply that the specification of those truth-­conditions already takes into account the different circumstances. The point of specifying such conditions is precisely to establish how, what one has said on an occasion, would fare under different circumstances. But here one uses the term “circumstances” in two different ways. “ ­ Circumstances” in the sense that we previously used the term referred to the specificities of a situation that would be relevant in deciding whether what was said is true or not. The idea being that the kind of considerations that are relevant to deciding the truth of a statement can change from situation to situation. If one now says that the truth-­conditions of what one said on an occasion, in specifying what other situations would make what one said true, take into account changes in circumstance, then one is referring to conditions the relevance of which is established independently of a particular situation. What is of particular value in Searle’s presentation of his claims is that he indicates that what he takes to be truth-evaluable is of the same nature than what his adversary takes to be truth-evaluable. The target of Searle’s argument is the view that we can ascribe truth-conditions to a sentence solely by drawing on the sentence’s meaning and assigning values, in a systematically constrained way, to the context-dependent elements in the sentence. This, Searle insists, is not possible because the truth-conditions of the sentence will depend on a host of background conditions – conditions that aren’t, and also could not be, specified in the sentence. The objection certainly revives an aspect of the Austinian claim

Introduction  9 about sentences not being in the business of being true: only s­ tatements, made under particular conditions that aren’t themselves thematized in the sentence, can be qualified as true or false. Yet, for Searle, once one ­ etermines takes into account the relevant background conditions and d what is said, one arrives at a content which can accomplish just what his adversary hoped meaning would accomplish without the help of the background assumptions. In framing his claim as a claim about what is required for a sentence to have “literal meaning”, Searle makes clear the stakes of the disagreement. What he objects to is not the very possibility of specifying truth-­ conditions in the way that is done by those authors who seek to determine what is said solely by relying on the syntactic and semantic rules of a language. What he objects to is, rather, the way in which such authors seek to arrive at such a specification, the specification that he calls the “literal meaning” of a sentence. That the “literal meaning” of a sentence depends on – and varies with a change in the – background assumptions means that the specification of the truth-conditions cannot be accomplished without taking into account factors that are entirely independent of (what one usually calls) the linguistic meaning of the sentence. It doesn’t mean, according to Searle, that no such specification is possible. The crucial point here is that Searle’s discussion of these issues ­completely severs their connection to questions about truth and thought. As we have seen, the claim that we find in Austin and Strawson to the effect that statements, and not sentences, are in the business of being true inevitably leads to questions about the very idea of being true. Claiming that only statements are in that business involves claiming that the circumstances in which one makes a statement plays a role in determining whether what one thereby states is true. Which then also means: whether what one said is true about another situation depends on the particular circumstances of that other situation. One might consider that these kinds of considerations involve c­ alling into question the very idea of a truth-evaluable content, and that cashing out Austin’s Claim in the way that I’ve done involves abolishing the very idea of a proposition or a thought.12 Whether that is the case or not is not a question I want to, or can, deal with here. The objective of this introductory essay is to deal with an issue that precedes – and is presupposed by – any debate about this question, namely the issue of whether there is a close connection between Austin’s Claim and the notions of truth and thought. The crucial point here is that the claim that only statements are in the business of being true is, inter alia, just what it seems to be when formulated in this way, namely a claim about what kind of thing can enjoy the status of being true (or false). It is a claim about the nature of truth-bearers. And it should thus hardly come as a surprise that the claim might involve a criticism of certain traditional conceptions of such a bearer.13

10  David Zapero I’ve already mentioned one way in which a commitment to a certain conception of such a bearer can obscure the stakes of Austin’s Claim. The idea that we can state an indefinite number of things on different occasions by using a sentence alike in meaning entails, trivially, that different ways for things to be make true what we say on those different occasions. Yet, one way of cashing out this idea is the following: on the different occasions, we give expression to different propositions – where that means: there are, on the different occasions, different ­specifications of conditions under which what we then said would be true. ­Circumstances thus play a role in determining which proposition is expressed on an occasion by using a given sentence, but – and this is the crucial point – they don’t play a role in determining when the proposition expressed would be true. Because a proposition just is a specification of a certain set of conditions under which what was said, on an occasion, would be true. If what was said were not true under precisely those conditions, it would not express that proposition. This kind of appeal to propositions presupposes that there is exactly one set of conditions under which what one said, on a given occasion, would be true. Of course, in one sense, this is trivial. What one said to be so will be true in exactly that range of situations in which things are as they were then stated to be. But it is an entirely different matter to consider that a unique set of conditions can be specified without reference to what was then said. It is, however, such a specification that is required for the aforementioned appeal to propositions. It must, in principle at least, be possible to specify in which other situations what one said would also hold, and one must be able to do so without presupposing that what one said would (or not) hold. In that sense, the specification of situations must play the role of an explanans: it must account for the fact that what one said would hold (or not) in a given situation. What one said would hold in another situation when the conditions that are specifiable independently of the fact that it did (or didn’t) hold are satisfied. Yet, it is the very possibility of such a specification, as we’ve seen, that Austin’s Claim seeks to call into question. That the circumstances can make any of an indefinite number of contributions to determining when what one said would be true means that there is no way of determining, in general, whether what one said on a given occasion would be made true in another situation. Most crucially, it remains open whether the conditions that made what one said true about the present situation would also make it true in the other situation. For it is an open matter whether those conditions would also be relevant in deciding whether what one said was true. Again, one may object by insisting that such an answer avoids the question. The question is whether what we said, under the present conditions, would also be made true about another situation. To say that that depends on the circumstances of the speaking, one may insist, is to

Introduction  11 ­ uestion: ignore the fact that the circumstances have been specified by the q it’s the circumstances of the present speaking. To put the worry here in Kaplanese: we are not simply asking what truth-value a certain sentence would have in another “circumstance of evaluation”; that question would indeed have no answer because the truth-value depends on the content which the utterance gives expression to and that in turn depends on the context in which the sentence is uttered. What we are asking, rather, is what the truth-value is, for a particular circumstance of evaluation, of a sentence uttered in a certain context. The idea that whether or not what was said would hold in another situation depends on the circumstances (“context of utterance”, in Kaplan’s terminology) neglects the fact that those circumstances have indeed been provided for – they are simply the utterances of the statement whose truth-value in other situations we are interested in. But this line of reasoning brings us back to the point of departure. The claim that the circumstances of a speaking can make any of an indefinite number of contributions to what is said on an occasion is not primarily a claim about “character” but about “content”. The idea is not just that there is no way of specifying, context-independently, what content is expressed by a certain character. That in itself, it could be argued, creates fundamental problems for a distinction like Kaplan’s between meaning (“character”) and a proposition (“content”).14 The crucial point, though, is that there is no way of determining context-independently whether or not a given content would be true or false under certain circumstances of evaluation. The claim being that there is no way of specifying those circumstances of evaluation in a way that is independent of the fact that the content is or isn’t true. Any attempt to provide such a context-­ independent specification will be confronted with the dilemma that what is relevant for the determination of the truth-value in one context of ­utterance is not relevant on another one – and that the same conditions can thus, in one case, make the statement true and, in the other, false. The point can be given a non-linguistic form. In representing things as being a certain way, one picks out one particular way for things to be – a way, which, if it obtains, will make the representation true. In representing the ball that I’m perceptually presented with as blue, I pick out a certain way for things to be, namely a way such that the relevant ball is blue. One may speak of partitions here in a way that can seem similar to the way in which one often speaks of propositions as involving a partition of possible worlds: in representing things as being a certain way, one makes a partition between the way for things to be that makes the representing true and an indefinite range of other ways for things to be which would make the representing false. The important point, though, is this: a way for things to be possesses a certain generality. That is, the decision about whether things are the relevant way will take into account certain things and ignore others; there will be things that

12  David Zapero matter and things that don’t. If one has represented a particular ball as being blue, there will be an indefinite number of considerations that are irrelevant towards deciding whether things are as one has represented them to be. The representation can be true both while an owl howls outside and after it has stopped howling. Whether the owl howls or doesn’t howl doesn’t affect the decision of whether the ball in question is blue. Representation thus involves delimiting a range of particular situations – all of which allow one to say that things are as one represented them to be. Of course, all instantiate that way for things to be (ex hypothesi), and in that respect one may speak of things being the same way. But it is, nonetheless, a number of different particular situations. The point that we’re interested in can then be put in the following way: the answer to the question of just what particular situations are part of the range delimited by a representation is something that depends on the particular situation at hand. What features of a given situation are relevant for deciding whether that situation instantiates a way for things to be is not something one can determine independently of a given situation. What features count towards deciding whether things being as they are instantiate a certain way for them to be depends on the particularities of the situation at hand; there is no way of specifying, independently of a situation, what counts and what doesn’t. In some cases, certain yellow spots on the ball may count towards deciding whether a particular ball is blue; in other cases it may not. One way of putting the point is to say that we cannot construct a theory of what it is for a particular situation to instantiate a particular way for things to be: we cannot provide an account that specifies the reach of a representation in terms of conditions that are identifiable independently of the representation’s reaching as it does. No such account is possible because different things will count for different situations; the features that, in one case, make a situation part of the range of situations delimited by a representation will, in another case, not suffice – or even be a reason to exclude – a situation from that same range. It can be tempting to respond that, if that were the case, the reach of a representation would be arbitrary or, at best, conventional. After all, if there is no one set of features – be it at least a vague or fuzzy set – that determines when a representation is true, what allows one to claim that things being as they are, independently of how we take them to be, make the representation true or false? A closely related worry will be: what even allows us to speak of the same representation being true or false in the various different situations? If it isn’t possible to specify a set of conditions under which a given representation is true independently of the fact that it is true, then how are we to individuate the representation? Whether Austin’s Claim has such consequences is, of course, not a matter to be decided here. It should be clear, though, from my presentation of the Claim, that it does at least purport to avoid any such

Introduction  13 consequences. The Claim, as I’ve presented it, seeks to do justice to what one may call the objectivity of truth, namely to the idea that being true is being as things are anyway, independently of the fact that one took them to be that way. I’ve reformulated Austin’s Claim in the following way: what it amounts to for things to be a certain way cannot be determined independently of a given situation because different considerations may matter on different occasions. Now the important point for present purposes is that the claim is one about what is involved in the instantiation of a way for things to be. The fact that different considerations are relevant on different occasions is supposed to leave room for the idea that, on all the occasions where a given way for things to be is instantiated, things are that way. The fact that being of a certain hue of blue may amount to different things on different occasions is supposed to leave room for the possibility that the things that fall under that concept do so simply in virtue of being that hue of blue. Put in slightly different terms, the fact that it should not be possible to specify a way for things to be in more basic terms is not taken to exclude the possibility that such ways are ways for things to be. The claim certainly involves a certain antireductionism about truth and representation: it excludes the possibility of accounting for the reach of a representation in more basic terms. But that antireductionism isn’t taken to call into question the objectivity of truth. The claim is meant to leave room for the idea that the truth of a representation depends solely on the representation, on the one hand, and on how things actually are, on the other. One may of course think that this intent to do justice to the objectivity of truth is just that: an intent. If one gives up on the possibility of specifying a set of (independently specifiable) conditions under which a representation would be true, one must – the idea would be – also give up on the claim that the truth of a representation is determined by how things are anyway. Once merit or relevance are allowed to play a role in deciding whether a representation is true – once truth is held to be a “dimension of assessment”15 – there is no way of keeping in place a full-blooded conception of objectivity. So one could argue, and perhaps correctly. But while its appeal to circumstance and to merit may turn out to be incompatible with the objectivity of truth, it is no part of Austin’s Claim to call that objectivity into question. Perhaps it does so by entailment, but that is then something that must be shown, not assumed. If one acknowledges that Austin’s Claim is, inter alia, one about truth and objectivity, one is led to reconsider the question of what criteria should be used to capture deep differences in attitude towards language and its formalization. That is, viewing that claim in the wider context in which I have set it inevitably raises questions about the fault lines that have generally been taken to be fundamental in the philosophy of language. On the picture that Strawson presents us, for instance, such authors as

14  David Zapero ­ avidson all fall on one Frege, early Wittgenstein, Russell, Carnap, and D side of a divide – as either pioneers or full-blooded ­spokespersons of the “theory of formal semantics” – and such authors as Austin, Grice, later Wittgenstein, and, presumably, Searle, all fall on the other side of that same divide – as “theorists of communication-intentions”. In the context of contemporary debates about the relation between semantics and pragmatics, the fault lines, as we’ve seen, are usually drawn in a similar way. Of course, no one seeks to deny the profound differences between the various authors on either side of the divide. In proposing such a categorization, one need not neglect the multiple disagreements between the various authors. The divide does, however, reflect a certain conception of what one takes to be the most fundamental issues or questions in the field. At the very least, it gives expression to the relative importance that one grants certain problems. Now, one could claim that by setting up the kind of divide that Strawson does – a divide that is still widely seen as pertinent to understand debates on these issues – one makes a quite uncontroversial decision: one gives particular weight to the question of how linguistic meaning relates to truth and one accounts for different philosophical attitudes towards language in terms of the general approach to this question. But, as we’ve seen, this is not quite right. In fact, this way of framing the debate is grounded in something more specific. It relies on the assumption that a question along the following lines allows to capture the fundamental fault lines between different ­philosophical attitudes towards language: can a theory of truth of the sort that Tarski provided for a formal language be used to account for the way in which the linguistic meaning of a sentence determines what is said when one uses that sentence to make a statement? That is, one grants fundamental importance to a question about whether a certain kind of transition is possible between what a sentence means and what is said when that sentence is used on an occasion. If one views the various positions in this light, it will seem as if the party which denies that such a transition is possible – the contextualist – is simply defending what we have dubbed Austin’s Claim. For what one will retain, in this Claim, is that it involves a denial that the relevant kind of transition is possible. What I’ve suggested, though, is that these two denials are fundamentally different in nature. They are responses to two substantially different questions. When the theorist of communication-intentions or, more recently, the contextualist, denies that appeal to a formal theory of the sort first proposed by Tarski can help us explain what is said on an occasion by using a given sentence, he insists on the (supposed) limits of any such explanation: such an explanation, he claims, will not by itself succeed in establishing a systematic connection between a given sentence and the statements that can be made with that sentence. Yet, while he denies  the possibility of establishing the sort of systematic connection

Introduction  15 which the “semanticist” seeks to establish, he does consider that, once other ­(extra-semantic) factors are taken into consideration, one will have specified the same thing that the semanticist was hoping to specify without appeal to such other factors. Most crucially, once one has taken into account the relevant pragmatic considerations and arrived at the real bearer of truth, namely a statement, one has a specification of the situations under which what was then said, on the relevant occasion, would be true. Yet, such a specification requires that one be able to establish whether what one said to be so on a given occasion would be true about other situations than the one in question without attending to the particular circumstances of those other situations. Or, put slightly differently, the specification presupposes that we can establish what it would involve to make the same statement – say that things were as we said they were said to be with respect to the present situation – without attending to circumstances of those other situations. And this is a crucial part of what Austin’s Claim targets: the idea that it should be possible to establish, without attending to the particular circumstances of another situation, what it would involve to say the same thing as one said about the present situation. In Strawson’s “Homeric struggle“, questions about truth still play a central role. The debate between the two parties is, at bottom, a debate about the nature of the truth-bearer. The communication theorist’s case against the theorist of formal semantics turns on the idea that the notion of a truth-condition can only be explicated with reference to the notion of a statement and that this notion requires, in turn, reference to a ­communication intention. [W]e cannot, the theorist maintains, elucidate the notion of ­stating or asserting except in terms of audience-directed intention. For the fundamental case of stating or asserting, in terms of which all ­variants must be understood, is that of uttering a sentence with a certain intention […] which can be incompletely described as that of letting an audience know, or getting to think, that the speaker has a certain belief”.16 In this respect, Strawson’s conception of the debate can be contrasted with more recent conceptions, such as Searle’s, in which the link to ­questions about truth is entirely lost and the notion of a truth-bearer no longer plays a role. Yet, despite the prominent role that questions of truth still play in the Homeric Struggle, the debate is framed as one about the nature of ­linguistic meaning – about whether the nature of the rules governing such meaning are ultimately explicable with reference to the notion of truth or rather with reference to the notion of communication. As Strawson

16  David Zapero himself points out, by viewing the debate in this way, one presupposes that the opposing parties share a certain idea of a truth-condition: For it is a truth implicitly acknowledged by communication-theorists themselves that in almost all the things we should count as sentences there is a substantial central core of meaning which is explicable ­either in terms of truth-conditions or in terms of some related notion quite simply derivable from that of a truth-condition, for example, the notion, as we might call it, of a compliance-condition in the case of an imperative sentence or a fulfilment-condition in the case of an optative. If we suppose, therefore, that an account can be given of the notion of a truth-condition itself, an account which is indeed independent of reference to communication-intention, then we may reasonably think that the greater part of the task of a general theory of meaning has been accomplished without such reference.17 On Strawson’s understanding of the debates about the relation of ­meaning and truth, the idea of a truth-condition is taken to be unproblematic. Both parties hold that the rules governing the meaning of a sentence can be explicated with the help of such a notion; for both, such meaning can be accounted for by a specification of the conditions under which what would be said, in using the sentence on an occasion, would be true. The issue on which they diverge is the question of whether that specification requires appeal to the idea of a communication-intention, or whether any such appeal is optional. If one considers that the debates about the relation between meaning and truth are also – or even primarily – debates about the legitimacy and purpose of the very idea of a truth-condition, one is led to reassess the divide that Strawson presents us with and that has been widely accepted. From such a vantage point, the question that Strawson takes to be fundamental – the question about the nature of the rules governing meaning – is a secondary matter, or, more precisely, it is an issue that is debated on one side of what one then takes to be a more basic divide. Questions about the nature of such rules are grounded in a conception of the connection between meaning and truth, which, on this other view of the issues, is the main target of the claim that only statements are in the business of being true. The question which allows to capture the fundamental disagreements is, on this alternative view, the question of the nature of a truth-bearer and, more specifically, the role that one ascribes to circumstances in an account of truth and objectivity. Naturally, this view that I have sketched is not uncontroversial. It ­involves a take on various issues – both exegetical and conceptual – that continue to be subject to much debate. Indeed, various contributors in the present volume take positions that are at odds with such an understanding of the issues. But the preceding sketch is not meant to give expression to a

Introduction  17 consensus that underlies all the contributions to the ­volume. It is instead meant as a presentation of the kind of approach to those controversies that has motivated the editors in putting together this essay collection. ­ eaning The claim according to which there is a categorial gap between m and saying has come to be widely regarded as being primarily a claim in the philosophy of language. The various contemporary endeavors to defend such a claim – endeavors that have, not implausibly, been dubbed “contextualist” or “radical contextualist” – are generally understood as being a position on the role that pragmatics should play in our theory of language and meaning. Of course, understanding the claim in this way doesn’t exclude acknowledging the possibility that it may have important consequences for other domains of inquiry. Yet, on such an understanding, those consequences will be exactly that – consequences of a claim that is conceptually wholly autonomous. The present essay collection takes a different approach to these issues. It seeks to explore the ways in which the aforementioned claim – as defended both by ordinary language philosophy and, more recently, by various contextualist endeavors – is grounded in considerations that transcend the philosophy of language. More specifically, the volume seeks to explore how that claim is inextricably linked to considerations about the nature of truth and representation. It is thus part of the objective of this volume to rethink the current way of framing the debates on these issues. By framing the debate in terms of an opposition between “ideal language theorists” and their semanticist heirs on the one hand and “communication theorists” and their contextualist heirs on the other, one brackets certain controversies and risks obscuring the undoubtedly very real oppositions that exist between different currents of thought. An emphasis on communication and intention can coexist with quite different conceptions of truth and saying. One may stress the social dimension of language while adopting a conception of truth according to which the objectivity of truth requires that reference to context and circumstances ultimately be eliminable. The profound differences between such a view and views which take the role of context to be irreducible thereby become obscured. By failing to acknowledge that issues about truth and representation are at the root of these controversies and do not just enter at a later stage one risks missing the central points of contention.18

Notes 1 Strawson, “Meaning and Truth.” 2 Ibid., p. 173. 3 Strawson is of course thinking here of Grice’s analyses, primarily in Grice, “Utterer’s Meaning, Sentence-Meaning, and Word-Meaning.” 4 Strawson himself touches on this dimension of the opposition in: ­“Logical Form and Logical Constants”; “'If’ and ‘⊃’”; “Carnap’s Views on ­Constructed Systems versus Natural Languages in Analytic Philosophy”; Introduction to Logical Theory, Chapter 8.

18  David Zapero 5 Bianchi, The Semantics/Pragmatics Distinction; Borg, Minimal Semantics; Cappelen and Lepore, Insensitive Semantics; Predelli, Contexts; Preyer and Peter, Contextualism in Philosophy; Récanati, Literal Meaning; Récanati, Truth-Conditional Pragmatics; Stanley, Language in Context. 6 Needless to way, there are a host of intermediary positions in the debate. One may go as far as acknowledging the existence of “top-down” pragmatic processes, as the contextualist does, but still try to hold on to the idea of a minimal, truth-evaluable content, namely a content that is entirely determined by the syntactic and semantic rules of the language. On this view, an appeal to ­extra-semantic, i.e. pragmatic, considerations is indispensable to determine what is said on an occasion; but the idea of a minimal proposition is still retained – on the basis, for instance, that it plays a role in the process of communication. Cf. Bach, “Conversational Impliciture” and Salmon, “The Pragmatic Fallacy.” 7 Austin, Sense and Sensibilia, p. 110. 8 Strawson, “On Referring,” p. 8. 9 On these differences, cf. Benoist, “Les actes de langage, entre invention et convention.” 10 Cf. Searle, “The Background of Meaning” and Searle, “Literal Meaning.” 11 “For a large class of unambiguous sentences such as ‘The cat is on the mat’, the notion of the literal meaning of the sentence only has application relative to a set of background assumptions. The truth conditions of the sentence will vary with variations in these background assumptions; and given the absence or presence of some background assumptions the sentence does not have determinate truth conditions. These variations have nothing to do with indexicality, change of meaning, ambiguity, conversational implication, vagueness or presupposition as these notions are standardly discussed in the philosophical and linguistic literature” (“Literal Meaning,” p. 214). 12 Cf. for instance Fodor, Hume Variations, pp. 96–111. 13 It is worth noticing that Austin, in the passage quoted above, explicitly ­targets the idea of a proposition. For purposes of legibility, we omitted the reference to propositions. But Austin explains that neither sentences nor propositions, “to use the term Ayer prefers,” are in the business of being true. And he then, in a footnote, adds the following remark about Ayer’s use of that term: “The passage in which Ayer explains his use of this term (p. 102 [of The Foundations of Empirical Knowledge]) obscures exactly the essential point. For Ayer says (a) that in his use ‘proposition’ designates a class of sentences that all have the same meaning, and (b) that ‘consequently’ he speaks of propositions, not sentences, as being true or false. But of course to know what a sentence means does not enable us to say that it is true or false; and that of which we can say that it is true or false is not a ‘proposition’, in Ayer’s sense” (Sense and Sensibilia, p. 110). 14 Cf. Travis, “Meaning’s Role in Truth.” 15 Austin, How to Do Things with Words, p. 149. 16 Strawson, “Meaning and Truth,” p. 181. 17 Ibid., p. 178. 18 Research for this paper was supported by funding from the Fritz Thyssen Foundation. I am deeply indebted to Charles Travis for comments and discussion of a previous version of this introduction.

Bibliography Austin, John L. How to Do Things with Words. Edited by James O. Urmson. 2nd ed., [repr.]. Cambridge, MA: Harvard University Press, 2009.

Introduction  19 ———. Sense and Sensibilia. Edited by Geoffrey James Warnock. Repr. ­London: Oxford University Press, 2010. Bach, Kent. ”Conversational Impliciture”. Mind & Language 9, no. 2 (June 1994): 124–162. Benoist, Jocelyn. “Les actes de langage, entre invention et convention”. In ­L angage ordinaire et métaphysique: Strawson, edited by Sandra Laugier and Jocelyn Benoist, 209–247. Paris: Vrin, 2005. Bianchi, Claudia, ed. The Semantics/Pragmatics Distinction. CSLI L ­ ecture Notes, no. 155. Stanford, CA: Center for the Study of Language and ­I nformation, 2004. Borg, Emma. Minimal Semantics. Oxford: Oxford University Press, 2008. Cappelen, Herman, and Ernest Lepore. Insensitive Semantics: A Defense of Semantic Minimalism and Speech Act Pluralism. Chichester: John Wiley & Sons, 2007. Fodor, Jerry. Hume Variations. Oxford: Clarendon Press, 2005. Grice, Herbert Paul. “Utterer’s Meaning, Sentence-Meaning, and Word-­ Meaning”. In Philosophy, Language, and Artificial Intelligence, edited by Jack Kulas, James H. Fetzer, and Terry L. Rankin, 49–66. Dordrecht: Springer Netherlands, 1968. Predelli, Stefano. Contexts: Meaning, Truth, and the Use of Language. Oxford: Clarendon Press, 2005. Preyer, Gerhard, and Georg Peter, eds. Contextualism in Philosophy: ­Knowledge, Meaning, and Truth. Oxford: Oxford University Press, 2005. Récanati, François. Literal Meaning. Cambridge: Cambridge University Press, 2004. ———. Truth-Conditional Pragmatics. Oxford: Clarendon Press, 2010. Salmon, Nathan. “The Pragmatic Fallacy”. Philosophical Studies 63, no. 1 (1991): 83–97. Searle, John. “Literal Meaning”. Erkenntnis 13, no. 1 (1978): 207–224. ­ ragmatics, ———. “The Background of Meaning”. In Speech Act Theory and P edited by John Searle, Ferenc Kiefer, and Manfred Bierwisch, 221–232. ­Dordrecht: Springer Netherlands, 1980. Stanley, Jason. Language in Context: Selected Essays. Oxford: Oxford ­University Press, 2007. Strawson, Peter F. “Carnap’s Views on Constructed Systems versus Natural Languages in Analytic Philosophy”. In The Philosophy of Rudolf Carnap, ­e dited by Paul Arthur Schilpp. The Library of Living Philosophers 11, 503–518. La Salle, IL: Open Court, 1991. ———. “‘If’ and ‘⊃.’” In Entity and Identity, 162–178. Oxford: Oxford ­University Press, 2000. ———. Introduction to Logical Theory. New York: Routledge, 2011. ———. “Logical Form and Logical Constants”. In Entity and Identity, 142–161. Oxford: Oxford University Press, 2000. ———. “Meaning and Truth”. In Logico-Linguistic Papers, 170–189. London: Methuen, 1971. ———. “On Referring”. In Logico-Linguistic Papers, 1–27. London: Methuen, 1971. Travis, Charles. “Meaning’s Role in Truth”. Mind 105, no. 419 (1996): 451–466.

2 What Is a Statement? Jocelyn Benoist

One striking feature of Austin’s philosophical perspective is the register in which it locates the truth-bearing function. The British philosopher jettisons the idea of an ideal truth-bearer, such as a ‘proposition’, and replaces it by a new kind of truth-bearer – one which is meant to be ­concrete and in some sense ‘real’: what he calls a ‘statement’. In the new framework of analysis that Austin develops, the notion of a statement is taken to be an elementary notion, whereas a notion like that of a proposition is considered to be a ‘logical construction’. In How to Do Things with Words, he suggests that we “consider a statement not as a sentence (or proposition) but as an act of speech (out of which the ­others are logical constructions)” (HTDTWW, 20). Admittedly, in the passage from which we have extracted the quotation, the ­recommendation is ­primarily meant to change our way of conceiving statements: the more we consider a statement not as a sentence (or proposition) but as an act of speech (out of which the others are logical constructions) the more we are studying the whole thing as an act. But that does not change anything about the fact that, by conceiving a statement as an act, one is led to see things like propositions as logical constructions which can be obtained by abstraction from the concrete performances that amount to statements. So one consequence of A ­ ustin’s (anti?)-Copernican turn in the analysis of the logical accomplishments of speaking agents is that statements come first. When he considers the various candidates for the business of being true in his landmark paper on “Truth” (1950), Austin abides by the perspective we find in How to Do Things with Words: there too, he credits statements with a privileged status. He suggests that, among our truth-ascriptions, the following forms of expression should be considered paradigmatic: It is true (to say) that the cat is on the mat. That statement (of his, &c.) is true. The statement that the cat is on the mat is true. (PP, 118)

What Is a Statement?  21 The second and third expressions refer explicitly to a statement. The first expression doesn’t explicitly involve reference to a statement but Austin’s analysis shows that it must involve a statement if the sentence is to be meaningful. So statements are the genuine truth-bearers in Austin’s analysis. The objective of my paper is to explore the import of this shift in the locus of truth. To do so, I will examine various difficulties that Austin faces. Indeed, by making truth a property of statements – and not of the abstract ‘propositions’ to which the logical tradition appealed – truth at first becomes something more concrete, even mundane. But once one looks a bit further it quickly turns out that it is not so clear what exactly a ‘statement’ is. There is at least here some need for elucidation. ‘Propositions’ and even – in a certain respect – ‘sentences’ are described as logical constructions obtained by abstraction from real statements. It might seem redundant to call statements ‘real’, because, contrary to propositions and sentences, statements seem to be real by definition: they are, after all, real performances (although we will see that we must qualify this assertion). It may thus come as a surprise that, at the very beginning of How to Do Things with Words, in the first footnote of the book, Austin writes that “the statement itself is a ‘logical construction’” (HTDTWW, 1). This remark raises a particular difficulty: how are we supposed to contrast ‘propositions’ with ‘statements’ – a contrast that initially seems to be one between something abstract and something concrete, between a logical constructions and reality – if both turn out to be logical constructions? One may of course retort that both things need not belong to the same level of logical construction, but that isn’t of any help here. The contrast was meant to turn on a generic opposition between something abstract and something concrete, something ­logically constructed and something real. Of course, this should be taken as a sign that Austin’s problem is not primarily to make ‘more concrete’ what the philosophical tradition may have conceived in too abstract or too ideal terms. As if we were in need of a more humane truth and it were the philosopher’s business to satisfy this need. Austin’s point seems to be a quite different one, namely that truth, in some not entirely obvious sense, is always human, however inhuman it might sometimes look; or, to be more precise, it is always related to the contextually situated exercise of the capacities of a speaking subject. As he says in “Truth”: “the statements fit the facts always more or less loosely, in different ways on different occasions for different intents and purposes” (PP, 130). He thus wants to have ‘truth’ understood as one amongst a variety of different kinds of fit, namely as the kind of fit that pertains to statements. This remark, however, also makes clear the real nature of truth as a kind of ‘fit’, i.e. something that obtains only in a particular way on a particular ­occasion – or at least on particular occasions – for a particular intent and purpose. In this sense, truth is human or at least essentially related

22  Jocelyn Benoist to an epistemic agent’s take under particular conditions – a take which might indeed be expressed precisely in a ‘statement’. There is thus no need to ‘humanize’ a truth that would not be of itself human, hovering instead in the Platonic heavens. Rather, philosophy should help us acknowledge the humanity of our truths, by teaching us to focus on their real location, i.e. on the kinds of human accomplishment that we call ‘statements’. The fact remains that these kinds of accomplishment, whilst being human, are nonetheless ideal. And this is the problem. What is a ‘statement’ exactly? The previously quoted sentence – “the more we consider a ­statement not as a sentence (or proposition) but as an act of speech (out of which the others are logical constructions) the more we are studying the whole thing as an act” (HTDTWW, 20) – gives a hint about the nature of the problem; for if a statement is an act, how could it come to be called ‘true’? It would seem that an act, as such, cannot be ‘true’ in the required sense of the term. It can be ‘true’ in the sense of being genuine, authentic, demonstrating adequate spontaneity etc., but it seems difficult to make sense of it being true in the logical sense of the term. Some people might in fact believe that the whole point of ordinary language philosophy, as opposed to earlier varieties of analysis that focused on the possibility of a formal language, is to do away with such a ‘logical sense’. But this view is of course completely mistaken: ‘truth’ in the logical sense of the term is precisely what is at stake in Austin’s eponymous essay. Now the quotation, in recommending that one consider the statement as an act, implies something further: namely, that the statement may not be an act from every point of view. We should learn to consider a statement not as a sentence or a proposition, but as an act of speech. It is, however, not so clear whether Austin is recommending a ­theoretical revision, which should replace previous definitions of ‘statement’, or whether he is instead recommending an aspect-switch, such that one might also consider a statement as an act. As a matter of fact, he is perfectly aware – a point we shall return to – that, on certain occasions, we happen to call a sentence or even a proposition ‘a statement’. ‘Statement’ is a term with many uses, and Austin is not so much selecting one arbitrarily, as constructing a technical philosophical usage, which piggybacks on a dimension associated with at least one kind of ordinary use of the term as a common target of truth- or falsity-ascription. Now, beyond this issue of demarcating a particular usage of the term within the multifarious space of its ordinary uses, the suspicion remains that even a purified concept of ‘statement’ could be essentially double-sided. Even if we should consider the statement as an act in order to notice something particular about it, it might be that something else can be found in every statement, in the very same sense of the term, that is not an act at all.

What Is a Statement?  23 We in fact find an indication of this possibility later on in the first footnote of How to Do Things with Words: the statement itself is a ‘logical construction’ out of the makings of statements. (HTDTWW, 1) This sentence certainly sounds paradoxical: a ‘statement’ is a “‘logical construction’ out of the makings of statements”. One natural way to remove the suspicion of vicious circularity here would be to distinguish two different senses of the word ‘statement’. On the one hand, there are statements as actual performances; on the other hand, there is ‘the’ statement as a logical construction, some kind of ideality – different, to be sure, from ‘the’ sentence or ‘the’ proposition, but just as ideal nonetheless. It seems to me, however, that this solution would lose sight of an important aspect of the problem, i.e. that the aforementioned definitional circularity (between performance and ideality) is an intrinsic feature of Austin’s suggested candidate for the title of ‘truth-bearer’. If we could really disentangle two senses of the word, we would no longer be dealing with a ‘statement’, but with a proposition on the one hand and with an utterance on the other – exactly the kind of dualism that the notion of statement, as Austin wants to understand it, is supposed to overcome. I would therefore prefer to talk of the double-sidedness of the – new – notion of ‘statement’ than of two substantially different senses. My aim here is to elucidate just this double-sidedness. How can a ‘statement’ be logically constructed “out of the makings of statements”? Prima facie, the shift from propositions to statements as the primary candidates for playing the role of ‘truth-bearers’ could be described as a shift from what might be called the content-view to the act-view. Let us have a look at the criticism of the notion of a ‘proposition’ that Austin advances at the beginning of the paper “Truth”. After reminding us of the use of the term in scientific – and, more specifically, in mathematical – contexts, in which it refers to some kind of general claim – a proposition to be established1 – rather than to an observation about a particular state of affairs, Austin addresses its characteristic use in modern philosophy of logic. “In philosophy,” he says, “‘proposition’ is sometimes used in a special way for ‘the meaning or sense of a sentence or family of sentences’” (PP, 119). A problem arises precisely when one wants to make the proposition in this sense the ‘truth-bearer’2 . This, of course, is just the intended function of this particular concept of a proposition. One might think of Bolzano’s ‘Satz an sich’ or Frege’s ‘Gedanke’, which these authors treat both as the meaning of a (logically) complete sentence and what is really true or false: the ‘truth-bearer’. The guiding thread of Austin’s analysis is that no entity can play both roles

24  Jocelyn Benoist at the same time: meaning is never enough in order to get truth. Or, to be more exact: each role concerns a quite different task – it is not as if there were ‘something to add’ to meaning in order to get truth: rather, meaning plays a role in grasping a truth only when one uses it in order to state a truth (which is, of course, one use among many others). That meaning cannot play the role of the truth-bearer is a crucial point. It means that we should disentangle two functions that the traditional ‘Platonic’ framework of analysis has incorrectly conflated. 3 This invalidates at least a certain notion of ‘content’, characterized by that dual functionality. Of course, this is not to say that every notion of content should be jettisoned, if by ‘content’ one just means the very idea of something that is true or false – that is, in one sense, the very idea of a ‘truth-bearer’. One can even take a step further and retain the following understanding of content: that there really is a certain ‘something’ that is true or false whenever there is truth or falsity, and that this ‘something’ is, in some sense, ‘what is said’. Once we have allowed for this notion of content, the relevant dispute is then over the correct interpretation of the phrase ‘what is said’ in this context. Austin’s whole reconsideration of the doctrine of the truth-bearer turns on this issue. One notion of ‘content’ is, however, surely rejected in his critical ­discussion: namely, that of a content that would, so to speak, precede the saying of what is said and explain its being true (or false) from ­outside – from above, as it were. On this notion of content, the content itself would first be true (or false) and then, second, what is said would ­therefore be true (or false), as though the truth (or falsity) of what is said were, so to speak, settled ahead of its being said. Now, this is exactly what the notion of a ‘proposition’ was supposed to do according to traditional philosophical analysis: propositions are true and false; therefore, when we express them, what we say is true or false. On such a view (the ‘content-view’), the truth-value of our verbal performances – or at least of those for which this notion is relevant – is an inherited property, which belongs primarily to nonlinguistic, ideal non-temporal entities: the propositions. A proposition can include a temporal determination in its content, for instance [Christopher Columbus discovered America in 1492]. This content is nonetheless, as content, timeless: it is what it is independently of its being expressed at any given time, and, thus, it is timelessly true. On this view, a real linguistic performance can, under certain conditions, just be the external and contingent manifestation of what is true (or false) and, by nature, ideal. Austin’s anti-Copernican revolution consists first and foremost in highlighting the irreducibility of the act-view when it comes to truth and falsity. An abstract and idealized idea of ‘content’ does not do the job of ‘truth-bearer’ correctly, at least if we want truths that are about something – which to Austin seems to be intrinsic to the idea of ‘truth’ as

What Is a Statement?  25 he uses it. Indeed, one can find nothing like aboutness in such ‘contents’. Meanings are not enough to hook words onto the world. Such a hook can be found only in an actual linguistic performance: in a speech act as such. Now there is a way to misunderstand Austin’s breakthrough here, that is, to think of the act as simply supplying the meaning with the historical anchoring that it lacks per se. Truth would still be the truth of a meaning – a kind of expression that, Austin emphasizes, is intrinsically incorrect4 – but of a meaning deictically referred to a state of affairs. A superficial reading of Austin’s characterization of the workings of a statement could (mis-)lead us to this result. He recognizes the existence of an essential duality in these workings. On the one hand, one has ­“descriptive conventions correlating the words (= sentences) with the types of situation, thing, event, &c., to be found in the world”, and on the other “demonstrative conventions correlating the words (= statements) with the historic situations, &c., to be found in the world” (PP, 121–122). Thus, it seems that the specific contribution of the statement as such is to provide the historic state of affairs about which what is said can be true or not. A statement is said to be true when the historic state of affairs to which it is correlated by the demonstrative conventions (the one to which it ‘refers’) is of a type with which the sentence used in making it is correlated by the descriptive conventions. (PP, 122) This might be read as if, to some extent, we already knew, by ­descriptive conventions alone, what should be true or false, but not that about which it is true or false. Such a reading would completely miss the point. In Austin’s view, the very notion of truth – or falsity – only makes sense at the junction between descriptive and demonstrative conventions, because there is no truth, in the required sense of the term, but when something is said about something. As a matter of fact, one can make a further step that might seem to go beyond the letter of Austin’s text, but that there are good reasons to believe is true to the spirit of his analysis. To be sure, he emphasizes in a footnote that although “both sets of conventions may be included together under ‘semantics’ […] they differ greatly” (PP, 122). However important this difference might be – I shall return to this issue presently – the fact remains that both sets of conventions should probably be seen, in some sense, as interdependent: they are connected in/by an actual performance, insofar as to describe is necessarily to describe something – and therefore needs a referent – and, conversely, the aboutness established by demonstrative conventions has no sense beyond assigning a possible description to its target. It is therefore possible to see

26  Jocelyn Benoist the two sets of conventions as setting up two complementary logical dimensions of one kind of act – acts that consist in describing states of affairs in speech: statements in the technical sense Austin intends. Of course, this is not to diminish the difference Austin marks when outlining the two sets of conventions: demonstrative conventions, he says, correlate words as statements with historic situations, whereas descriptive conventions correlate words as sentences with types of situation. Along these lines, one might believe that only demonstrative conventions are relevant to statements as such. This would certainly be mistaken, however, since there can be no statements independently of the exercise of a describing capacity, meaning that descriptive conventions are integral to the possibility of acts of stating. Conversely, if descriptive conventions pair sentences with types of situations, they govern sentences as they are used in making statements. Conventions as such concern uses: sentences are not by themselves correlated with any situations. They become correlated with types of situation whenever they are used in a statement. To see the bearing of this nontrivial point, it helps to pay attention to a crucial aspect of the ‘fit’ between statements and the world, which Austin emphasizes in the central part of the same paper: “the statements fit the facts always more or less loosely, in different ways on different occasions for different intents and purposes” (PP, 130). In this context, he is concentrating on the fact that, in a wide range of cases, ‘truth’ is not necessarily the most adequate characterization of the accomplishment of a successful statement: We say, for example, that a certain statement is exaggerated or vague or bald, a description somewhat rough or misleading or not very good, an account rather general or too concise. In cases like these it is pointless to insist on deciding in simple terms whether the statement is ‘true or false.’ (PP, 129–130) It is however noteworthy that the aforementioned remark applies to true statements as well: true statements, if they want to fit the facts ‘exactly’ (as opposed to ‘loosely’), do fit the facts by different standards of ‘exactness’, in different ways on different occasions for different intents and purposes. What is true in everyday life is not true in physics, and conversely. If one takes into account this problem of selecting the right ‘descriptive convention’, and, even more importantly, of the right use of a descriptive convention in some given circumstance, it becomes clear that descriptive conventions are utterly relevant to statements. In this regard, contextualism, far from being a further doctrinal element, is an essential aspect of Austin’s relocation of the truth-bearing function from the sentence or proposition to the statement. In order to assess the truth of something that is said (i.e. in order for the very idea of such

What Is a Statement?  27 assessment to make sense), one needs to know who said it, on which occasion and how. Thus, the shift of analysis to statements, far from implying some re-anchoring of an otherwise homeless truth, in fact involves the complete reframing of ‘truth’ as something that might not be obtained ­independently of actual attempts at describing something. The fact that they are real acts is essential in this regard: everything depends on the specificity of the actual performances. Of course, the interpretation of deixis plays a pivotal role in this respect. By rejecting any purely indexicalist account of this phenomenon – as if demonstration could ever be completely encoded in language – the Austinian analysis refers the truth of what is said to “a non-verbal origin […], which is the point of utterance of the statement” (PP, 122). This does not necessarily mean that what is said is supposed to be true at the very time and place of utterance, but that the statement as such always locates what it is about relative to this time and place, which serves for it as some kind of absolute origin. Thus, in this framework, a statement is treated as something that happens, a historic event. ‘Propositions’ as such were supposed to be timeless – even in the case of timeless dated contents. ‘Statements’, on the contrary, are essentially datable and their being-dated cannot be made a pure aspect of their ‘content’ – cannot be integrally semanticized. The fact remains that the statement has to be made and that this making is as such a historical event. Independently of this event, there is no connection with the context, narrow or wide, and there is no truth. Thus, with his account of deixis, Austin makes statements radically external: what they achieve cannot be interpreted from the inside on the basis of an alleged ‘content’. What goes for the accomplishment of speech acts in general, goes for truth too: we should have in view “the total speech act [in this case: the statement] in the total speech situation” (HTDTWW, 148). ‘Truth’ as such, like any other normative accomplishment of a speech act, can make sense only on that level. This is what I would like to call ‘the act-view’. On this view, truth is both historical – related to concrete performances – and pragmatic – ­related to uses in the circumstances of these performances. A problem remains, however. However much of a role the act as such may play in the fact that what is true is true, it cannot be itself said to be ‘true’, at least not in the same sense. Again, the act can be authentic, true to its performer, or true in the sense of what one might expect someone to do – that is, to say – in these circumstances. This ‘pragmatist’ account would still miss something – the heart of the matter: “the trite but central point that truth is a matter of the relation between words and world” (PP, 130), and that ‘words’, as such, are not an act, but rather the material used by an act, and, at the same time, in some sense, the result of an act. We are back to the aforementioned double-sidedness

28  Jocelyn Benoist of statements, and perhaps even of speech acts in general, or at least of those kinds of speech-acts that have a product. In other words, what is true or false is what is said, in the sense of: what is said by a particular speaker, at a particular time, in particular circumstances and in a particular way. However, ‘what is said’ is by no means to be confused with this saying itself. In this regard, there is a tension in the notion of a statement: between the performance and what one may call the accomplishment. If a statement is essentially ‘made’, one cannot simply and seamlessly assimilate what is made and its making. These are at least two different aspects of a statement. The heart of the matter lies in the difficult question of the identity of ‘a’ statement. According to the act-view, it might seem there should be as many statements as performances, since the set-up of a definite statement – in particular of its way of referring to the world – seems to be performance-dependent. Thus, it would seemingly make no sense to ask whether two or more linguistic performances, on different occasions, are ‘the same statement’ or not. They surely cannot be: on each occasion, even when using the same words in the lexical sense of the term, we produce a different statement merely as a consequence of their being uttered in different situations, However, if ‘statement’ here is a name for ‘what is said’ in the sense of that kind of ‘what is said’ that can be true or false, this assumption is certainly incorrect: we often consider the same thing to have been said in two different utterances, such that they make ‘the same statement’. And Austin does not want to suggest anything else. It is indeed a very important fact that in some cases different linguistic performances may be described as being ‘the same statement’ – as, in other cases, different performances can be ‘the same order’, or ‘the same promise’. Speech acts in general are not reducible to utterances and this is true of statements as well, as ‘what can be true or false’. That question of ‘sameness’ is nonetheless, as ever, a tricky one. As Austin notes, “the same statement” can be made “on two occasions or by two persons”, but “for this the utterance must be made with reference to the same situation or event”: The ‘same’ does not always mean the same. In fact it has no meaning in the way that an ‘ordinary’ word like ‘red’ or ‘horse’ has a meaning: it is a (the typical) device for establishing and distinguishing the meanings of ordinary words. Like ‘real’, it is part of our apparatus in words for fixing and adjusting the semantics of words. (PP, 120) In Austin’s technical vocabulary, ‘same’ is an adjuster-word. 5 ­Adjuster-words are “words […] by the use of which other words are ­adjusted to meet the innumerable and unforeseeable demands of

What Is a Statement?  29 the world upon language” (SS, 73). This means that, in particular, ‘the same’ has a special role to play in the business of ‘statements’, i.e. in their fit to a situation. “Situations are practically bound to crop up sometimes with which our vocabulary is not already fitted to cope in any tidy, straightforward style” (SS, 74). Now, adjuster-words like ‘the same’ help our statements (amongst other things) precisely to establish a fit in these situations: do we not say that two things are ‘the same’ only when it is not absolutely obvious that they are – they may be treated as ­different – and we have to decide that they are ‘the same’ for the purpose of our statement? Thus, what if we press the question of ‘sameness’ in regard to ­statements, and so ask under which conditions it makes sense to talk of ‘the same statement’ in two different sets of circumstances? One might first note that everything depends, of course, on what is meant by ‘statement’. In a striking passage of How to Do Things with Words, Austin highlights both the ambiguity of the word ‘statement’ and, correspondingly, the variable identity of so-called ‘statements’. Both issues are of course tightly connected, since to define identity conditions for statements just is to define what ‘statements’ are in each case. The context of this discussion is interesting because it is about locutionary acts. These represent a level of Austinian analysis in How to Do Things with Words that is generally ignored and that, in the rare cases when it is not, is liable to fuel misunderstandings. In his analysis of locutionary acts – or better: of the locutionary ­dimension of speech acts – Austin makes a distinction between the phonetic act, the phatic act, and the rhetic act. The phonetic act is merely the act of uttering certain noises. The phatic act is the uttering of certain vocables or words, i.e. noises of certain types, belonging to and as belonging to, a certain vocabulary, conforming to and as conforming to a certain grammar. The rhetic act is the performance of an act of using those vocables with a certain ­more-or-less definite sense and reference. (HTDTWW, 95) The phatic act, as such, is the act of uttering a “pheme” (HTDTWW, 92), whereas the rhetic act is the act of uttering a “rheme” (HTDTWW, 93). In his subsequent analysis, he insists that, when different phemes are used with the same sense and reference, we might speak of rhetically equivalent acts (‘the same statement’ in one sense) but not of the same rheme or rhetic acts (which are the same statement in another sense which involves using the same words). (HTDTWW, 98)

30  Jocelyn Benoist This means that we should already distinguish two different meanings of ‘statement’: one according to which a statement is merely identified by its sense and its reference; and another that requires that the very same words be used for there to be the same ‘statement’. This remark is highly interesting because it opens the door to the idea that, on some understanding of what a statement is, there might be one single statement by using different words. If, as is surely the case, statements can be made only by using words, this relativizes what one might call the literalism of statements. A statement is made by using words, but this does not preclude one’s using any words useful in the relevant circumstances. On at least some interpretations of what a statement is, a given statement can always be made with a range of different words. Now, a further ambiguity in the notion of a ‘statement’ resides in the fact that it may be used at the locutionary level – as in the previous ­analysis – or at the illocutionary level – as at the end of How to Do Things with Words or throughout the essay “Truth”. In this regard, one should be careful not to oversimplify Austin’s analysis: one must not ignore its multiple layers. A rhetic act is defined by its capacity to determine a sense and ­reference. It is, of course, noteworthy that, as such, it is already a speech act or, more precisely, one dimension of a speech act. Thus, Austin says: “the pheme is a unit of language: its typical fault is to be nonsense—­ meaningless. But the rheme is a unit of speech; its typical fault is to be vague or void or obscure, etc.” (HTDTWW, 98). However, the fact that locution is a fully-fledged dimension of the speech act does not mean that the distinction between this dimension and the others – in particular the illocutionary dimension – would collapse in any way. To be sure, “to perform a locutionary act is in general, we may say, also and eo ipso to perform an illocutionary act” (HTDTWW, 98), but this does mean that both are just identical: the two dimensions remain logically distinct. Thus the fact that, beyond the apparent diversity of words, we have the same sense and reference, is still not enough for us to identify two different utterances as ‘the same statement’ in the illocutionary sense of the term. As Austin says in the final part of How to Do Things with Words, “the truth or falsity of a statement depends not merely on the meanings of words but on what act you were performing in what circumstances” (HTDTWW, 145). One could object that meaning is perhaps not ‘sense and reference’ (at least not ‘reference’), and that, on the one hand, the question of what it refers to is pivotal in the determination of the statement as an illocutionary act (as is extensively shown in the paper “Truth”), and, on the other hand, that “sense and reference (naming and referring) themselves are […] ancillary acts performed in performing the rhetic act” (HTDTWW, 97). Thus, it might seem that the distinction, introduced in How To Do Things With Words, between what pertains to language (the pheme) and what pertains to speech

What Is a Statement?  31 (the rheme) could at least partially overlap with the distinction that is central to “Truth”, namely that between what is semantic in the narrow sense of the term (belongs to the ‘sentence’) and what belongs to the statement as ‘what can be true or not’. There is certainly no doubt that the capacity for terms to refer cannot be treated as completely independent of the fact that the statement in which they are used refers, as such, to a situation – in this sense, ‘reference’ as it is dealt with in “Truth”, is a property of statements, probably like of some other speech acts. The fact remains that sense and reference are not enough to fully account for a statement’s capacity to say something about the world. For anything to be assessable, it needs to be determined not only what it is about, but how it is about what it is about. Trying to view statements as they are understood in the essay “Truth”, as merely rhetic acts, just means ignoring the contextualist dimension of Austin’s perspective. What matters is not only sense and reference, but ways of referring and ways of saying something about what one is referring to. And this can be done only at the level of the statement as illocutionary act. In this regard, just as with other speech acts, what matters is only “the total speech act in the total speech situation” (HTDTWW, 148): ‘total’ in the sense of: in its ­full-fledged illocutionary dimension. Now, we can nonetheless retain something from the remarks about phemes and rhemes. In one sense of the word ‘statement’, one can talk of ‘the same statement’ when there is not the same rheme – because not the same words – but just “rhetically equivalent acts” (HTDTWW, 98). The remark is still illuminating if we return statements to the class of illocutionary acts. In “Truth”, Austin says: “the same sentence is used in making different statements […], it may also be used on two occasions or by two persons in making the same statement” (PP, 120). We should complete this analysis by adding that different sentences too may be used on different occasions or by different persons in making the same statement. And we can even add: in order to make the same statement on different occasions we do, at times, need different sentences. This is a consequence of Austin’s first point: the very same sentence used on different occasions may turn out to be the instrument of different statements; so, in order to maintain the identity of the statement throughout the variation in the circumstances of utterance, it may sometimes be necessary to adapt the tool to the circumstances and to make use of another sentence instead. This analysis, like the general remarks made at the outset of “Truth”, presupposes that a statement as such is not riveted to an utterance: it makes sense to recognize “the same statement” in different utterances. How is this possible, when the reference to the utterance situation seems to be part of what the statement is? Austin suggests a condition that seems to guarantee the identity of a statement throughout variations in the circumstances of utterance: “the utterance must be made with

32  Jocelyn Benoist reference to the same situation or event” (PP, 120). He says this about one statement made on diverse occasions by using the same words, but it is hard to see why the condition should not apply to one statement made on diverse occasions by using different words. Now, this means that there are two different ways to identify a ­statement. On the one hand, a datable statement is a performance. To say this does not amount to mistaking the statement for the mere utterance by which it is made. The statement is precisely what is made by this utterance. It still contains an essential reference to the time of the utterance by which it is made – the “non-verbal origin”, that is, “the point of utterance of the statement” (PP, 122). As far as this reference is to the point of its utterance, it is not transferable. Another utterance makes another statement. At the same time, however, if one adopts the point of view of the kind of commitment intrinsic to the act, it is clear that the very same commitment can, in general, be made through other performances, on different occasions. Thus, it will make sense to talk in this case of ‘the same statement’ made on different occasions insofar as whilst the circumstances of utterance change, we can, in different circumstances, establish a referring connection to the very same situation using a new speech act, and say ‘the same thing’ of that same situation as we did in the previous utterance. In so doing, of course, we necessarily make use of the present circumstances of utterance, just as we always do whenever we make a statement. However, what matters is only the situation to which we refer and the way in which we refer to it in saying what we say. Thus, we can define a kind of identity for statements as acts to be performed on different occasions, as something repeatable. It is essential that there should be such an identity because a statement possesses its logical properties only insofar as it has this kind of identity. For instance, when Austin highlights that a statement can be about anything in ‘the world’, including words, except “the actual statement itself which on any particular occasion is being made about the world” (PP, 121), this does not mean that a statement cannot ever be about its own utterance. In fact, the statement can be about the event of some one performance of the statement, just as it can be about any event in the world. But the statement cannot be about itself: it cannot be about itself as the ideal, repeatable statement it is. The impossibility here is not so much the impossibility of making such a statement – as if, in being made, it modifies itself, such that it becomes unable to be about what it was meant to be about (i.e. itself); it is, rather, the impossibility of making some performance where there isn’t anything to be performed: there is no statement to be made that would be about itself. For a statement is defined by its reference to a world to which it does not belong and remains so defined on whatever occasion of its utterance. This non-self-­ referentiality is a property of the statement as an ideal structure, not something related to its actuality.

What Is a Statement?  33 We may even take a further step along these lines and allow for p ­ ossible statements, beyond the ones that have actually been made. And this is what Austin does. After having given what is not so much a definition as an external characterization of a statement – “a statement is made and its making is an historic event, the utterance by a certain speaker or writer of certain words (a sentence) to an audience with reference to an historic situation, event or what not” (PP, 119–120) – he observes in a footnote that “‘historic’ does not, of course, mean that we cannot speak of future or possible statements” (PP, 120). This remark may seem to drive a wedge between the statement and that performance that seemed so essential to it as a speech act. However, it of course does nothing of the sort. A possible statement – like any statement in general – is constitutively referred to the possibility of its utterance. In order to make sense of its capacity to refer – and so to be a statement – we must allow for the possibility of its performance. The fact that it can be performed on different occasions – that is, that it is relatively independent from any particular performance – and the fact that we can even envisage future performances (as something that might be said, not as something that is said) does not detach it from performances in general: it remains something to be performed and assessable only as performed – as something that is said about something by someone on a certain occasion. Without this sense of performance, we are left merely with the shadow of a statement – which might look more or less like what we initially called a ‘proposition’. We thus reach the point I have been steering towards. We have seen the need to ascribe some ideality to statements: they are not mere ­occurrences but retain their identity between utterances. One might suspect that this all simply brings us back to the notion of a ‘proposition’. ­However, the ideality at issue is the ideality of something that ­obtains only in a performance. In a brilliant essay, the Polish philosopher ­Kasimierz Twardowski (1912) draws a distinction between ‘actions’ and ‘products’, which might prove helpful in the present context. A statement, after all, is something that is made, and if speech acts, according to Austin’s famous series of lectures, are things that are done with words, it is important that, in some cases at least, these doings are makings (i.e. yield a product). Yet, in the case of statements, we should perhaps not take this phrase too literally: ‘making’, in this context, should be interpreted in the sense of an achievement. Managing to state something is certainly not nothing. Once one has stated something, what one states can be true or false – that is the point. But that is another story. What matters is that what is stated – ‘the statement’ – only makes sense as something that has been stated. Unlike a proposition, the ideality of a statement does not explain our doings from above, as it were; rather, the ideality is the mere result of our doings, of what we do when we try to speak the truth.

34  Jocelyn Benoist

Notes 1 In this sense, ‘proposition’ is a good equivalent to the use of German ‘Satz’ understood as ‘Lehrsatz’. 2 Thus, the real source of metaphysical ‘shadows’ in the philosophy of ­language and the philosophy of mind – or more exactly between both – is the mingling of two different functions, as emphasized by Travis 2000. 3 See what Travis calls ‘Platonism’ (Travis 2000, pp. 12–16). 4 “We never say ‘The meaning (or sense) of this sentence (or of these words) is true’” (p. 119). 5 Lecture VII of Sense and Sensibilia might prima facie mislead us into thinking that adjuster-words are to be found only on the side of approximation terms (like ‘like’, “the great adjuster-word” (SS, 74)), but ‘exactness’ ­(‘sameness’) is not of a different nature than approximation (‘likeness’): it is nothing but an adjustment norm.

Bibliography Austin, J. L., Sense and Sensibilia (= SS), Reconstructed from the Manuscript Notes by G. J. Warnock, Oxford: Oxford University Press, 1962. Austin, J. L., How to Do Things with Words (= HTDTWW), ed. by J. O. Urmson and Marina Sbisà, Oxford: Oxford University Press, 1976. Austin, J. L., Philosophical Papers (= PP), ed. by J. O. Urmson and G. J. Warnock, Oxford: Oxford University Press, 1979. Travis, Charles, Unshadowed Thought. Representation in Thought and ­L anguage, Cambridge, MA: Harvard University Press, 2000. Twardowski, Kazimierz, “Actions and Products” (1912), trans. by Arthur ­Szylewicz in Kazimierz Twardowski, On Actions, Products and Other ­Topics in Philosophy, ed. by Johannes Brandl and Jan Woleński, Amsterdam-­ Atlanta: Rodopi, 1999, pp. 103–132.

3 Ordinary Language Philosophy needs Situation Semantics (Or, Why Grice needs Austin) Krista Lawlor ‘Ordinary Language Philosophy’ conjures different ideas in different imaginations, and major figures in the movement defend very different views. H. P. Grice sees Ordinary Language Philosophy as a method of conceptual analysis, and also as a base for defending common sense against skepticism.1 In both respects his view is shared by J. L. Austin. 2 Grice’s own ordinary language response to the skeptic fails, by his own admission. In this paper, I argue that two further supports are needed to construct an ordinary language response to skepticism, both of which are furnished by Austin. The first support comes from Austin’s semantic insight: situation semantics. The second support comes from a substantive claim about the content of the concept of knowledge – namely, Austin’s condition on knowledge, to the effect that the skeptic needs to offer a concrete reason for us to be concerned about skeptical hypotheses. When we build Austin’s condition into our situation semantics for knowledge claims, we can resolve the so-called skeptical paradox. 3 In what follows, I will set out Grice’s vision of Ordinary Language Philosophy as a method of conceptual analysis, and give a brief exposition of Grice’s own failed attempt to respond to the skeptic. Then we’ll see how Austin’s understanding of truth gives rise to situation semantics. A situation semantic account of knowledge claims that incorporates Austin’s condition provides the means to address the skeptical paradox. I’ll pose a dilemma for those who would follow Grice’s methodological principles but reject situation semantics.

Grice’s Ordinary Language Philosophy: Conceptual Analysis In Studies in the Way of Words, Grice makes explicit his commitments as an ordinary language philosopher. First, he affirms two methodological principles that are his starting points: 1 It is, in my view, an important part, though by no means the whole, of the philosopher’s task to analyze, describe, or characterize (in as general terms as possible) the ordinary use or uses of certain expressions or classes of expressions (Studies, 172).

36  Krista Lawlor Grice adds: … if I philosophize about the notion of cause, or about p ­ erception, or about knowledge and belief, I expect to find myself considering, among other things, in what sort of situations we should, in our ­ordinary talk, be willing to speak (or again be unwilling to speak) of something as causing something else to happen, or again of someone as seeing a tree; or again of someone as knowing rather than merely believing that something is the case (Studies, 172). Grice’s second methodological principle is this: 2 It is in my view the case that a philosophical thesis which involves the rejection as false, or absurd, or linguistically incorrect, of some class of statements which would ordinarily be made, and accepted as true, in specifiable types of situation is itself almost certain (perhaps quite certain) to be false… (Studies, 172). These two principles are the pith of Ordinary Language Philosophy (hereafter ‘OLP’). Grice next directs our attention to the aim of OLP. 3 Conceptual analysis of expression e (or class of such expressions) aims at a general characterization of the types of case in which it is appropriate to apply e (Studies, 174). We start by assuming we are already in a position to use e, that is, ­already in a position to deploy and withhold use of e. If we do not judge ourselves to be in a position to use e, then of course we are not in a ­position to analyze the concept e expresses. This assumption in place, the stages of conceptual analysis are as follows: first stage: we ‘botanize’ – we survey cases in which it is appropriate to deploy expression e, without prejudging where to draw conceptual boundaries; once the variety of uses of e is assembled, nearby expressions are contrasted; we also survey cases where use is withheld. For example, the philosopher aiming at an analysis of ‘knows’ begins with some botanizing, classifying various uses of the word without prejudice: one may speak of ‘knowing the answer’, ‘knowing a person’, ‘knowing too well’… and so on. Linguistic botanizing is a “foundational” part of conceptual analysis. Second stage: form a general characterization of the types of case in which it is appropriate to apply e. Third stage: test the general characterization. Imagine cases where the term so characterized would apply (fail to apply) and judge whether the use licensed (prohibited) by the general characterization conforms to one’s pre-theoretic judgment.4 Grice says the aim of conceptual analysis is a general characterization of types of case of appropriate usage. Conceptual analysis is explication, which can but need not take the form of providing necessary and sufficient conditions for the correct application of a term. Why is having a general characterization of usage or explication important? An objection to OLP is that its practitioners aim to turn philosophy into lexicography

OLP needs Situation Semantics  37 or empirical linguistics. Grice replies that the objection misunderstands what Ordinary Language philosophers are aiming to do – they are engaged in “conceptual analysis”, not lexicography. Grice claims that the philosopher engaged in conceptual a­ nalysis seeks to “give the meaning” of e in a different sense than giving a ­“dictionary definition” of e (Studies, 176). We might think of the “ ­ dictionary ­definition” as giving the linguistic meaning of the expression – the ­meaning e has in isolation from any particular sentence. ­Conceptual analysis delivers linguistic meaning, but it also delivers more than linguistic meaning. What conceptual analysis gives those who already know how to use e correctly is an explicit understanding of what they know (Studies, 176). It delivers what Austin calls an “explanatory definition” (Papers, 189). That is to say, analysis gives us further conceptual content, distinct from linguistic meaning – for instance, the linguistic meaning of ‘triangular’ might be ‘having three sides and three corners’, while the content of the concept triangle arguably contains more information (say about triangles having angles summing to 180, or being the object of study in Geometry, or…). Conceptual analysis, according to Grice, aims to give the meaning of an expression in the wider sense of giving its conceptual content.5 Grice’s picture of conceptual analysis is largely shared by Austin. When Austin speaks in explicit terms about methodology, it is clear he shares Grice’s two methodological principles.6 Austin might moderate the tone of the second principle, emphasizing that “ordinary language is not the last word….” (Papers, 185): There may be plenty that might happen and does happen which would need new and better language to describe it in. Very often philosophers are only engaged on this task, when they seem to be perversely using words in a way which makes no sense according to ‘ordinary usage’. There may be extraordinary facts, even about our everyday experience, which plain men and plain language overlook. (Papers, 69) Austin explicitly allows that language and concepts grow; sometimes our theoretical purposes require innovation. Austin would also insist on a wide range of situations – including unusual or extraordinary situations, at the second stage when testing one’s characterization of meaning: Ordinary language blinkers the already feeble imagination. It would be difficult, in this way, if I were to say, ‘Can I think of a case where a man would be neither at home nor not at home?’ This is inhibiting, because I think of the ordinary case, where I ask ‘Is he at home?’ and get the answer ‘No’: when certainly he is not at home. But supposing

38  Krista Lawlor I happen first to think of the situation when I call on him just ­after he has died: then I see at once it would be wrong to say either. So in our case, the only thing to do is to imagine or experience all kinds of odd situations, and then suddenly round on oneself and ask: there, now would I say that … A new idiom might in odd cases be demanded. (Papers, 68–69) ­ ustin This brings us to a still more important point of difference between A and Grice, namely their views about natural language semantics. They agree that usage reveals commonsense commitments. As Grice says, “… the appeal to the structure and content of languages or l­anguage as such [is] a key to unlock the storehouse of Common Sense…” ­(Studies, 382). But Austin strongly emphasizes that ordinary commitments are revealed by language as used on particular occasions to talk about ­particular situations: …it is not enough simply to examine the words themselves; just what is meant and what can be inferred (if anything) can be decided only by examining the full circumstances in which the words are used.7 A sentence is a package of words that has different significances in ­different utterances: …the question of truth and falsehood does not turn only on what a sentence is, nor yet on what it means, but on, speaking very broadly, the circumstances in which it is uttered.8 While Grice seems to think that we can limn the use of an expression in terms of cases to which it does or does not apply, Austin insists that expressions apply or fail to apply to given cases only in given circumstances. That the truth or falsity of what we say does not depend on sentence meaning alone, but is determined by sentence meaning in circumstances, is the heart of Austin’s “situation semantics.” Situation semantics is explored and defended by a number of ­philosophers, including both Francois Recanati and Charles Travis, as “radical contextualism” and “occasion sensitivity” respectively.9 Travis sees some deeper philosophical foundations of occasion sensitivity in Wittgenstein’s work, specifically in the idea that every representation needs interpretation; the very nature of representation is such that a representation cannot determine its own truth conditions.10 I believe Austin is motivated to his view more by observations of the role played by the notion truth in our everyday lives. Let us look a bit more into Austin’s thinking.

OLP needs Situation Semantics  39

Austin on Truth First, what is evaluated for truth, according to Austin? It seems to be fairly generally realized nowadays that, if you just take a bunch of sentences…impeccably formulated in some language or other, there can be no question of sorting them out into those that are true and those that are false; for (leaving out of account socalled ‘analytic’ sentences) the question of truth and falsehood does not turn only on what a sentence is, nor yet on what it means, but on, speaking very broadly, the circumstances in which it is uttered. Sentences as such are not either true or false.11 Sentences are not true or false – that includes sentences with p ­ articular sentence meanings. The same sentence, with the same meaning, may be used to make a statement that is true about one situation, and false about another. So far, this is not shocking: suppose “I am hungry” means the speaker is hungry; everyone will agree that we have to know who is talking and when, in order to know whether any particular utterance of this sentence is true or false. Austin’s point goes beyond this sort of straightforward dependence of utterance truth-value on sentence meaning and ­circumstance or context of utterance. Let’s unpack Austin’s view a bit more. Sentences are used to make statements that are true of one situation and false of another. (That is to say, restricting our attention to statements about empirical matters – we set aside statements of mathematics and other possibly ‘analytic statements’.) Let’s make this clearer by disambiguating Austin’s talk of ‘statements’ into talk of an utterance’s descriptive content and talk of the proposition the utterance expresses given the situation: The descriptive content of one’s utterance is (a) ­determined by the sentence’s meaning, and (b) is constant across different utterances of the sentence, but (c) is not evaluated as true or false. The proposition expressed by one’s utterance given the situation is (a) determined by sentence meaning combined with the situation of the utterance, and (b) is not constant across different utterances of the sentence, and (c) is ­evaluated as true or false.12 The basic claim of situation semantics is that the truth or falsity of an utterance depends on its descriptive content and the situation one talks about. These together fix the proposition one’s utterance expresses given the situation. How this works: A statement is said to be true when the historic state of affairs to which it is correlated by the demonstrative conventions (the one to which it ­‘refers’) is of a type with which the sentence used in making it is ­correlated by the descriptive conventions (Papers, 122).

40  Krista Lawlor An utterance is associated at once with a situation (by a language’s demonstrative conventions) and a type of situation (by its descriptive conventions), and is true just in case the former belongs to the latter. For example, my utterance of “The cat is on the mat” is true. (You weren’t here when I said it, so I will tell you the facts of the case: I was looking at Franklin the cat, and he was lying on the doormat.) “The cat is on the mat” is true, as said by me on this occasion, because the situation s I refer to includes Franklin the cat lying on the doormat in front of me, and because my words as I use them on this occasion are correlated by convention with a type of situation S, and s is of type S. Roughly, one’s statement is true because the situation one talks about is the type of situation that is described by the language one uses. More precisely, for my statement to be true the situation s I talk about must be the same type of situation as type S, where S is the type conventionally linked to the descriptive content I express. What is it for s to be the same type as S? Austin answers: ‘Is of a type with which’ means ‘is sufficiently like those standard states of affairs with which’. Thus for a statement to be true one state of affairs must be like certain others, which is a natural relation, but also sufficiently like to merit the same ‘description’, which is no longer a purely natural relation. (Papers, 122, note 2) My utterance is true if the situation s I talk about is sufficiently like such standard cases to merit the same description as is associated with the type S. In his “Truth” essay, Austin does not say a great deal more about the relation of being sufficiently like, except to say that it is a “non-­ natural” and “conventional” relation. To say “This is red” is not the same as to say “This is like those”, nor even to say “This is like those which were called red”. That things are similar, or even ‘exactly’ similar, I may literally see, but that they are the same I cannot literally see – in calling them the same color a convention is involved additional to the conventional choice of the name to be given to the color which they are said to be Se, 122.13 The relation of sameness of situation s and type S – the key relation that determines whether s is appropriately described by one’s utterance – is non-natural. Calling the relation “conventional” reinforces the fact. What else does Austin mean in calling it “conventional”? Austin is more forthcoming in How to Do Things with Words. He is clear that interpretation settles the question of whether a state of affairs is describable one way or another: …in the case of stating truly or falsely, just as in the case of advising well or badly, the intents and purposes of the utterance and its

OLP needs Situation Semantics  41 context are important; what is judged true in a school book may not be so judged in a work of historical research. Consider … ‘Lord Raglan won the battle of Alma’, remembering that Alma was a soldier’s battle if ever there was one and that Lord Raglan’s orders were never transmitted to some of his subordinates. Did Lord Raglan then win the battle of Alma or did he not? Of course in some contexts, perhaps in a school book, it is perfectly justifiable to say so – it is something of an exaggeration, maybe, and there would be no question of giving Raglan a medal for it…‘Lord Raglan won the battle of Alma’ is exaggerated and suitable to some contexts and not to others; it would be pointless to insist on its truth or falsehood.14 We should read Austin’s last remark about the pointlessness of insisting on the truth or falsehood of the sentence “Lord Raglan won the battle of Alma” not (just) as a reiteration of his claim that sentences are not true or false. He is making the point that judgment is required in order to determine whether one’s utterance is true or false. And judgment must take into account a wide variety of features of the context of utterance, i.e. any and all features relevant to a sound interpretation of the speaker’s intents and purposes. In the absence of such features, there is no point insisting on truth or falsehood. Austin goes on to draw a parallel between evaluating an utterance for truth or falsity with evaluating a bit of advice as good or not. Our evaluation of advice depends on the circumstances in which it is offered. Why doesn’t the linguistic meaning of a sentence determine its truth? Austin spends little time on this question. (He prefers other questions, such as, why are maps not true or false, even though they also are judged in terms of their accuracy? When does a map become a diagram?) He does remark that languages evolve to cover more and more situations with fewer and more economical terms.15 Expressions must be flexible. Having fewer and more flexible expressions makes language learnable, and also better suited to a wide range of unforeseeable situations. Let’s call this Austin’s Hypothesis: Austin’s Hypothesis: the function of a meaningful expression in a ­natural language is to provide a constraint on interpretation, which constraint can be easily learned and also easily reasoned from, to reach an interpretation of the utterer’s intended message in the context of utterance. One might wonder how we can get by with terms whose linguistic meanings do not determine specific truth conditions. If the linguistic meaning of our words only encodes general features or standard cases or what have you, of blue things, tall things, or cases of knowledge, it might look mysterious that we can effectively communicate with words, “blue”, “tall”, “knows”. Here it helps to remind ourselves that we are as a species especially equipped to cooperate; we are designed to coordinate both our actions and our interpretations of one another’s ­signals.

42  Krista Lawlor Michael Tomasello’s work is suggestive in this regard.16 Tomasello tells a story in which language is a late-comer in our communicative endeavors, a tool for providing additional constraints on interpretations we are ­already pretty well equipped to make thanks to our evolved dispositions toward cooperation, mutual recognition, shared intentions, shared desires to share emotional states, and so on. Linguistic capacities enter on the heels of our developed capacities for interpreting pointing gestures and other more abstract or iconic gestures. A linguistic token (word, phrase) stands in for an abstract gesture, itself already a fairly flexible (i.e. reusable) constraint on interpretation. Syntactic effects then layer over individual word use, to greatly expand expressive power. ­Tomasello’s story – admittedly speculative – shores up Austin’s hypothesis: the idea is that the function of linguistic elements is to provide further constraints on interpretation, the use of which exploits our ready-made capacities to arrive at suitable interpretations of each other. Flexible, reusable representations (pointing gestures and iconic gestures) have been part of our communicative equipment from the start, able to function well for us because of our essentially cooperative ­communicative nature. Flexible linguistic items prove useful given our robust interpretive ­resources. These interpretive resources are partly a matter of our shared conceptual resources. The linguistic meaning of our words provides a general constraint on interpretation, while further conceptual content helps determine how ordinary purposes interact with the general constraint provided by linguistic meaning, so as to establish the truth conditions of particular utterances in the situation talked about. For example, the linguistic meaning of “blue” encodes general features of our use of “blue”, notably, that it is said of visible phenomena that are blue. Further conceptual content of the concept blue fixes how our ordinary purposes interact with the general constraint provided by linguistic meaning, so as to establish the truth conditions of a particular utterance of “X is blue.” Conceptual content about lakes, water, and blueness for instance, helps us to judge that “Lac Leman is blue” is true when uttered as a description of the scene from the train, but not when uttered looking into a bucket of water drawn from Lac Leman.17 Shared conceptual content, then, is also part of the story of how we succeed in communicating, despite the fact that the linguistic meaning of our words only encodes general features or standard cases.

Ordinary Language Philosophy and the Skeptical Paradox For Grice and Austin, the justification of OLP lies in its fruitfulness, and mounting a defense of common sense against the skeptic is a significant part of being fruitful.18 Grice suggests OLP might form the basis of a response to the paradox-mongering skeptic, viz. someone who asserts what

OLP needs Situation Semantics  43 “a layman might be expected to find at first absurd, shocking, and repugnant” (Studies, 154). For instance, Grice says, a skeptic might claim: We do not know the truth of any statement about material things. How might this claim be supported? Grice notes the skeptic’s claim follows from a string of “alleged entailments” (Studies, 164). Filling in for Grice here, let’s have in mind as a concrete example of the much ­discussed “skeptical paradox” or “argument from ignorance”:19 (a) I know I have hands (b)  If I know I have hands, then I know what follows from this: that I am not a handless brain in a vat, systematically deceived into thinking the world is as it now appears. (c)  But I do not know that I am not a handless brain in a vat, systematically deceived into thinking the world is as it now appears. The skeptic argues that each of these is a claim of common sense. (a) Is a claim about ordinary material items, easily formed and verified by the senses. (b) Is an instance of a single-premise closure principle, roughly to the effect that if one knows P, and knows that Q follows from P, then one knows Q. (c) Is an exhibition of sensible epistemic modesty – one acknowledges that one does not know that one is not globally deceived, a brain in a vat. How could one know one is not a brain in a vat, given that the hypothesis that one is globally deceived is consistent with all one’s evidence? But these three claims cannot be true together. (a) and (b) together force the negation of (c). So, by entailment, says the skeptic, something is incoherent about the way we think of knowledge. 20 Early on, Grice believes he has a reply to the skeptic. 21 The sort of skepticism Grice takes most seriously seems to be of a form that portrays knowledge claims as conceptually false (Studies, 166–170).22 Later he comes to doubt his reply works. Sparing many details, here is his brief summary reply to the skeptic (Studies, 348): …the skeptic’s position can be represented as follows: that a certain ordinary sentence s does, at least in part, mean that p, second, that in some such cases, the proposition that p is incoherent, and third, that standard speaker intended their hearers incoherently to accept that p, where incoherently is to be read as specifying part of what the speaker intends. That position [the skeptic’s position] may not perhaps be strictly speaking incoherent, but it certainly seems wildly implausible. To flesh this out using our case: Grice has us imagine a speaker U saying, “I know I have hands” which means, at least in part that I know have hands. Now, Grice reasons, according to the skeptic the concept knows expressed by U is incoherent, and so is the proposition I know I have hands. But then according to the skeptic, U intends her hearer H ­incoherently to accept

44  Krista Lawlor this proposition. The skeptic’s own position is then, if not incoherent, at least wildly implausible – speakers do not intend their hearer’s to accept incoherent claims. At least not in the normal course of things. Grice comes to see that his reply fails: the skeptic can avoid the argument by an extensional interpretation of the appearance in this context, of the adverb ‘incoherently.’ … it would be possible for s to mean (in part) that p and for p to be incoherent and also for the standard speaker to intend the hearer to accept p which would be to accept something which is in fact incoherent though it would be no part of the speaker’s intention that in accepting p the hearer should be accepting something which is incoherent. 23

Austin’s OLP Resolution of the Skeptical Paradox Let’s proceed to a brief exposition of Austin’s situation semantics, and how it resolves the skeptical paradox. 24 As we have seen, one’s utterances are true or false in light of what ­Austin calls “the descriptive content” of one’s claim and the situation one is talking about. Your utterance (a) I know I have hands is true if there is a match between its descriptive content and the s­ ituation you are talking about. What is the ‘descriptive content’ of a knowledge claim? It is built up from the linguistic meaning of the sentence parts. We arrive at the linguistic meaning of ‘knows’ and the conceptual content of the concept of knowledge via conceptual analysis. Let us suppose for the sake of argument that when completed, our conceptual analysis produces a general characterization of the use of “knows that p” along the lines of “believes truly that p on the basis of conclusive evidence”, or “can rule out all the alternatives to the true claim that p….”. This is the linguistic meaning of “knows that p”. So, the sentence “S knows p” expresses the descriptive content:

. The descriptive content that you express with your claim (a) requires (i) that p be true (that you do have hands), and (ii) that you be in a position, in virtue of your evidence (looking at your hands), to rule out all the alternatives to p. Conceptual analysis also delivers further conceptual content (what Austin calls an “explanatory definition”), distinct from linguistic meaning. Conceptual content works with linguistic meaning to fix the truth or falsity of our utterances. The linguistic meaning of our words ­provides a ­general constraint on interpretation; it fixes general features of our use

OLP needs Situation Semantics  45 of “knows”. Beyond linguistic meaning, conceptual content fixes how ordinary purposes interact with the general constraint provided by linguistic meaning, so as to guide interpretation of the truth conditions of our utterances (or application conditions for our concepts) in the situation talked about. In other words, linguistic meaning helps determine the descriptive content of a claim, and conceptual content helps determine how we assess the descriptive content of one’s utterance for truth. When we check to see whether the descriptive content of an utterance is true of the situation talked about, we will observe further conditions or constraints provided by the relevant concepts. Let’s see how this goes in the case of knowledge claims. Without ­completing a full analysis, Austin provides evidence that it is part of the concept of knowledge that not every conceivable defeater of p is an alternative. Some defeaters are not worth considering. One needs concrete reason to consider them (Papers, 84, 88, 98). Roughly, Austin’s ­condition on knowledge: S knows p on the strength of reason or evidence that eliminates all the relevant or reasonable alternatives, and if challenged, one needs a concrete reason for additional alternatives to be relevant.25 When we check to see whether the descriptive content of a ­knowledge claim is true of the situation talked about, we will observe Austin’s condition. To illustrate, let’s return to the skeptical paradox. First, let’s stipulate about the case that the situation you are talking about, when you say, “I know I have hands”, is comprised of specific items, including both the hands you are looking at, and a set of alternatives to the proposition that I have hands. Suppose the only relevant alternatives to this claim are that your hands were amputated, or you have prosthetics. Suppose your ­visual evidence is the distinctively lifelike appearance of your hands as you flex and examine them. This evidence eliminates these alternatives in the situation. Your knowledge claim is true because its descriptive content is true of the situation you are talking about. The descriptive content you express with (a) is . This content is true of the situation, because (i) in the situation talked about that you have hands is true, and (ii) your evidence eliminates all the alternatives. Your evidence eliminates the set {amputated, prosthetics} and that just is eliminating all the ­alternatives in the situation you are talking about. The foregoing illustrates the Austinian semantics for knowledge claims. A few points are worth marking: “knows” means nothing less than “has conclusive reasons to believe the true claim that…” or some equally demanding thing. It means this on every occasion of its use. (Or  more carefully, every occasion in which it is used to claim propositional knowledge.) The truth of one’s utterance depends on what ­situation one talks about, including specifically, what alternatives are relevant in the situation.

46  Krista Lawlor On the situation semantic account of knowledge ascription, both of your utterances are true: (a) I know I have hands (c) But I do not know that I am not a handless brain in a vat, systematically deceived into thinking the world is as it now appears. Your lack of knowledge in (c) owes to the fact that different alternatives are relevant in the situation in which you are concerned about whether your visual perception tells you about material objects. For instance, the alternative that you do not have hands, but vat-images-of-hands. Your visual evidence is consistent with this skeptical alternative, so perforce does not rule this alternative out, so you don’t know that you aren’t a brain in a vat. So (c) is true. Fortunately you needn’t know you’re not a brain in a vat in order to know (a) that you have hands. What about the closure principle, (b) – it has some intuitive pull, doesn’t it? And (b) generates the demand that if one knows one has hands, one must know one isn’t a handless brain in a vat, which lands us back in a paradox. In reply: We need to be careful about closure principles. The general principle at work in (b) is a closure principle: [K(ϕ) & K(ϕ ⇒ ψ)] ⇒ K(ψ) As David Lewis observes, we must be careful in the application of such principles: closure principles, like many other principles, hold within a given context. It is a mistake to suppose that closure holds across different contexts or settings in which utterance truth is evaluated. 26 ­Austinian semantics pays attention to changes in the epistemic situation talked about – specifically, a change in the set of relevant alternatives – as such changes affect utterance truth. So following Lewis, we say that closure applies only if the situation remains stable. We apply a closure premise only when the set of relevant alternatives is unchanging. (Or, when no new relevant alternatives are introduced by the implicated proposition.) A more carefully stated closure principle explicitly incorporates Lewis’s observation, restricting it to cases where the two propositions are evaluated with respect to the same situation: [K(ϕ) & K(ϕ ⇒ ψ) & Restriction (ϕ, ψ)] ⇒ K(ψ) This closure principle does not generate paradox. 27 One cannot use it to force the extraordinary conclusion that one knows one is not a brain in a vat. To firm up our discussion, let’s recap some central points by running through a more mundane example.

OLP needs Situation Semantics  47

Duck Identification Looking across the water you see a Gadwall duck, floating peacefully. You reason that it’s not a Wood Duck (it’s too dull) or a Mallard (the bill is noticeably thinner). It is in fact a Gadwall, and your evidence eliminates all the relevant alternatives (which are in this case, let’s suppose, just the Wood Duck or the Mallard). “It’s a Gadwall, I know it – look at the distinctive markings” is a true knowledge ascription, given that it is a Gadwall, and your evidence is sufficient to eliminate all the alternatives. Do you also know it is not a very convincing decoy? Its not being a decoy follows from its being a Gadwall. (The latter proposition implies the former.) Yet it is natural to judge you need more evidence to know that it’s not a decoy, natural to judge that the set of relevant alternatives changes when this proposition is a candidate for knowledge (perhaps it’s not a wooden decoy, but plastic yard decor?). “So it’s not a decoy, then? But hasn’t Mary been testing her decoys this morning?” asks your friend. You do not know it isn’t a decoy. Austin’s insight that the truth of a knowledge claim depends on the situation combines with his relevant alternatives account of knowledge. Each target proposition brings its own alternatives to the table, and these alternatives may vary with the occasion. You can know it’s a Gadwall, when its being a decoy isn’t part of the situation you are considering. That doesn’t mean that on this same occasion you also know that it is not a very convincing decoy. To know such a thing, you’d have to have more or different evidence to eliminate some new alternatives (its being an animatronic ­robot, or a child’s clever toy, or whatever alternatives would be reasonable in the circumstances). In the ordinary run of things, you do not need to know that it is not a very convincing decoy when you know that it is a Gadwall. We’ve seen how Austin’s OLP with situation semantics lets us resolve the skeptical paradox. It tells in favor of a semantic theory that it provides a resolution of the paradox. That said, the literature on the skeptical paradox and the semantics of knowledge ascriptions is large, and a full defense of Austin’s position is a large project. In the remainder of this paper, I will confine myself to responding to a line of objection that might be traced to Grice. For our purposes, what is interesting is that the objection reveals that situation semantics better conforms to Grice’s methodological principles than other semantic accounts.

Objection: Grice’s Distinction between the Said and the Implicated Grice attacks ordinary language philosophers who would equate ­meaning with use, or simply “read the meaning of an expression off its use.” Notoriously, P.F. Strawson comes under Grice’s fire in this ­regard.

48  Krista Lawlor Strawson argues that the meanings of ordinary language sentence ­connectives, such as “and”, “or”, and “if … then…”, are not the same as the meanings of the connectives we find in first order logic. 28 We cannot translate “and” as “&”, Strawson argues, because our uses of “and” can suggest causal or temporal relations, while “&” does not. (For example, with “Jones put on his pajamas and went to bed” we mean he changed first and then got into bed.) Grice’s reply to Strawson turns on his distinction between what is said and what is implicated. What is said with “Jones put on his pajamas and went to bed” is just the truth-functional claim (Jones put on his pajamas & Jones went to bed), while what is implicated is that Jones first put on his pajamas and then went to bed. Grice thus explains Strawson’s case – we do mean Jones changed first and then went to bed, in the sense that we implicate it – while maintaining that logical connectives capture the meaning and the truth-conditions of the natural language sentences. 29 For some philosophers, Grice’s attack on Strawson is devastating; and it serves as a powerful example of the danger lying in store for ordinary language philosophers. Grice’s work thus gives rise to the oft repeated lore: OLP takes commonsense judgments about what one says or c­ ommunicates as directly relevant to the truth-conditions of one’s utterance. But, the lore continues, this is a mistake; Grice has made us more careful about usage and meaning, more careful about how what one says or communicates factors into what is said (which is fixes truth-conditions of one’s utterance) and what is implicated (which does not affect truth-conditions). For other philosophers, Grice’s distinction between what is said and what is implicated would block Strawson’s argument, if we had a better grip on the notion of what is said. Unfortunately, we do not have a rigorous theoretical explication of the notion of what is said. 30 Grice tries to provide one, and fails. And, some philosophers add, not only does Grice leave the distinction in an unsatisfactory state, its very coherence is questionable.31 My own view is that the lore is hyperbolic. Ordinary Language ­Philosophy is not devastated by Grice’s distinction between what is said and what is implicated. Grice himself does not think so. Rather than reject ordinary language methods, he says that OLP needs augmentation, with a theory which will enable one to distinguish between the case in which an utterance is inappropriate because it is false or fails to be true, or more generally fails to correspond with the world in some favored way, and the case in which it is inappropriate for reasons of a different kind (Studies, 4). However, it is true that Grice’s particular understanding of what is said puts pressure on Austin’s way of pursuing OLP, as it puts pressure on situation semantics. As we’ve seen, the core tenet of situation ­semantics is that an utterance’s truth conditions are not determined by expression meaning alone, and only get determined by the situation

OLP needs Situation Semantics  49 talked about. (“I know that’s a Gadwall” is not true or false, even after we fix the speaker’s identity, evidence, and the reference of “that.”) Expression meaning determines what Austin calls the “descriptive content” of an utterance, but the descriptive content does not determine conditions for the truth of the utterance. For Grice, what is said is, roughly, the overlap of what is conventionally meant by one’s sentence with what one means in uttering it.32 And for Grice what is said does determine conditions for the truth of the utterance.33 Charles Travis’s work defends situation semantics in part by ­questioning the coherence of Grice’s notion of what is said. I won’t recap Travis’s work – it spans several books. Very briefly, Travis’s position is that what is said with an utterance is not invariant, but is open to interpretation on occasion. He thus questions Grice’s attempt to articulate a notion of what is said that is simultaneously truth-conditional and invariant on different occasions of use.34 Travis thereby reclaims commonsense judgments about what we say or communicate for the purposes of OLP analysis. This is in briefest form Travis’s way of bringing “occasion sensitivity” to the rescue of OLP. My contribution here to the project of rehabilitating OLP is through a defense of situation semantics, focusing on the specific case of knowledge claims. I will argue that if we pursue Grice’s distinction and reject situation semantics for knowledge claims we will encounter a dilemma. We either maintain Grice’s position that what is said determines truth-­ conditions or we maintain his second methodological principle. We can’t have both.

A Dilemma for Grice The dilemma takes some set-up. Start by considering the simple invariantist position about knowledge claims. The simple invariantist says the truth-conditions for knowledge claims are invariant with respect to the situation talked about or the context of utterance. If Fred says he knows it’s a Gadwall, and John says that Fred knows it’s a Gadwall, the same proposition is expressed (very roughly, we can imagine where G is the bird in question, the proposition ).35 This proposition is true provided Fred meets the invariant conditions for knowing – his evidence meets “modest standards” required for ­knowing, say, and is false otherwise. Consider this case. Decoy Fred, a competent ornithologist, gets a good look at what is in fact a ­Gadwall across the lake. His evidence is very good – sufficient to ­eliminate all the alternative kinds of waterfowl it might be. He says, “I know that’s a Gadwall.”

50  Krista Lawlor It is natural to suppose that in such a case, Fred’s utterance is true – he is knowledgeable. Next, suppose John has all the evidence Fred has, and they both are looking at the same duck. John believes that his wife, Mary, a decoy artist, has just launched several cleverly painted pieces of work into the lake. In fact, she has delayed her launch, but John is unaware of this. “Do you know if that is a Gadwall?” asks a bystander, Otto. Unsure, given his wife’s morning launch, John says, “Don’t know, not sure.” Otto has heard that Fred is an ornithologist, and so suggests that they ask him. “Fred won’t know, either”, says John. The invariantist will hold that one of John or Fred is wrong. If Fred knows it’s a Gadwall on the basis of his evidence, and John has the same evidence, then John’s claim, “I don’t know, and Fred doesn’t either” is false. If on the other hand John’s claim is true, and neither he nor Fred know it’s a Gadwall, then Fred’s claim, “I know it’s a Gadwall” is false. One of our judgments about the truth of their respective claims must be false, according to the simple invariantist. The invariantist faces some uncomfortable questions. Suppose the ­invariantist insists that the same thing is said no matter whether the ­target situation is ordinary (with relevant alternatives, it’s a Mallard, it’s a Wood Duck) or extraordinary (it’s a Mallard, a Wood Duck, or a decoy), then at least one of John’s or Fred’s claims is false. Either the ordinary set is relevant to the truth of knowledge claims in the circumstances at hand or the extraordinary set is. If the ordinary set, then John’s claim that Fred doesn’t know is false. If the extraordinary set, then Fred’s claim is false, and John’s is true. Which is it? How do we make the choice? Either choice will look arbitrary. Moreover, in making a choice, we run afoul of Grice’s second ­methodological principle. Recall: (2) It is in my view the case that a philosophical thesis which involves the rejection as false, or absurd, or linguistically incorrect, of some class of statements which would ordinarily be made, and accepted as true, in specifiable types of situation is itself almost certain (perhaps quite certain) to be false… [172] Simple invariantism “rejects as false a class of statements ordinarily made and accepted as true” – namely it rejects those statements that we make when we are guided by variable standards of relevance. It r­ ejects as false one of John’s or Fred’s statements (interestingly, the invariantist will have to decide which). It also rejects as false our further reflective statements, in which we state that what John (Fred) said was true. ­Situation semantics has the advantage of counting intuitively true ­utterances as true.

OLP needs Situation Semantics  51 The invariantist might try to identify the invariant content of a ­ nowledge claim, what it says, with a Kaplanian character. Compare: I k say I’m hungry, and you say you’re hungry, and in one sense what is said is the same, namely, that the speaker is hungry.36 Here Grice will face the uncomfortable questions that indexical contextualists face. Consider this case. Report Fred the ornithologist rightly claims to know that the duck is a G ­ adwall. Cassie the casual observer also rightly claims to know the bird is a ­Gadwall. Not much is at stake for Cassie, Fred has a lot at stake in making a correct identification. More alternatives are relevant in Fred’s situation than in Cassie’s. The indexical contextualist says the following: Fred’s standards are more demanding and “know” has a concomitantly stronger meaning in his mouth than in Cassie’s, When Fred is talking, according to the indexical contextualist, the meaning of “knows” makes a semantic contribution to the proposition expressed something like “knows relative to high standards.” Likewise, according to indexical contextualist the meaning of “knows” in Cassie’s mouth is weaker, something like “knows relative to ordinary standards”. The propositions each express with “I know it’s a Gadwall” differ. Cassie expresses, roughly, that it is a Gadwall, and my evidence meets ordinary standards; Fred expresses, roughly, that it is a Gadwall, and my evidence meets high standards. So Fred’s report, “Cassie said she knows it’s a Gadwall,” involves a more demanding meaning than Cassie herself uses. Or more carefully: Fred’s report, “Cassie said she knows it’s a Gadwall”, expresses a different proposition than the one Cassie herself expresses. It seems Fred cannot make a true report of Cassie’s knowledge claim, then. This is a problem for indexical contextualism. And again we run afoul of Grice’s second ­methodological principle. Situation semantics accords with Grice’s second methodological principle: the meaning of “knows” and the propositional content (the “descriptive content”) of knowledge claims is stable across Fred’s and Cassie’s utterances, despite different situations of utterance. The descriptive content in Cassie’s utterance and Fred’s utterance is that the bird is a Gadwall and I can eliminate all the alternatives to its being a Gadwall. So Fred can make a true report of Cassie’s utterance, “Cassie says she knows it’s a Gadwall” without misrepresenting its content.37 Prima facie, invoking Grice’s distinction between what is said and what is implicated in order to reject situation semantics leads to a dilemma. Both invariantist and contextualist means of characterizing a constant truth-functional content for knowledge claims lead to failures to accord with ordinary judgments about the truth or falsity of utterances.

52  Krista Lawlor Invariantists and contextualists can maintain that meaning determines what is said, which in turn determines truth-conditions, and chalk up these ­failures as cases where ordinary users of the language are just wrong, but in doing so they must reject Grice’s own second ­methodological principle. I should note the bounded scope of what I have argued. I have argued that OLP augmented with situation semantics opens up the space for a compelling and natural response to the skeptical paradox. But responding to the skeptical paradox is at best an incomplete defense of ­common sense.38 A full defense requires us to vindicate Austin’s condition on knowledge against a competing skeptical conception of what the concept of knowledge requires. For present purposes, I save a full discussion of ­Austin’s condition and its defense for another occasion.39

Conceptual Analysis and Situation Semantics What is the connection between conceptual analysis as Grice and Austin pursue it and situation semantics? For Austin, there is a deep connection – for Austin situation semantics is fundamental to doing conceptual analysis. For Grice conceptual analysis and situation semantics have no ­connection – situation semantics is not in view for Grice, even as he advocates for an OLP approach to conceptual analysis. Grice would suggest that to do analysis we need to identify what our usage of a term ‘know’, or ‘blue’ tells us about the application conditions of the concept. And to do that, we have to distinguish that in our usage that reflects application conditions, and that which does not, but say, reflects other conditions (of appropriateness or what have you). We have to identify which claims we hold true, as opposed to merely appropriate but possibly false. As Grice says, we need to augment OLP with “a theory which will enable one to distinguish between the case in which an utterance is inappropriate because it is false or fails to be true, or more generally fails to correspond with the world in some favored way, and the case in which it is inappropriate for reasons of a different kind” (Studies, 4). Grice acknowledges that this in turn requires nothing short of “a search for a systematic philosophical theory of language” (ibid), which he sees as a long way off. A systematic theory of language includes both a semantic theory and an account of the interaction of pragmatic and semantic ­aspects of our language. But of course Austin agrees with Grice that we must be sensitive to the interaction of pragmatic and semantic features of our utterances. In fact, it is interesting to note that in Austin’s discussion of Moore-­ paradoxical utterances, he invokes something like Grice’s distinction ­between ­logical implication and pragmatic implicature. Austin notes that there are ­different senses of the word ‘implies’: … of course ‘the cat is on the mat’ does not imply ‘Austin believes the cat is on the mat’: nor even ‘the speaker believes the cat is on the

OLP needs Situation Semantics  53 mat’—for the speaker may be lying. The doctrine which is produced in this case is, that not p indeed, but asserting p implies ‘I (who assert p) believe p’. And here ‘implies’ must be given a special sense: for of course it is not that ‘I assert p’ implies (in the ordinary sense) ‘I believe p’, for I may be lying. It is the sort of sense in which by asking a question I ‘imply’ that I do not know the answer to it. By asserting p I give it to be understood that I believe p. (Papers, 64)40 The ‘ordinary sense’ of ‘implies’ is one in which to say that a proposition p implies a proposition r is to say that p is not compatible with not-r (Papers, 63). The sense of ‘implies’ in which I ‘give it to be understood’ that r follows from p does not require the incompatibility of p’s truth and r’s truth. The point I want to emphasize with this example is that Austin agrees with Grice that in deploying ordinary language methods, whatever our philosophical aims, we must be sensitive to the distinction “between the case in which an utterance is inappropriate because it is false or fails to be true, or more generally fails to correspond with the world in some favored way, and the case in which it is inappropriate for reasons of a different kind” (Studies, 4). Moreover, it is important to remember that while Austin does not have a complete systematic philosophical theory of language in the sense of a complete account of how pragmatic and semantic aspects of our language interact (in this he is not alone), as we have seen, he does have a powerful, novel, and important semantic program for natural language – situation semantics. Grice seems unaware of Austin’s contributions to semantics and to the interface of semantics and pragmatics. Unfortunately, Grice instead invites a large misunderstanding of Austin’s contribution. Grice writes … I do not regard it as certain that such a [systematic] theory can be worked out, and I think that some [ordinary language philosophers] were skeptical of just this outcome; I think also that sometimes they were unimpressed by the need to attach special importance to such notions as that of truth. (ibid, my emphasis) It becomes clear that in making this remark, Grice has Austin in mind: I am very much afraid that he [Austin] was trying to have his cake and eat it; that he was arguing in favor of using various inadmissibilities of application in respect to adverbs or adverbial phrases such as ‘voluntarily,’ or ‘deliberately,’ and ‘under constraint,’ and so forth, as a basis for determining the meaning of these expressions

54  Krista Lawlor (the boundaries of the concepts which they express), while at the same time endeavoring to put on one side the question whether such ­applications would be inadmissible because they would be false, ­because they would lack a truth-value, or for some other reason. It seems to me very doubtful, to say the least, whether this ­combination of procedures is itself admissible. [13] According to Grice, Austin is another ordinary language philosopher who would simply read meaning off use, because he was “unimpressed by the need to attach special importance to the notion of truth.” Nothing could be farther from the truth than to say Austin was not impressed by the need to attach special importance to the notion of truth. (One might note that his semantic theory is given in a paper entitled “Truth.”) Austin understands that conceptual analysis – the disciplined search for linguistic meaning and conceptual content – needs data, and our judgments about the truth or falsity of what we say is precisely what reveals linguistic meaning and the content of our concepts. This is why he cautions philosophers to bear in mind the situation people talk about, when we try to discover the truth-conditions of our utterances. On ­Austin’s view, utterances are true or false of the situation talked about, and only if philosophers understand this fact will they assign the correct truth-conditions for natural language utterances. This is why for Austin situation semantics is fundamental to doing ­conceptual analysis. Ordinary Language Philosophy needs situation semantics if it is to achieve its aims. One of its aims is a clearer understanding of our ­commitments, via a clearer understanding of linguistic meaning and conceptual content. Another aim is the defense of common sense. For reaching both ends, situation semantics provides the key. Once equipped with a powerful semantic account that accords with Grice’s ­methodological principles, ordinary language philosophers can pursue compelling ­defenses of common sense.

Notes 1 See Grice, Studies in the Way of Words, 340 ff. Note that Grice himself doesn’t think Ordinary Language Philosophy wrong-headed. He clearly thinks some ordinary language arguments are bad—see for instance S­ tudies, chs. 8–9, but in Studies, ch. 10 Grice identifies himself as an ordinary ­language philosopher, and he goes to some length to articulate and defend ordinary language methods and methodology. See Grice Studies, 182 ff. 2 Austin is also interested in reorienting philosophical problems, away from traditional skeptical problems about the very possibility of freedom, or knowledge, and toward questions about kinds of action, or cognitive accomplishment. See my “Ordinary Language and Common Sense: Wittgenstein and Austin” forthcoming Cambridge Companion to Common Sense Philosophy, ed. Peels and van Woudenberg.

OLP needs Situation Semantics  55 3 For more about Austin’s response to the skeptical paradox see ch. 2 of ­A ssurance: An Austinian View of Knowledge and Knowledge Claims. 4 Grice’s description of the stages of analysis is obviously rough. Somewhere in this process we must be on the lookout for expressions that are paronymous, or otherwise misleading about their extensions. Austin makes much of such difficulties in his essay “The Meaning of a Word” (1979, 55–75). Moreover there are questions we might have about the process: what to do if people disagree in their judgments of appropriate use, for instance? Grice handles this question and other objections to OLP in Studies, 173–180. 5 To suppose that lexical meaning and conceptual content are distinct is not to take any deep position on difficult matters of where semantic competence begins and ends. For one view of the latter sort, see Glanzberg, ­“Explanation and Paritiality in Semantic Theory.” 6 See Austin, Sense and Sensibilia. “A Plea for Excuses” in Austin, ­Philosophical Papers. 7 Sense and Sensibilia, 41. 8 Ibid. 111. 9 Recanati, Literal Meaning. Travis, Occasion-Sensitivity. For further ­references to Austin’s semantics, see Longworth, “John Langshaw Austin.” Some interpreters of Chomsky assign a similar view about the relation of meaning and truth. See Pietroksi, “Meaning before Truth.” 10 See Travis, Unshadowed Thought. Travis of course sees Austin as ­articulating “the core idea of occasion-sensitivity” Travis, Occasion-Sensitivity, 2. 11 Austin, Sense and Sensibilia, 110–111. 12 I draw here on Barwise and Etchemendy, The Liar. 13 Sense and Sensibilia, 122. 14 Austin, How to Do Things with Words: The William James Lectures ­Delivered at Harvard University in 1955, 143–144, my emphasis. 15 See “Truth” in Austin, Philosophical Papers, 124–126. 16 Tomasello, Origins of Human Communication. 17 See Travis. Occasion Sensitivity, 173. 18 See “Retrospective Epilogue” in Grice’s Studies. 19 Cohen, “How to Be a Fallibilist”; DeRose, “Solving the Skeptical Problem.” 20 Schiffer, “Contextualist Solutions to Scepticism.” We might notice that if the concept of knowledge is incoherent, then the skeptic’s claim, which deploys the concept as well, is itself incoherent. If so, then the skeptic’s claim that we do not know the truth of any statement about material things is better framed as the claim that nothing really answers to the concept of ‘knowledge.’ 21 See Grice, Studies, chs. 8–9, especially pp. 164–169. Grice’s own condensed version of his reply to the skeptic, summarized below, occurs on p. 348, in the Retrospective Epilogue of his Studies. It can be difficult to discern the summary reply in the earlier discussions, but see p. 169, where Grice identifies a difficulty he hopes “will in the end prove fatal”. 22 See “Moore and Philosophers’ Paradoxes.” The incoherence claim of the skeptic is very different than other skeptical challenges Grice considers in “Common Sense and Skepticism” (Studies, 147–153). In this latter paper, Grice distinguishes two skeptical challenges: (a) a challenge from Descartes’ evil genius to the effect that no empirical proposition is ever more probable than any other; (b) a challenge that no evidence is ever good enough to render a proposition knowable. Grice in his late recapitulation in his Retrospective Epilogue is more concerned about a third skeptical challenge posed by incoherence, i.e. the challenge from “Moore and Philosopher’s Paradoxes”. It is this challenge that strikes me as both more urgent in Grice’s mind, and also more closely tied to the contemporary skeptical paradox I’m concerned with here. Thanks to David Hills for discussion of this point.

56  Krista Lawlor 23 Grice more or less leaves the anti-skeptical argument in this unsatisfactory state. In the Retrospective Epilogue to his Studies, he briefly voices hope that a commonsense reply might still be mounted, suggesting that perhaps the way forward likes in something like Austin’s condition – ordinary beliefs are justified unless and until the skeptic produces concrete reason to doubt them. 24 I discuss the skeptical paradox and Austin’s response at greater length elsewhere, so this is a condensed recap. See Lawlor, Assurance: An Austinian View of Knowledge and Knowledge Claims. 25 What determines relevance? That is a long story. Lawlor (2013) gives an account of relevance as reasonableness. 26 Lewis, “Elusive Knowledge.” One might object: like indexical contextualism, Austin’s situation semantics must posit error and confusion in ordinary users of the language: If both (a) and (c) are true, why do ordinary users get taken in by the paradox at all? In answer, it is the skeptic’s use of closure that is confusing, not the truth conditions of our knowledge claims. 27 I am not wedded to the principle stated but something like it will do the trick. The general idea is a principle that places a restriction on the relation between propositions. Recently, there has been a lot of interest in restricting closure principles to cases where no new subject matter is introduced. See Lewis, “Relevant Implication”; ­G emes, “A New Theory of Content I: Basic Content”; Yablo, “Precis of Aboutness”; Fine, “Angellic Content.” 28 Strawson, Introduction to Logical Theory. 29 See Grice’s “Prologue” in Studies. 30 Neale, “Paul Grice and the Philosophy of Language.” 31 See “On what is strictly speaking true” ch. 1 in Travis, Occasion-Sensitivity. 32 Grice, Studies in the Way of Words, 120 ff. 33 This is rough, and can only be rough, since Grice’s account of what exactly is ‘what is said’, is itself rough and unfinished. His repeated pleas throughout Studies to see the naturalness of the notion aside, Grice’s own expressions of tentativeness about how to draw the distinction between what is said and what is meant, and his admission that he has given no rigorous way of drawing the distinction, suggest caution on our part. 34 Travis, Occasion Sensitivity. 35 Rysiew, “Contesting Contextualism.” 36 Keith DeRose invokes OLP in favor of indexical contextualist response to the paradox. DeRose, “The Ordinary Language Basis for Contextualism, and the New Invariantism.” 37 For more indirect reports and belief reports. See MacFarlane, “The Assessment Sensitivity of Knowledge Attributions”; Travis, Occasion-Sensitivity; Lawlor, Assurance. 38 As critics of semantic contextualism are quick to point out. See Kornblith, “The Contextualist Evasion of Epistemology.” 39 See my “Ordinary Language and Common Sense: Wittgenstein and Austin” for a reply to Stroud, The Significance of Philosophical Scepticism. 40 Austin’s discussion of this distinction is in the paper “The Meaning of a Word”, originally published in 1940.

Bibliography Austin, J. L. How to Do Things with Words: The William James Lectures ­Delivered at Harvard University in 1955. Oxford: Clarendon, 1962.

OLP needs Situation Semantics  57 ———. Philosophical Papers. Oxford: Clarendon, 1979. ———. Sense and Sensibilia. Edited by G. J. Warnock. 2nd ed. Oxford: Oxford University Press, 1962. Barwise, Jon, and John Etchemendy. The Liar: An Essay on Truth and Circularity. Oxford: Oxford University Press, 1987. Cohen, Stewart. “How to Be a Fallibilist.” In Philosophical Perspectives 2 (1988): 581–605. DeRose, Keith. “Solving the Skeptical Problem.” Philosophical Review 104 (1995): 1–52. ———. “The Ordinary Language Basis for Contextualism, and the New ­I nvariantism.” Philosophical Quarterly 55, no. 219 (2005): 172–198. Fine, Kit. “Angellic Content.” Journal of Philosophical Logic 45, no. 2 (2016): 199–226. https://doi.org/10.1007/s10992-015-9371-9. Gemes, Ken. “A New Theory of Content I: Basic Content.” Journal of ­Philosophical Logic 23, no. 6 (1994): 595–620. Glanzberg, Michael. “Explanation and Paritiality in Semantic T ­ heory.” In  Metasemantics: New Essays on the Foundations of Meaning, ­259–293. Oxford: Oxford University Press, 2014. Grice, Paul. Studies in the Way of Words. Cambridge, MA: Harvard University Press, 1989. Kornblith, Hilary. “The Contextualist Evasion of Epistemology.” In Skepticism, edited by Sosa and Villanueva, 24–32. Boston, MA: Blackwell, 2000. Lawlor, Krista. Assurance: An Austinian View of Knowledge and Knowledge Claims. Oxford University Press, 2013. Lewis, David. “Elusive Knowledge.” Australian Journal of Philosophy 74 (1996): 549–567. ———. “Relevant Implication.” Theoria 54, no. 3 (1988): 161–174. Longworth, Guy. “John Langshaw Austin.” Stanford Encyclopedia of ­Philosophy. Accessed February 1, 2016. . MacFarlane, John. “The Assessment Sensitivity of Knowledge Attributions.” In  Epistemology: An Anthology, edited by Kim Sosa, 779–800. Oxford: Blackwell, 2008. Neale, Stephen. “Paul Grice and the Philosophy of Language.” Linguistics and Philosophy 15, no. 5 (1992): 509–559. Pietroksi, Paul. “Meaning before Truth.” In Contexualism in Philosophy, ­edited by G. Preyer and G. Peter, 255–302. Oxford: Oxford University Press, 2005. Recanati, François. Literal Meaning. Cambridge: Cambridge University Press, 2004. Rysiew, P. “Contesting Contextualism.” Grazer Philosophische Studien 69, no. 1 (2005): 51–70. Schiffer, Stephen. “Contextualist Solutions to Scepticism.” Proceedings of the Aristotelian Society 96 (1996): 317–333. Strawson, P. F. Introduction to Logical Theory. London: Methuen and Co. Ltd., 1952. Stroud, Barry. The Significance of Philosophical Scepticism. Oxford: Clarendon, 1984.

58  Krista Lawlor Tomasello, Michael. Origins of Human Communication. Cambridge, MA: MIT Press, 2008. Travis, Charles. Occasion-Sensitivity: Selected Essays. Oxford: Oxford University Press, 2008. ———. Unshadowed Thought. Cambridge, MA: Harvard University Press, 2001. Yablo, Stephen. “Precis of Aboutness.” Philosophical Studies 174, no. 3 (2017): 771–777.

4 Beyond Unnatural Doubts Lessons from Wittgenstein Michael Williams

Introduction The most important skeptical problems and arguments belong to one or the other of two categories. Agrippan problems revolve around the apparent threat of an infinite regress reasons. Cartesian problems, while they sometimes incorporate Agrippan elements, differ from pure ­Agrippan skepticism in making essential use of skeptical hypotheses. These concern possible situations in which most of my beliefs about the world are undetectably false, or so we are supposed to think. In Unnatural Doubts,1 I focused on Descartes’s skeptical problem ­regarding knowledge of the external world, subjecting it to a diagnosis inspired in part by remarks in Wittgenstein’s On Certainty. 2 This diagnosis led me to conclude that whether someone knows or justifiably believes this or that is a deeply contextual matter. So although my inquiry began with an examination of a particular skeptical problem, it led me to a conception of knowledge and justification which provided a response not only to the full range of broadly Cartesian skeptical problems but to Agripppan skepticism too. I still think that this story is broadly correct, but it could use some further development. According to my diagnosis, skepticism is an artifact of the quest for theories of knowledge that are impossibly general. So one way of stating my conclusion would be that the antidote to epistemological skepticism is skepticism about epistemology. In Unnatural Doubts and for some time afterwards, I took this to mean that if skeptical problems could be dissolved, nothing much needed to be said, or even could be said, about everyday knowledge, at least not at the level of generality to which much traditional epistemology aspires. I now think that this leaves a gap in my treatment of Cartesian skepticism that should be filled. And it can be. My version of epistemic contextualism, I shall argue, is best understood as a kind of ‘knowledge first’ epistemology, though in a way that does not detract from my skepticism concerning knowledge as an object of theory. I came to this conclusion partly through an exploration of pragmatist approaches to understanding epistemological and semantic concepts and partly through a return to Wittgenstein and Austin, both of

60  Michael Williams whom are ‘knowledge first’ epistemologists in the sense I have in mind. 3 However, in this paper I shall focus on some things we can learn from Wittgenstein, whose ideas about knowledge and justification have been widely misunderstood. I shall begin by revisiting the anti-skeptical strategy I pursued in ­Unnatural Doubts, addressing some objections due to Duncan Pritchard. Then, drawing on my reading of Wittgenstein, I shall explain how knowledge-first contextualism strengthens my diagnosis, thereby offering a more complete response to skepticism.

Diagnosing Skepticism Cartesian skepticism challenges us explain how we know anything whatsoever about the external world, even that there is one. The challenge has proved hard to meet. As Barry Stroud argued in The ­S ignificance of Philosophical Scepticism, if the skeptic is allowed to pose the challenge as he means it to be understood, it is hard to resist the conclusion that it cannot be met. A symptom of the problem’s peculiar difficulty is the tendency of encounters with it to generate theories of knowledge that turn out to be difficult to distinguish from the skepticism they supposedly address.4 What we need is a diagnosis that brings to light whatever mistaken – albeit superficially attractive – ideas make the skeptic’s question seem compelling. But a fully satisfactory diagnosis had not been given and it was far from clear that one ever would be. In Unnatural Doubts I responded to Stroud’s pessimistic conclusion. My idée maîtresse was that while external-world skepticism must be an ­‘intuitive’ problem – that is, a problem implicit in our ordinary ways of thinking about knowledge and justification – closer examination shows that it is nothing of the sort. Far from being a consequence of what Stroud called “platitudes we would all accept,” Cartesian skepticism is hand-in-glove with an empiricist-foundationalist conception of empirical knowledge. All arguments for Cartesian skepticism insinuate, without exactly arguing for, the idea the ultimate basis for my beliefs about the world is provided by my ‘perceptual (or sensory) experience’: my ‘immediate awareness’ of how things appear to me to be. I argued that this piece of contentious theorizing radically distorts our ­ordinary ways of thinking about knowledge and justification. The diagnostic strategy I applied to external-world skepticism can be generalized to the entire list of modern skeptical ­problems. As I wrote: Other sceptical problems about particular “kinds” of knowledge seem to depend on making analogous cuts between what we are allowed to know, at least for the sake of argument, and what we hope (or like

Beyond Unnatural Doubts  61 to think) we know. The problem of other minds sets ­knowledge of the inner lives of other people against the “evidence” provided by the “external” behavior; the problem of induction contrasts knowledge of general, future or otherwise unobservable facts with observational knowledge gained to date; the problem of historical knowledge challenges us to relate beliefs about the past to evidence available in the present … Sceptical arguments begin by partitioning propositions into [privileged and problematic classes. Propositions in the (at least relatively) privileged class are taken to provide the (ultimate) evidence for those in the problematic class and sceptical arguments challenge us to explain how they manage to do this. This challenge is not easy to meet, which is why the problematic classes are problematic.5 In each case, the problem arises from the fatal interaction between a division of our beliefs into evidence-requiring and evidence-providing kinds and a suitably designed skeptical hypothesis. Other people behave as I would behave if I were in severe pain: but how do I know they aren’t zombies, who feel nothing? Up to now, the world has operated in more or less predictable ways, stones falling to earth and the sun rising daily: but how do I know that the course of nature won’t suddenly change? I have memories of my childhood and there is overwhelming evidence for the Earth’s longevity: but how do we rule out the possibility that the Earth came into existence five minutes ago, complete with our delusive memories and what we take to be evidence of its having existed time out of mind? In each case, the skeptic makes it looks as if our ultimate evidence for beliefs of a certain kind is compatible with their being false. If all such skeptical problems depend on partitioning our beliefs into evidence-requiring and evidence-providing kinds, attacking the rationale for such partitions threatens to call into question the modern skeptical problematic as a whole, and with it epistemology as traditionally conceived. But although philosophers like Austin, Wittgenstein, and Sellars had done just that,6 interest in Cartesian skepticism did not die. Philosophers intrigued by the problem found ways of presenting it that seemed to bypass any dependence on foundationalist presuppositions. My aim in Unnatural Doubts was to show that this dependence could be disguised but not eliminated. Stroud, for one, was not convinced. He argued that all diagnoses of the kind I favor are not just unconvincing: they get things upside down. The quest for foundations is not the source of skepticism but the result of “a certain intellectual goal, a certain kind of understanding of knowledge in general”. The goal of the traditional epistemological project is “to understand how any knowledge at all is possible – how anything we currently accept amounts to knowledge” or “to understand with complete generality how we come to know anything in certain specified domains”: for example, how we come to know facts about the

62  Michael Williams external world.7 As he wrote, “What some philosophers see as a poorly motivated demand for ‘foundations’ for knowledge looks to me to be the natural consequence of seeking a certain intellectual goal, a certain understanding of human knowledge in general”.8 Stroud did not deny that pursuing this goal will involve seeking a basis for our knowledge of the world that does not take knowledge of the world for granted: this is how we come to look for a basis for such knowledge in ‘experience’.9 But if for whatever reason we deny that worldly knowledge can be ­understood this way, we will still not have the kind of general understanding of knowledge that philosophers have striven for; and this too, he thought, would leave us dissatisfied.10 This objection does not touch my diagnosis. My first and most ­fundamental claim in Unnatural Doubts is that there is no such thing as knowledge of the external world.11 The category “knowledge of the external world” owes its theoretical integrity to the doctrine of the general epistemic priority of experiential knowledge over knowledge of the ‘external’ world. Take that doctrine away and nothing remains for us to succeed or fail in understanding. As an object of theoretical investigation, knowledge of the external world belongs with phlogiston and demonic possession. Perhaps once upon a time the concept of knowledge of the external world looked like one that would guide some fruitful inquiries, but that time had come and gone.12 The traditional epistemological project is founded on the idea that ­beneath the shifting standards of everyday epistemic assessment there lies a context-independent order of epistemic priority, according to which experiential knowledge is and must be the basis for knowledge of the world. This doctrine is an articulation of what I called ‘epistemological realism’. To reject epistemological realism is to embrace what I called a ‘contextual’ understanding of knowledge, according to which […] the epistemic status of a given proposition is liable to shift with situational, disciplinary and other contextually variable factors: … independently of such influences, a proposition has no epistemic status whatsoever.13 Now following Descartes’s lead, many philosophers generate the ­skeptical problem reached by considering a ‘best’ or ‘paradigm’ case of perceptual knowledge: a case such that if knowledge fails here, it is hard to see how we could know anything about the world around us. Such cases concern, or purport to concern, familiar objects in what we take to be favorable conditions of perception. But on the contextual view I recommend, being in a position to know depends on a would-be knower’s actual situation. Being in a position to know is a matter of being favorably placed in the world and equipped with the recognitional capacities or background knowledge to take advantage of the ­opportunity. ­Epistemologists ­puzzled

Beyond Unnatural Doubts  63 by skepticism want to examine a case of knowing in the way that a chemist tests a sample of sulfuric acid. This would be fine if knowledge of the world had a ‘nature’ in the way that chemical elements and compounds have natures. But to view knowledge as deeply circumstance-dependent is to reject the very idea of our having an ‘epistemic position’. To reject this idea is to cast doubt on the idea that a ‘case’ of knowing can represent knowledge of the world ‘as such.’ Take away the idea of our having a situationally invariant epistemic position, and ‘knowledge of the world’ is like ‘things that happen on Tuesdays’. In neither case should a lack of theoretical understanding disappoint us. Calling my conception of knowledge and justification ‘contextual’ was perhaps unfortunate, given that ‘contextualism’ has come to be identified with so-called ‘attributor contextualism’. While similar in some superficial respects, the two forms of contextualism depend on very different ­notions of context. Attributor contextualists hold that ‘know’ is a context-sensitive term in the sense that the truth value of a knowledge-attribution varies with the context in which the attribution is made. This variation stems from variation in the epistemic standards governing the attributor’s context, which may be higher or lower according to the range error-possibilities in play. Considering a skeptical hypothesis, such as the brain-in-­a-vat scenario, creates a context in which making a true knowledge claim requires me to be sensitive to or able to rule out this remote possibility. Arguably, this is a standard I cannot meet: a claim to know that I am not a brain in a vat must be false. But if I cannot know that I am not a brain in a vat, then (by closure) I cannot know anything that I know to entail that I am not a brain in a vat, so all claims to know everyday facts go false too. By contrast, in everyday contexts, where skeptical considerations don’t get a mention, standards are more relaxed. In these contexts, knowledge ascriptions involving simple everyday propositions are generally true. In the eyes of its proponents, an important virtue of attributor ­contextualism is that it explains how skeptical arguments can seem irreproachable. How can I know that I have hands, if I don’t know that I am not a handless brain-in-a-vat? Well I can’t in the following sense: in a skeptical context, closure ensures that even the most ordinary claims to knowledge go false. This has no implications for everyday knowledge-­attributions in everyday contexts: in such contexts, I know lots of ­everyday things, along with their known everyday consequences. However, if we notice that an everyday proposition entails the falsity of a skeptical hypothesis, we bring a skeptical possibility into the context of attribution, making it a high-standards. Failure to keep track of the context-shift induced by raising a skeptical question creates the illusion that skeptical reflections undermine everyday knowledge-claims.

64  Michael Williams From my standpoint, this supposed virtue of attributor contextualism points to its signal defect, which is to suppose that a change of epistemic context can be effected by merely considering skeptical error-­possibilities, either in themselves or as possibilities we supposedly know not to obtain. For attributor contextualists, epistemic context is conversational context, whereas for me context is multidimensional and includes the would-be knower’s actual situation. To be sure, dialectical factors – for example, objections that need to be dealt with – are an aspect of a person’s epistemic context. But nothing in my view encourages the idea that merely considering a supposed error-possibility can be an obstacle to knowledge. Quite the contrary: my view was and is that the supposedly ‘skeptical’ context is not created by a mere conversational move but by an implausible epistemological theory.14 If this is right, attributor contextualists are wrong to suppose that­ considering skeptical hypotheses raises the standards governing ­attributions of knowledge in everyday or scientific undertakings. ­Pondering Russell’s hypothesis that the Earth came into existence five minutes ago would not introduce higher standards for engaging in historical or cosmological research: it would raise the question of whether research in these areas is even a possible source of knowledge. Focused disciplinary research necessarily excludes the sort of questions that ­concern ­traditional epistemology. In presenting this argument, I drew on Wittgenstein’s thought that “it belongs to the logic of our scientific investigations that certain things are in deed not doubted”, since they are the ‘hinges’ on which our inquiries turn (341). As I put what I took to be Wittgenstein’s point, For a subject like history, there is more to method than abstract ­procedural rules. This is because the exclusion of certain questions (about the existence of the Earth, the complete and total unreliability of documentary evidence, etc.) amounts to the acceptance of substantial factual commitments. These commitments, which must be accepted, if what we understand by historical inquiry at all, have the status, relative to that form of inquiry, of methodological necessities.15 But while I stand by this point, as far as it goes, I did not fully appreciate what Wittgenstein was saying. I can bring out the problem by considering Duncan Pritchard’s objections to what he sees as my anti-skeptical appeal to Wittgenstein’s ‘hinge commitments’ construed as methodological necessities. According to Pritchard, Williams accepts the general Wittgensteinian approach that all rational evaluation takes place relative to arational hinge ­commitments… [He] rejects the ‘totality condition’ that is implicit in the traditional epistemological enterprise, such that it is possible to rationally

Beyond Unnatural Doubts  65 evaluate all our belief at once, and he does so because he believes that all rational evaluation takes place relative to ‘methodological necessities’ that are not themselves subject to rational evaluation. Thus he maintains that the very idea of a fully general rational evaluation is incoherent. Moreover, he argues as a consequence of his claim that radical skepticism is not the paradox it purports to be, but rather trades on dubious theoretical claims masquerading as common sense.16 But while in agreement with the general tendency of my anti-skeptical diagnosis, he still thinks that I go badly wrong in identifying ­Wittgensteinian hinge commitments with methodological necessities. Thus: Which inquiries one undertakes is a matter of choice … and hence on this view which methodological necessities one has at a certain point in time can be a matter of choice too. But can we really make sense of our hinge commitments being optional in this way? Isn’t the commitment in play meant to be visceral, an ‘animal’ commitment?17 A further way to bring out the supposed oddity of my appeal to methodological necessities is to ask … what sort of inquiry would lack any of the methodological ­ ecessities that Williams claims is distinctive of history … [O]nce n one reflects on the matter, it is hard to think of a specific inquiry that doesn’t, for example, presuppose the reality of the past.18 Pritchard’s objections reflect a fundamental misunderstanding of my strategy. With respect to knowledge of the external world, I reject the totality condition because it cannot be dissociated from a conception of the structure of epistemic justification that is wildly at variance with our epistemic practices, whether in everyday life or in scientific and scholarly inquiries. Of course, devotees of the traditional epistemological project are well aware of the discrepancy but they account for it, as did Descartes, by suggesting that ordinary justifications are ‘local’ and as such constrained by practical interests. If we prescind from all such interests, we can consider error-possibilities that we ordinarily ignore, thereby raising epistemic standards. This is the defensive move that my appeal to history is intended to resist. Historical research need have no practical implications; and though research in the natural sciences often has important practical applications, this is not always so; and when practical applications emerge, they need not have been intended or foreseen. The opposite of ‘practical’ is ‘theoretical’, not ‘purely epistemological’. What about the charge that I make hinge commitments optional? It depends what we count as hinges. Systematic practices of inquiry emerge

66  Michael Williams over time, shaped by human innovations and decisions. At any given stage of inquiry, there will be commitments which, by determining what is to be explained (or not) and how, define the subject of investigation. To take a banal example, in Aristotle’s physics the fundamental question concerning motion is why anything moves at all. For Newton, the question is why objects deviate from rest or from rectilinear motion at constant velocity. Newton’s first law (the principle of inertia) doesn’t just figure in explanations of particular celestial and terrestrial motions: it determines the phenomena to be explained. Similarly, the principle of universal gravitation determines how apparent deviations from the laws of motion, such as perturbations in a planetary orbits, are to be accounted for: find some suitably massive and appropriately situated body (Jupiter!). But while such commitments, which might be seen as playing a hinge-like role within the Newtonian research program, are subject to revision and in that sense ‘optional’, not all the commitments that determine the direction and character of our inquiries and epistemic assessment are optional. All inquiry, mundane or systematic, takes place against a rich background of everyday knowledge, much of which is certainly not a matter of choice. Pritchard goes wrong because he thinks that I take the everyday knowledge that is the indispensable background to all inquiry to be on a par with the theoretical presuppositions that play a constitutive role in scientific research and which can be revised or abandoned. Nevertheless, his objections point to a lacuna in my argument, which is that I did not say enough about everyday knowledge. Wittgenstein shows how to fill the gap, though not if he is read in the usual way.

‘Hinges’ as Knowledge There is an approach to interpreting On Certainty that has become more or less orthodox. I have called it “the framework reading”.19 It is now often called ‘hinge epistemology’, after some striking passages that are taken to encapsulate Wittgenstein’s most fundamental insight. On this approach, Wittgenstein’s response to skepticism turns on three fundamental ideas: i There are basic certainties: propositions or judgments that “lie apart from the route travelled by inquiry” (OC, 88). They are the ‘riverbed’, the channels through which our activities of inquiry and epistemic assessment flow (OC, 96), alternatively, the ‘hinges’ of inquiry, which must stay fixed if the door is to turn (OC, 341). ii Because they constitute our epistemic ‘frame of reference’, thereby giving our practices of inquiry and epistemic assessment their characteristic forms and directions, they are not themselves known or epistemically justified.

Beyond Unnatural Doubts  67 iii Recognizing the inescapable need for such sui generis basic ­certainties is the key to dissolving skeptical problems, Agrippan and Cartesian. Granted, this broad brush picture of knowledge and certainty has some basis in Wittgenstein’s text; but it demands further explication. It is easy to see why it might be taken to have anti-skeptical potential: if hinge commitments are an inevitable feature of our (or any) belief-­ system, the project of subjecting our beliefs to wholesale assessment is a non-starter. What is not so obvious is how insisting on the need for hinge commitments amounts to an answer to skepticism rather than a surrender. Pritchard, who defends a variant of the hinge strategy, i­dentifies what he calls a ‘core problem’ for hinge epistemology: that it is hard to see just what is so anti-skeptical about the claim that the structure of rational evaluation has, at its core, arational commitments. Isn’t that just what the radical skeptic claims? But if so, how is this view to be distinguished from radical skepticism, exactly?20 Competing versions of hinge epistemology reflect different responses to this core problem. But the problem is insoluble. Any response to skepticism that rests on the view that everyday certainties are not known is bound to be either concessive, revisionary of our ordinary understanding of epistemic concepts, or both. Let me illustrate the difficulty by briefly considering three prominent versions of this supposedly ­Wittgensteinian idea. Strawson thinks that Wittgenstein’s anti-skeptical stance is a kind of ‘naturalism’ closely related to that of Hume. According to Strawson, the skeptic demands reasons for fundamental beliefs that we hold without reasons. This is a mistake because, though lacking rational grounding, such beliefs are unavoidable. They are ‘natural’ in Hume’s sense: holding them belongs to human nature. 21 But as Strawson is well aware, on the theoretical level, Hume’s psychological anti-skepticism is explicitly concessive. Although extreme skepticism can have no influence on practical life, Nature being “too strong for principle”, skeptical arguments are theoretically invulnerable. 22 This is certainly not Wittgenstein’s view. The idea that there is a distinctively philosophical doubt, behind ­familiar everyday ways of doubting, is an illusion (OC, 19). Crispin Wright takes a similar albeit less psychological line. Though incapable of evidential warrant, our hinge commitments (Wright sometimes calls them ‘cornerstones’) are rationally warranted by virtue of being presuppositions of rational inquiry. Wright thus takes seriously Wittgenstein’s remark that exempting certain propositions from doubt “belongs to the logic of our scientific investigations” (OC, 342), which Strawson is inclined to downplay. But though less psychological than

68  Michael Williams Strawson’s, Wright’s reading is equally concessive. According to Wright, skepticism points to a deep truth. Skeptical arguments establish “the limits of justification.” Since attempts to transcend them would not lead to greater rigor but to “cognitive paralysis” – a thoroughly Humean claim – we must learn to live within them. As Wright candidly notes, his ‘Wittgensteinian’ response to skepticism is “an exercise in damage limitation”. 23 Danielle Moyal-Sharrock proposes to avoid such problems by a­ rguing that hinges, or basic certainties, are not propositional commitments at all but rather ways of acting. While this might seem to be a radical move, Moyal-Sharrock’s insistence on the ‘animal’ character of this ­commitment-in-action reveals her suggestion as yet another ­neo-Humean reading of Wittgenstein. As Pritchard notes, that hinge commitments normally go without saying does not exclude their being put into words. Moyal-Sharrock would agree; but she would point out that making hinge commitments propositionally explicit, as the skeptic insists on doing, undercuts their hinge-like function: put into words, they are no longer hinges. This may be true, but it is far from clear why it should bother the skeptic, who readily concedes that that we cannot take skepticism seriously in everyday practical affairs. Although highlighting the core problem for what he takes to be ­Wittgenstein’s account of the structure of rational justification is to Pritchard’s credit, his own suggestion fares no better. According to Pritchard, hinge commitments are not beliefs “in the knowledge-apt sense.”24 Pritchard fleshes out his “non-belief” ­reading by arguing that our particular hinge commitments are “codifications” of “the entirely general hinge commitment that one is not radically and fundamentally mistaken in one’s beliefs”.25 Pritchard call this general commitment the uber hinge proposition. This is just a verbal m ­ aneuver. Pritchard states his core problem in terms of arational commitments and the skeptic’s conclusion is that they are not ­“knowledge-apt”, whatever we call them. Nor would the skeptic question our commitment to the über hinge proposition. Of course, in everyday practical life we neither worry nor could worry that we might be radically and fundamentally mistaken in our beliefs; but if such pervasive error is possible, as a ­theoretical problem skepticism remains untouched. In light of the failure of these Humean readings of Wittgenstein, let us take a closer look at the passages from On Certainty that give ‘hinge epistemology’ its name. Here they are, as typically quoted: 341. …[T]he questions we raise and our doubts depend on the fact that some propositions are exempt from doubt, are as it were like hinges on which those turn. 342. That is to say, it belongs to the logic of our scientific ­(Wissenschaftlichen) investigations that certain propositions are in deed not doubted.

Beyond Unnatural Doubts  69 343. But it isn’t that the situation is like this: We just can’t investigate everything, and for that reason we are forced to rest content with assumption. If I want the door to turn, the hinges must stay put. Two features of these passages strike the eye. This first is that ­Wittgenstein’s topic is not general skepticism but ‘scientific’ investigations, in the German sense of ‘scientific’ that covers all systematic research: so history as much as chemistry or physics. The second is the ellipsis. The missing words are “That is to say”. Evidently, the passage that introduces the ‘hinge’ metaphor is a gloss on its predecessor: 340. We know, with the same certainty with which we believe any mathematical proposition, how the letters A and B are pronounced, what the colour of human blood is called, that other human beings have blood and call it blood. Far from being arational commitments, ‘hinge’ propositions are known to be true, with the same certainty as any mathematical proposition! All Wittgenstein claims in these much-discussed remarks is that systematic investigations depend on a background of everyday knowledge. The conclusion is inescapable: Wittgenstein’s appeal to ‘hinges’ presupposes answers to both Cartesian and Agrippan skepticism and thus cannot be his response to either. Hinge epistemologists will reply that the suggestion that hinges are known is an aberration. Wittgenstein’s predominant view is that it makes no sense to think of hinge commitments as things we know. Here is an emphatic statement of the point in question: 10. I know that a sick man is lying here? Nonsense … So don’t I know then …? Neither the question nor the assertion makes sense. Any more than the assertion ‘I am here’, which I might yet use at any moment, if a suitable occasion presented itself. Now even if this passage means what hinge epistemologists take it to mean, it is not obvious why it should be privileged over a passage like 340, in which Wittgenstein has no qualms about making a general comment on how the basic certainties that make up everyday knowledge provide the necessary background for systematic investigation. Perhaps epistemological discussion provides a “suitable occasion” for such a reminder. However, hinge epistemologists read too much into this remark. According to Wittgenstein it is the question and assertion that lack sense. In other words, he is not concerned with knowing but with claiming to know. In normal circumstances, my saying ‘I know…’ with respect to some banal and obvious fact is ‘nonsense’ in that I fail to perform a significant speech-act. Granted, Wittgenstein does not theorize

70  Michael Williams speech acts as Austin does, but this passage has a distinctly Austinian flavor. As Krista Lawlor points out, a common speech-act function of saying ‘I know …’ is to offer assurance. 26 But with bedrock certainties, assurance is not in prospect, for they are things that anyone knows, if I know them. In the situation Wittgenstein envisages, you and I come upon some person who, as we can both see, is obviously in bad way. In such a case, you have no need to ask for assurance and I am in no position to offer it. Indeed – a point that will prove to be important – I would have trouble understanding what you wanted from me, just as you would have difficulty making sense of an apparent attempt to offer assurance, should I make one unprompted. But suppose I had been by myself and, hearing what I thought might be a cry for help, had looked into the room to investigate. Recounting the incident, I say “When I entered, I  saw right away that there was a sick man lying there”. Here, ‘saw’ means ‘realized’, that is, ‘came to know’. So can I now state truly that I came to know something that I didn’t know at the time? There is no mystery here: that I know something does not mean that I can always felicitously claim to know it. The example of “I am here” should be understood along similar lines. The same goes for facts we expect anyone to know. This is why Moore’s insisting that he knows his commonsensical platitudes to be true strikes Wittgenstein as off-key: 84. Moore says that he knows that the earth existed long before his birth. And put like that, it seems to be a personal statement about him … Now it is uninteresting whether Moore knows this or that, but is interesting that, and how, it can be known. … I believe e.g. that I know as much about this matter (the existence of the earth) as Moore does, and that if he knows that it is as he says, then I know it too … Wittgenstein does not think that ‘common knowledge’ is an oxymoron. Recall that in 340, which introduces the hinge passages, he does not make a claim about what he knows but about what we know. Still, we might wonder how Wittgenstein can look askance at Moore’s claims. After all, Moore himself insists that everyone knows what he knows, or (when the facts involve some personal reference) certain corresponding facts. The answer is that Moore is addressing philosophers who doubt or even deny his propositions of common sense. That is, Moore enters his knowledge-claims in a philosophical context in which the legitimacy of skeptical doubts has already been conceded. Such claims misfire. I cannot assuage the skeptic’s doubts by assuring him that I (or even we) know this or that. Nevertheless, we may all know what the skeptic purports to doubt.

Beyond Unnatural Doubts  71 There is a further objection to consider. Moyal-Sharrock claims that “Wittgenstein does not take … [bedrock] certainty to be a knowing ­because he adheres to the standard view of knowledge as justified true ­ ittgenstein belief…”27 No direct textual evidence supports this claim: W nowhere offers an ‘analysis’ of knowledge. But even if the standard view could be attributed to him, we would have to explain how it should be understood. Moyal-Sharrock takes it that the ‘standard view’ of knowledge takes “justified” beliefs to be beliefs that are supported by evidence. Since hinge commitments provide the framework that enables justifying beliefs by evidence, they cannot themselves be so justified, from which it follows on her understanding of the standard view that they are not known. Now Wittgenstein himself may be inclined to use ‘justified’ this way. However, we should recall an important remark from P ­ hilosophical Investigations: “To use a word without j­ustification (Rechtfertigung) does not mean to use it wrongfully” (Unrecht; ­A nscombe, ‘without right’)28 If we understand “justified belief” as encompassing beliefs which, though neither based on evidence nor evident in themselves, are rightfully held, the standard view is no bar to holding that basic certainties are things we know. But there is another possibility. Perhaps ­Wittgenstein repudiates the standard view, as Moyal-Sharrock understands it, because he thinks that it gets things backwards. Knowledge cannot be equated with justified true belief because justification by evidence depends on ­knowledge: that much of our evidence consists in everyday (or otherwise unproblematic) knowledge, as the hinge passages suggest.

Knowledge as Objective Certainty On Certainty opens with a comment on G. E. Moore’s “Proof of an External World”. Moore devotes most of his discussion to clarifying what we mean (or should mean) by ‘external things’. External things, he decides, are things to be met with in space, as opposed to things like free-floating after images, which are merely presented in space. The former, unlike the latter, are open to public view, can be approached, viewed from different angles, and so on. Then, having settled on a definition, Moore holds up his hands and announces “Here is one hand and here is another”. 29 This is the proof’s only explicit premise. However, while stating it Moore makes certain the gestures which show that hands are external things, as he has defined them. Wittgenstein is intrigued but not impressed: 1. If you do know that here is one hand, we’ll grant you all the rest. When one says that such and such proposition can’t be proved, of course that does not mean that it can’t be derived from other propositions: any proposition can be derived from other propositions. But they may be no more certain than it is itself.

72  Michael Williams Initially, Wittgenstein is not so much interested in the proof itself as in the question of whether Moore’s is really a proof or just a deduction. As Moore notes, a proof must meet three conditions: its conclusion must differ from its premises, the premises must entail the conclusion, and the premises must be known to be true. Wittgenstein happily concedes that the proof satisfies the first two conditions: the question is whether it satisfies the third. Moore thinks it does. He insists that “Here is one hand and here is another” is something that he does know to be true, even though he cannot explain how. This peculiar feature of Moore’s procedure is what catches Wittgenstein’s attention: there seem to be things that we know without being able to explain how we know them. Wittgenstein’s opening remark sets the stage for everything that ­follows. The Agrippan trilemma is already hovering in the background. Our ability to give reasons for what we believe, thereby explaining how our beliefs amount to knowledge, is limited. Very soon, we come upon things that, like Moore, we want to insist that we just know: basic certainties. Such certainties cannot be proved, since nothing is more certain than they are. But if they are not self-evident either, how can they amount to knowledge? It is easy to see why hinge epistemologists take Wittgenstein to teach that basic certainties are not – cannot be – known. Knowing goes with having conclusive reasons: being able to prove. But having such reasons always presupposes a background of hinge commitments that are beyond proof. And there are passages where Wittgenstein makes an explicit connection between knowing with being able to prove. For example: 243. One says ‘I know’ when one is ready to give compelling grounds. ‘I know’ relates to a possibility of demonstrating the truth … But if what he believes is of such a kind that the grounds he can give are no surer that his assertion, then he cannot say he knows what he believes. However, note once more that the topic here is saying “I know”. Granted, Wittgenstein’s suggestion, in this regard, is very much in the spirit of Austin’s thought that to say “I know” is to imply (or perhaps ‘implicate’) that one is in possession of a conclusive reason. But Wittgenstein does not claim that one says “I know” only when one is in a position to give compelling grounds. Consider: 272. I know = I am familiar with it as a certainty. This is the closest Wittgenstein comes to a ‘definition’ of knowledge. Knowledge, it seems, goes together with certainty. Possessing compelling grounds is one path to certainty. Is there another? And what kind of certainty does knowledge require?

Beyond Unnatural Doubts  73 Let us take the second question first. Since one way of obtaining it is being able to demonstrate the truth of a proposition, the certainty that goes with knowing is not mere conviction but rather objective certainty. If this is right, Wittgenstein inclines towards an “infallibilist” conception of knowledge. Knowledge must be mistake-proof. Knowing requires more than truth and justification, where justification falls short of establishing the truth of what is believed: it requires the impossibility of having gone wrong. This is just what Wittgenstein thinks: “‘I know that’ means ‘I am incapable of being wrong about that’. But whether I am so needs to be established objectively” (16). This view of what knowledge requires has a good deal of intuitive plausibility: “I know that I live in Baltimore, though I might be wrong” just sounds fishy. But doesn’t it set the bar for knowing impossibly high? In fact, doesn’t it open the door to skepticism? Not as Wittgenstein develops it. If knowledge is infallible in the sense just indicated, possessing conclusive reasons is one way of knowing. But if basic certainties are things we know, there must be another way. What is it? Wittgenstein is well aware that his point about the unprovability of basic certainties invites this question and immediately suggests an answer: 2. From its seeming to me – or to everyone – to be so, it doesn’t follow that it is so. What we can ask is whether it can make sense to doubt it. 3. If someone says “I don’t know if there’s a hand here” he might be told “Look closer” – this possibility of satisfying oneself is part of the language-game. It is one of its essential features. Assuming that basic certainties involving worldly propositions are not self-evident, we might seem to be caught in a dilemma. On the one hand, they cannot be proved; on the other hand, their seeming to be so, even to everyone, yields only subjective certainty. (This is the sort of certainty that attaches to Pritchard’s ‘arational commitments’.) But there is a way between the horns. A proposition can be objectively certain because we cannot make sense of doubting it. Wittgenstein does not say what he means by ‘making sense’. 30 Does he mean that doubting would be unintelligible or just utterly unreasonable or impractical? As it turns out, the answer is ‘both’, 31 but his initial focus is on what it takes for an expression of doubt to be intelligible. Wittgenstein suggests that understanding an expression of doubt requires understanding why someone might be uncertain. 32 Consider Moore’s demonstration: if I’m uncertain whether Moore has held up his hand, this must be because, for example, I wasn’t in a position to take a good look. Once this defect is remedied, no room for doubt remains: I am in a position to know what’s what. Supposing me to be close enough to see and more or less normally sighted, no one would

74  Michael Williams have a clue what I meant if I still claimed not to know whether there was a hand there. We have the beginnings of a diagnosis of Agrippan skepticism. The Agrippan skeptic’s founding intuition is that merely by taking myself to know this or that I invite the question “How do you know?” If this were so, the entire justificatory burden would fall on the one who takes himself to be knowledgeable, nothing being required of the doubter. This is not how the game of doubting is or even could be played. 495. One might simply say ‘O, Rubbish’ to someone who wanted to make objections to the propositions that are beyond doubt. That is, not reply to him but admonish him. Groundless expressions of doubt can properly be dismissed. Sometimes doubts are just unreasonable. But although there is no bright line, unreasonableness shades into unintelligibility. The certainty that is grounded in the absence of intelligible doubts has nothing to do with practical limitations. Knowledge is mistake-proof, but not every false belief is a mistake. 71. If my friend were to imagine one day that he has been living for a long time past in such and such a place etc. etc., I should not call this a mistake, but rather a mental disturbance, perhaps of a t­ ransient kind. 74. Can we say that a mistake doesn’t only have a cause, it also has a ground? I.e., roughly: when someone makes a mistake, this can be fitted into what he knows aright? We can. Mistakes may be rationally corrected (given a willingness to ­listen etc.); mental disturbance not so much. But even this kind of possibly transient mental disturbance has its limits. 80. The truth of my statements is the test of my understanding of these statements. 81. That is to say: if I make certain false statements, it becomes uncertain whether I understand them. Unless my judgments about non-controversial matters are for the most part unproblematically true, there is a question as to what, if anything, I mean to say. Without making mostly true judgments, I can’t make false ones. Pritchard thinks that, reflecting our commitment to the über hinge proposition, we ignore the possibility that we are radically and fundamentally mistaken in our everyday beliefs. But such pervasive error is not a possibility we ignore: it is not a possibility at all. But getting lots of things right is a precondition for getting anything wrong. Wittgenstein’s point is not, or not just, about the structure of rational justification: it is about the connection between truth and meaning.

Beyond Unnatural Doubts  75 What about situations in which going wrong is not utterly i­nconceivable? They arise against a background of cases where error is not conceivable. But there is a further point to be made. Although knowledge is mistake-proof, I am not. The difference between knowing and merely believing (perhaps falsely), is not a matter of being in different mental states. The grammar of reporting on our mental states grants us qualified first person authority. If say, sincerely, that I doubt that Liverpool will win next Saturday, the question “How do you know you doubt?” is not usually in place. “I know” doesn’t work like that, though Moore’s assurances that he knows that he is gesturing with his hands suggest that he thinks it does. 21. Moore’s view really comes down to this: the concept ‘know’ is analogous to the concepts ‘believe’, ‘surmise’, ‘doubt’, ‘be convinced’ in that the statement ‘I know …’ can’t be a mistake. And if that is so, there can be an inference from such an utterance to the truth of an assertion. And here the form ‘I thought I knew’ is being overlooked … 42. One can say ‘He believes it, but it isn’t so’ but not ‘He knows it, but it isn’t so’. Does this stem from the difference between the mental states of belief and knowledge? No. One may for example call ‘mental state’ what is expressed by tone of voice in speaking, by gestures etc. It would thus be possible to speak of a mental state of conviction, and that may be the same whether it is knowledge or false belief … I can’t determine what I know by introspection. Knowledge is ­ istake-proof, but objectively so. The expression “I thought I knew” m must not be overlooked. Having no reason to think that I could be going wrong, I may blamelessly take myself to know, though, as it happens, I don’t know. In taking myself to know, thus to be in a situation where I cannot have gone wrong, I sometimes commit to more than is guaranteed by it being reasonable for me so to take myself. Circumstances have to co-operate: “It is always by favour of nature that one knows something” (505). If it turns out that circumstances were not as I, however blamelessly, took them to be, then I will have to retract: “I thought I knew”. Commitments acquired by taking ourselves to know, when they turn out not to be fulfilled, are the engine of self-correction.

Skepticism as a Language Game of Its Own The Cartesian skeptic accepts that doubts need grounds and provides them in the form of skeptical hypotheses. Admittedly, skeptical scenarios invoke possibilities that we do not and for everyday, practical purposes could not consider. But this does not show their theoretical irrelevance,

76  Michael Williams much less their impossibility. By Wittgenstein’s own s­ tandards, if we can’t rule out such possibilities, then we don’t really know what we think we know. Wittgenstein expects this challenge and is unmoved. 4. … [W]hat about such a proposition as ‘I know that I have a brain’? Can I doubt it? Grounds for doubt are lacking! Everything speaks for it, nothing against it. Nevertheless it is imaginable that my skull should turn out empty when it was operated on. Grounds for doubt are not as easy to come by as aficionados of ­skeptical scenarios suppose. My being able to imagine a situation in which my cranium is found to be empty gives me no grounds for my thinking that perhaps I don’t have a brain. We can imagine all sorts of things. That is to say, we can picture them or make up stories that are not obviously inconsistent. But so what? That Wile E. Coyote, in a cartoon, runs off a cliff and hangs in the air until he looks down is no reason to doubt the law of gravity. The brain-in-a-vat ‘hypothesis’ is no different. It is not a hypothesis: it is a scenario, and an extremely thin one at that. It gives no reason for doubting the existence of things that I can quite plainly see. Such imaginings have no support from, and often are in sharp conflict with, our entrenched common sense and scientific knowledge. This is why everything speaks in favor of (and nothing against) Wittgenstein’s having a brain, even if no one has seen it. Skeptical scenarios are not hypotheses: they are fairy-tales, on a par with the tale of sleeping beauty. How does she manage to sleep for a hundred years until awakened with a kiss? Don’t ask, you’re spoiling the fun. Such goings on seem possible not because they are possible but because we don’t ask awkward questions. Our sense of what is really possible is shaped by our knowledge of how they are. Wittgenstein takes note of this in his discussion of the tribesmen who say they go to the Moon (OC, 106–108). Wittgenstein allows that ‘going to the Moon’ might be how they interpret their dreams, but on a literal interpretation, their claim is utterly incredible. 108. … Not only is nothing of the sort ever seriously reported to us by reasonable people, but our whole system of physics forbids us to believe it. For this demands answers to the questions ‘How did he overcome the force of gravity?’ ‘How did he live without an ­atmosphere?’ and a thousand others which could not be answered. One reason not to credit the tribesmen’s exploits is that no one has ever seen them on the Moon. But an even more powerful reason is that, at the time, going to the Moon was impossible. Though scientists had ideas

Beyond Unnatural Doubts  77 about how one might go to the Moon, the technology had not been fully developed. Perhaps going to the Moon would soon be possible, but at the moment it wasn’t. We might think of the brain-in-a-vat scenario this way, though it is closer to the fantasy of unaided lunar voyaging. Although it may seem more up to date than Descartes’s Evil Demon fable, which is a naked appeal to magic (or theology), we have no idea whether it represents a genuine possibility or is just science fiction: magic in ‘scientific’ garb. The skeptic’s response will be that the scenario is only meant to ­represent a ‘logical’ or ‘metaphysical’ possibility. So much the worse for the idea of ‘metaphysical’ possibility. The skeptic’s appeal to this ill-­ defined notion is no more than a maneuver to keep alive a conundrum that does nothing to illuminate our understanding of epistemic concepts as they are deployed in either everyday life or systematic inquiry, except perhaps by helping us see how such concepts don’t work. These reflections bring us to the pivotal objection. These remarks on everyday doubting miss the point. When Wittgenstein says that everything speaks in favor of his having a brain, he is simply ignoring the distinctively philosophical question: how we know anything whatsoever about the ‘external’ world? Wittgenstein expects this reaction: 19. The statement ‘I know that here is a hand’ may … be continued: ‘for it’s my hand I’m looking at’. Then a reasonable man will not doubt that I know. – Nor will the idealist; rather he will say that he was not dealing with the practical doubt which is being dismissed, but there is a further doubt behind that one. – That this is an illusion must be shown in a different way.33 The skeptic is right to this extent: we cannot dispose of skepticism (or idealism) simply by confronting the skeptical conclusions with our ­ordinary ways of talking about knowledge and doubt. However, for the purpose of exposing the illusory character of philosophical doubt, the skeptic’s (idealist’s) admission that it lies “behind” everyday “practical” doubting is very significant: 24. The idealist’s question would be something like: ‘What right have I got not to doubt the existence of my hands?’ (And to that the answer can’t be: I know that they exist.) But someone who asks such a question is overlooking the fact that a doubt about existence only works in a language-game. Hence, that we should first have to ask: what would such a doubt be like?, and don’t understand this straight off. If philosophical doubt is not an extension or intensification of ordinary doubting, it must be a language game of its own. How is it played?

78  Michael Williams To ask this question is to engage in what I called ‘theoretical diagnosis’. Wittgenstein continues: 25. One may be wrong even about ‘there being a hand here’. Only in particular circumstances is it impossible. – ‘Even in a calculation one can be wrong – only in certain circumstances one can’t’. Objective certainty is deeply contextual in the sense that it is deeply­ circumstance-dependent. This idea is the key to combining a ­knowledge-­­­first approach to epistemology with skepticism about knowledge as an object of systematic theory. It is also the key to diagnosing Cartesian skepticism. To see this, consider a common way of putting the ‘paradox’ that ­Cartesian skepticism supposedly presents: the so-called ‘Argument from Ignorance’. Let O be some ‘ordinary’ proposition, say the proposition that I have hands; and let SK be the skeptical hypothesis that I am a handless brain-in-a-vat. We have: (AI-1) I know that O. (AI-2) I don’t know that not-SK. (AI-3) If I don’t know that not-SK, I don’t know that O. Obviously, these claims cannot all be true, but which should we reject? Finding AI-1 and AI-2 unassailable, many philosophers locate the source of the problem in AI-3. They admit that AI-3 can also seem beyond reproach, since it is an instance of some principle of epistemic closure: for example, the principle that if I know that P, and know that P entails Q, then I know that Q. But denying this principle generates what Keith de Rose calls “abominable conjunctions”: in this case, that I know that I have hands but don’t know that I am a handless brain-in-a-vat. 34 So we have a paradox that must be resolved by limiting one or another form of the closure principle. The ‘hinge epistemology’ reading of Wittgenstein embodies this understanding of the skeptical problem. 35 The issue of closure is a red herring. For Wittgenstein as for me, the crucial misstep is taken in the first premise. This premise builds in two assumptions. The first is that an ‘everyday’ proposition, as a candidate for being known, can be identified simply by some broad feature of its content: it is about a determinate object or situation that can be recognized without specialist knowledge. The second is that there is no evident reason to doubt the proposition’s truth: as far as we can tell, the situation appears to be conducive to our knowing what we take ourselves to know. Both assumptions are wrong. The proper objects of epistemic evaluation are what Austin called “statements”: claims made (or commitments undertaken) by particular speakers in particular ­circumstances.36 Whether a statement expresses knowledge, or a

Beyond Unnatural Doubts  79 commitment amounts to knowledge, depends on what is claimed, by whom, where, and when. Being in a position to know engages all four aspects of a person’s epistemic circumstances. To further develop this theme, Wittgenstein turns to knowledge in mathematics. Philosophers have often traced mathematical knowledge to a special “inner” process, such as the exercise of rational insight. But the ‘inner process’ or ‘state’ is unimportant: “What is interesting is how we use mathematical propositions” (38). An essential aspect of the use of such propositions is that, in certain circumstances, “a calculation is treated as absolutely reliable, as certainly correct” (39). With respect to “12 × 12 = 144,” there is no question of my saying “I see what you mean, but I’m not sure I agree.” If I think you might be mistaken, I haven’t learned my multiplication tables. That is why an expression of doubt fails to make sense here. The circumstantial absence of intelligible doubts obviates the need for a special faculty, giving us insight into ‘necessary’ truths: 43. What sort of proposition is this: ‘We cannot have ­miscalculated in 12 × 12 = 144’? It must surely be a proposition of logic. – But now, is it not the same, or doesn’t it come to the same, as the s­ tatement 12 × 12 = 144? Exempting some propositions from doubt is built into how we play the game. As calculations become more complicated, the possibility of making a mistake creeps in and so the need for checking may arise. But no rule determines when doubts become sensible: 44. If you demand a rule from which it follows that there can’t have been a miscalculation here, the answer is that we did not learn this through a rule, but through learning to calculate. The deeply circumstantial character of basic knowledge – non-evidential objective certainty – means that it cannot be reduced to rule. However, its resistance to theory is no reason for regret. “Nothing is lacking. We do calculate according to a rule, and that is enough” (46). The logico-­ linguistic situation is essentially the same for propositions about things in the world. As we saw in connection with Moore showing his audience his hands, in some situations, the notion of making a mistake gets no grip, in which case neither does doubt. But while such situations can be recognized, they cannot be specified by a rule. The skeptic – we might as well say the traditional epistemologist – will deny Wittgenstein’s claim that mathematical propositions and propositions about the world around us can be certain in the same way. ­Mathematical propositions, s/he will say, express necessary truths, which in simple cases are grasped by rational insight. Propositions about

80  Michael Williams the world around us are contingent: by definition possibly false and ­therefore ‘empirical’ in the sense that knowing them to be true (or false) involves perceiving things and events in the world around us. However, the skeptic understands perceptual dependence in terms of dependence on the evidence furnished by ‘perceptual experience’. On the basis of perceptual experience alone, we have knowledge only of how things appear to us. As skeptical hypotheses show, such evidence never delivers the absolute certainty given by ­rational insight. Wittgenstein’s conception of certainty casts doubt on the skeptic’s ­bifurcation of ‘propositions’ into ‘a priori’ and ‘empirical’. We have ­already seen that the ‘a priori’ propositions like “12 × 12 = 144” have nothing to do with rational insight, they are simply taken on board in the course of learning to calculate. Their ‘necessity’ is no more mysterious. It is a matter of their having become ‘fossilized’ (657): so widely useful that they have become deeply sedimented in our form of life. As for the empirical side of the bifurcation, facts about things around us can be known with certainty when it makes no sense to doubt them. Accordingly, there can be knowledge of contingent facts that does not depend on more basic evidence. Of course, we have to be able to recognize things we see. But given the requisite abilities and propitious circumstances, we can just tell how things are.37 ‘A priori’ and ‘empirical’ judgments are supposed to have wholly disparate epistemic characters, the former being (potentially) absolutely certain in a way that the latter never can be. In fact, in the right circumstances, the statement “Here is one hand” is not only just as certain as “12 × 12 = 144”, it is certain in the same way. The skeptic’s thought that propositions about things in the world ­depend on perceptual evidence leads straight to a further bifurcation within the class of ‘empirical’ propositions. Experiential propositions resemble simple a priori propositions in that, thoughtfully entertained, they too are mistake-proof. This is because although things can appear to be other than the way they are, appearances cannot: the ‘appears’ operator does not iterate. Being thus exempt from skeptical doubt, experiential propositions express knowledge that I would have even if I were a brain-in-a-vat or a victim of Demon deception.38 This is why experiential knowledge is epistemologically prior to knowledge of the world. The problem is to show how the former can provide an adequate for the latter: Cartesian skepticism is an underdetermination problem. But the circumstance-­dependent character of basic world-involving certainties, which is built into our ordinary practices of epistemic assessment, means that there is no such circumstance-independent order of epistemic priority. 250. My having two hands is, in normal circumstances, as certain as anything that I could produce in evidence for it. That is why I am not in a position to take the sight of my hands as evidence for it.

Beyond Unnatural Doubts  81 “Here is a hand” can be every bit as certain as “It seems to me as if there were a hand here”. The idea that knowledge of things around us is everywhere dependent on experiential knowledge – knowledge of how things appear – is a myth.39

Totality and Epistemological Realism Wittgenstein’s account of knowledge as objective certainty fleshes out the diagnosis of Cartesian skepticism that I offered in Unnatural Doubts. ­According to that diagnosis, the source of the skeptical problem is the idea that “empirical knowledge” answers to a circumstance-independent structure of epistemological priority. I called this idea “epistemological realism” because it takes knowledge of the world to possess an inherent structure. In the end, our standards for possessing knowledge are answerable to this structure: it is not answerable to them. So if we can think up error-­ possibilities, no matter how thinly described, that we cannot rule out given the constraints imposed by the structure of empirical knowledge, then we don’t really know facts about the world, though for practical purposes we may reasonably take ourselves to possess such knowledge. But to say this is to mischaracterize epistemic norms, which we are both responsible to and responsible for. As Wittgenstein reminds us, “Whether a proposition can turn out false … depends on what I make count as determinants for it” (5). Like Stroud before him, Duncan Pritchard challenges this diagnosis. Following the path of ‘hinge epistemology’, Pritchard takes Wittgenstein to have offered an account of ‘the structure of rational evaluation’ according to which hinge commitments are not themselves subject to rational evaluation because they make such evaluation possible. It follows that there is no possibility of rationally evaluating all our beliefs at once. All rational evaluation is essentially local. No skeptical problem emerges from our failure to come up with the illusory global legitimation of our claim to knowledge of the world that traditional epistemology aims to provide. As Pritchard notes, so far this Wittgenstein-inspired diagnosis is similar to mine. However, he thinks that I go wrong in taking Wittgenstein’s account of the structure of rational evaluation to entail the rejection of epistemological realism. Rejection of what I call ‘the totality condition’ on a philosophical understanding of worldly knowledge is an obvious consequence of what he sees as our shared commitment to the essential locality of rational evaluation. Rejection of epistemological realism is not. Perhaps all rational evaluation is essentially local but that (sic) within this constraint there are nonetheless invariant (i.e. context independent) epistemic categories? For example, perhaps it is both the case that all rational evaluation is local and that propositions regarding ­ ropositions one’s mental states have an epistemic priority relative to p relative to propositions concerning one’s environment.40

82  Michael Williams Pritchard’s argument is an ignoratio elenchi for, as this passage shows, his notion of context is very different from mine. For Pritchard, local context is doxastic context: ‘rational evaluation’ is informed by whatever hinge commitments and background knowledge are locally relevant. This deeply anti-Wittgensteinian idea of context ignores the circumstance dependence of basic, non-inferential knowledge. Pritchard is blind to the possibility that some claims and beliefs are exempt from rational evaluation not because they are arational commitments but b ­ ecause, in the circumstances, we cannot make sense of doubting them. Accordingly, he gets my position backwards. I reject the totality condition because I reject epistemological realism. Emphasizing the circumstance-­dependence of basic knowledge makes the case against epistemological realism clearer and more compelling. The totality condition is a condition on a philosophical account of knowledge of the external world. Like many epistemologists, Pritchard writes as if the external world were just the world around us. It isn’t: the ‘external’ world is the world external to the (Cartesian) mind. My argument was and is that epistemological realism – the idea that there are “context independent categories”, so that “propositions regarding one’s mental states have an epistemic priority relative to propositions concerning one’s environment” – both constitutes knowledge of the external world as an object of theory and at the same time presents the apparent underdetermination of knowledge of the world by its evidential basis as the fundamental problem for a theory of knowledge to solve. Cartesian skepticism and the very idea of knowledge of the external world as a kind of knowledge go hand in hand.

Conclusion Cartesian skepticism is an artifact of the traditional epistemological ­project, as we inherit it from Descartes. It is hostage to the ­Cartesian metaphysics of mind and meaning, or what Jay Rosenberg calls “the myth of mind apart”.41 Obviously, the myth is one of the later W ­ ittgenstein’s prime targets. As I read Wittgenstein, his epistemological reflections in On Certainty proceed against the background of his conception of meaning, set out in Philosophical Investigations, as rooted in linguistic practices, which are learned and take place in our interactions with other speakers and the world we share.42 Perhaps the fundamental mistake of hinge epistemology is the assumption that skepticism, as an epistemological problem, must yield to a narrowly epistemological solution: an account of the structure of rational justification, or rational entitlement, or something along these lines. This assumption makes it difficult if not impossible to appreciate the lessons that Wittgenstein has to teach.43

Beyond Unnatural Doubts  83

Notes 1 Williams [1996]. 2 Wittgenstein [1969]. Subsequent references in main text and footnotes given by section number. 3 I should also acknowledge the important work of Charles Travis. See Travis [2008]. 4 Stroud [1984]. See Chapter 1 for Stroud’s understanding of how skepticism arises and Coda for his pessimistic conclusion. 5 Williams [1996], p. 52. 6 Austin [1962], Quine [1980], Sellars [1963], Wittgenstein [2009]. 7 Stroud [2000a], p. 101. 8 Ibid., p. 104. 9 Stroud himself (ibid., p. 102) sees Descartes’s skeptical problem as turning on the claim that “there is no strictly sensory information the possession of which necessarily amounts to knowledge of the material world”. Strictly sensory information is information we would have even if we were victims of the malin genie, i.e. information about appearances. 10 Ibid, p. 121. 11 Williams [1996], p. xii. 12 Descartes draws on material from ancient Academic skepticism but transforms it. See Williams [2010]. In Williams [2015], I argue that Agrippan skepticism in its contemporary form is also fundamentally different from its ancient forebear. 13 Williams [1996a], p. 119. 14 For further discussion, see Williams [2004b]. 15 Williams [1991], p. 123. 16 Pritchard [2016], p. 103. 17 Ibid., p. 106. 18 Ibid. 19 For further discussion of the structure of the argument developed in sections 1–65 of Wittgenstein [1969], see Williams [2004a]. 20 Pritchard [2016], p. 71. 21 Strawson [1985], Ch. 1. 22 Hume [2000], p. 123. For discussion of Hume’s skepticism in relation to his “naturalism”, see Williams [2005, 2008]. 23 Wright [2004], p. 206. For further consideration of Wrights anti-skeptical strategy, see Williams [2012]. 24 Pritchard [2016], p. 90f. 25 Ibid., p. 95. 26 Lawlor [2008], Ch. 1. 27 Moyal-Sharrock [2016], p. 98. 28 Wittgenstein [2009], § 289. 29 Moore [1970a], pp. 165–166. 30 The German reads “ob man sinvoll bezweifeln kann”. Used, adverbially “sinvoll”, which can mean, among other things, “meaningfully”, “appropriately”, or even “practically”. Anscombe’s “makes sense” is comparably polysemic. 31 The idea that ‘knows’ can indicate the absence of intelligible doubts is present in Wittgenstein [2009], § 247. Discussing the thought that only the person involved can know what was his intention, Wittgenstein remarks parenthetically that in such a claim “‘know’ means that the expression of uncertainty is senseless” (sinnloss).

84  Michael Williams 32 Austin imposes a ‘definite lack’ condition on “How do you know?” ­questions, when asked pointedly [Austin [1961a], p. 52]. Wittgenstein is thinking along the same lines. 33 Although, Moore [1970a, 1970b] are generally taken to be attempts to ­refute skepticism, Moore’s intended targets in these essays are various forms of ­Idealism: Berkeleyan, Kantian, and neo-Hegelian. Moore corrects this common misunderstanding in Moore [1968]. Berkeleyan subjective idealism, Moore’s target in Moore [1970a], has an intimate connection with C ­ artesian skepticism but Moore is quite clear that proving that external things exist and that proving that we know for certain that external things exist are ­distinct undertakings. Wittgenstein is aware of Moore’s concern with idealism. In Wittgenstein [1969], the only occurrence of the word ­‘scepticism’ is found in the phrase “the scepticism of the idealist” (37). 34 De Rose [1995]. 35 Pritchard [2016], Ch. 3, offers a clear example. 36 Austin[1961b], pp. 87–88. 37 Cf. McGinn [1989], p. 133: … inferring and calculation…is not to be explained by the spurious idea of a common intuition into the absolutely necessary workings of a rigid calculus. Rather, our mutual participation in the practice of inferring and calculating is to be seen as the result of our shared natural propensity to respond to the training we receive in techniques for making transitions between propositions in the very same way … There is no foundation for our techniques that could, from the point of view of a non-participant in our practice justify the steps we take or the applications we make. But anyone who cannot be brought to respond as we do will be deemed mad or incompetent, incapable of acquiring an immensely important human skill. As McGinn rightly says, for Wittgenstein the certainty of everyday judgments about things around us is to be explained “in the same way”. ­McGinn’s book remains one of the most insightful treatments of Wittgenstein on skepticism. However, I disagree with McGinn about how Cartesian skepticism arises. See Williams [1996], p. 176f. 38 Empiricists have often thought that, among empirical judgments, e­ xperiential reports (judgements about appearances) are unique: though contingent, they are like mathematical propositions in that, if formulated with a clear understanding of their content, they cannot be false. For a classic deconstruction of this idea, see Sellars [1963], ¶ 32–34, pp. 164–167. 39 We have seen that Wittgenstein suspects Moore of thinking that knowing and believing are different mental states. While I am not so sure that he is right about Moore, the view has certainly been held. Charles Travis quotes Cook Wilson: There is … no common mental attitude to the object about which we know or about which we have an opinion. Moreover, it is vain to seek a common quality in belief, on the ground that the man who knows that A is B and the man who has that opinion both believe that A is B. Belief is not knowledge and the man who knows does not believe at all what he knows; he knows it. (Quoted in Travis [2008], p. 290) But Cook Wilson also holds that, as mental attitudes, knowing and ­believing must be introspectively distinguishable: [knowledge cannot be one of] two states of mind … the correct and the erroneous one … quite indistinguishable to the man himself. [For] as the man does not know in the erroneous state of mind, neither can he know in the other state. (Ibid., p. 293)

Beyond Unnatural Doubts  85

40 41 42 43

Travis calls this “the accretion”. Combined with an infallibilist conception of knowledge, it leads to the disastrous result that the only things we really know about are our mental contents. It is possible that Wittgenstein also takes the accretion, which he thinks of as Moore’s mistake, to be a source of the temptation to make facts about our own sense-data the only facts we really know. Much depends on how we understand (90), a difficult passage in which Wittgenstein traces the retreat to sense-data to a picture of perceptual knowledge with roots in “a primitive sense” of “know”. How this picture, the primitive sense of “know”, and Moore’s mistake are supposed to be related is far from clear. Pritchard [2016], p. 105. Rosenberg [1980], Ch. IX. Wittgenstein [2009]. My approach to the problem of skepticism in Unnatural Doubts is also in some ways overly epistemological. In that book, I draw a distinction between two kinds of diagnosis, therapeutic and theoretical, in order to distance myself from responses to skepticism that struck me as coming too quickly to the conclusion that skepticism is nonsensical. Surely, I argued, we understand skepticism well enough to see how certain theories of knowledge can be plausibly presented as responses to the problem; so a better approach was to excavate the theoretical presuppositions that give the entire skeptical problematic whatever sense it has. However, even at the time I recognized that distinction is not knife-edged, since the theoretical presuppositions of the skeptical problematic might very well extend to matters concerning meaning, thought and representation, though the theme remained ­insufficiently developed. Wittgenstein argues that “There are physical objects” (his version of the conclusion of Moore’s proof) is nonsense (36). “Physical object” is a ‘logical’ concept, having do with how things like hands and tables are individuated, thus with “instruction in the use of words.” In waving his hands while intoning “Here is one hand and here is another, so two external things exist”, Moore isn’t proving anything: he is teaching his audience the use of the term ‘external thing’, using his hands as exemplary cases. However, Wittgenstein is far from supposing that this is the end of the matter. He writes: 37. But is it an adequate answer to the skepticism of the idealist, or the assurances of the realist, to say that ‘There are physical objects’ is nonsense? For them after all it is not nonsense. It would, however, be an answer to say: this assertion, or its opposite is a misfiring attempt to express what can’t be expressed like that. And that it does misfire can be shown; but that isn’t the end of the matter. We need to realize that what presents itself to us as the first expression of a difficulty, or of its solution, may as yet not be correctly expressed at all. Just as one who has a just censure of a picture to make will often at first offer the censure where it does not belong, and an investigation is needed to find the right point of attack for the critic. The right point of attack is provided by the theoretical diagnosis I have e­ xplored in this paper.

Bibliography Austin [1961a]. J. L. Austin, “Other Minds,” in J. L. Austin, Philosophical ­Papers (Oxford: Clarendon Press), 76–116. Austin [1961b]. J. L. Austin, “Truth,” in J. L. Austin, Philosophical Papers ­(Oxford: Clarendon Press), 85–101. Austin [1962]. J. L. Austin, Sense and Sensibilia (Oxford: Oxford University Press).

86  Michael Williams De Rose [1995]. Keith DeRose, “Solving the Skeptical Problem,” Philosopical Review, 104, 1–52. Hume [2000]. David Hume, A Treatise of Human Nature, edited by David Fate Norton and Mary J. Norton (Oxford: Oxford University Press). Lawlor [2008]. Krista Lawlor, Assurance (Oxford: Oxford University Press]. McGinn [1989]. Marie McGinn, Sense and Certainty (Oxford: Blackwell). Moore [1968]. G. E. Moore, “Reply to My Critics,” in Paul Arthur Schilpp, The Philosophy of G.E. Moore, Vol. 2, 3rd edition (Lasalle, IL: Open Court 1968), 535–677. Moore [1970a]. “Proof of an External World,” in G.E. Moore, Philosophical Papers, 3rd edition (London and New York: Allen and Unwin/Humanities Press), 127–150. First published in 1925. Moore [1970b]. “A Defence of Common Sense,” in G.E. Moore, Philosophical Papers, 3rd edition (London and New York: Allen and Unwin/Humanities Press), 32–59. Moyal-Sharrock [2016]. Danielle Moyal-Sharrock, “The Animal in Epistemology,” in Hinge Epistemology, special issue of the International Journal for the Study of Skepticism, 6, 97–119. Pritchard [2016]. Duncan Pritchard, Epistemic Angst (Princeton, NJ: Princeton University Press). Quine [1980]. W. V. Quine, “Two Dogmas of Empiricism,” in W.V. Quine, From a Logical Point of View, 2nd revised edition (Cambridge, MA: Harvard ­University Press), 20–46. Rosenberg [1980]. Jay Rosenberg, One World and our Knowledge of It ­(Dordrecht, Holland: Reidel). Sellars [1963]. Wilfrid Sellars, “Empiricism and the Philosophy of Mind,” in ­Science, Perception and Reality (London and New York: Routledge/­ Humanities Press), 127–196. Strawson [1985]. P. F. Strawson, Skepticism and Naturalism: Some Varieties (New York: Columbia University Press). Stroud [1984]. Barry Stroud, The Significance of Philosophical Scepticism ­(Oxford: Clarendon Press). Stroud [2000a]. Barry Stroud, “Understanding Human Knowledge in General,” in Stroud, Understanding Human Knowledge (Oxford: Oxford University Press), pp. 99–121. Stroud [2000b]. Barry Stroud, “Epistemological Reflection on Knowledge of the External World,” in Barry Stroud, Understanding Human Knowledge ­(Oxford: Oxford University Press), pp. 22–138. Travis [2008]. Charles, Travis, “A Sense of Occasion,” in Charles Travis, ­Occasion-Sensitivity (Oxford: Oxford University Press), pp. 290–315. Williams [1996a]. Michael Williams, Unnatural Doubts (Princeton, NJ: ­Princeton University Press). Williams [1996b]. Michael Williams, “Understanding Human Knowledge ­Philosophically,” Philosophy and Phenomenological Research, 56:2 (June 1996), 345–358. Williams [2004a]. Michael Williams, “Wittgenstein’s Refutation of Idealism,” in Denis McManus, Wittgenstein and Scepticism (Abingdon: Routledge, 2004), pp. 76–96.

Beyond Unnatural Doubts  87 Williams [2004b]. Michael Williams, “Knowledge, Reflection and Sceptical Hypotheses,” Erkenntnis, 61:2–3 (November 2004), 143–172. Williams [2004c]. Michael Williams, “The Unity of Hume’s Philosophical ­Project,” Hume Studies, 20:2 (November 2004), 265–296. Williams [2008]. Michael Williams, “Hume’s Skepticism,” in The Oxford Handbook of Skepticism (Oxford: Oxford University Press), pp. 80–107. Williams [2010]. Michael Williams, “Descartes’s Transformation of the S­ ceptical Tradition,” in The Cambridge Companion to Ancient Scepticism, edited by Richard Bett (Cambridge: Cambridge University Press), pp. 288–313. Williams [2012a]. Michael Williams, “External World Scepticism and the Structure of Epistemic Entitlement,” in The Possibility of Philosophical ­Understanding, edited by Wai-hung Wong (New York: Oxford University Press 2012), pp. 43–61. Williams [2012b]. Michael Williams, “Wright against the Sceptics,” in Mind, Meaning and Knowledge: Themes from the Philosophy of Crispin Wright, edited by Anna Maria Coliva (New York: Oxford University Press), pp. 352–376. Williams [2015]. Michael Williams, “The Agrippan Problem, Then and Now,” International Journal for the Study of Skepticism, 1:2, 124–137. Williams [2018]. Michael Williams, “Wittgenstein and Skepticism: Illusory Doubts,” in Skepticism from Antiquity to the Present, edited by Diego ­Machuca and Baron Reed (London: Bloomsbury), pp. 481–505. Wittgenstein [1969]. Ludwig Wittgenstein, On Certainty, edited by G. E. M. Anscombe and G. H. von Wright, translated by Denis Paul and G. E. M. ­A nscombe (New York and Evanston: Harper). Wittgenstein [2009]. Ludwig Wittgenstein, Philosophical Investigations, ­G erman text with English translation by G E. M. Anscombe, P. M. S. Hacker and Joachim Schulte, Revised 4th edition by P. M. S. Hacker and Joachim Schulte (Chichester: Wiley-Blackwell). Wright [2004]. Crispin Wright, “Warrant for Nothing (and Foundations for Free)?” Proceedings of the Aristotelian Society, (supplementary volume) 78, 167–212.

5 Meaning and Ostension From Putnam’s Semantics to Contextualism François Recanati

Putnam is known for having demonstrated the existence of a new form of context-dependence, namely that which characterizes natural kind terms (and possibly others as well). Terms like ‘tiger’ and ‘water’ are indexical, Putnam says, since their conditions of application (i.e. the p ­ roperty or properties something has to possess in order to be in the extension of the term) vary with the context of use – in a suitably broad sense of ‘context’. A term like ‘water’ or ‘tiger’ is conventionally associated with a stereotype, i.e. a cluster of properties known to any competent user of the language. The stereotype enables one to identify paradigmatic exemplars or instances of the category in one’s local environment. But the property which determines whether or not something belongs to the extension of ‘water’ or ‘tiger’ is not the cluster in question: it’s a different property which the local exemplars happen to instantiate as a matter of empirical fact. In our environment, the local exemplars that fit the tiger-stereotype are animals, but in a different environment, they would be robots controlled from Mars. In our context, therefore, ‘tiger’ refers to a certain type of animal, while in the counterfactual context the same word, associated with the same stereotype, would refer to a different type of entity sharing only superficial characteristics with our tigers. The story is well-known, and I will not dwell on it. What I will f­ ocus on in this chapter is the relation between Putnam’s semantics and a body of views I call ‘contextualism’.1 Contextualism generalizes context-­ sensitivity: it claims that sentences carry content only in the context of a speech act. Sentences, in vacuo, are not truth-evaluable. This view was put forward by ordinary language philosophers in the mid-twentieth century, and it has re-surfaced in more recent times in the works of philosophers like John Searle, Charles Travis, and myself. I think Putnam’s semantics has strong affinities with contextualism, and I want to argue this point.

I A good place to start is Waismann’s classic paper ‘Verifiability’, and more specifically the passage where he introduces the notion of ‘open texture’. 2 That passage is representative of a view (and a type of example) which

Meaning and Ostension  89 Waismann shared with Austin and Wittgenstein, and which has been revived by Travis and Searle. According to this view, truth-­conditional content is essentially unstable and context-dependent. As we shall see, the thought-experiment conducted by Waismann in that passage is similar in some respects to Putnam’s twin-Earth thought-experiment. What Waismann says in introducing open texture can be paraphrased as follows. It seems that we have no problem assigning truth-conditions to ordinary sentences. To do so we describe a state of affairs, i.e. a type of situation, the obtaining of which is necessary and sufficient for the sentence to be true. That seems easy enough to do. Thus we think we know the truth-conditions of ‘There is a cat next door’, or ‘This is gold’ (or ‘This is a man’): we can specify a state of affairs s such that the utterance is true if and only if s obtains. But this is an illusion – we can’t really. Given an utterance u and the state of affairs s, which is its alleged truth-maker, it is always possible to imagine a world in which s obtains, yet it is not the case that u is true (with respect to that world). To show that, one has only to embed the state of affairs s within a larger situation, by providing further details about an imagined world in which s obtains. If the world in question is sufficiently unlike our world (for example, if the ‘cat’ talked about turns out to speak Latin, or grows to a fantastic size, or changes into a fish), we shall be at a loss when it comes to deciding whether the statement ‘There is a cat next door’ is actually true, with respect to that world, even though the state of affairs we initially specified obtains. Is a world in which there is an animal next door exactly like a cat in all respects except that it speaks Latin a world in which there is a cat next door? We don’t know, because “most of our empirical concepts are not delimited in all possible directions”. 3 Or, to take another classic example: The notion of gold seems to be defined with absolute precision, say by the spectrum of gold with its characteristic lines. Now what would you say if a substance was discovered that looked like gold, satisfied all the chemical tests for gold, whilst it emitted a new sort of radiation? ‘But such things do not happen’. Quite so; but they might happen, and that is enough to show that we can never exclude altogether the possibility of some unforeseen situation arising in which we shall have to modify our definition. Try as we may, no concept is limited in such a way that there is no room for any doubt. We introduce a concept and limit it in some directions; for instance, we define gold in contrast to some other metals such as alloys. This suffices for our present needs, and we do not probe any farther. We tend to overlook the fact that there are always other directions in which the concept has not been defined. And if we did, we could easily imagine conditions which would necessitate new limitations. In short, it is not possible to define a concept like gold with absolute

90  François Recanati precision, i.e. in such a way that every nook and cranny is blocked against entry of doubt. That is what is meant by the open texture of a concept.4 We find something analogous to Waismann’s embeddings in Searle’s writings on the ‘background’. In ‘Literal Meaning’, Searle enquires into the truth-conditions of ‘The cat is on the mat’. It’s easy to describe the sort of state of affairs that would make the sentence true (with respect to a particular assignment of values to indexical expressions). But once we have described such a state of affairs, we can embed it within an extraordinary situation: Suppose that the cat and the mat are in exactly the relations depicted only they are both floating freely in outer space, perhaps outside the Milky Way galaxy altogether… Is the cat still on the mat? And was the earth’s gravitational field one of the things depicted…? What I think is correct to say as a first approximation in answer to these questions is that the notion of the literal meaning of the sentence “The cat is on the mat” does not have a clear application, unless we make some further assumptions, in the case of cats and mats floating freely in outer space.5 Now Waismann has an explanation for the fact that, in extraordinary situations, ‘words fail us’, as Austin puts it. Here it goes: If I had to describe the right hand of mine which I am now holding up, I may say different things of it: I may state its size, its shape, its colour, its tissue, the chemical compound of its bones, its cells, and perhaps add some more particulars; but however far I go, I shall never reach a point where my description will be completed: logically speaking, it is always possible to extend the description by adding some detail or other. Every description stretches, as it were, into a horizon of open possibilities: however far I go, I shall always carry this horizon with me. (…) [This] has a direct bearing on the open texture of concepts. A term is defined when the sort of situation is described in which it is to be used. Suppose for a moment that we were able to describe situations completely without omitting anything (as in chess), then we could produce an exhaustive list of all the circumstances in which the term is to be used so that nothing is left in doubt; in other words, we could construct a complete definition, i.e. a thought model which anticipates and settles once and for all every possible question of usage. As, in fact, we can never eliminate the possibility of some unforeseen factor emerging, we can never be quite sure that we have included in our definition everything that should be included, and thus the process of defining and refining an idea will go on without ever reaching a final stage.6

Meaning and Ostension  91 Waismann’s point is simple enough. A term is used in, or applies to, ­situations of a certain type – situations with which we are acquainted. (Or at least, our mastery of the term is somehow connected to our having experienced the type of situation in question.) To define a term is to describe the type of situation in question. The problem is that the situations in question, like Waismann’s right hand or any aspect of empirical reality, cannot be completely described: they possess an indefinite number of features, some of which may never have been noticed and perhaps will never be noticed by anyone. When we describe an empirical situation, we make certain features explicit, but an indefinite number of other features remain implicit and constitute a sort of hidden ‘background’. (Here I use Searle’s term, on purpose.) Now the applicability of a term to novel situations depends on their similarity to the source situations, i.e. to the situations by association with which the term has acquired the meaning it has. For the term to be (clearly) applicable, the target situation must be similar to the source situations not only with respect to those features which easily come to mind and constitute the ‘explicit’ definition of the term (the ‘tip’, to use the iceberg metaphor), but also with respect to the hidden background. If the two situations considerably diverge with respect to the latter, it is unclear whether or not the term will be applicable, even though the explicit conditions of satisfaction are satisfied. Other examples can be provided where the divergence between the source situation and the target situation affects not only the hidden background but also the explicit part (the tip of the iceberg). Thus Searle imagines that ‘Snow is white’ is uttered as a description of the following situation: Suppose that by some fantastic change in the course of nature the earth is hit by an astronomical shower of radiation that affects all existing and future water molecules in such a way that in their crystalline form they reflect a different wave length when in sunlight from what they did prior to the radiation shower. Suppose also that the same shower affects the human visual apparatus and its genetic basis so that snow crystals look exactly as they did before. Physicists after the shower assure us that if we could see snow the way we did before, it would look chartreuse but because of the change in our retinas, which affects our observation of snow and nothing else, snow looks the same color as ever and will continue to do so to ensuing generations… Would we say that snow was still white?7 Putnam’s twin-Earth thought-experiment has the same structure.8 It is a case of divergence between source situations and target situation, where the divergence affects the tip and not merely the hidden background. For if we are to define water, we shall say that it is a liquid with such and such phenomenal properties, and that its chemical structure is H 2O, in

92  François Recanati the same way in which the whiteness of snow has to do both with its ­reflecting a certain wavelength and with its looking to us a certain way. We can imagine an extraordinary situation such that the two things that go together as part of the tip get separated, e.g. a situation in which the liquid which has all the superficial characteristics of water is not H2O but XYZ. In such cases we don’t know what to say because there is some measure of similarity between the source situations and the target situation, but we have no contextual clue as to which dimension of similarity matters. That is not the conclusion Putnam himself draws from his thought-­ experiment. According to Putnam, twater is definitely not water. Instead of saying that we don’t know whether or not twater counts as water, as the contextualists would do, Putnam claims that we do know: necessarily, since water is H 2O, twater is not water. I will come back to this difference between Putnam and the contextualists in a moment. First, however, I want to mention a couple of other respects under which ­Waismann’s position diverges from Putnam’s.

II Waismann’s argument for the unstability of application conditions can be objected to on two grounds. First, it presupposes ‘descriptivism’, i.e. the idea that the only way to define words is to do so descriptively. ­Second, it conflates the semantic and the epistemological. In these two respects, Waismann’s position seems to be significantly different from that of contemporary philosophers of language like Putnam and Kripke. Let me start with the first objection. All that Waismann’s ­argument ­establishes, arguably, is that words cannot be defined in purely ­descriptive terms. But this does not imply that words cannot be defined at all, or that they do not possess definite (stable) conditions of application. The reason why words cannot be defined in purely descriptive terms is that they have an irreducibly referential dimension: they, as it were, point to real situations in the world – situations which have an indefinite number of features and cannot be exhaustively described. Why not simply incorporate this referential dimension into our statements of truth- and application-conditions? Why not follow Putnam and define a cat as an animal belonging to the same species as this specimen (or those specimens, where the specimens in question are normal cats found in the local environment)? Why not define gold as this metal (while pointing to a piece of gold)? By explicitly incorporating ostensive reference to the actual environment into the definition of predicates like ‘gold’ and ‘cat’, as Putnam’s semantics does, it seems that we can overcome the alleged unstability of truth-conditions. ‘There is a cat next door’ will be true, in the imagined situation (where a catlike animal next door speaks Latin), provided the catlike animal belongs to the same species as our cats (the cats to be found in ordinary situations).

Meaning and Ostension  93 The second objection is the following. Of course, we do not know whether or not the strange animal we imagine would be considered a cat by the scientists, just as we don’t know whether or not a gold-like metal emitting a new sort of radiation would count as gold. But our epistemic limitations do not prevent the word ‘gold’ (or the word ‘cat’) from having a definite content and a definite extension: gold is anything that is the same metal as this, and a cat is any animal of the same species as those. Our epistemic limitations, so much emphasized by Waismann, have no bearing on the properly semantic issue. These objections, inspired by Putnam’s work, are well-taken but they are not decisive. In particular, they do not threaten Waismann’s conclusion regarding the unstability of truth- and application-conditions. They would threaten that conclusion if, by appealing to ostensive definitions, we could determine a stable content (i.e. stable conditions of application) for the word thus defined. But we cannot. For a stable content to be determined by an ostensive definition à la Putnam, the dimension of similarity to the demonstrated exemplars must itself be fixed.9 But what fixes the dimension of similarity? On Waismann’s picture, words are associated with situations of use, that is all. To apply the word to or in a novel situation, that situation must be similar to the source situations; but we cannot survey in advance all the possible dimensions of similarity between the source situations and possible target situations: open texture again. According to Putnam (op. cit.), the predicate ‘water’ means something like: ‘same-L[iquid] as the transparent, odorless, thirst-quenching stuff to be found in lakes and rivers in the local environment’, or more simply ‘same-L as that’ (pointing to a sample of water). If we change the demonstrated stuff (i.e. the liquid which satisfies the stereotype of ­water in the local environment) we may thereby change the extension of ­‘water’. ­‘ Water’, therefore, has an indexical component, Putnam says. Yet there is another form of context-sensitivity in play here, which ­Putnam’s ­ostensive definitions do not properly capture: the dimension of ­similarity itself is not given, but contextually determined. In some contexts the chemical composition of the demonstrated stuff will be relevant, in other contexts only functional properties will be relevant.10 Accordingly XYZ will count as water in some contexts simply because it is the liquid which is odorless, colorless, quenches thirst, and can be found in lakes and rivers. If the conversation bears on the issue of what currently fills a certain bottle, milk or water, then plainly the answer ‘It’s water’ will be true, even if it is XYZ rather than H 2O.11 Putnam himself expresses awareness of this point: x bears the relation same-L to y just in case (1) x and y are both liquids, and (2) x and y agree in important physical properties…. Importance is an interest-relative notion. Normally the ‘important’

94  François Recanati properties of a liquid or solid, etc., are the ones that are structurally important; the ones that specify what the liquid or solid, etc., is ultimately made out of… From this point of view the important characteristic of a typical bit of water is consisting of H 2O. But it may or may not be important that there are impurities… And structure may sometimes be unimportant; thus one may sometimes refer to XYZ as water if one is using it as water… Even senses that are so far out that they have to be regarded as a bit ‘deviant’ may bear a definite relation to the core sense. For example I might say ‘did you see the lemon’, meaning the plastic lemon. A less deviant case is this: we discover ‘tigers’ on Mars. That is, they look just like tigers, but they have a silicon-based chemistry instead of a carbon-based chemistry… Are Martian ‘tigers’ tigers? It depends on the context.12 We now have in hand all the ingredients which I take to be constitutive of contextualism: the demonstrative (referential) dimension of meaning; the central role of similarity in determining extension (hence truth-­ conditions); and finally the context-dependence of similarity relations.

III On the contextualist picture, words are not primitively associated with abstract ‘conditions of application’, constituting their conventional meaning (as on the Fregean picture). The conditions of application for words must be contextually determined, like the reference of indexicals. What words, qua linguistic types, are associated with are not abstract conditions of application, but rather particular applications. The set of applications a word is associated with I call its ‘semantic potential’. In the spirit of Wittgenstein, consider what it is for someone to learn a predicate P. The learner, whom I’ll call Tom, observes the application of P in a particular situation S; he associates P and S. At this stage, the semantic potential of P for Tom is the fact that P is applicable to S. In a new situation S’, Tom will judge that P applies only if he finds that S’ sufficiently resembles S. To be sure, it is possible for S’ to resemble S in a way that is not pertinent for the application of P. The application of P to S’ will then be judged faulty by the community, and Tom will stand corrected. The learning phase for Tom consists in noting a sufficient number of situations which, like S, legitimate the application of P, as opposed to those, like S’, which do not legitimate it. The semantic potential of P for Tom at the end of his learning phase can thus be thought of as a collection of legitimate situations of application; that is, a collection of situations such that the members of the community agree that P a­ pplies in or to those situations. The situations in question are the source-­situations. The future applications of P will be underpinned, in Tom’s usage, by the

Meaning and Ostension  95 judgement that the situation of application (or ­target-situation) is similar to the source-situations. In this theory the semantic potential of P is a collection of source-­ situations, and the conditions of application of P in a given use, involving a given target-situation S’’, are a set of features which S’’ must possess to be similar to the source-situations. The set of features in question, and so the conditions of application for P, will not be the same for all uses; it is going to depend, among other things, on the target-situation. One target-situation can be similar to the source-situations in certain respects and another target-situation can be similar to them in different respects. But the contextual variability of the conditions of application does not end there. Even once the target-situation is fixed, the relevant dimensions for evaluating the similarity between that situation and the source-situations remain underdetermined: those dimensions will vary as a function of the subject of conversation, the concerns of the speech participants, etc. One particularly important factor in the contextual variation is the relevant ‘contrast set’. As Tversky has pointed out, judgements of similarity are very much affected by variations along that dimension.13 If we ask which country, Sweden or Hungary, most resembles Austria (without specifying the relevant dimension of similarity), the answer will depend on the set of countries considered. If that set includes not just Sweden, Hungary, and Austria but also Poland, then Sweden will be judged more like Austria than Hungary; but if the last of the four countries considered is Norway and not Poland, then it is Hungary which will be judged more like Austria than Sweden. The explanation for that fact is simple. Poland and Hungary have certain salient historical and geopolitical features in common which can serve as basis for the classification: Hungary and Poland are then put together and opposed to Austria and Sweden. If we replace Poland by Norway in the contrast set a new principle of classification emerges, based on the salient features shared by Norway and ­Sweden: in this new classification Hungary and Austria are back together. Tversky concludes that judgements of similarity appeal to features having a high ‘diagnostic value’ (or classificatory significance), and that the diagnostic value of features itself depends on the available contrast set. So the set of similarity features on which sense depends itself depends upon the relevant contrast set, and the relevant contrast set depends upon the current interests of the conversational participants. It follows that one can, by simply shifting the background interests ascribed to the conversational participants, change the truth-conditions of a given utterance, even though the facts (including the target-situation) don’t change, and the semantic values of indexicals remain fixed. Charles Travis has produced dozens of examples of this phenomenon of truth-­ conditional shiftiness over the last 40 years, and his examples often ­involve ­manipulating the relevant contrast-set.14

96  François Recanati In this framework the background-dependence of truth-conditions emphasized by Searle is accounted for by appealing to the global character of the similarity between target-situation and source-­situations. As Waismann stresses the source-situations are concrete situations with an indefinite number of features. Some of these features are ubiquitous and their diagnostic value in a normal situation is vanishing.15 They belong to the most general and immutable aspects of our experience of the world: gravity, the fact that food is ingested via the mouth, etc. When we specify the truth conditions of a sentence (for example the sentence ‘The cat is on the mat’), or the conditions of application of a predicate (for example the predicate ‘on’ in that sentence), we only mention a small number of features – the ‘foreground’ features – because we take most of the others for granted; so we do not mention gravity, we presuppose it. ­Nevertheless, gravity is one of the features possessed by the situations which are at the source of the predicate ‘on’; and there is an indefinite number of such features. These background features of the source-­situations can be ignored inasmuch as they are shared by the situations of which we may wish to speak when we utter the sentence; but if we imagine a target-situation where the normal conditions of experience are suspended, and where certain background features of the source-situations are not present, then we shatter the global similarity between the target-situation and the source-­situations. Even if the target-situation has all the foreground features which seem to enter into the ‘definition’ of a predicate P, it suffices to suspend a certain number of background features in order to jeopardize the application of P to the target-­situation. That shows that the semantic potential of P is not, as in Fregean semantics, a set of conditions of application determined once and for all, but a collection of source-situations such that P applies to a ­target-situation if and only if it is relevantly similar to the source-situations. A caveat: as Searle himself emphasizes, the fact that the target-­situation does not possess certain background features of the source-situations does not automatically entail the non-applicability of the predicate P. It can be that the background features which the target-situation does not possess (for example gravity) are contextually irrelevant and do not affect the application conditions of the predicate.16 For the same sort of reason, the possession by the target-situation of what I have called the foreground features of the source-situations is no more a necessary condition for the application of the predicate than it is a sufficient condition. For a predicate (or a sentence) to apply to a target-situation that situation must resemble the source-situations under the contextually relevant aspects. So a predicate can apply even if the target-situation differs markedly from the source-situations, as long as, in the context and taking into account the contrast set, the similarities are more significant than the differences. Thus, in certain contexts, as Putnam notices, the predicate ‘lemon’ will apply to plastic lemons, or the word ‘water’ to XYZ.

Meaning and Ostension  97 I conclude that Putnam’s position, suitably radicalized, leads directly to contextualism. The important contextualist idea – which forms the core of Austin’s theory of truth – is twofold: first, words are associated with worldly situations (at whatever level of abstractness and fine-grainedness), i.e. entities at the level of reference rather than the level of sense; second, the sense of words, i.e. their conditions of application, depends upon similarities between those situations and the target-situation in which or to which the word is applied. Both ideas are present in Putnam’s semantics. The only thing that has to be added to get contextualism is a third idea which is also (though less centrally) present in Putnam, viz. the idea that the relevant similarity relations themselves are not fixed once for all but depend upon the context.17

Notes 1 My use of ‘contextualism’ to refer to a particular view in the philosophy of language goes back to Direct Reference (1993: 260). It is to be distinguished from the use of ‘contextualism’ to refer to a particular view in epistemology. 2 Friedrich Waismann, ‘Verifiability’, in Flew, Anthony (ed.), Logic and ­L anguage, 1st series (Oxford: Blackwell, 1951), pp. 119–123. 3 Ibid., p. 120. 4 Ibid. 5 John Searle, “Literal Meaning”, in Erkenntnis 13 (1978), p. 211. 6 Waismann, “Verifiability”, pp. 121–123. 7 John Searle “The Background of Meaning”, in Searle, John, Kiefer, ­Ferenc, and Bierwisch, Manfred (eds.), Speech Act Theory and Pragmatics ­(Dordrecht: Reidel, 1980), p. 230. 8 Hilary Putnam, “The Meaning of ‘Meaning’”, in his Philosophical Papers, vol. 2: Mind, Language and Reality (Cambridge: Cambridge University Press, 1975), pp. 223–227. 9 As McKay and Stern point out, ‘to determine the extension of the natural kind term, we need, in addition to the sample, at least some indication of the breadth of the term – the respects in which other individuals [in the extension] must be related to the sample.’ (Thomas McKay and Cindy Stern, “Natural Kind Terms and Standards of Membership”, in Linguistics and Philosophy 3 [1979], p. 27) 10 A reader of Saul Kripke’s Naming and Necessity (Oxford: Blackwell, 1980) might object as follows: ‘We are concerned only with what words mean in our language, not with what they mean in other possible situations. Now, in our language water is H 2O, and This is water is true if the demonstrated stuff is H 2O.’ This pseudo-Kripkean objection is mistaken and easy to rebut. I am talking about our language. Our language is such that we can imagine contexts in which it would be true to say ‘This is water’ of some stuff if the stuff in question had certain phenomenal/functional properties, whether or not the thing in question was H 2O. 11 Noam Chomsky, New Horizons in the Study of Language and Mind ­(Cambridge: Cambridge University Press, 2000), p. 41. 12 “The Meaning of ‘Meaning’”, pp. 238–239. 13 See Amos Tversky, “Features of Similarity”, in Psychological Review 84 (1977), 327–352.

98  François Recanati 14 See Charles Travis, Saying and Understanding (Oxford: Blackwell, 1975); The True and the False (Amsterdam: Benjamins, 1981); The Uses of Sense (Oxford: Clarendon Press, 1989); Unshadowed Thought (Cambridge: ­Harvard University Press, 2000); Occasion-Sensitivity (Oxford: Oxford University Press, 2008), etc. The following example, inspired from Austin, is taken almost at random from a list of Travis-examples compiled by Claudia Bianchi (then a graduate student of mine): Fred is walking with his young nephew beside a pond where a decoy duck is floating. Pointing to the decoy, he says, “That’s a duck”. Again we might ask whether what he said is true or false. But again, the above description is not enough for us to tell. If Fred has just finished laughing at a sportsman who blasted a decoy out of the pond, and if he has been trying to show his nephew how to avoid similar mistakes, then what he said is false. But suppose that Fred and his nephew are attending the annual national decoy exhibition, and the boy has been having trouble distinguishing ducks from geese. Then what Fred said may well be true. It would also be true had Fred said what he did in pointing out the fact that all the other ducks were poor copies (perhaps on the order of Donald Duck). (Saying and Understanding, p. 51) 15 See Tversky, “Features of Similarity”, p. 342: The feature ‘real’ has no diagnostic value in the set of actual animals since it is shared by all actual animals and hence cannot be used to classify them. This feature, however, acquires considerable diagnostic value if the object set is extended to include legendary animals, such as a centaur, a mermaid or a phoenix. 16 It is easy to imagine a context with respect to which it would be definitely true to say ‘the cat is on the mat’ of the gravitationless situation described by Searle: For example, as we are strapped in the seats of our spaceship in outer space we see a series of cat-mat pairs floating past our window. Oddly, they come in only two attitudes. From our point of view they are either as depicted in Figure 1 or as would be depicted if Figure 1 were upside down. “Which is it now?”, I ask. “The cat is on the mat”, you answer. Have you not said exactly and literally what you meant? (Searle, “Literal Meaning”, p. 212) 17 This chapter is a lightly revised version of a talk I gave in a conference on Hilary Putnam (“Les Défis d’Hilary Putnam”) in Paris in March 2005. The talk itself was based upon chapter 9 of my book Literal Meaning, published the year before. I am grateful to Christiane Chauviré, organizer of the conference, for inviting me to give the talk in question, and to Eduardo Marchesan for proposing to include it in this volume.

Bibliography Austin, John “Truth”, in his Philosophical Papers, 2nd edition, Oxford: Oxford University Press, 1970, 119–133. Chomsky, Noam New Horizons in the Study of Language and Mind, ­Cambridge: Cambridge University Press, 2000. Kripke, Saul Naming and Necessity, Oxford: Blackwell, 1980.

Meaning and Ostension  99 McKay, Thomas and Stern, Cindy “Natural Kind Terms and Standards of Membership”, in Linguistics and Philosophy 3 (1979), 27–34. Putnam, Hilary “The Meaning of ‘Meaning’”, in his Philosophical Papers, vol. 2: Mind, Language and Reality, Cambridge: Cambridge University Press, 1975, 215–271. Recanati, François Direct Reference, Oxford: Blackwell, 1993. Recanati, François Literal Meaning, Cambridge: Cambridge University Press, 2004. Searle, John “Literal Meaning”, in Erkenntnis 13 (1978), 207–224. Searle, John “The Background of Meaning”, in Searle, John, Kiefer, Ferenc, and Bierwisch, Manfred (eds.), Speech Act Theory and Pragmatics, Dordrecht: Reidel, 1980, 221–232. Travis, Charles Saying and Understanding, Oxford: Blackwell, 1975. Travis, Charles The True and the False: The Domain of the Pragmatic, ­A msterdam: Benjamins, 1981. Travis, Charles The Uses of Sense: Wittgenstein’s Philosophy of Language, ­Oxford: Clarendon Press, 1989. Travis, Charles Unshadowed Thought, Cambridge, MA: Harvard University Press, 2000. Travis, Charles Occasion Sensitivity, Oxford: Oxford University Press, 2008. Tversky, Amos “Features of Similarity”, in Psychological Review 84 (1977), 327–352. Waismann, Friedrich “Verifiability”, in Flew, Anthony (ed.), Logic and ­L anguage, 1st series, Oxford: Blackwell, 1951, 117–144.

6 The Role of Intention in Truth1 Eduardo Marchesan

It is often thought that there is a gap between semantics, which d ­ etermine the truth conditions of a sentence, and pragmatics, which allow a speaker to express something other than what is said. In assessments of this distinction, criticism of the notion of “what is said”, particularly as described by Paul Grice (1989), has figured prominently. In Grice’s view, what a speaker says when he or she makes an assertion corresponds to the expression of a propositional content codified in the meanings of the words that compose the uttered sentence. Thus, saying something would be the same as verbalizing contents equivalent to the truth conditions of the sentence. Any discrepancy between the spoken content and the sentence in its abstract form would be restricted to the possible gap in the semantic content produced by the presence of indexical constituents. Despite the apparent banality of the description of this category, it is widely thought to be important because it is centered on the verbalization executed by the speaker, in the sense that it prioritizes the speaker’s intention. When uttering a sentence, the speaker may equally wish to express exactly what the sentence means (i.e. the speaker wanted to say what was said) or to express something that is extrapolated from the meaning of that sentence. In the second case, what the speaker wished to express is conveyed as a conversational implicature, e.g. A asks B how his colleague C is getting on in his new job whereupon B replies “Oh, quite well, I think; he likes his colleagues, and he hasn’t been to prison yet” (Grice, 1989, p. 24). Obviously, B wants to convey something quite different from the sentence that he originally said, such as “C tends to be a troublemaker” or “C is dishonest”. Notwithstanding the primacy of the speaker’s intention, the fact is that in both cases, the content of what is said is treated as being fundamentally stable. Even in cases where implicatures are in play, the possibility of the hearer recognizing these implicatures involves her understanding of the utterance, i.e. understanding the literal meaning of the spoken sentence. Indeed, that literal content is precisely the foundation for the implicature, and therefore for the possibility of understanding the speaker’s unspoken intention because it involved the recognition that what the speaker had said seemed somewhat out of place (or, as Grice would

The Role of Intention in Truth  101 say, the speaker had flouted a conversational maxim), which enabled the hearer to detect that something else was being expressed. In this way, Grice’s well-known model established a compromise ­between the semantic and pragmatic domains, since he based the ­centrality of the speaker’s intention and the belief that he intended to express, on the possibility of mobilizing a content whose description rested entirely within the realm of semantic studies. By organizing the intersection of the two domains in such a way Grice’s model actually enabled their strict separation; a separation that could be formulated as the thesis that pragmatic factors had no influence whatever on the truth conditions of what was said. Criticism of the semantic notion of what is said threatens to ­undermine the classic distinction between semantics and pragmatics just described. This criticism was mainly proposed by two contextualist tendencies, which are sometimes referred to as Radical Contextualism.2 Both tendencies endeavor to show that the possibility of configuring a given propositional content (i.e. the content of what is said) depends on extralinguistic factors that are not susceptible to the strict semantic rules codified in the language. Despite their important similarities, these two tendencies are based on different presuppositions, gathered from different legacies of the philosophy of ordinary language. On the one hand, the thesis inherited from the Oxford tradition whereby an affirmative sentence could only be recognized as true or false as part of a speech act gained ground as the way to demonstrate the unfeasibility of describing what is said on Grice’s terms. On the other hand, no consensus has been found regarding the interpretation of that same thesis or of its main categories. Although the two contextualist traditions mentioned here are not explicitly opposed to one another, they bear fundamental divergences whose roots lie in the debates between Peter F. Strawson and John L. Austin regarding the place of the concepts of intention and convention as defining speech acts, i.e. as being responsible for delimiting the nature of the speech act performed by a speaker on a given occasion. 3 As these categories have reappeared in contemporary philosophy, they offer competing accounts of the factor responsible for fixing the content of what is said. Thus, the distinction between these two traditions engender important consequences for the understanding of the nature of truth and objectivity. In this text, I set out to explore the divergence between these two contextualist currents, which are often treated as equivalent by their adversaries, especially considering two related problems: the determination and the interpretation of what is said, i.e. problems regarding the possibility of fixing and recognizing the truth conditions put forward by an assertive speech act in a given context. My main focus here is to inquiry the role attributed to the notion of speaker’s meaning in fixing the content of what is said once the strictly semantic reading of that notion is abandoned.

102  Eduardo Marchesan To that end, my strategy will be to describe how the intentionalist strand of contextualism checks the traditional reading of what is said by radicalizing Grice’s precept, where what is said should be read as a variety of what Grice calls non-natural meaning. That is, the precept according to which what is said should be read based on the way the speaker makes it possible to descry the intended content to be expressed. This is a radicalization engendered by the identification of a fundamental tension organizing the Gricean category. Taking François Recanati (1989) as one of the main exponents of intentionalist contextualism, I endeavor to show how the explosion of the abovementioned tension stems from the impossibility of using the implicature model to analyze cases in which modulation phenomena are present, i.e. situations in which the utterance of a sentence alters some of its non-indexical constituents’ literal meanings. For example, when a speaker utters the sentence, ‘all the books are on the table’, with the intention of expressing that a certain set of books is on the table even though that is not what was actually said. One of the consequences of this contextualist dismantling of Grice’s model is the contamination of what is said by the speaker’s intention, which means that pragmatic ­categories – essentially the speaker’s meaning – are responsible for fixing the truth conditions for the statement. Based on the criticisms engendered by the difficulty of setting an ­objective criterion for interpreting the content of what is said if its determination is founded on the speaker’s meaning, I endeavor to show an appeal to the notion of convention in the second contextualist strand. This appeal, which is an attempt to meet the requirements for the objectivity of what is said, ends up denying any possible explanatory role for categories, such as the speaker’s meaning, considering the fixing of the truth conditions of what is said. Such a denial is shown to be based on the irrelevance of psychological matters for the objective determination of the content of what is said – and on a conflation of the problems of fixation and interpretation in the intentionalist strand. Therefore, the priority given to the mentalist notion of intention in that process is checked. The conclusion we get to is the impossibility of treating what is said as a variety of non-natural meaning, which places conventionalist contextualism outside of Grice’s pragmatic tradition.

What Is Said, Intentions and Truth Conditions In its alliance with the central thesis of literalism, Grice’s model states that what is said would not be identical with the meaning of the sentence only in those cases where the sentence contains open constituents that require an act of utterance to fix their reference. In other words, what is said would only be sensitive to the kind of contextual influences that are associated with indexicals. Whenever the sentence is devoid of indexical constituents, what is said would be identifiable with the meanings of

The Role of Intention in Truth  103 the words pronounced in the sentence, which are correctly combined according to the rules of syntax. Therefore, any non-indexical variation in the sentence’s literal meaning would be external to what is said and only captured by Grice’s inferential communication model. Thus, in the example, ‘all the books are on the table’, the quantifier operating on the referential expression, ‘the books’, would produce a constituent which, if taken literally, would mean that all the books in existence were on the table. In strict terms, that would be what was being said when those words were pronounced. In this model, the possible restriction of that content in a real communicative situation such as ‘all the books [that I borrowed from you] are on the table’, is read in the following way: the hearer understands the content of what is said, but realizes that it is unlikely to be what the speaker actually wanted to say during this situation. By confronting what is said with the situation in which the utterance is made, the hearer realizes that, if taken literally, what is said, would violate a conversational maxim and, due to this realization, infers what the speaker effectively wishes to say. In that scenario, sentence use corresponds to the speaker’s m ­ anipulation of meaning in which the spoken sentence would be pitted against a situation. The sentence is verbalized in such a way that its content, i.e. its literal meaning plus the fixing of the values of the indexicals, is assessed not in terms of being true or false but of being plausible or implausible in that situation. Thus, the speaker’s manipulation of the sentence has nothing to do with any manipulation of its content. That is, the speaker is incapable of altering the semantic value of any of the sentence’s constituents. Therefore, the content of what is said is seen as essentially s­ table.4 Should the speaker wish to allude to something more than the meaning of the explicitly pronounced words, then that understanding could only be achieved by a rational effort of expression and comprehension that would explore precisely the inadequacy of the utterance in relation to the context. The communicative intention reveals itself in the form of the deliberate act of mobilizing a sentence that is to a certain extent inadequate, i.e. an act that is equivalent or interpreted as being equivalent to the expression of something the speaker actually had in mind. However, sentences like ‘all the books are on the table’ generate ­insuperable problems for the way the Gricean model explains the expression of what the speaker intends to say in the case of conversational implicatures, i.e. as an inference based on what is said. In examples of that kind, the application of Grice’s model estranges both the speaker and hearer from the mobilized content of speech. On being asked whether he had actually said that all the books in existence were on the table while meaning to say that all the books he had borrowed were on the table, the speaker would reply that it was not the case at all (the hearer would have the same impression). From the speaker’s perspective, given his immediate intuition, he would not acknowledge that he had actually said

104  Eduardo Marchesan anything about all the books in existence. That situation contrasts with a standard example of implicature in which, on being asked whether she knows how to cook, the speaker replies, ‘I’m French!’ In this last situation, those involved would certainly not deny that the speaker had said that she was French while meaning that obviously she knew how to cook, i.e. while intending to pragmatically imply that she knew how to cook. If one accepts a reading based on the traditional Gricean model, then such cases present an essential difficulty in the supposition that the speaker must employ a literal content of which he is unaware, i.e. he would be mobilizing words that he does not recognize as his own with the intention of revealing to the hearer what he means to say. Analyzing Sperber and Wilson’s (1986) example in which the speaker says ‘I have had breakfast’ to express the idea that he had already had breakfast on the morning when he uttered the sentence, Recanati (1989) says: if what the speaker says is that he has had breakfast at least once in his life, then the speaker does not know what he says, because he does not know that this is what he says (were he to be told, he would be very surprised). (p. 313) Taking Grice’s model to its ultimate consequence in this kind of ­example leads to the bizarre situation whereby the communicative intention, as derived from the voluntary utterance of a content in a given situation, would suppose that the speaker had no knowledge of the content in question, which jeopardizes the intentional expression of the implicature itself. Given that the expressive possibilities that extrapolate the merely literal meanings of the words presuppose an intentional act in which the content that is mobilized plays a fundamental role, i.e. if the speaker says X in order to mean Y, then not knowing the uttered content dismantles any possibility of regarding the implicature as being that which the speaker deliberately intended to communicate. Thus, the speaker would be using something that he did not know. The supposed incompatibility of what is said and the situation in which it is uttered ends up being engendered fortuitously. If it exists, the speaker is unaware that he produced it because he does not know what a literalist analysis would conclude that he had said. Something similar occurs with the hearer’s understanding. Grasping the implicature is the result of an operation that begins by recognizing the inadequacy of the content of what is said in a given situation. In this way, the inadequacy in question is that which suggests to the hearer that there must be a reason for the speaker’s use of a given sentence, i.e. the speaker must have uttered those words with the intention of expressing something quite different from what was said. If the hearer is

The Role of Intention in Truth  105 unaware of the content inadequacy, then obviously she cannot base her ­understanding on it and suppose that there must be other reasons motivating the speaker’s speech act. Therefore, her lack of knowledge of what was said would short circuit the operation that would have allowed her to infer what the speaker meant to say. It would not be possible to account for the evident fact that the hearer captures the implicature ­because of the collapse of the explanation for the interpretative activity. The difficulties that arise when analyzing such cases reveal the ­fundamental tension present in the implicatures model, which stems from the two incompatible principles that organize what is said. Grice recognizes what is said as being a variety of what he calls non-natural meaning, i.e. a kind of meaning based on the possibility of the speaker exposing his communicative intentions to his audience, i.e. expressing what he intends to say.5 When uttering a sentence, the speaker can either make it clear to his hearer that what is said corresponds to what is meant or, if a conversational implicature is involved, make it clear that he intends to convey something different from what is said. Viewed in that light, the category is essentially founded on the speaker’s meaning even in those cases where there is no implicature involved. Simultaneously, Grice seeks to reconcile that idea with the literalist principle which, as has been shown in the preceding paragraphs, reduces what is said to the strictly literal meaning of the words that are spoken.6 This principle generates the tension we refer to here: assuming the literality of what is said removes any possibility of treating it as a kind of non-natural meaning because, in the examples analyzed above, what is said would be inconsistent with that which the speaker deliberately wishes to state. The intentionalist version of contextualism arises precisely from the theoretical decision to address that problem by preserving the first principle of recognizing what is said as a variety of non-natural meaning; thus, contextualism abandons the obligation to ensure that what is said is equivalent to the literal meaning of the sentence actually pronounced.7 This decision means acknowledging the primacy of the speaker’s intentions in delimiting the utterance because of the necessity of preserving the intuition that the content accessible to the consciousness of the participants in an interaction must correspond to what is said. The main exponent of that tendency, Recanati (1989) reformulates the notion of what is said having the speaker and the hearer’s spontaneous knowledge as a criterion; in his own words, it is “an appeal to common sense” (p. 310). This observation underlies Recanati’s (1989) availability principle: “In deciding whether a pragmatically determined aspect of utterance meaning is part of what is said, that is, in making a decision concerning what is said, we should always try to preserve our pre-theoretical intuitions on the matter” (p. 310). If on pronouncing the sentence, ‘I have had breakfast’, both speaker and hearer recognize that the intended content is that the speaker had breakfast in the morning of that day, then that

106  Eduardo Marchesan is what is being said. However, if we were to stipulate that in the strict literal meaning of the words, the truth conditions in play are that the speaker has had breakfast on at least one occasion in his life, then up to a point this observation would be a legitimate stipulation.8 However, this stipulation does not alter the fact that if he is unaware of the literal proposition in play, then, as Recanati (1989) points out, “this is not what he says” (p. 313). In this contextualist version, what is said then bears a content that goes beyond the mere meaning of the words, which is precisely what leads to the equivalence between the intuitive interpretation of what is said and the propositional content of the utterance in the case of an assertion. That is, the specification of the scope of the quantifier, e.g. ‘all the books [I borrowed from you]’ is part of the truth conditions of what is said and not just the speaker’s intentional content. What happens here is precisely the fusion of the communicative intention, as formulated in Grice’s model as that which the speaker means to say through ­implicature, with the propositional content.9 If the thesis of contextual influence on the truth conditions of an ­assertion is anchored in equating the intuitive apprehension of what is said with the propositional content, its radical nature is strongly apparent in two of its consequences (the second consequence being the one that most clearly reveals the radical nature of the contextualist strand in question and on which its intentional bias will be founded). Firstly, the corollary of eliminating any kind of inferential process at the level of what is said10 is the exclusion of the idea that literal content needs to be expressed when a sentence is pronounced. As I have endeavored to show, this elimination corresponds to acknowledging that the necessary presence of the sentence’s literal meaning would make any explanation of the expression of what the speaker means to say unviable. If what is said can correspond to a modulated propositional content, such as ‘all the books [I borrowed from you]’, there is no priority whatever attributed to the literal meaning of the sentence. Even if one assumes that the sentence does have a literal meaning that corresponds to a proposition with definite truth conditions, there is no guarantee that the literal value is the actual value in question when the sentence is verbalized in a concrete situation.11 Therefore, it follows that, when a sentence is taken in the abstract, i.e. not associated with a particular situation, it is not possible to determine the truth conditions put forward. The central aspect of this idea is that it is also applicable to sentences that do not contain indexical elements, which leads to the second consequence referred to above. As is well known, the inferential model’s assumption of a ­propositional content variation as a function of the various occasions on which an affirmative sentence is pronounced was intended to acknowledge the ­evident fact that a sentence like ‘I am French’ does not mobilize the same truth conditions when it is uttered from the mouths of different speakers.

The Role of Intention in Truth  107 Indeed, that variation was precisely what justified the formulation of a category like what is said. Expressions present in the sentence require an act of utterance to fix their references and, accordingly, to fix the semantic contribution they make to the proposition; thus, a distinction comes into play between the literal meaning of the sentence as a potential indication of the propositional content and what is said as the level at which it was possible to express a full proposition. However, as Recanati (2004, 2010) points out, even if the constituents in question are sensitive to the context and capable of varying the truth conditions of what is said, this sensitivity is already inscribed in their very meaning in the form of a gap. In other words, given the vacuity typical of their meaning, those expressions necessarily require an act of utterance to fix their references, which eliminates the speaker’s influence considering her stipulation or not of the possibility of contextual variation in her ­expression’s semantic value. Negating the priority of the literal meaning – i.e. acknowledging that in the cases referred to it is not only unnecessary but impossible to s­ uppose that what is said mobilizes the propositional literal content – requires non-indexical constituents to be uttered concretely so that their contribution to the sentence’s truth conditions can be fixed. However, insofar as Recanati assumes that those constituents have a determined semantic value, i.e. a specific contribution to be made to a proposition by their literal meaning, which is one that does not allow for any gap, any variation of its meaning in a given context would be optional. In principle, the composition of the sentence constituents’ literal meanings can generate a complete proposition.12 Even in the case of a sentence like ‘all the books are on the table’, it is possible to imagine circumstances in which the semantic value that ‘all the books’ contributes would be something like ‘all the books that exist’. For example, we could imagine a situation in which almost all the libraries in the world were in ruins and the electronic versions of books had all been destroyed so that the speaker, having brought together the only books that remained, says to his audience that they are on the table. The optional nature of the meaning modulation means that the speaker’s intention comes into play as a principle that determines whether the modulation is present or not. If the modulation is intuitively identified, but we admit (through an analysis) that there is no gap in the constituents in play that would make the contextual contribution obligatory, then there is a certain principle of freedom, so to speak, that allows the speaker to fix the content in question according to his intention. In Recanati’s (2015) words, that makes the said constituents a “semantic entry point for speaker’s meaning”13 (p. 12). Here, “entry point” refers to the existence of possible variations, which are fundamentally open by nature, in the specification of ‘all the books’. That is, there is no preexisting set of senses that the expression can acquire from its linguistic meaning.

108  Eduardo Marchesan For example, the mass term ‘rabbit’ will be preferentially interpreted as meaning rabbit fur in the context of ‘He wears rabbit’ and as meaning rabbit meat in the context of ‘He eats rabbit.’ This [is] not a matter of selecting a particular value in a finite set; with a little imagination, one can think of dozens of possible interpretations for ‘rabbit’ by manipulating the stipulated context of utterance; and there is no limit to the number of interpretations one can imagine in such a way. (Recanati, 2004, p. 24) In these considerations, which were derived, as I have tried to show, from the radicalization of Grice’s model and the preservation of the ­utterance’s non-natural nature, we can clearly see the absolute indetermination that comes into play considering the proposition to be expressed when a sentence is taken in the abstract. This indetermination embodies the contextualist thesis that eliminates the possibility of systematizing the generation of the truth conditions of an affirmative sentence based on a reading of its semantic components as if they were strictly linguistic items. For my purposes, the most relevant aspect of that thesis concerns its positive bias towards recognizing the indetermination of meaning as being a point of entry for the speaker’s intention when performing a speech act. I will discuss this issue in the next section.

Speaker’s Meaning and Contextual Contribution A central issue that imposes itself at this point concerns elucidating the idea of radical indetermination as a point of entry for the speaker’s influence, especially in its connection with another issue, i.e., the contextual aspects that are relevant to the interpretation of the utterance’s content. Once more, this question involves the comparison of the contextual influences that the indexicals are open to with the contextual influences in play in intentionalist contextualism. Considering that David Kaplan’s (1989) distinction between character and content is operating in the case of indexicals, the delimitation of the appropriate contextual element for fixing what is said is codified in the meaning of that type of constituent. The examples are all quite well-known: i.e. given that the first-person pronoun’s reference is fixed by a semantic rule like “the pronoun refers to whoever pronounces the sentence” then the mere utterance of ‘I am French’ automatically generates the appropriate referential content for the expression, which fixes the truth conditions of what is said. The generation of the referential content in this case concerns the fact that the aspect of reality that is to be selected as the content of ‘I’ is perceptible to anyone who knows the language being spoken to the extent that the comprehension of what is said, in this analysis, does not correspond in any way to an interpretive process (at least not from a pragmatic

The Role of Intention in Truth  109 perspective). There can be no interpretation because there is no question that can possibly be asked regarding what the appropriate contextual contribution would be, i.e. there is no room for questions about what the speaker meant to say. In contrast, in the case of ‘all the books are on the table’, the meaning of the constituent ‘all the books’ does not fix the pertinent elements that correspond to the specification of its semantic contribution to what is said merely because of the utterance of the sentence. What books on the table would make the speaker’s assertion true? This kind of indetermination corresponds to the impossibility of linguistic meaning delimiting the aspect of reality that would be pertinent for the truth of a specific utterance.14 This difference in the way of selecting the relevant items for the truth of what is said motivates a distinction between the narrow and wide contexts in the cases of indexicals and other context-sensitive expressions, respectively. This distinction specifically intends to show how, in principle, the wide context is unlimited, i.e. in the case of non-­ indexical context-sensitive expressions, linguistic meaning is incapable of operating as a restricting factor considering reality.15 As Recanati (2001) observed, context-dependence in the case of non-indexical constituents is defined by the fact that “any piece of contextual information may be relevant” (p. 85). Given that in certain situations, it is clear that what is said mobilizes a certain degree of specification – or in some extreme cases, no specification at all, which is a form of delimiting the relevant set of books, i.e. all of them – there must be some extralinguistic factor in play capable of achieving the said delimitation. In this contextualist current, because of the preservation of the Gricean principle that what is said is a variety of non-natural meaning,16 the relevant extralinguistic factor is the speaker’s intention, which gives pride of place to the category of the speaker’s meaning. Recanati (2001, p. 86) observed that: semantic interpretation by itself cannot determine what is said by a sentence containing such an expression: for the semantic value of the expression – its own contribution to what is said – is a matter of speaker’s meaning, and can only be determined by pragmatic interpretation. As set out in these terms, unlike what occurs in the case of indexicals, this idea poses a problem regarding interpretation as shown by Recanati above. The speaker’s speech acts and her intentions delimit the pertinent elements enabling the establishment of the truth conditions of what is said; thus, the question that arises concerns how the hearer will be able to apprehend the speaker’s intended delimitation. In theory, the wider context of the speech act is supposed to operate as the factor that allows the hearer to ascertain the speaker’s intention, since,

110  Eduardo Marchesan as Kent Bach (1997) observed, it corresponds to “anything that the hearer is to take into account to determine (in the sense of ascertain) the speaker’s communicative intentions” (p. 39). As he says, “items of information” (p. 39) are provided contextually, which allows the hearer to grasp the content of what is said. Thus, if he knows that the speaker borrowed books from him, then he most certainly will understand that the books referred to are precisely the borrowed ones. Thus, the context, i.e. the fact that the books have been borrowed, imposes the modulation of ‘all the books’, which constrains the interpretation of what is said. However, given that the wide context was defined by its unlimited character, the problem seems to concern not the idea that there is pertinent information necessary to the interpretation of what is said, but the question of how the hearer can establish which information is relevant for the proper understanding of a given utterance. I will try to develop this problem in the next section.

Pragmatic Interpretation and Psychological Determination The problem mentioned in the last part of the preceding section is r­ elated to one of the issues raised against the contextualist thesis. Given that the content of what is said responds to a principle of open variation, an instability would enter into play and impede any possibility of communication, i.e. of sharing a specific content. That sharing obviously presupposes that the ways of fixing whatever is pertinent for the truth conditions of the utterance are the same in the expression as in the comprehension. However, in view of the hypothesis that the fixing occurs as a function of what the speaker means to say, how is it possible to guarantee that the way the speaker circumscribes the specification of ‘all the books’ will be the same as that of the hearer interpreting his assertion? In the case of Cappelen and Lepore (2005), their attack seems to be formulated first in the form of a difficulty related to the idea that a given variety of knowledge would correspond to aspects of the context that are relevant for determining what is said.17 For example, knowledge related to whoever the speaker is talking to, to what information they have about the speaker, or to what the perceptible aspects of the environment the conversation takes place in are, would be of fundamental importance for a precise interpretation of what is said. In the way ­Cappelen and Lepore (2005) construe the problem, the difficulty concerns the possibility of apprehending the content of an utterance in the passage from one context to another, such as the difficulty someone who had not witnessed the conversation would have in understanding it. However, the same point also seems to be valid for hearers participating in a communicative exchange in the same context. To be able to understand what is said in this case, the hearer must be able to mobilize whatever the speaker knows about her and their shared environment, as well as the cultural references they have

The Role of Intention in Truth  111 in common. Thus, the difficulty seems to be understood in terms of the quantity of knowledge required: If [Radical Contextualism] were true, then for us to understand what you said by an utterance of ‘Philosophy is fun’ we would have to know what knowledge has been triggered by previous conversations you have been engaged in, we have to know whom you are talking to, what you know about them, what knowledge you can assume is shared between you and your audience, the nature of your mutually shared perceptual environment, and so on and so on. That’s a lot of requirements just to figure out what you said. When the full ­[Radical Contextualist] story is told, it will turn out to be a miracle every time anyone manages to figure out what someone had said. But there are no miracles. People do not need to access all of this knowledge in order to figure out what has been said. (Cappelen and Lepore, 2005, p. 124) However, the question that arises in the case of the contextualist ­interpretation of what is said is not so much about the quantity or ­variety of knowledge required, or the possibility of accessing such knowledge on a given occasion. Instead, it is about the knowledge pertinent to understanding what is said. As I mentioned in the previous section, if ‘all the books are on the table’ is a way of saying that all the books that the speaker borrowed from the hearer are on the table, then the supposition that both of them must be aware that certain books were borrowed to ensure a proper understanding of what the speaker meant to say18 seems to be banal (which is also valid for the knowledge that lending and borrowing books is pervasive in the community where they live, or other similar suppositions). The same observation would be valid for a situation in which the speaker says to C that he had told B that all the [borrowed] books were on the table. C would need to be aware of the loan of the books to apprehend the content of what is said. Here, the fundamental aspect for the contextualist thesis is precisely that a specific piece of knowledge and not any other knowledge is relevant for what is being said on a given occasion, i.e. such knowledge guides the proper interpretation of what is said. Indeed, this question seems to have given rise to Cappelen and Lepore’s second criticism when they declare that an even more serious problem overshadows the contextualist thesis: the impossibility of explaining how speaker and hearer manage to converge on the same content. That, at least, is how Recanati (2010, p. 6) translates that criticism: Some people say: if you accept free pragmatic processes in the ­determination of semantic content, semantic content leaps out of control – it is no longer determined by the rules of the language

112  Eduardo Marchesan but varies freely, à la Humpty Dumpty. Communication (content ­sharing) becomes a miracle since there is nothing to ensure that communicators and their addressees will converge on the same content.19 What such a line of questioning reveals appears to be the need for something that guarantees that the content of what is said, i.e. the relevant content or the relevant contextual delimitation, should be the same for both speaker and hearer: Since there are no miracles, and communication happens all the time, there must be something that ensures that communicators and their addressees converge on the same content: it is our job as ­language theorists to find out what that is. (Recanati, 2010, p. 6) In Recanati’s (2004, 2010) contextualist version, that guarantee is a psychological factor. He suggests that the participants not only have the same code in common, but they also have “the same psychological makeup” (p. 7). On more than one occasion, Recanati quotes a passage from Bolinger (1968, p. 230) that encapsulates the fundamental precept: Speakers not only share the same code, but also share the ability to see the same resemblances between what their code already designates and what they would like it to designate, and so to make the old forms reach out to new meanings. There is some similarity between the content mobilized by the code, i.e. what it designates in the case of a referential constituent, and the intended designation in a speech act (in our example, between ‘all the books’ and all the books of a certain pertinent set). Therefore, the main psychological aspect is the shared ability to recognize the similarity in play, i.e. to recognize which books are actually pertinent when ‘all the books’ is pronounced. In a somewhat different way, this problem reappears as the main subject in critiques of the availability principle. If the intuition of the speaker–hearer pair is the criterion for establishing the truth conditions of an assertion, would that not mean that the interlocutors can freely attribute their particular understanding to a given utterance in such a way that, in the case of an incompatibility between what is said and what the two of them think is being said (resulting from a mistake or from incomprehension), there will be no method for deciding which of the two interpretations is pertinent to the utterance?20 In this case, the problem is presented as the risk of a loss of objectivity because the interlocutors’ interpretation of the content of what is said would diverge from what a third observer would identify as having been said. That problematic

The Role of Intention in Truth  113 point could also be formulated in terms of a divergence between the ­interlocutors insofar as it presupposes that in the contextualist proposal, the speaker would be considered as having authority over what she says which, up to a point, would be insusceptible to outside assessment. ­Accordingly, we could suppose that in situations in which the speaker and the hearer diverged regarding their interpretations of what was said, the speaker could avoid admitting that she had said something. That is, if her hearers affirmed that the content she had said was X, it would be possible for her to deny that and declare that actually what she had wanted to say was Y. 21 Recanati (2004) offers a similar retort to this criticism that he takes “what is said [to be equal] to [the interlocutors] understanding of what is said” (p. 19) by vehemently denying that any such equating is involved in his formulations. Instead, he insists, the identity established in his theory is one between the content of the utterance and that which a “normal interpreter” (Recanati, 2004, p. 19) would understand from what was said in a given context. Here, a “normal interpreter” is taken to be precisely someone who “knows which sentence was uttered, knows the meaning of that sentence, knows the relevant contextual facts” (Recanati, 2004, p. 19). As Recanati (2004) sees it, “Ordinary users of the language are normal interpreters, in most situations. They know the relevant facts and have the relevant abilities” (p. 19). Thus, normal interpreters are characterized in terms of their knowledge and skills, which would correspond somewhat to their psychological makeup. Accordingly, what is said does not vary according to the understanding of individual subjects, but according to the understanding of a particular type of individual. By belonging to a certain psychological type, an individual can recognize the similarity between the content of a word and the intended content, i.e. the content that the speaker intends to publicize. To put this observation in terms of contextual contribution, it is due to his psychology that the hearer not only knows the relevant contextual information for fixing the value of the modulated constituent, but, more importantly, is able to know which information is relevant for interpreting what is been said. Thus, both criticisms concerning the convergence to the same content and the objectivity of such content are dealt with in the same way, i.e. by an appeal to a shared psychological constitution. Recanati seems to be saying that if you are a normal interpreter, then your intuition is equivalent to what is said because it works as the relevant criteria for objectively determining what is said. Additionally, a normal interpreter shares the same psychological makeup as their interlocutor, which allows them to share the same intuitions. Thus, the interpreter knows which piece of contextual information is relevant for interpreting what was said. Therefore, fixation and recognition are equated in this version: the speaker can fix the content of what is said in the same way as the hearer ­understands it because of their shared psychology.

114  Eduardo Marchesan However, if we consider a situation in which the speaker is not a ­ ormal interpreter, i.e. she does not know or cannot know the relevant n contextual information, what happens to the content of what is said? If she does not know the relevant information, then her intuition cannot be the criteria for establishing the truth conditions of the utterance, which means that her communicative intentions cannot fix a content for what is said. Thus, the speaker’s meaning would be ineffective. Would there still be a content for what had been said? Or, to be more precise, would there still be a content that we – as normal interpreters – could recognize as that which was objectively being said? When considering the problem of objectivity, Recanati (2004, p. 20) says: There are situations … where the actual users make mistakes and are not normal interpreters. In such situations their interpretations do not fix what is said. To determine what is said, we need to look at the interpretation that a normal interpreter would give. In a situation like that, there seems to be something to be recognized as the right content, but it would be independent of what the speaker intended to express. Thus, the speaker’s intentions could not be the factor responsible for fixing it. Simultaneously, if the right hearer is the one who can vouch for the objective content in play, it would be strange to take him as fixing it, i.e. determining the content of someone else’s utterance. Therefore, in a situation like that, even if the right recognition is a matter of psychological makeup, fixation is not, because both speaker and hearer are not intentionally responsible for it. An example of that kind would show that the problems of determining the truth conditions of a sentence and recognizing such a determination are fundamentally distinct in the sense that they are related to different extralinguistic factors. Consequently, the principle according to which the speaker’s meaning is one of the main categories in the process of fixing the truth conditions of an assertion would be in peril.

The Business of Being True Let us examine an example proposed by Travis (2008). 22 One of the assistants in a pizza restaurant heats up the oven in the kitchen while preparing to bake the pizzas. At a certain moment, he goes up to the chef who is waiting to bake the pizzas and declares, ‘the oven is hot’. The oven needs to be at the right temperature for the pizzas to bake properly. Any temperature lower than 220° would have disastrous consequences for the pizzas. Let us imagine that the assistant is inexperienced and states that the oven is hot when it is at a temperature of only 140°. In this case, he would be saying something false even though the oven is hot enough to burn someone’s hand if they were to inadvertently touch

The Role of Intention in Truth  115 it. Here, just as in Recanati’s examples, the meaning of the words that compose the sentence ‘the oven is hot’ as taken in the abstract outside of a specific context, is not capable of circumscribing a specific mode in which a certain object, namely, an oven, should be for the sentence to be true. Dozens of other situations requiring different conditions for the utterance to be true could be imagined: e.g. a 140° oven could be hot enough to prepare other dishes, which would make the sentence ‘the oven is hot’ true in these cases. The oven could be switched off, but in an extreme situation, such as a power cut during a blizzard, someone might suggest that the structure of the oven made it a good place in which to shelter delicate electronic equipment, which would mean that even though the oven was not switched on, the utterance of ‘the oven is hot’ could be true. Like Recanati’s cases, this example suggests an inquiry into what could fix the appropriate truth conditions for an utterance of ‘the oven is hot’. Here too, given the variation identified regarding the utterance’s possible content, the objectivity of the truth conditions used in this occasion becomes questionable. What would be the ramifications of a psychological-type answer to such a question in the case of the pizza restaurant assistant’s sentence? If the question formulated is treated as being about the possibility of the speaker–hearer’s convergence regarding the content of what is said, then the supposition that they share the same psychological makeup would indeed facilitate the idea that they both share not only the necessary knowledge about making pizzas, e.g. the right oven temperature, materials that should be used to heat it, how long the pizzas should remain inside the oven, or the frequency with which they need to be rotated, but, more importantly, also a series of presuppositions that sustain those items of knowledge, such as what constitutes a properly baked pizza, or what constitutes properly baked dough, or which kinds of food are normally eaten raw and which are not. Recognizing that what is said by the utterance ‘the oven is hot’ in the context imagined here is, or should be, that the temperature of the oven is 220° presupposes a psychological configuration that is not limited to a mere set of explicit knowledge items but is also based on a series of shared practices implicitly presupposed in a communicative exchange. Someone who lived in a community that was not in the habit of baking dough, e.g. one that did not make dough or was in the habit of eating it unbaked, would either ignore what the assistant said or at least find it difficult to understand that his words could mean that the oven was at a temperature of 220°. However, even if a shared psychological makeup is the necessary condition for the existence of speaker–hearer convergence on the same content, the supposition of that sharing is not sufficient to say what the content is towards which the two of them should converge. Let us suppose, as Travis (2008) does, that at the very same moment, in the same

116  Eduardo Marchesan pizza restaurant the same assistant says to the same chef ‘the oven is hot’, but this time with the intention of warning him not to touch the oven surface. Or yet, we could perhaps suppose that the sloppy chef was in the habit of drying some of his clothes by stretching them out on the pizza oven walls when they were not so hot as to burn them, perhaps at somewhere around 140°. The interlocutors’ psychological makeup is of fundamental importance in enabling the recognition of what is being said in each of those cases, namely that the oven temperature is 140°, but it does not offer the means to fix the particular condition the oven should be in. The psychology of the hearers presupposes such fixing, which is precisely what there is to be recognized. Therefore, even if an appeal to the psychological dimension is an effective way of explaining the possibility of a hearer–speaker convergence existing regarding the content of what is said, it seems that it cannot ameliorate anxieties associated to the doubt regarding the objectivity of the said content. In other words, it cannot explain how the relevant contextual content towards which the two converge regarding the predicate ‘is hot’ is 140°. The distinction between the two questions is captured by Travis (2018a, 2018b) when he says that there is that which is “in the business of being true” – a thought (gedanke) in Frege’s terms, i.e. “that which makes truth turn on how things are” (2018a, p. 3) – and that which guarantees an entry point into such business – the uttered words. In Travis’s contextualism, the words someone speaks are, just as Frege considered them, the sensitive element that enables the recognition of what is being said. Treating this recognition as a psychological matter considers that given understandings are not available to everyone, but only to those who share a certain framework, or certain forms of life, as Wittgenstein described it. However, the crucial consideration is the fact that the psychological configuration of someone who is excluded from a certain form of life to the point of not being able to recognize the content of what is said by those that are within it, or finding it extremely difficult to do so, would not have any influence in determining the content of what is said even in situations where what is in question is what the excluded speaker said. Let us imagine a typical scene in a movie from the “fish out of water” genre, in which a person coming from an isolated community suddenly finds himself having to live in a cosmopolitan city like New York. ­A fter obtaining a job in a pizza restaurant, that person would experience basic difficulties because, let us suppose, his community was not in the habit of baking dough. However, they did eat food made from dough, such as bread or pizza, but only slightly baked, i.e. they consumed it in a state that generally would be considered uncooked. On his first day at work, the displaced individual is ordered to warn the chef once the oven is hot enough for the pizzas to be baked. Given that an oven hot enough to bake dough in his homeland is an oven at 140° (which he is perfectly

The Role of Intention in Truth  117 aware of because his old job in the isolated community was to cook these unbaked doughs), when the assistant sees that the oven is at this precise temperature, he says ‘the oven is hot’. The chef then introduces the pizzas into an oven that is not hot enough to bake them. Obviously, the results are very bad. From the perspective of what the assistant intended to say, i.e. of the belief he intended to express, his utterance that ‘the oven is hot’ sought to express the idea that the temperature of the oven was 140°. That was the idea he believed he had expressed. Despite the assistant’s expectation, however, the chef understood his sentence to mean that the oven was at 220°. Admittedly, one cannot consider the assistant to be a ‘normal speaker’ in this context, which would explain the divergence of the perceptions in play. However, what is relevant here concerns the fact that any person engaged in the task of baking a pizza would have understood that the assistant’s utterance meant that the oven was at a temperature of around 220°. Considering that pizza dough requires that temperature to become what we consider to be baked properly, there is no other possibility for what ‘is hot’ means when pronounced by someone undertaking the action of baking a pizza. In other words, someone involved in the process of making pizzas cannot comprehend any propositional content other than “the oven is at 220°” when making this statement. As Stanley Cavell (1976) observed in his classic essay, the assistant must mean that the oven is at 220°. What the assistant believes himself to be expressing and the fact that he is intuitively prepared to recognize something different from the pizza chefs when he pronounces his sentences are relevant elements for explaining the misunderstanding. This explains the non-convergence of the two hearers considering the content in play, but fails to have any effect on the objective evidence that the content of what is said on the described occasion corresponds to ‘the oven is at 220°’. If the pizzas placed in the oven were spoiled or if there was a delay caused in the chef’s pizza delivery, the cause of these difficulties is what the assistant said. Therefore, he would be responsible for the consequences stemming from the content of his utterance even if he was unaware of that content at the moment of its utterance. 23 In those terms, Travis’s formulations recover the centrality of the notion of convention as described by Austin (1962) when he discussed the felicity conditions necessary for an effective performance of a speech act. In Travis’s reading of Austin’s (1962) notion of convention, it operates precisely as an objectivity-fostering factor with respect to the contents of what is said. Given that making pizzas presupposes baking dough, which requires an oven at a temperature of at least 220°, then saying that an oven is hot during the pizza-preparation process is necessarily the same as saying it is at 220°. However, as the example shows, the notion of convention is not understood merely in terms of explicit ritualized rules, i.e. in terms of strictly delimited procedures to be followed,

118  Eduardo Marchesan such as those performed to baptize a ship or to declare two persons to be joined in matrimony. Here the conventions broadly refer to whatever is necessary for a specific action to be performed, i.e. so that a certain action can be recognized as such, or, so that a given description can be attributed to what a speaker does. Treating the conventions defined above as if they were factors ­fostering objectification of the utterance necessarily excludes any possibility of ­attributing relevance to a category, such as the speaker’s meaning, at least in its aspect as a category, which corresponds to the strong temptation to suppose that a speaker may fix the standards by which the accuracy of his words is to be judged merely by intending that they express ‘the proposition that (such and such),’ or by making this evident enough. (Travis, 2008, p. 105) Thus, as what there is to be said, a thought is that which is expressed by the words uttered by someone (irrespective of his psychology) who occupies a given position on a given occasion, such as a kitchen assistant heating the oven to bake pizza. In this sense, there is nothing that would justify the appeal to a concept such as speaker’s meaning: “In the normal case, I suggest, the purposes someone’s words should serve to be true are just the purposes the words that person used should serve, in the circumstances in which she spoke, to be true” (Travis, 2008, p. 105). The coincidence between the content of what a person says in a specific circumstance and the content of the words that person uses in the said circumstance is, as I see it, the reason for the use of the term “purpose”, which establishes an important contrast with the notion of intention as a mechanism for expressing mental contents. The speaker’s purpose is also the purpose of the action performed so that, as the example of the pizza shows, even if the assistant intends to express a belief such as “the oven is at a temperature of 140°” by saying ‘The oven is hot’, given that the purposes of the action he is engaged in involve making a pizza, his belief is irrelevant to the truth conditions set into play by what is said. Thus, this conventionalist version of contextualism seems to ­overthrow Grice’s principle whereby what is said operates as a kind of non-natural meaning, i.e. operates in terms of letting the hearer grasp the intention of the speaker. As I have endeavored to demonstrate, preserving the principle according to which what is said represents a variety of non-natural meaning leads to a conflation of the questions of fixation and understanding, making them both seem to depend on the psychological makeup of the interlocutors. Disentangling them generates a position concerning the ‘nature’ of truth that sees it as independent of psychological considerations and reveals a fundamental distinction between these two kinds of radical contextualism.

The Role of Intention in Truth  119

Notes 1 This text was written with the support of the Fundação de Amparo à Pesquisa do Estado de São Paulo (FAPESP), who awarded me a two-year post-doctoral scholarship. 2 These contextualist tendencies are united by the negative thesis they share, i.e. the meaning of the words is not sufficient to fix the truth conditions of a sentence outside a given context. 3 As is well known, the classical loci of those debates are Austin (1979a, 1979b) and Strawson (1949, 1964). 4 Provided that there are no actual ambiguities present or that they have been duly resolved. 5 As is well known, a further step is involved in the expression of a ­communicative intention: the possibility of the hearer realizing that the speaker wants his intention to be recognized. Jocelyn Benoist (2009) explored the possibility of an infinite regression as suggested by Grice’s formulation, while Recanati (1986) performed a rereading of the notion of ­communicative intention in an endeavor to get around the associated difficulties. 6 Grice’s reason for maintaining the literalist principle was to negate the thesis of the philosophy of ordinary language that a sentence only acquires content in the context of a speech act, i.e. that which Grice (1989) referred to as the conflation of meaning and use (see his “Prolegomena”). 7 To some extent, the acknowledgement of tension in Grice’s model organizes the field by considering the assessment of the schism between semantics and pragmatics insofar as this acknowledgement is present throughout the extension of the continuum of available stances. Even literalist stances presume the existence of this incompatibility and endeavor to resolve it by stipulating that the truth conditions of what is said remain, stricto sensu, those associated to the literal meaning of the spoken sentence. However, this supposes that there is a difference between the literal meaning and the so-called intuitive truth conditions for spoken sentences. 8 This legitimate stipulation leads to Emma Borg’s (2012) perspective, i.e. the study of the intuitive truth conditions and truth conditions stricto sensu would occur in parallel. In those terms, literalism would not clash with contextualism, or at least not with the contextualism that accepts the possibility of a given propositional content generated based on the composition of the literal meaning of the constituents of a sentence. However, this position is not shared by all contextualists (see note 12 in this text). 9 Considered as a set of phenomena comprising the distortion of the literal meaning and capable of influencing the truth conditions of the utterance, modulation is equated with the speaker’s meaning. Presenting the position against which he is arguing, Recanati (2010) says: “Insofar as modulation  – hence speaker’s meaning – enters into the determination of satisfaction conditions, it is not incumbent upon semantics to account for satisfaction conditions and the content of speech acts more generally” (p. 43). 10 As Recanati sees it, the criticism that destroys the possibility of treating cases of modulation as implicatures does not imply negating the possible existence of other cases that could be treated in terms of inferred implicatures. Even if the content of what is said is equated to the intuitions of the interlocutors – thus being sensitive to the speaker’s intention – it can nonetheless be used as a way of producing an implicature, i.e. its modulated truth conditions might be considered to flout a conversational maxim on a given occasion. Such cases would suggest a sort of duplication of the notion

120  Eduardo Marchesan

11

12

13

14

of intention in Recanati’s model, given that both the content of what is said and the implicated content are founded on the intention of the speaker. The difficulty here would seem to be how to account for that distinction since the hearer apparently wishes to express the content of what she said and she wishes to express the implicated content. In that case, would she mean both contents? In Grice’s model, this duplication did not take place: when what is said generated an implicature, the speaker meant the implicature, but not what she said (what she said worked merely as a means for getting to the desired content). This is what Recanati considers the nonpriority of the literal ­interpretation. However, even though he negates that priority at the level of the whole ­proposition, he does not do so at the level of each of the sentence’s constituents. In his view, the subsentential constituents do have a literal semantic value, but one that is not necessarily the contribution made to the full proposition expressed. Recanati (1995, 2004, 2010) sees that level as being precisely the one in which operates the modulation phenomena, i.e. a subpersonal level at which the speaker is unaware of the literal meaning of the constituents. In this way, modulated constituents could be part of the composition of the meanings that form a proposition, which makes it unnecessary to suppose the priority of a full literal proposition as an explanation for the comprehension of the intended content. Thus, this version of contextualism accepts the possibility of supposing that an utterance whose constituents contributed their literal meanings to the propositional content is capable of putting forward a precise set of truth conditions. On the one hand, this idea establishes a distance between this kind of contextualism and the perspectives of other authors like Charles Travis, for whom the possibility of supposing that a combination merely based on linguistic significances could generate a content susceptible to a ‘true or false’ evaluation is impeded. On the other hand, those peculiar cases clearly reveal how the speaker’s intention in Recanati’s proposal is the fundamental basis for fixing the content of the utterance even if we consider that the semantic values in play in an assertion are those codified in the respective linguistic system. In those cases, according to Recanati (2010), a mod function, defined as the modulated meaning to be contributed to the proposition (and equated with the speaker’s meaning), operates on the literal meaning of the constituents with an identity value. Indexicalist perspectives resist entry points to the speaker’s meaning insofar as it strips any possibility of systematically determining the truth conditions of the utterance. Those perspectives endeavor to restore Grice’s category of what is said by identifying an indexical-type gap in the modulated constituents. Supposing the existence of that gap means recognizing, on the one hand, the contextual sensitivity of such expressions but, on the other hand, it means endeavoring to recover the linguistic rules’ control regarding the pertinent contextual delimitation. Maintaining the compulsory aspect in fixing the meaning of a context-sensitive constituent is what led Borg (2012) to descry an alliance between literalism and indexical perspectives: even though literalism intends to avoid the multiplication of gappy elements as part of an abstract sentence, recognizing their existence does not constitute a risk for the thesis of the fundamental determination of the semantic rules or for the compulsory nature of their application. Here, I would insist that the literal meaning of the expression, “all existing books”, is just one of the possible contributions to be made to what is said and has no prevalence over other possible senses.

The Role of Intention in Truth  121 15 It is worth noting that the inability of meaning to delimit the pertinent ­element for the truth of what is said does not constitute an absolute denial that there is any restriction in play as some of contextualism’s critics, like Stanley (2007) would have it. Obviously, the meaning of ‘all the books’ restricts the relevant referents to certain books. The point is that the restriction is insufficient and it leaves all sorts of different possibilities open as to which are the relevant books. In that light the principle of freedom referred to previously is merely relative. 16 “The availability of what is said follows from Grice’s idea that saying itself is a variety of non-natural meaning… Non-natural meaning works by openly letting the addressee recognize one’s primary intention (for example, the intention to impart a certain piece of information, or the intention to have the addressee behave in a certain way), that is, by (openly) expressing that intention so as to make it graspable” (Recanati, 2004, p. 13). 17 Cappelen and Lepore (2005) present the following list attributed to ­B ezuidenhout (2002, p. 117), in which she sets out the contextual features that can determine what is said: i

Knowledge that has already been activated from the prior discourse context (if any). ii Knowledge that is available based on who one’s conversational partner is and on what community memberships one shares with that person. iii Knowledge that is available through observation of the mutual ­perceptual environment. iv Any stereotypical knowledge or scripts or frames that are associatively ­triggered by accessing the semantic potential of any of the expressions currently being used. v Knowledge of the purposes and abilities of one’s conversational partner (e.g. whether the person is being deceitful or sincere, whether the person tends to verbosity or is a person of few words, etc.). vi Knowledge one has of the general principles governing conversational ­exchanges (perhaps including Grice’s conversational maxims, culturally ­specific norms of politeness, etc.). 18 A proper understanding of the speaker is, as Recanati suggests, a question present even in those analyses that suppose that access to what the speaker meant to say is inferential. 19 In this case, the problem of the impossibility of convergence seems to presuppose a prior problem related to determination: it is not just a question about whether the content will be or could be apprehended, but about the doubt as to whether there could be something recognizable as to the content of what is said (I will come back to this issue). The problem involves the risk of skepticism based on an apparent loss of objectivity on the part of the propositional content in question, i.e. a risk that motivates the literalist insistence on equating what is said with the meaning of the sentence taken as a factor of content objectification. 20 In this example, a model would be the case when the speaker means to say that Tim is tall, but instead she points to Paul and says “He is tall”. Also mistaking Paul for Tim, the hearer believes that the speaker said that Tim is tall. The tension stems from the fact the speaker objectively said “Paul is tall”, but both she and the hearer believe that what was said was “Tim is tall”. 21 This is a risk inherent to mentalist perspectives as Stanley Cavell (1995) shows in his reading of Austin’s (1962) citation of Hippolytus in How to Do Things with Words. Having made a promise with no intention of keeping

122  Eduardo Marchesan it someone could say “My tongue swore to, but my heart (or mind or other backstage artiste) did not” (Austin, 1962, p. 9), which supposedly would free the speaker from the commitment made in the promise. That being so, an irremovable doubt hangs over all promises, indeed, it is inherent to the act of promising, i.e. not whether the promise will be kept, but whether the promise has indeed been made. See Moati (2011) for a presentation of ­Austin and Cavell’s position against mentalism. 22 Here, our focus will be on an example centered on predication rather than on referential constituents. Given the contextualist’s intent to generalize their thesis to embrace all sentence constituents, any eventual difference in the treatment of those two constituent types will not be relevant here. For a discussion of predication, see Recanati (2015). 3 Obviously, the blame for the poorly cooked pizzas could be attenuated 2 according to the actual circumstances. For example, if the pizza chef had been in charge of the assistant’s training and had not considered the relevant cultural differences, then the chef shares the blame for the pizzas with the assistant or takes full responsibility. Those considerations drive Austin’s (1979c) discussion regarding excuses and the centrality attributed to them by contemporary readers, such as Benoist (2011) and Sandra Laugier (2004), who emphasized the actantial aspect of the speech act as being fundamental to the debate on fixing the truth conditions of what is said.

Bibliography Austin, John Langshaw. 1962. How to Do Things with Words. Cambridge: Harvard University Press. ———. 1979a. “Truth.” In Philosophical Papers, edited by James Opie Urson and Geoffrey Warnock, 117–33. Oxford: Clarendon Press. ———.1979b. “Unfair to Facts.” In Philosophical Papers, edited by James Opie Urson and Geoffrey Warnock, 154–74. Oxford: Clarendon Press. ———. 1979c. “A Plea for Excuses.” In Philosophical Papers, edited by James Opie Urson and Geoffrey Warnock, 175–204. Oxford: Clarendon Press. Bach, Kent. 1997. “The Semantics-Pragmatics Distinction: What It Is and Why It Matters.” In Pragmatik. Linguistische Berichte (Forschung ­Information Diskussion), edited by E. Rolf, 33–50. Wiesbaden: VS Verlag für Sozialwissenschaften. Benoist, Jocelyn. 2009. “Grice sous la lune.” In Sens et sensibilité, edited by Jocelyn Benoist, 179–99. Paris: Cerf. ———. 2011. “Des intentions finies.” In John Austin et la philosophie du ­langage ordinaire, edited by Sandra Laugier and Christophe Al-Saleh, 269–85. Hildesheim: Olms. Bezuidenhout, Anne. 2002. “Truth–Conditional Pragmatics.” Noûs 36 (s16): 105–134. Bolinger, Dwight. 1968. Aspects of Language. San Diego, CA: Harcourt, Brace and World. Borg, Emma. 2012. Pursuing Meaning. Oxford: Oxford University Press. Cappelen, Herman, and Ernie Lepore. 2005. Insensitive Semantics. Oxford: Blackwell. Cavell, Stanley. 1976. “Must We Mean What We Say?” In Must We Mean What We Say?, edited by Stanley Cavell, 1–43. Cambridge: Cambridge University Press.

The Role of Intention in Truth  123 ———. 1995. “What Did Derrida Want of Austin?” In Philosophical Passages: Wittgenstein, Emerson, Austin, Derrida, edited by Stanley Cavell, 66–90. Cambridge: Blackwell. Grice, Paul. 1989. Studies in the Way of Words. Cambridge, MA: Harvard University Press. Kaplan, David. 1989. “Demonstratives.” In Themes from Kaplan, edited by Joseph Almog, Howard Wettstein, and John Perry, 481–563. Oxford: Oxford University Press. Laugier, Sandra. 2004. “Acte de langage ou pragmatique?” Revue de ­métaphysique et de morale 2 (42): 279–303. Moati, Raoul. 2011. “Langage ordinaire et vouloir dire: Stanley Cavell d ­ éfenseur critique d’Austin.” In La philosophie du langage ordinaire: histoire et actualité de la philosophie d’Oxford, edited by Christophe Al-Saleh and Sandra Laugier, 469–92. Hildesheim: Olms. Recanati, François. 1986. “On Defining Communicative Intentions.” Mind & Language 1 (3): 213–41. ———.1989. “The Pragmatics of What is Said.” Mind & Language 4 (4): 295–329. ———. 1993. Direct Reference: From Language to Thought. Oxford: Blackwell. ———. 1995. “The Alleged Priority of Literal Interpretation.” Cognitive ­Science 19 (2): 207–32. ———. 2001. “What is Said.” Synthese 128 (1–2): 75–91. ———. 2004. Literal Meaning. Cambridge: Cambridge University Press. ———. 2010. Truth–Conditional Pragmatics. New York: Oxford University Press. ———. 2015. “From Meaning to Content: Issues in Meta-Semantics.” Accessed January 12, 2018. www.academia.edu/22576845/From_Meaning_to_ Content_Issues_in_Meta-Semantics Sperber, Dan, and Deidre Wilson. 1986. Relevance: Communication and ­C ognition. Oxford: Blackwell. Stanley, Jason. 2007. Language in Context: Selected Essays. Oxford: Oxford University Press. Strawson, Peter Frederick. 1949. “Truth.” Analysis 9 (6): 83–97. ———. 1964. “Intention and Convention in Speech Acts.” The Philosophical Review 73 (4) 439–60. Travis, Charles. 2008. “Meaning’s Role in Truth.” In Occasion-Sensitivity: ­Selected Essays, edited by Charles Travis, 94–108. New York: Oxford. ———. 2018a. “What Words are For.” Accessed June 13, 2015. http://kcl­.­ academia.edu/CharlesTravis/Papers. ———. 2018b. “Their Work and Why They Do It.” Accessed May 29, 2018. www. academia.edu/34764341/THEIR_WORK_AND_WHY_THEY_DO_IT1

7 Is Seeing Judging? Radical Contextualism and the Problem of Perception Sofia Miguens

Radical Contextualism and the Problem of Perception Charles Travis’ basic claim regarding perception is that perception does not have representational content: the senses are, in the Austinian metaphor he favors, silent.1 More often than not this claim is considered within the framework of ongoing discussions in the philosophy of perception, where one central question is whether perception has representational content. 2 Yet pursuing the discussion in terms of the representational content of perception risks begging the question of the nature of representation in general, which is one central question that Travis’ radical contextualism addresses. Also, restricting the discussion of the proposal to the field of philosophy of perception makes one disregard the fact that the silence of the senses view of perception comes off as an application of a general radical contextualist outlook. Radical contextualism is an approach to a range of questions regarding truth, thought and language, and a case built in several fronts. Globally it amounts to a view of representation as it is pursued by thinkers such as ourselves, a view which is currently formulated by Travis around a particular reading of Frege.3 Such a view of thinkers and how they represent is the background against which the silence of the senses view of perception should be understood. My main goal in this article is not to discuss Travis’ radical ­contextualism in all its ramifications but rather to shed light on his views on perception by setting them against a background of recent discussions regarding perception and representation. I want to show how the silence of the senses thesis flows from – or at least is dependent upon – a position regarding the subpersonal/personal distinction. In order to do that I will take the supposedly less charged question of whether ­seeing is judging, which I borrow from D. Dennett,4 as my leading thread. I will thus go through two recent debates on perception and representation in the philosophy of perception – the first one between Tyler Burge and John McDowell and the second between John ­McDowell and Charles Travis himself. 5 One immediate purpose is to

Is Seeing Judging? Radical Contextualism  125 show that there are radically divergent understandings of ­representation at play in the field. Suffice it to consider that it is quite frequent, and basically non polemical, within cognitive science, to speak of perceivers’s subpersonal states, as well as of perceptual experiences, as representations, whereas from a philosophical viewpoint the question remains open as to whether such subpersonal states and experiences do indeed represent things as being a certain way, namely veridically or non-veridically. Many philosophers are more than happy to simply pick up from cognitive science the use of the term ‘representations’. Yet, as I hope a picture of the two debates will help bring forth, this practice helps hide questions the nature of representation poses. The first debate I am interested in was a debate over perceptual psychology and disjunctivism and took place between a representationalist (Tyler Burge)6 and a disjunctivist (John McDowell). It concerns mostly what I will call Question 1: should one say of a perceivers’s subpersonal states that they are representations? The reference to D. Dennett’s proposal of a personal/subpersonal distinction will be introduced in this context. The second debate took place between two disjunctivists (John McDowell and Charles Travis) and was on the ‘Myth of the Given’.7 It concerns mostly what I will call Question 2: should one say of perceivers’ perceptual experiences, as opposed to his (or her) judgements, that they represent things as being a certain way? Both debates will help me bring forth questions regarding the nature of representation which I believe matter for understanding radical contextualism’s bearing on the problem of perception. As I analyze the two debates, I come across discussions concerning ­so-called disjunctivism and try to uncover the participants’ motivations for defending (or attacking) it. Still other aspects are relevant thoughout the debates for characterizing the background against which Travis’ radical contextualist approach to perception should be read. In the case of the first debate I am interested in making clear what philosophy of perception is not for people such as McDowell and Travis, in contrast with people like Burge. Namely, philosophy of perception is not cognitive science of perception. This should be apparent in McDowell’s opposition to Burge, which I believe would be fully endorsed by Travis. In the case of the second debate I want to trace the crucial difference between ­McDowell’s and Travis’ respective approaches to perception to contrasting conceptions of judgement, a Kantian conception in the first case and a Fregean conception in the second. This is the final step for identifying the answer the radical contextualist gives to my probe-question, the question whether seeing is judging. A Fregean conception of judgement helps Travis spell out the implications, for thinkers’ dealings with the world, of the ‘silence of the senses’ view of perception and leads him to claim that seeing per se is not judging.

126  Sofia Miguens

Perception: Two Controversies Centering on the Notion of Representation The Tyler Burge-John McDowell Controversy The controversy between John McDowell and Tyler Burge over perceptual psychology and disjunctivism was prompted by a 2005 Philosophical Papers article intitled “Disjunctivism and perceptual psychology”.8 In the article Burge accuses a number of disjunctivists (Paul Snowdon, John McDowell, Gareth Evans, Mike Martin and John Campbell) of defending a view that is incompatible with what is known from empirical science of perception,9 scolds disjunctivists for failing to understand fallibility and ultimately accuses them of giving expression to a parochially British concern with an outmoded problem, that of overcoming an empiricist veil-of-ideas skepticism. His final verdict, as restated in 2011, is that Disjunctivism ignores science in specifying ordinary psychological kinds and in doing epistemology. It is a doctrinal and methodological aberration. Philosophical progress will continue to pass it by.10 Of all the disjunctivists mentioned by Burge here I will consider John McDowell. I want to bring out his particular motivation for being a disjunctivist as he responds to Burge. Needless to say, getting clear about what one means by ‘disjunctivism’ is in itself a problem – after all, positions range from considering it ‘mandatory’ (which McDowell does) to seeing it as an ‘aberration’, like Burge does. As Burge defines disjunctivism, and he is extremely careful in doing it, since he believes the philosophers he criticizes are not, disjunctivism is an idea (a quite bad one) about psychological explanation and type-sorting of mental states: Disjunctivism claims that there is never an explanatorily relevant mental state type in common between (and specific to) a veridical perception and a referential perceptual illusion.11 Such denial of a common type (a ‘common factor’) clashes with the assumptions of what Burge calls ‘the science’, i.e. the psychology of vision: The idea that perceptual states causally depend only on proximal stimulations, internal input, and antecedent psychological conditions is basic to the method of the science (…) So the methodology of all serious empirical theory of vision guarantees that given types of visual state can be veridical in some circumstances and non-­veridical in others.12

Is Seeing Judging? Radical Contextualism  127 In other words, according to Burge psychological explanation simply requires commitment to mental states and types of mental states, and these are typed in terms of representational content.13 Of course representational content is ‘an abstraction that helps type psychological kinds’.14 As for representations, for Burge they are tokened, structured, instances of, or vehicles, like inner symbols, expressing representational contents.15 Anyway perception is, according to Burge, representational in that it purports to be about something and to represent it as being a certain way.16 In this it contrasts with what he calls sensing. It is clear that Burge’s view, and criticism, of disjunctivism is tied to his view of psychological explanation. It is as a claim concerning explanatorily relevant mental states that disjunctivism is ‘an aberration’. Yet when McDowell defends disjunctivism, he defends it as, in its most simple guise, an idea about appearances. This is an idea expressed by a disjunction and an idea which bears crucially on epistemology. The disjunction is the following: an appearance is either a case of things being thus and so in a way that is manifest to the subject or a case of its merely seeming to the subject that that is how things are.17 Clearly, there is lack of agreement on what disjunctivism is. In p ­ articular, does it concern appearances to the subject or inner vehicles of representation? Much in the analyses that follow will concern this. One ­cautionary note before I proceed: one problem of taking McDowell as an example of a disjunctivist as I will do here, is that there are many strands in his current position on perception. Some of these concern interpretations of Kant, Hegel or Sellars, and thus issues which are not of central interest to many of the participants in the mainstream discussions in the philosophy of perception.18 Also his formulation of disjunctivism involves not only perceptual experience but also (and maybe primarily), knowledge. So if one takes it that the central claim of disjunctivism is about the nature of perceptual experience, McDowell is not exactly the paradigm example of a disjunctivist.19 I will concentrate on two points for now, which are closer to more general worries of the discussion in the philosophy of perception than the history of philosophy disputes McDowell is also involved in: the methodological distinction between epistemology of perception and cognitive science of perception and the representationalist (or conceptualist) view of perceptual experience. 20 It is the first, I believe, that is at the heart of the clash between McDowell and Burge. The second element (the idea that perceptual experiences have representational content) is something Burge simply takes for granted and agrees with, overlooking the fact that McDowell has withdrawn earlier stronger claims (in response to Travis, 21 see McDowell 2009a, Avoinding the Myth of the

128  Sofia Miguens Given) and in fact does not think of it as Burge does, given that his own r­ epresentationalism is formulated for appearances to the subject and not for inner vehicles of content. I believe McDowell’s motivation for being a representationalist regarding perception is rather different from the issues at stake in the clash with Burge. The analysis of the second debate is meant to show this.

Perception: Presentiment or Presence McDowell’s 1994 article The Content of Perceptual Experience 22 is one of the pieces discussed by Burge in Disjunctivism and Perceptual Psychology. It focuses on (early) Daniel Dennett (i.e. on an article on consciousness prior to the 1991 book Consciousness Explained23). Criticizing Dennett gives McDowell a pretext to put forward his own disjunctive conception of perceptual experience. 24 The disjunctivist conception of experience comes out as an alternative to the view of perception (or maybe one should say the lack thereof) that goes with Dennett’s theory of content and consciousness. For present purposes the main idea in Dennett’s view of perception is that perceivings, or experiencings, are seemings, seemings to a subject that things are a certain way. As for what seemings are, Dennett’s position is blunt: seemings are judgements. In other words, for Dennett perceiving, e.g. seeing, is straight judging. 25 Here are some examples from Consciousness Explained of what Dennett means by his claim that seeing is judging. 1 The Marilyn wallpaper. 26 I walk into a room and the wallpaper is a regular array of hundreds of Warhol-like Marilyns. I cannot possibly foveate and thus clearly see all Marilyns as Marilyns (as opposed to blobs). Still, I see a Marylin-covered wall. According to Dennett, such seeing, i.e. seeing the Marilyn-covered wall, is believing that the wall is covered with Marylins. 2 The pink neon ring.27 In the illusion known as the neon-spreading effect I look at a grid in which there are only black and red lines and I see a pink neon ring spreading over the region delimited by red-black intersections. There really isn’t a pink ring, as it were, ‘in the world’ – there are only black and red lines. For me, now, to be seeing the ring (i.e. it seeming to me that there is a pink ring) simply is judging that (or believing that) there is a pink ring. Again, there isn’t such a thing, as it were, ‘out there’: one brain circuit is misled to distinguish a bounded region, another comes up with a colour discrimination to label the region. This seeing-(or perceiving)-is-believing move does a lot of work in ­Dennett’s Multiple Drafts Model of consciousness. It is what ultimately

Is Seeing Judging? Radical Contextualism  129 allows Dennett to deal with decisive challenges for his theory of ­consciousness such as the Libet-cases of ‘delayed awareness of intention’ (where subjects’ reports of conscious decisions supposedly lag behind cortical initiation of action, 28 or the phi-phenomenon, i.e. cases of apparent movement.29 Dennett deals with such cases by dealing with subjects’ reports. Such reports are linguistic reports of seemings. It is such seemings that are taken by Dennett to be judgements. So a judgement in Dennett’s sense is a case of a subject taking it, and stating it, that things are a certain way for him/her. It is to this idea of judgement that he appeals in order to counter the temptation of speaking of ‘filling in’ by the brain to account for the cases above: This idea of filling-in is common in the thinking of even ­sophisticated theorists, and it is a dead give away of vestigial Cartesian materialism. What is amusing is that those who use the term often know better, but since they find the term irresistible, they cover themselves by putting it in scare quotes.30 Although McDowell is much less interested in cognitive science per se than Dennett is, he apparently professes a somewhat similar position about perceptual experience. In Mind and World he claims that “That things are thus and so is the content of the experience”, 31 or, more recently, in a more explicitly Sellarsian formulation, that experiencings or seemings are takings-as, or claims. 32 In other words, both McDowell and Dennett take perceptual experience to have representational content, which has the form ‘that things are thus and so for a subject’. In fact, what comes after the passage of Mind and World I just quoted above is: and it can also be the content of judgement. 33 Still, McDowell’s main criticism of Dennett in the 1994 article is that Dennett’s idea of seeing as judging amounts to a presentiment view of perceptual experience. In Dennett’s way of puting things, perceptual experiences turn out to be ‘less than encounters with objects’, they turn out to be something ‘inside our heads’. This is the first thing I want to look at. According to McDowell, Dennett’s position rests (in McDowell’s terms, which I believe are quite important for understanding his debate with Burge), on a ‘causal-enabling account of mindedness’.34 A causal-enabling account of mindedness concerns what Dennett calls the subpersonal level of agents, or perceivers. The personal/­subpersonal distinction was put forward by Dennett as regarding two ways of explaining human behavior and thinking of the mind. It contrasts the level of people, their sensations and activities (the personal level), with

130  Sofia Miguens the level of their brains, nervous systems and cognitive machinery (the ­subpersonal level). McDowell praises Dennett for claiming that we have access to ­content, i.e. to personal-level content, which, in Dennett’s view, is what is reported in subjects’ reports of how things seems to them, and not to subpersonal content-bearers, i.e. vehicles of content. Subpersonal vehicles of content are not, according to Dennett, introspectible by the perceiver. Yet McDowell sees a problem. As he later put it in his response to Burge, In my paper [on Dennett] I urge that a good theory of the workings of a perceptual system yields accounts of what enables a perceiver to get into perceptual states, not accounts of what it is for a perceiver to be in those states.35 (McDowell 2011: 250) This is a way of spelling out the personal/subpersonal distinction which is not in Dennett – it amounts to making the contrast explicit between a causal-enabling approach to mindedness and a constitutive account of mindedness. A causal-enabling approach to mindedness yields an account of what enables a perceiver to get into perceptual states; in other words, its object are the workings of a perceptual system. Mistaking it for something else, namely mistaking it for what McDowell calls a constitutive account of mindedness, makes one oblivious of what it is for a perceiver to be in those states. This is not a Nagelian observation about the what-it-is-liketo-be of consciousness. That is not McDowell’s point. McDowell’s point is that a causal-enabling account of mindedness naturally leaves out the very mark of perceptual experience, which is its openness to the layout of reality. Openness to things being as they are, e.g. to presentness and particularity, is exactly what McDowell thinks is missing in Dennett’s account. That this is missing is what makes Dennett’s view amount to an identification of experiences with presentiments. To take McDowell’s own exemple, let us say that I utter ‘That cat is asleep’36 [and there is a cat in front of me]. According to Dennett what is happening here is that I find myself in a position to judge thus; I am struck, without knowing why, by (my own) taking things to be a certain way; I know nothing about the etiology of such ‘presentiment’, I am just ‘told something’ by my inner processing machinery. And that is all there is in my taking things to be thus and so. McDowell thinks this cannot be right. Perceptual experience does not have to be carried out to its object ‘by an hypothesis’;37 in a perceptual experience the object of experience is simply directly there for the thinker. Dennett’s judgings-that may be something else – maybe thoughts, maybe fiction – but they are not perceptual experience. Dennett’s problem, McDowell claims, is that he conceives of perceptual experience as less than encountering objects, less than being

Is Seeing Judging? Radical Contextualism  131 acquainted with, or being presented with, the here and now. McDowell’s point is not a phenomenological point, rather, it comes out of an interpretation of the personal/subpersonal distinction. He thinks Dennett’s own interpretation of his own (good) personal/subpersonal distinction keeps him from acknowledging the level at which perceptual experience ­ erceptual expecomes into play, which is what interests McDowell. P rience belongs in the personal level, not in the subpersonal level. This is not about doing justice to the subject’s phenomenology, it is a point about the level at which philosophy is concerned with perception. And this is the personal level, which is the level at which the world is present for the subject. In visual experience it is the environment that is present, not “a computation of a representation of part of the environment from a pair of arrays of intensities and wavelengths”.38 And he goes on: “whatever may be true about the information-processing that goes on in the visual system (…) it is the relevant tract of the environment that is present to consciousness, not an image of it”.39 This is what matters for philosophers according to McDowell, when they are interested in the nature of perceptual experience or also, as he himself definitely is, in knowledge. I want to suggest that in the exchange McDowell thinks Burge too is (mis) taking a causal-enabling account of perception (what he repeatedly calls ‘the science’) for a constitutive account. He is convinced that that is the only way to go, i.e. that there is only one account of perception, not two. This makes him impervious to the fact that all of McDowell’s claims concerning perceptual experience and the epistemology of perception belong in a constitutive account and it is only as such that they should be understood. McDowell focuses on the ‘openness to the layout of reality’, the ‘presence to one’ and ‘encounters with objects’, not on the workings of a perceptual system. Such, for McDowell, is the territory of philosophy of perception, in contrast (not in competition) with ‘the science’. Needless to say, Burge is bound to disagree with the very idea of such a division of labor. What else could perceptual psychology be, if not the science of perception? This is something Burge repeatedly asks. By that he means that perceptual pscyhology is not limited to s­ ubpersonal states; the subject’s experience is part of its object: The empirical psychology of vision is in the business of explaining the processes involved in individuals’ perceptual representations of the physical environment. McDowell has given no genuine argument for regarding psychological explanation as non-literal. Similarly, he has given no argument for seeing it as concerned merely with the transactions of a subsystem and not with the empirical nature of visual perception by animals and humans. His claims about the science rest on a string of misunderstandings that elementary familiarity

132  Sofia Miguens with the science would have prevented. (…) The main principles of the science have beeen in place since Helmholtz. Its maturation into a complex, well-established and mathematized body of knowledge has been evident over the last thirty years. The science is about the laws of visual perception, not about something else.40 In other words, McDowell’s distinction between causal-enabling and constitutive accounts of mindedness simply does not make sense to Burge. This is actually quite relevant because McDowell’s distinction is not just a matter of who does what. Rather, it reflects directly on the fact that in the controversy McDowell keeps going back to the idea that ‘good cases’ involve more than veridical representation (he speaks of ‘a more demanding condition that experiences being veridical’). But Burge is not in a position to see what that means. Exploring the epistemological significance of the good cases (i.e. the epistemological significance of ‘openness to the layout of reality’, or ‘encounters with objects’, in the good cases) is crucial for McDowell. For Burge, in contrast, good cases are simply cases of representations, not per se any different from non-­ veridical perceptual representations. Here is how this reflects on McDowell’s response to Burge’s criticism number one in the controversy. According to Burge, the disjunctivists’ denial that there are explanatorily relevant common states between veridical experience and the bad cases simply flies in the face of the science. It is up to cognitive science of perception to tell us whether perceptions and illusions are the same as inner representational states. And it does tell us that they are. The background for this claim is what Burge sees as “The bearing of the empirical psychology of vision on the individuation of perceptual states”.41 Considering that the paradigmatic problem of visual psychology is “to explain how information contained in these [retinal] arrays is transformed into representations (perceptions) of physical entities in the distal environment,” given the fact that “the information available (…) significantly underdetermines the environmental distal causes of those registrations”, one is faced with a problem of ‘underdetermination’. ­Underdetermination takes a variety of forms, such as ambiguous figures, visual illusions, or visual completion.42 Perceptual systems basically usually get things right, so the problem is in fact overcome by the visual system operating under certain principles in the formation of perceptions. These principles constitute biases that convert proximal stimulation (…) into perception of the environment. Psychology is commited to some of the information being processed being representational – genuinely perceptual in that it represents objective matters and can be mistaken as well as veridical.43

Is Seeing Judging? Radical Contextualism  133 Two of Burge’s examples here are lightness constancy and the determination of the slant of a textured surface from perceptual representations of features of the texture of the surface.44 Thus the Proximality Principle (i.e. the idea that on any given occasion the total proximal input together with internal input into the system suffices to produce a given type of perceptual state) holds.45 A corollary of the Proximality Principle as it bears on non-veridical perceptual states is that “Since relevantly different environmental distal conditions could yield the same type of proximal stimulation, a given type of perceptual state can be produced by different distal conditions”.46 Under such light Burge cannot but see the disjunctivists’ denial of the common factor as ungrounded philosophical apriorism: “There is no getting around the fact that the basic kinds in perceptual psychology are (…) representational. Commitment to representations (…) as marking perceptual abilities is deeply embedded in the theory’s objectives, methods and explanations”.47 Step one of McDowell’s rejoinder is that his disjunctivism regarding perceptual experience does not have to deny any common factor and it doesn’t – appearing to one that things are a certain way is a common factor in perceptions and illusions. Also, he grants Burge’s claims about the Proximality Principle. But they concern ‘the science’, not the epistemology of perception. Burge does not see that, whereas ‘the science’ is concerned with the subpersonal level of perceivers, ‘encountering objects’, and thus ‘openness to the layout of reality’, is the foothold for the epistemology, an epistemology (i.e. disjunctivism) which is, McDowell says, ‘mandatory’. Of course such claim builds on a causal-enabling versus constitutive distinction which Burge does not accept: he thinks that what he calls ‘the science’, i.e. perceptual psychology, does concern the whole subject, i.e. the whole animal and not merely ‘transactions of a subsystem’. The fact that perception does makes knowledge about things available by placing them in view for us is crucial for understanding McDowell’s response to Burge’s accusation number two. Accusation number two concerns fallibility: ‘disjunctivists don’t understand fallibility’, he says. So I will go now into McDowell’s disjunctivist account of how openness to the layout of reality (our ‘encounters with objects’48) should bear on how fallibility is conceived. Notice that when Burge speaks of fallibility he speaks (mostly)49 of veridical and non-veridical states of perceivers, i.e. of inner representations, subpersonal states taken to be representations. In contrast, McDowell’s views on fallibility are built around capacities of subjects and appearances. Now, such notions are not subpersonal notions. But before I look at what this means when it comes to the Burge-McDowell discussion of fallibility, I want to bring Dennett in again. I want to look at the appearance-reality distinction that goes with his conceptualism because it will help me bring out what McDowell is fighting against with his own way of approaching fallibility. 50

134  Sofia Miguens

What Can Go Wrong? Seemings and Fallibility McDowell often speaks of experiencings, such as seeings, as seemings. A vocabulary of seemings allows one to consider both perceptual experience and epistemology. McDowell himself, in fact, always has an eye on the nature of empirical knowledge when he is considering perception. 51 The nature of empirical knowledge is something Dennett is not so much interested in (in matters epistemological, McDowell attributes to him “a pre-Humean epistemological optimism”52). Still a view of the status of seemings is an essential component of his view of content and consciousness given that such a view revolves around reports of seemings. I want to suggest that McDowell’s way of dealing with seemings is a way of avoiding an epistemological oddity which is part and parcel of Dennett’s view of content and consciousness: first-person verificationism. The issue is the following: is there such a thing as things seeming to me a certain way where I may be wrong about that? For Dennett this cannot be: in his interpretation of the personal/subpersonal distinction, once the difference between personal and subpersonal levels for a thinker, or a perceiver, is in place, there is, one basic principle regarding the personal level. There is no such phenomenon as really seeming over and above the phenomenon of judging in one way or another that something is the case. 53 This applies to reports of mental life (e.g. in the examples above: ‘I see the Marylin-covered wall’, ‘I see a pink ring’, ‘I am raising my arm now’, ‘I see a light moving from left to right on the screen’). This is in fact the principle embedded in Dennett’s use of the formulation ‘seeing is believing’. In Dennett’s own words, judgment is the key of phenomenology (by which he means the mental life of a thinker at the personal level): phenomenology is made of judgment, there is nothing more to ­ henomenology than that.54 p Specifically this means that where it concerns her own awareness of things being a certain way, it is impossible for the subject to decide between appearance and reality. All she can do is say how things seems to her. This is what Dennett calls ‘first person verificationism’. Interestingly, he sees first person verificationism as a Wittgensteinian position. Interestingly, McDowell’s (very different) disjunctivist position about appearance-reality distinction concerning ‘inner space’ is also supposed to be Wittgensteinian. And in fact so is Travis’ position. 55 Undoubtably there is dispute here concerning Wittgenstein’s inheritance where it ­concerns inner space. Anyway, even if McDowell and Dennett share a vocabulary of seemings for the personal level, there is one all-important difference: whereas for Dennett, perceptual seemings amount to being told (by inner

Is Seeing Judging? Radical Contextualism  135 information-processing machinery) that things are thus and so, and then the idea that there is no elbow room for the subject to distinguish ­between appearance and reality comes in, McDowell’s ‘space of seemings’ is completely different. McDowell thinks of seemings, such as perceivings, as acts of capacities of a subject in an environment. Now, conceiving of seemings, such as perceivings, as acts of capacities of a subject in an environment, McDowell is precisely avoiding Dennett’s subjectivizing of seemings, i.e. the restriction of the appearance-reality distinction to a subjective realm. In fact, fighting such restriction is one crucial motivation for his disjunctivism. So it is not just the disregard for presence and particularity of experiential encounters with the here and now that makes Dennett’s view of perception a presentiment view, it is also such a restriction to the internal of the appearance-reality distinction. These are, in fact, two sides of the same coin. Anyway, stressing, as McDowell does, that it is capacities that can be fallible means, first, that it is not representational inner states that should be regarded as being straight veridical or non-veridical. Also, stressing that it is capacities that can be fallible has that further intent of doing away with what he calls ‘a ­Cartesian conception of the inner world’. A conception of seemings may be Cartesian or not. McDowell wants to do away with any Cartesian conception of seemings. According to a Cartesian conception of seemings, the inner world is a region of the world whose elements include states of affairs consisting in appearing to the subject that things are a certain way. An extra clause is decisive: that things thus appear to the subject on an occasion is the whole truth about the relevant state of affairs (McDowell 2011: 244). This Cartesian conception of seemings may then come together with the apparently innocent idea that fallibility concerns a contrast between veridical and non-veridical representational states. McDowell’s alternative view is that although perceivers are sometimes mistaken regarding seemings, that does not entail that when they are not mistaken, they are ‘anywhere short of the facts’ – in some cases (the good cases) that things are ­ anifest to the subject. This is how this thus and so is simply being made m idea bears on how one conceives of the appearance-reality distinction: But suppose we say – not at all unnaturally – that an appearance that such and such is the case can either be a mere appearance or the fact that such and such is the case making itself perceptually manifest to someone. As before, the object of experience in the deceptive case is a mere appearance. But we are not to accept that in the non-­deceptive case too the object of experience is a mere appearance and hence something that falls short of the fact itself. On the contrary, the appearance that is presented to one in those cases is a matter of the fact itself being disclosed to the experiencer. So appearances are no longer conceived as in general intervening between the ­experiencing subject and the world. 56

136  Sofia Miguens Thus the disjunction that summarizes McDowell’s perceptual ­ isjunctivism is formulated as an idea about how to conceive of d appearances: We can express the idea [perceptual disjunctivism] with a disjunction: an appearance is either a case of things being thus and so in a way that is manifest to the subject or a case of its merely seeming to the subject that that is how things are. (McDowell 2010: 244) The intuition can also be formulated in terms of factoring: It is part of the point of my disjunctive conception of experience that having an aspect of objective reality perceptually present to one entails having it appear to one that things are a certain way. But that is not to say that having an aspect of objective reality perceptually present to one can be factored into some non-mental conditions and an appearance conceived as being the mental state it is independently of the non-mental conditions. The factoring fails; the state is the appearance it is only because it is a state of having something perceptually present to one.57 The point may be formulated in terms of appearances, of factoring, and also in terms of seemings. In the last case we have the following: perceptual experiences afford us seemings or appearings, i.e. takings-to-be that things are a certain way. Yet there is no genus seeming seeings, of which real seeings and merely seeming seeings are species (McDowell 2013a). The second ones are ‘experiences the subject would innocently take to be (e.g.) seeings’. If there was such a genus seeming seeings, species membership would make no epistemic difference. There is an alternative way, though, of conceiving the relation between seeings and merely seeming seeings, which is to take it that when an experience is a real seeing it does make an environmental reality present to the subject, and this is what makes all the difference epistemologically. It has a distinct epistemic significance without which no proper account of perceptually based knowledge is possible. When things are given to one and the subject’s conceptual capacities are operating, the subject is in a position to know that things are a certain way. One can sometimes know by seeing – and this is knowing in no weak, defeasible sense, of knowing. In McDowell’s words, experiences that are e.g. seeings afford conclusive warrants. This is what the good case provides, and this is one thing McDowell means when he says that more than veridical representation is at stake. McDowell’s story about fallibility is thus a story about capacities, and such a story gives us an alternative to what he thinks is a faulty inference from fallibility:

Is Seeing Judging? Radical Contextualism  137 A perceptual capacity is (…) a capacity – of course fallible – to get into positions in which one has indefeasible warrant for certain beliefs. 58 It is a confusion to think that the idea of fallibility can intelligibly carry over to exercises of fallible capacities, as in Burge’s remark ‘I believe that all perceptual representations apply fallibly to their referents in any given instance’.59 Burge can only say this given the subpersonal touchstone of his c­ onception of what representation is. We may see McDowell’s point about capacities, in his discussion of ­fallibility, as a point about grammar – according to such a point, ­capacities are capacities of an agent and fallibility is a property of such capacities, not of exercises thereof. One could also, going back to ­Dennett, put it in terms of cognitive theory: fallibility concerns the animal as a whole as a semantic engine, at the personal level, and that should be the foothold for epistemology. Dennett’s celebrated distinction between syntactic and semantic ­engines is, according to McDowell in his 1994 article, a potentially rich idea which Dennett himself ended up subverting. According to the distinction, only (whole) animals are semantic engines – i.e. the level at which the world is meaningful; their subpersonal parts are syntactic engines: unthinking, unjudging, unintelligent. This should mean, ­McDowell thinks, that only the animal (e.g. the person) can ever be ‘told’ something. The question then becomes (and he means this as a criticism of Dennett, in the 1994 article) how could then my inner processing machinery ‘talk’ to me (the whole me, as it were, in contrast to information flow taking place among functional subsystems)?: What could an information-processing device really tell an animal? What could an information-processing device really tell anything? It is essential to realize that the answer to this question can be, in fact is, “Nothing”.60 If an animal is ‘told’ something it is ‘told’ something by its environment, not by its inner cognitive machinery. Apart from the specific criticism of Dennett, one very important point here for the McDowell-Burge controversy is that if in fact the animal is the sole semantic engine, then internal representations cannot be regarded as actual representations: Nobody knows how to make sense of an animals control m ­ echanism, and connect it conceptually with the competence it is supposed to explain except by describing it as if it were what we know (…) it is not really, a semantic engine, interpreting inputs as signs of environmental facts and as output, directing behaviour so as to be suitable to those facts in the light of the animal’s needs or goals.61

138  Sofia Miguens This is why one should not mistake subpersonal representations (taken as such in the theory of cognition) for representation proper. If only semantic engines have dealings with content, then only semantic engines, as a whole, represent (the environment). But this is precisely one point which exasperates Burge about the 1994 article: “The appeal to representational content is not a metaphor in perceptual psychology”,62 he says; one may, one should, in fact one is committed by perceptual ­psychology to take certain inner states as representations. The point of the Proximality Principle is precisely that representations are needed to formulate laws of perceptual psychology. It is not just the case that Burge’s and McDowell’s views on perception and representation clash: they could not see each other’s position as a reasonable one. It is important to keep in mind that while M ­ cDowell’s view of fallibility – a central element of his perceptual disjunctivism and also of his epistemology of perception – allows for Burge being right about the science, it starts from the idea that being inner states of perceptual systems is not the only kind of intelligibility cognitive states have. One should sort acts of capacities in terms of epistemic significance, even if they belong together in terms of the kind of intelligibility they have for cognitive science, and then start doing epistemology. As possessors of epistemic significance, experiencings (e.g. seeings) are not veridical or non-veridical inner states of cognitive systems; they are intelligible as acts of capacities of perceivers in an environmentally present reality and it is as such that they matter for epistemology. It is thus ultimately the conviction that there is a difference between the tasks of epistemology and cognitive science that underlies McDowell’s claim that “the position I have recommended is not the position Burge attacks”.63 Whereas for McDowell what ‘the philosophy’ (to introduce an analogue for Burge’s term ‘the science’) is involved with – e.g. understanding the epistemological significance of openness to reality in good cases or getting clear about the appearance-reality distinction – is different than what cognitive science is involved in, for Burge, in his own terms, there is only one type of intelligibility. Anyway after answering his own question in the first Agnes Cuming Lecture (Can cognitive science determine epistemology?) negatively, he goes after a variation of the question: why does cognitive science need epistemology? His answer is that cogitive science needs epistemology for its conceptual apparatus to be so much as ­intelligible – one has to understand what representing is in order to understand what cognitive scientists may call representations. One first take-home message regarding uses of representation when conceiving of perception that we may get from this controversy is the following. If and when we take subpersonal states of perceivers to have representational content, to be representations, as Burge does, we are replacing a constitutive-account with a causal-enabling account of mindedness. We may very well believe that is the thing to do (certainly Burge

Is Seeing Judging? Radical Contextualism  139 himself does not see the point of the distinction). Still, if we do it, we lose touch with the fact that perception’s openness to the layout of reality is not only the mark of perception but also the touchstone for epistemology. A second take-home message comes up in the discussion of fallibility. In assuming that the contrast between veridical and non-veridical states of perceivers is the reference for conceiving of fallibility, we might be unknowingly making extra (‘internalizing’) assumptions regarding the ­appearance-reality distinction. And these might be assumptions we would otherwise dismiss as Cartesian. McDowell’s alternative is then to focus on capacities of an agent in an environment, and to see perceptual capacities as, granted, fallible capacities, but, still, capacities to get into positions in which one may have an indefeasible warrant for certain beliefs. Another important point to keep from this first controversy is that talk of ‘two different kinds of intelligibility’ does not involve rejecting the science, or even the Proximality Principle: it is ultimately a matter of taking cognitive science and philosophy (in this case philosophy of perception and epistemology) to be projects of a different nature. This is one thing that is definitely at the core of McDowell’s response to Burge. All of these points come off a reading of the personal/subpersonal distinction. In fact uses of the distinction make it possible to see that different conceptions of what doing philosophy is are at work in the field of philosophy of perception, which does not make things easier. The McDowell-Travis Debate McDowell’s motivations for disjunctivism that I have identified so far concern doing justice to presentness and acquaintance as marks of perception, detaching the appearance-reality distinction form the subjective realm, and acknowledging the bearing of the good cases on a conception of empirical knowledge. In general terms, Charles Travis would be with him in all this. Yet there is something else in McDowell’s current conception of perceptual experiencings – there is the representationalism, i.e. the idea that perceptual experiences (not the perceiver’s subpersonal states) represent things as being a certain way. Burge does not take issue with this; what he thinks does not stand scientific scrutiny is the disjunctivism, the representationalism about perceptual experience he simply finds uncontroversial. Yet this is precisely where the two disjunctivists, McDowell and Travis, part ways. I will now concentrate in understanding what their parting ways has to do with Travis’ radical contextualism. I will try to show that the difference between their positions can be traced to their different conceptions of judgement. Before going into some details of the McDowell-Travis more recent debate, let us stop for a moment and look at the different answers to the two questions that I posed at the beggining that we have got until

140  Sofia Miguens now. The questions were: (1) Should one say of a perceivers’s subpersonal states that they are representations? (2) Should one say of a perceivers’ perceptual experiences, as opposed to his (or her) judgements, that they represent things as being a certain way? The answers we have, and will be dealing with, are the following:

1 2

Burge

McDowell

Travis

Dennett

Yes Yes

No Yes

No No

No Yes

what interests me now is understanding the difference between ­McDowell and Travis where it concerns their answers to Question 2. Triangulating the McDowell-Travis Debate: Kantian and Fregean Conceptions of Judgement Burge’s conception of the job of philosophy of perception is not a­ cceptable either for McDowell or for Travis. For neither of them are the problems of philosophy of perception and the problems of cognitive science of perception to be identified. Philosophical discussions of perception do not concern perceivers’ subpersonal states – they start, as it were, at the personal level. This is where the McDowell-Travis debate is situated. Yet the very fact that McDowell and Travis have opponents such as Burge in common and the fact that their debate takes place against a background of wide philosophical agreement makes it much harder to identify the sources of their divergences. what is it that they ultimately disagree about? Representation and ‘Shared Form’ we know one thing from the start: in spite of background agreement, namely on the (negative) idea that a perceiver’s subpersonal cognitive states are not representations proper, McDowell’s representationalist conception of perceptual experiencings (i.e. the idea that perceptual experiences themselves represent things as being a certain way) is simply at odds with Travis’ silence of the senses view of perception. how can this disagreement be spelled out and better understood? Both McDowell and Travis address the question of perception in the context of an investigation of the conceptual capacities of agents (at the personal level, one might add). here is an example of McDowell doing it in his article “Conceptual capacities in perception”: A zebra can be described, but that is no reason to suppose that the zebra itself has a form it shares with a description, or with the thought a description expresses.64

Is Seeing Judging? Radical Contextualism  141 In order to understand this excerpt one has to have in mind the fact that some critics who accuse McDowell of idealism do it precisely because they attribute to him a sharing of a form (the same form in world and thought). They see this as a ‘projection of subjectivity’ and thus as an idealistic stance. In this particular passage McDowell is denying such a reading, by refusing the fact that an object in the world may be ­described implies that the wordly object (the zebra) partakes of the form of thought and description (he is, by the way, discussing criticisms of his view by Michael Ayers). But if it is not a sharing of a form that he proposes, what is it then? According to McDowell, perceivings (seemings) are claim-like, claims which are not yet judgements. Now ‘claim’ is a term of w. Sellars, which McDowell thinks is “wrong in the letter but right in spirit”, as he puts it in his article Avoiding the Myth of the Given.65 he uses the term to speak of experiences as intuitions. It is important here to always keep in mind that McDowell’s approach to perception has its eyes on epistemology. In the words of French philosopher Jocelyn Benoist, McDowell’s approach to perception has in fact an all too dominating epistemological purpose (Benoist speaks of “La misère du theoreticisme”, and describes it as “confondre perception et connaissance perceptive”66 – in other words he believes that McDowell permanently con ates perception and perceptual knowl edge). For McDowell, anyway, an account of perceptual experiencing is key in an account of knowledge in that we cannot understand the relations in virtue of which any judgement is warranted except as relations within the space of concepts (this is a core thesis of Mind & World, McDowell 1994). Only representations enter such relations, so the fact that perceptual experiences are representational is crucial for knowledge. But it is also the case that McDowell has reformulated his Mind and World position, the position according to which such representational content is propositional content. he is no longer commited to the idea of propositional content of experience; the article Avoiding the Myth of the Given is a particularly clear expression of the shift. he does not claim anymore that perceptual seemings have propositional content; only judgements and assertions have propositional content. Yet he still claims that perceptual experiences have content: they have intuitional content. McDowell’s intuitional content is as an interpretation of Kant’s Anschauung. It is in fact the german word Anschauung that McDowell translates as a ‘having in view’.67 his current claim is that perceptual experiences have intuitional content; that is why they are representational. Before the more recent exchanges and reformulations, McDowell representationalism had been a target for Travis, namely in his 2004 Mind article The Silence of the Senses. what Travis targeted then was the commitment to the determinateness of seemings which according to him necessarily came with representationalism concerning perceptual experience (McDowell was not the sole target of such criticism, so were e.g. Christopher Peacocke, gilbert harman, John Searle, Michael Tye and Colin Mcginn). As McDowell himself put it once (and he is very

142  Sofia Miguens good at formulating the opponent’s theses), for Travis if a rock might look like a crouching animal and also look like a rock this better not all be the content of the same perceptual experience.68 In The Silence of the Senses what supported the objection to the representationalist’s ­commitment to such determinatness was, in a very Austinian vein, a linguistic analysis, i.e. an analysis of our ways of speaking of looks, seemings and appearances. The conclusion of the analysis was that there simply is not one sense of looks that would serve the purpose of the representationalist. We speak of looks or appearances in many senses. In particular, there are (in the most recent 2013 terminology69) perceptual (e.g. visual) appearances and conceptual appearances. Imagine that we say “The upper line looks longer” as we look at the Müller-Lyer lines. And then imagine that we said, watching TV on the night of the last French presidential election, and before knowing the full results, “It looks like Macron is going to win the election”. Travis’ point is that these are totally different phenomena, which should not be conflated. Yet such conflation is what Travis thinks is bound to happen when we speak of a perceptual experience as a particular seeming. When we speak of a perceptual experience as a particular seeming we are doing it as if there was one thing which is the one and only way things look in a particular perceptual experience. Of course one might counter: but isn’t there really such a thing, at least sometimes? Let us consider the Müller-Lyer illusion again. Should we really not think that there is a look here (one and one only)? (A look which is, by the way, not a conceptual look, but one objective visual look.) Here is Benoist, quoting Merleau-Ponty, as a help for resisting such temptation: dans l’illusion de Müller-Lyer les segments ne sont ni égaux ni ­non-égaux, c’est dans le monde objectif que cette alternative s’impose.70 He means that only with judgement is such an alternative there; it is there, thus, in the objective world. This is something Benoist attributes to Travis and agrees with. Anyway, one point of Travis’ is that the variety of senses of looks and seemings precludes what the representationalist needs to get his case off the ground. Sticking to the non-­determinateness of what one is presented with, and thus to the idea that perception basically puts things in view is the suggestion; this is (part of what) the ‘silence of the senses’ view proposes. That the role of judgement is such is one reason why, according to Travis, seeing has to be essentially unarticulated. But of course for McDowell this is The Myth of the Given. One interesting question here is how much does McDowell’s contrasting conviction (the conviction that seeing is essentially articulated) owe to his Kantianism.

Is Seeing Judging? Radical Contextualism  143 McDowell’s Kant It is well known that McDowell’s reading of Kant was very much ­influenced by P. F. Strawson and then, more and more, by Willfrid ­S ellars.71 Let us consider the Kant-inspired formulation of what constitutes the Myth of the Given: we succumb to the Myth of the Given, ­McDowell says, by not acknowledging that the understanding is at play in sensibility itself. But where does McDowell’s interest in Kant really lie? Although McDowell may care about transcendental arguments (see e.g. his article “The Disjunctive Conception of Experience as Material for a Transcendental Argument”, in Engaged Intellect 72), and one might trace transcendental arguments to Kant, he does not care much about Kant’s global transcendental framework, or Kant’s view of subjectivity as it includes, say, the difference between understanding and reason, or the topic of synthesis, or the Critique of Pure Reason inventory of the forms of propositional unity (i.e. the categories). What McDowell cares about when he appeals to Kant is unity, the unity of judgement as it relates to intuition. Hence the recurrent quote from the Critique of Pure Reason is The same function which gives unity to the various representations in a judgement also gives unity to the mere synthesis of various representations in an intuition; and this unity in its most general expression, we entitle the pure concept of the understanding. (I. Kant, Critique of Pure Reason A 79/B104–105, § 10 Transcendental Analytics) We fall prey to the Myth of the Given by not acknowledging that only the unity of (content) of judgement could make for the unity of content of intuitions. Now McDowell’s own (officially, Kantian) story about experiencing and judging goes the following way. Experience reveals that things are thus and so; he calls this seemings. As we experience, capacities that belong to reason (he speaks of conceptual capacities) are actualized. They are actualized in the experiencing itself – but this does not mean we should think of perceptual experience as ‘putting significances together’. All we need to acknowledge is that perceptual experiences are ­actualizations – not exercises – of conceptual capacities. This means that there is a potential for discoursive activity already there in intuition having its content – and for that to be so not all concepts need be at play, but some concepts have to be. Yet content whose figuring in (such) knowledge is owed to the (further) recognitional ­c apacity need not be part of the experience itself. Experiencing puts me in a position to know something non-inferentially. Although according

144  Sofia Miguens to the view above experience does have (conceptual) content, it does not have ­propositional content, nor need it include everything the e­ xperience ­enables its subject to know non-inferentially. What McDowell means by the idea that experiencings have intuitional content is that the unity of intuitional content is given; it is not a result of our putting significances together, as discoursive unity is; but it is not provided by sensibility alone either. This is what matters for him. Seeing capacities of reason as actualized even in our unreflective perceptual awareness is, for McDowell, the best antidote to an intellectualistic conception of human rationality.73 Of course if in this anti-­intellectualist view, seeings are seemings and thus (proto) judgings, one might wonder whether this is actually the best antidote to an intellectualist conception of human rationality (Hubert Dreyfus, himself a paladin of anti-­ intellectualism in philosophy and cognitive science, criticized ­McDowell, in the debate between them which took place a few years before his death,74 for seeing humans as ‘24hour rational animals’). What I am getting at here is the fact that in Kant himself what holds sensible and discursive unity together is the ‘I think’ of apperception. This in turn leads him to explore the originally synthetic unity of ­apperception – and once apperception is in, what we are dealing with is a view of consciousness, of consciousness as synthesis. Of course synthesis in ‘consciousness as synthesis’, the synthetic unity of consciousness, is synthesis in a different sense than synthesis of concepts in a judgement (namely it involves time). Be that as it may, it is unity that McDowell is interested in. He believes there is a task for unity: the task of the unity of judgement. But just to finish sketching McDowell’s current view of judgement, what exactly is judgement, ‘the paradigmatic exercise of theoretical rationality’, as he calls it? Judging is making it explicit to oneself. Judgements are inner analogues to assertions. The capacity to judge is a capacity for spontaneity, for self-determination in the light of reasons recognized as such. This is what is, for McDowell, the most important trait of judgement. A knowledgeable perceptual judgement has its epistemic entitlement in the light of the subject’s experience.75 This is what McDowell is most interested in: how perceptual experience represents things is not under one’s control76 yet minimally it must be possible – in view of a particular seeming – to decide whether or not to judge that things are a certain way. Travis’ Frege One key to understand the dispute about perception that takes place between McDowell and Travis is to see that whereas the problem of unity is a problem for McDowell it is a non-existing problem for Travis. In contrast with McDowell’s Kantian idea that given Erkenntnisse (cognitions) are brought to the subjective unity of apperception and made

Is Seeing Judging? Radical Contextualism  145 into a unity, Travis sees judgement is rather a stance of an agent in an environment. A judgement takes place where there is room for being exposed to error; that happens when an agent is presented with (what there is in) an environment. ‘Environment’ in Travis’ terminology simply means ‘what there is to be met with’. In Frege Father of Disjunctivism Travis put it like this: Frege saw that we needed an environment, and thus perception, and not merely sensation, if there is something for logic to be about. Not that logic applies only to environmental thoughts but rather that only given an environment for thinkers can the notion of judgement gain a foothold. (Travis 2013: 89) Jerry Fodor once put his conception of mind in a nutshell by saying: no representations, no computations; no computations, no mind. Travis’ conception of mind and thought could be put in a nutshell by saying ‘no environment, no judgement, no judgement, no logic’ (and, of course, if ‘no logic’ then ‘no thought’). The reason for environment being so crucial is the fact that, for Travis, judgement involves a particular kind of correctness: truth. According to Frege, and Travis follows him there, truth is the very business of logic. As Frege put it in Der Gedanke, logic deals with the laws of being true. No truth, no logic. Travis’ suggestion is that there is no such thing as this kind of correctness, i.e. truth, for the non-environmental: To be a judgement just is to be subject to a kind of correctness (i.e. truth), which is a particular kind of correctness (contrasting, for example, with being justified). Explaining what kind of correctness truth is and explaining what sort of attitude judgement is are one and the same enterprise. (Travis 2013: 71) Yet Travis’ story is until now a story about logic, judgement and truth – how does it become a story about perception? And how is it possible that in this story Frege ends up not being a conceptualist regarding perception? Admittedly Travis’ reading of Frege is very unorthodox if we think that Frege is usually, e.g. by Michael Dummett,77 taken to have been, inasfar as he was concerned with perception, namely in The Thought, a conceptualist regarding perception. That is how the relevant passage of Der Gedanke that brings in the role of the non-sensible (nicht sinnliches) element, is usually read: Sense impressions do not reveal the outer world to us. Having i­mpressions is not seeing things….It is necessary but not sufficient. Something nicht sinnliches has to be added. This is what unlocks

146  Sofia Miguens the outer world. Without it each of us would be locked in an ­inner world. Besides the inner world we must distinguish the external world of sensible perceptible things. (but) To recognize any of these domains we need something not sensible.78 This something nicht sinnliches is for Travis a Fregean thought. ­According to Travis’ proposal a thought always contains something reaching beyond the particular case, by means of which the particular case is presented as falling under a generality. Nothing less than this makes for something truth-evaluable. But it is precisely because a thought is, ­according to Frege, involved in perception that e.g. an interpreter of Frege such as Michael Dummett sees Frege as a conceptualist regarding perception.79 How come Travis sees things differently? A further step is needed here. For Travis such a step is a (also Frege-­ inspired) distinction between what he calls the conceptual and the non-conceptual. The conceptual he identifies as ‘ways for things to be’. The non-conceptual are the particular ways things are (he also calls it ‘the historical’). The conceptual is the domain of logic. One very important further point here concerns a distinction between logic and rationality. Logic, which Travis characterizes as ‘moves within the conceptual’, is not sufficient for accounting for what he calls reason’s reach, i.e. for rationality of agents. The rationality of agents is reach to the non-­conceptual. Such reach is to be done by judgement, and judgement only. So judgements are the doings of agents while reaching to the non-­conceptual – they are not anything like steps in a formal domain. Anyway, only against this background are we then entitled to the ­following reading of seeing (in which Travis takes Frege’s following words literally): But don’t we see that this flower has five petals? One can say that, but then one uses the word ‘see’ not in the sense of a bare experience involving light, but one means by this a connected thought and judgement.80 This is Travis’ alternative to McDowell’s reading of seeing as i­ nvolving understanding in sensibility. One main point there is that judgement as the doing of an agent is the core of ‘reasons’s reach’, the reach from the conceptual to the non-conceptual. Of course from ­McDowell’s Kantian view point there is no such thing as the Travisian non-­conceptual (the ways things are) to reach to; this would be a kind of noumenon. But in Travis’ account of representation, representation proper is necessarily representing-as done by a thinker, or agent, as he/she judges, and as such it takes place under the guise of such reaching from the conceptual to the non-conceptual. This is the core of the radical contextualist approach to representation. Representation proper

Is Seeing Judging? Radical Contextualism  147 is necessarily representing-as, which is a three party affair: there has to be (1) the representer, i.e. a thinker, representing (2) a stretch of the non-­conceptual (what is represented-as) as (3) a way for something to be (so involving the conceptual, i.e. ‘ways for things to be’). Short of this third party affair there is no such thing as representation proper. Hence there are e.g. no subpersonal ‘representations’ or representational content of perceptions. For the radical contextualist there ­simply is no representation proper if it is not representing-as by a thinker, who is an agent in an environment. I will not go deeper into this view, or into its clearly ontological ­implications, which have to be substantiated. All I wanted to point out at this point was that Travis recruits Frege for thinking of how we even get something truth-evaluable into a picture of thinking and judging and this leads him to representing as it is done by thinkers, or agents. One ­basic idea here is that the ‘environment’ is needed for there to be accuracy conditions; this will do away with the idea that experiencings themselves have anything like accuracy conditions. Experiencings do not have accuracy conditions, experiencings (just) bring our surroundings into view. Sight affords awareness of what is before the eyes, it puts opportunities on offer – that is all. In terms closer to philosphy of language: for there to be representing (representing-as, in Travis’ terminology) there have to be speech acts as issued by authors, for which such authors take responsability. And such representings-as are historic happenings. Only in such circumstances (speech acts, agents, environment) may we speak of anything being true or false, e.g. of a perceptual judgement being true or false. The encounters with ‘what there is to be met with’ are occasions for linguistic expression. In the case of perceiving, the non-­determinateness of experience (silence of the senses) makes for alternative partitioning in judgement; only the uttering of a judgement makes for determination of what otherwise is simply what we are ­presented with. To sum up: according to radical contextualism any story about representation proper has to be a story about thinkers, and in that sense a story about the personal, not about the subpersonal, level. Such thinkers are agents, they go about representing-as what they encounter – so what matters here is openeness to the world at the personal level. This means – and we know this from other parts of Travis’ work81 – that not only subpersonal representations do not represent in the proper sense, but also thoughts themselves do not represent, as sentences themselves do not represent; only thinkers represent things as being certain ways, on occasions. Where does that leave us where it concerns the question whether seeing is judging? Travis’ position is that only perceptual judgements of agents on occasions represent things as being a certain way and may then and as such be true or false. The senses are silent, i.e. they do not judge, do not represent, only thinkers do, on occasions.

148  Sofia Miguens

An Ambiguity Regarding Appearances I would like to finish by testing how McDowell’s Kant-inspired c­ onception of judgement and Travis’ Frege-inspired conception of judgement position us, respectively, when it comes to an ambiguity regarding the notion of appearances. This is an ambiguity which bears on metaphysical discussions, i.e. on whether one has a general realist or an idealist stance on thought-world relations.82 I believe a difference comes out in McDowell’s and Travis’ r­ espective (and apparently innocuous) linguistic preferences for ‘seemings’ and for ‘looks’. This is the ambiguity. When we speak of ‘appearances’ we might mean (1) a mere appearance, i.e. a seeming to me; (2) my being appeared to. Let us imagine Travis in his Austinian mode speaking of looks. Let us evoke Austin’s example of the lemon in Sense and Sensibilia – a piece of soap which looks exactly like a lemon.83 So there is a soap lemon in front of me; it looks like a lemon, it looks like a real lemon; I think it is a lemon. So what? Are my senses misleading me? There is nothing wrong with my experience, it is neither a hallucination, nor a real lemon, it is a soap imitation. The only mistake is in the thinking, the judging, there is nothing wrong in reality, or in my experiencing. Looks are looks of things, for anyone who might be there to be appeared to. This is sense number two of appearances above. Let us now consider McDowell’s view of seemings as claims (‘claim’ is Sellars’ word, which is wrong in the letter but right in spirit, he says). Granted, McDowell takes seemings to be acts of capacities of a subject in an environment. But it is also true that when he is for example considering the Müller-Lyer illusion (e.g. in Mind and World) he wants to say that there is the seeming and there is the room for deciding, the judging: things seem to be a certain way and the subject does not believe things are as they seem to be. Insofar as McDowell takes seemings to be claims he is tempted to see them as seemings to a subject – and this is precisely where he risks facing a particular problem. Consider senses 1 and 2 of appearance above. It is only with the idea that judging involves the unity of apperception that the first sense of appearance (an object as a seeming to me, and thus a mere appearance) sneaks in. But it is precisely because of that that the risk of ‘internalizing the object of representation in representation’ comes in (I borrow the expression from Béatrice Longuenesse,84 who uses it for Kant). In other words: if one takes a seeming to be a judging, and if one takes it that a judging as such is a matter of bringing representations under the unity of self-consciousness, then self-consciousness will be involved in the very fact that there is a seeming. The result will be internalizing the object of representation in representation.

Is Seeing Judging? Radical Contextualism  149 Internalizing the object of representation in representation is obviously not McDowell’s goal – after all he is a disjunctivist, and, if anything, disjunctivism aims at rejecting the restriction of the appearance-reality distinction to a subjective realm. But his appeal to (Kantian) unity to account for the exercise of rational capacities makes him run that risk. It is thus that the respective Kantian and Fregean allegiances of McDowell and Travis come to make a difference. Because McDowell’s view of judgement is ultimately Kantian, self-­ consciousness is its touchstone. In contrast, Travis’ Frege-inspired view of judgement is deeply anti-Kantian. This is not so (at least not immediately) because of any questions regarding structure of judgements and propositions, or regarding analyticity, i.e. the kind of questions logicians and historians of logic might be interested in when comparing Kant and Frege. The view is deeply anti-Kantian because of the ­environment-constraint for logic (to be), given the environment-­ constraint for judgement. Of course, together with the environment-­ constraint for logic (which is a constraint for logic to be, i.e. for there to be such a thing as logic), T ­ ravis is putting forward an idea about rationality: the idea of rationality’s reach to the non-conceptual. He wants to contrast rationality and logic. Logic’s reach is a reach (solely) within the conceptual, but in any story about representation, thinkers, or agents, enter and where thinkers enter there is rationality. A thinker as agent, and not self-consciousness, is the touchstone for Travis’ view of thought-world relations. This means that the radical contextualist picture of thought-world relations will ­focus on rationality (reason’s reach) first and on logic (or logics) next.85 I am not here claiming that all this stands, or stands as it is. All I intend to stress is that there is no central role in Travis’ radical contextualist picture of logic and rationality for the ‘I think’ of consciousness. Not anyway within the view of judgement. There is no role in judgement for the ‘I think’ of consciousness according to Travis because there is no role in judgement for combination or synthesis, synthesis into a unity of anything which would be ‘previous’ to the judging. This something previous would have to be concepts previously possessed. But, precisely, according to Travis’ radical contextualist, concepts come only after judgements, i.e. after the doings of an agent in an environment.86

Conclusion In which way is Travis’ view of perception a form of contextualism as it is usually considered in the philosophy of language? The fact is that Travis’ radical contextualism, as he currently formulates it, concerns the metaphysics of representation and agency and not just linguistic meaning. Yet the assumption is that the debate around linguistic meaning cannot be isolated from such questions. Anyway Travis’ position

150  Sofia Miguens about linguistic meaning counts as contextualism because it indexes linguistic meaning to a speaker’s judging things to be so on a specific occasion. This means, of course, that there will be no recipe for determining what a speaker puts forward as true on such a specific occasion; words are used as tools by the speaker to make recognizable how the speaker represents the world, so the representing is done not by sentences per se (or, for that matter, by what Travis calls thoughts): it is done by the speaker. Only thinkers represent, for the reasons analyzed above. Representing-as is, as Travis’ puts it, the preserve of thinkers. Sentences do not represent. Thoughts do not represent. Thinking otherwise is a kind of anthropomorphization of sentences, or thoughts – which is, of course, what the contextualist thinks the non-­contextualist is doing. Such is the core of the radical contextualist picture of representation. I took a long detour in this article to try to make clear what the ­questions regarding perception and representation are (and are not). The reason for doing it was, as I said at the beginning, the fact that more often than not, in current philosophy of perception, one starts off by discussing representational content and whether perception has it, thus ignoring the questions I am interested in before they are even posed. In particular, because the question of representation in general is not directly posed, work in the philosophy of perception is simply taken to be on the same page with cognitive science of perception, or in some sense continuous with it. This is not at all unexpected, of course, in times of Quinean naturalism. The analysis of the first debate, the ­McDowell-Burge debate, was meant to isolate a viewpoint from which to reject such Quinean continuity between science and philosophy, which many people take to be a totally uncontroversial consequence of naturalism. There is one sense in which the work of philosophy of perception should not be taken to be on the same page with the cognitive science of perception; the personal/ subpersonal distinction may be instrumental in formulating what is at stake. Philosophy of perception is concerned with the personal level of mindedness and with openness of mind to the world at such a level, not with subpersonal workings of cognitive systems. Subpersonal workings of cognitive systems belong in cognitive science, not philosophy. Such is a way of seeing things which Travis shares with McDowell. But if they share such a viewpoint, then where does their difference as regards perception and representation come from? The analysis of the second debate was meant to trace the difference to different conceptions of judgement. And there we find Travis’ position to be, as it were, in a further sense anti-naturalist – not just in the sense of rejecting Quinean continuity between cognitive science of perception and philosophy of perception and using the personal level to mark that difference, but also because at its center lies of a view of language and truth and not a view of (self) consciousness. And language and truth are much harder to ‘naturalize’ than

Is Seeing Judging? Radical Contextualism  151 consciousness is – or, at least, the problems posed by ‘­naturalization’ are different. Perhaps that accounts for why Travis, in contrast, with McDowell, never calls himself a naturalist. But that is a question I will not pursue here.

Notes 1 Travis, Charles. “Frege, Father of Disjunctivism.” In Perception – Essays after Frege, Charles Travis, 59–89 (Oxford: Oxford University Press, 2013). 2 See Brogaard, Berit. Does Perception Have Content? (Oxford: Oxford ­University Press, 2014) for an overview of the discussions. 3 According to Travis (see Travis, Charles “Frege, Father of Disjunctivism.”), Frege is the father of disjunctivism. Disjunctivism is a position Travis espouses, and whose nature will be discussed in this article. For overviews of disjunctivism see Byrne, Alex and Heather Logue. Disjunctivism – ­C ontemporary Readings (Cambridge, MA: MIT Press, 2009) and H ­ addock, Adrian and Fiona Macpherson. Disjunctivism: Perception, Action, Knowledge (Oxford: Oxford University Press, 2008). It should be noted that at the time of Travis’ 2004 Mind article “The Silence of the Senses” (Travis, Charles. “The Silence of the Senses”. In Perception – Essays after Frege, Charles Travis, 23–58 (Oxford: Oxford University Press, 2013); originally in Mind 113 (449) (2004): 59–94), J. L.Austin was the main historic reference. 4 Dennett, Daniel. Consciousness Explained (Boston, MA: Little Brown, 1991). 5 My sources will be the following: Burge, Tyler. “Disjunctivism and P ­ erceptual Psychology”. Philosophical Topics 33, no. 1 (2005): 1–78; ­McDowell, John. “Tyler Burge on Disjunctivism”. Philosophical Explorations 13, no. 3 (2010): 243–255; Burge, Tyler. “Disjunctivism Again”. Philosophical Explorations 14, no. 1 (2011): 43–80; McDowell, John, “Avoiding the Myth of the Given”. In Having the World in View, John McDowell, 256–272 ­(Cambridge, MA: Harvard University Press, 2009); McDowell, John. “Can Cognitive Science ­ gnes Cuming Lecture Determine Epistemology? Response to Tyler Burge”. A I. www.youtube.com/watch?v=m8y8673RmII, 2013a; McDowell, John. ­ ecture “Are the Senses Silent? Response to Charles Travis”. Agnes Cuming L II. www.youtube. com/watch?v=fBQHEGg5JSo, 2013b; Travis, “The S­ ilence of the Senses”; Travis, “Frege, Father of Disjunctivism”. 6 Burge is a representationalist in the sense that he takes mental states such as veridical perceptions, illusions and hallucinations to be all mental representations, and all on the same foot as such. 7 As we will see, one of the two disjunctivists (McDowell) is a representationalist in a different sense from the sense above. 8 Burge, “Disjunctivism and Perceptual Psychology”. For Burge’s epistemology of perception, see also Burge, Tyler. “Perceptual Entitlement”. P ­ hilosophy and Phenomenological Research 67, no. 3 (2003): 503–548. The debates involved many exchanges, with detailed proposals on both sides; naturally here I will be considering only some of aspects. 9 In fact, amounting to completely ungrounded philosophical apriorism, ­comparable to Hegel’s claim “that the number of planets is seven” (Burge, “Disjunctivism and Perceptual Psychology”, 29). 10 Burge, “Disjunctivism Again”, 71. 11 Burge, “Disjunctivism and Perceptual Psychology”, 25.

152  Sofia Miguens 12 Burge. “Disjunctivism and Perceptual Psychology”, 22–23. 13 Burge. “Disjunctivism and Perceptual Psychology”, 3. 14 Burge. “Disjunctivism and Perceptual Psychology”, 67. 15 Burge. “Disjunctivism and Perceptual Psychology”, 67. 16 Burge. “Disjunctivism and Perceptual Psychology”, 3. 17 McDowell, “Tyler Burge on Disjunctivism”, 244. 18 But they are in fact quite relevant for the second debate I consider in this a­ rticle, since McDowell and Travis definitely have very different philosophical ‘heroes’. 19 See Snowdon, Paul, “The Formulation of Disjunctivism”. Proceedings of the Aristotelian Society, New Series 105 (2005): 129–141 for a thorough and extremely clear spelling out of the central claim of disjunctivism (as a claim in the philosophical theory of perceptual experience) and also for an identification of the peculiarities of McDowell’s position. 20 This is not an ad hoc choice, since these are in fact the topics of ­McDowell’s two 2013 Agnes Cuming Lectures at UCDublin (Can Cognitive Science ­Determine Epistemology? Response to Tyler Burge, and Are the Senses Silent? Response to Charles Travis). The lectures are not published, but the ­video-recordings (of the full lectures) are available online. Of course ­McDowell is a rather peculiar disjunctivist in that he is a representationalist. Cf. Soteriou, Matthew, 2014, “The Disjunctive Theory of Perception”. The Stanford Encyclopedia of Philosophy (Summer 2014 Edition), Edward N. Zalta (ed.). http://plato.stanford.edu/archives/sum2014/entries/ perception-­disjunctive/ for comparisons among disjunctivists regarding representationalism. 21 McDowell, “Avoiding the Myth of the Given”. 22 McDowell, John. “The Content of Perceptual Experience”. Philosophical Quarterly 44, no. 175 (1994): 190–205. Reprinted in a slightly different form in Mind, Value and Reality, John McDowell, 341–358 (Cambridge, MA: Harvard University Press, 1998). 23 Dennett, Daniel. “Toward a Cognitive Theory of Consciousness”. In ­Brainstorms – Philosophical Essays on Mind and Psychology, ­Daniel ­Dennett, 149–173 (Montgomery, VT: Bradford Books, 1978). Since ­Dennett’s basic approach remains unaltered (See Miguens, Sofia. Uma ­Teoria Fisicalista do Conteúdo e da Consciência – Daniel Dennett e os debates da filosofia da mente (Porto: Campo das Letras, 2002)), I will refer to his 1991 book Consciousness Explained. 24 In other places McDowell’s presentation of his disjunctive view is prompted otherwise, e.g. by Myles Burnyeat’s comparison between classical and ­Cartesian skeptic positions on the nature of appearances (see McDowell, John. Singular Thought and the Extent of Inner Space. In Subject, Thought, and Context, edited by Philip Pettit and John McDowell, 137–168 (Oxford: Clarendon Press, 1986). This article is the other target of Burge 2005). ­A nother starting point is the Wittgensteinian discussion on criteria (see ­McDowell 1998, Criteria, Defeasability and Knowledge. 25 He more often says ‘believing’ (see Dennett, Consciousness Explained, Chapter 11, Dismantling the witness protection program, Seeing is believing). But in Dennett, Consciousness Explained, 364, in the Dialogue with Otto, he simply equates judging and believing: You seem to think there is a difference between thinking (judging, deciding, being of the heartfelt opinion that) something seems pink to you and something really seeming pink to you – but there is no difference. There is no such phenomenon as really seeming – over and above the phenomenon of judging one way or another that something is the case.

Is Seeing Judging? Radical Contextualism  153 26 27 28 29 30

31 32 33 34

Dennett, Consciousness Explained, 354–355. Dennett, Consciousness Explained, 351, 363. Dennett, Consciousness Explained, 162–166. Dennett, Consciousness Explained, 114. In an elementary case of the phi-phenomenon subjects report something like ‘I see a light move across the screen’ when there are only two separate flashes lit in rapid succession. Dennett, Consciousness Explained, 344. For instance to account for the phi-phenomenon or the neon spreading illusion. He also uses the idea to resist what he sees as overinterpretations of the Libet cases (i.e. using them to deny free-will). McDowell, “The Content of Perceptual Experience”, 26. McDowell, “Can Cognitive Science Determine Epistemology? Response to Tyler Burge”. McDowell, “The Content of Perceptual Experience”, 26. In the exchange Burge repeatedly rejects the idea that perceptual psychology is a question of a causal-enabling, as opposed to a constitutive, approach to perception: The aim of the theory is to explain the structure of human and animal perception. The theory does not just explain a mechanism of perception or a set of enabling conditions for perception. The theory does not confine itself to providing an account of a causal chain of non-perceptual processes that precede or lie in the background of an individual’s ­perceiving – and then stop there. The theory incorporates what is known about the accuracy and inaccuracy of whole-organism perception. (Burge, “Disjunctivism and Perceptual Psychology”, 21)

McDowell, “Tyler Burge on Disjunctivism”, 250. McDowell, “The Content of Perceptual Experience”, 343. McDowell, “The Content of Perceptual Experience”. McDowell, “The Content of Perceptual Experience”, 342. McDowell, “The Content of Perceptual Experience”, 342. Burge, “Disjunctivism and Perceptual Experience”, 50. Burge, “Disjunctivism and Perceptual Experience”, 4, 21. Burge, “Disjunctivism and Perceptual Experience”, 11. Burge, “Disjunctivism and Perceptual Experience”, 12–13. Burge, “Disjunctivism and Perceptual Experience”, 14–18. Burge, “Disjunctivism and Perceptual Experience”, 22. Burge, “Disjunctivism and Perceptual Experience”, 22. Burge, “Disjunctivism and Perceptual Experience”, 19–20. In fact McDowell mostly prefers ‘facts’; I will not go into the ontological aspects of the discussion here. 49 I am forcing the contrast here. Burge’s full story involves a difference between pattern-based representations and occurent-based representations (Burge, “Disjunctivism and Perceptual Experience”, 2); the role of normal conditions and normal world for perceptions within that framework might bring him closer to McDowell’s story about capacities. One very interesting point of Burge here is about objectood as the mark of representation in his sense: 35 36 37 38 39 40 41 42 43 4 4 45 46 47 48

I believe that what is central in distinguishig perceptual representation from mere registration, or even mere sensation, is a certain type of objectification. To count as a perceptual system, the system must have objectifying capacities. Such capacities are perhaps most vividly exemplified by what is called representational constancies. These are capacities

154  Sofia Miguens sistematically to represent a given property or object as the same despite significant variations in proximal stimulation. (Burge, “Disjunctivism and Perceptual Experience”, 10) 50 See Miguens, Sofia. “Trois perspectives sur la distinction apparence-réalité dans la expérience consciente”. Rue Descartes 63 (2010): 18–30. 51 This is one importante criticism of J. Benoist. See Miguens forthcoming, Les problèmes philosophiques de la perception. 52 For McDowell this ultimately means pre-Kantian, since Kant took Hume’s point seriously (i.e. the idea that ‘there is no satisfactory route’ from the idea that ‘no experience is intrinsically an encounter with objects’, but rather ‘glimpses of objective reality’, to ‘the epistemic position we are in’, cf. ­McDowell 1998: 344). As McDowell sees things, a Kantian viewpoint means precisely conceiving of every experience as being intrinsically an ­encounter with objects. 53 Dennett, Consciousness Explained, 364 54 Dennett, Consciousness Explained, 366. 55 As is Travis’ too (Travis, Charles, “While under the Influence”. In ­C onsciousness and Subjectivity, edited by Sofia Miguens and Gerhard Preyer, 391–412 (Frankfurt: Ontos Verlag, Reprinted in Travis, 2013)). In Criteria, Defeasability and Knowledge (McDowell, John, “Criteria, ­Defeasibility, and Knowledge”. In Meaning, Knowledge, and Reality, John McDowell, 369–394 (Cambridge, MA: Harvard University Press, 1998)) the disjunctivist approach is put forward under the guise of a Wittgenstein-­ inspired dispelling of skepticism. 56 McDowell, “Criteria, Defeasability and Knowledge”, 386–387. 57 McDowell, “Tyler Burge on Disjunctivism”, 251. 58 McDowell, “Tyler Burge on Disjunctivism”, 245. 59 McDowell, “Tyler Burge on Disjunctivism”, 245, quoting Burge ­“Disjunctivism and Perceptual Psychology”, 30. McDowell’s italic in the quote. 60 McDowell, “The content of Perceptual Experience”, 350. 61 McDowell, “The content of Perceptual Experience”, 350–351. 62 Burge, “Disjunctivism and Perceptual Psychology”, 47. 63 McDowell, “Tyler Burge on Disjunctivism”, 244. 64 McDowell, John. “Conceptual Capacities in Perception”. In Having the World in View, John McDowell (Cambridge, MA: Harvard University Press, 2009), 142. 65 McDowell, “Avoiding the Myth of the Given”, 267. 66 Benoist, Jocelyn. Le bruit du sensible (Paris: Cerf, 2013), 9. 67 McDowell, “Avoiding the Myth of the Given”, 260. 68 McDowell, John, “Are the Senses Silent? Response to Charles Travis”. Agnes Cuming Lecture II. www.youtube. com/watch?v=fBQHEGg5JSo (2013b) 69 Travis’ 2013 book Perception includes a rewritten version of the 2004 Mind article. 70 Benoist, Le bruit du sensible, 92. Benoist is citing Merleau-Ponty’s La Phénoménologie de la perception (Paris: Gallimard, 1945), 12. 71 See Haag, Johannes. “Analytic Kantianism: Sellars and McDowell on ­S ensory Consciousness”. Con-Textos Kantianos – International Journal of Philosophy 6 (deciembre 2017): 18–41 (special issue edited by Sofia Miguens and Paulo Tunhas). 72 McDowell, John, “The Disjunctive Conception of Experience as Material for a Transcendental Argument”. In The Engaged Intellect, John McDowell, 225–240 (Cambridge, MA: Harvard University Press, 2009).

Is Seeing Judging? Radical Contextualism  155 73 McDowell, “Avoiding the Myth of the Given”, 271. 74 See Shear, Jonathan, ed. Mind, Reason and Being in the World – the ­McDowell-Dreyfus debate (London: Routledge, 2013). 75 McDowell, “Avoiding the Myth of the Given”, 257. 76 McDowell, John. Mind and World (Cambridge, MA: Harvard University Press, 1994), 11. 77 Dummett, Michael. Origins of Analytical Philosophy (Cambridge, MA: ­Harvard University Press, 1993). 78 Beaney, Michael ed. The Frege Reader (London: Blackwell, 1997), 325–345, 343. Das Haben von Gesichtseindrücken ist zwar nötig zum Sehen der Dinge, aber nicht hinreichend. Was noch hinzukommen muß, ist nichts Sinnliches. Und dieses ist es doch gerade, was uns die Außenwelt aufschließt; denn ohne dieses Nichtsinnliche bliebe jeder in seiner Innenwelt eingeschlossen.

79 80

81 82 83 84 85

86

Frege, Gottlob. “Der Gedanke”. Beiträge zur Philosophie des deutschen ­Idealismus 2 (1918), 58–77. In Logische Untersuchungen, hrsg. Günther Patzig, Göttingen, Vandenhoeck und Ruprecht, 1993. Dummett, Michael. Origins of Analytical Philosophy (Cambridge, MA: ­Harvard University Press, 1993). Frege, Gottlob. Logik 1897. In Nachgelassene Schriften, edited by H. ­Hermes, F. Kambartel and F. Kaulbach (Hamburg: Felix Meiner, 1983), 137–163, 149. In the original: “Aber sehe ich denn nicht, dass diese Blume fünf Blumenblätter hat? Man kann das sagen, gebraucht aber das Wort ‘sehen’ dann nicht in dem Sinne des blossen Lichtempfindens, sondern man meint damit verbunden ein Denken und Urteilen.” Travis, Charles. Occasion-sensitivity (Oxford: Oxford University Press, 2008). In the case of the McDowell-Travis debate, Travis’ contention is that realism is on the radical contextualist side, i.e. his side. Austin, J. L. Sense and Sensibilia (Oxford: Oxford University Press, 1962), 50. Longuenesse, Béatrice. Kant and the Capacity to Judge (Princeton, NJ: Princeton University Press, 1998), 399. The interesting difference one finds here and which is very much worth ­exploring is then the difference between the questions regarding rationality (of real agents, in a world) and the questions regarding logics (abstract, and in fact alternative systems of logic). Of course neither for Travis nor for McDowell is a view of judgement a ­descrition of (subpersonal) psychological machinery at work. But there is an important contrast between the two here: according to Travis’ Frege, ­concepts are not building blocks for judgements (in the ontology arising from his view of logic and language, concepts are at the level of Bedeutung, i.e. reference, not of Sinn, sense). In contrast in McDowell’s Kantian way of seeing seemings as ‘representing’, as rational capacities operative in experiencing, inevitably involving apperception, concepts are ‘mental’ in a very different sense.

Bibliography Austin, J. L. Sense and Sensibilia. Oxford: Oxford University Press, 1962. Beaney, Michael (ed.). The Frege Reader. London: Blackwell, 1997. Benoist, Jocelyn. Le bruit du sensible. Paris: Cerf, 2013.

156  Sofia Miguens Brogaard, Berit. Does Perception Have Content? Oxford: Oxford University Press, 2014. Burge, Tyler. “Perceptual Entitlement”. Philosophy and Phenomenological ­Research 67, no. 3 (2003): 503–548. Burge, Tyler. “Disjunctivism and Perceptual Psychology”. Philosophical Topics 33, no. 1 (2005): 1–78. Burge, Tyler. Origins of Objectivity. Oxford: Oxford University Press, 2010. Burge, Tyler. “Disjunctivism Again”. Philosophical Explorations 14, no. 1 (2011): 43–80. Byrne, Alex and Heather Logue. Disjunctivism – Contemporary Readings. Cambridge, MA: MIT Press, 2009. Dennett, Daniel. “Toward a Cognitive Theory of Consciousness”. In ­Brainstorms – Philosophical Essays on Mind and Psychology, Daniel Dennett, 149–173. Montgomery, VT: Bradford Books, 1978. Dennett, Daniel. Consciousness Explained. Boston, MA: Little Brown, 1991. Dummett, Michael. Origins of Analytical Philosophy. Cambridge, MA: ­Harvard University Press, 1993. Frege, Gottlob. “Der Gedanke”. Beiträge zur Philosophie des deutschen ­Idealismus 2 (1918): 58–77. In Logische Untersuchungen, hrsg. Günther Patzig, Göttingen, Vandenhoeck und Ruprecht, 1993. English translation: Beaney, Michael ed. The Frege Reader. London, Blackwell, 325–345, 1997. Frege, Gottlob. Logik 1897. In Nachgelassene Schriften, edited by H. Hermes, F. Kambartel and F. Kaulbach, 137–163. Hamburg: Felix Meiner, 1983. Haag, Johannes. “Analytic Kantianism: Sellars and McDowell on Sensory Consciousness”. Con-Textos Kantianos – International Journal of Philosophy 6 (deciembre 2017): 18–41 (special issue edited by Sofia Miguens and Paulo Tunhas). Haddock, Adrian and Fiona Macpherson. Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press, 2008. Kant, Immanuel, Critique of Pure Reason. Translated by Norman Kemp Smith. London: McMillan, 1929 (edition quoted by J. McDowell). Longuenesse, Béatrice. Kant and the Capacity to Judge. Princeton, NJ: ­Princeton University Press, 1998. McDowell, John. “Singular Thought and the Extent of Inner Space”. In Subject, Thought, and Context, edited by Philip Pettit and John McDowell, 137–168. Oxford: Clarendon Press, 1986. Reprinted in McDowell 1998c. McDowell, John. Mind and World. Cambridge, MA: Harvard University Press, 1994. McDowell, John. “The Content of Perceptual Experience”. Philosophical ­Q uarterly 44, no. 175 (1994): 190–205. Reprinted in a slightly different form in Mind, Value and Reality, John McDowell, 341–358. Cambridge, MA: Harvard University Press, 1998. McDowell, John. Mind and World. Cambridge, MA: Harvard University Press, 1994. McDowell, John. “Criteria, Defeasibility, and Knowledge”. In Meaning, ­Knowledge, and Reality, John McDowell, 369–394. Cambridge, MA: ­Harvard University Press, 1998.

Is Seeing Judging? Radical Contextualism  157 McDowell, John, “Avoiding the Myth of the Given”. In Having the World in View, John McDowell, 256–272. Cambridge, MA: Harvard University Press, 2009. McDowell, John. “Conceptual Capacities in Perception”. In Having the World in View, John McDowell, 127–144. Cambridge, MA: Harvard University Press, 2009. McDowell, John. “The Disjunctive Conception of Experience as Material for a  Transcendental Argument”. In The Engaged Intellect, John McDowell, 225–240. Cambridge, MA: Harvard University Press, 2009. McDowell, John. “Experiencing the World”. In The Engaged Intellect, John McDowell, 243–256. Cambridge, MA: Harvard University Press, 2009. McDowell, John. “Sensory Consciousness in Kant and Sellars”. In Having the  World in View, John McDowell, 108–126. Cambridge, MA: Harvard University Press, 2009. McDowell, John. “Tyler Burge on Disjunctivism”. Philosophical Explorations 13, no. 3 (2010): 243–255. McDowell, John. “Can Cognitive Science Determine Epistemology? ­Response to Tyler Burge”. Agnes Cuming Lecture I. www.youtube.com/ watch?v=m8y8673RmII, 2013a. McDowell, John. “Are the Senses Silent? Response to Charles Travis”. Agnes Cuming Lecture II. www.youtube. com/watch?v=fBQHEGg5JSo, 2013b. McDowell, John. “Perceptual Experience: Both Relational and Contentful”. European Journal of Philosophy 21 (2013): 144–157. Merleau-Ponty, Maurice. La Phénoménologie de la perception. Paris: ­Gallimard, 1945. Miguens, Sofia. Uma Teoria Fisicalista do Conteúdo e da Consciência – ­Daniel Dennett e os debates da filosofia da mente. Porto: Campo das Letras, 2002. Miguens, Sofia. “A crítica de J. McDowell à concepção dennettiana de ­percepção”. Revista de Filosofia Aurora (Brasil) 22, no. 30 (Jan–Junho 2010): 195–214. Miguens, Sofia. “Trois perspectives sur la distinction apparence-réalité dans la expérience consciente”. Rue Descartes 63 (2010): 18–30. Miguens, Sofia, “Les problèmes philosophiques de la perception”. In Lire Le Bruit du sensible de Jocelyn Benoist, edited by R. Moati and D. Cohen-­ Lévinas. Paris: Éditions Hermann (forthcoming). Miguens, Sofia and Susana Cadilha. John McDowell – uma análise a partir da filosofia moral. Lisboa: Colibri, 2014. Shear, Jonathan, ed. Mind, Reason and Being in the World – The ­McDowell-Dreyfus Debate. London: Routledge, 2013. Soteriou, Matthew. “The Disjunctive Theory of Perception”. In The Stanford ­Encyclopedia of Philosophy, edited by Edward N. Zalta (Summer 2014 Edition). http://plato.stanford.edu/archives/sum2014/entries/perception-disjunctive/. Snowdon, Paul. “The Formulation of Disjunctivism”. Proceedings of the Aristotelian Society, New Series 105 (2005): 129–141. Travis, Charles, Occasion-Sensitivity. Oxford: Oxford University Press, 2008. Travis, Charles, “The Silence of the Senses”. In Perception – Essays after Frege, Charles Travis, 23–58. Oxford: Oxford University Press, 2013. Originally in Mind, 113 (449) (2004): 59–94.

158  Sofia Miguens Travis, Charles, “Frege, Father of Disjunctivism” In Perception – Essays after Frege, Charles Travis, 59–89. Oxford: Oxford University Press, 2013. Travis, Charles, “While under the Influence”. In Consciousness and Subjectivity, editd by Sofia Miguens and Gerhard Preyer, 391–412. Frankfurt: Ontos ­Verlag. Reprinted in Travis 2013, extensively revised. Travis, Charles, “The Preserve of Thinkers”. In Does Perception Have Content? edited by Berit Brogaaard, 138–178. Oxford: Oxfrod University Press, 2014. Travis, Charles. “Their Work and Why They Do It”. In Context, Truth and Objectivity, edited by David Zapero and Eduardo Marchesan. New York: Routledge, 2018.

8 Externalism and Context-Sensitivity David Zapero

Much work in contemporary philosophy of mind has been dedicated to analysing a particular kind of error that we can expose ourselves to when thinking about the world. Roughly, the scenario is the following. It can turn out that a given object doesn’t satisfy the specifications with which we sought to make it available to thought but that, despite this, our thought, on that occasion, was about the relevant object. In such a scenario, the specifications that we use to bring an object into view don’t just lead us astray; they don’t just end up bringing into view an object that is different from the one that we sought to bring into view. If they had done that, the object that we would have thought about would have satisfied the specifications we drew on; it would just not have been the object that we were intending to bring into view. That is, we would have failed to draw on the appropriate or right specifications, namely the ones that would have allowed us to think about the object we were hoping to think about; but the specifications themselves would not be wrong. (Obviously, this presupposes that there is a way of determining, independently, what we were “intending” or “hoping” to think about. I leave aside the difficulties that surround this issue, or, rather, the difficulties tied to this way of formulating the issue.) The scenario that has attracted so much interest is different in this respect. In such a scenario, we also fail to draw on the right specifications – those that the “intended” object would satisfy – but we succeed, nonetheless, in thinking about the relevant object. And this then means that the specifications are wrong in a more obvious sense. It turns out that the object that the specifications bring into view doesn’t satisfy those very specifications. There is a similar kind of scenario for concepts. It can turn out that the things which we sought to think about with the help of a certain concept don’t, in fact, satisfy the principles which guide our application of that concept – and yet, for all that, our concept does allow us to think about the relevant things. In that kind of case, the principles which guide our application of the concept – and which we might have believed provide a definition of the concept – turn out to be wrong in a similar sense than in the scenario described above. It doesn’t just turn out that the concept applies to things that aren’t those that we believed the concept

160  David Zapero applied to (or if nothing satisfies the relevant principles, that the concept applies to nothing at all). The concept does, in fact, allow us to think about the things we were hoping to think about – say, samples of gold. It just turns out that that the principles that guided our application were wrong: the things that fall under the concept don’t, in fact, satisfy those principles (e.g. they aren’t yellow in their pure state). That is, we succeed to pick out samples of gold with our concept of gold, it just turns out that our ‘definition’ of that concept was wrong. One may, of course, deny the very possibility of there being such e­ rrors. One way of doing so is by denying that, in the aforementioned scenario, we are in fact thinking about the relevant object. One may consider that,  on the level of thought, there can’t be the kind of mismatch that we have presented – a mismatch between the specifications used to bring an object into view and the object actually brought into view. One may retort: we can talk about objects that we don’t discriminate correctly, but no analogous failure is possible in thinking.1 This isn’t, however, the predominant position on these issues. For most authors, the errors that we have described are possible, and they are testimony to a particularly profound way in which our thought is world-involving. Yet, most authors do hold that such scenarios are a reason to believe that our conceptual or discursive capacities play less of a role in enabling us to engage in thought than we may have otherwise considered. They hold that, in the kind of situations which are susceptible to the error outlined, what enables us to think about the relevant object is not – or not p ­ rimarily – our discursive capacities but the existence of an ­informational or causal connection. Only if one assumes that our discursive c­ apacities enable thought in such situations – they claim – do the ­situations seem enigmatic. Not all authors that acknowledge the possibility of such errors have however drawn that conclusion. There is also a set of authors for whom the existence of such errors is not so much a reason to restrict the role that conceptual capacities play in thinking as a reason to reassess our ­account of how such capacities operate. 2 In the present essay, I explore how this way of explaining the world-involving character of our thought can lead to, and converge with, certain contextualist ideas. Charles Travis is the author who has done most to bring out these connections and my essay thus constitutes a reflection on this strand of his work. The purpose is not to vindicate the relevant way of understanding the world-involving character of thought but to see how it converges with certain ideas about context-sensitivity that might at first seem quite remote.

World-Involving Capacities An answer to the question of what falls under a given concept obviously depends on what things are to be found in the domain to which the concept is applicable. What falls under our concept of being red, say,

Externalism and Context-Sensitivity  161 depends crucially on what red items the world provides us with. Perhaps that if the Paris commune of 1871 had been victorious there would, at least in France, be quite a few more flags which fall under that concept. What falls under the concept of being red depends on what the world has on offer, and that of course depends on historical developments. But acknowledging this triviality about the relation between our concepts and the world still leaves open another, closely related issue, namely the question of when something counts as falling under a certain concept. This, one may consider, doesn’t in any way depend on how the world is. Indeed, one may go a step further and even deny that there is here a separate issue to be attended to. When something counts as falling under a given concept, one may say, simply depends on the state of the relevant thing. Something counts as falling under the concept red when it is, in fact, red. The question of whether it is relevant, on a certain ­occasion, to treat something as falling under the concept of being red just boils down to the question of whether the thing in question does in fact fall under that concept. We are simply led back to the previous issue of whether something falls, or doesn’t fall, under a given concept. But the denial that there is here a separate issue – we have tentatively framed that issue in terms of when something counts as falling under a given concept – can be understood as being itself a stance on the matter. Presumably, the question of what it takes for something to fit a given concept is not an unintelligible one. It can just seem to be an entirely banal question if one considers that an answer to that question must be internal to the concept at stake. Indeed, on a widespread conception of what a concept is, it is simply part of a concept’s being the one it is that it should entirely settle the issue of when something counts as fitting it. The concept of being red settles, on such a view, just when a certain object counts as being red. Or perhaps the concept of being red doesn’t achieve that task, because it is too coarse or vague, but a more finegrained concept does – that, for instance, of being a particular shade of red. What is required for an object to count as being that shade of red is – on the view in question – settled entirely by the concept of being that shade. It’s a part of that concept that it should resolve all questions as to what counts as fitting it. That then also means that, on this view of concepts, the world plays no role in such matters. The facts as to when something would fit a given concept are what they are independently of how the world turns out to be. Whether reality turns out to be one way or another, reality has no impact on what it amounts for something to fit a given concept. The concept of being (a shade of) red settles the matter from the outset, so to speak. To the extent that it is well delimited, the concept entirely determines when something is red. Indeed, once one allows for that idea, it is quite natural to deny the very existence of a separate issue. Since one considers that it is intrinsic to a concept that it should determine when

162  David Zapero something fits it, the “when” question is reducible to the “whether” question. There is then no separate issue to be dealt with. This view is closely associated to the notion of an extension, and an adherence to the view often manifests itself in an appeal to that notion. 3 The idea that concepts have extensions – a determinate set of objects that fall under the concept given the state of the world – can seem just as innocent as the “when” question can seem trivial. For a concept to have an extension just is, one may say, for there to be a definite set of things, at a given time, which have the property that the concept is a concept of. The reasoning behind this being: what falls under a given concept just depends on two things, namely on the concept and on how the world is at a given time. Given how the world is at any given time, a concept delimits a determinate set of objects. The idea of an extension only becomes problematic if one allows for an idea along the following lines: what counts on some occasions as having a certain property doesn’t, on other occasions, count as having that same property. Someone who has died her hair red may, on some occasions, but not on others, count as having red hair. In those different situations, we of course deploy the notion of having red hair differently: in one case it may refer to the person’s natural hair color, in another to something like the person’s currently perceptible hair color. But one may consider that while “having red hair” is understood differently in the two different cases, this difference need not be one between two concepts. Indeed, one may consider that the same concept – the concept of having red hair – is deployed in the different situations, but that different things count as falling under it. If one allows for such an idea, concepts cannot come with extensions attached to them for every state of the world. (They cannot be Carnapian intensions.) For what falls under a given concept depends on the occasion of the concept’s deployment. Whether this means giving up the commitment from which the idea of an extension is usually thought to derive – the commitment to the idea that falling under a concept should be a two-party enterprise, depending solely on the concept and the world – we leave open for the time being. The important point is simply this: the very idea of an extension is wedded to a view of concepts on which the world has no role to play in deciding when something counts as fitting a concept. It is important to note, particularly for our purposes, that this issue about concepts can be recast as an issue about thoughts, that is, as an issue about the kind of stances that are true or false. The relation between a concept and what falls under it (or doesn’t fall under it) is of course fundamentally different from the relation between a thought and what makes it true (or false). But one could hold that those two relations have something in common, namely that in both cases it is a relation between something general and something particular. Just like the relation of falling under a concept is a matter of a certain generality applying to

Externalism and Context-Sensitivity  163 particular objects, the relation of instantiating a thought (of making it true) is a relation in which a certain generality (the thought) reaches to certain particular states of affairs. I leave aside the various issues that such a parallel raises. We’ll come back to some of these issues later. The important point here is that, by drawing on this parallel, the view of concepts that we have presented could be recast as a view about thoughts. The view would take, roughly, the following shape: the world plays a crucial role in deciding whether a certain thought is true (or false), but it plays no role at all in determining when a thought counts as being true (or false). What it involves, for a thought, to be true (or false) on a given occasion is entirely independent of how the world turns out to be. One way of calling into question this view is by inquiring into the ­nature of the capacities on which we draw when deploying a concept. One may ask whether there are particular capacities that are indispensible in deploying concepts – or whether, in principle at least, one could do without any given capacities. Put somewhat differently: would it be possible to construct a theory, which would allow someone to correctly apply a given concept without possessing any given capacity? Can, for instance, the visual recognition of a person be accounted for by a ­psychological theory in such a way as to allow someone who doesn’t possess any conceptual capacity to deploy the relevant concept? Does the empirical psychologist, in his theory of visual recognition, provide a theory of a conceptual capacity? Presumably, such a psychological theory must satisfy at least two ­major constraints. Firstly, for us to be able to judge whether the theory succeeds in explaining the visual recognition of certain objects, we must specify just which objects or individuals are at stake – or, put somewhat differently, what counts as being that object or individual (in a variety of different possible scenarios). Otherwise we would lack a standard to which we can appeal when deciding whether the theory is successful, whether in a given situation an object is in fact the object to be recognized. In other words, the theory must specify what the capacity is meant to achieve in a way that doesn’t rely on that very capacity. Failure to provide such a specification would mean that the explanandum of the theory hasn’t been determined enough to allow us to judge when the theory would have (correctly) accounted for the recognitional capacity. Secondly, the workings of the capacity must be specified by a set of principles the application of which doesn’t require the possession of that very capacity. This is simply for the theory to live up to its ambitions of accounting for the transition from what there is in the environment to the recognitional capacity’s output, namely its identification of the relevant object. If the theory is to account for that transition in a non-­ circular manner, the principles it provides us with must allow us to make that transition without drawing on the capacity in question. In other words, the principles have to be something like functions whose

164  David Zapero arguments are environmental factors, which we can identify without ­relying on the capacity, the deployment of which those very functions are meant to explain. For instance, if the recognitional capacity is one of identifying Pablo Iglesias, then it must rely on features the identification of which is available to creatures that don’t have the capacity to identify that individual. Such things as beards and long hair may be candidates of what that theory can rely on. It is just as important, though, to acknowledge what such a theory need not do. Since what matters, for such a theory, are simply the ­results – accounting for a certain capacity’s outputs – there is no need for it to rely on the essential features of the object in question. If the aim is to identify rhinoceros, the principles need not rely on the features that are essential to being a rhinoceros (whatever they may be). That means that the theory need not be able to account for the capacity’s workings in environments that are hostile to the capacity because rhinoceros have all taken on a radically different appearance (while still remaining ­rhinoceros). The theory need not exclude the possibility of error in such extreme cases: it seeks merely to be successful in normal environments. Of course, one question that is important for our purposes is whether such a theory could, potentially, capture such essential features. One could concede that such theories don’t usually operate on the basis of essential features but consider that they could, potentially, specify functions that work even in inhospitable environments. That is, one could consider that recognitional capacities could (potentially) solely draw on features that are essential to the thing in question. If that were indeed possible, one could exclude the following scenario: something qualifies as a rhinoceros according to the theory in question because it fulfils all of a set of criteria but it isn’t, for all that, a rhinoceros. If the theories that account for recognitional capacities were indeed able to do this, the conceptual capacities we draw on when deploying a concept would be eliminable. It would turn out that a recognitional capacity of the sort studied by the psychologist works just like a proper conceptual capacity. The important point here is that, on one understanding of what ­concepts are, a concept is entirely immune to unforthcoming environments. Even in environments where the appearance of rhinoceros has radically changed, only rhinoceros fall under the concept of being a ­rhinoceros. So even if all rhinoceros have been dehorned and disguised so as to make them thoroughly unrecognizable to poachers – and all other people, for that matter – they would still all fall under the concept of being a rhinoceros (if indeed they were still rhinoceros). In fact, to say that concepts are immune to inhospitable environments is misleading: what falls under a concept – on this understanding of “concept” – just doesn’t depend at all on the forthcomingness of the environment. For, on this understanding, concepts are individuated in terms of a way of being. The concept of being a rhinoceros is individuated in terms of a

Externalism and Context-Sensitivity  165 certain way for something to be, namely that of being a rhinoceros. The more traditional term here would be that of a property: concepts are of a certain property. One may of course use the term “concept” differently. But if one uses it this way, then the question of what counts as the same concept is decided simply by what is one (same) way for something to be. Whether it is a dehorned Namibian rhinoceros or some other extravagant exemplar unrecognizable as a rhinoceros to most mortals doesn’t change the matter: if it is a rhinoceros, it falls under the concept of being a rhinoceros. Reichenbach emphasized one corollary of this point in advancing the following claim: if one relies on a given set of presumptions in deciding what fits a given concept, one better hope that those presumptions are true – but there is no way of ruling out entirely the possibility that they should be false. An example would be the concept of being six feet long.4 A presumption that one relies on when deploying that concept is that objects do not change length merely by changing spatial position. Indeed, if such changes did take place, it would be quite indeterminate what a measurement of six feet – or any such measurement of length, for that matter – amounts to. One would lack a certain constancy that is presupposed in speaking of measurement as we do. Yet, one can hardly claim that the presupposition in question is necessarily true: it could have turned out that space has a geometry that makes it false. So while that presumption had better be true, it could turn out to be false. The point made by Reichenbach is closely related to the idea that ­concepts are of a way for things to be. If concepts are individuated in terms of a way for things to be, there is, at least potentially, room for the possibility that the principles that govern our application of a given convcept should turn out to be false. Since what counts is whether something is one same way of being, there can be a divergence between what we take to be characteristic of that way of being and what is actually characteristic of that way of being. It can turn out, for instance, that there is no one way for things to be that satisfies all the principles that we rely on when deploying a certain concept. Such was, notably, the case of the concept of gold. There was no one metal, it turned out, which had the various properties which we once thought. It turned out that gold, in its pure state, doesn’t have a golden color at all, but rather something close to grey. Of course, in the case of gold, it turned out – or so most people would claim – that there was a certain metal that we talked about both after and before we knew it not to be golden-colored. It just turned out that certain presumptions about that way for things to be were wrong. Another possibility would be that our presumptions should not be satisfied because there was no one way for things to be. In that case, a putative concept turns out to be a bogus concept. It isn’t that certain presumptions are false, but rather that there is nothing that that concept is a concept of.

166  David Zapero One may deny that there is indeed room for the kind of divergence that Reichenbach was interested in. The denial can be grounded in a thought along the following lines: nothing obliges us to introduce amongst the presumptions guiding the application of a concept the supposition that the concept applies to the same “stuff” or kind of thing. Indeed, what bars us from customizing our concept so as to ensure that there isn’t the kind of divergence that Reichenbach is interested in? What bars us, for instance, from introducing a concept of gold (call it “gold*”) which solely depends on a certain set of phenomenal qualities – those that we attributed to gold before discovering that it wasn’t golden in its pure state? The discovery that there is no one metal which in its pure state satisfies all the relevant properties wouldn’t make any of our presumptions false because, according to this proposal, gold need not be one given metal. Gold would be whatever satisfies a certain set of phenomenal criteria, namely those according to which we used to judge whether something was gold. Of course, this isn’t how people thought about gold in the eighteenth century. They had the expectation that the different samples of gold were all samples of one kind of stuff. But that doesn’t matter for the purposes of the current proposal. The proposal is meant to call into question the idea that there is always potentially a gap between what a concept is a concept of and the presumptions that guide our application of the concept. To call into question that possibility, one may introduce the idea of a concept that is of a way for things to appear. All the items that look like what we took to be (pure) gold at a certain point in time would be gold*. The discovery that different samples were actually different substances or that each sample was a mix of different substances wouldn’t make any of our presumptions about gold* false since the only thing that matters is how something looks to us – and that can’t (allegedly) be overturned by any discovery. This is of course a certain variation on a traditional defence of analyticity. What hinders us from stipulating a concept so that there can be no divergence between what the concept is a concept of and what we take the concept to be a concept of? Nothing forces us to assume, for instance, that everything that looks a certain way is, in fact, one same substance. So why not allow for a concept that foregoes such assumptions and thereby ensures that the presumptions guiding our application of the concept also fix what the concept is a concept of? We will return to the issue below. For the time being we leave it aside to inquire into the consequences of allowing for such a possible divergence. One conclusion one may draw from the existence of scenarios of the kind that Reichenbach pointed to may be the following: whenever we discover that all our presumptions about a given concept aren’t satisfied, it turns out that the putative concept was not in fact a concept at all, but rather a pseudo-concept – a Scheinbegriff. This is the position that Feyerabend adopted in the early 1950s. If it had turned out that objects

Externalism and Context-Sensitivity  167 vary in length when they change their spatial position, we would find out that certain concepts of measurement were in fact pseudo-concepts. If the theory that guided the application of a concept turns out to be wrong it means that there was no one way for things to be. So it means that we weren’t in fact dealing with a concept at all: we were all along operating with a pseudo-concept. There is, though, as we’ve seen, another lesson that one may draw from Reichenbach’s idea. One may allow for the possibility that certain presumptions about the application of a concept should turn out to be wrong. In allowing for that eventuality one takes the concept to be an authentic one, but one simply considers to have been wrong, in some respects, about what the concept was a concept of. Thus, the concept is the same before and after the discovery in question – we just learn something about the relevant way for things to be. On this view, the possibility that certain of our ideas about gold turn out to be wrong doesn’t mean that our concept of gold wasn’t a concept of gold all along. It was just that we just had some wrong ideas about what it was a concept of. The example of gold of course goes back to Leibniz, who presents us with the following reflections in his Nouveaux Essais: For example, a way might be found of counterfeiting gold so that it would pass all the tests we have so far; but one might then also discover a new assaying method which would provide a way of distinguishing natural gold from this artificial gold. There are some old documents which attribute both of these discoveries to Augustus, Elector of Saxony, but I am not in a position to vouch for this. If it did happen, however, it could lead us to a more perfect definition of gold than we have at present; and if artificial gold could be made in large quantities at low cost, as the alchemists claim it could, this new test would be important, because it would enable mankind to retain the advantages which natural gold has in commerce, because of its rarity, in providing with a material which is durable, uniform, easy to divide and to recognize, and valuable in small quantities.5 We’re presented with a situation in which, after having found a method to counterfeit gold, we discover an experiment to distinguish the counterfeited material from natural gold. Now while Leibniz doesn’t pronounce himself explicitly on the matter, his presentation of the situation leaves little doubt that for him the concept of gold was, all along, the concept of a certain substance which we learned how to identify better when we discovered an experiment to distinguish it from a “phenomenal equivalent”, namely the counterfeited gold. So, like Putnam, who rehabilitates this example, Leibniz seems to hold that there can be a divergence between what the concept is a concept of and what we take it to be a concept of. That we shouldn’t have been

168  David Zapero capable, during a certain time, to distinguish between the counterfeited gold and the natural stuff doesn’t mean that applying the concept to the counterfeited samples wasn’t illegitimate. Certain phenomenal qualities and scientific procedures guide our application of the concept but they don’t fix irrevocably what the concept is a concept of. That’s just done by a way for things to be. And the fact is that the counterfeited material, while phenomenally similar or perhaps even equivalent to gold, wasn’t the same stuff as natural gold. So that even if we were incapable of distinguishing it from gold for a while, this didn’t make it legitimate to take it to be gold. Of course, the question that now arises – and it is closely related to the previous one we raised, about analyticity – is what determines just what is one same way for things to be. In the case of natural kinds such as gold this seems fairly obvious. Our talk about gold, one may say, was talk about a certain kind of stuff and the identity of stuff is a matter of atomic number. But even in these cases, one may raise the question: what allows us to sacrifice certain criteria that are central to our idea of what a certain concept is a concept of – forego, for instance, the idea that ­yellow is characteristic of the things that fall under the concept gold – and consider that the concept we end up with is the same concept as what we had before? We are working with the supposition that concepts are of a way for things to be – so if the concept gold is the concept of a certain metal, then, even before any of the scientific discoveries are made, what the concept is a concept of must clearly be defined in terms of atomic number. But what allows us to say that the concept we deployed before making the discoveries was a concept of the stuff we later identified as having a certain atomic number? An initial answer to the question might be that we generally applied the concept of gold to a certain metal with the atomic number 79. But that can’t constitute a complete answer. Why should we consider that, even though the concept doesn’t any longer satisfy certain constraints that guided our application of the concept, it is still the concept of the same stuff? That is, what allows us to decide that the fact that we would only apply the concept to a certain golden-colored solid is not relevant but the fact that that solid was (mostly) composed of a certain element is? The question is, in other words, what allows us to decide, when nothing fits all our presumptions guiding the application of a concept, whether the concept is simply bogus or whether some of the presumptions are wrong. How many presumptions can we discard before we must accept that we are no longer dealing with the same concept? How much can turn out to be wrong about how we determined a certain way for things to be before we admit that there was no such way for things to be? Drawing on the approach that Putnam adopts to these issues, Travis has advanced the following claim. Just what counts as one same way for things to be depends on whether what we speak of, once we have

Externalism and Context-Sensitivity  169 dropped all the presumptions that we cannot retain, is still r­ ecognizable as the same thing we spoke of before dropping those presumptions. Whether the concept of gold before the aforementioned discovery is the same concept as after that discovery depends on whether what we are left with, once we recognize that no relevant metal is yellow in its pure state, is still recognizably gold. Thus, on such a view, there is no formal account or theory to be given of these matters. As Travis puts it, these questions, about what counts as one way for things to be, is “theory-­ resistant”. The matter can only be decided on a case-by-case basis, by looking at whether what we are left with, once we have dropped certain suppositions, is still recognizable as the same thing. For the case that we’ve been discussing, this means: whether talk about gold is still recognizable as talk about the same thing as what we talked about before we discovered that gold, in its pure state, isn’t yellow. Since such a response to the problem appeals to the idea of a shared sense of what is the same, one may have the impression that it eludes the real issue. The problem, one may insist, is just how to determine what counts as one way for things to be. The issue is the identity of the way for things to be. What we find reasonable or recognizable is another matter – and can hardly be offered as a response to the problem. Yet, it is crucial to note that Travis does not seek to change the subject of the conversation by appealing to a shared sense of recognizability. The appeal to that sense is meant to be an answer to the issue that we just outlined: what we can count as recognizably being the same is meant to be the way in which we can decide what is essential to a certain way for things to be. The thesis is of course not that our sense of what is recognizable determines what is the same, but rather that that sense allows us to establish what is, in fact, one same way for things to be. Whether and how such a sense of recognizability could allow us to do that is, of course, something that has to be shown.

Generality Unconstrained We have recast the question of whether the world plays a role in ­determining when something fits a concept as a question about the relation between conceptual capacities and the recognitional capacities studied by the empirical psychologist. If conceptual capacities are irreducible to such recognitional capacities, then, for any set of principles that one may provide to account for what fits a concept, it is possible that an object not fall under the concept even though, according to the principles, it should (or vice versa). If indeed it is possible that a recognitional capacity to recognize rhinoceros should, when functioning correctly, identify something as a rhinoceros which, nonetheless, is not a rhinoceros, the world has a role to play in deciding when something falls under a concept. For a concept, in the sense that we are interested in, is

170  David Zapero constitutively of a way for things to be. The fact that something should be hard to recognize as a rhinoceros or that it doesn’t satisfy a given set of presumptions of what we think is a rhinoceros doesn’t change the fact that, if it is a rhinoceros, it falls under the concept of rhinoceros. Thus, any possible divergence between recognitional and conceptual capacities implies that what it amounts to fit a concept cannot be determined ­without taking into account how things actually are. Now the irreducibility of conceptual capacities depends, most ­crucially, on two issues. There is, firstly, the issue of whether we can forge a concept which excludes the possibility of a clash between the presumptions guiding its application and the characteristics of that which the concept is a concept of. Any such concepts would succeed in making  the world irrelevant in determining what amounts to something’s fitting the concept. The issue here is one about what forces us to revise, in certain cases, the presumptions that guide our use of a concept. It may turn out that there is no one metal that satisfies all the criteria that we held, up to a point, to be true of gold. But what bars us from having a concept which doesn’t presuppose that everything that falls under it is of the same stuff? Couldn’t we, in that case, continue using a concept irrespectively of any discoveries we may make? The issue is closely related to another one about how to individuate that which concepts are of, namely ways for things to be. The irreducibility of conceptual capacities to recognitional ones depends, as we saw, on the impossibility of determining, in terms of a set of principles, what it amounts for something to count as the same way of being. This claim about the “theory-resistance” of conceptual capacities may also be put in the following terms: how the way things are articulates into particular ways for things to be depends on the occasion for distinguishing such ways. There is no “general” answer to the issue because what counts as one particular way for things to be can only be established in the context of a concrete situation. Let me make some remarks about this formulation; I’ll return in more detail to the central notions in the last section. “The way things are” refers here, in a generic fashion, to whatever allows to decide about the value of a truth claim. It is whatever plays just that role: whatever allows us to decide whether things are, in fact, as a given truth claim purports them to be. If a truth claims purports things to be thus and so (Barack to be smiling), then “the way things are” is what settles whether that is indeed the case (that is, whether Barack is indeed smiling). “World” or “reality” may be alternative terms for this notion. In contrast, “ways for things to be” are that which concepts are of. It is a range of situations or things that are the same with respect to a certain feature, which is precisely the relevant way for that range of things to be. A more traditional term here may be that of property. Concepts, as we’ve been using the term, are of a certain property. And properties, or ways for things

Externalism and Context-Sensitivity  171 to be, possess a certain generality: they are, precisely, a way of grouping a range of things. That is, the grouping isn’t accountable in terms of the particular items that, in fact, compose that group. The concept of rhinoceros delimits a certain set of animals in a way that is irreducible to what, in fact, are the members of that group. The way that that concept delimits a set of animals is irreducible to what animals in fact fall under it: it remains the same concept whether there are more, less or no rhinoceroses. Indeed, it groups the same way even if an indefinite set of other things about the world were different. Concepts are, in that respect, categorially different from what falls under them. They always transcend any set of particular things by providing us with a way of grouping an indefinite number of particular things. Of course, the idea that what counts as a way for something to be depends on the particular occasion for judging the thing in question is perfectly trivial. Indeed, put in this general way, the idea is an utter banality. Whether it is true to say that the Danube is blue of course depends on the context in which one speaks. The same stretch of river on the same sunny afternoon may count as blue in the context of certain conversations and not in the context of others. The same piece of pain perdu might count as burnt in one context and not in another – depending, for instance, if it was prepared by the restaurant chef or his three-year-old daughter. But such variations are, for most people, a reason to distinguish between a term and a concept. “Burnt”, they will insist, doesn’t refer to the same concept on the two occasions. Indeed, how should it? If a concept is a way of delimiting a set of objects, then we clearly cannot say that the same object both falls and does not fall under the concept. On this understanding of the matter, concepts serve to disambiguate. The term “burnt” is ambiguous to the extent that it can refer to different concepts: on one occasion we employ it to give expression to one concept of being burnt (a lenient variant), on another we employ it to give expression to another (a strict variant). This line of reasoning has a parallel on the level of whole thoughts. Thoughts are conceived as enabling us to overcome an analogous kind of ambiguity. One and the same sentence about a certain pain perdu can be both true and false in different contexts because different thoughts are expressed by that sentence. Thoughts allow to distinguish between two different sets of truth-conditions – and thereby allow to account for the fact that one and the same situation can be said to be a case of pain perdu being burnt and not a case of pain perdu being burnt. Thoughts serve to extract content from circumstances. We’ll reserve the term “proposition” for this conception of a thought: propositions are what allow us to fix a definite set of truth-conditions where language fails (or seems to fail) to do so. The view of concepts and thoughts that we are interested in can be ­understood as resulting from a denial that concepts and thoughts serve to disambiguate in this way. On the view that we are interested

172  David Zapero in, concepts and thoughts cannot help to solve the ambiguities that one finds (or rather, that one can have the impression exist) in language. Not only can they not help to disambiguate, it is not their business to perform any such task. To that extent, the ambiguities that one finds at the level of language carry over into thought. Thought can’t help resolve whatever problems language raises. (“No interpolation between a sign and its fulfilment does away with the sign.”6) Which of course still leaves open the question of whether language does indeed raise the relevant problems. That is, whether the context-dependence of our understanding of words used on an occasion is indeed a problem in the way the ­“propositionalist” view thinks it is. If it is, then thought can’t help resolve it – that’s the claim that we are interested in here. The claim that the “ambiguities of language” carry over into thought may initially seem absurd. How could those problems re-emerge if propositions are designed to disambiguate the relevant cases? A proposition, we are told, is true just where the world is thus and so. How could there be an ambiguity as to whether the proposition is true under certain conditions if it is defined in terms of such conditions? The proposition is just a way of delimiting a range of situations, namely those that make the proposition true. There is, one may insist, no room for questions about just what the truth-conditions of a proposition are. The idea of delimiting a range of situations may perhaps be ­unproblematic. But the important point is that propositions are meant to eliminate the kind of ambiguities that one can find at the level of words. And that involves more than just delimiting a certain range of situations. Indeed, propositions are meant to disambiguate between different ways of understanding a sentence by distinguishing between different sets of truth-conditions, which the relevant sentence could serve to express. That way, once it has been determined which proposition has been expressed on an occasion, it has been determined when what was said would be true. “When what was said would be true” involves specifying independently of the occasion on which the words were used when what was said would be true. Which is why the relevant determination has often been construed in terms of equivalences of the form “the proposition that p is true if and only if p”. The idea is: what counts as a situation in which what was said is true is specified independently of the occasion in which one speaks that way. One may thus distinguish between two ideas. On the one hand, there is the idea that propositions allow to disambiguate by delimiting a determinate range of situations in which what was said, by using a certain sentence, was true. On the other hand, there is the idea that that delimitation can be done in a certain way, namely that it can be specified independently of the occasion on which the relevant sentence is used. It is the combination of these two ideas that gives rise to the idea of a proposition that we are interested in. Propositions are meant to achieve

Externalism and Context-Sensitivity  173 their task of disambiguation in a very particular manner: by specifying, independently of the occasion on which a sentence is deployed, when what was said, by using that sentence, would be true.7 The problem is not that propositions involve a picture of truth which includes no reference to language – if by language we mean such things as words and sentences. Rather, the problem is that that picture seeks to abstract in a certain way from the fact that only what we say, on certain occasions, can be true or false. (“What is true or false is what human beings say; and it is in their language that human beings agree. This is agreement not in opinions, but rather in form of life.”8) Of course, those who appeal to such propositions don’t seek to deny that it is what we actually say, in given situations, which is true or false. Trivially, to determine what someone said on an occasion, we have to take into account certain ambient factors. If, for instance, we don’t have a particular cat in view, we may not be able to determine what someone says when affirming that Garfield has made a mess on the living room table. But one can allow for this triviality and still consider that it is possible to abstract in a certain way from the fact that saying something is always saying it on a particular occasion and in a particular manner. One may acknowledge the banality and still consider that those facts, about who says what where and how, are just the shackles of the quotidian. One may consider, most crucially, that there is a way of specifying what is said on an occasion that allows to determine whether such would also hold in other possible situations. That is what a proposition is meant to do: specify what is said in such a way that it provides a general condition the application of which allows to decide whether what was then said holds in other situations. Travis’s claim is that no general condition can accomplish any such task: for a general condition of that kind and a given state of the world, there is no way of determining whether the condition has been fulfilled. The crucial point is that the facts that are relevant for making the decision may not always be the same. The facts that allow one to state on some occasions that the condition holds may not allow one to state that on other occasions. Different facts or considerations may be relevant on different occasions, and what makes the condition hold on some occasions may not do so on others. It should have transpired that this point is, in a different guise, the claim about the theory-resistance of conceptual capacities that we have discussed previously. We have given some quotidian examples, but ­Wittgenstein explored the point in detail, perhaps for the first time, in a mathematical context. One example that he returns to on various occasions is the decision of introducing imaginary numbers, i.e. the decision of extending the square root to negative numbers. Surely √–1 must mean just the same in relation to –1 as √1 means in relation to 1! This means nothing at all.9

174  David Zapero There is, naturally, something in common to any pair of numbers of which the first is the square root of the second. One thing all such pairs have in common is that the second number of the pair is the result of multiplying the first number by itself. But once we reach −1, nothing of what we know about what is required for a number to be the square root of 1 tells us what is required for a number to be (or not be, for that matter) the square root of −1. The considerations that allow us to decide whether a certain number stands in the relevant relation to −1 do not follow from any general account of what the square root relation is. Whether there is a square root of −1 is underdetermined by any general account we may have of what the relation amounts to. It rather depends on the kind of arithmetic we choose to do. This means, in the present context, that the problems about ambiguity that propositions were meant to solve cannot in fact be solved by such a notion. The questions about what it amounts for certain words to be true carry over into thought. Thought cannot help eliminate those questions, because at the level of thought the same kinds of questions emerge. Grasping a thought also requires a certain sensitivity to the ­surroundings – and only those equipped with the relevant sensitivity will be able to grasp what it amounts for things to be as a given thought represents them. Grasping or entertaining a thought is knowing when things would be as that thought represents them (i.e. knowing when that thought would be true). So Travis’s claim amounts to the following point: such grasping or entertaining requires an appropriate sensitivity to the kind of facts which, on an occasion, are relevant to deciding whether things are as the thought represents them. More specifically, to entertain a particular thought to the effect that a certain object a is a certain way G, it will not suffice to possess the concept G and the capacity to identify the object a. Those two abilities are not enough to entertain the relevant thought because they don’t allow us to distinguish between the thought in question and other thoughts equally about a and about being G. Indeed, Travis’s claim involves calling into question the idea that thoughts have an intrinsic structure by which they can be identified. There are various thoughts to the effect that a is G, each of which differs as to when things would be as it represents them. All those different thoughts are thoughts of a’s being G – they just differ in what they count as being a and in what they count as something’s being G. To entertain one of those thoughts, one thus needs to possess the right ­understanding of what it amounts for something, namely a, to be G; one needs to grasp the understanding that is at stake in the particular thought in question. The same goes for the capacity of identifying a. One must have the relevant understanding of what it would amount to for something to be a. Only when one has this relevant understanding of what makes something a and what makes something G will one have a capacity to

Externalism and Context-Sensitivity  175 entertain or grasp the relevant thought to the effect that a is G; for only then will one have distinguished that thought from other ones to the ­effect that a is G. Only having the capacities to identify a in general and to know, in general, what it is for something to be G will not suffice. One will be able to grasp the relevant thought only when these capacities are complemented by an appropriate sensitivity to the surroundings. The idea that thoughts have an intrinsic structure has been put f­ orward most eloquently by Evans, who introduces that idea in his Varieties of Reference under the title of the Generality Constraint. That Constraint is meant to cash out the idea of an intrinsic structure by accounting for the grasping of a thought as a joint exercise of distinct abilities. I should prefer to explain the sense in which thoughts are structured, not in terms of their being composed of several distinct elements, but in terms of their being a complex of the exercise of several ­distinct conceptual abilities. Thus someone who thinks that John is happy and that Harry is happy exercises on two occasions the conceptual ability which we call possessing the concept of happiness’. And similarly someone who thinks that John is happy and that John is sad exercises on two occasions a single ability, the ability to think of, or think about, John.10 Evans proposes to think of the abilities at work in an elementary thought, to the effect that a certain object falls under a certain concept, in the ­following way: When we say that a subject’s understanding of a sentence, ‘Fa’, is the result of two abilities (his understanding of ‘a’, and his understanding of ‘F’), we commit ourselves to certain predictions as to which other sentences the subject will be able to understand; furthermore, we commit ourselves to there being a common, though partial, explanation of his understanding of several different sentences. […] Thus, if a subject can be credited with the thought that a is F, then he must have the conceptual resources for entertaining the thought that a is G, for every property of being G of which he has a conception.11 The Constraint thus involves the following claim: if one possesses a ­concept G, one will be capable of entertaining thoughts, about any of a range of objects with which one is acquainted, to the effect that that object is G. Notice that Evans formulates the claim as a predictive one: the claim involves the prediction that someone who understands a thought to the effect that a is F will also be able to understand a thought to the extent that b is F. There is no room for possessing the concept F and then failing to understand what that amounts to when applied to some other object that one is acquainted with.

176  David Zapero Indeed, the claim is a substantial one because the capacity in ­question – to apply a concept to any of a range of objects – is not part of the definition of what it is to possess a concept. Possession of a concept is defined independently. I possess a concept G if I am able to entertain a thought to the effect that some item is G. The Constraint amounts to the idea that if I possess a concept in that sense, if I can think of some item that it is G, I will also be able to think of an indefinite number of other objects with which I am acquainted that they are G. The same is applicable to what Evans calls Ideas of objects. If I am able to think about some object a that it is a given way, then I must also be able to think of that object that it is any other of an indefinite set of ways for something to be – I must also be able to subordinate it under other concepts that I possess and that are applicable to that object. The Constraint thus involves the idea that possession of concepts (and Ideas of objects) can be defined independently of the deployment of such a capacity in the grasping of whole thought. Of course, one could object something along the following lines. The idea of possession of a concept that Evans appeals to doesn’t entirely divorce possession of a concept from the grasping of a thought. For it construes that possession as being nothing more and nothing less than an ability that is involved in the grasping of a range of whole thoughts. Possessing the concept of being happy, for instance, just is the ability to grasp all those thoughts to the effect that someone is happy. What one possesses when one possesses that concept is thus not some item or element that is separable from the grasping of whole thoughts. Rather, one possesses a capacity that is common to the grasping of a certain range of thoughts. And that means that possession is still construed in terms of the thinking of whole thoughts. “Possessing a concept” just singles out a certain capacity that one deploys in a range of thoughts that one is able to grasp. The fact that one may distinguish a common feature in that range shouldn’t make one think that it is a separable feature. Apart from the thinking of those thoughts, the very idea of such a capacity is entirely idle. The capacity only comes into view when one considers whole thoughts. At that level, one is able to detect a certain common feature, which is then best understood as a capacity that is exercised in each of those different thoughts. Indeed, that is the important difference between construing the ­structure of thoughts in terms of elements and in terms of capacities. An account in terms of elements involves an atomistic perspective, since the elements are, presumably, identifiable independently of the contribution they make to a thought. But this is precisely not the case when one provides an account in terms of capacities. The point of drawing on the notion of a capacity is that a capacity is only distinguishable in the context of an activity – in this case, the grasping of (whole) thoughts. Once we have brought whole thoughts onto the scene, we may distinguish certain

Externalism and Context-Sensitivity  177 commonalities between thoughts and thereby identify different ranges. One range that we may identify involves the concept of being happy: it will include all those thoughts in which truth is made to depend on whether or not an object is happy. The capacity involved in possessing the concept of being happy is then just that: what is common to the grasping of that range of thoughts. This idea plays a central role, in a slightly different guise, in ­Wittgenstein’s Tractatus. He also allows for the idea that thoughts (or, rather, Sätze) have an intrinsic structure, but he seeks to reject Russell’s endeavors to cash out that idea in terms of independently graspable simples or particulars. The point is already presented in the early parts of §2, where Wittgenstein writes: 2.0121 It would seem to be a sort of accident, if it turned out that a situation would fit a thing that could already exist entirely on its own. If things can occur in states of affairs, this possibility must be in them from the beginning. […] Just as we are quite unable to imagine spatial objects outside space or temporal objects outside time, so too there is no object that we can imagine excluded from the possibility of combining with others. If I can imagine objects combine in states of affairs, I cannot imagine them excluded from the possibility of such combinations. 2.0122 Things are independent in so far as they can occur in all possible situations, but this form of independence is a form of connexion with states of affairs, a form of dependence. (It is impossible for words to appear in two different roles: by themselves, and in propositions.)12 Wittgenstein allows for the idea of a simple, but a simple, on his conception, isn’t separable from the wholes of which it can be a part. In fact, the simple is nothing more and nothing less than something common to a range of thoughts: it is a structural feature that characterizes a certain range. One may thus distinguish or identify simples in whole thoughts, but, pace Russell, one may not entirely dissociate them from whole thoughts. The idea that thoughts have an essential structure doesn’t, on this view, involve an atomistic perspective. This then also entails a certain holism about such capacities – a point that Evans doesn’t deal with in his presentation of the Generality Constraint. Since the relevant capacities are only identifiable in terms of ranges of thoughts and those ranges necessarily involve more than one capacity, the possession of any capacity is inextricably linked to the possession of others. In fact, on the view in question, the different capacities are features we carve out of the domain of all possible thoughts. Wittgenstein (but not Evans) is explicit about this.

178  David Zapero A proposition can determine only one place in logical space: ­ evertheless the whole of logical space must already be given by it. n […] (The proposition reaches through the whole of logical space). (§3.42) Yet, while such a perspective still allows for an internal connection ­ etween the possession of a concept and the grasping of a thought, it asb cribes a certain primacy to the former over the latter. The fact that whole thoughts need to be in the picture for capacities to be identified doesn’t change anything about the fact that those capacities – that is, the range of relevant thought-parts – are identifiable independently of any given thought. To be able to determine whether any given thought falls inside a certain range, one must have a conception of the feature common to that range that is independent of the thought at hand – or, for that matter, independent from any other particular thought. Of course, one starts with a certain domain of thoughts and sees what they have in common. One is for instance struck by the fact that the thoughts expressed by sentences such as “John is happy” and “Harry is happy” have something in common. In that sense, the domain of thoughts comes first. But to find (supposedly) common features one must be able to extract from those different thoughts something that is identifiable independently of any particular thought. Otherwise one would lack the means to answer determinately the question whether any given thought falls into a range of thoughts (that allegedly share a feature) or not. (This, too, is something about which the Tractatus is particularly explicit. The relevant structural features, what Wittgenstein calls things or simples, “exist anyway”, independently of what may actually be the case.) Yet, this is precisely what Travis seeks to call into question: that there is the possibility of distinguishing such capacities independently of the grasping of any particular thought. There is something beyond any number of such distinguishable capacities that is required to be able to grasp a particular thought. To know when a given thought would be true one must also have a suitable sensitivity to the surroundings and one must be capable of recognizing what facts or considerations are relevant in determining, on a given occasion, whether things are as the thought presents them. Which is why there is indeed an indefinite number of thoughts to the effect, say, that a is F. The two relevant capacities, of being able to identify a and to think something as being F, leave it underdetermined just when a would count, on a particular occasion, as being F. ­ ossession This claim has, in turn, an important impact on what the p of a concept can amount to. Concepts, on this view, are carved out of particular whole thoughts. What it amounts for something to be F ­depends on the particular thought at stake. Different thoughts involve different understandings of what it is to be F and no general account of what it is to be F will allow us to make a transition to the particular case:

Externalism and Context-Sensitivity  179 ­ articular none will allow us to answer the question of whether a certain p ­ bject a, on a certain occasion, is F. Depending on the occasion, differo ent facts and considerations will be relevant for deciding whether something is F. And just which are relevant depends on the understanding of F at hand – that is, on the thought from which one has carved out the concept of being F. One may, naturally, be able to find a set of features that are in ­common to all the various cases of being F. But which facts or considerations are relevant for deciding whether the concept is applicable to some further case isn’t derivable from any account of what is common to those cases that fall, up to now, under the concept. Look at what falls under a concept as much as you like, it will not provide you with a means to decide whether some further object does or does not fall – counts or doesn’t count as falling – under the relevant concept. The crucial point is this: what falls under a concept, call it the particular case, doesn’t determine just what other particular cases may also count as being that same way. Of course, whether another particular case falls under the concept depends on whether it is the same in the relevant respect. That presumably is the only thing that can determine whether a particular case falls under a concept. But the point that we have reached here is not about what determines whether a particular case falls under a concept, but about when it does. Study the previous cases as much as you like, the claim goes, they won’t provide you with an answer to the question of whether this new case also falls under the concept. Study the existing rhinoceroses as much as you like, that will not allow you to determine whether some unexpected candidate that you come across in the jungle of Mato Grosso is also a rhinoceros. In deciding whether it is you will of course start off by appealing to a set of features that all hitherto discovered rhinoceros may have. But there will remain the possibility that the new candidate doesn’t have those features and yet, for all that, may still be a rhinoceros. We touch here upon an idea that is at the root of Travis’s claim about the theory-resistance of conceptual capacities. The basis of that claim is provided by a certain picture of the kind of generality that is characteristic of concepts. The idea being that that generality is sui generis in the sense that it cannot be accounted for in terms of what falls under a ­concept. Look at what falls under a concept as much as you like, it will not allow you to say to what other particular cases a concept reaches. The generality of a concept is such that its applicability to particular cases cannot be accounted for in more basic terms.

Context-Sensitivity and Answerability The suspicion that this idea about the nature of conceptual capacities raises is of course that the reach of a concept should be conventional or even arbitrary. If we aren’t able to account for the reach of a concept

180  David Zapero in terms of a set of features that the things falling under the concept ­ ossess, then what allows us to claim that all of those things are the p same in a certain respect, namely in the way that is supposed to be captured by the concept? Indeed, a concept, in the sense in which we have been using the term, is meant to capture a certain way for things to be. It serves to delimit all those things which are that particular way. Yet, how can they capture a way for things to be if what falls under the concept is not determined by the possession of a feature or property that we can define independently? Indeed, if a concept’s application cannot be explained by something’s possessing independently identifiable features or properties, then what allows us to claim that that application is determined by how things are – and not just by how we happen to classify them? This suspicion is of course closely related to the question that we already presented at the beginning of the previous section, namely the question about what could justify Travis’s criterion for identifying ways for things to be. Why should what is recognizable (to us) as an instance of some property actually be an instance of that property? How can our particular, parochial capacities decide what are different instances of one same way for something to be? It seems that only two things should play a role in deciding whether something falls under a given concept: the thing being considered and the relevant concept. The worry is that Travis’s claim may force us to allow some further party into the picture – something like our particular sensitivities, or our ways of viewing things. The counterpart of this worry, at the level of thought, is a worry about what parties are involved in the relation of being true. As we already mentioned, the problem of determining whether a thought is true can be considered to be the same, in a certain respect, to the problem of determining whether something falls under a concept: the former, like the latter, is a problem of determining whether a particular case falls under a certain generality. Put in the terms that Travis often uses: the problem is one of determining the reach of that generality, of whether the generality at stake reaches to the particular case in question. The worry is thus, ultimately, that the theory-resistance of conceptual capacities should involve calling into question the idea that truth depends only on the truth-bearer on the one hand and the world or reality – the way things are – on the other. Here is one passage in which this seems to be the case: Truth, from our vantage point, is a product of three factors: the above two [how words described things as being and the way those things in fact are] and the character of the occasion on which given words were used so to describe things. For a statement to be true, it must relate in the right way to a pair of factors: the way the world it describes is; and the circumstances of its making.13

Externalism and Context-Sensitivity  181 It can seem that a further factor is being introduced into the equation. Truth depends not only on the relevant statement on the one hand and on the world on the other, it also depends on the circumstances in which the statement is made. But it would be a mistake to think that this talk about three factors is in conflict with the idea according to which truth is a “two-party enterprise”. To see, however, that there is at least no intention to contest that idea it is helpful to consider the setting of these kinds of statements, and, more generally, the setting in which Travis first developed the claim about the theory-resistance of concepts. In the background of the discussion of truth from which I’ve quoted is the criticism of a certain conception of the relation between linguistic meaning and actual statements. The target is a picture which Grice did much to promote: a picture according to which, modulo certain quite particular factors, the meaning of words fixes what is said when one uses those words on a particular occasion. This picture involves establishing a close connection between linguistic meaning and truth, since the meaning of sentences used to make declarative statements are considered to determine, modulo the relevant factors, when those statements would be true. While one will of course have to take into account the context to establish just what someone has said when she used a certain sentence on a certain occasion, there is, on the targeted view, a systematic link between the meaning of the sentence and what is said (and, thus, in the cases of declarative utterances, a systematic link between the meaning of the sentence and the truth-conditions of what is said). There is, thus, on this view, the possibility of providing a theory of that connection. One finds in Travis’s criticisms of this view his claim about the ­theory-resistance of conceptual capacities in yet another guise – but that is not something that I can deal with in any detail here. The important point for our purposes is the role that the idea of circumstances or occasions plays in that criticism. The criticism is that no systematic link can be established between the meaning of a sentence and what it can be used to say – and, more specifically, that no link can be established between the meaning of a sentence and when, in a declarative use, the corresponding statement would be true. The meaning can make an indefinite number of contributions to the conditions under which what was said, in using that sentence, would be true. The notion of circumstances serves, in the first instance, to point to this theory-­ resistant relation between meaning and saying. We have to look to the circumstances of the utterance to determine what is said not because the meaning of the sentence needs to be “completed” – because the arguments of certain variables (indexical expressions, etc.) need to be plugged in – but, on the contrary, because without taking into account the circumstances it remains entirely undetermined when what has been said would be true.

182  David Zapero The appeal to circumstances in the quote above has to be understood in this context. If a further, third factor is needed for there to be truth it’s because the first factor – how words described things as being – is not in fact a truth-bearer. Circumstances are adduced as a further party in the relation of being true because the first party does not yet constitute a determinate content. The kind of determinate content that can be either truth or false only comes into view when how words described things as being is viewed in the right circumstances. Indeed, once that is done, we are back to the two-party conception of truth. There is the determinate thing that is said on the one hand and the world, or how things are, on the other. The concern, in the essay from which I have quoted – and in many other places, as we’ll see – is to get clear about just what can play the role of being true or false. Or, to paraphrase another of Travis’s ­essays, just what can strictly speaking be true. It is thus important to qualify one of the claims I made earlier on about the relation between thought and language. In tracing the corollaries of the idea that conceptual capacities are theory-resistant, I claimed that, on the view that Travis develops, certain features from language carry over into thought. Indeed, if one looks to thought to eliminate a certain kind of context-dependence that is characteristic of language, one will be disappointed. The kind of dependence on context that characterizes particular utterances carries over, in a decisive way, into thought. And, thus, any attempt to make the grasping of thoughts independent of sensitivity to occasion must flounder. But this talk of “features carrying over” could give the impression that the view at stake involves a ­rapprochement of language and thought. And that would be a mistake. In fact, the point about features carrying over from language into thought must be understood against the background of an attempt to show that the gap between language and thought is deeper that it is generally considered to be. The point can, tentatively, be put in the following way: thought cannot solve the problems one might find in language because it is in a different business than language. The fact that certain features carry over has to do with the fact that thought isn’t the kind of thing that could possibly make language any clearer or less ambiguous. If that’s what one turns to thought for, then thought can be of no help. To flesh out some features of the conception of thought that follows from the general claim that I’ve been discussing let me take up again the criticism of the Generality Constraint. The central idea was the following: what facts are relevant for the application of a concept varies from occasion to occasion. So knowing what it would amount for a certain object a to be F doesn’t yet tell you what it would involve for another object to be that way. It could turn out that, for that other object, different considerations or facts are relevant in deciding whether it is F or not. Indeed, it could turn out that that other object is F even though it

Externalism and Context-Sensitivity  183 doesn’t possess the features which are considered to make object a an F. The point, as we saw, is an antireductionist one: it involves the idea that a way for something to be isn’t analyzable in terms of more basic features or facts. Of course, in a trivial sense, all objects that are F have something in common: precisely the fact that they are all instances of F. The antireductionism comes in when trying to account for this. On the view that Travis defends, an account of the relation of falling under a concept cannot appeal (successfully) to features that are more basic than the feature to be accounted for. For any theory involving more basic features one could find exceptions – cases which are F but which don’t possess all the ingredients the theory claims are conditions for being F. This antireductionist idea goes hand-in-hand with another ­anti-­reductionist idea. That one shouldn’t be able to account for the ­relation of falling under a concept in terms of more basic facts means that thoughts do not have an intrinsic structure. The reason is that the contribution that a concept can make to a thought is not determinate in the way that would be required for thoughts to have an intrinsic structure. A thought’s being about an object a and its being about something’s being F does not, together, determine when that thought would be true; the two things do not suffice for picking out one particular thought because, when taken independently of the role it plays in any particular thought, the concept of being F can amount to various different things. There are thus multiple thoughts to the effect that a is F as long as one draws only on those two factors. This way, one is led from the claim about the theory-­resistance of conceptual capacities to a picture of thought a­ ccording to which thoughts have no intrinsic structure. On the conception of thought that emerges in this way, thoughts have no components – if by components one understands parts that are identifiable independently of the whole thought. The whole thought comes first: it is only with respect to a whole thought that one can identify a particular part. Thought-elements must thus be construed as partial contributions to what the thought as a whole does, namely presenting things as thus and so. Thus, unlike language, thoughts have no structure or syntax. Indeed, for their achievement to be taken as basic means that they cannot be individuated by anything other than that achievement, namely the way they represent things. A thought, in representing things as a certain way (in representing Barack as smiling, a flag as being red, or a morning as being gloomy) establishes a certain connection between truth and the way things are. It makes truth depend, in a determinate way, on how things are. In representing the morning as gloomy, the thought raises a determinate question about how things are, namely about whether the world is such that the morning in question is gloomy. In that sense, its achievement is simply to delimit a certain range of ­situations, namely those which make the given thought true.

184  David Zapero Of course, all depends on what morning is in question and, crucially, on what kinds of things count as gloomy. What is gloomy to some may be arcane or foreboding to others. So we may distinguish between different understandings of gloomy. Yet, if the question of what is gloomy is meant to be one about how things are, namely about how a certain morning is – if the question is one to which the answer admits of truth – then we must make those distinctions by distinguishing the kind of situations which would count as being gloomy. That is, the distinctions cannot be made for being gloomy in general; fixing a relevant understanding of gloomy requires an entire thought and thus a certain something – for instance, a certain morning – which is qualified as gloomy. One may understand various different things by the statement that the morning is gloomy, and thoughts just are those different understandings. Different understandings of what it is for this morning to be gloomy are different ranges of situations: different ranges of particular cases in which the relevant understanding is made true. So if thoughts are just individuated by what they achieve, if they are just the delimitation of a range of situations, no structure, form or empirical feature can distinguish one thought from another. Thoughts are thus, on this conception, invisible. They are, indeed, not the kind of thing that could have visible traits or features. Any such features or traits must be external to the thought because they are external to what it achieves. They may play a role in the attempt to make a thought recognizable (to someone), but they cannot be part of the thought itself. This invisibility of thoughts is a point that Frege, famously, laid much stress on: Without wishing to give a definition, I call a thought something for which the question of truth arises. So I ascribe what is false to a thought just as much as what is true. […] A thought is something immaterial and everything material and perceptible is excluded from this sphere of that for which the ­question of truth arises.14 One may think it ironic, given our starting point, that we should be led to such a conception of thought. Many have taken Frege’s conception of thought to be the paradigm case of “Cartesian internalism”: a conception of thought according to which content can be individuated without taking into account the environment. The idea of a third realm of thoughts, ontologically secluded and without any connection to the world of empirical objects, seems to be just the conception of thought that is targeted by such authors as Putnam when they show that thoughts are necessarily world-involving. But it isn’t clear that Frege’s claim about the invisibility of thoughts needs to be understood in such a way. It could be understood as being, in

Externalism and Context-Sensitivity  185 the first instance, an antireductionist claim about the kind of ­achievement that thoughts involve. On such a reading, Frege is claiming that that achievement cannot be accounted for in more elementary terms, e.g. by appealing to something like form, structure or empirical features. This is of course not the place to take a stance on such controversial matters as Frege’s conception of thought. The important point for our purposes is that such an antireductionism about thought – be it Frege’s or not – is a corollary of a certain antireductionism about our conceptual capacities – it is the corollary of another antireductionist claim according to which such capacities are theory-resistant because they are not accountable in terms of principles that could be grasped without the possession of the relevant capacity. And that anti-reductionism, we have seen, is in turn the mark of the world-involving character of conceptual capacities, of the fact that the world has a role to play in deciding what is relevant in deciding when something falls under a concept.15

Notes 1 Evans summarizes perspicuously this line of argument in his Varieties of Reference, pp. 73–76. 2 Of chief importance in this respect are: Evans, The Varieties of Reference; ­McDowell, “On the Sense and Reference of a Proper Name”; M ­ cDowell, ­“Singular Thought and the Extent of Inner Space”; Putnam, “It Ain’t ­Necessarily So”; Putnam, Representation and Reality; Travis, The Uses of Sense. 3 Travis, “Pragmatics”, pp. 121–123. 4 The example is taken from Travis, “The Face of Perception”, p. 55. 5 Leibniz, New Essays on Human Understanding, pp. 312–313. 6 Wittgenstein, Philosophical Occasions, 1912–1951, p. 59. 7 In his Unshadowed Thought, Travis calls these two aspects the word-like and the world-like face of propositions (which, there, are called “shadows”). Cf. Travis, Unshadowed Thought, chapter 1. 8 Wittgenstein, Philosophical Investigations, §241. 9 Wittgenstein, Philosophical Investigations, p. 398. 10 Evans, The Varieties of Reference, p. 101. 11 Evans, The Varieties of Reference, p. 101 and p. 104. 12 Wittgenstein, Tractatus Logico-Philosophicus, §§2.0121–2.0122. 13 Travis, “Meaning’s Role in Truth”, p. 458. 14 Frege, “The Thought”, p. 292. 15 Research for this essay was supported by funding from the Fritz Thyssen Stiftung. Thanks also to Alex Englander for discussion about a previous version

Bibliography Evans, Gareth. The Varieties of Reference. Oxford: Oxford University Press, 1982. Frege, Gottlob. “The Thought”. Mind 65, no. 1 (1956): 289–311. Leibniz, Gottfried Wilhelm. New Essays on Human Understanding. Edited by Peter Remnant and Jonathan Bennett. Cambridge Texts in the History of Philosophy. Cambridge: Cambridge University Press, 1996.

186  David Zapero McDowell, John. “On the Sense and Reference of a Proper Name”. Mind LXXXVI, no. 342 (1977): 159–185. ———. “Singular Thought and the Extent of Inner Space”. In Subject, Thought, and Context, edited by Philip Pettit and John Henry McDowell, 137–168. Oxford: Oxford University Press, 1986. Putnam, Hilary. “It Ain’t Necessarily So”. The Journal of Philosophy 59, no. 22 (1962): 658. ———. Representation and Reality. Cambridge, MA: MIT Press, 1988. Travis, Charles. “Meaning’s Role in Truth”. Mind 105, no. 419 (1996): 451–466. ———. “Pragmatics”. In A Companion to the Philosophy of Language, edited by Bob Hale and Crispin Wright, 87–107. Oxford: Blackwell, 1997. ———. Unshadowed Thought: Representation in Thought and Language. Cambridge, MA: Harvard University Press, 2000. ———. The Uses of Sense: Wittgenstein’s Philosophy of Language. Oxford: Clarendon Press, 2001. ———. “The Face of Perception”. In Hilary Putnam, edited by Yemima Ben-Menahem. Contemporary Philosophy in Focus, 53–82. Cambridge: Cambridge University Press, 2005. Wittgenstein, Ludwig. Philosophical ­Occasions, 1912–1951. Edited by James Carl Klagge and Alfred Nordmann. Indianapolis, IN: Hackett Pub. Co, 1993. Wittgenstein, L. Tractatus Logico-Philosophicus. Edited by David Pears and Brian ­McGuinness. London: Routledge, 2001. ———. Philosophical Investigations. Edited by G. E. M. Anscombe, P. M. S. Hacker, and Joachim Schulte. Rev. 4th ed. Chichester: Wiley-Blackwell, 2009.

9 Contextualism and the Twilight of Representationalism1 Avner Baz

This paper argues that, though contemporary ‘contextualists’ about ­linguistic sense have tended to be representationalist in how they think about language, they ought not to be: thought through, contextualism shows that, and how, our words may do their work for us, and may be beholden to the world in which we find ourselves, and speak, even apart from representing that world truly or falsely. Moreover, only by breaking away from the representationalist conception of language would we be able to truly appreciate the depth of insight, and reap the therapeutic benefits of ordinary language philosophy as exemplified, however ­differently, in the works of Austin and Wittgenstein.

Stage Setting: Representationalism and Contextualism I start with an elucidation of what I mean by the two ‘isms’ in my title, beginning with ‘representationalism’. ‘Representationalism’, as I’ll be using it, refers to the idea that language is first and foremost an instrument for forming and communicating representations, or ‘descriptions’, of the world, or, in other words, an instrument for storing and conveying information about the world. “The foremost thing we do with words” David Lewis asserts, “is to impart information” (1980, 80). 2 The representations of ‘the world’3 formed by means of language are taken by the representationalist to be, in principle, assessable in terms of truth and falsity, and to be identifiable by what worldly arrangement(s) would make them true, or by their ‘truth-conditions’ (which normally are “specified” just by reproducing (‘dis-quoting’) the representing sentence). Thus, the representationalist takes it, in the words of Kukla and Lance, that “the most fundamental, important, and common thing we do with language is use it to make propositionally structured declarative assertions with truth-values” (Kukla and Lance 2009, 10). The representationalist would not deny that there are many different kinds of things that we do with words other than describing or otherwise representing things as being some way or another. He insists, however, that the representational function of language is primary and fundamental – that it

188  Avner Baz is as instruments of representation that refer to ‘items in the world’ that words are usable in all of the ways they are usable. Though I will mostly focus on representationalism about language, in the background there will be a broader position on which our basic relation to the world – the basic relation of human minds to their world, to echo McDowell (1994) – is one of representing it truly or falsely, judging it to be one way or another. The idea is that all of our relations with the world – or anyway all of the ones that are not merely causal, and are rationally assessable – are based on that sort of ur-relation.4 ‘Contextualism’, as I’ll be using the term, refers to the idea that the sense, or ‘semantic content’ of an utterance – even before we consider all that the speaker might mean (imply, ‘implicate’) in uttering her words – is a function, not merely of whatever it is that the uttered words bring with them from one utterance to another – call it ‘their meanings’ – and of how they are put together, but also of the context in which the utterance is made. Another way of putting the contextualist idea would be to say that the contribution a word makes to the overall sense or ‘semantic content’ of an utterance is a function, not merely of its meaning and of how it is combined with other words, but also of the context in which the utterance is made. This second way of formulating the contextualist position is really equivalent to the first, for, on the contextualist understanding, the sense of a word, on an occasion – what contribution it makes to the overall sense of an utterance – is indeterminate as long as, or to the extent that, the overall sense of the utterance is indeterminate. This, I should note, is just the basic insight of gestalt psychology, ­applied to the perception of linguistic sense: the perceived (or perceivable) significance of the parts is a matter of their contribution to the perceived (or perceivable) significance of the whole, and remains perceptually indeterminate as long as, or to the extent that, the significance of the whole is perceptually indeterminate. The perceived significance of the whole – here, the utterance or speech-act – is in turn affected by its perceived context. The truth of semantic contextualism is, thus, a truth about perceived sense in general, not just linguistic sense. Having clarified what I mean by ‘representationalism’ and by ­‘contextualism’, I now note that contemporary contextualists have tended to be representationalists: they have taken the sense of an uttered word, on an occasion, to be its contribution to the sense of the utterance, and have argued that the sense of an uttered sentence – again, even before taking into consideration all that the speaker might (be said to) have meant in uttering it – is context sensitive, but they have still tended to think of the sense of an utterance in terms of its truth-­ conditions. ­Correspondingly, they have tended to take the basic function of language to be that of ‘describing’ the world, or conveying ‘information’ about it. Thus, David Lewis still thinks of the sense of an uttered sentence in terms of its truth-conditions (1980); Robin Carston speaks

Contextualism and Representationalism  189 of the sense of uttered words in terms of “the thought or ­proposition” ­expressed (2002, 29); Recanati is happy to describe his position as “truth-­conditional pragmatism” (2010); Keith DeRose glosses all that we may do with sentences of the form “N know(s) (or do(es) not know) that such and such” in terms of “describing situations” (2005, 174); and Charles Travis argues that the role of a sentence… is not to be the expresser, in its language, of such and such thought, but rather to be usable in many different circumstances for expressing any of many thoughts, each with its own condition for truth. (1991, 242)5 Travis’ representationalism manifests itself not only in his tendency to think of sentences as instruments for expressing Fregean thoughts, and to think of sense first and foremost in terms of truth-conditions, but also in how he understands, and employs, the central Wittgensteinian notion of ‘nonsense’. An utterance is nonsensical, for Travis, when it is made in a context unsuitable for fixing the truth-evaluable representation, or Fregean thought, it expresses: Suppose that ‘know’ may make any of many distinct semantic ­contributions to wholes of which it is a part, and varies its contribution from one speaking to another. Then, describing someone as he is at a time, we would, on some occasions, say something true in saying him to know that X is a tree, and, on other occasions, say something false in saying that… In that case, circumstances of a speaking of ‘N knows…’ may confer on it a supplement to the content provided by the meaning of the terms alone. For some such supplements, the result will be stating truth; for others it will be ­stating falsehood. But some circumstances may fail to confer a supplement of either of these sorts. Words produced in such circumstances would have a content supplementable in either way. But a content still so supplementable can require neither truth nor falsity. (1997, 97, my emphases)6 For Travis, then, the conditions of sense are conditions of giving this or that description of the world, making this or that statement, expressing this or that thought, not conditions of, for example, describing the world, or stating something, or expressing a thought, as opposed to doing any other thing with our words or doing nothing at all with them. The variety of things we (can) do with words – the variety of ways we could mean them – and the conditions of doing with them some particular thing or another or meaning them in some particular

190  Avner Baz way or another, therefore do not come into view in his account, just as they tend not to come into view in the writings of other ­contemporary ­contextualists (and anti-contextualists). When, in Thought’s Footing, Travis discusses what he there calls “personal meaning”, which he glosses in terms of “how someone meant her words”, he cashes that out in terms of “meaning one’s words to represent, in some specific way, things as being some specific way” (Travis 2006, 119). That the words are in the business of representing the world, truly or falsely, is thus simply presupposed.7

Austin’s (and Wittgenstein’s) Non-Representationalism Since contextualists such as Travis (1989, 1997) and DeRose (2002, 2005) have taken and presented themselves as following Wittgenstein and/or Austin, it is worth noting that there are good reasons for thinking that neither Wittgenstein nor Austin would have been happy with the contextualist’s commitment to the representational-referential and truth-conditional understanding of linguistic sense. Early on in the Investigations Wittgenstein likens words to tools, or instruments – Werkzeuge, things to do work with, the sort of things that may be found in a toolbox (Wittgenstein 2009 (‘PI’), 11 and 14). And he clearly means this comparison to encourage us to think of the meaning of a word as a matter – not of what independently-existing worldly item or set of items it refers to or represents, but – of what work we do, or may do with it, the intersubjectively-significant difference(s) we (can) make by means of it (in combination with other words); and he wants us to appreciate the variety of things, or kinds of work, we do with words, and even with one and the same word (cf. PI, 182). Later in the Investigations Wittgenstein says that the key to the dissolution of certain sorts of traditional philosophical difficulties is to “make a radical break with the idea that language always functions in one way, always serves the same purpose: to convey thoughts” (PI, 304). I take the reference to Frege (and to the younger Wittgenstein) in this remark to be fairly obvious: the radical break he is calling for is with what I’ve been calling ‘representationalism’. Austin’s anti-representationalism is, if anything, even clearer. After all, at the opening of How to Do Things with Words he draws our attention to ‘performative’ utterances that, despite having a ‘declarative’ form, he thinks are obviously unsuitable for assessment in terms of truth and falsity (Austin 1999, 6); and towards the end of that set of lectures he professes an inclination to “play Old Harry with the true/false fetish” (Austin 1999, 151). And then, in ‘Other Minds’, he compares the performative ‘I promise’ and ‘I know’ – at least in one of its uses (a use, I note, that is actually quite special and uncommon) – and argues that the latter too is not a “descriptive phrase” (Austin 1979, 103). He then warns his

Contextualism and Representationalism  191 audience of the “descriptive fallacy”, and goes on to say that “even if some language is now purely descriptive, language was not in origin so, and much of it is still not so” (Austin 1979, 103). As with many other moments in Austin, I wish he had spelled out more clearly what he meant when he said that language was not in ­origin purely descriptive; for, as I see it, he is getting at a deep truth, not only about the evolution of natural language, but also about the acquisition of a natural language. In Baz (2017), I examine empirical findings about children’s acquisition of philosophically troublesome words such as ‘know’ or ‘think’. The findings, I argue in that book, support a non-representationalist, broadly pragmatist understanding on which those words are acquired as instruments for making different kinds of significant moves in intersubjectively significant situations, not as instruments for representing ourselves and others as being, or not being, in some particular state – the state of knowing, or not knowing, this or that, for example. What children need to learn to pick out and follow in order to become competent users of (for example) ‘know’ and cognates, I there propose, is not some type of worldly constellation in which ­knowledge presumably consists and to which ‘know’ and its cognates presumably refer – not even context-­ dependently. What they acquire is better thought of as several different and variously related patterns of ­u sage of constructions featuring these words – each with its ordinary and normal worldly conditions and its ordinary and normal ­significance(s). The philosopher’s ‘knowledge’, I then propose, refers to a philosophical construction, which may well be related in various ways to our ordinary and normal use of ‘know’ and cognates, but which has blinded us to important aspects and dimensions of that use, and has entangled us in difficulties that are extraneous to it. Coming back to Austin’s questioning of the descriptive or representational fallacy, I note that in a footnote to “A Plea for Excuses”, having said that ordinary language – as it reveals itself in an examination of “what we should say when, and so why and what we should mean by it” (Austin 1979, 181) – should provide ‘the first word’ for philosophy, he implores his readers to “forget, for once and for a while, that other curious question ‘Is it true?’” (Austin 1979, 185). DeRose, having proposed that Austin may be seen as contextualism’s “granddaddy”, cites the above footnote from “A Plea for Excuses” and says that he finds it “troubling” (DeRose 2002, 196). It is clear why DeRose should find this footnote troubling: contemporary ‘contextualism’ is essentially a theory about the truth-conditions of ‘assertoric’ (‘declarative’) sentences.8 ­DeRose notes, however, that the footnote is actually consistent with Austin’s writings on epistemology, in which Austin “avoid[s] issues of whether our epistemological claims (especially claims about what is and is not known) are true or false” (DeRose 2002, 196).

192  Avner Baz And that is true. In “Other Minds” Austin says things like the following: If you say ‘That’s not enough’ [as a way of challenging a claim that such and such and the basis given as its support], then you must have in mind some more or less definite lack… If there is no definite lack, which you are at least prepared to specify on being pressed, then it’s silly (outrageous) just to go on saying ‘That’s not enough’. (1979, 84) He also says that what’s enough will be a matter of what is “within reason”, and will depend on “present intents and purposes” (Austin 1979, 84). One could see in these passages of Austin’s all of the main ingredients of contemporary contextualism about ‘knowing (that such and such)’, except that he seems content to speak in terms of when we would be ‘right’ to say we know (Austin 1979, 98), or ‘justified’ in saying this (Austin 1979, 101), and, as in the above quotation, in terms of what would or would not be ‘silly’ or ‘outrageous’ or ‘within reason’ to say. He never asks when it would be true to say of someone, who has asserted that there is a goldfinch in the garden, that he knows there’s a goldfinch in the garden. He never says that if the knowledge claimer has said enough for all intents and purposes to support his claim, then he just knows, or at any rate has claimed something true. In thus describing the functioning of ‘know’ in discourse, Austin goes against the grain of the representationalism of contemporary contextualism (and anti-­ contextualism, or ‘invariantism’). I want next to indicate why I believe we would do well to follow him in this.

The (Real) Austinian Overcoming of Cartesian Skepticism Whatever the differences between them, for all contextualists about ‘know’, the word is taken to refer to a state in which someone takes it that such and such and the question is whether their basis for thus taking things – their ‘justification’, if you will, or their ‘evidence’ – is good enough for them to count as knowing that such and such.9 The ­context-sensitivity of ‘know’ is then supposed to be the context-­ sensitivity of the correct answer to that question. On a contextualist account such as Lewis’ (1996) or Travis’ (1989, 1997), for example, ‘know(s)’ truly applies to some pair of proposition assumed to be true and potential knower who believes that proposition, just in case the potential knower (or ‘her experience’, or ‘her evidence’) has ruled out (‘eliminated’, ‘discharged’) all relevant alternatives ­(Travis says “real (as opposed to mere) doubts”) – where what alternatives are relevant is context-dependent. There are also contextualist accounts

Contextualism and Representationalism  193 on which the perceived stakes in p being, or not being, true, affect the truth-conditions of utterances of the form ‘N knows that p’ (see DeRose 1992, 2011; Cohen 1998, 1999). These accounts are more vulnerable to the anti-contextualist (‘invariantist’) charge that the contextualist intuitions are affected by the pragmatics (‘assertability conditions’), rather than semantics (‘truth-conditions’), of ‘know’ and its cognates.10 Lewis, Travis, DeRose and Cohen all claim that recognition of the context-sensitivity of ‘know’ and cognates is key to overcoming what they call ‘skepticism’. This seems to me mistaken for several reasons. First, they run into trouble with the traditional skeptic because of how they understand the ‘factivity’ (or ‘infallibility’) of knowledge (if you know that such and such, then such and such). For Austin, the idea that knowledge is ‘factive’ or ‘infallible’ captures correctly something, not about the truth-conditions of ‘know’, but about the competent use of the word, or what may be called its felicity-conditions: ‘When you know, you can’t be wrong’ is perfectly good sense. You are prohibited from saying ‘I know it is so, but I may be wrong’, just as you are prohibited from saying ‘I promise I will, but I may fail’. If you are aware you may be mistaken, you ought not to say you know, just as, if you are aware you may break your word, you have no business to promise. (1979, 98) But for the contextualist, who still thinks about the word as either ­applying or not applying to any given pair of knower and proposition – context-sensitively, but regardless of how it is used, or meant – this Austinian understanding does not come into view. Both Travis and Lewis propose to build the factivity or infallibility of knowledge into the ­elimination-of-relevant-alternatives requirement, which is in turn supposed to participate in fixing the truth-conditions of utterances featuring ‘know’. Lewis simply says that “the possibility that actually obtains is never properly ignored” (1996, 554), and then adds that a possibility that “saliently resembles actuality” may not properly be ignored either (1996, 557). Travis puts the idea this way: If there are facts to make this a case of F’s not obtaining, then for any claim that A knows that F (or, equivalently, any occasion for judging whether A knows that F), there are some facts which show some doubt to be real for that claim or on that occasion which A has not discharged. (1989, 162; see also 173) In thus building the factivity or infallibility of knowledge into their ­account of what makes alternatives (or doubts) relevant (or real),

194  Avner Baz however, Lewis and Travis undermine their ambition to offer a ­satisfying response to the skeptic. For though on their account we may – if we are lucky – “know a lot” (Lewis 1996, 549), we cannot (truly be said to) know that we know those things, because we cannot, by hypothesis, rule out any number skeptical possibilities; and any of those possibilities might, for all we know, turn out to be actual – we have not, after all, eliminated it – and therefore, by Travis’ and Lewis’ light, might turn out to be relevant. This difficulty might seem to be a mere technicality; but it actually points to a much deeper flaw in the contextualist response to skepticism. The very idea that we could overcome traditional forms of skepticism such as Cartesian skepticism about ‘the external world’, or about ‘other minds’, by way of an account of the semantics of the word ‘know’, or for that matter by way of semantic contextualism more generally, is, I believe, incredible. Each of those forms of skepticism is rooted in a compelling picture that has us barred, in principle, from accessing something – the external world, for example, or another person’s mind. For anyone captivated by such a picture, the idea that its skeptical upshot could somehow be blocked by way of an account of the semantics of ‘know’ is bound to seem not only hopeless, but entirely misguided. In order to overcome those forms of skepticism – which could be expressed without even using the word ‘know’! – one would need to shake the hold of the pictures that underlie them. One would need to radically transform the way we look at our relation to the world, and to other people. If there is a potentially satisfying response in “Other Minds” to ­skepticism about the ‘external world’ and about ‘other minds’, I accordingly wish to propose, it is found in Austin’s reminder that making a knowledge claim, or challenging one, is a human act that relies for its intelligibility on the history of our interpersonal practices, and which may intelligibly or felicitously be performed only within a suitable worldly context. There are worldly conditions for performing such an act; and performing it – whether felicitously or infelicitously – exacts worldly commitments. In making or challenging a knowledge claim, I place myself in relation to other people and open myself to their criticism, or rejection, or ridicule, or ingratitude, for example, to which I will then necessarily respond in one way or another (for whatever I do will be a response to the other’s response). And I also take the risk of turning out to have misled someone, for example, and of becoming responsible for that. Making myself vulnerable in these ways, as Nancy Bauer has insightfully pointed out (Bauer 2015), is part of the illocutionary force of such speech acts, part of what I do in uttering my words. As Austin reminds us, to say ‘I know that such and such’, in a context in which the question has arisen of whether such and such, is to “take a new plunge” (Austin 1979, 99) – not an epistemic new plunge, but a personal or ethical new plunge. We bind ourselves to others and stake our reputation in

Contextualism and Representationalism  195 a new way in saying ‘I know’, Austin says (Austin 1979, 99). And that too is part of the illocutionary force of our words; that too is something we do in uttering our words. The other might then reject our offering, insult us “in a special way” by refusing to accept it (Austin 1979, 100); and, if she does, we will, of necessity, respond to that act of hers in one way or another. But it would not be possible for her to insult us in that way if we did not perform the act of offering her our assurance. Even when all goes well and our knowledge claims, or challenges of other people’s claims, are as unremarkable and insignificant as can be, they are still moments in our history and in the history of others, still part of what we each carry with us to the next moment. But of course, this is true of any speech act we perform or might ­perform. The performance of any speech act requires suitable worldly conditions and exacts worldly and interpersonal commitments. Even the choice not to perform some speech act is only possible within a suitable worldly context and is something for which we could be held responsible. My proposal, therefore, is that if there is a potentially truly satisfying response to Cartesian skepticism in Austin’s work, that response has nothing in particular to do with his reminders about how the word ‘know’ functions in discourse or even with his response to sense-data theories in Sense and Sensibilia. If there is a truly satisfying response to Cartesian skepticism in Austin’s work it is best represented by the set of lectures that constitutes How to Do Things with Words. To be sure, there is very little in How to Do Things with Words of the existential fervor that I have extracted from “Other Minds”. But the general counter-Cartesian lesson that any act of expression (including the act of suppressing expression) is already beholden to a worldly background and positions its performer significantly in a world shared with others is surely one of the most important lessons of that set of lectures. The world, Austin reminds us, provides both the setting that makes possible any speech act I might perform – even if I perform it in the privacy of my study, on paper or even only in thought – and the setting for the act’s unfolding significance. It is in forgetting or repressing his knowledge of – that is, intimate familiarity with – this fact, which is most immediately a fact about himself at that very moment, that the philosopher “casts himself out from the garden of the world we live in” (Austin 1979, 90). Later I will propose that this anti-skeptical lesson is at the same time an anti-representationalist lesson.

‘Knowing’ without Representing I have argued that extant contextualist accounts of ‘knowledge’ ­cannot plausibly be taken to liberate us from traditional forms of skepticism, at least not in the way contextualists have envisioned. In order to truly overcome traditional skepticism, I have proposed, we will need to

196  Avner Baz overcome the representationalism that underwrites that skepticism, and acknowledge the way in which both the world and others are always ­already there for us, prior to any representation. The next thing I want to propose is that contextualists’ commitment to representationalism has also led them to significantly misrepresent the functioning of ‘know’ and its cognates in everyday discourse. And here too, Austin seems to me to point us in the right direction. Focusing once again on ‘relevant-alternatives’ contextualist accounts of ‘knowledge’, I begin by noting that though there seems to be an undeniable truth in the idea that some distinction between relevant and irrelevant doubts or alternatives is at least sometimes pertinent to our use of ‘know’ and its cognates, no clear and plausible story has been offered by anyone about how exactly the distinction gets effected on particular occasions in which those words are used; nor has a clear and plausible story been offered about how, or in what sense, relevant alternatives, or doubts as to p, might get ‘eliminated’ or ‘discharged’ by N, or by N’s ‘evidence’ or ‘experience’.11 The good news is that we do not really need such a story in order to know all there is to know about the ordinary and normal use of ‘know’ and its cognates. The story would only appear to be needed on a representationalist conception of what using our words intelligibly involves and requires – a conception on which there need to be truth-conditions (context sensitive or not) for every ‘application’ of ‘know’, and a truthvalue to each such ‘application’ that is establishable from the theorist’s perspective. And that conception is not actually vindicated by our ordinary and normal linguistic practices. As I note in Baz (2012), the distinction between relevant and irrelevant alternatives normally belongs in the same sorts of contexts in which the ‘How do you know?’ question – the question on which Austin focuses in “Other Minds” – normally belongs. In those contexts, someone asserts that such and such, or assures someone else that such and such, and then tells the other – either in response to her ‘How do you know?’ or without her prompting – what his basis is for taking it, and for asserting, that such and such. And, as Austin reminds us in “Other Minds”, not just any doubt or alternative to such and such could then reasonably or competently be invoked to challenge that person’s assertion or basis; and, moreover, alternatives that in one context would reasonably be taken to be relevant, could in another context reasonably be dismissed as irrelevant. This is the truth in (context-sensitive) ‘relevant alternatives’ accounts of ‘knowledge’. But it is not a truth about what philosophers have called ‘truth-conditions’. It is a truth about what may generally be called reasonableness in the employment of our words – I have earlier spoke about it, following Austin’s How to Do Things with Words, in terms of felicity-conditions. And one crucial difference between the A ­ ustinian question of reasonableness and the question of truth and falsity as the

Contextualism and Representationalism  197 representationalist typically understands it, is that the former is not just context-sensitive, but importantly also allows (within reason) for no-fault disagreements, whereas the latter is supposed to have one, and only one, objectively establishable correct answer (on which, it should be noted, philosophers have not been able to agree). Austin’s deep insight in “Other Minds” is that, if we truly are interested in a “restatement of our common knowledge about our practices of linguistic communication” (Lewis 1980, 79), and interested specifically in reminding ourselves of how ‘know’ “functions in talking” (Austin 1979, 79), then the truth about reasonableness is all the truth we need. If you assure me or otherwise assert that such and such, and give me your basis for taking it that such and such, and I accept your basis as good enough ‘for present intents and purposes’, and do not or cannot raise further reasonable doubts, and then all goes well, and such and such turns out to be the case, then that’s the end of the story – as far as the work ‘know’ has done for us goes.12 Whether or not you knew that such and such, or said something true in saying you knew (assuming you actually said ‘I know’), is a question that makes no contact with the normal practice in such situations and has no bearing on that practice. And if not all goes well, and it turns out that, by some turn of events neither of us has expected, it is not the case that such and such, then, under normal circumstances, I would have no (good) grounds for complaint against you; and that is what matters – as far as our use of ‘know’ is concerned – in such moments. The insistence that in that case the assurer didn’t (couldn’t) really know that such and such, because knowledge is ‘factive’, would then be an expression, once again, not of part of our common knowledge about our practices of linguistic communication, but of representationalist theoretical commitments extraneous to those practices. The upshot is that the contextualist’s commitment to a representationalist understanding of ‘know’ has blinded him to how the word actually functions “in the language that is its natural home” (PI, 116), and has entangled him in difficulties that are extraneous to our linguistic practices. The contextualist’s difficulties, just like those of the traditional (‘invariantist’) philosopher, are difficulties not with the phenomenon, or set of phenomena, we refer to or invoke when we competently employ ‘know’ and cognates; rather, they are difficulties with a philosophical construction that – just like Wittgenstein’s ‘beetle in the box’ (PI, 293) – drops out of consideration as irrelevant whenever ‘know’ is put to some intelligible use or another, under worldly conditions suitable for that use.

Contextualism and the Twilight of Representationalism And yet, for all of the ways in which the contemporary contextualist still thinks from within the framework of representationalist commitments and entangles himself in difficulties that are due entirely to

198  Avner Baz those commitments, in recognizing the role of context in fixing the ­contribution words make to the overall sense of utterances, contemporary contextualism points toward a more radical break with the representationalist conception of language than it itself has accomplished, or envisioned. To begin to see this, note that if we choose to continue to talk about context-sensitive terms as having a reference, that reference – and this is something that Travis saw clearly from very early on (cf. Travis 1989, 116–117) – cannot plausibly be thought of as an item, or set of items, that may be identified independently of our use of the term, and to which the term – as such, and irrespective of how it is used – refers. Context-sensitive terms do not have context-insensitive extensions. In this way, contextualism is an effective way of driving home the Kantian warning against the idea that our concepts may apply to the world ‘as it is in itself’ – that is, as it is apart from its coming into view, and coming to make sense to us, under suitable conditions. If the contextualist with respect to ‘know’ (for example) is right, we cannot identify instances of knowledge – thought of as existing in the world independently of our employment of ‘know’ and its cognates – and study those instances directly, for there are no such things to identify and study. (Travis would say the same thing about instances of something being green, or containing milk!) With perhaps very few exceptions, no relation between a potential knower and an empirical fact, taken by itself and context-independently, constitutes a case of knowledge simpliciter, or may truly be said to be, context-insensitively, ‘a case of knowledge’.13 We can therefore sensibly attempt to study and elucidate, not cases of knowledge, but only what we (may properly) count, in some contexts but possibly not others, as cases of ‘knowledge’. It might seem that we could accept this basic contextualist contention – however radically it goes against all of the attempts in recent years to undo, or dismiss, the so-called ‘linguistic turn’ in philosophy and to revive traditional, pre-Kantian metaphysics14 – while continuing to hold on to the idea that context-sensitive words such as ‘know’ and its cognates are first and foremost instruments of representation whose sense is essentially a matter of their referring – context-sensitively, but irrespective of how they are used, or meant – to worldly constellations. All we need, it would seem, is to allow that what any of those words refers to or represents includes the contextual factors that – on our favorite ­contextualist account – affect the contribution the word makes to the overall sense (‘semantic content’) of utterances in which it features (where ‘sense’ is still primarily understood in terms of ‘truth-conditions’). Thus, for example, suppose we accept a ‘relevant-alternatives’ contextualist account, on which ‘know(s)’ truly applies to some pair of a true proposition (or Fregean thought) and a potential knower who believes that proposition, just in case the potential knower (or her ‘experience’,

Contextualism and Representationalism  199 or her ‘evidence’) has ruled out (eliminated, discharged) all relevant a­ lternatives to that proposition – where what alternatives are relevant is context-dependent. And let us set aside all of the difficult-and-­hithertounanswered questions mentioned above, about what it is exactly that is supposed to rule out the relevant alternatives, what constitutes the ruling out of an alternative, and what makes some alternative relevant or irrelevant on an occasion. It might seem that we could hold on to the basic representationalist picture, by letting the worldly constellation to which ‘know’ refers include a variable that ranges over all of the alternatives to p that are – or that properly count as – relevant in the context of application, and the requirement that all of those alternatives be properly eliminated by the potential knower, if she is to count truly as knowing that p. In other words, we could seek to preserve the representationalist picture by saying that ‘know’ refers to or represents a state in which the holder of some true belief, p, or her experience or evidence, has ruled out all of the alternatives to p that are relevant in the context in which it is used. However, once we expand the intended reference of ‘know’ to include contextual factors, or types of factors, that affect the contribution the word makes to the overall sense of utterances in which it features, it becomes misleading to say that we use ‘know’ to refer to or speak of or represent that. For that now includes contextual features that affect how ‘know’ is to be understood, on some occasion of its employment; and those features may not themselves plausibly be thought of as part of what is spoken of or represented by ‘know’, on that occasion. Those contextual features contribute to the determination of what ‘know’ means, on that occasion; but they may not plausibly be said to be part of what it means – not if what a word means, on an occasion of its employment, is understood as a matter of what it refers to, or represents. What emerges is that by the contextualist’s own lights, our words – and so we, as their (competent) employers – are beholden not only to the world we use them to speak about, or to represent, but also to the world in which we use them. And this means that even if, as speakers, we do sometimes relate to our world in ways that may aptly be thought of in representational terms – which seems undeniable, if only in the trivial sense that we sometimes say things with respect to which it makes sense to ask questions such as ‘What (or who) are you referring to, or talking (or thinking) about?’ and ‘What are you saying (or thinking) about it?’ – our basic or primordial relation to the world may not aptly be thought of in those terms. For (even) in representing things to ourselves or to others, we are necessarily beholden and responsive – on pain of incompetence and incomprehensibility – to a worldly background that is not itself, there and then, represented by us or spoken about. Earlier I suggested that the above is one of the deepest and most ­neglected lessons of Austin’s How to Do Things with Words. It is also one of the deepest lessons of the phenomenological work of Heidegger

200  Avner Baz and Merleau-Ponty, for whom our relation to our world is not to be understood first and foremost in representational terms. “The world”, ­Merleau-Ponty writes in the Preface to the Phenomenology of P ­ erception, “is not what I think, but what I live through” (Merleau-Ponty 1996, xvi-xvii) – strikingly echoing Emerson, who wrote this almost a century earlier: “I know that the world I converse with in the city and in the farms is not the world I think” (Emerson 2000, 326). Later on in the Phenomenology of Perception Merleau-Ponty writes, The Kantian subject posits a world, but, in order to be able to assert a truth, the actual subject must in the first place have a world, that is, sustain round about it a system of meanings whose reciprocities, relationships, and involvements do not require to be made explicit in order to be exploited. (1996, 129) But of course, Merleau-Ponty could equally correctly have said this about any utterance, or thought – whether world positing or not, whether truth-evaluable or not. And this means that, on his understanding, the relation between our words and our world is not essentially, let alone necessarily, representational, or assessable in terms of truth and falsity. For our words not to be “frictionlessly spinning in the void”, as ­McDowell (McDowell 1994) puts it, they need not represent anything truly or else falsely: what we might call their justification (or lack thereof) – their rationally-assessable anchoring in experienced reality – need not take that form. Nor, therefore, should we think of what philosophers have called ‘judgement’ as the fundamental relation between us and the world. Not every time that the world draws out or solicits from us a linguistic (or other) response that may be found apt (competent, intelligible, reasonable, acceptable…), or not, are we aptly described as having judged, or as giving voice to a judgement. So it might be fine, or philosophically harmless, to take it, with Travis, that the fundamental question of perception is “How can perceptual experience make the world bear (rationally) for the perceiver on what he is to think and do?” (Travis 2013, 3). But if we take that question, as Travis does following Kant, Frege and McDowell, to be a question about what’s involved in the formation, understanding and assessment of Fregean thoughts, or Kantian Erkenntnisse, then we are going to over-intellectualize, and fundamentally distort, our beingin-the-world. We will then suppose, with Kant at some of his moments15 and with Frege, that “by the step with which I secure an environment for myself I expose myself to the risk of error” (Frege 1956, 306) – that is, that it is only by way of judgement, or the formation of true or false ‘cognitions’ or ‘thoughts’, that I move from being “shut up in [my] inner world” (Frege 1956, 309) to a world shareable with others. And then

Contextualism and Representationalism  201 we’ll forget that positioning ourselves by means of words – where that includes, but is in no way restricted to, the formation of true or false empirical judgements – is an activity that takes place in a world that is not, certainly not there and then, the object of such judgements, but which is nonetheless already shared with others, and to which we find ourselves always already responding, and beholden.16

Notes 1 This paper is dedicated to Charles Travis, from whose work I have ­benefited more than I’ve benefited from the work of any other contemporary philosopher. 2 More recently, Schoubye and Stokke have similarly asserted: “Discourses [are] goal directed activities whose fundamental aim is discovering what the actual world is like” (Schoubye and Stokke 2016, 767). Schoubye and Stokke say they are following Robert Stalnaker in taking that to be the fundamental aim of (all?!) discourses (Schoubye and Stokke 2016, 767). 3 I note that the notion of ‘the world’, as mostly used in contemporary analytic philosophy, is a very problematic notion. While it quite clearly means to refer to the objective world – the world as paradigmatically described and explained by the natural sciences – in effect it refers to what Kant would call ‘the world as it is in itself’; for the objective world is not the represented world – that is, the world to which our representations purport to be faithful – but rather the world as (objectively) represented. But, as Kant has taught us to see, philosophical talk about the world as it is in itself is ultimately nonsensical and a major source of philosophical headaches. By contrast, when I use (rather than quote) ‘world’, or talk about ‘worldly conditions’, I mean to refer to the phenomenal world – the world as perceived and responded to prior to becoming the object of empirical judgements, or Kantian ­Erkenntnisse. I elaborate on this issue in Baz forthcoming. 4 Note that this broader form of ‘representationalism’ is not the view that our perceptual experience provides us with representations. So the representationalism I mean to question in this paper is not the representationalism ­Travis argues against in “The Silence of the Senses”, but rather is one that he himself at least allows (see Travis 2013, 48, 315) – one on which the representing is done by us, but somehow necessarily, pervasively, and continuously. 5 In this way, all contextualists have participated in what Kukla and Lance call ‘the declarative fallacy’ (Kukla and Lance 2009, 11). 6 See also Travis (1989, 151; 1991, 243). 7 This is why Travis’s examples work best when the words he puts in the mouths of his protagonists are naturally understood as meant to provide others with some useful piece information, as for example when they are meant to describe a car that the participants are looking for in a parking lot. The examples work less well when the words are most naturally heard as meant to do something else, as for example when Pia says to Luc, “You know where the milk is”, when (out of laziness, or chauvinism, or both) he asks her for milk for his coffee, or when Pia, having just finished spraying her reddish maple with green paint, exclaims, “The leaves are green now”. I discuss these two examples, and show that Travis’s representationalist rendering of them is forced, or anyway hard to make clear sense of, in Baz (2008, 2012). 8 Grice, whose theory of ‘implicature’ is premised on the possibility of ­separating the truth-evaluable content of an uttered sentence from what,

202  Avner Baz in the Austinian sense, is being done with it, also complains about Austin’s footnote (Grice 1989, 13). 9 Note that I make no use here of the notion of ‘belief’. Travis is a ‘disjunctivist’, and denies that knowledge is belief+something (see Travis 2005). I myself am generally very suspicious of the philosophical notion of ‘belief’, which I take to be a piece of jargon that all-too-often comes in the way of truly understanding the phenomena it purports to refer to. But, in any case, nothing in the argument that follows depends on attributing to the contextualist an understanding of knowledge that takes it to consist of some representational mental state that may be identified independently of anything ‘external’ to the subject, and shared by knowers and non-knowers, plus something else. 10 Another significant advantage of the ‘relevant-alternatives’ contextualist ­account – besides the fact that, as I point out below, it does get something right about the ordinary and normal functioning of ‘know’ in some contexts – is that it is supported, not merely by intuitions about cases, but also by the serious difficulty faced by the invariantist of coming up with a context-­ insensitive way that is at once principled and plausible of drawing the distinction, within all of the countless alternatives to p that a would-be knower has not eliminated, between those that need to be eliminated by her and those that need not be eliminated by her in order for her to truly count as ‘knowing that p’ (see Travis 1989, 168; 1991, 245). 11 Travis offers no account, as far as I can tell, of how the distinction between real and mere doubts gets effected on particular occasions, or of what is required for ‘discharging’ a doubt. Lewis also does not tell us how one’s ‘experience’ is supposed to ‘eliminate’ an alternative; and he notoriously proposes that an alternative becomes relevant whenever it is ‘attended to’ (Lewis 1996, 559), which would make it extremely easy to disown an unwanted piece of knowledge! 12 And if I do not accept your basis as giving us sufficient assurance that such and such, then that – that is, whether your basis is good enough (for present intents and purposes) for us to proceed as if such and such, rather than whether you (may truly be said to) know – is the issue on which we disagree. And this sort of disagreement, once again, is one that allows for cases in which neither party is wrong, or mistaken. 13 Travis allows that ‘I exist’ may truly count as known in every context (Travis 1989, 146). But ‘I exist’, whatever exactly it might express, arguably does not express an empirical proposition. 14 In recent years, mainstream analytic philosophers such as Scott Soames (2003), Timothy Williamson (2007), Ernest Sosa (2007), Jason Stanley (2008), and Herman Cappelen (2012), have made considerable efforts to convince us that the so-called ‘linguistic turn’ in philosophy was a mis-turn, because, as they claim, it is not true that philosophical problems are all at bottom linguistic problems. What philosophers are mostly interested in, they have insisted, are things and their natures, not words and their meanings. By the light of contextualism, the irony here is that this very idea – that it should be possible not merely to distinguish, but to separate in practice, philosophically interesting phenomena from our concepts of those p ­ henomena – itself rests on a misguided conception, or picture, of language. 15 What I am here attributing to Kant seems to me to be true of the ­“Transcendental Analytic” of the first Critique. Kant’s representationalism gives way at certain points in the “Transcendental Dialectic”, especially in those moments in which reason’s demand for the ‘unconditioned’ or ‘­absolute’ – the demand that, when combined with the idea of the empirical world as a thing in itself, generates the antinomies – is presented not merely as the

Contextualism and Representationalism  203 ­ emand to t­ ranscend the ‘conditions of sensibility’, but also, even primarily, d as the demand to transcend the temporal unfolding of empirical investigation, or, as Kant refers to it, ‘the successive synthesis of the manifold of intuition’ (Kant 1998, A 417/B 444). Later on, Kant refers to this successive synthesis – to the natural home, as it were, of our transcendental categories and empirical concepts – as ‘the advance of experience (Fortschritt der Erfahrung)’ or as ‘empirical advance (empirische Fortschritt)’ (Kant 1998, A 493/B 521; see also A 479/B 507). Kant thereby acknowledges, in effect, that the natural home of those categories and concepts is a shared practice in which we e­ ngage against the background of a world we already share with others. 16 I’d like to thank David Zapero for a set of penetrating comments on the penultimate draft of this paper. Earlier drafts of the paper were presented to audiences at the University of Amiens, the University of Porto, and Haifa University. I’d like to thank those audiences for the discussions that ensued. In particular, I’d like to thank Bruno Ambroise, Eduardo Marchesan, Sofia Miguens and Charles Travis.

Bibliography Austin, J. L. (1979). Philosophical Papers. New York: Oxford University Press. Austin, J. L. (1999). How to Do Things with Words. Cambridge, MA: Harvard University Press. Bauer, N. (2015). How to Do Things with Pornography. Cambridge, MA: ­Harvard University Press. Baz, A. (2008). “The Reaches of Words”. International Journal of ­Philosophical Studies 16: 31–56. Baz, A. (2012). When Words Are Called For. Cambridge, MA: Harvard ­University Press. Baz, A. (2017). The Crisis of Method in Contemporary Analytic Philosophy. New York: Oxford University Press. Baz, A. (forthcoming). “Bringing the Phenomenal World into View”. In ­Wittgenstein on Objectivity, Intuition, and Meaning. James Conant and ­Sebastian Greve (eds.). New York: Cambridge University Press. Cappelen, H. (2012). Philosophy without Intuitions. New York: Oxford ­University Press. Carston, R. (2002). Thoughts and Utterances. Malden, MA: Blackwell. Cohen, S. (1998), “Contextualist Solutions to Epistemological Problems: ­S cepticism, Gettier, and the Lottery”. Australasian Journal of Philosophy, 76: 289–306. Cohen, S. (1999), “Contextualism, Skepticism, and the Structure of Reasons”. Philosophical Perspectives, 13: 57–89. DeRose, K. (1992), “Contextualism and Knowledge Attributions”. Philosophy and Phenomenological Research 52: 913–929. DeRose, K. (1995), “Solving the Skeptical Problem”. Philosophical Review 104: 1–52. DeRose, K. (2002), “Assertion, Knowledge, and Context”. Philosophical ­Review 111: 167–203. DeRose, K. (2005), “The Ordinary Language Basis for Contextualism, and the New Invariantism”. Philosophical Quarterly 55: 172–198. DeRose, K. (2011), “Contextualism, Contrastivism, and X-Phi Surveys”. ­Philosophical Studies 156: 81–110.

204  Avner Baz Emerson, R. W. (2000), “Experience”. In The Essential Writings of Ralph Waldo Emerson. New York: The Modern Library, 307–326. Frege, G. (1956), “The Thought”. Mind 65: 289–311. Grice, P. (1989), Studies in the Way of Words. Cambridge, MA: Harvard ­University Press. Kant, I. (1998), Critique of Pure Reason, Guyer, P. and Wood, A. (eds. and trans.). New York: Cambridge University Press. Kulka, R. and Lance, M. (2009), Yo! and Lo!. Cambridge, MA: Harvard ­University Press. Lewis, D. (1980), “Index, Context, and Content”. In Stig Kanger and Sven ­Öhman (eds.), Philosophy and Grammar. Dordrecht: Reidel, 79–100. Lewis, D. (1983), Philosophical Papers, vol. 1. New York: Oxford University Press. Lewis, D. (1996), “Elusive Knowledge”. Australasian Journal of Philosophy, 74: 549–567. McDowell, J. (1994), Mind and World. Cambridge, MA: Harvard University Press. Merleau-Ponty, M. (1996), Phenomenology of Perception, C. Smith (trans.). New York: Routledge. Recanati, F. (2004), Literal Meaning. Cambridge: Cambridge University Press. Recanati, F. (2010), Truth-Conditional Pragmatics. New York: Oxford­ ­University Press. Schoubye, A. and Stokke, A. (2016), “What Is Said?” Nous 50: 759–793. Soames, S. (2003), Philosophical Analysis in the 20th Century, Volume 2: The Age of Meaning. Princeton, NJ: Princeton University Press. Sosa, E. (1998), “Minimal Intuition”. In Rethinking Intuitions: The Psychology of Intuition and Its Role in Philosophical Inquiry. M. DePaul and W. Ramsey (eds.). Oxford: Rowman and Littlefield, 257–269. Sosa, E. (2007), “Intuitions: Their Nature and Epistemic Efficacy”. Grazer ­Philosophische Studien 74: 51–67. Stanley, J. (2008), “Philosophy of Language in the 20th Century”. In Routledge Guide to 20th Century Philosophy. M. Dermont (ed.). New York: Routledge, 382–437. Travis, C. (1989), The Uses of Sense. New York: Oxford University Press. Travis, C. (1991), “Annals of Analysis”. Mind 100(398): 133–144. Travis, C. (1997), “Pragmatics”. In A Companion to the Philosophy of ­L anguage. B. Hale and C. Wright (eds.). Oxford: Blackwell, 87–107. Travis, C. (2000), Unshadowed Thought. Cambridge, MA: Harvard University Press. Travis, C. (2005), “A Sense of Occasion”. The Philosophical Quarterly 55: 286–314. Travis, C. (2006), Thought’s Footing. New York: Oxford University Press. Travis, C. (2013), Perception. New York: Oxford University Press. Williamson, T. (2007), Philosophy of Philosophy. New York: Oxford ­University Press. Wittgenstein, L. (2009), Philosophical Investigations. Anscombe, G. E. M., Hacker, P. M. S. and Schulte, J. (trans.). Malden, MA: Wiley-Blackwell.

10 Their Work and Why They Do It1 Charles Travis

All linguistic phenomena which only concern the mutual interplay of speaker and hearer, in which, for example, the speaker takes account of the expectations of the hearer and tries to bring him onto the right track before uttering a sentence, have nothing corresponding to them in my formal language, because in a judgement only that comes into ­consideration which affects its possible consequences. (Frege, 1879: 3)

Frege was right to exclude from Begriffsschrift “all which only ­concerns the mutual interplay of speaker and hearer”. For the task of B ­ egriffsschrift was to formalise a class of features of being true as such; of the representation of things, truly or falsely, as being some given way. The central protagonist in this business is what Frege calls a “thought” (Gedanke); precisely, as he puts it, “that by which truth can come into question at all” (1918: 60; 1919: 273). For reasons he makes clear, such items could not be objects of sensory awareness. So a theory of the ­phenomenon of being true (as is Begriffsschrift parts 1 and 2) will have to deal, in one way or another (depending on the form the theory takes) in (visible) ­Vertreter for those features of a thought whose distribution in the corpus (the domain of thoughts) identifies what is intrinsic to this phenomenon. Thoughts (Gedanken) thus have no features by which to come before the eyes; a correlate of the fact that it is no part of their business to make anything recognisable. Their role is not to lead anyone onto a track towards Verständigung. Not to make recognisable what thought was ­expressed; rather to be something thus to be recognised. The role of words is quite another matter. For, in terms of Frege’s ­distinction between the psychological and the logical (the business of holding true (Fürwahrhalten) and that of being true (Wahrsein), language belongs fully to the psychological. Words are devices with which (in concert with other factors) a being such as us might aim to achieve, to create, the representing of something as being something ­(representing-as), thus to gain entry into the business of being true or false. So what Frege rightly prescinds from for Begriffsschrift’s purpose,

206  Charles Travis what merely facilitates recognition, is all that matters to words’ business (or main business). For to author representing (thought-expression) is eo ipso to make this recognisable as expression of the thought it is. (Not yet to expand on the notion recognisable, but very roughly, to indefinitely many, and to those for whom Verständigung, an understanding, might be in the cards.) Perhaps what was engaged in that business might also dabble on the side in the business itself of being true. But such would just be moonlighting, and there is no reason a priori to think that such occurs. This last remark will not strike all and sundry as correct, or at least obvious. For some think that it is precisely by their role in the business of being true that words carry out their mission of aiding recognition. Whether this thought is, or could be, correct is the main topic of this essay. Our starting point is just that achieving thought-expression is words’ first business, their raison d’être. Which opens this last question to serious investigation. The question, though, does point to one ­ulterior motive for the present exercise, to which I should at once confess. Since about 1968 I have been defending an idea which I call occasion-­ sensitivity. Suppose we speak, as we might, of a well-formed, meaningful, predicative expression of a language as speaking of a way for things (or for a suitable n-tuple of things) to be. For example, we might speak of the English ‘takes tea at 3’ as speaking of a person (or suitable object) as taking tea at 3. So for one to take tea at 3 would be a way for one to be. Then the principle is: the ways for something to be there are for us to speak of admit of understandings, where what it would be for a relevant item to be the way in question, and what would count as something so being varies from one such understanding to another, and where for any such understanding there are occasions on which the way in question might rightly be so understood. So, for example, if the sentence ‘Pigs fly’ speaks of pigs as fliers, there are various things it might be understood to come to for pigs so to be, where different of these would differ in what, on them, would count as pigs so being. Such rules words of a language out of direct involvement in the business of being true (on which, again, more anon). I once thought I might convince others of this principle simply by ­exhibiting it. But with age I came to appreciate that what can be seen by looking depends on the intellectual climate in which such looking is to be done. What is proof depends on one’s starting point. Whether (e.g. English) words, in what they speak of, are thus distanced from the business of being true is not something which can be investigated in this way if the idea that they might be strikes enough philosophers as too extravagant for words. Here, then, a different tack. The present idea is to try to identify something in the thought of a being such as us which would require occasion-sensitivity in the way such a being thus thought things; thus occasion-sensitivity in language fit for such thought’s expression.

Their Work and Why They Do It  207 Before occasion-sensitivity can be simply observable, it seems, we need a good reason to expect it. The lead ideas in this search for reason will be twofold: first, that we are Cartesian thinkers in Descartes’ 1637 sense; second, that we are ­fi nite, and hence fallible, thinkers. That first idea finds current expression, for one, in Hilary Putnam’s maxim, “Reason can transcend whatever it can survey (or comprehend)”. Another expression is in John McDowell. Put in terms of the conceit of a rational wolf, it looks like this: A rational wolf would be able to let his mind roam over possibilities of behaviour other than what comes naturally to wolves. … Suppose, now, that [he] finds himself in a situation in which some behaviour would come naturally to him, say … hunting with the pack. Having acquired reason, he can contemplate alternatives, he can step back from the natural impulse and direct critical scrutiny at it. (1996/1998b: 170–171) And it is just another form of the same point when he writes, [I]t is only this misconception of the deductive paradigm that leads us to suppose that the operations of any specific conception of rationality in a particular area – any specific conception of what counts as doing the same thing – must be deductively explicable; that is, that there must be a formulable universal principle suited to serve as major premise in syllogistic explanations of the sort I considered above. (1979/1998a: 62) The idea is: as Cartesian thinkers (as I will label this), whatever it is we are doing, we have the capacity to step back and entertain reasons which we can see to bear rationally on whether this is how we should be doing it, or what we should be doing at all. In particular, whatever, in given circumstances, might make it recognisable to us that things were thus and so, we are always open to seeing such recognisability as exploiting the hospitality to it of that environment, thus hospitality liable to lapse with a change of circumstance, or to prove only illusory as things stand. This idea will be central in what follows. But I do not mean for it yet to be taken as axiomatic, or as self-evident. The second idea, that we are finite, hence fallible, thinkers, I do take to be evident enough, though in philosophy there is a long history of looking for a loophole. I take it as at least to be so of our efforts at thought-expression and its understanding (recognising what thought was expressed). Our efforts may, of course, misfire: what we say, or when things would be as we represented them,

208  Charles Travis might be other than what we meant this to be; what we take to be so as to what thought was expressed may be other than what is so of this. How words would have been to be understood, ought to have been taken, in given circumstances need be neither how they were understood by (say, intended) hearers, nor how they were meant to be understood by a perhaps deluded speaker. But also, a speaker may be under illusions as to the nature of what he is relying on in the circumstances for achieving recognition; illusions he may share with his intended audience – a point to become clearer in section 4. At the same time, as Frege argued at some length, liability to error is inherent in the possibility of truths and falsehoods überhaupt. For, to put it in one way Frege does, it is intrinsic to that countable, a thought, that “The same thought is graspable by multiple people” (1915: 270). From which it follows that it is intrinsic to there being ways for things to be, thus to be represented (truly of falsely) as being, that, for any given such way, any given thinker, or collection of same, is always susceptible to getting it wrong as to whether things are actually such as to be that way. There is, and can be, no form of privileged access to the way things are which grants immunity to error. As for finitude, as Frege once put it, It would be part of an all-encompassing knowledge of a Bedeutung that for any given Sinn we could say right away whether that Sinn belonged to that Bedeutung. We never attain to this. (1892: 27) A thought is a countable, carved from some mass, thought (how things may be thought to be). It represents things as bringing some way for them to be – say, such that pigs fly, or that Pia is chairing; such way itself carved from a mass, things being as they are. (‘Things’ here not a plural countable, inviting a question ‘Which?’, but rather what blocks such questions; simply all that truth might turn on, henceforth to be written ‘thingsm’.) Such paths to countables open up, not just the possibility of representing truly or falsely, but also that very different sort occurring within domains of factive meaning. Finitude and fallibility at least suggest that there is a distinction to draw between what, in that episode, it would have been right to suppose as to what thought was expressed – in particular, as to what it would be for things to be as that thought represents them – and what is in fact so as to this, thus as to what way for thingsm to be was spoken of by those words in that mouth. And insofar as, e.g. English words may be said to speak as such of some way for thingsm, or for a thing, to be, mutatis mutandis for them: there is what they would have been (or would now be) supposed to speak of, and then – a different matter – there is what they in fact do speak of. Hilary Putnam has elaborated this last point.

Their Work and Why They Do It  209 J. L. Austin brought the first into view. The present aim is, with their help and with Frege’s, to show how these are two facets of one point, which, further, ought to hold. 1. Recognisability: One principle announced already is this: when it comes to thought-expression – to the authoring of representing-as – to be is per se to be recognisable. Why should this be? It is a familiar point by now that representing things as thus and so need not be, often is not, representing them so to be. Questions of the truth of our thought-­ expression do not always, inevitably, arise. Not all ­representing-as is eligible for representing either truly or falsely. One may, e.g. represent things as such that Sid will show up sober for the meeting in representing such as to be made so. Nonetheless, wherever there is things being/having been represented as being such-and-such way, there is truth-­evaluability. If that Sid will show up sober is represented as to be made so, there is at least the question what would count as his having so shown up, when it would be true to say that such is how things were. So there is some reason, in a discussion like the present one, to focus on cases of representing-to-be – on, so to speak, the assertive case. With which an explanation of the importance of recognisability may begin with Frege on holding true. One thing Frege insists here is that to hold a thought true cannot be to predicate something of it. Such simply yields anew a thought, still awaiting being held true or not. In fact, truth being an identity under predication, it yields the very same thought all over again. Thus, he tells us, One moves around on the same level, but one does not step from the one level to the next. (1892: 35) One moves around (or rather treads water) in the realm of representations. But one does not yet concern himself with that which these representations represent as thus and so – in the case of truth outright (that of any whole thought) just thingsm. One does not yet treat this as being as it would be were the thought in question true. Frege later elaborates the above thus: The word ‘true’ thus appears to make the impossible possible, namely, to let what corresponds to assertive force appear as a contribution to the thought. And this attempt indicates, although it misfires, or rather more precisely through its misfiring, what is ­peculiar to logic. … ‘true’ really only makes a failed attempt to single out logic in that what this really comes to does not lie at all in the word ‘true’, but rather in the assertive force with which a sentence is uttered. (1915: 272)

210  Charles Travis Logic answers the question how one must think to reach the goal truth (insofar as such is answerable by what being true is as such) (1897: 139). It tells us how to stay on course, once on it, towards the goal truth – the truth as to whatever may concern us, notably, as to what to do to reach whatever goal we may pursue. Logic is one guide as to what to do (or how to think). More generally, any representing-to-be is of what may guide us. The force with which a mere ‘as’ after ‘represent’ becomes ‘to be’ is that of offering guidance in pursuit of goals. To assert is to commit (as only a thinker could). If Pia tells Sid that the cellar is bare, she pretends to, and offers, authority as to how things are. Following Frege, she thereby underwrites (his) so treating things accordingly. If such would dictate an urgent trip to the nearest beverage depot, then such is dictated. Or Pia so commits. To underwrite is to assume liability. If the cellar is in fact well stocked, Pia thus incurs blame for wasted effort. So, following Frege, to assert – to represent to be – is eo ipso to a­ ssume responsibility, liability to praise, blame, or other sorts of revenge or reward. One stops nowhere short of offering a guide to conduct. With which (as already indicated) issues of fairness and justice enter the very business of achieving thought-expression. Pia described the cellar as ‘bare’. There is what she meant by that. (What is not wine, more specifically bubbly (by default Billecart-Salmon) is not to be reckoned contents of a cellar, as she understands that word. (A cellar is no mere hole in the ground.)) There is what Sid understood by it. (For him a cellar is where to look for one’s usual quaff.) But where fairness is at stake the issue becomes what (a relevant) one should have understood, in the circumstances, as to what Pia was doing in speaking those words as she did; how her words ought to have been taken, in other words what was then recognisable (to a reasonably endowed observer). Here, then, a connection in the nature of the thing itself: to have expressed (given) thought is for such to be what was thus made (suitably) recognisable as the thought expressed. What thought was expressed is thus, to that degree at least, a matter of circumstance; in the verbal case, of what in them would have been made recognisable in speaking the words used as was then done. What has just been said does not yet quite rule out a semantics for a language presupposing language’s direct participation in the pure business of being true, and not just as aid to a speaker in gaining entry into such business. It does, though, provide occasion and means for bringing such assumption into question. The slight element of caginess one might detect in this will soon unfold itself. 2. The Ready-Made View: The thought (der Gedanke) is designated by Frege as principal protagonist in the business as such of being true. It is party of the first part in an exclusively two-part enterprise, that of ­being true (or false). Its mission is to demand something of party of the second. The mission of party of the second, thingsm, is simply to be.

Their Work and Why They Do It  211 In so doing it fulfils the demand or party of the first part, or again fails to. A thought so conceived is, as Frege points out, ‘invisible’, that is, not a possible object of sensory awareness. For if two thoughts are distinguished only by their different roles in that two-party enterprise, by the different demands they put on party of the second part, then features which belonged to objects of visual awareness – colour, shape, etc. – would not distinguish one such item from another. Moreover, as Frege insists, a thought is decomposable – structurable – in any of many different ways, none of which in general, as Frege put it, “may claim objective priority” (Vide 1882a: 118). So we must not think of such a thing as a domain of thoughts, the whole domain generated from some ‘vocabulary’ of antecedently existing thought-parts by some given set of ‘syntactic’ rules. So a thought is structurable in many equally permissible ways, and we must not think of it as having some one intrinsic hallmark structure. In all of which a thought contrasts with a sentence. Where a thought is ‘invisible’, a sentence had better not be. For, at the least, it is designed to be a content bearer. It is recognisable as the sentence it is simply by its place in the syntax of its language. There may, typically will, be such a thing as what the sentence ‘speaks of’, or ‘says’ (on which more in the next section). But to say what this is to take an extra step beyond exhibiting its syntactic structure. So, unlike a thought, there is, for each sentence, some one structure such that for something to be an instance of that sentence is for it to be structured thus. Thoughts (significantly) do not have instances. They occur, or are expressed, or thought. But whether it is a given thought which was expressed in some given words is an issue not settled by some structure the thought must then have been presented as having. That in a sentence which distinguishes it from a thought is just what one would expect of a content bearer, and of something whose role it was to make the expression of thought recognisable. As that in a thought which distinguishes such items from sentences is just what one would expect of that whose raison d’être lay in the business of being true. A thought’s business is not to make anything recognisable, but rather (at most) to be, or identify, what would be made recognisable where thought-expression was achieved. It is thus a bit surprising that it should have come so naturally to so many philosophers, at least in the last century, and so far this one, to assume that words of a language (in which for thinkers to express thoughts) should be, per se, in the business of being true; that a sentence should (as a rule) be an expression of a thought, or at least a proper thought-part. Such, though, has been, and, I think, is still, the received view. To be sure, in the years since Frege (and since Carnap) that view has become a bit nuanced. First, though, the unvarnished version. On this version, a (typical) intelligible declarative sentence is, modulo lexical or syntactic ambiguity, an expression of some given thought. A thought,

212  Charles Travis Frege tells us is “that by which truth can come into question at all” (Vide, 1918: 60). It thus poses a determinate question of truth, makes truth turn in a determinate way on how things are. So, too, for (such) a sentence on this view. A syntactically well-formed part of the sentence, correspondingly, partially poses a question of truth, makes truth turn in part in such-and-such way on how things are. Such is something it does in being the expression it is, whatever the sentence in which it occurs. The syntactically well-formed parts of a sentence in concert, and in that construction, pose precisely the question of truth that the whole sentence does. For a sentence to express the thought it does (on a reading if ambiguous) is for it to mean what it does (on that reading). So the sentence is identified as meaning what it does – as a thought is identified as the thought it is – in identifying when it would be true. Mutatis mutandis for proper thought-parts. Since Frege (if not before) it has been clear to everyone that, au pied de la lettre, this cannot be quite right. The usual, or at least most plausible, remedy is to suppose that what has the form of an expression of a thought, namely, a whole sentence, is, despite appearances, actually, to a logician’s eye, the expression of a proper predicative thought-part (a predicative sub-thought so to speak). In other terminology, it is in fact an open sentence. For any such sentence, there is a given n, n > 0, such that what the sentence expresses is what would be true or false of a suitably composed n-tuple. For example, if the sentence is ‘The cellar is bare’, what it might be true or false of is an ordered pair made up of a cellar and a time. But it may be, for all it matters here, that, for a typical sentence, the relevant n is considerably greater than 2. For a sentence to mean what it does, on this approach, would then be for it to map the n-tuples it does into the truth-values it does, given that thingsm is as it is. This general approach owes a debt to Carnap. So, with no disrespect for him implied (or deserved), I will dub it the utopian socialist approach (or view). The reason for this will emerge only at the end here when we consider certain claims on behalf of the conception that it is mandatory if we are to be scientific. Part of the present brief is that the business of language is ­F ürwahrhalten; that it is a tool for achieving thought-expression. The case to be made cuts equally against either version of the Ready Made view. (One gets the simple version from the Carnapian by just setting n at 0.) It will help to see what is at stake if we think of the issue as primarily, or even entirely, about predicative thought-parts, proper or not, that is, about what a sentence, if really open, predicates (on that aspect of the verb on which predicating is something a sentence might do) of whatever sort of n-tuple might fill the opening(s). If sentences were, as such, each an expression of some given thought, as per the simple view above, then to use a sentence as meaning what it did (and thereby to express a thought) would be, eo ipso, to express the

Their Work and Why They Do It  213 thought it does. Conversely, if there is some thought you wish to express (in a given language), what you need to do is to find some sentence of that language which, per se, expresses it. You would thereby make your thought-expression recognisable as expression of the thought it is to one who knew what the sentence means (or its well-formed parts do) – to one, that is, who spoke the language and thus recognised the sentence ­ utatis as meaning what it does, and who recognised you as so using it. M mutandis for the utopian socialist approach. Here language would make it recognisable what thought you expressed of given items in some relevant n-tuple, and it would only remain for a hearer (or the speaker himself) to work out just what n-tuple that was. In either case, on this view, so much is what language has to contribute to making recognisable what thought was expressed in a given instance. For this reason I will dub either of these view the ‘Ready-made View’ of language’s contribution to thought-expression, RMV for short. 3. Turbidities: Though they are so far but straws in the wind, we have already seen some reasons not to suppose that the business of language, or anyway primary business of language, is the business of being true (Wahrsein). Why, then, does it come so naturally to so many to suppose, generally without argument, that language is in that business, and, moreover, in any given instance, identifiable as the language it is by its role in this? Perhaps there is no concise or comprehensive answer to this question. Some, I suspect, see such as mandated by the requirements of logic, or its laws, on whatever it might govern – demands Frege’s countable, the thought, was designed specifically to meet. So there is a conflation of thoughts and sentences – one aided, perhaps, by that versatile German word, ‘Satz’. Others have seen such as mandated if the phenomena of thought-expression, and of Verständigung, are to be a tractable topic for a science – a view which might merit the title ‘utopian socialism’. We will return later to both these themes. At the start, though, it will be prudent to be on guard against some notorious grammatical traps by which one might be seduced. These traps have to do, most centrally, with aspect. Such issues arise, not just with language, but also with the notion of a thought. One can say: a thought represents thingsm as being thus and so (and it is thus that it is eligible for being true outright). But we represent things as thus and so in authoring representing, or merely in thinking of things as being thus and so. Surely a thought cannot ­author representing. Nor can it think. An author bears responsibility for his authorship. He commits to something (most straightforwardly in the assertive case). Surely a thought cannot commit. Nor is a thought a thinker, an entity which might hold some attitude towards things being thus and so. Which just means: a thought represents in a different aspect of the verb ‘represent’ than that which fits a thinker, something on this order: for a thought to represent things as such that aardvarks eat ants is

214  Charles Travis for it to be a thought which is expressed just where someone represents things as such that aardvarks eat ants. Now consider such words as ‘Gentlemen take thirds’, or ‘Death ­cancels all obligations’. One might say, not incorrectly: they speak of, respectively, gentlemen taking thirds, and death cancelling all obligations. In them, ‘Gentlemen’ speaks of gentlemen, ‘Death’ of death, ‘take thirds’ of (some plurality of items) taking thirds, and so on. One might also say, not necessarily incorrectly, that the sentence ‘Gentlemen take thirds’ says gentlemen to take thirds, and so on. Again, though, we need to note the aspects. Sentences do not speak, nor do they write. Nor, a sentence not being a thinker, do they commit to anything. The sentence ‘Gentlemen take thirds’ is not guilty of propagating falsehood simply because a II.1 is the new gentleman’s third. Again, though, those verbs ‘speak of’ and ‘say’ need not suggest any such things if only we attend to aspect. And here, for a start, we might follow the above pattern for thoughts. The sentence ‘Gentlemen take thirds’, one might reasonably suppose, is for something. Specifically, it is for speaking of gentlemen as third-takers (or, again, for reminding some swot of what it would be to be a gentleman); or, equally, for speaking of gentlemen taking thirds. There is something else it may be said to be for: it is for (inter alia) saying that gentlemen take thirds. If such is what you want to say, then this is the sentence for you. So use these words for what they are for and you will speak of gentlemen as third-takers. Do so with assertive force and you will say gentlemen to be third-takers. Such is roughly what the verbs ‘speak of’ and ‘say’ say in one aspect they are liable to take; one on which they might apply, even truly, to sentences. The aspect in question is much the same as that in which one might say of a switch on the wall above the night table that it turns on the reading light above the bed. Well-behaved switches do not take on such initiative. If a switch did so, it would be time to replace it. But all one need say in so speaking of it is: flip the switch and if there is no power outage or etc., then you will turn on the reading light above the bed (if there is such a light). This last points to one finer level of distinction. There on the wall is the switch. And there a bit further on the wall, above the bed, is a reading light. Now, ‘The switch turns on the reading light’, can bear either of two understandings: first, that it turns on that reading light – for each light its switch; second, that it turns on the reading light (whatever particular light may play that role at a particular time), no matter what light this may be (again, all in working order, etc.). In our normal commerce with reading lights such a distinction seldom needs marking. In the matter of the relation between language and its use in thought-expression the parallel distinction does need marking. For one might easily enough fail to see how attending to aspects might really defuse any Drang one might feel towards taking words as such to be in the business of being true. For one might respond to the point

Their Work and Why They Do It  215 just made on these lines: ‘Sentences don’t say things to be so; speakers do. Good. Guns don’t shoot people, people do. A nicety (if even that much) to no purpose.’ If the sentence, ‘Gentlemen take thirds’ speaks of gentlemen as taking thirds, in that just-indicated aspect of the verb, this thought runs, then if you use it for what it is for, and thus use it assertively, you will say gentlemen to take thirds. You will speak of that way for thingsm to be and represent it as a way things are. You will speak truth or falsehood according as that is, or is not, a way things are. For gentlemen to take thirds is a way things either are or are not. So to do what you would thus eo ipso do would be, as the case may be, either to speak truth, or to speak falsehood. Wherever the sentence is used (assertively) for what it is for, one thus always speaks truth or, as the case may be, always falsehood. Which is a perfectly good sense in which the sentence itself, in saying what it does in that aspect of the verb, is either true or false – roughly the same sense in which a thought may be either true or false. In which case, after all, language is in the business itself of Wahrsein (being true). This, though, simply begs a main question at issue in what follows. The sentence ‘Gentlemen take thirds’ speaks of gentlemen as taking thirds. So if you use it for what it is for, then you speak of gentlemen as taking thirds, a certain way for things to be. If, moreover, you do so assertively, then in speaking of things as that way, you thereby express a truth or, casu quo, a falsehood. What you do is what it would then be to say things to be such that gentlemen take thirds. You say what you would then say in speaking of things being that way. But is this always the same thing no matter what the circumstances in which you do so? Compare: that switch is for turning on the reading light over the bed. So if you flip the switch you turn on the reading light over the bed. The same light no matter what the circumstances in which you do so? This need not be so. On one night you might turn on the 1-watt led, on another the klieg light, depending on what light then plays the role of the light over the bed. Might it be on some such reading of the verbs ‘speak of’, or ‘say’, that it is true that the sentence ‘Gentlemen take thirds’ speaks of gentlemen as taking thirds? The answer to this depends, inter alia, on in just what ways the circumstances of an episode of thought-expression might contribute to making recognisable, so determining, just what thought was expressed. Such is the topic we have only just begun to investigate. 4. A Model: One use for a definite description (primary or not) is in the expression of a singular thought. Its role on such use is to work towards making recognisable with respect to what object the thought thus expressed is singular. By a singular thought I mean one of some object that that object is thus and so: to be that thought is thus, inter alia, to be of that object; hence, but for that object’s existence, there would have been no such thought at all. Thus there is, say, a thought of Sid that he takes tea at three, expression of which might, on some occasion, be

216  Charles Travis made recognisably a thought of him in speaking the words, “That man who just ordered another pint”. The points I want to stress in this section are, first, that there need be no direct connection between the work thus done by a description and the work done in the thought expressed in making itself about some given object, or in making some object the one it is about – that is, in making its truth turn in part on how that object is. And second, that there is reason to think there is no such direct connection. There is the business of being true, and the business of arriving in the business of being true (or of representing truly or falsely); the business of a thought expressed, and the business of expressing it. Language (words), we are supposing here, are in that second business, and only incidentally, if at all, in the first. The work required for this last business may be very ­different from that required for that first. Such is a point rightly insisted on by, among others, David Kaplan (1989) – for the last 60 years our leading utopian socialist semanticist – though in service of an agenda I disclaim. It is an important point for a number of reasons. For one thing, it works to break the hold of the idea that when someone expresses a countable bit of thought, there is some one thought (one thinkable) which is, full stop, the one expressed; so that what thought was expressed in some given episode of thought-­ expression is fixed entirely by the particular circumstances of that episode (inter alia, the words used and what they mean or (in the case of nonce-senses or ambiguity) then meant. But such is not that import which presently matters. Rather, I will develop the point so as to use it as a model for how, in general, means, notably verbal ones, of thought-­ expression relate to the thought thus expressed. The virtue of this case for that purpose lies largely in the fact that a description so used makes something explicit as to how the words contribute to recognisability. The first question, then, is: Just what? Time for an example. Pia is at the dean’s start-of-year reception. She stands by a potted palm, surrounded by a small coterie of friends. Over their aperitifs they watch, with some detached amusement, the activity at the buffet. Pia remarks, “That man at the buffet wolfing the last of the foie gras is writing a review panning Benno’s new book”. Pia might then rightly, fairly, be perceived as supposing there to be an object which, in the circumstances, one would suppose to be standing wolfing, and, by virtue of which status, would (or should) be taken by those in whom she aims for understanding as the one she is representing as in process of panning Benno’s book. Suppose there is such an object, say, Sid. Then Sid is the one one might then suppose to be standing wolfing, and who, by virtue of so appearing would, or should, be supposed by Pia’s party (or those in whom she aims for understanding) as the one she is representing as engaged in panning Benno’s book. So one who understood her as above, and was sufficiently aware of the circumstances, would take

Their Work and Why They Do It  217 Sid to be the one she was representing (and, moreover, did represent) as panning Benno’s book. Which is at least a prima facie case that for Pia, and thus her words, to have borne the understanding she/they did – for her to have been so to be understood – is, in fact, for her to have ­represented Sid to be in the course of panning Benno’s book. If all is well so far, then Pia expressed a certain thought, of Sid, that he was panning Benno’s book. Now that point of my convergence with Kaplan applies. Sid standing and wolfing (or at least appearing to do so) is what, in the circumstances, makes him recognisable as the one with respect to which the thought expressed is singular. That thought, a singular thought, would not have existed had Sid not. It does not follow that it would not have existed had Sid not been standing wolfing, etc. No such considerations need be doing any work in that thought’s being singular with respect to Sid. Their work was rather, in something like the way just scouted, to make recognisable what thought it was that Pia was expressing. The contingencies of circumstance she thus exploited need have nothing to do with what it is for that thought to be the thought it is. Equally for the words she used to call attention to some relevant contingencies. Such sketches a way words may work towards making recognisable what thought was expressed in a given instance of thought-expression. In outline the story works like this. A given use of given words is to be understood as, inter alia, supposing there to be something (here an ­object) which a relevant audience would (or at least ought to) take as identified in being what those words would speak of in those circumstances (used as speaking of what they then would in speaking of (meaning) what they anyway do; thereby as identifying an item on which the thought expressed makes truth turn; thereby identifying in part the thought being expressed. The role of a definite description on this use is to provide some of what is to be taken to be being supposed of the item thereby identified as the one the thought is about. It is not to identify any further feature of the thought thus made recognisable as the one expressed. ­Notably, if the identity of that thought turns in part on the way in which, in it, the item in question is made the one on which truth thus turns, it is no business of the description used to make this identifiable. Nor is there any reason a priori to think that such an ingredient in the thought would match in any way the content of the description as so used. If, e.g. the thought expressed is of Sid that he was busy panning ­Benno’s book, then the description, ‘The man at the buffet wolfing …’ will have done its work if it makes this much recognisable as to what thought was expressed. It is no part of the bargain, nor need it be so, that the way the thought expressed makes its truth hostage to how Sid is by deploying the concept being that very object, the man now at the buffet wolfing. The business of making recognisable that a certain thought was expressed and the thought’s business of bring truth into question in a

218  Charles Travis determinate way (making it turn in a determinate way on how things are) are two ­entirely different enterprises, with no need for the one to mirror the other (and, as we shall see, good reason for it not to). So suppose, for example, that Sid was not in fact wolfing foie gras, but only a reasonable facsimile thereof, or that he was not wolfing, but only creating a diversion by which surreptitiously to slip the foie gras into his breast pocket for later. Or that it would be a stretch to call what he was doing wolfing. Or that it was not the last of the foie gras, two entire new ones hiding behind the ice swan. For all of which, the circumstances may well have been such that a relevant audience (e.g. Pia’s) would have taken her use of the words to be supposing that Sid would be taken to be the one then picked out by that description – even if some of that audience in fact knew that he was not. In which case, no matter (for a description’s role in this use) that this one does not in fact fit what it picked out. Why stand for such unruliness? A simple answer: Justice and fairness. Zoë, one of the party, makes the mistake of inviting both Sid (the one at the buffet) and Benno to the same dinner on Thursday. Has she been warned? There was, anyway, what Pia told her. But what was that? If Zoë is bad at faces, then perhaps though she understood what Pia said, she just did not recognise the man at the buffet as Sid. Or if, in those circumstances, no one would have supposed Pia to be using those words to identify that man (when his ruses would have fooled no one into thinking the foie gras was actually going in his mouth). In which case, Pia gets no credit for warning anyone of anything. But it would be unfair to distribute credit and blame in such a way that Zoë could claim truly not to have been told on grounds that in point of fact what Sid was eating was soy substitute. A rule of language which so dictated (were there such) would be unjust. We now have two notions of bearing an understanding. There is the understanding Pia’s words bore, or ought to have been taken to bear, in the circumstances in which she uttered them. And then there is the understanding they bear sub specie æternitatis, correctly viewed from without the circumstances of their production. The first sort of understanding is in terms of what, in the circumstances, would have been (to be) supposed as to how words were working to make recognisable what thought was expressed; and thus, inter alia, as to what was thus being spoken of (what, or how, what was spoken of was to be supposed to be). The second would be simply in terms of how, in the thought expressed, truth was made to turn on how things are; where to understand this would be to grasp correctly just what the items are which were represented as being some way or other, and just what the ways are which they were thus represented as being. In our sample case it would be to grasp adequately just who, in the thought, was represented as ‘busy panning Benno’s book’, and what it would be for one to be as this person was thus represented.

Their Work and Why They Do It  219 An understanding of the first sort is correct or not, inter alia, by the standards of justice and fairness. It corresponds, inter alia, to a correct assignment of commitments, or responsibility, to the speaker on the basis of that speaker having so spoken. It is a basis for judgements as to when minds met, or ought to have. When two thoughts compete for the title the thought expressed, such an understanding is the basis, if any, on which there may be a victor: there is one just where there is something in such an understanding which so decides. By contrast, there are two thoughts, so competing or not, only where each differs from the other in how it represents things to be; thus in what it would be, in each case, for things to be as represented. Two very different matters, to be decided by two very different sorts of understandings. Language’s contribution is to the first sort of understanding. Words, in meaning what they do, need contribute to supplying no more of an understanding of this. The understanding required for the business of Wahrsein is of a very different sort. Language, if it supplied this, would be just dabbling on the side. 5. Predicating: Our next task is to apply the last section’s model to predication. We can at least find words to do this in that model itself. The question would then be what might lie behind these; what phenomena might require them. But to deploy these words, let us start with an example. It is Zoë’s shout. She returns from the bar with the orders: a lager for Sid, of course, an IPA for Pia, and, also of course, a g&t for her good self. As this last suggests, Zoë does not know her lagers from her ales. She is thus, by now, unsure which pint is Sid’s, which Pia’s. Pia helps out. “That (pointing at a glass) is lager”. Pia, I think we may suppose, has expressed a thought. In it something is predicated of the contents of a certain glass. The words she used are tasked with helping to make recognisable what thought this is. As that task is apportioned, we may also reasonably suppose, it falls to the lot of the words, ‘is lager’ to make recognisable what was predicated of the relevant stuff. Pia’s words are then to be understood as supposing there to be something (here a way for a (liquid) substance to be), where the words ‘is lager’, used as in her mouth, would work towards fixing what, in the circumstances, would be to be supposed as to what this something is, and where, first, they would contribute what they do in meaning what they do (as did the words ‘at the buffet wolfing’), and, second, the something in question would thus be that way-to-be (if any) of which all this would have been, in those circumstances, to be supposed. Those English words are for such use. Pia, we are supposing, is using them for what they are thus for. What they work towards fixing is what, in those circumstances, one would speak of in speaking of something as being lager. What would thus be fixed is what, in those words, was predicated of the liquid in that glass. Not that Pia’s words must have fit such description. The point is to identify a use predicative English words

220  Charles Travis would have in and by meaning what they do (and on a given reading, may they have more than one in English). One could describe Pia’s act in this rather tortuous way. But why do so? In the case of ‘The man standing wolfing …’, for all that Pia was to be understood as above, any of many objects may have thus emerged as the one she spoke of, depending on how the world was then arranged. There is no requirement, of any of these, that it be in fact standing wolfing (etc.). By contrast, one might think, in the present case there is no room for a parallel contrast: use the words, ‘is lager’, to speak of what they are for speaking of, and, eo ipso, you have spoken of being lager. And one would speak of a substance as just the same way wherever one spoke of it as being that. But the point of the framework is to make room for just this last not to be so. It anyway makes room for two points to apply. The first is this: if the work of language is in achieving thought-expression, then the contribution it needs to make to a given such effort is one to what the speaker was then to be understood to be doing. Such would include, notably, an understanding as to what he was speaking of, inter alia, as the way relevant objects were. But such an understanding might consist in grasping what was then to be understood to be given to be supposed as to what this was. The speaker may be to be understood as aiming at asserting truth. The words he used have a dedicated role to play in making recognisable just what thought, if true, this truth would be. He may be to be understood as using them in that role. But it is not eo ipso part of playing the role the words thus do that these words must also play the role of that thought itself, or some proper part on a relevant decomposition – the role of making truth turn (in part) on how things are. Recognisability arrives, if at all, with what would, with right, have been supposed on the occasion as to how the speaker was representing things as being (given that enough was to be supposed). What was in fact said cannot transcend what would have been said where such was to be supposed. But the step from there to what is, in fact, so as to this is always (in principle) substantial. To understand, or grasp, the English ‘is lager’ as speaking of what it does – in the circumstances in which we now live – is to grasp (enough of) what it would, or might, be correct to suppose this to be. Such, ineluctably, is the sort of understanding an expression of a language bears (a point to be expanded on further anon). So it would be a happy accident, if accident it were, that just such an understanding is what those words, ‘is lager’, are required to supply in their role in thought-­expression. It might be, for all that, that what it would, or might, now be right to suppose as to what way for stuff to be being lager is, leaves just one thing which would, or could, be what this would be supposed of, at least given that our environment (the world of grains, yeasts, and beverages) is what it is. This idea is not yet in dispute. But it is at least so that what this

Their Work and Why They Do It  221 one thing is (if there is such a one thing) is underdetermined merely by what it would have been supposed to be. Again, a point to expand anon. In any case, the conceptual dynamics at work in determining what this one thing is (if there is such) are just those which, applied to particular episodes of aiming at thought-expression, make room for the answer to depend on just what such episode we have in mind. Thus the second point. The words, ‘is lager’, presumably speak of (a stuff) as being lager. Such is something they may be presumed to do independent of any episode of their use. What they thus speak of would, on the present model, be fixed by what, occasion-independently, would be to be supposed as to what lager, or being lager, is. Suppose that, on an occasion, a speaker uses these words of some given stuff. Suppose he uses them as meaning what they do, as for (speaking of) what they are for (speaking of). What he thus speaks of, what he thus says the relevant stuff to be, is then what one would speak of on that occasion in speaking of something as being lager. It is not automatic that what this is fixed independent of the occasion. And there is at least some reason to think otherwise. What might pass for lager down the pub (or anyway Sid’s pub) might not fit the bill in more refined society. Nor might what so counted down the Kneipe so count in Sid’s circles. (Again more anon.) The general point so far, then, is that there is a (conceptual) gap between what would have been (to be) supposed as to what given words speak, or spoke of, and what is, in fact, so of this. For speaking of an object, the gap is between what would have been (to be) supposed as to by what the object spoken of was identified as the object it is. For speaking of a way for an object (or for thingsm) to be, it is between what would have been supposed as to what it would be for an item to be that way, and what this would in fact be. That ‘speak, or spoke’ above just indicates that the point applies both to words of a language (where, perhaps it is more familiar) and to words in the mouth of an author of representing-as. The question why there should be such a gap has been answered in detail by Hilary Putnam, and before him by Leibniz (vide 1703, book 3, chapter 6, especially 269–270). The idea could be seen as an elaboration of Frege’s insistence that one cannot define things into existence. The master insight in Putnam’s elaboration of this idea (and Leibniz’) is that the things of which we speak and think (what our concepts are of) are generally identified (or supposedly identified) as the things they are by a whole made of multiple strands, where, at least as a rule, these strands are not immune to falling apart, should thingsm prove to differ enough from what one would have supposed. To take one of Putnam’s examples (1962), suppose we view Euclidean geometry as a theory of space. (As a mere piece of mathematics, the following would not apply.) A notion straight line would be defined within this theory. A straight line would be, (a) the shortest distance between two points, and (b) a member of

222  Charles Travis a family of lines such that for any pair in the family, the segment of any (straight) line orthogonal to and cutting both would be a constant length. Thus, on the definition, straight line is a concept with multiple strands. We are now able to suppose that the geometry of space is, in fact, Riemannian (roughly that of the surface of a sphere). If so, then the strands come apart. In such a space there would be nothing which satisfied both a and b. In either Euclidean or Riemannian geometry there is such a thing as a straight line. But in Riemannian geometry there is no such thing as a Euclidean straight line. So, for sure, the notion straight line is to be preserved. But once we get that far (its space is Riemannian) our hands are tied: a straight line does not – cannot – have feature b. What we were speaking of all along in speaking of straight lines would certainly (rightly) have been supposed to have feature b. But, as it turns out, it does not. Thus the need for the space made in the general point so far. A simpler example, given by both Putnam and Leibniz, concerns what it is for some stuff to be gold. Here it is important that one of the strands in the concept to be gold is that gold has a certain distribution in the world, much of which familiar to us. (Some of us, e.g. need only to look at a certain finger on our left hand.) Other strands would once have included the idea that gold is a yellow metal. As the chemistry of the world has turned out to be, that idea about yellow must be dropped: must be on pain of the idea that gold is a familiar metal, or so much as a familiar stuff. The general point is about words, and their relation to what they speak of. Thoughts, those countables by which (as Frege put it) truth can be brought into question at all, determinate truths or falsehoods, do not admit of understandings or interpretations, certainly not multiple ones. They are contents for a thought-expression to have, items to be expressed, not vehicles, or bearers, of such content. Nonetheless there is a corollary for thoughts here. Suppose that we can identify a thought. For sake of illustration, suppose that I do so in now speaking of ‘the thought that the stuff in Sid’s glass is lager’. There is what now (in the Anglophone world) would be to be supposed as to what lager is, as to when some stuff would so be. Such would now rightly be supposed of the just-mentioned thought, that is, of how, in it, the stuff in Sid’s glass is represented as being. But lager is a conspicuously multi-strand notion. Hence the Putnam-Leibniz point applies. What it in fact would be for things to be as per the thought which, by hypothesis, I just mentioned need not be what it would now (with right) be taken to be. If one can define lager into existence, such is not to define into existence something which fits that definition. That much is still liable to require proof. There is a deeper reason, in the sort of thing a res cogitans (a thinker like us) is, why our concepts should be like that. We will come to this later.

Their Work and Why They Do It  223 At this point a certain possibility (though so far no more than that) opens up. So far, wherever someone speaks of some stuff as being lager, there is a gap between what would thereby be to be supposed as to what, or how, he is saying that stuff to be and what in fact he said. It is at least the gap between what is now to be supposed, occasion-independently, as to what it is for a stuff to be lager (or simply what lager is) and what is in fact so of this. It is conceivable, e.g. that that particular strain of ‘yeast’ which makes lager lager is not really a true yeast, or that a conspiracy of brewers has all along been foisting off on us as lager some artificial chemical substance, not actually brewed at all. But, for all said so far, the gap might be at most that. It would be if, but only if, it were the same gap for any case of someone speaking of stuff as being lager, using the word ‘lager’ for what it is for. Having recognised the gap, though, we see that such need not be so. What might make it not so, for all said so far, is that what words mean is, or is liable to be, only one factor among many on which what is said in given words (in a mouth) may depend. There is also, notably, the work done by that packet of notions, responsibility, commitment, justice, fairness, and their kin. Suppose, for example, that that stuff in Sid’s glass (above) were so adulterated as not to be eligible for the title ‘lager’ according to EU regulations, but which will, post-Brexit, once again be so eligible (within the British Isles) – part of the hard rewon British freedom. Taking a sideways-on view (sideways-on to the UK, that is), is the stuff in Sid’s glass as Pia said it was in speaking to Zoë above? Thinking of the point of Pia’s thought-expression (helping Zoë see how to distribute the drinks), and of for what Pia should therefore be held responsible one might well be inclined to say “Yes”. Now imagine an Australian beer connoisseur in a pub in the UK with his English host, pointing to a tap and asking the host (of the stuff there dispensed), “Is this lager?” (the stuff being that adulterated stuff above). One might feel some justice in the response, “Well, some people here call it that. But actually, no”. Such an account might, on occasion, squarely hit the mark. Taking all this as correct au pied de la lettre, the gap sketched above between what would be to be supposed as such as to what lager is and what is in fact so of this is not to be filled in the same way for every case of the words ‘is lager’ having been used to speak of what they do. Which would realise the above possibility. So far, though, it remains here but a possibility. 6. Understanding Understandings: For words to mean what they do is for there to be something they are for in the general business of ­representing-as (of thought-expression). In particular, for a predicate of, say, English to mean what it does is, inter alia, for there to be something it is for speaking of, thus which it may be said to speak of (in the suitable aspect of ‘speak’). It speaks of some way for an object, or n-tuple thereof, to be. In the business of achieving thought-expression, what words are for is what they are recognisably, to be understood to be, for. So such

224  Charles Travis words as ‘is pure cotton’, ‘takes tea at three’, ‘brewed with hops and malt’, are for speaking of being pure cotton, taking tea at three, being brewed with hops and malt. To know what words mean is (at least) to understand them, so to speak. So to understand them is, as a rule, to understand them punkt. The verb ‘understand’ has a success use, and a neutral one. On its success use it takes a simple direct object. One may understand what Pia said, or Kant. Or one may understand Pia, which on one use, just is to understand what she said. On the neutral understanding, one might understand something in a certain way; as, or to, such-and-such. One might, e.g. understand Pia as saying, or having said, such-and-such. One  might understand a given law as applying in the case at hand. Or one might understand given words (in a mouth, or perhaps of a language) as saying such-and-such. Whether so understanding the relevant item is understanding it is, so far, grammatically, left open. To understand someone’s words as one did is, plausibly, to understand them in a particular way. Sometimes the question, ‘In what way?’ has an answer (thus suggesting, rightly or not, that ‘way’ here is a countable). For example, Sid complains, “They starched my shirts”; Pia takes him to have said that their laundry have starched the shirt’s collars (whereas what in fact bothered Sid is that they didn’t stop there). Here two understandings those words might bear. “Starch the shirts”, Pia told the laundry; they came back rigid throughout. One who understood might see that not to be what was asked for. Here understanding extends to a proper sense for what might count, what not, as things being as spoken of (in this case as things were to be). Often, then, thinkers’ understandings of given words are countable just insofar as senses for when things would count as being as thus represented are (however far that is). So far our talk is of words in a mouth. What of words of a language? Take the English verb ‘to rif’. Here ‘know’ seems more le mot juste: one may know, or not, what this means. And one might count as knowing this if he knows that it speaks of: to let go (fire, no offence mind you) through a reduction in force. A reader who just learned this might anyway have known what ‘rif’ speaks of: (one thing called) riffing (whatever that is). Which may well fail to count as knowing what the word speaks of, thus nor what it means. Such a reader might by now well know this. But then only if he now has a an ‘n’. (‘an’ adequate capacity to recognise when it was this which was spoken of on an occasion in speaking of someone having, or having been, riffed. And here subtleties arise, e.g. as to le mot juste. Let us compare two Dutch words, ‘weggebonjourd’ and ‘afgepoeierd’. To be weggebonjourd is to be let go, to be let know that one’s presence is no longer required (nor remunerated). A touch of irony sits in the word. So, while to be ­weggebonjourd need not be to be riffed (taking ‘rif’ au pied de la lettre) – one might be singled out for special treatment – so, too, perhaps, to be

Their Work and Why They Do It  225 riffed need not be to be weggebonjourd. The word ‘afgepoeierd’, given its etymology, raises slightly different issues: it refers to a practice of brushing powder, or other substances, from the shoulder of one’s garment. Thus a suggestion of ‘the brush off’. Which means that it has a wider application than ‘weggebonjourd’, but also, since not every case of being weggebonjourd need be being given the brush off, in some respect a more restricted one. Now the question: how sensitive to such issues must a capacity be in order for its possessor thereby to understand given words, or to know what they mean, or what they speak of? Presumably such capacity is not reserved for the exquisitely sensitive alone. But then, one would expect, the general answer to such questions would be: ‘Such is negotiable’. Again, consider the words, ‘decide’ and ‘determine’. Sometimes these are useable interchangeably: ‘the outcome of this game decides/­ determines the champion’. But often not. A judge may decide in favour of the plaintiff, accordingly determine (fix) damages at, say, €5,000,000. Zoë may determine that there are six pigs in the pen by counting. But she cannot decide (though she can conclude) that there are six pigs. Deciding how many pigs there are is not her prerogative. Such are the subtleties a fluent speaker of a language would be expected to grasp (that is, to recognise when called for) in understanding these words, or what they are for. Such capacities, called for in a fluent speaker, are also called for by the Putnam-Leibniz point. As turned out, not so long after Leibniz, there is no heavy yellow metal with whose presence in the world we are in fact familiar (in the expected way). Depending on what can be brought into existence by definition, this might have meant that, as turns out, there is no such thing as gold. (Bad news for certain markets.) One (a cartesian thinker) who understood what the word ‘gold’ (or its correlates in other tongues) was for speaking of, though, would be able to recognise that such is not the case. Part of grasping what this word means is a capacity to recognise that what in fact emerged through the march of science is that gold, in its natural state, is not yellow. Such is just part of grasping what the word ‘gold’ is for. In given circumstances (e.g. giving Benno his gold watch), or at a given stage of history (e.g. pre-1703), there is what one who then counted as able to speak, and understand, talk of gold – e.g. one able to use the English ‘gold’ for what it is for – would then rightly understand as to what such talk was of – what was to be supposed as to when it would be this which was being spoken of. Such is an understanding the word (and talk of the stuff) might bear historically, in then-current circumstances of its use. And then there is, ahistorically, what is true as to what it would be to be that which the word speaks of (or the relevant talk is of). Such is something to be recognised (a success); an understanding ‘gold’ would bear in light of how thingsm are, full stop.

226  Charles Travis And there is this relation between the two sorts of understanding: one who (success use) ­u nderstood ‘gold’ (or talk of gold) historically – who saw (near enough) what was to be supposed at that stage of things as to when it would be that which was spoken of would be, inter alia, one thus equipped ­(modulo the usual perturbations) to recognise what was to be understood, ahistorically, as to when it would, in fact, be gold which was spoken of, given awareness enough of whatever it proved to be that, should coincide with or differ from what was then to be supposed. And there is this relation: just as the object spoken of on an occasion in words, ‘That man at the buffet wolfing …’ is the one (if any) of whom, in those circumstances, it was understood as being supposed that he was at the buffet (illusory as mirrors may have made that) wolfing (illusory as the frantic hand gestures may have made that), what the word ‘gold’ in fact speaks of, used for what it is for in English, is that of which what would have been supposed pre-1703, or by the architects of Fort Knox, was in fact being supposed. Such expresses Putnam’s point. So far we have been speaking of words of a language. Let us now turn to understanding what was said (words in a mouth). Here we begin with a specific historical occasion for understanding, with right or not. At the beach house Pia remarks to Zoë, “Sid drinks Bloody Marys”. How would this to have been understood in those circumstances? Suppose that, hearing the remark, Zoë duly prepares only two gin and tonics, and, after a frantic search for celery, one Bloody Mary. Sid, on receiving the Bloody Mary, is not best pleased. It turns out that Pia only meant to be giving an example of the lengths to which Sid would be willing to go to have something to drink where lager was unavailable. At the beach house (of course) there was lager in abundance. Sid had seen to that. Otherwise, he thought, he might be compelled to drink some rubbish such as, say, a Bloody Mary. The inevitable question: are things as Pia said they were? Since it is an historical understanding which is our entrée to what was said, we can begin with what would, with right, have been supposed of Pia’s words in the circumstances in which she produced them. With sufficient background, her would-be casual remark may have been so to be taken. Without such a background, or with a different one (e.g. Zoë: “I’ll just fix some drinks for us to take on the deck”; Pia: “Sid drinks Bloody Marys”), they might have been rightly understood as Zoë took them. The words used, ‘Sid drinks Bloody Marys’, would not have changed meaning from one occasion to the other. On the contrary: it is part of these different circumstances having the different effects they had that what these words speak of (note the tense) remains constant throughout. Similarly for the words, ‘Please starch the shirts’. Similarly, one would suppose, in general. If circumstances can have this effect, then what was raised in the last section as mere possibility now emerges as the case.

Their Work and Why They Do It  227 Someone who understands words in a given way may be said to have an understanding of them, to wit, his. The determiner here suggests a countable. How seriously may such countables be taken? On this, one might start here: to understand words (success) is to have a capacity to recognise what they are for/how they were used (see above); to understand words in a given way, or as one does (success-neutral), is to have a capacity to recognise what would be so of them so understood. How, then, might one count capacities? It is time to note several distinctions between different notions of a capacity. One is this: a capacity may be conceived as identifiable, individuated, by the particular way in which it effects, or enables, whatever it does. The object of a capacity so thought of would just be whatever (if anything) is thus effected or enabled. ­A nother is this: a capacity is to be identified by what it is a capacity to recognise (e.g. a syntactically well-formed bit of English). Such is taken to be determinate independent of the capacity. How the capacity enables what it does is then left open – perhaps a subject for scientific investigation. This distinction will become relevant in the subsequent sections. For the moment let us restrict attention to this second notion of a capacity. So a capacity to recognise when it would be being gold (or what ‘is gold’ speaks of) which was on the table would be identified by what would thus be to be recognised (whether in given circumstances or überhaupt) as to when it would be this which was on the table. To have such a capacity would be eo ipso to understand relevant talk about gold (insofar as it was that); and vice-versa. So one way to identify the object of such a capacity would be to spell out the relevant understanding – what there would be for the capacity to make recognisable – without any reliance on the notion of the capacity itself. One would, in effect, provide some sort of condition for being that which the word ‘gold’ is for speaking of, or that which talk of gold is of, without incurring circularity by any reference to those, if any, who had the capacity. But, as will emerge in the next section, as a rule, at least, such an approach must fail. (To adumbrate, it is just this which persuaded some, in the roaring 1970s and 1980s, to take up some version of homophony in semantics.) What, then? Here we might try lifting a leaf from the study of s­ yntax (as made explicit by Noam Chomsky (1965 (vide1.4)). As large, and intricately structured, as the body of English speakers may be, we may take it as non-controversial, of a large body of candidate English expressions, that these are, or are not, syntactically well-formed in English. For example, there are the sentences of this essay. (Some may question whether all of these are well-formed. But we know how to take things from there.) As a rule, with exceptions, one need not appeal to anything outside English itself to prove that something is or is not a well-formed English expression. We certainly do not need (as proven by the fact that we do not have) anything like a decision procedure for doing this. There

228  Charles Travis is, anyway, an imperfectly determinate, but indefinitely e­ xpandable, body of thinkers with (near enough) a capacity to recognise these things. And this body, for all its imperfections, is enough for it to remain noncontroversial, most of the time, for most purposes, whether a given string is syntactically well-formed English or not. Here in order of logical priority, it is the capacity and its actual ­possession which comes first, what it is a capacity to recognise is what is thus identifiable as to be recognised. Another insight we (or I) owe to Chomsky is this: this order of priority still leaves us on the indicated side of our first distinction: a capacity identified by how it works, or a capacity identified by what it is a capacity to do. That what a capacity is a capacity to do is identifiable in the way just indicated is not to commit to anything as to by just what it is effected or enabled; nor even to there being anything specifiable by which such may be, or is to be, achieved. The present point can thus be put as follows. A capacity possession of which would be understanding what ‘gold’ is for (or what talk of gold is about) is in the same boat as a capacity to recognise syntactically wellformed English. It does not presuppose some independent account of what it in fact would be for it to be that, what ‘gold’ speaks of, which is on the table. It is enough that there are enough who, non-controversially, are able to engage in such talk, who possess the capacity in question (individuated by what it is a capacity to recognise. Enough, that is, that what would need to be non-controversial when it comes to particular judgements as to what is to be understood as to this is non-controversial, fortunately (since nothing of the sort is in the cards) does not require any procedure for determining answers to the question thus arising. In the above I made a remark on what may move one to ­homophony in semantic theory. Which, in the roaring 1970s, moved some to some form of truth-conditional, or post-Carnapian (utopian socialist) ­semantics. If, as per present brief, language is in the business of Fürwahrhalten only, and not that of Wahrsein – simply an aid in getting into that second business – then this second goal cannot be reached. But, it is worth remarking, the quest for homophony need not reach its goal in that way. The ideas here are detachable. Suppose we think of a semantic theory of English as something which pairs with each well-formed English expression a property which an expression would have just in case it meant what that one does. (There may be rough edges, or loose threads, around the notion to mean what … does. But such is orthogonal to present matters.) What might that property be? If the present brief is right, nothing like having a given truth- or satisfaction-condition. But then, as in the above, other possibilities lie before our eyes. The word ‘gold’, for example, speaks of a certain stuff, to wit, gold. The words ‘is gold’ speak of some object as being (consisting of) gold. Similarly, the words, ‘takes tea at three’ speak of (something) as taking tea at three. If we watch our aspects, we might say: ‘is gold’ says something to be gold. Such is

Their Work and Why They Do It  229 ­ omophony all the homophony one needs. And it respects the idea that h is ­mandated if, as suggested above, in present matters the object of a relevant capacity is not specifiable (nontrivially) by any condition on being what this is a capacity to recognise, independent of some reference to possessors of the capacity. As res cogitans, language users, we occupy the epistemic position Chomsky describes. We dispose of a wealth of non-controversial facts about the words we speak. One might well take this as a condition on speaking of ways for things to be at all. An idea next to unfold. 7. The Paradox of the Intrinsic: This ‘paradox’ concerns two features of Frege’s notion of a thought (that countable, der Gedanke). The first begins with the idea that a thought is to be just that by which truth can come into question at all. A given thought is thus identified by no more nor less than what distinguishes its way of making truth turn on how things are from any other’s; thus by when there would be a case of thingsm being as it represents this. So, it seems, it is intrinsic to a thought to be true of just those particular cases which it is true of. Or, conversely, wherever there is something the thought would be true of, that this is such a something is intrinsic to that thought being the thought it is. That it is intrinsic to a thought to be true of just what it is, built into the notion of the objectivity of truth (at least as Frege conceives this). Objectivity requires that truth be an exclusively two-party enterprise: party of the first part (by definition, a thought) puts a demand on party of the second – so to speak, proposes. Party of the second (the way things are), just in being (as it is) disposes: satisfies the demand of party of the first, or does not. Where there is truth or falsehood, objectivity excludes third parties. This demand distributes through a decomposition of a thought to any proper predicative part of the decomposition. If the thought is, say, that Sid is dishevelled, decomposed into a proper part which makes being dishevelled what relevant items must be for there to be truth, then, by the same token as with the whole thought, it is intrinsic to that proper part to predicate truly, casu quo falsely, of just what it in fact does. And what is intrinsic here is located in, provided by, what is intrinsic to that way for an item to be: for it to be dishevelled. So we may see the point here as about concepts (of ways for an item to be). Now the second feature. This is built into what is, to logic’s eye (and Frege’s) the whole point of a concept: put one way, to provide a distinction between a relation an object may bear to a way for it to be, and one way for a thing to be may bear to another. In terms of concepts, an object may fall under, or satisfy, a concept; a concept may be subordinate to another (whatever falls under it falls under the other). For this, as Frege notes, we need something which remains aloof from the objects it relates to in a way that a proper name in Frege’s sense (what identifies an object for truth to turn on) is not aloof from what it names. Without Sid,

230  Charles Travis no vocable, ‘Sid’, could name him: there would not be that route to being a name at all. By contrast, at least in general, it is not as such intrinsic to a concept that those very cases of a thing being such as to fall under it are the cases that they are. The concept merely generalises in its proprietary way, capturing and rejecting accordingly just those cases of an object’s being as it is which there in fact are. Thus, the concept of being dishevelled might still have been just the concept it is, distinguished by just that which distinguishes it from other concepts, had Sid not been just as he is (and hence not fallen under it); for that matter had Sid not been, full stop. So, it seems, it cannot be intrinsic to the concept that Sid falls under it. Nor that his being as he is be a case of something falling under it (since, for all it matters to the concept, there need not have been that case at all). Something intrinsic to the concept dishevelled must thus mandate, require, of something which might not have been while all intrinsic to the concept remained just what it is. There need not have been just that – the way Sid in fact is – to be a case or not of dishevelment. But given that there is just that, it must be intrinsic to an object being that way that this is a case of dishevelment. The question is how could this be? What intrinsic to the concept could make it so? Such is the paradox of the intrinsic. All of which is not to say that there would have been a given concept no matter how things were. Without the Camellia sinensis, there would be no such thing as drinking tea, hence no concept thereof. Thus there is such a concept only by grace of evolution. But since there is, we may point to what Pia is now doing and ask whether this is a case of an object falling under it. On the one hand, given what she is doing, there is nothing to decide an answer but the concept in question. Given the objectivity of truth, it seems, such must be so. So its being so must be intrinsic to the concept. But had Zoë not cancelled lunch at A Belota, Pia might now be sipping a glass of her preferred Billecart-Salmon. There would then have been no such case as above to instance the concept to be drinking tea or not. So it cannot be intrinsic to the concept that this case so instances. A misunderstanding, one might think. It is not intrinsic to the concept to be drinking tea that Pia should be doing just what she now is; for there to be just that case to be or not one of someone drinking tea. But, the idea would be, given that there is that case, it is intrinsic to the concept that this be, casu quo not, one of someone drinking tea. For, given the case, there is nothing but the concept to decide this. But just what about a concept might do this? With which question the plot thickens. Suppose one were to say: the concept to be drinking tea is governed by a law/rule/principle of application (a statable condition), which determines when something would be a case of this. The principle might be homophonous: there is a case of this just where it is a case of someone drinking tea. Whatever its other merits or demerits, such might at least

Their Work and Why They Do It  231 pass for true. But what help is it? Suppose we now ask of a particular case whether it satisfies this condition. It does so just in case it is one of something falling under the concept to be drinking tea. We have thus made no progress in answering the question we were trying to answer. The question was how it could be intrinsic to the concept to be drinking tea that Pia, as she now is, is a case of someone so doing. The present answer: it could be if it were intrinsic to our homophonous condition to be satisfied by Pia as she now is. In plain Scots English, ‘Thanks, pal.’ Suppose, then, that the condition is non-homophonous, e.g. that the candidate object should be imbibing an infusion of leaves of the camellia sinensis in hot water. Which is to say: just in case the object falls under this further concept, for one to be imbibing an infusion, etc. So if we knew how it could be intrinsic to a concept überhaupt for a case to be one of something falling under it, where that case need not have existed, perhaps we might know how such could be intrinsic to this further concept. But if we knew that much, we need not have asked our question in the first place. The problem precisely is how such could be intrinsic to a concept überhaupt. And this is not to be solved along any route like the one now scouted. It may now look as if, to steal a phrase of Frege’s from a different context, we are looking for something to make the impossible possible; are on a snipe hunt. But after all we do engage in representing-as, often enough truly or falsely – e.g. as to who is drinking tea/lager/­BillecartSalmon. So there must be a mistake somewhere in whatever makes this look mysterious or impossible. Before identifying the mistake, though, let us consider the problem in another form. For this we need a notion which I will call Freedom of Thought. This phenomenon arises in the above-mentioned aloofness from its extension which is a concept’s hallmark, that is, in the first element in our two-part ‘paradox’. An example. Suppose you know that there is such a thing as for someone to be ‘waddling’, but have no idea what waddling is. Sid, as it happens, has been letting himself go a bit lately. As it also happens, he now approaches us, waddling through the Jardim dos Ingleses. I point and say: “there’s a case of someone waddling (if there ever was)”. You now know one fact about the concept to be waddling. Suppose this is all you know. Now here comes Pia gliding through the Jardim dos Ingleses, thoughts of bubbles dancing in her head, about to be replaced by the Dinge an sich. Do we have here another case of someone waddling? For all you know so far the answer might as well be yes as no. For there are ways of generalising over and beyond Sid’s case on which Pia’s would be another case of the same (such is one conceivable understanding of same), and there are ways of generalising which capture Sid’s case but reject Pia’s (as we actually know, generalising over waddlers being one such). So far, on the information given, and only on that, you have no right to an opinion as to which sort of way of generalising to be waddling involves.

232  Charles Travis Here the move you are not in a position to make is from one case to two. But the point holds for any move from n cases to n + 1 (an inductive base). Suppose you are given a thousand cases of someone waddling, but this is absolutely all you know as to what being a waddler is; you have no right to suppose any more than this. Now the 1001th candidate case comes along (as may be through the Jardim dos Ingleses). Is it a case of someone waddling? Well, there are generalisations which capture those first thousand and also this 1001th, and there are generalisations which capture the first thousand but reject this. Perhaps, e.g. for all you know, waddling is standing in some very complex relation of gait to times and tides. The operative fact here is one of the aloofness of a concept from its extension. For any set of cases which are ones of something being what the concept is of being, for it to be that concept is not eo ipso for it so to generalise. Something else is needed to identify the concept as the one it is; something which one do what one might have hoped a ‘law of application’, a statable condition, might do, but which we have seen it cannot do for the present purpose. Happily, we already have means to disarm the paradox. The key notion is that of a capacity, defined by what it is a capacity to do (here to recognise). We, or many of us (some more or less loosely defined, but extendible, body of thinkers) have such a capacity in re waddling. We are able cogently to discuss waddling – e.g. know how to use the verb ‘waddle’ to speak of that which it is for speaking of. We know, can recognise, inter alia, when to count something, and when not, as a case of (someone) waddling; what would be rightly so counted (when); so, too, what might bear for or against so counting a case. For there to be such a corpus of thinkers so abled is, inter alia, for there to be, for given such thinkers on given occasions, facts as to what does so count which are non-controversial, indubitable. Now, on the one hand, that such thinkers as we thus are would be prepared to recognise what Sid is now doing, making his way through the Jardim dos Ingleses, as waddling does not even begin to explain, or identify, what it is which makes what he is doing a case of this. To identify this would be something such as to give details of his gait. But on the other, given that we are so abled, the object of the capacity, when there would be a case of an item waddling, may be made identifiable, at least in any given part, by what those with the capacity would thus be prepared to acknowledge (the usual idealisations applying). If the paradox is thus disarmable, there is an important moral to be drawn. This moral lies in what it is to be a thinker of the sort we are. On the one hand, to be such a thinker is to be human (apologies to any Martians in the audience). On the other, it is to be a rational being, a res cogitans in a sense spelled out by Descartes in 1637 (vide especially pp. 90–92), and more recently by such philosophers as Hilary Putnam (e.g. 1988), John McDowell (e.g. 1996) and Matthew Boyle (e.g. 2016).

Their Work and Why They Do It  233 Part of this package, Descartes tells us, is that we are discursive beings, never in principle uitgepraat – bereft of more to say, coherently however banally. The other more important part, on his account, is the ability in principle to step back and subject whatever we are doing, or planning, however we are then guiding our career, to open-minded critical ­scrutiny; to pose non-rhetorical questions whether this is to be done, or to be done as we have been attempting it. As Putnam puts the point, reason can transcend whatever it can survey. When it comes to cases, such is just what is illustrated in the Putnam point. There is, for example, the way we are prepared to treat being gold, or being a given path in space, at a given stage of history. But we are open and sensitive to reasons, should any emerge, why this is not the way such should be done. We can see things to have bearing on such questions. And though, to repeat, what makes something a case of waddling (or gold, or a straight line) is not our being prepared to acknowledge it as such, the fact that we are res cogitans (as I will put it, Cartesian thinkers) does have implications for what the objects of our capacities, of the sort now being discussed, are likely to be. For though in exercising these capacities on any given occasion we may rely on capacities which may well be shared by non-rational creatures, the fact that we can step back from this cogently, as we can, suggests a different shape for their objects. For example, we may share with a sheep dog the capacity to recognise a sheep at sight: there is that distinctive ovine look. But unlike the sheep dog we can also recognise that that sheep are so identifiable, even that animals in general are identifiable in such ways, depends on circumstances which are hospitable to this, so that it is at least conceivable that there should be circumstances in which they are not. We are then prepared to recognise the facts, should there be such, which would show our ways with sheep (or the just-mentioned ones) not the way to go about things. A sheep dog might or might not be impressed by (scented) silicon ‘sheep’. It is for a Cartesian thinker to see whether these are really, or merely ersatz, sheep. In circumstances in which one can, and we do, recognise a sheep by its ovine look, there is exercise of a capacity posing a tractable explanatory problem for cognitive psychology. What abstract visual features characterise this ovine look to which we are thus sensitive? By contrast, it is the very nature of Cartesian thought that our capacity to recognise all we are prepared to recognise, no matter in, or of, what circumstances such recognition is called for, should be expected to be theory-resistant; insusceptible to explanation on the model on which an ability to recognise sheep by looks (in circumstances in which this might be done) may be susceptible to such. Cartesian thought as a whole, one should expect, is theory-resistant. The theory-resistance of Cartesian thought is reflected, one would ­expect, in our actual capacities to recognise what would count, what not,

234  Charles Travis as waddling, or lager, or a gold watch. How might Cartesian ­thinking relate to Freedom of Thought? Suppose that we, or some body of Cartesian thinkers, have something to discuss: some given phenomenon, a way for an item to be. Call this way ‘being F’. Such a body of thinkers thus share the capacity to recognise when it is being F that is on the table, so when a case of a thing being as it is would, or might count as a case/not a case of being F. Now suppose that a question arises, on an occasion, or in given circumstances, as to whether such-and-such is, or does count, as a case of an item being F. Suppose that, in the circumstances, there is some corpus of non-controversial cases of what would/would not so count. These cases, as such, do not identify what being F would be. For, by Freedom of Thought, there are ways of generalising over cases which capture all of these and also the case now in question, and ways which capture all of these but reject this case. Thus far it is undetermined what our body of Cartesian thinkers ought to find. We might, then, look to networks, or sets, of rational relations between being F and other ways for an item to be – relations of entailment, or at least of contributing to, bearing on, what cases would count, what not, as being F. Such might help on occasion. But any such (supposed) rational relation is itself open in principle to rational scrutiny by a Cartesian thinker. So far, the possibility remains that, as may happen with the relation between what it would be for something to be a sheep and a certain ovine look, this is an occasion for scrutiny of this supposed relation between being F and being something else. What remains for us to look at is what the relevant body of thinkers would, or is prepared to find, or to acknowledge. If that body is us, we may ourselves pose the question what it would be right to say in the case at hand. What we are looking at this body of thinkers for is not grounds for finding one way or the other, but rather merely to identify what the object of the relevant capacity is. And now, Cartesian thinking being what it is, there is no reason why the answer we thus discover is: one might say yes, and one might say no. That is, one could count this as a case of something being F, one could count it as one of not; nothing in that two-part enterprise which is truth per se requires the one finding or the other. In which case what one might also find on looking is that whether the right answer is ‘yes’ or ‘no’ depends on the circumstances in or of which the question is being posed. Such sort of sensitivity to circumstance is just what one would expect of a Cartesian thinker – one equipped to recognise when it is time to stop thinking one can identify a sheep by its looks. Consider a case of Wittgenstein’s (1953: §99). “I locked the man up in the room. Only one door is left open.” Was the man then not locked up? Suppose that one door opened up onto a sheer 50 meter drop onto some rocks. Was the man locked up or not? Suppose the man was suicidal, or a notorious secret agent, or a human fly? How would a Cartesian

Their Work and Why They Do It  235 thinker – how would we approach such issues? What sort of answer might one thus expect? With the idea of a Cartesian thinker, a res cogitans, now in place, what was floated as a mere possibility two sections back – in the case at hand, that one might count the man as locked up, or as not, all depending on what one counted as being locked up; neither answer ruled out as such by what it is, punkt, for someone to be locked up; or that here, for these purposes, such would count as being locked up, there, for those, it would not – now begins to look just what one would expect. The ‘paradox’ of the intrinsic, insofar as it is a paradox, appears as such only when we think of a concept as, of necessity, relating to its instances (to cases of something being what the concept is a concept of) via some sort of specifiable rule of, or condition on, its application to particular cases – a Gesetz der Anwendung. Such a rule or condition is, or has the content of, a thought; a content by which things are presented as falling under further concepts (or, in the homophonous case, the same one all over again). With which we embark on a regress. The idea that the application of a concept can be identifiable as the object of a capacity to recognise the concept’s applications – without prejudice to what it would be for there to be a case of something falling under it – together with the idea of a Cartesian thinker, offer one way to stop the regress. The price for this, though, is that what is now to be recognised as to what a concept’s applications would be must match up with the shape of Cartesian thought; if not full stop, then at least insofar as concepts are what a Cartesian thinker (such as us, for example) might mention, or present things as falling under. The next section begins to unfold the price thus exacted. 8. Occasions: One can, as a rule, tell a pig by looking, recognise one at sight. There is that certain peculiarly porcine look. Insofar as this is so, and, as may well be, theory-receptive – insofar as the look is suitably identifiable – one might, in principle, build a bit of cognitive protheses: an electronic pig-detection device (PDD). Suitably exposed to the look – good enough lighting, angle, a medium free enough of disturbing factors – the device would register ‘pig’, and mutatis mutandis for the look’s absence. It would thus alert us to cases of porcine presence. The PDD, of course, would rely on an environment receptive to pig detection by these means. And since, as we Cartesian thinkers know, to be a pig is not eo ipso to have this look, nor vice-versa, it is certainly conceivable that we should stray into, or our environment should mutate into, an environment inhospitable to this. Suppose one were to ask how, in general, one could detect inhospitable environments (for the PDD), and then, for any such environment, how one could come to see by what in it pigs were distinguishable from other things. The theory, or design, of the PDD (wisely) leaves such questions open, unaddressed. To speak in other terms, while there might well be a theory of what it is, things being as they are, which makes pigs identifiable by their looks, there is no

236  Charles Travis reason to think (and considerable reason not to) that an entirely general account of what a Cartesian thinker would be prepared, and equipped, to recognise, in any circumstances, as to what identifies a pig as such (much less a case of something being a pig as such) is in the cards; that such a general phenomenon is theory-receptive. Someone might complain that two different things have been ­conflated here. There is, on the one hand, the question how one can tell when it is a pig he has to deal with (to which an answer might be ‘by the snout’). And, on the other, there is the question how one can recognise what would count as an item being a pig; as what would so count. As said already, no one (I suppose) thinks that to be a pig just is to look like one. From which we see that such pairs as these are, or are liable to be, quite different questions. But the general shape of Cartesian thinking, at least as appealed to so far, shows up in either arena. Probably most of us nowadays believe, however vaguely, that what it is for something to be a pig is a matter of something called ‘the pig (Sus domesticus) genome’, whatever that is. I do not doubt any of this, and leave it to zoology to fill in the details of that ‘matter of’. Nor do I doubt that this does settle something important for us as to what it would be for something’s being as it is to be a case of something being a pig; that science now tells us what it is to be a pig – within the usual limits of science’s ambitions in conceptual matters. (As Putnam pointed out, chemistry, in telling us what water is, does not tell us whether that stuff in the cup is to count as (slightly adulterated) water, or rather as a mixture, tea. Similarly, zoology does not, so far as I know, undertake to settle whether what is hanging on that meat hook is a (dead) pig, or rather just a largish chunk of meat.) Still, though, in showing us what it does show us as to what it would ­ artesian be for something to be a pig, zoology relies on the capacity of a C thinker to recognise, in the light of the facts, when what a zoologist tells us would be about pigs, in other terms, when what it is telling us is the truth as to that of which we had supposed what we had all along supposed as to when it would be a pig (or what, in literal sense, ‘pig’ speaks of) that was at issue. And what zoology cannot exclude in principle (though as Cartesian thinkers we might be able to see it as not so) is that there is something of which it is now supposing what it does of which what it thus supposes is not so. Theory-resistance sets in when it comes to a general explanation of how a Cartesian thinker might be able to see this to be so if it were, or not so if it were not. With any novel candidate a thinker may encounter for what would/ might count as a pig (or a Mimosa, a starched shirt, or waddling) comes the means, if perhaps not the occasion, for posing a novel sort of question: whether this is what would count as a pig (Mimosa, or etc.). However, and by whatever, such questions may have been answered in already decided cases, Freedom of Thought, or rather that plus Cartesian

Their Work and Why They Do It  237 thinking, guarantees at least two candidates for what being a pig (casu quo a Mimosa) might be. There is a way for an item to be which agrees with to be a pig (or whatever) on all decided cases (at least as to these being cases or not of so being) and which, moreover, is also instanced by the novel case, and a way for an item to be which, again, agrees on all decided cases, but is not instanced by the novel case. Which sort of way for an item to be is to be a pig (casu quo Mimosa)? Such a question need not always arise. It is not as if one must always pose and address it, at least explicitly, before he succeeds in recognising – it to be a pig (or Mimosa, etc.) which he has to deal with. But there can be occasion for addressing it. And if, on occasion, one need not bother, it also belongs to the capacity of a Cartesian thinker to recognise such things; when one’s standing ways with such matters may stand. In the quotidian, then, Cartesian thinking may be passive, or at least tacit. But such questions as above do sometimes need addressing, as, e.g. in Putnam’s cases. Though our capacity to address such issues may be theory-resistant, there is something to be said about the form of problem we would thus confront. It is a form which first came into sight in this essay with Pia remarking on Sid at work panning Benno’s book. There is what one would have supposed in the circumstances as to what would have been at issue, what the topic would have been, in talk, or thought, of something being a pig (Mimosa, etc.) The general form of the question arising is: How is what one would have been supposed to be speaking of in speaking of pigs (or of something as a pig) to lead us, in the circumstances which in fact obtain, to that of which one in fact so speaks? What is shown by what would have been to be supposed as to what one is in fact committed to in speaking of something as a pig? Answering such a question calls on those same capacities peculiar to a Cartesian thinker which allow us to see when, and why, to stop relying on the PDD to pick out pigs; and then by what else, in the circumstances, such might be accomplished. One might also note that, significantly, the form of question here is the same as the form of question which leads us from what would with right have been supposed as to how a speaker was aiming to achieve thought-expression in given words on an occasion, and as to what he was aiming to say to identifying what thought was in fact expressed – from the fact that Pia was to be understood as supposing a certain man to be identifiable, or then taken to be identifiable, by the (supposed) fact of wolfing foie gras to the fact of her having expressed a thought of Sid that he was panning Benno’s book. All this part of a Cartesian thinker’s ability to stand back from his, or his fellows’ engagement with the world to reflect critically on the efficacy, more generally, the workings of such actual deployment of, inter alia, his/their capacity for thought. There is the question what Pia ought to have been understood to be doing in deploying words as she did, and then the question to what she is thus in fact rightly held to be ­committed.

238  Charles Travis It is, we may suppose, a feature of her deployment that those words were to be understood as speaking of what one who knew English might take them to speak of. So far, perhaps, what would, anyway, then have been supposed as to what they speak of (note the tense), independent of her having spoken them or not. But to state something is to proffer presumed authority as to how things stand in some given matter. It is to vouch for things being some way for them to be, to underwrite treating things accordingly as part of treating them as being as they are. It is to make a commitment, to assume responsibility. Still assuming Pia to have used the words she did for what they are for, as speaking of what they do, still, what this would rightly have been supposed to be is clearly just one factor among others leading from her speaking as she did to the commitment, the pretence to authority, she is thus reasonably, fairly, taken to have made. The words, ‘is tea’, ‘is a pig’, ‘is starched’, ‘is waddling’, are each for speaking of some (presumed) way there is for a thing to be. For each there is what the way in question is to be, or might reasonably be, taken to be. There is then a path from such things to what it in fact would be (1) for a thing to be the way in question, (2) for a way for a thing to be to be that one. Such a path may lead to indefinitely many different termini depending on the circumstances in which it leads anywhere at all. To say this is just to say that the ways in question here for a thing to be are the objects of a capacity peculiar to a Cartesian thinker (a res cogitans). (By which, to repeat, the paradox of the intrinsic is disarmed.) A Cartesian thinker is precisely one sensitive (in principle at least) to indefinitely many ways in which the vagaries of circumstance may matter to an outcome in matters such as these; thus to recognise of Pia’s use of the words, ‘The shirts are not starched’, used for what they are for in speaking English, that, for all that, what she is rightly held committed to may vary with the authority-as-to which, in the circumstances, she might reasonably have been taken to claim. Recognise, that is, if such is so. But when one is speaking to Cartesian thinkers (something we always do, monologues directed at pets aside), the just-mentioned capacity is one available to be exploited. In which case there is no reason why one should not so arrange things, or just be so placed, as to be correctly taken to be exploiting it. Which is all there need be to make language designed for Cartesian thinkers – thus for us – occasion-sensitive. 9. The Laws of Truth: The main conclusion has now been reached. Our initial question was what work words do, that is, what role ­language plays in the business of thought-expression – of achieving entry into a thought’s (a thinkable’s) quite different business, that of being true or false. So it would be understandable if one regarded the conclusion we have reached as merely about language. The idea would be: perhaps what words speak of – what some expression speaks of in its language – is

Their Work and Why They Do It  239 such that in speaking of this, one (a thinker) may, on occasion, speak of any of many distinguishable ways for a thing to be (each varying from the other in when something would so be); but where those words have been used to speak of some given such way – to say, say, what is either true or false – one might not in speaking of that way (on one occasion or another) speak of any of many ways for a thing to be, each varying in when something would so be. The idea in other terms: what (predicative) words speak of in a language may, perhaps, be such that what they would speak of in a mouth varies with the mouth and the occasion; but the way a thought represents a thing as being cannot be such that in representing a thing as that way a thought (or a thinker) might represent any of various things as so of it, depending on the thought (or the thinker and an occasion). True (and importantly), the present topic is thought-expression, and not the phenomenon as such of being true; in Frege’s idiom, one half of the phenomenon of Fürwahrhalten (as one might say, holding-forth as true); not the phenomenon of Wahrsein. But the core point permeates the phenomenon of thought-expression too thoroughly to be held within the just-envisaged bounds. It is a general point about the idea of a way for a thing (or n-tuple, or simply thingsm) to be. Consider the English words, ‘is lager’. Mention what way for stuff to be they speak of, and however you do this – perhaps in using the words, ‘of being lager’ – you will have mentioned something to which the core point applies: in speaking of that on an occasion, a thought-expresser may represent something as any of many different ways. Suppose, now, that Sid does use those words, say, in complaining to the barman, “This (pointing at the contents of his glass) is not lager” (thus using the words ‘is lager’ to speak of what they anyway do). He thus expresses a certain thought; represents the stuff in that glass as not a certain way there is for stuff to be. Suppose the way he thus speaks of is just one of the many ways one might say something to be in speaking of what the English ‘is ­lager’ does. For example, perhaps it is such as for such things as ‘double bock’ not to count as so being. One might, anyway, mention that way, e.g. as follows: ‘being lager on the understanding of so being on which Sid spoke of this’. Now the core point applies to what was mentioned: there is what that mentioning would be to be understood to speak of, and then there is what one might speak of in speaking of something of which all that was to be understood. In Cartesian thought, the first of these items is always susceptible to ­underdetermining what the second would be. To each philosopher, perhaps, his own use of the word ‘concept’. But let us use the word here as follows: wherever there is such a thing as something being A, there is, correspondingly, the concept of being A; a concept is intrinsically, essentially, of being just that which it in fact is (so that, inter alia, if there were no such thing as being that, there would

240  Charles Travis be no such concept). In this last clause we have departed from Frege’s use of the word ‘Begriff’ from 1891 on. But not far enough to avoid a major conflict. For Frege repeatedly insists, We have the requirement here for concepts that for each object it be determinate whether it falls under the concept or not … without whose fulfilment it would be impossible to set up logical laws for them. (1891: 135) A definition of a concept (a possible predicate) must be complete. It must determine univocally, for each object, whether it falls under the concept (whether the predicate can be spoken of it truly) or not. There must thus be no object for which the definition leaves it questionable whether this falls under the concept, even if for us, with our defective knowledge, it may not always be possible to decide the question. … The law of excluded third is really only another form of demanding that a concept be sharply bounded. An arbitrary ­object Δ either falls under the concept Φ, or it does not fall under this, ­tertium non datur. (1903: 69 (§56)) So, it seems, whatever might be so of two concepts which nonetheless differed in what would fall under them cannot identify any one concept at all. Frege read this as a law of truth. And it is not uncommon so to read some formula generated by a f­ ormal ­ oldfarb logic and assigned in it the status logical truth (Vide, e.g. G 2010: 71). But suppose we ask how this last fact should come to be. To start with a particularly clear example, consider the ­Begriffsschrift of 1879. Commenting on it in 1882, Frege wrote, If it is a matter of relating two judgeable contents, A and B, one aims to capture the following cases: A and B A and not B not A and B not A and not B. These cases can be either affirmed or denied. (1882b: 34) A constant thus introduced is defined only for cases where all that it relates is truth-valued. Try to substitute something else for the ‘A’ or ‘B’ in a formula governed by such a constant, ‘AconstB’, and you simply get an undefined, uninterpretable, expression, not a well-formed expression

Their Work and Why They Do It  241 of the calculus. Let ‘vel’ be a constant by definition true in all cases but the last (where false), and ‘neg’ a one-place constant defined as reversing truth-value. Unsurprisingly, ‘Avel(neg-A)’ is a theorem. It is an artefact of the way the constants were introduced. By the early 1890s, Frege had abandoned his 1879 conception of a constant. But the general point survives the change. The change itself is signalled here: [W]hat is at issue in logic is not how thoughts issue from thoughts without reference to their truth value, … the step from thought to truth value, … more generally, the step from Sinn to Bedeutung, must be taken. … [L]ogical laws are first of all laws in the realm of Bedeutung, and relate to Sinn only mediately. (1892–1895: 133) What logic immediately treats of is thus, for example, a certain ­function, V, from ordered pairs of truth-values ({The True, The False}²) into {The True, The False}, to wit, that function which maps all such pairs into The True except for the pair , which it maps into The False. Mediately, logic thus tells us something about triples of thoughts, G, H, J, which relate to each other thus: for G to be true (for it to represent things as they are) is eo ipso for V(the value of H, the value of J) to be The True. Let N be that function from {The True, The False} into itself which flips truth-values. Then for any thought G, V(the value of G), the value of N(the value of G) is The True. This last can be seen as a form of tertium non datur. If the logic contains a constant, v, and a constant, n, which correspond to V and N in the obvious way, then (taking ‘P’ to be a symbol in the logic which ‘indicates (a thought) indefinitely’, ‘Pv(nP)’ will be an expression in that logic of a logical truth. All of this, of course, on the assumption that whatever is indicated by those indefinitely indicating letters, ‘G’, ‘H’, ‘J’, ‘P’, is truth-valued. So again, tertium non datur is a theorem in just that within, or for which, it is assumed to hold. None of the above is to say that so treating logical constants is in any way arbitrary or capricious or anything less than well-motivated. But it is occasion to ask what the motivation is. For Frege, at least, we can find this in the task he assigns logic, to wit: first, to answer the question ‘How must one think to reach the goal truth?’, but then, second, to do so only insofar as an answer is given simply by what being true is as such (Vide, e.g. 1897: 139). Further, One justifies a judgement either by tracing it back to already a­ cknowledged truths or without appeal to other judgements. Only the first case, consequence, is the object of logic. (17 Kernsatze zur Logik, Kernsatz 13)

242  Charles Travis Logic’s central concern is thus the phenomenon of truth-preservation, that is, with when a move from given thoughts to a given thought is such that, provided the first-mentioned are all true, so, too, the last-­ mentioned. All this restricted, that is, to cases where such is so merely by virtue of what being true is as such (by virtue, e.g. of the fact that a question of truth is a yes-no question). Given this, one might ask to what a law of truth would be sensitive; between what the relevant law-like relations might hold. We have already seen one answer to that question: in first instance (by the above) those relations would hold of n-tuples of truth-values. Another route to an answer (at the level of mediate concern) would be to begin with some given thought (decomposed in some given way), and, step by step, prescind from whatever is not logic’s concern. Let the thought be that Sid is enjoying a liquid lunch. Clearly logic has no interest in the particulars of being liquid, or a lunch, or of enjoyment. Nor in who, or what, Sid is. Similarly for any other feature by which this thought is distinguished from any other thought (or, more properly, by which any element in this decomposition is distinguished from any other element which might occupy its place). Prescinding from all this we are left with roughly this: the representing of some object as such-and-such (some way for an object to be). Such is no thought at all. But it is a form for truth, or for falsehood, or for both, to take. And it is by just this feature that it, as such, is, or may be, of interest to logic. It is of interest insofar, and only insofar as, completions of it would be either a truth or a falsehood. In any other case, were there such, questions of truth-preservation (of the sort logic addresses) would not arise. Reason enough to build the sort of assumption scouted above into a logic (that is, a theory of the laws of being true). Reason enough to expect from what is so built an expression of tertium non datur. A law of being true is, eo ipso, a law in the domain of the truth-valued. That tertium non datur is such a law is no surprise. How logic applies to the things we in fact think, or as Cartesian thinkers might think, is another question, not for logic itself to answer. For any way we (Cartesian thinkers) can represent something as ­being, any way we might speak of, and for any candidate case of a thing so being, there are two ways of generalising over cases, each of which agrees elsewhere as much as you like with whatever is fixed there as to what counts as the way in question, what not, on one of which this candidate would be a case of a thing so being, on the other of which it would not. In speaking of some way for a thing (or things) to be, might someone have been so to be understood that for every such choice point, he was to be understood as speaking of what generalised in the one way rather than the other? If he expressed either truth or falsehood, must he have been so doing? Both would be tremendously strong assumptions, neither with much plausibility if not compelled by logic itself. What of logic itself? Just as mechanics’ concern

Their Work and Why They Do It  243 is with given quantities, say force, mass, velocity, and how these might distribute over given sets of ­eligible objects, so logic’s concern is with given values, true and false, and how these might distribute over given sets of items susceptible to them; ways what was truth-valued might relate to what else was. Nothing in this compels one to make that just-mentioned strong, otherwise unmotivated, assumption. It is not a truth of mechanics that a ball of given mass and initial velocity, rolling down a plane of given incline, will reach the bottom. Nor is it, barring stipulation, a truth of logic that a thought would be true (false) no matter what. 10. Theory Resistance: What makes pigs recognisable to us at, or by, sight? How do we achieve such feats of recognition? With proper respect for boundary conditions, answers to these questions might be in the cards. Such would specify certain abstract visual features of pigs. It might turn out (again, with proper respect for boundary conditions) that (nearly enough) wherever those features are present, so, too, a pig; and that our capacities in fact exploit this. Such a theory might remove a (small) mystery. By contrast, suppose someone asked how we are able (where we are) to see (recognise) the thing to do or think: how, in ­general, one is ever able to see such things. If we are Cartesian thinkers, then a general answer to such a question is not in the cards; the contrast between theory-receptiveness and theory-resistance. The question how to tell a pig at sight, or how we in fact do so, is, or may well be, ­theory-receptive. A question how, in general, we are able to see what we can see as to the thing to do is theory-resistant. Note the role in this for boundary conditions. In the world as we find it, there is, by and large, a way to tell by sight when it is a pig one faces. Most of us have mastered this. So far, theory is in the cards. Let us drop the reference to boundary conditions. How could one ever recognise a pig at sight, in any possible conditions? Now theory-receptiveness ­disappears. How might a pig ever be made recognisable to us at sight? Who knows? Similarly, how can one recognise whether it is a good idea to offer Sid another pint? Well, there is that glazed-over look his eyes get when he has had enough. But, with an eye to all the factors which might matter to what the thing then to do would be, a general theory of the matter seems a bit much to expect. The reason for dwelling on such banalities is that some have suggested that to give up on the utopian socialist view of language and its relation to thought-expression would be to plunge into mysticism and deepest darkness. Herman Cappellen and Ernie Lepore, for example, write: The simple idea is this: If RC were true, it would be miraculous if people ever succeeded in communicating across diverse contexts of utterance. (2005: 123)

244  Charles Travis ‘RC’ abbreviates radical contextualism, essentially the denial of utopian socialism, particularly the present form of such. Such is also what Jason Stanley calls ‘the pessimistic view’: On the pessimistic view, there is stability to word meaning and the significance of the syntactic structure of sentences. But in general there is no … systematic way of proceeding from knowledge of the extra-linguistic context and knowledge of the meaning of individual words and modes of combination into sentences to a grasp of the information about the world that is conveyed by an utterance of a sentence, it is mysterious how language users could so smoothly move from linguistic comprehension to action. In short, if the pessimistic view were correct, the connection between speech and action would be inexplicable. (Stanley 2007: 10–11) The present case for occasion-sensitivity rests on these ideas: that the role of language is as an instrument in the business of achieving thought-expression, not in the business of being true (or false) itself; that the thought in question is thought of, by and for Cartesian thinkers as that notion has been set out above, and that the thought expressed in a given case is fixed, first, by how the words used ought then to have been taken, thus what ought to have been supposed as to what they spoke of, and then (the Putnam point) of what this would have been being thus supposed. For Cappellen, Lepore and Stanley (henceforth CLS), what emerges from this framework amounts to mysticism, in place of the science one might have expected. Such is to misunderstand both mysticism and science. On which, here, I will make just two remarks. First, CLS insist that a certain phenomenon of human cognition be theory-receptive, namely, the full, unrestricted, capacity of a fluent human speaker of a human language to grasp, to recognise, what was said in using words of that language for what they are for in speaking that language. But the rule is that a human capacity for recognition of such-and-such is, in its entirety, theory-resistant. Compare, for example, these two questions: 1 How do we, as things stand, generally manage to recognise a pig at sight? 2 How is a human (ever) able to recognise an encountered object as a pig (or as not)? The first phenomenon may well be theory-receptive. The second, if ­Descartes is right, will not be. Who knows, in advance, in what circumstances we might be called on to stand back from our relations with pigs and ask again how, in these, recognition is to be accomplished?

Their Work and Why They Do It  245 In understanding what was said, even if in fact words were used for what they are for in a language, we are sensitive to a host of factors, among which those on which depends whether the words were to be taken as so used. Who knows in advance how, on occasion, we might be called on to step back from our usual way of dealing with these? What a surprise if there were theory-receptiveness here! Second, suppose that a human fluent speaker’s competence to ­recognise what there was to recognise as to what would have been said in use of given words of a given language as for what they were for in speaking it were theory-receptive. On any plausible view of the object of this competence, what would have been accomplished would just be elimination of theory-resistance from one of a battery of competences drawn on (as a rule) in understanding what was said; the remainder of the battery still as theory-resistant as one would expect them to be. That supposed ­mystery CLS regard occasion-sensitivity as making for would still remain unabated. Such is recognised by such a reasonable utopian semanticist as David Kaplan in what he stresses to be the goals of the semantic theory at which he aims. A central idea in such a theory would be this: often, in a natural language, a complete sentence to syntax’s eye is really an open sentence to the eye of the semanticist: it predicates something of some n-tuple with yet-to-be-specified values, e.g. an n-tuple of a speaker, a time and a place. What the theory would accomplish is this: specify values of the n-tuple, e.g. Kaplan, Mulholland Drive, noon, 1 January 2000, and you get what would be said in such an open sentence, used for what it is for, etc., in saying Kaplan, Mulholland Drive and that time to be as the relevant open sentence speaks in its language of such a triple being. A theory which told us that much would be a success by Kaplan’s lights. He explicitly disavows any ambitions in re saying in what circumstances a speaking of those words would have been about him, or Mulholland Drive, or that time (beyond the truism that it would be just where it was so to be understood). But a fluent speaker’s unrestricted capacity to recognise what was said in given speakings of those words would encompass capacities to recognise those things insofar as it is one to recognise when truth, or falsity, was spoken. That capacity remains as theory-resistant as ever. If theory-resistance is per se a mystery, mystery remains. So viewed, there is certainly no more reason to expect our capacities to recognise how things ought to be taken to have been represented in a use of given words to be more theory-receptive than our (unrestricted) capacity to distinguish pigs from other things. In the last century philosophy of language took two peculiar turns. One was to embrace enthusiastically Frege’s rare errors (issuing, for one, in the idea of Begriffsschrift as a language, albeit only of ‘pure thought’), while discarding his most illuminating insights. The other was to ­assign undue authority to a thinker (or what he meant) over what thought was expressed by given words in his mouth. The first error can

246  Charles Travis make it seem innocuous to say with Stanley that what words mean plus syntactic structure must fix truth conditions, or at least conditions on truth of that of which what they say is predicated. The second error can tempt to misguided, inevitably ill-fated, reductionism. It also oversimplifies the object of a capacity to recognise (and grasp) what thought was expressed in a given instance. The corrective to the first error is to take Frege seriously in disclaiming language as his topic, thus placing it on the psychological side of the psychological-logical distinction. It then becomes at least not obvious that truth has the role in words meaning what they do that Stanley (among many) assigns it. With which exploring the nature of Cartesian thinking (that of thinkers of our sort) can uncover good reason not to suppose it to do so. The corrective to the second error is also found in Frege with the idea that ‘true’ simply gestures at assertive force; at the proffering of presumed authority as to how things are. With which we regain roles for responsibility, fairness, justice. With which the object of a (human) capacity to recognise what thought was expressed is reshaped, so that both occasion-sensitivity and that theory-resistance which seems mysticism to CLS become no more than one would expect.

Note 1 Apologies to Donald Barthelme (“Our Work and Why We Do It”, Sixty ­Stories, New York: G. P. Putnam’s Sons, pp. 317–322.) This essay is in memory of Peter Bosch who shared with me, and with Frege, a concern to separate the logical from the psychological. I was started on the main line of thought here by probing questions from two graduate students then at King’s College London, Peter Sutton and Peter Ridley. Thanks, too, to David Zapero for considerable help and inspiration.

Bibliography Boyle, Matthew, 2016. “Additive Theories of Rationality: A Critique”, The ­European Journal of Philosophy, v. 24, n. 3, pp. 527–555. Cappellen, H. and Lepore, E., 2005. Insensitive Semantics, Oxford: Basil Blackwell. Chomsky, Noam, 1965. Aspects of the Theory of Syntax, Cambridge, MA: The MIT Press. Descartes, René, 1637/2000. Discours de la Méthode, Paris: Flammarion. Frege, Gottlob, 1879. Begriffsschrift, Eine der Arithmetischen Abgebildete ­Formelsprache des Reinen Denkens, Halle: Louis Nebert. ———, 1882a. Letter to Marty, 29.8.1882, in Gottlob Freges Briefwechsel, G. Gabriel, F. Kambartel, and C. Thiel, eds., Hamburg: Felix Meiner, 1980. ———, 1882b. “Booles logische Formelsprache und meine Begriffsschrift”, in Nachgelassene Schriften, pp. 53–59. ———, 1891. “Funktion und Begriff”, in Funktion, Begriff, Bedeutung, G. Patzig, ed., Göttingen: Vandenhoeck und Ruprecht, 1986, pp. 17–39.

Their Work and Why They Do It  247 ———, 1892–1895. “Ausführungen Ueber Sinn und Bedeutung”, Nachgelassene Schriften, pp. 128–136. ———, 1893. Grundgesetze der Arithmetik, volume 1, Jena: Hermann Pohle. ———, 1897. “Logik”, in Nachgelassene Schriften (2nd edition), H. Hermes, F. Kambartel, and F. Kaulbach, eds., Hamburg: Felix Meiner, 1983, pp. 137–163. ———, 1903. Grundgesetze der Arithmetik, volume 2, Jena: Hermann Pohle. ———, 1915. “Meine grundlegenden logischen Einsichten”, Nachgelassene Schriften, pp. 271–272. ———, 1918. “Der Gedanke: Eine Logische Untersuchung”, Beiträge zur Philosophie des deutschen Idealismus, part 2, v. 1, pp. 58–77. ———, 1919: “Die Verneinung”, Beiträge zur Philosophie des deutschen ­Idealismus, v.1,1919, pp. 143–157. ———, n.d. “17 Kernsätze zur Logik”, Nachgelassene Schriften, pp. 189–190. Goldfarb, Warren, 2010. “Frege’s Conception of Logic”, in The Cambridge Companion to Frege, New York: Cambridge University Press, pp. 63–85. Kaplan, David, 1989. “Demonstratives”, in Themes from Kaplan, Almog, et al., eds., Oxford: Oxford University Press, pp. 481–564. Leibniz, Gottfried Wilhelm, 1703 (1968). Nouveaux Essais Sur l’Entendement Humain, Paris: Garnier-Flammarion. McDowell, John, 1979. “Virtue and Reason”, in Mind, Value and Reality, Cambridge, MA: Harvard University Press, 1998a, pp. 50–73. ———, 1996. “Two Sorts of Naturalism”, in Mind, Value and Reality, ­Cambridge, MA: Harvard University Press, 1998b, pp. 167–197. Putnam, Hilary, 1975. “It Ain’t Necessarily So”, The Journal of Philosophy, v. LIX, n. 22 (25 October 1962), reprinted in Mathematics, Matter and Method (Philosophical Papers, v. 1), Cambridge: Cambridge University Press, pp. 237–249. ———, 1988. Representation and Reality, Cambridge, MA: The MIT Press. Stanley, Jason, 2007. Language in Context, Oxford: Oxford University Press. Travis, Charles, 2017. “Views of My Fellows Thinking (Counting Thoughts)”, Dialectica, v. 71, n. 3, pp. 337–378. Wittgenstein, Ludwig, 1953. Philosophical Investigations, Oxford: Basil Blackwell.

Contributors

Avner Baz, Tufts University Jocelyn Benoist, Université Paris 1 Panthéon-Sorbonne Krista Lawlor, Stanford University Eduardo Marchesan, Universidade de São Paulo Sofia Miguens, Universidade do Porto François Recanati, Institut Jean-Nicod, École Normale Supérieure Charles Travis, King’s College, University of London Michael Williams, Johns Hopkins University David Zapero, Universität Bonn

Index

ambiguity 171–173 analyticity 166–168 assertion 100, 106, 109–110, 112, 114, 141, 144; and commitment 209–210, 213–215; and authority 210, 238, 245–246 Austin, J. L. 4–10, 12–15, 20–33, 35, 37–42, 44–49, 52–54, 59, 61, 70, 72, 78, 89–90, 97, 101, 117, 124, 142, 148, 187, 190–197, 199, 209 availability principle 105–106, 112–113

independence 11, 62, 82, 198; and knowledge 59, 62–64, 78, 82, 192–199; and natural kind terms 88, 92–97; narrow and wide 27, 109–110. convention 11–12, 102–103, 117–118, 179–180; demonstrative conventions 25–26, 39–40; descriptive conventions 25–26, 39–40 conversational implicature see pragmatic implicature

Benoist, J. 141–142 Burge, T. 124–133, 137–140, 150 capacity 136–137: conceptual capacity 136, 140, 143, 160, 163–164, 169–170, 173, 179–182, 185; linguistic capacities 42; recognitional capacity 143, 163–164, 169

definite description 215–219 Dennett, D. 124–125, 128–131, 133–135, 137 DeRose, K. 189–191, 193 Descartes, R. 59, 62, 65, 77, 82, 207, 232–233, 244 disjunctivism 125–128, 132–136, 138–39, 149

Cappelen, H. 110–111 Carnap, R. 14, 162, 211–212, 228 Chomsky, N. 227–229 communication 1–3, 14–17, 41–42, 48–49, 103, 112, 197; see also communicative intentions concept 239–240; application of a 52, 159–185, 198, 217, 235; conceptual content 42, 44–45, 54, 144; to fall under a 13, 159–185, 229–232, 235, 240; non-conceptual 146–147, 149; possession of a 174–178; pseudo-concept 166–167 conceptual analysis 35–37, 44, 52, 54 consciousness 128–131, 144 content see proposition context; context dependent elements 2–4, 8, 94, 107–109; context

Evans, G. 126, 175–177 extension 88, 93–94, 162, 198 externalism 88, 92, 97, 159–185 fallibility 126, 133–139 felicity-conditions 117, 193, 196 Frege, G. 14, 23, 94, 96, 116, 124–125, 140, 144–149, 184–185, 189–190, 198, 200, 205, 208–213, 221–222, 229, 231, 239–241, 245–246 generality 11–12, 146, 162–163, 169–180 Grice, H. P. 14, 35–38, 42–44, 47–54, 100–106, 108–109, 118, 181 Hume, D. 67–68, 134

252 Index illocution see performative indexicality 3, 27, 51, 90, 93–95, 100, 102–103, 106–109, 181 intension 162 intention 1–3, 14–17, 100–110, 114, 116, 118; communicative intentions 1–3, 14–17, 103–106, 110, 114 invariantism see literalism judgment 125, 128–129, 139–150, 200–201, 219, 228 Kaplan, D. 11, 51, 108, 216–217, 245 knowledge: as certainty 71–75; as circumstance-dependent 63, 75, 78–82; of the external world 59–60, 62, 65, 82; knowledge ascription 46–47, 63; knowledge claims 35, 43–45, 47, 49–51, 63, 70, 194–195 language: formalized languages 1–3, 13–14, 22, 169, 205; natural language 3, 41, 48, 53–54, 191, 245; natural language semantics 38; ordinary language 48, 191; rules of a 1–4, 7–9 Lawlor, K. 70 Leibniz, G. 167, 221–2 Lepore, E. 110–1, 243–4 Lewis, D. 46, 187, 188, 192–4, 197 Literalism 50, 102, 192 Logic 145–6, 149, 209–10, 241–3 McDowell, J. 124–44, 146, 148–51 meaning: as character 11, 51, 108; linguistic meaning 1, 3–4, 9, 14–15, 37, 41–42, 44–45, 54, 107, 109, 149–150, 181; literal meaning 7–9, 90, 100, 103, 105–107; non-natural meaning 102, 105, 108–109, 118; speaker’s meaning 102–109, 114, 118 Merleau-Ponty, M. 142, 200 mind 129, 145, 188; mindedness 129–130, 132, 138, 150; personal/ subpersonal distinction 124–125, 129–131, 133–134, 137–140, 147, 150 modulation 102, 107, 110 Moore, G. E. 70–3, 75 Moyal-Sharrock, D. 68, 71 objectivity 13, 16–17, 101, 229–230 occasion-sensitivity 38, 49, 206–207, 244–246

ordinary language philosophy 4, 17, 22, 35–36; method of 35–54; ostension 92; see also indexicality perception 46, 124–128, 131–135, 138–142, 144–147, 149–150, 200; epistemology of 127, 133, 138; perceptual experience 80, 127, 129–131, 133, 139, 141–144; perceptual psychology 125, 131, 133, 138; and representational content 124–128, 132–133, 135–142, 145–150; science of 125, 127, 131–132, 140, 150 performative 30–31, 190, 194–195; see also felicity-conditions pragmatic implicature 52, 100, 102–6 pragmatics 3, 7, 17, 53, 100–101, 193 Pritchard, D. 60, 64–8, 73–4, 81–2 proposition 2–4, 6–11, 20–27, 33, 39–40, 42, 44–45, 49, 51, 78, 88–89, 93, 100–118, 124, 127–130, 134, 138, 141–144, 147, 149–150, 171–174, 182, 188, 198, 211, 217, 222, 235, 240; see also truth-bearer Putnam, H. 88–9, 91–3, 97, 167–8, 184, 207–8, 221–2, 225–6, 232–3, 236–7 Recanati, F. 38, 102, 104–9, 111–5 reference 30–31, 97, 198–199 Reichenbach, H. 165–6 representation 11–13, 17, 38, 42, 187–192, 195–200; and perception; see perception and representational content Rosenberg, J. 82 Russell, B. 14, 64, 177 Searle, J. 7–9, 14–5, 88–91, 96, 141 Sellars, W. 61, 127, 129, 141, 143 self-consciousness 148–149 semantics 2–3, 7, 14–15, 53, 100–101, 193, 210, 228 sentence: sentence meaning, see meaning; sentence use 2–5, 7–8, 10–11, 14, 16, 26, 31, 37–39, 103–104, 150, 172, 181, 191, 212, 214–215, 226, 244–245 skepticism 35, 43, 59–61, 63, 66, 67–69, 73, 75, 77–78, 82, 126; Agrippan skepticism 59, 67, 69, 72, 74; Cartesian skepticism 59–61, 67, 69, 75, 78, 80–82, 192–196 situation-semantics 35, 38–39, 44–54

Index  253 speech act 7, 25, 27, 30–33, 88, 101, 117, 195 Sperber, D. 104 Stanley, J. 244, 246 statement 2, 5–6, 8–9, 11, 14–16, 20–23, 25–33, 39–40, 78, 92, 102, 181, 184; see also speech act Strawson, P. F. 1–5, 13–6, 47–8, 67–8, 101, 143, Stroud, B. 60–2, 81 thought 9, 116, 118, 124, 145–150, 159–160, 162–163, 171–172, 174–180, 182–185, 189, 198, 200, 205–220, 222, 229, 231, 235, 237–239, 241–246; thoughtexpression 2, 171, 206–207, 209–217, 220–223, 237–239, 243–244; structure of a 174–177, 183–185, 211 Tomasello, M. 41–2 Travis, C. 38, 49, 88–9, 95, 114–8, 124–5, 139–51, 160, 168–9, 174, 178–85, 189–94, 198, 200 truth: truth-bearer 9, 15–16, 20–21, 23–24, 180, 182, 222;

truth-conditions 2, 7–9, 13, 16, 38, 41–42, 45, 48–49, 52, 54, 89–90, 92–96, 100–103, 106–110, 112, 114–115, 118, 171–172, 181, 187–189, 191, 193, 196, 198, 246; truth-preservation 242; laws of 145, 238–242 Twardowski, K. 33 understanding 178, 184, 206, 214, 218–220, 222–224, 227 utterance: as distinct form a speech act 28; circumstances of the 5, 11, 31–32, 39, 41, 49, 102, 103, 181; content of the 4, 11, 45, 106, 108, 110, 113, 117, 188 Waismann, F. 88–93, 96 what is said 4–7, 9, 11, 14, 24–25, 27–28, 48–49, 51–52, 100–118, 173, 181, 223; and indexicality 102–108; and interpretation 110–118; objectivity of 112–117 Wittgenstein, L. 14, 59–61, 64–82, 89, 94, 173, 177–8, 190, 197, 234 Wright, C. 67–8