Language, World, and Limits: Essays in the Philosophy of Language and Metaphysics 0198823649, 9780198823643

Citation preview

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Language, World, and Limits

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Language, World, and Limits Essays in the Philosophy of Language and Metaphysics

A. W. Moore

1

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

1 Great Clarendon Street, Oxford, ox2 6dp, United Kingdom Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries © A. W. Moore 2019 The moral rights of the author have been asserted First Edition published in 2019 Impression: 1 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by licence or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this work in any other form and you must impose this same condition on any acquirer Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America British Library Cataloguing in Publication Data Data available Library of Congress Control Number: 2019930483 ISBN 978–0–19–882364–3 Printed and bound by CPI Group (UK) Ltd, Croydon, CR0 4YY Links to third party websites are provided by Oxford in good faith and for information only. Oxford disclaims any responsibility for the materials contained in any third party website referenced in this work.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

In memory of Pamela Sue Anderson (1955–2017)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Contents Preface Publisher’s Acknowledgements Introduction

ix xi 1

Part I. Language 1. How Significant is the Use/Mention Distinction? (1986)

11

2. The Underdetermination/Indeterminacy Distinction and the Analytic/Synthetic Distinction (1997)

17

3. What are these Familiar Words Doing Here? (2002)

39

4. The Bounds of Nonsense (2019)

56

5. Transcendental Idealism in Wittgenstein, and Theories of Meaning (1985) and Postscript (2010)

71

6. The Bounds of Sense (2006)

90

Part II. The World and Our Representations of it 7. A Note on Kant’s First Antinomy (1992)

107

8. Bird on Kant’s Mathematical Antinomies (2011 and 2012)

112

9. Solipsism and Subjectivity (1996)

119

10. One or Two Dogmas of Objectivism (1999)

132

11. Apperception and the Unreality of Tense (2001)

143

12. The Metaphysics of Perspective: Tense and Colour (2004)

158

13. Realism and the Absolute Conception (2007)

165

14. One World (2016)

182

Part III. Ineffability 15. Being, Univocity, and Logical Syntax (2015)

195

16. Ineffability and Religion (2003)

210

17. On Saying and Showing (1987)

224

18. Ineffability and Nonsense (2003)

244

Bibliography Index

261 275

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Preface These essays are reprinted with relatively minor amendments. Many of the amendments are purely cosmetic. Some, such as the addition of some cross references and the introduction of some standardization, are for the sake of the volume. In a few cases I have corrected what I now see as simple philosophical or exegetical mistakes. Two cases are worth singling out. There is much in Essay 5 with which I now disagree. But I have not tried to rectify it. Rather I have included a postscript in which I indicate what my discomfort with it is, why I think that the main contention of the essay would have survived even if I had tried to rectify it, and why I think that its primary lesson concerning Wittgenstein’s later work would actually have been reinforced. Similarly, though on a smaller scale, I have added a new footnote to Essay 16 (note 16) in which I disavow some material elsewhere in the essay and reference a section in Essay 11 that I now take to undermine it. I think that it is instructive to leave the offending material in Essay 16 intact, not least to highlight what is at stake in the relevant section of Essay 11. The rest of Essay 16, by far the bulk of it, is unaffected. I have made no attempt to eliminate (sometimes verbatim) repetition from one essay to another: this is partly to accentuate interconnections between the essays, partly to ensure that each essay remains self-contained. As for the interconnections, I have an introduction whose main rationale is to elucidate these. I thank Peter Momtchiloff, philosophy editor at Oxford University Press, for his advice, encouragement, and support. I also thank the editors and publishers of the volumes in which these essays first appeared for permission to reprint them. A. W. Moore

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Publisher’s Acknowledgements We are grateful for permission to include the following copyright material in this book. Essay 1 is reproduced from Moore  A.W.  ‘How Significant is the Use/Mention Distinction?’ Analysis, 46 (4): 173–179, by permission of Oxford University Press. Copyright © 1986, Basil Blackwell Oxford, doi: 10.1093/analys/46.4.173 Essay 2 is reprinted by permission from Springer Nature: Kluwer Academic Publishers, Erkenntnis, 46 (1): 5–32, Moore  A.W.  ‘The Underdetermination/ Indeterminacy Distinction and the Analytic/Synthetic Distinction’. Copyright © 1997, doi: 10.1023/A:1005382611551 Essay 3 is reprinted with permission from Moore A.W. ‘What are these Familiar Words Doing Here?’ Logic, Thought and Language, ed. Anthony O’Hear, Royal Institute of Philosophy Supplements, 51: 147–171. Copyright © 2002, The Royal Institute of Philosophy and the contributors, doi: 10.1017/S1358246100008122 Essay 5 is reproduced from Moore A.W. ‘Transcendental Idealism in Wittgenstein, and Theories of Meaning’. The Philosophical Quarterly, 35 (139): 134–155, by permission of Oxford University Press. © The Editors of The Philosophical Quarterly, doi: 10.2307/2219340. The appendix is from The Later Wittgenstein on Language (2010), ed. Daniel Whiting. Copyright © 2010, Palgrave Macmillan; reproduced with permission of Macmillan Publishers Limited through PLSclear. Essay 6 is reproduced with permission from Moore A.W. ‘The Bounds of Sense’. Philosophical Topics, 34 , numbers 1&2 (Spring/Fall 2006). Copyright © 2006 by the Board of Trustees of the University of Arkansas. Reprinted with the permission of The Permissions Company, Inc., on behalf of the University of Arkansas Press, www. uapress.com. Essay 7 is reproduced from Moore  A.W.  ‘A Note on Kant’s First Antinomy’. The Philosophical Quarterly, 4 2(169): 480–485, by permission of Oxford University Press. Copyright © 1992, The Editors of The Philosophical Quarterly, doi: 10.2307/2220288 Essay 8 is reprinted with permission from Moore A.W. ‘Bird on Kant’s Mathematical Antinomies’. Kantian Review, 16 (2): 235–243. Copyright © 2011, Kantian Review, doi: 10.1017/S1369415411000094 Essay 9 is reproduced with permission from Moore A.W. ‘Solipism and Subjectivity’. European Journal of Philosophy, 4 (2): 220–234. Copyright © 2008, John Wiley and Sons, doi: 10.1111/j.1468-0378.1996.tb00075.x Essay 10 is reprinted from Moore A.W. ‘One or Two Dogmas of Objectivism’. Mind, 108 (430): 381–394 by permission of Oxford University Press. Copyright © 1999, Oxford University Press, doi: 10.1093/mind/108.430.381 Essay 11 is reproduced from Moore A.W. ‘Apperception and the Unreality of Tense’. Time and Memory: Issues in Philosophy and Psychology, eds. Christoph Hoerl and Teresa McCormack. Copyright © 2001, Oxford University Press; reproduced with permission of Oxford University Press through PLSclear.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

xii

Publisher’s Acknowledgements

Essay 12 is reproduced with permission from Moore  A.W.  ‘The Metaphysics of Perspective: Tense and Colour’. Philosophy and Phenomenological Research, 68 (2): 387–394. Copyright © 2007, John Wiley and Sons, doi: 10.1111/j.1933-1592.2004.tb00350.x Essay 13 is reprinted with permission from Moore A.W. ‘Realism and the Absolute Conception’. Bernard Williams, ed. Alan Thomas. Copyright © 2007, Cambridge University Press. Essay 14 is reproduced with permission from Moore A.W. ‘One World’. European Journal of Philosophy, 24 (4): 934–945. Copyright © 2016, John Wiley and Sons, doi: 10.1111/ejop.12201 Essay 15 is reproduced from Moore  A.W.  ‘Being, Univocity, and Logical Syntax’. Proceedings of the Aristotelian Society, 115 : 1–23. Reprinted by courtesy of the Editor of the Aristotelian Society: © 2015, doi: 10.1111/j.1467-9264.2015.00381.x Essay 16 is reproduced with permission from Moore A.W. ‘Ineffability and Religion’. European Journal of Philosophy, 11 (2): 161–176. Copyright © 2003, John Wiley and Sons, doi: 10.1111/1468-0378.00181 Essay 17 is reprinted with permission from Moore A.W. ‘On Saying and Showing’. Philosophy, 62 (242): 473–497. Copyright © 1987, The Royal Institute of Philosophy, doi: 10.1017/S0031819100039048 Essay 18 is reproduced from Moore A.W. ‘Ineffability and Nonsense’. Aristotelian Society Supplementary Volume, 77 (1): 169–193. Reprinted by courtesy of the Editor of the Aristotelian Society: © 2015, doi: 10.1111/1467-8349.00108 The publisher and author have made every effort to trace and contact all copyright holders before publication. If notified, the publisher will be pleased to rectify any errors or omissions at the earliest opportunity.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Introduction The rationale for collecting these essays together is that they are all concerned, at some level, with the nature, scope, and limits of representation. By representation here I mean the act of representing how things are, or the act of producing a representation. By a representation I mean anything which has content—that things are thus and so—and which, because of its content, is true or false. The essays in Part I deal with linguistic representation. One thesis that surfaces at various points in these essays is the following: (1)

Some things are ineffable.

The essays in Part II deal with representation more generally, and with the character of what is represented. They all touch more or less directly on the distinction between ‘perspectival’ representations, that is representations from a point of view, and ‘absolute’ representations, that is representations from no point of view. One thesis that surfaces at various points in these essays is the following: (2)

Nothing is ineffable.

How is the resulting tension between Part I and Part II to be resolved? The essays in Part III, deriving their inspiration from the early work of Wittgenstein, indicate a way. We can construe the ‘things’ of which (1) says that some are ineffable as states of knowledge or understanding, and the ‘things’ of which (2) says that none are ineffable as facts or truths. This allows us to affirm both that, however things are, they can be said to be that way, in line with (2), and that there are nevertheless insights of a kind that are beyond expression, in line with (1).

1. Part I: Language In Essays 1 to 3 I consider some of the things that we do with words and some of the ways in which we might characterize these things. More specifically, I consider some of the distinctions that we might draw as a means to characterizing them. Among these are: (i) distinctions that we find it very natural to draw; (ii) distinctions that we are inclined to draw only as a result of philosophical reflection; and (iii) distinctions that we are disinclined to draw—in fact whose drawing we are inclined to repudiate— as a result of philosophical reflection. Self-conscious attention to our linguistic behaviour leads us to try to disentangle these, and, where a distinction belongs to both

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

2

Language, World, and Limits

category (ii) and category (iii), a fortiori where it belongs to both categories by virtue of the same philosophical reflection, to try to determine what should give: the distinction; or the deliverances of the relevant philosophical reflection; or the power of these deliverances to incline us one way or another. All of these possibilities are illustrated in the first three essays. In the course of these essays I argue that some of the distinctions in question are sufficiently well suited to their purposes that we need feel no compunction about continuing to draw them, while others are genuinely problematical. Roughly, the more natural it is for us to draw one of these distinctions, the more likely it is to be of the former kind. Essay 3 is a kind of focal point for the first three essays: it recapitulates some of Essay 1 and some of Essay 2. Its main burden is that we can hold fast to a relatively common-sense conception of various things that we do with words despite the power of philosophical reflection to make us doubt whether we can do these things at all. The distinctions at stake in Essays 1 to 3 are mostly distinctions among representations. Self-conscious attention to our linguistic behaviour is also liable, sooner or later, to make us consider the distinction between representations and non-representations—which, on a suitable construal of sense, and relative to a suitable domain, is the distinction between sense and nonsense. In any enquiry of this kind, this is the most fundamental distinction of all. There is material on it in both Essay 2 and Essay 3. But it does not dominate until Essays 4 to 6. It is a distinction that clearly falls into category (i) above: that is to say, it is a distinction that we find it very natural to draw. Yet philosophical reflection exposes problems involved in the drawing of it—if not to the extent of placing it in category (iii), then certainly to the extent of raising awkward questions about it. There seem to be reasons of principle why we shall never be able to attain and express any general philosophical understanding of why the distinction should be drawn in the way in which it should. Or so I argue in Essay 6. Essays 4 and 5 provide a kind of prolegomenon to Essay 6. In Essay 4 I explore the domain within which this distinction between sense and nonsense is drawn. Or rather, I explore the domain within which a distinction between sense and nonsense is drawn. Essay 4 has a concern with one kind of nonsense. The distinction at stake in  this essay is the distinction (albeit not expressed in these terms) between our producing representations and our merely seeming to do so. And the essay has an additional exegetical concern: how Wittgenstein treats of this distinction in his Tractatus, at the end of which he famously admits that he himself has been doing the latter, in other words that he himself has been giving the mere impression of making sense. One of the main lessons of this essay, if it is successful, is that sense and nonsense, so construed, are not two species of an independently intelligible genus, but a phenomenon and its non-independently-intelligible impostor. But this still leaves us without a general philosophical account of the phenomenon: a general philosophical account, in other words, of what it takes for something to be the genuine article. Relatedly, it leaves us without what I call in Essay 5 a philosophically substantial theory of meaning. I argue in Essay 5 that due assimilation of Wittgenstein’s later work must make us despair of ever having such a thing—if, that is, Bernard Williams’ claim that there is a kind of transcendental idealism in Wittgenstein’s later work is correct. In the essay I make clear my sympathy for this claim. In a postscript, written

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Introduction

3

for the reprint of the essay, I indicate my lack of sympathy for it. But I also argue that my change of mind does not in the end matter. For the main contention of the original essay, as I now conceive it, is that a philosophically substantial theory of meaning is precluded by Wittgenstein’s later work because it would force us to answer certain questions that we cannot answer without explicitly endorsing the transcendental idealism in question—which we have no satisfactory way of doing. According to the Wittgenstein of the original essay, the fault would lie, not with the transcendental idealism, but with our attempt to state it, and hence with the theory of meaning, for forcing us to make the attempt. According to the Wittgenstein of the postscript, the fault would lie with the transcendental idealism too, though it would lie principally with the theory of meaning, for putting us in this position in the first place. Either way, the theory of meaning would be at fault. And to conclude this is to conclude something that is a variation on the previously proclaimed theme: there seem to be reasons of principle why we shall never be able to attain and express any general philosophical understanding of why the distinction between sense and nonsense should be drawn in the way in which it should. At the end of Essay 6 I moot the idea, for which Essay 5 has already prepared us, that the second conjunct in the phrase ‘attain and express’ is critical: the attainment of such an understanding may not be what is beyond us, only its expression. This would be our predicament if any attempt to express such an understanding automatically meant failing to respect the very distinction (between sense and nonsense) that is its subject matter. Among the reasons for thinking that this is our predicament is the fact, as I argue in both Essay 5 and Essay 6, that an analogous predicament afflicts Kant’s prototypical version of transcendental idealism. And if we do indeed find ourselves stewing in these Kantian juices, as I believe we do, then one immediate consequence is of course thesis (1): that some things are ineffable.

2. Part II: The World and Our Representations of it Kant’s transcendental idealism reclaims our attention in Essays 7 and 8. In these two  essays I consider Kant’s treatment of questions concerning the age, size, and composition of the physical universe. Kant’s treatment of these questions is grounded in one of the most significant aspects of his transcendental idealism: his belief that we are incapable, in principle, of representing what the physical universe is like except from some point of view; or, more specifically, his belief that there is one particular point of view such that only representations from that point of view can count as representations of what the physical universe is like in the first place. Kant’s transcendental idealism thus connects with the question whether representations that are not from any point of view, or absolute representations as I called them above, are possible at all. I have argued elsewhere, in broad opposition to Kant, that they are.1 The issues that arise here—about the very distinction between absolute representations and perspectival representations, about what it takes for a fact to be represented perspectivally and whether it takes that fact to be itself in any sense 1 Moore (1997), Ch. 4.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

4

Language, World, and Limits

perspectival, and about the assumptions on which my argument for the possibility of absolute representations rests—animate Essays 9 to 13. Essay 9 is concerned with knowledge. In particular it is concerned with what I call in the essay solipsism: the extreme sceptical view (roughly) that I have no knowledge except of how things seem to me, or (equivalently?) that my knowledge extends no further than my experiences. After mooting a less rough formulation of the view, and after following a rather long and tortuous path, I conclude that solipsism, so formulated, can be resisted; but also that the formulation in question is not, after all, a satisfactory one. This leads to speculation about the status of solipsism itself. Perhaps it does not have a satisfactory formulation—say because it is incoherent, or because it is ineffable. The second of these possibilities is especially interesting given the thrust of Part I. In fact, however, my own inclination is to endorse the first possibility. I do not deny that there are connections with ineffability. But I believe that they lie elsewhere. As for the connections with the notion of a point of view, I argue towards the end of the essay that one of the objections to solipsism is that it makes a mystery of my subjectivity, and that it does this because it makes a mystery of my self-conscious grasp of my own most fundamental point of view on the world. It is in Essays 10 to 13 that I am most directly concerned with the question whether representations from no point of view are possible—as I have already indicated I think they are. But I am as keen to contextualize the question, and to disentangle it from other questions with which it might all too easily be conflated, as I am to justify my own view of the matter. Thus I spend much of Essay 10 distancing myself from someone who initially appears to be an ally, namely Thomas Nagel, who argues in Nagel (1997) that some of our beliefs are indeed from no point of view. I think he arrives at this conclusion too quickly. In particular I think he significantly underestimates how hard it can be to tell that a belief is from some point of view. Essay 11 pursues the potential for slippage here, with a focus on temporal points of view. I try to explain (with further reference to Kant) why the following question is an open one: given a tensed representation, which is just to say a representation from a temporal point of view, is there anything in reality that corresponds in any suitably direct way to its tense? Open question or not, my own view is that the answer is no. I think it would be an abnegation of the unity of reality to think otherwise. More generally, I think that, if reality is unified, then representing perspectivally how things are had better not be thought of as representing how things are perspectivally. In Essay 12 I turn my attention to the significance of the question whether absolute representations are possible. I argue that, if we were able to demonstrate that they are, and if, moreover, we were able to demonstrate that any fact can be conveyed by such a representation (which would be a strictly stronger conclusion, although, as it happens, still a conclusion to which I would subscribe), then we would have successfully undertaken what Barry Stroud calls ‘the philosophical quest for reality’. This sets me apart from Stroud himself, whose aim in the book in which he introduces this label2 is to convey scepticism about whether any such quest can ever be successfully undertaken. The ideas about tense canvassed in the previous essay inform my argument in Essay 12, as do certain counterpart ideas about colour. 2 Stroud (2000).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Introduction

5

The inspiration behind much of the material in Essays 10 to 12 is the work of Bernard Williams. My reasons for thinking that absolute representations are possible are broadly Williams’ reasons for thinking it: here I am adverting to Williams’ wellknown argument for the possibility of what he calls an ‘absolute conception of reality’, a conception of reality that consists of just such representations. In Essay 13 this argument of Williams’ comes to the fore. But my aim in Essay 13 is not to defend the argument (which, as I remarked earlier, I have tried to do elsewhere). My aim is rather to see what we can learn from the argument about its underlying realism, famously expressed by Williams as follows: ‘knowledge is of what is there anyway’.3 (It is a common mistake, incidentally, to think that by ‘knowledge of what is there anyway’ Williams means absolute knowledge. But that makes a complete mockery of his argument: it casts the principal premise of the argument as not much different from its conclusion. All knowledge, for Williams, is knowledge of what is there anyway: what distinguishes absolute knowledge is that it is knowledge of what is there anyway as it is anyway, not as it is from this or that point of view.) Part of what it is to accept such realism, I claim, is to be committed to the very unity of reality that I paraded in Essay 11. I did not argue for that unity in Essay 11. On the contrary, I suggested that to assume it is to adopt a basic unargued assumption to which there are alternatives— which is why I took the question about tense that I raised in Essay 11 to be an open one. The same message can be heard here. But granted such a basic assumption, we can put it to significant use. Most significant, in the context of Essay 13 itself, is that we can use it to obtain an especially firm grip on some of Williams’ own characteristic views about ethics. But even more significant, in the context of the overall dialectic of these essays, is that we can use it in the way in which Williams himself uses it, to argue for the possibility of absolute representations—and also, relatedly, to argue against the existence of ineffable facts (more below). However, ‘granted such a basic assumption’ is the operative phrase. The assumption serves as a kind of conceptual datum. The unity of reality, to repeat, is part of that datum; and it is the focus of Essay 14. Although I do not believe that such unity can be argued for, I use this essay to motivate it. I do this by considering, within a Kantian framework, what such unity consists in and then urging that to reject it would be to incur a problematical commitment to the existence of things that cannot be expressed from ‘our’ point of view. I think we should deny the existence of things that cannot be expressed from ‘our’ point of view (just as I think we should deny the existence of things that cannot be expressed from no point of view: see my comments on Essay 12 above). A fortiori I think we should deny the existence of things that cannot be expressed, period. I have more to say about this in Part III. What this denial comes to, of course, is the endorsement of thesis (2): that nothing is ineffable.

3. Part III: Ineffability Essay 15 is hard to classify. It might just as easily have appeared either in Part I or in Part II. It involves consideration of the nature of reality, via consideration of our linguistic representations of reality. My reason for including it in Part III is that the 3 Williams (1978), p. 64.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

6

Language, World, and Limits

tension between theses (1) and (2), which is what primarily links the essays in Part III, is felt within Essay 15 itself. This essay is concerned with what it would take to repudiate the transcendent, where the transcendent is construed as that whose most fundamental character is beyond all but analogical expression. The repudiation of the transcendent, so construed, clearly entails the repudiation of that whose most fundamental character is beyond all expression. That is to say, it entails the repudiation of the ineffable: it entails the endorsement of (2). Yet there is a hint, at the very end of the essay, that it involves an acknowledgement of the ineffable. There is a hint, in other words, that it involves an endorsement of (1). The three remaining essays in Part III address the tension between these two theses through an exploration of the very idea of ineffability. All three essays indicate how the tension is to be resolved. The resolution is as advertised above: (2) is a thesis about facts or truths; (1) is a thesis about states of knowledge or understanding. Note that in some of these essays there is talk of ‘insights’, and in some of them talk of ‘states of enlightenment’: but I do not mean anything other by these phrases than certain states of knowledge. And where, in Essay 17, I say that such states are always grounded in states of understanding, I would be happy to cast the entire discussion, including this reference to what such states are grounded in, in terms of states of knowledge. For I believe that states of understanding are themselves states of knowledge of a certain kind. True, in Essay 18 I leave open the possibility that they are not. But this is a caginess adopted for essay-specific purposes only. It is supposed to help signal the relative strength of my various commitments. As I explain in note 35 of that essay, if I could be persuaded to surrender my belief in ineffable knowledge, then I would rather accede to the possibility of states of understanding that are not states of knowledge than surrender my belief in ineffable understanding.4 Essay 16 is concerned with religion. More specifically, it is concerned with religious language. It is in this essay that I give the fullest reasons for thinking that no truths are ineffable: see especially sections 1 and 2. But I also go on to give reasons for thinking that some states of knowledge are ineffable. I then argue that religious language is very often what accrues from a (necessarily unsuccessful) attempt to put these states of knowledge into words. In Essay 17 I proceed with reference to the early work of Wittgenstein—supplemented, in the final section, with reference to his later work. I try to show how both the early work and the later work help us to appreciate what it takes for a state of knowledge to be ineffable. And indeed it is in this essay that I give the fullest reasons for thinking that this is what some states of knowledge are: see especially sections 3, 4, and 7. This essay contains several links with other essays. Thus section 5 is largely about unity, the unity that was at issue in Essays 11 and 14, the unity, in other words, of reality. I suggest that our insight into this unity is an ineffable insight. And in section 8 there is a link back to Essay 5: I revisit the idea that there is an inducement in Wittgenstein’s later work to put what is ineffable into words, in particular an inducement that results from reflection on our own linguistic understanding. Part of my subsequent change of mind about Essay 5 is that I now think that Wittgenstein himself provides us with the resources to resist this inducement. But that does not stop 4 The caginess is not replicated elsewhere: see Moore (1997), pp. 161 and 183 ff.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Introduction

7

the inducement from being there—and, at the end of Essay 17, I provide some evidence that Wittgenstein himself is not always firm in his resistance to it. In Essay 18 I again proceed with reference to the early work of Wittgenstein. This final essay reinforces my resolution of the tension between (1) and (2). But it also thereby provides a resolution, indeed the same resolution, of what has come to be a fierce exegetical debate about the Tractatus. A more or less traditional exegete is liable to say that, in the Tractatus, Wittgenstein holds that some things are ineffable. More recent exegetes, or many of them anyway, are liable to say that he holds no such thing. We are now in a position to say that, suitably understood, both are right. The ‘things’ to which the more traditional exegetes are alluding are states of understanding; and Wittgenstein indeed does hold that some of these are ineffable. The ‘things’ to which the more recent exegetes are alluding are truths; and Wittgenstein indeed does not hold that any of these are ineffable. On the contrary, he gives us reason to deny it. He gives us reason, in sum, to endorse both (1) and (2).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

PA RT I

Language

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

1 How Significant is the Use/Mention Distinction? Abstract It is argued that the use/mention distinction, if it is to be a clear-cut one, cannot have the significance that it is usually thought to have. For that significance attaches to the distinction between employing an expression in order to draw attention to, or to talk about, some aspect of the world, as determined by the expression’s meaning, and employing it in order to draw attention to, or to talk about, the expression itself—and this distinction is not a clear-cut one. In the final section of the essay this argument is extended to cast doubt on a rather glib appeal to the use/mention distinction that is frequently made in the philosophy of language.

1. The distinction introduced Let us take for granted that there is a distinction to be drawn between using an expression and mentioning it. And let us say that an expression is employed when it is put to any kind of service. Then I shall assume each of the following: that using an expression (in the relevant quasi-technical sense) involves employing it in such a way as to exploit (one of) its meaning(s); and that mentioning an expression involves employing it together with certain other linguistic devices that are themselves being used to demonstrate, or ‘to point to’, some aspect of this employment. These assumptions take their cue from Donald Davidson’s essay ‘Quotation’.1 Davidson there says that quotation marks are used to refer to expressions by demonstrating particular tokens of them. I have deliberately said something much looser than this. For one thing, as Davidson acknowledges, quotation marks are not the only device that enables us to mention expressions. Also, for reasons that will emerge, I prefer not to commit myself on the question of just what is being demonstrated when such devices are used. Note that it is perfectly possible to employ an expression without either using it or mentioning it. Consider, for example, saying ‘Cheese’ when having a photograph taken. More interestingly, it is possible for an employment of an expression to be both a use and a mention of it. For example, one would be both using and mentioning the quoted words if one asserted the following sentence: 1 Davidson (1984b). Cf. also Goddard and Routley (1966), p. 21, and Whiteley (1957).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

12

Language, World, and Limits (1) Quine describes the totality of our knowledge and beliefs as ‘a man-made fabric which impinges on experience only along the edges.’

Again, one would be both using and mentioning ‘Arthur’ if one said: (2)

Arthur was so named after an uncle.

There are many ways of mentioning an expression other than by using quotation marks or anaphoric phrases such as ‘so named’. One way has just been illustrated: displaying an expression on a separate line of print (here with a numerical label). Writing an expression in italics is another way. Employing an expression after such a phrase as ‘the word’ or ‘the expression’ is another very common way. For example, one can say that ‘transcendental’ is an esoteric word by using the following sentence: (3)

The word transcendental is esoteric.

Another common way of mentioning an expression is simply to employ it as a singular term. One could just as well have used: (4)

Transcendental is an esoteric word.

If the expression is one that would normally be used as a singular term, an ambiguity arises. But the ambiguity will usually be easy to resolve. Compare how ‘Elizabeth’ would be employed in a typical use of: (5)

Elizabeth has arrived.

with how it would be employed in a typical use of: (6)

My name is Elizabeth.

No doubt there are other ways of mentioning expressions besides these. The little that I have said so far leaves it unsettled whether a whole range of uses also count as mentions—for example, uses of expressions in scare-quotes or uses of expressions in italics when they are being defined. (Consider the scare-quoted ‘to point to’ and the italicized ‘employed’ in the opening paragraph of this essay.) I have certainly not provided a comprehensive and determinate catalogue of the devices that can be used to mention expressions. Nor am I interested in doing so. I think that anybody attempting to devise such a catalogue would find that, if it could be done at all, then eventually more or less arbitrary stipulations would be called for. For my main contention is this: the kind of significance that the use/mention distinction is usually thought to possess must in fact, if the distinction is to be a clear-cut one, reside elsewhere.

2. The impossibility of the distinction’s both being clear-cut and having the significance that it is usually thought to have It is usually thought that the significance of the use/mention distinction consists in its coincidence with another distinction: the distinction between employing an expression in order to draw attention to, or to talk about, some aspect of the world,

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

How Significant is the Use/Mention Distinction?

13

as determined by the expression’s meaning, and employing it in order to draw attention to, or to talk about, the expression itself. This latter distinction is significant because it is sometimes easy to confuse these ways of employing expressions—with serious consequences. It is undeniable that invoking the use/mention distinction will often serve to combat such confusion. Imagine, for example, somebody’s challenging scientific pretensions to have discovered that heat is mean kinetic energy on the grounds that the latter, unlike the former, is a piece of technical terminology used only by a small group of initiates. But notions like what one is employing an expression for, what one is drawing attention to, and what one is talking about are grounded in the pragmatics of communication. How they apply is a messy, complex, indeterminate matter, heavily dependent on particular circumstances and varying in degree from one case to the next. (What is Dylan Thomas employing the following words for when he writes, ‘the sloeblack, slow, black, crowblack, fishingboat-bobbing sea’?) If the use/mention distinction is to be clear-cut—if there is to be a range of easily recognizable and specifiable devices that register when an expression has been mentioned—then, even if the distinction has an interesting bearing on what expressions are employed for, it cannot be right to insist on the coincidence outlined above. And this in turn means that it cannot be right to credit the use/mention distinction with the significance that it is usually thought to have. The sentences (7) to (12) below help to confirm this. (7) The city where Kant lived was then Königsberg but it has now changed to Kaliningrad (8) Christopher can never remember his four-times table; he thinks that three fours are sixteen (9) This is sepia (10) Do you know the woman who works there called Catherine? (11) ‘This is ridiculous,’ she thought to herself (12) In a certain sense of ‘know’, nobody knows anything Think about typical uses of these sentences. Imagine (9), for example, being used while pointing to something in order to provide an ostensive definition, in response to the question: what colour does the word ‘sepia’ pick out? (On one fairly radical view, whereby the meaning of any given expression is continually evolving, this would not be fundamentally different from a supposedly non-defining use of (9).) Now consider what the underlined expressions would be being employed for. This would certainly not depend in the requisite way on whether they were being used or mentioned—provided, that is, that it were transparent whether they were being used or mentioned, the quotation marks in (11), for example, signifying the latter. For excepting (12), the cases where it would be most natural to say that the expressions were being employed to draw attention to, or to make some claim about, themselves ((7) to (9)), would be precisely the cases where they were not being mentioned. And in (12) the two underlined expressions would naturally be said to be operating in tandem, in such a way that little further significance (beyond the demands of grammar) could attach to the fact that one, but not the other, was being mentioned.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

14

Language, World, and Limits

Of course, one could deny that it was transparent which expressions were being mentioned in these cases. After all, sentences can have a misleading surface grammar (for example, they can contain ellipsis), and this means that an expression can be mentioned surreptitiously—or only apparently. So one could insist on a coincidence between the use/mention distinction and a distinction based on what expressions are being employed to draw attention to or to talk about, even in the face of examples such as these. For instance, one could say that ‘Königsberg’ was being mentioned in the first example and that it only seems otherwise because of an ellipsis. (The city was then called Königsberg.) On the other hand, one might be reluctant to admit such radically false appearances, perhaps because one felt that it would require undue ad hocery. (Think about accommodating a use of ‘Catherine’ in (10) to talk about Catherine herself.) One might think that it must always be transparent when an expression has been mentioned (ambiguities apart). But one could still retain the form of the coincidence by developing a suitably technical and non-pragmatic sense of drawing attention to, or talking about, something—for example, referring to it, where this is a notion of reference determined by a theory of strict and literal truth conditions. (This would involve denying that somebody using (7) had, strictly speaking, said something true, though he or she may have succeeded in conveying a true thought.) It is no part of my contention to deny that one could take such tacks as these. What one could not do is maintain both of the following: that the use/mention distinction is a clear-cut one; and that it has the significance usually attributed to it. If one took the first tack above and tied the distinction to what people are talking about in the relevantly vague and pragmatic sense, then the use/mention distinction would itself be correspondingly indeterminate. Consider: Jim remembers almost nothing of the physics he once knew, but he still knows that electrons are elementary particles with negative charge. Did I just make a claim about electrons? Or about the word ‘electrons’? Or both? What was I drawing attention to when I employed the word ‘electrons’? Is there a fact of the matter? If not, there is no fact of the matter, on this approach, whether I just mentioned the word ‘electrons’. If one took the second tack and introduced a technical sense of talking about things, then one would remove the use/mention distinction from the arena of pragmatics and also, thereby, from the arena where the confusions to which it was supposed to be pertinent arise. For example, as a matter of pragmatics, one could use (8)  to convey a truth about certain arithmetical expressions—a truth which arguably could not be conveyed in any other way—even though, on this approach, one would not have talked about those expressions in the relevantly technical sense and might indeed have said something literally false. The fact remains that one would be suffering from the kind of confusion that is of concern here if one’s purpose in thus using (8) was, say, to undermine the thesis that it is impossible to think something that is impossible, such as an arithmetical absurdity. (The relevant kind of confusion has to do with what one is talking about in a pragmatic, non-technical sense.) This approach therefore has the consequence that the use/ mention distinction does not impinge directly on such confusion and so does not have that particular significance, precisely the significance that it is usually thought to have.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

How Significant is the Use/Mention Distinction?

15

3. Repercussions I have said that if there is a range of easily recognizable and specifiable devices that register when an expression has been mentioned, then the fact that an expression has been mentioned cannot be related in any neat way to what it has been employed for. Corroboration for this comes from further reflection on what one of the relevant devices would be: quotation marks. Davidson suggests that we read a pair of quotation marks as ‘the expression a token of which is here’.2 An expression is supposed to be a purely syntactic entity. But surely there are times when it would be more appropriate to read a pair of quotation marks as ‘the word a token of which is here’, where a word is construed as partly semantic. Consider, for example, saying: (13)

‘Transcendental’ is an esoteric word

(14)

It would be an insult to call somebody ‘clever’,

or not meaning thereby that it would be an insult to use that expression. Indeed sometimes quotation marks are used to pick out something purely semantic, as in a typical use of: (15)

The answer to this question is ‘No’

where one might just as well have used: (16)

The answer to this question is no.

Davidson could obviously admit that quotation marks are capable of being used in these ways, by insisting that they are ambiguous; and he might go on to say that only by using quotation marks in the way he describes does one strictly mention an expression. But this would be sufficiently ad hoc to require justification. (It would not be outrageous, however. We might want to single out a distinct use of quotation marks for, say, referring to essays by their titles.) A likelier proposal, which could still be offered within the framework of a demonstrative theory, is that quotation marks resemble such indexicals as ‘this’ and ‘that’ in leaving open a wide range of things that they can be used to demonstrate, the demonstratum on any occasion of use being fixed by appeal to context.3 If this is roughly correct, the possibility arises that they can be used, in particular, to demonstrate actual employments of expressions, or meaningful utterances, and this in turn suggests that they have a use very much like the use of ‘that’ which Davidson outlines in his paratactic account of indirect discourse in his essay ‘On Saying That’.4 On this construal, there could be a perfectly legitimate use of: (17)

‘Grass is green’ is true

that had exactly the same import as a typical use of: 2 Davidson (1984b), p. 90. 3 Cf. Goddard and Routley (1966), p. 3 and pp. 16–17 and Goldstein (1984), p. 4. 4 Davidson (1984c).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

16

Language, World, and Limits (18)

It is true that grass is green.

And it seems to me that there could be. However that may be, we see here once again that what one is employing an expression for when one mentions it is heavily contextdependent and capable of merging imperceptibly into what one is employing it for when one uses, but does not mention, it—provided that it is appropriately transparent when an expression has been mentioned. (If it is not, use and mention themselves merge imperceptibly into each other.) This reinforces my main contention. But it also casts doubt on a rather glib appeal to the use/mention distinction that is frequently made in the philosophy of language, and I shall close by making some brief comments about that. Consider the two following sentences: (19) Mary knows that ‘eiπ + 1 = 0’ is true (20) Mary knows that eiπ + 1 = 0 It would be typical of what I have in mind to insist, simply on the strength of the use/mention distinction, that somebody who asserted both of these sentences would be making two very different claims, the difference being one which (for familiar reasons) has a vital bearing on questions of understanding. I do not want to deny that these sentences could be used to make two different claims. But I do want to insist, in the light of the discussion above, that they could equally well be used to make the same claim. So provided we can be confident that somebody asserting (19) would be mentioning the mathematical expression, and that somebody asserting (20) would not be, then simply appealing to this fact does not ensure that there would be any difference between the two claims that the person was making. Consequently, if we do want to consider uses of (19) and (20) to make two different claims, much more needs to be said about what we are considering and what philosophical significance it is supposed to have than is provided by some cursory appeal to the use/mention distinction. This is especially significant because of a certain tension that will be felt (a tension that is typical of what will be felt in other similar cases). The firmer a grip we have on the difference between the two claims, then the more we shall be viewing the linguistic entity to which Mary is being related by the first claim as an abstract and theoretical construct (a sequence of typographical shapes, say) and the less we shall think of it as something that ordinarily impinges on people who handle the language of mathematics with understanding, so that the less, in turn, we can be confident that what we are considering is able to throw much light on what it is to handle the language of mathematics with understanding. A similar tension will be felt ‘from the bottom up’ with respect to the second claim. For we shall feel some pressure to view that claim in such a way that it does not treat of Mary’s linguistic practices or of what she knows about the language of mathematics, and this is, to put it mildly, not easy. Be that as it may, we are brought to realize that here too the use/mention distinction might not have quite the significance that it has customarily been supposed to have.5

5 My thanks are due to Anita Avramides, Naomi Eilan, Andrew Rein, and an anonymous referee for Analysis.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

2 The Underdetermination/ Indeterminacy Distinction and the Analytic/Synthetic Distinction Abstract Two of Quine’s most familiar doctrines are: that there is a distinction between underdetermination and indeterminacy; and that there is no distinction between analytic and synthetic truths. An argument is given that these two doctrines are incompatible. In terms wholly acceptable to Quine and based on the underdetermination/ indeterminacy distinction, an exhaustive and exclusive distinction is drawn between two kinds of true sentences, which, it is argued, corresponds to the traditional analytic/ synthetic distinction. An appendix is used to develop one aspect of the underdetermination/ indeterminacy distinction, as construed here, and to discuss, in passing, some of Quine’s more general views on truth.

Two distinctions that have played a major rôle in the work of W.V. Quine are that between underdetermination and indeterminacy and that between analytic and synthetic truths. Associated with these two distinctions are two of his most familiar doctrines. He endorses the former distinction and he rejects the latter. I hope to show that these doctrines are incompatible.1 In section 1 I shall say how I understand each of the two distinctions. In section 2 I shall try to show that Quine’s two doctrines are at least in tension with each other. In section  3 I shall argue for their incompatibility.

1. The two distinctions

1.1 The underdetermination/indeterminacy distinction The distinction between underdetermination and indeterminacy applies where there are two accounts of a certain matter satisfying the following three conditions: (i) the two accounts are incompatible, that is to say there is a sentence (with a fixed interpretation) that is true according to one account and false according to the other; (ii) the two accounts are empirically equivalent, that is to say they are both compatible 1 This idea is not new. See e.g. Rorty (1980), pp. 192 ff. But I think my argument is new.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

18

Language, World, and Limits

with the same possible evidence; and (iii) one of the two accounts is true.2 The distinction to be drawn is between cases where there is, and cases where there is not, a fact of the matter concerning (iii), in other words a fact to determine the truth of one account and the falsity of the other. In cases of both kinds the choice between the two accounts (by which I mean, simply, the matter of which is true and which is false) is underdetermined by the evidence. But if there is a fact of the matter, the choice is merely underdetermined. If there is not, it is indeterminate. Quine not only accepts this distinction, he also has clear views about where to draw it. Thus he insists that, if the two accounts in question belong to the natural sciences, the choice between them is merely underdetermined; if they are two accounts of what the predicates in a given language denote, the choice between them is indeterminate.3 I shall return to these paradigms from time to time as convenient pegs on which to hang different parts of the discussion. But I am less concerned with whether or not  there is a fact of this or that particular matter than with the very idea of the distinction. Both the distinction itself and Quine’s views about where to draw it are grounded in his naturalism, that is in his conviction that, to arrive at a theory of what the world is like, we have neither need nor room for a philosophical propædeutic to ordinary scientific endeavour. Thus the notion of evidence itself is understood in terms of what we can learn, through that endeavour, about the stimulation of our senses. Quine’s picture is as follows. Even given an inventory of every individual sensory stimulation, past, present, and future, there are different stories to be told—in some cases, different in such a way that no possible evidence could decide between them. Some of these stories, but not all of them, differ with respect to the facts. That is, in slightly cruder terms, they differ with respect to what is going on in the world. For example, some differ in how they reckon the broad structure of the universe: they have different geometries and they say correspondingly different things about the shrinking and stretching of bodies as they move about in space.4 Others differ in their basic conceptual apparatus: they presuppose objects of different kinds. Others again differ in respects that have become familiar through philosophical discussions of scepticism: there are stories according to which our senses have been deliberately manipulated to

2 My appeal to truth in this formulation is already liable to raise some eyebrows. I have in mind an extremely thin conception of truth, whereby the truth predicate is a device of disquotation (as Quine himself puts it in Quine (1970a), p. 12). To call a sentence or an account true is a surrogate for coming straight out with it. Coming straight out with the account may in turn be a matter of making some arbitrary choice. For instance, if the account is an account of what the predicates in a given language denote, then coming straight out with it may mean opting for one manual of translation instead of opting for some equally acceptable alternative: see Quine (1990a), p. 51. If so, then calling the account true is simply a way of registering which choice one has made. There is much more to be said about this, however. I am well aware that these comments will not suffice to appease everyone’s worries. I have therefore added an appendix in which I address the worries more fully. 3 There are numerous references. But see e.g. Quine (1969d), pp. 302–303. Two essays that are especially good on Quine’s distinction, and on where Quine wants to draw it, are Friedman (1975) and Gibson (1986). Also very helpful is Hookway (1988), esp. Pt. III, pp. 125–182. 4 Cf. Quine (1990a), pp. 96–97.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

19

mislead us, undetectably, about what is going on out there.5 One of these accounts is the truth. The facts are as that account says they are. If someone asks, ‘But which account is the truth?’, we have no recourse but to cite whichever account we have arrived at by ordinary scientific endeavour. Thus Quine’s naturalism. However, even given that account, indeed given a complete inventory of the facts, still there are certain matters that remain unresolved. Still there are different stories to be told. Some of these stories differ in what they say about certain very abstruse issues in mathematics, issues that are impervious to anything that is going on in the world.6 Others differ in what they say about the facts of linguistic behaviour: they do not differ about what the facts are, but they differ in how they represent the facts.7 It is because we have the linguistic and conceptual wherewithal to deal with issues that go beyond the facts in this way that there is such a thing as indeterminacy, as opposed to mere underdetermination by the evidence. This picture is a physicalist picture. The facts are physical facts, facts about the distribution of whatever fundamental states are recognized by physics. For something to be going on in the world just is for those states to be undergoing some kind of redistribution. (Not that, where the choice between two accounts is merely underdetermined, the true account need make explicit reference to these states, nor yet that there need be a paraphrase of it that does so. To expect as much would be to harbour a radical kind of reductionism. The point is simply this. Only the true account is compatible with the actual distribution of physical states.) Let us follow Quine and define ‘physical equivalence’ as compatibility with all the same distributions of fundamental physical states.8 Then what Quine is claiming, in these terms, is that physical equivalence cuts finer than empirical equivalence; and that identity cuts finer still.9 This is what enables him to draw his distinction. That there can be empirical equivalence without physical equivalence allows for mere underdetermination; that there can be physical equivalence without identity allows for indeterminacy. Someone may protest, ‘No; the physicalism is independent. The original picture was quite neutral concerning what the facts are like.’ But really, the physicalism is nothing over and above the original naturalism, or scientism as it might also be called. This is because physics is here being understood à la Quine, as that science whose business is to identify the fundamental states whose redistribution constitutes anything’s happening.10 Nothing is being added to the picture when it is said that all 5 For further discussion of the relationship between underdetermination and scepticism see Bergström (1993). 6 Cf. Quine (1990a), §40, pp. 94–95. Cf. also Quine (1986c), p. 430 and Quine (1995), pp. 56–57. 7 We should not forget that there are such facts. It is to these that an assignment of denotations to predicates (say) is answerable. To say that there is an indeterminacy here is not to deny these facts. It is simply to say that they do not force the choice between empirically equivalent but incompatible assignments. Cf. Quine (1986c), p. 429 and Quine (1986d), pp. 459–460. I shall amplify on this in section 2. 8 Quine (1981a), p. 23. (I have slightly simplified Quine’s definition.) 9 Identity of what? I have been talking about ‘accounts’. I shall introduce a more refined notion in section 3. Quine talks sometimes about ‘theories’, sometimes about ‘theory formulations’. He adopts the latter terminology whenever he either wants to allow empirical equivalence to suffice for identity of theory or wants to forestall that question on the grounds that it is basically verbal. See Quine (1981b), p. 24 and Quine (1990a), p. 96. 10 See esp. Quine (1981f), p. 98. See also, in more detail, Quine (1977). For a good summary of the implications of this view of physics for indeterminacy, see Quine (1986b), §V, pp. 187–188.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

20

Language, World, and Limits

the facts are physical facts. Nor is anything being excluded—not even the existence of Cartesian souls. It is just that if, included in the various happenings in the world, were indeed the antics and vicissitudes of Cartesian souls, then it would behove physics to recognize and to deal with suitably immaterial states.11 Now, given the existence of underdetermination, the following epithet makes perfectly good sense: ‘empirically warranted, but false’. (‘Empirically warranted’ here just means ‘compatible with all the evidence’.)12 Quine has been troubled by this. He is reluctant to discredit any account that is empirically warranted. So he considers an expedient that he attributes to Donald Davidson. Given an account of things A which, despite being empirically warranted, is incompatible with the truth, the incompatibility can be resolved by reconstruing certain key terms in A as mere homonyms of the terms that appear in the true account. Reconstrued, A can be regarded as true.13 This is all very well. The fact remains that, unreconstrued, it is false. Anything incompatible with the truth is false, howsoever warranted. Just as the existence of underdetermination entails that there can be accounts that are empirically warranted but false, so too the existence of indeterminacy—as I have presented it—entails that there can be accounts that are physically warranted, but false. (‘Physically warranted’ means ‘compatible with all the facts’.) This is much more radical. Indeed my exegesis will strike many as absurd—as will the doctrine itself. To discuss this now would be too great a diversion. I defer discussion to the appendix.

1.2 The analytic/synthetic distinction Now to the analytic/synthetic distinction.14 I shall base my comments on his essay ‘Two Dogmas of Empiricism’, the locus classicus.15 There has been much discussion about how far, if at all, Quine has modified his views since then.16 But his early account is still an ideal point of departure; and I think my main contentions apply even given the modifications, if such they be. 11 Cf. the final paragraph of Quine (1986c), pp. 430–431. But see also Hookway (1988), pp. 72–74 for the idea that this conception trivializes physics and that the trivialization can be resisted. 12 This seems to go against what Hookway says in Hookway (1988), p. 210. But Hookway is talking about theories, understood as having a coarser grain than theory formulations (see n. 9). What is not clear is whether it makes sense to assess theories, on this construal, as true or false. 13 See Quine (1986a), pp. 156–157 and Quine (1990a), §§41–42, pp. 95–101. Quine does not cite anything by Davidson in support of his attribution. But cf. Davidson (1984i), p. 237, where he seems to be taking for granted that a theory cannot be false if it is empirically warranted. Moreover, Davidson, unlike Hookway, is not talking about theories in the coarse-grained sense of the term (see n. 12). N.B. There are additional issues about interpretation raised here, in particular issues that arise in connection with assessing other people’s accounts. But these need not detain us now. See further section 3. 14 N.B. I understand the analytic/synthetic distinction to be a distinction between different kinds of true sentences. And by ‘sentences’ I mean what Quine calls ‘eternal sentences’, linguistic types that can be classified as true or false without reference to individual utterances of them (see e.g. Quine (1960), §40, pp. 191–195). I am not sure that Quine is entitled to think that there are such things. More to the point, I am not sure that Quine is entitled to think that there are such things given other things he thinks, in particular given his rejection of the analytic/synthetic distinction: see Moore (1996), §3.1; cf. Voss and Sayward (1976). But I shall let that pass, since my concern here is purely expository. 15 Quine (1961b). 16 See e.g. the apparent concessions at Quine (1960), p. 56 and Quine (1970a), p. 81. Dummett, whose discussion of Quine’s views we are about to consider, is a good example of someone who thinks that Quine has retracted some of his earlier more radical claims: see Dummett (1978e), esp. §1, pp. 375–384.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

21

That Quine is hostile to the analytic/synthetic distinction is well known. But does he deny its coherence? Michael Dummett, commenting on ‘Two Dogmas of Empiricism’, insists not. Dummett writes: In the last third of the article, Quine employs notions in terms of which it is quite straightforward to define ‘analytic’ and ‘synthetic’: in these terms, an analytic sentence is one such that no recalcitrant experience would lead us to withdraw our assignment to it of the value true, while a synthetic one is one such that any adequate revision prompted by certain recalcitrant experiences would involve our withdrawing an assignment to it of the value true. The position arrived at at the conclusion of the article is not in the least that there would be anything incorrect about such a characterisation of the notions of an analytic and a synthetic sentence, but simply, that these notions have no application: as thus defined, there are no analytic sentences, and there are no synthetic ones.17

Yes and no.18 Obviously Dummett, in this passage, is not construing sentences purely phonemically. If he were, there would need to be some explicit caveat to discount change of meaning. Otherwise it would be entirely trivial that ‘analytic’ and ‘synthetic’, thus defined, had no application. Dummett must therefore be construing sentences as having a semantic component—as being identified, in part, by their meaning. Now, consider a construal that would be acceptable to Quine. (There must be such a construal. Quine has persistently railed against uncritical talk of meaning, but he must be prepared to accept a notion of sentence-meaning adequate to support the notions of empirical and physical equivalence, and indeed those of truth and falsity.) On such a construal, Dummett is quite right: Quine’s position is that there are no analytic sentences and no synthetic ones, as defined. But this is not the end of the matter. For, on such a construal, the proposed definitions are not faithful to the traditional distinction. The traditional distinction presupposes a much more robust conception of meaning. Anyone who accepts the traditional distinction will admit that, given an analytic sentence, a Quinean story can be told in which this sentence is rejected in the face of recalcitrant experience: all that has happened, such a traditionalist will say, is that the sentence has undergone a change of meaning. Very well, consider a construal of sentences which, by incorporating this more robust conception of meaning, renders the proposed definitions acceptable to the traditionalist. Now Quine’s position is that the definitions are incoherent. The heart of the dispute between Quine and the traditionalist is in fact whether this more robust conception of meaning can be sustained.19 Quine, in rejecting the analytic/synthetic distinction, is ultimately rejecting this conception. His own rival conception is holistic. ‘The unit of empirical significance,’ he writes, ‘is the whole of science.’20 That is, it is not individual sentences, but suitably inclusive sets of sentences, that are compatible or incompatible with the evidence.21 Moreover, given any such set, and given any member of the set, it is always possible to remove that member and, through suitable further deletions and additions, to preserve empirical content. As it were, the organism can survive the loss of any one organ. There is no such 17 Dummett (1978e), p. 375. 18 Some of what I am about to say is anticipated by Dummett later in his discussion, ibid. pp. 411 ff. Cf. also Carnap (1990). 19 See esp. Quine (1961b), §5, pp. 37–42. 20 Ibid. p. 42. 21 Cf. Quine (1981d), p. 70.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

22

Language, World, and Limits

thing as ‘the’ set of sentences that must be true for any given evidence to hold, nor, conversely, subject to a qualification that I shall mention in section 2, as ‘the’ evidence that must hold for any given sentence to be true. While the truth of an individual sentence depends partly on its meaning and partly on the evidence, there is no way of separating out these two components.22 Much of this, that is much of what has been said here about the evidence, applies equally to the facts. For instance, it is equally true that what are compatible or incompatible with the facts are not individual sentences, but suitably inclusive sets of sentences. It is also true that, while the truth of an individual sentence depends partly on its meaning and partly on the facts, there is no way of separating out the two components. For the traditionalist, on the other hand, further resolution is possible. Associated with each sentence are those facts that would render it true and those facts that would render it false. This allows for the limiting case of a sentence that no facts would render false. Such a sentence is analytic.23 Any other true sentence is synthetic. As long as the analytic/synthetic distinction is understood in these terms, then Quine has to be seen as challenging its very coherence. There is a related distinction of degree which he accepts. This is a matter, roughly, of how much trouble it would be to change one’s mind about whether a given sentence was true. But Quine does not think it is possible to make sense of the clear-cut distinction that the traditionalist claims to have drawn.

1.3 Options for what to say if accepting the former distinction entails accepting the latter I shall argue that, if one accepts the underdetermination/indeterminacy distinction, then one must accept the analytic/synthetic distinction. I shall remain non-committal about what to make of this. All sorts of options are available, beyond leaving Quine to stew in his own juice. These options fall roughly into three categories. Those in the first category involve a modus ponens: to accept the former distinction, and thereby to accept the latter. Those in the second category involve a modus tollens: to reject the latter distinction, and thereby to reject the former. Those in the third category involve a reductio ad absurdum: to acknowledge that there are Quinean reasons for accepting the former distinction and rejecting the latter, which is what Quine wants to do, but thereby to abandon some other doctrine of Quine’s, perhaps, but not necessarily, one of the two doctrines under consideration. I shall say no more about options in the first category. But it is worth commenting briefly on options in the other two categories. Let us turn first to the ‘modus tollens’ options, which involve rejecting the underdetermination/indeterminacy distinction. These options themselves fall roughly into three sub-categories, corresponding to three different ways of rejecting the distinction. One can: maintain that in all the relevant cases there is a suitable fact of the matter; maintain that in none of the relevant cases is there a suitable fact of the matter; or give up talking in these terms. The options in the first of these sub-categories, insofar as they still presuppose a broadly Quinean framework, require physics to underpin any empirically underdetermined 22 Cf. Dummett’s references to what he calls Quine’s ‘inextricability thesis’, in Dummett (1978e), p. 387. 23 One thinks here of the familiar definition of ‘analytic’ as ‘true by virtue of meaning’.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

23

truth. Let us grant that truth about denotation is a case in point. It is not required that there be a definition of ‘denotes’ in fundamental physical terms. (That would be the reductionist demand renounced earlier.) What is required is that, given incompatible, empirically equivalent accounts of what the predicates in a given language denote, one true and one false, there should be recognizably physical facts that make the difference. Someone may say, in opposition to this, ‘But it is part of our very understanding of denotation that there are no such facts. The only candidates are facts about behaviour and dispositions to behaviour. But these still leave the matter undetermined.’ We can ask such a person, ‘What about other familiar physical facts, for example neurophysiological facts or facts about causal relations between uses of predicates and what they denote?’ If they balk at this, we can ask, ‘What about unfamiliar physical facts, as mooted above in connection with Cartesian souls?’ If they still balk, we can ask, ‘How sure are you that there is any truth about denotation, in other words that the word “denotes” is not just incoherent?’ If they are completely sure, then certainly they have reason not to embrace this kind of option. But, by now, this is a large ‘if ’.24 According to options in the second sub-category, factuality stops at the evidence. So if truth in physics is empirically underdetermined, then it, no less than any other empirically underdetermined truth, is indeterminate. Of course, options in these two sub-categories may turn out to be notational variants on one another. For that reason alone, somebody wanting to endorse the modus tollens may think it best to adopt an option in the third sub-category, dropping all talk of factuality: beyond the evidence, there is just whatever it takes to smooth over the rough edges, fill in the holes, round out the system—whatever is the most appropriate metaphor. There is an option in the third sub-category that is of interest because there is some evidence that Quine himself has been tempted by it. This is to deny that there is a clear-cut distinction between mere underdetermination and indeterminacy, but to concede enough of a distinction of degree to yield the ersatz analytic/synthetic distinction of degree that Quine has always accepted.25 Someone might say that this is not an option that Quine has been ‘tempted by’, it is what he has always thought, in other words that I am wrong to attribute to him a commitment to anything sharper. But think about it. Is Quine really going to concede that there is no difference, in clarity and distinctness, between his own belief in the indeterminacy of translation and Frege’s belief in the analyticity of arithmetic, or Kant’s in its syntheticity? Let us turn next to the third main category of options, those that involve a reductio ad absurdum. Dummett has argued that the thesis of the indeterminacy of translation, which, in the form in which Quine espouses it, presupposes the underdetermination/ indeterminacy distinction, entails that there is no analytic/synthetic distinction.26 So one option in the third category would be to combine Dummett’s argument with my argument to undermine the thesis of the indeterminacy of translation. Quite generally, any doctrine of Quine’s that requires both acceptance of the underdetermination/

24 For further discussion of the various issues raised here see Field (1972) and Friedman (1975), esp. §§III and IV, pp. 360–373. See also Rorty (1971–72), esp. p. 459. (Rorty is keen to endorse the modus tollens.) 25 See Quine (1986c), p. 430. 26 Dummett (1978e), esp. §7, pp. 409–416.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

24

Language, World, and Limits

indeterminacy distinction and rejection of the analytic/synthetic distinction27 can be undermined in this way.

2. Tension between Quine’s two doctrines concerning these distinctions This section is intended as a kind of warm-up. I shall try to lend support to my view of the connection between the two distinctions. Richard Rorty has said that the notion of ‘being about the world’, which is more or less equivalent to that of factuality, is ‘the notion . . . which the positivists used to explicate . . . “analytic”’.28 This is the sort of thing I have in mind. But complications abound. One complication that needs to be emphasized straight away is this. The kind of thing about which there may be no fact of the matter, strictly speaking, is not this or that phenomenon, or the truth of this or that sentence, or even the truth of this or that account. It is rather the choice between one account and another. It is sometimes said that Quine does not believe that there are any facts of the matter concerning denotation.29 This is wrong. Suppose that a particular predicate denotes rabbits. Then there is a fact of the matter concerning whether this predicate denotes rabbits. Had the facts been different in certain ways, the predicate would not have denoted rabbits. Indeed had the facts been different in certain ways, there would have been no predicate (for example, if there had been no language-users). The thing concerning which there is no fact of the matter, on Quine’s view, is whether this predicate, and the various other predicates in the language, denote what they denote as opposed to what they denote according to some empirically equivalent alternative account.30 It follows that we had better not try to ‘extricate’ the notion of factuality from that of indeterminacy and then try to foist upon Quine a distinction between those accounts that are, and those accounts that are not, ‘factual’, or ‘about the world’. Still less had we better do this at the level of individual sentences. That would be to beg all sorts of questions against Quine’s holism. If the underdetermination/indeterminacy distinction does yield the analytic/ synthetic distinction, as this abortive attempt would have shown, then, even so, it does not do so as quickly as that. It is clear, all the same, that no sharp semantic distinction is going to sit well with Quine’s holistic rejection of the analytic/synthetic distinction. Once it is conceded that no sentence, or virtually no sentence,31 has its own separable empirical content, nor its own separable physical content, but rather that sentences can work together in all sorts of syncretic ways to say how things are, then there is bound to be something 27 Consider here some of Rorty’s claims, also concerned with the thesis of the indeterminacy of translation, at the beginning of Rorty (1971–72), pp. 443 ff. 28 Ibid. p. 459. (Rorty of course means that the positivists explicated ‘analytic’ in terms of being about the world, not that they explicated it as being about the world. The latter is pretty much the reverse of what they did.) 29 Cf. McGinn (1984), p. 152. Cf. also Heal (1989), p. 35. N.B. Though I disagree with Heal on this, I think Heal’s book contains excellent material on Quine’s position. 30 This picks up on n. 7. See again the material cited there, esp. Quine (1986d), the paragraph straddling pp. 459 and 460. And cf. Quine (1990b). 31 I shall soon mention the qualification to which I referred in section 1.2.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

25

suspicious about recognizing some issues, but not all of them, as somehow impervious to physics. By what right do we say, given a choice between two rival accounts of a certain matter, that it is not the business of physicists to make sense of any resolution? After all, we might come to accept the truth of a sentence that explicitly casts the issue in physical terms, that is in terms of whatever fundamental states are recognized in physics. Here it is important to remember that Quine’s physicalism is not reductionist (see section 1.1). The distinction between mere underdetermination and indeterminacy is certainly not aligned to any distinction between what is, and what is not, already cast in physical terms. Nor, importantly, does our acknowledging that there is a fact of some matter require prior insight into how the matter is related to physics. On the contrary, it may be our acknowledging that there is a fact of this matter which leads to our acknowledging that there is such a relation—that there is work here for physicists to do.32 But how do we come by the former knowledge? How can we know that the choice between two biological theories is merely underdetermined, whereas the choice between two manuals of translation is indeterminate? I am not saying that there are no answers to these questions. I am saying that there are no answers that do not invite restoration of the analytic/synthetic distinction. Consider a case of indeterminacy. Of two incompatible and physically equivalent accounts one is true. How do we tell which? Presumably by appeal to such considerations as those of elegance, simplicity, and utility. These considerations also help us in choices that are merely underdetermined. That is partly constitutive of how we view the facts. In a case of indeterminacy, however, the facts are irrelevant. It is as if our exercise of these considerations itself determines the truth.33 In choosing one of the accounts as the truth, we are smoothing over the rough edges in whatever way we think fit, and our choice has nothing to answer to. But then are we not involved in a piece of linguistic legislation that is of precisely the kind that traditionalists would say sustains analyticity? Are we not creating pockets of truth which, unlike truth that depends partly on meaning and partly on the facts, depends solely on meaning? Here is a tell-tale sign. If, having made our choice, we come across others who have made the opposite choice, we are debarred from saying that they are wrong. For there is nothing (there is no fact) for them to be wrong about. We have to say instead that they have conferred different meanings on some of the terms involved. And this is just the kind of thing that a traditionalist would say when talking about an apparent denial of some analyticity (see section 1.2).34 But is there any one sentence in this case that Quine has to treat as analytic? I think there is. More precisely, I think there is a sentence S that satisfies the following condition: if we grant Quine his holistic conception of meaning and a correspondingly pragmatic conception of what is involved in telling whether a given sentence is true 32 Cf. again the material cited in n. 11. 33 There are interesting comparisons and contrasts with what Hilbert said in connection with the use of ideal elements in mathematics: see Hilbert (1967). Cf. again the material by Quine cited in n. 6. 34 For discussion of issues closely related to those in this paragraph see Field (1975). On the uncomfortability for Quine of distinguishing between choices that amount to linguistic legislation and choices that are answerable to the facts, cf. Quine (1961b), pp. 44–46. Cf. also Quine (1981g), pp. 93–94. For a related argument that Quine has problems here see Rorty (1971–72), pp. 453–454.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

26

Language, World, and Limits

or false, then, by his own reckoning, S cannot be rejected as false without undergoing a change of meaning, on what must be a more robust conception of meaning. (Hence Quine is subject to a kind of reductio ad absurdum.) Call the two alternative accounts in the envisaged case I1 and I2. These are to be thought of as individual sentences, long and complex enough to have their own empirical, and indeed physical, content and thus to be candidates for both empirical and physical equivalence. (This is the qualification to which I referred in section 1.2. I do not know whether Quine is entitled to admit such sentences. But he does.35 This device will incidentally play no rôle in my main argument, in the next section.) Let I1 be the truth. Then the sentence S in question—the sentence that I think Quine has to regard as analytic—is I2 → I1.36 To reject this sentence as false would be to reject I1 in favour of I2. And indeed, given Quine’s holistic conception of meaning, we can imagine circumstances in which that is precisely what we would do. These would not be circumstances in which we encountered evidence incompatible with I1. Such evidence would be incompatible with I2 as well, given that the two accounts are empirically equivalent. Rather they would be circumstances in which we felt, in the light of new evidence, that I2 was after all the more elegant and the more serviceable alternative (say). Now a similar thing could happen in a case of mere underdetermination. Consider such a case, and call the two alternative accounts U1 and U2. Let U1 be the truth. Then, as before, we can imagine circumstances in which we would reject U1 in favour of U2. For example, suppose that, according to U1, the universe does not have a centre, whereas, according to U2, it does (the accounts have compensatingly different geometries).37 Then we can imagine a steady accumulation of evidence that conferred privileged status on one particular point in the universe and convinced us that U2 was, as in the indeterminacy case, the more elegant and the more serviceable alternative. Indeed if I1 and I2 are two accounts of what the predicates in a given language denote, and if, according to I2, there is a predicate in the language that denotes whatever is at the centre of the universe, but not according to I1, then we could well be imagining the same evidence in both cases.38 But the point is this. There is, for Quine, a crucial difference between the two cases. Going over from I1 to I2 is unlike going over from U1 to U2, or equivalently rejecting I2 → I1 is unlike rejecting U2 → U1, in that it cannot be regarded as a change of view. There is nothing for it to be a change of view about. So it has to be regarded as a change of meaning.39 35 See Quine (1986c), p. 427 and Quine (1986e), p. 620. One of my concerns is that Quine may be begging important questions about finite axiomatizability (if I may use a precise term rather loosely). Note that in the first of these passages Quine also grants empirical content to what he calls observation sentences. I think this is a departure from the position of Quine (1961b), but I shall not argue the point here. 36 → is to be understood purely truth-functionally. (Note that the example could just as well have been, simply, ¬I2 . But I think I2 → I1 will illustrate the argument more vividly.) 37 Cf. again the passage cited above, n. 4. 38 Cf. how Quine argues from the underdetermination of physical theory to the indeterminacy of translation in Quine (1970b). For a further example of how the indeterminacy case and the case of mere underdetermination could interact in this way, imagine evidence that persuaded us to interpret other beings as manipulating us in various ways—I am assuming the indeterminacy of interpretation—and that thereby led us to adopt what had hitherto seemed just a crazy sceptical alternative to our view of the facts. 39 Here and elsewhere in this essay I am prescinding from all complications concerning the meaning of →: a change of logic would clearly add further problems, but we can afford to ignore these.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

27

It has to be regarded as a change of meaning because, even for Quine, there are two things that make a true sentence true: meaning and fact. To be sure, Quine has familiar reservations about saying what this comes to. These reservations are a direct product of his holistic conception of meaning. In particular, as I emphasized in section 1.2, he denies that there is any way of separating out the contribution that meaning and fact each make to the truth of a sentence. Relatedly, he denies that the meaning of a true sentence is some clearly delineated thing, and that the fact of the matter is some other clearly delineated thing to which it corresponds. But so long as the idea that meaning and fact conspire to make true sentences true is freed of these misconceptions, Quine has no quarrel with it.40 Indeed the idea has the status, as Quine himself suggests, of a platitude. But granted the platitude, and granted that going over from I1 to I2, or equivalently rejecting I2 → I1, cannot be regarded as a change of view about the facts, then it must be regarded as a change of meaning. The platitude leaves no alternative. However, although this is something that Quine himself is committed to, the conception of meaning in question cannot be the original Quinean conception. It must be something more robust. For on the original Quinean conception, the move from I1 to I2 would certainly be in accord with meaning (that is, it would not involve any change of meaning). Besides, nothing in the original Quinean conception distinguishes the two cases. Precisely the same pragmatic considerations would guide the move from I1 to I2 as would guide the move from U1 to U2. Quine must therefore resile. He is driven to regard I2 → I1 in just the way in which a traditionalist would regard an analyticity. I2 → I1 is an isolated example, however, and a very artificial one at that. I am still short of establishing that the underdetermination/indeterminacy distinction yields the analytic/synthetic distinction as a general, principled distinction. There is nothing yet to show that, given Quine’s rejection of the latter, he has as much reason to reject the former as he does, say, the distinction between philosophy and science, or Carnap’s distinction between internal and external questions.41 That is what I shall now argue.

3. Incompatibility between Quine’s two doctrines concerning these distinctions My strategy will be to try to generalize from key features of the example just considered. This will again mean trying to subject Quine to a kind of reductio ad absurdum. As before, the starting point will be to grant Quine his holistic conception of meaning, together with all that it entails, plus the underdetermination/ indeterminacy distinction, together with all that it entails. Then, in terms wholly acceptable to Quine, an exhaustive and exclusive distinction will be drawn between two kinds of true sentences. This distinction, I shall argue, corresponds to the traditional 40 See Quine (1970a), pp. 1–3. 41 Carnap (1956). See Quine (1961b), pp. 43–44 and Quine (1960), §56, pp. 270–276. Cf. in this connection various strands in Hylton (1982). (My essay overlaps with Hylton’s at several points, though there are also significant points of divergence.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

28

Language, World, and Limits

analytic/synthetic distinction with its attendant and more robust conception of meaning. The idea that I need to capture is that there are, among true sentences, those whose rejection, with due account taken of the rejection of other sentences, can never just be attributed to a difference of view about the facts. It must be attributed, at least in part, to a difference of meaning, on what must be the more robust conception of meaning. These are the analytic sentences. Furthermore, I need to capture this idea using purely Quinean resources, and in completely general terms, not just with respect to a cluster of specially constructed examples. I shall begin with some minor technicalities. I have been talking in this essay about ‘accounts’. But what exactly are these? Sets of sentences? No; not if an account is something according to which a sentence can be false. For it does not in general make sense to say that a given sentence is false ‘according to’ a given set of sentences (all presumed true). This is because, even if the given set contains the negation of the given sentence, that counts for nothing when, in suitably holistic vein, we are envisaging revisions to our logic whereby both a sentence and its negation may be true. An account must be more than just a set of sentences then. It is, in fact, more useful here to introduce a somewhat broader concept. Let us define an evaluation as an assignment, to every sentence in our language, of a truth value, or, more formally, as a function from the set of sentences in our language into {⊤, ⊥}. I stipulate all sentences to ensure that no evaluation lacks content through treating of too few sentences; also, relatedly, to ensure that it always makes sense to describe two evaluations as empirically or physically equivalent. But I do not include any closure requirements. Nor do I include any requirements of consistency. In both cases this is for the reason just alluded to: we are accepting the holistic principle that even our logic may be revised. Because of this lack of restrictions, some evaluations, indeed most evaluations, though extensive enough to have content, nevertheless do not. Or, as I shall rather say, they have ‘null’ content. They are incompatible with all possible evidence. It is a trivial feature of evaluations with null content that they are all empirically and physically equivalent.42 Let me emphasize a point that is related to something I said in section 1.1. All the sentences concerned here are sentences in our language, unreconstrued. There is no question of having to decide whether a term that we use is being used by others in some unfamiliar way. No doubt, to build on an example of Quine’s, if ever we come across people who accept exactly the same sentences as we do, but with the words ‘molecule’ and ‘electron’ switched, we shall say that their use of these two words is the reverse of ours rather than that they are mistaken.43 Even so, construed as we would construe it, their account of the world contains numerous falsehoods, by our reckoning. 42 The definition of incompatibility between evaluations is the same as the definition of incompatibility between accounts given at the beginning of section 1.1: two evaluations are incompatible if there is a sentence that is true according to one and false according to the other. This might seem to involve an illegitimate presupposition of our own logic. Will there not be some evaluations according to which some sentences are both true and false? Indeed there will. But the definition of incompatibility begs no questions. And our aversion to incompatibility—our resolve not to accept two incompatible evaluations—remains as part of the naturalism. 43 Quine (1981b), pp. 28–29.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

29

In the context in which Quine introduces this example, he argues that this false account of the world is nevertheless empirically equivalent to ours. I suspect that this indicates a significant concession to atomism. On his original holistic conception, I suspect, the account comes out as having null content. But I do not need to argue for that now. For now, my point is simply that we are to consider such accounts as they are, not as we might charitably suppose them to be.44 Remaining faithful to Quine’s holism, and exploiting the notion of the physical that goes with the underdetermination/indeterminacy distinction, we might now try the following definition of analyticity: a sentence is analytic if it is true, and if any empirically warranted evaluation in which it is false is physically equivalent to the truth. (Now that the notion of an evaluation is to hand, ‘the truth’ can be thought of as simply that evaluation which assigns truth to all and only true sentences.) Equivalently, and somewhat more formally: (D1) S is analytic = Def. (i) S is true, and (ii) any empirically warranted evaluation in which S is false is physically warranted. The idea behind this definition would be as follows. Clause (ii) says that, so long as an arbitrary evaluation v in which S is false is empirically warranted, then it cannot misrepresent the facts. In other words, given that S is in fact true, the choice between v and the truth must be indeterminate. In other words again, rejecting the truth in favour of v cannot be regarded as a change of view. It must be regarded as a change of meaning, specifically a change of meaning in S. This seems to be a suitable generalization of what we found with the artefact considered in section 2. In fact, however, Quine’s holism entails that no sentence fits the bill. Take any true sentence S. Then, given the holism (and given the underdetermination of physics), there are bound to be some empirically warranted evaluations in which S is false and which, because they differ from the truth in various other ways, are factually incorrect, that is to say not physically warranted. Clause (ii) is far too strong. How then should it be weakened? One intuitively attractive suggestion is to restrict attention to those evaluations that differ from the truth only with respect to the truth value of S. Only? Well, not strictly of course. They must also differ with respect to the truth values of all the other sentences that have to be re-evaluated in order to preserve empirical content. But therein lies the hitch. The holism means that there are no such things as ‘the’ other sentences that have to be re-evaluated in order to preserve empirical content. All sorts of empirically content-preserving changes can accommodate the falsity of S.—But perhaps one of these changes can be singled out as minimally disruptive?—There is no reason to think so. There is no reason even to think that there are uniquely reasonable criteria of minimal disruption. However, suppose there are. Even then, there is no reason to think that any interesting distinction turns on whether or not the resultant evaluation is physically warranted. Perhaps it never is. Perhaps it always is. 44 For further material relevant to these issues see Quine (1969d), §5, pp. 308–311. I am assuming that ‘language has settled the sentences and what they mean’ (see p. 309). The reason why I think that Quine has made a concession to atomism is related to what I said in n. 35: his argument here presupposes that same notion of observationality. N.B.  The idea of an evaluation is nicely underpinned by Quine’s notions of assent and dissent: see Quine (1960), pp. 29 ff.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

30

Language, World, and Limits

Reflection on this last possibility gives the lie to another natural suggestion for weakening clause (ii): to relax the universal quantification to an existential quantification. That gives: (D2) S is analytic = Def. (i) S is true, and (ii) some empirically warranted evaluation in which S is false is physically warranted. (Actually, since physical warranty entails empirical warranty, the phrase ‘empirically warranted’ in the new clause (ii) can be deleted.) For a true sentence to satisfy this definition is for it to be possible to reject the sentence without a change of view, that is for there to be a way of rejecting the sentence that must, willy-nilly, be regarded as a change of meaning. The problem now is that the new clause (ii) is liable to be too weak. Granted Quine’s holism, every true sentence is liable to fit the bill: every true sentence is liable to be rejectable without a change of view, provided that enough compensatory re-assignments are made elsewhere to cast the change as one that must, willy-nilly, be regarded as a change of meaning. I say ‘liable to’ because there is nothing in the underdetermination/indeterminacy distinction itself to prevent Quine from saying that certain privileged truths cannot be rejected without a change of view. He may, as far as the underdetermination/indeterminacy distinction goes, say this about any truth that does not involve such words as ‘denotes’, or other similarly troublesome vocabulary. He may indeed say it about any truth that is not one of the small cluster of artificial cases: I2 → I1 and its ilk. The problem with this is that it is tantamount to saying that the truths in question are synthetic. It sets them apart from the artefacts (and perhaps from other sentences too) as lacking that which makes the latter analytic. And even if Quine is committed to nothing more than the relatively innocuous claim that all but a tiny selection of grotesquely artificial truths are synthetic, that is bad enough for him. Quine’s hostility has always been to the analytic/ synthetic distinction itself, not to any particular view about where it lies. (This is why he has always disowned the view that every true sentence is synthetic.45) So I shall take for granted that the new clause (ii) is too weak, and that, granted Quine’s holism, every true sentence does indeed satisfy it. Incidentally, the objection to (D1) and (D2) is not just that they are indiscriminate. The objection is that they are indiscriminate in a way that violates their own rationale. True, we know that (D1) must be wrong because of the artefacts: we know that there are some sentences that Quine is committed to regarding as analytic. But that does not indicate what is wrong with (D1). What is wrong with (D1) is that, although it specifies a sufficient condition for S’s being analytic, this condition cannot possibly be necessary given how much scope there is, even without flouting any evidence, for rejecting S at the same time as misrepresenting the facts. Similarly, what is wrong with (D2) is that it specifies a condition that cannot possibly be sufficient given how much scope there is for rejecting S at the same time as correctly representing the facts. What we are witnessing here is a basic consequence of Quine’s holism: namely, how large the range of empirically warranted evaluations is, and how much, therefore, is demanded by universal quantification over the range, how little by existential quantification. This prompts the following thought. Perhaps there is a more suitable 45 See e.g. Quine (1970a), pp. 98 ff.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

31

definition which involves mixed quantification, a definition in which clause (ii) takes the following form: ‘for any empirically warranted evaluation v there is an empirically warranted evaluation v* such that – S – v – v*’. I think this suggestion is right. In fact, I think that mixed quantification of this kind is precisely what is required to capture the idea that the rejection of an analyticity always involves some change of meaning. (I shall have more to say about this below.) However, I shall deviate from the suggestion in one respect. I shall propose a definition that involves quantification over all evaluations, not just empirically warranted evaluations. This is because I believe that what would make the definition work for evaluations that are empirically warranted will make it work also for evaluations with any other empirical content. Here is the definition: (D3a) S is analytic = Def. (i) S is true, and (ii) any evaluation in which S is false is physically equivalent to one in which S is true. And of course: (D3b)

S is synthetic = Def. (i) S is true, and (ii) S is not analytic.

There are two things to note about this definition straight away. First, the mixed quantification ensures that it will never be possible, simply by adducing a single evaluation, either to prove or to disprove that a given true sentence is analytic by this definition. The only way to secure such a proof, or such a disproof, would be by reflection on the contribution that the sentence makes to the content of any arbitrary evaluation. It is as if the definition forces us, in spite of any holism, to think about sentences in a neo-atomistic way. The second thing to note is the rôle of physical equivalence in the definition, or at least one aspect of that rôle. Consider: if the expression ‘is physically equivalent to’ in (D3a) were replaced by ‘is empirically equivalent to’, then, trivially on a holistic view, every true sentence would satisfy the specified condition for analyticity. If it were replaced by ‘is identical to’, then, even more trivially, none would. So it is only if physical equivalence lies between empirical equivalence and identity that the definition stands a chance of effecting an interesting distinction. But, as I observed in section 1.1, to say that physical equivalence lies between empirical equivalence and identity is, in effect, to affirm the underdetermination/ indeterminacy distinction. Hence, if I am right to think that the distinction that the definition effects, on that basis, is the traditional analytic/synthetic distinction, then we are shown very graphically how the one distinction yields the other. But is not (D3a) vulnerable to pretty much the same objection as (D2)? Take any true sentence S and any evaluation v in which S is false. If v has null content, then it can trivially be transformed into an evaluation in which S is true, without change of physical content—without change of view. If, on the other hand, v does not have null content, then are we not still guaranteed that a view-preserving re-assignment to S is possible, provided that enough compensatory re-assignments are made elsewhere to cast the change as one that must, willy-nilly, be regarded as a change of meaning? I think not. The objection to (D2) was based on the idea that any true sentence, be it a highly general scientific formula, or a detailed account of various local goings-on, or something else again, could be rejected in holistic vein as a result of what a traditionalist would regard as a change of meaning. But that simple idea cannot be

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

32

Language, World, and Limits

adapted without further ado to the situation under consideration here—where v, the evaluation in which S is false, is already at one remove from the truth. Let us suppose, to allow the objection its most favourable case, that v is empirically warranted. Even so, the physical content of v, underdetermined as it is by its empirical content, is liable to differ radically from the physical content of the truth: v might tell a story in which the very structure and constitution of the universe are of some fundamentally alien kind, or in which the most bizarre sceptical possibilities are realized, so that there is no longer even any locus for those various local goings-on. Given that all the sentences being evaluated are sentences in our language, as we understand them, there is no reason whatsoever to suppose that the very same physical content can attach to an evaluation in which S is restored to its original status as true. It may well be that, by the time other compensatory re-assignments have been made in order to re-establish even that much overlap with the truth, some of v’s eccentric physical content will automatically have been lost. This highlights the basic idea behind (D3a). In applying the definition to any true sentence, we confront the following question: what can guarantee that the rejection of this sentence is never vital to telling some particular story, a story according to which the facts are thus rather than so? The only answer, as far as I can see, is: the sentence’s being true in a way that is impervious to the facts, however they may be; in other words, the sentence’s being true by virtue of meaning, on the traditionalist’s conception of meaning. Any holistically inspired rejection of the sentence has then to be seen as involving a change of meaning. And whatever story is told by means of the rejection, there is a meaning-preserving way of telling the same story that allows the sentence to remain true. An example may help to clarify this. Consider a sentence that would traditionally be regarded as analytic, say: (1)

All bachelors are unmarried.

And consider a sentence that would traditionally be regarded as synthetic, say: (2)

All bachelors are less than ten feet tall.

Suppose that (D3a) and (D3b) are in line with tradition here. In other words, suppose that, according to (D3a), (1) is analytic while, according to (D3b), (2) is synthetic. How so? There are all sorts of evaluations in which each of them is false, some of which have quite different physical content from the truth. For example, in the case of (1), there is an evaluation v1 in which each of the following sentences is true: Among men, the correlation between being misogynistic and being unmarried is very high Among men, all and only bachelors are misogynists A tiny number of married men are misogynists and in which (1) is accordingly false. In the case of (2), there is an evaluation v2 according to which the universe has some bizarre geometry, empirically indistinguishable from our own, whereby things rapidly increase in size as they move away from the

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

33

centre of the earth, and in which (2) is false because of those bachelors currently in aeroplanes. What makes (1) analytic, according to (D3a), is that its rejection is incidental to the physical content of any evaluation in which it is false, and thus, for example, to the physical content of v1: exactly the same content attaches to at least one evaluation, v1* say, in which (1) is true. Intuitively, the only way to explain this is by appeal to the fact that the rejection of (1) involves a change of meaning. Any story told by means of that rejection could just as well have been told without the change of meaning. Thus it may be that, in v1*, the sentence ‘Among men, all and only bachelors are misogynists’ is false. The difference between v1 and v*1 is basically a matter of whether the application of ‘bachelor’ is tied to being unmarried or to being misogynistic. There is no fact of the matter to determine which of them is the truth. It does not devolve on physicists to make sense of the issue’s being resolved one way rather than the other. (1) is analytic, according to (D3a), because any rejection of it generates just such an indeterminacy. By contrast, what makes (2) synthetic, according to (D3b), is that its rejection is vital to the physical content of at least one evaluation in which (2) is false. Let us suppose that v2 is a case in point. Intuitively, this is because the story v2 tells is so bizarre that there is no way of telling that story without rejecting (2). In order for an evaluation v2* to make (2) come out true, and to have the same physical content as v2, it must differ in so many other compensatory ways from v2 that it cannot, after all, have the same physical content as v2. That is, in order to ensure that all bachelors are less than ten feet tall, v2* must stay sufficiently close to the truth to prevent itself from spinning the same wild geometrical yarn as v2. And the only way to explain this is by appeal to the fact that v2 does not involve any change of meaning in (2): (2) has its own factual content, repudiation of which cannot always be masked by suitable compensatory meaning changes that enable it still to come out true. (D3a), I suggest, is a suitable generalization of what we found with the sentence I2 → I1 constructed in section 2. This sentence, to recapitulate, was so constructed that its rejection could not be regarded as a change of view. The choice between I1 and I2, on which any rejection of I2 → I1 would have to turn, was indeterminate. That is, concerning whether I1 was true as opposed to I2, there was no fact of the matter. The assignment of falsehood to I2 → I1, therefore, would never be crucial to the physical content of an evaluation. If, in a given evaluation, I2 → I1 were false, there would always be a physically equivalent evaluation in which it was true. It is this feature that has now been culled from the example and identified as the distinguishing feature, among truths, of analyticities. (D1) and (D2) already involved generalization from features of the example in section 2. But they failed because they inappropriately fastened on one peculiarity of the example, or rather, on one peculiarity of one particular extension of the example, namely that in which the subsequent rejection of I2 → I1 involved only change of meaning.46 (D1) picked out truths whose rejection always involved only change of meaning. This proved to be too demanding: any sentence could be rejected as a result of a change of meaning and of view. (D2) picked out sentences whose rejection sometimes involved only change of meaning. This proved to be not demanding enough: any sentence could be rejected as a result of a change of meaning alone. What was 46 As I indicated in section 2, rejection of I2 → I1 could certainly involve a change of view as well. Cf. n. 38.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

34

Language, World, and Limits

required, as I indicated, was a definition that picked out truths whose rejection always involved some change of meaning. But to capture the force of the ‘some’, in holistic terms, it was necessary to look to relations between evaluations—in particular, the relation of physical equivalence which holds between evaluations whose differences are, precisely, differences of meaning. Hence (D3a). Of course, the conception of meaning involved here is the traditionalist’s conception of meaning which emerges from the definitions, not the holistic conception of meaning which is fed into them. It is the notion of the factual implicit in the underdetermination/indeterminacy distinction which combines with the latter to produce the former. The facts are thought of as going beyond the evidence, but not far enough to fix an answer to every question that we can raise. So it becomes a real issue, what contribution the facts make to determining the truth, or certain parts of the truth; and, holism notwithstanding, we can sometimes separate the contribution made by the facts from the contribution made by meaning. This must, however, be a more robust conception of meaning than is involved merely in giving the truth its empirical warrant. That is, it must be a more robust conception of meaning than Quine’s holistic conception. It is this that is captured in the proposed definition of analyticity. ‘But how exactly,’ someone will ask, ‘are we supposed to apply the definition? Consider a problematic case such as Quine’s “Everything green is extended”.47 Is every evaluation in which this is false physically equivalent to one in which it is true?’ I have no idea. Nor, for that matter, do I have any idea whether ‘All bachelors are unmarried’ satisfies this condition, or whether ‘All bachelors are less than ten feet tall’ fails to. I earlier made these proposals, in order to see how the definition worked if they were true. But I made no effort to show that they were true. What I have tried to argue in this essay is that asking whether a given sentence satisfies the specified condition is the same thing as asking whether it is analytic, on the traditional conception. But I offer no independent criterion for telling whether a sentence does satisfy the condition. And I hold no special brief for the traditional conception. I am free to endorse the modus tollens from section 1.3. However, that is by the by. My concern has been to establish an incompatibility between two of Quine’s doctrines.

Appendix My argument in this essay has depended on attribution to Quine of the following view. There can be truth and falsity even where there is indeterminacy. In other words, of two incompatible accounts, each of which is compatible with the facts, one can be true and the other false. Truth outstrips factuality. ‘Small wonder,’ it will be said, ‘that Quine seems to be committed to the existence of analytic truths. Surely he has been misrepresented. Surely, when he says that there is no fact of a given matter, he means that there is no truth of the matter. The choice between the two relevant accounts (as it may be, the choice between two manuals of translation) is supposed to be a practical affair, like the choice between driving on the left and driving on the right. Truth and falsity do not come into it.’ 47 Quine (1961b), p. 32.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

35

I do not know of anything in Quine’s writing that decides between this alternative interpretation and mine.48 Later in this appendix I shall argue that it does not ultimately matter: if I am wrong on this exegetical point, only the letter of my argument is affected, not its spirit. But there is plenty to suggest that Quine is at least committed to the view that I have attributed to him, even if he does not explicitly endorse it. First, observe that the sentences that are at stake in cases of indeterminacy are, on narrowly grammatical criteria, declarative sentences. (‘“Rabbit” denotes rabbits’ is an example.)49 There are passages in which Quine suggests that, given his disquotational theory of truth, such grammatical criteria are decisive for truth-aptitude. Here is a revealing passage on Austin: [Suppose] we think of truth in terms of Tarski’s paradigm. The paradigm works for evaluations . . . as well as for statements of fact. And it works equally well for performatives. ‘Slander is evil’ is true if and only if slander is evil, and ‘I bid you good morning’ is true of us on a given occasion if and only if, on that occasion, I bid you good morning. A performative is a notable sort of utterance, I grant; it makes itself true; but then it is true. There are good reasons for contrasting and comparing performatives and statements of fact, but an animus against the true/false fetish is not one of them.50

Still, it might be protested that the grammatical criteria are decisive only in the case of sentences that make sense. Quine cannot be expected to acknowledge something as true or false just because of its grammatical form, if it incorporates an expression that is meaningless. He cannot, for instance, be expected to acknowledge something as true or false if it incorporates the expression ‘is synonymous with’. Will he not say that the sentences that are at stake in cases of indeterminacy do not make sense? ‘Making sense’ here is not straightforward. There is a danger of question-begging. The sentences in question have a perfectly acceptable use, even by Quine’s lights. (They have a use in a way in which sentences about synonymy, or indeed sentences of phlogiston theory, do not.) There are plenty of places in which Quine himself is seen using them.51 Not only that. To accept the disquotational theory of truth is itself to endorse the use of such sentences. For ‘“Snow is white” is true if and only if snow is white’ is a case in point: there are physically equivalent theories of truth on which this sentence is rejected.52 But perhaps Quine is taking for granted a suppressed relativization to the homophonic manual of translation, which restores determinacy? There are two points to be made in reply to this. First, it is unclear, in Quine’s own terms, whether any such relativization does restore determinacy.53 Secondly, Quine allows that sentences of this kind can be used without relativization as a way, simply, 48 The closest I can find to evidence decisive in favour of the alternative interpretation is Quine’s enthusiastic reception of Gibson (1986), where Gibson glosses ‘indeterminate’ as ‘neither true nor false’ (p. 152). 49 N.B. Throughout this appendix ‘sentence’ must be understood in a more relaxed sense than in the main text (see n. 14). Otherwise sentences will be by definition truth-apt, and the argument will be begged in favour of my interpretation. 50 Quine (1981e), p. 90. (But contrast Quine (1970a), p. 10.) 51 See e.g. the passage from Quine (1990b), cited in n. 30 and Quine (1981a), p. 20, for affirmations that ‘rabbit’ denotes rabbits. 52 There is a fascinating discussion of possible internal tension here in Boghossian (1990). 53 See Davidson (1984i).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

36

Language, World, and Limits

of ‘opting’ for a particular manual.54 What we need is an explanation of what it is to ‘opt’ for a manual, which (i) does justice to the fact that opting for a manual involves using declarative sentences, (ii) shows how opting for a manual nevertheless falls short of stating truths, and (iii) does not beg the principal question. Of course, an account can be provided which does beg the principal question. It can be said that truth stops where the facts stop. But then we are back at square one. Does Quine perhaps think that the sentences involved in cases of indeterminacy are truth-apt but neither true nor false? There is certainly a gap between being truth-apt and being either true or false, provided we reject bivalence. But can that which is truth-apt and which has the kind of use that these sentences have fall short of being true?55 Quine is in any case wedded to bivalence. He has always been attracted to bivalence by its power, its simplicity, and its familiarity. He is well aware of the pressures that bear on it. In particular he acknowledges some pressure coming from indeterminacy of various kinds. But he denies that either rules the other out. Here is another revealing passage: Bivalence requires us . . . to view each general term . . . as true or false of objects even in the absence of what we in our bivalent way are prepared to recognize as objective fact. At this point . . . the creative element in theory building may be felt to be getting out of hand, and second thoughts on bivalence may arise. For those of us who are inclined still to rest with bivalence for its undeniable merits, this heightened awareness of the props that sustain it can still be salutary.56

There are three further points that I wish to make before resting my case in favour of my interpretation of Quine. First, this interpretation provides by far the simplest gloss on what is meant by incompatibility in cases of indeterminacy. What is meant by incompatibility in such cases is precisely what is meant by incompatibility in cases of underdetermination: two accounts are incompatible when there is a sentence that is true according to one and false according to the other.57 Other glosses on what is meant by incompatibility are available, apt for other interpretations. But they are, by comparison, ad hoc. Secondly, although the examples we have focused on have been examples concerning truth and denotation, they are not the only examples. Remember that Quine also thinks that there is indeterminacy in the higher reaches of mathematics.58 To say that there is no truth or falsity in these matters would not be at all in keeping with his views about the enmeshment of mathematics with the rest of science. 54 See the passage from Quine (1990a), cited in n. 2. 55 Perhaps it can, by being false. Is it possible that Quine thinks all the sentences in question are false? No. In some cases there are alternatives which, granted truth-aptitude, are subcontraries, for instance that a given predicate denotes rabbits and that it does not denote rabbits. 56 Quine (1981g), pp. 94–95. Cf. Quine (1970a), pp. 83–86 and Quine (1987a). My opponent may say that the indeterminacy with which Quine is concerned in this passage, illustrated by the example of the table, is very different in kind from the indeterminacy with which we are concerned. Is it? Let a mable be a table with an extra molecule. Then cannot Quine’s worry in this passage be captured more or less as follows? There is no fact of the matter concerning whether ‘table’ denotes tables rather than mables. 57 If we understand incompatibility in the same way in cases of both kinds, then we lend considerable force to the underdetermination/indeterminacy distinction. 58 See the passages cited in n. 6. (This connects with his claim, in the second passage, that factuality comes in degrees, something particularly hard to accommodate on the alternative interpretation.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Two Distinctions in Quine

37

My third point is simply that the view that I am attributing to Quine is by no means absurd. It may have seemed absurd. We tend to hear ‘It is a fact that . . . ’ as equivalent to ‘It is true that . . . ’. But once we concede that there is more to be said about the factual, then we need not shrink from the idea that, given our linguistic resources, we can frame questions that are not settled by the facts and need to be settled instead by fiat. Thus, for example, it is true that electrons have negative charge, and that positrons have positive charge; even so, there is no fact of the matter to determine that it should be this way round.59 But the most important point that I want to make in this appendix is this. Even if I am wrong on the exegetical issue, I still have an argument to indicate internal tension in Quine. More specifically, I have an argument to show that, given the underdetermination/indeterminacy distinction, there is a distinction, among declarative sentences that are ‘acceptable’ (in a suitably neutral phrase), between those whose rejection must involve a change of meaning and those whose rejection need not. (I2 → I1 is an example of the first kind, provided that → is allowed to figure in such contexts; ‘“Rabbit” does not denote any animal that weighs more than 100 pounds’ is an example of the second kind, as of course is ‘No rabbit weighs more than 100 pounds’.) This is no longer a distinction among truths. But I do not see why it should be any the more congenial to Quine for that. Indeed on some conceptions—not on what I have been calling the traditional conception, but on conceptions that are traditional enough—the distinction’s not being a distinction among truths brings it that much closer to the analytic/synthetic distinction. Friedrich Waismann, who viewed sentences such as ‘Red and green exclude each other’ as ‘rules of grammar’, and who argued for the ‘autonomy of grammar’, wrote, ‘If we give the rules of . . . grammar we are not making any assertion,’ and again, ‘What misleads us, when we are looking for justification for rules, is thinking that what we want are proofs of their truth.’60 There is a related point. On this alternative interpretation, Quine must believe each of the following about opting for a manual of translation: it can consist in the use of declarative sentences; it does not consist in stating truths; it provides a framework within which there is room for a distinction between what is and what is not in accordance with the facts.61 Does this not bring him too close for what he ought to regard as his own comfort to Carnap with his doctrine of external questions? Consider the following quotation from Carnap. [The settling of external questions] . . . does not need any theoretical justification because it does not imply any assertion of reality. We may still speak . . . of ‘the acceptance of the new entities’ . . . [But this] must not be interpreted as referring to an assumption, belief, or assertion of ‘the reality of entities’. There is no such assertion. An alleged statement of the reality of the 59 Of course, put like that, the view appears un-Quinean. Just so. Here, I think, we are seeing a manifestation of precisely the tension in Quine’s views which I claim to have located. 60 Waismann (1965), pp. 35 and 40 respectively, his emphasis. (The example concerning red and green is given on p. 137.) Waismann’s position is, of course, essentially Wittgensteinian. 61 See again n. 7. We must not forget that, despite any indeterminacy, an assignment of denotations to predicates still has to answer to the facts. An entirely homophonic assignment would do so. An assignment for English that was entirely homophonic save that it included the sentence ‘“Rabbit” denotes helicopters’ would not. Opting for the homophonic assignment for English would enable one to use that sentence as a way of misrepresenting the facts.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

38

Language, World, and Limits

system of entities is a pseudo-statement without cognitive content. To be sure, we have to face at this point an important question; but it is a practical, not a theoretical question; it is the question of whether or not to accept the new (entities). . . . The acceptance cannot be judged as being either true or false. . . . 62

I am not saying that, on this alternative interpretation, Quine is committed to Carnap’s doctrine. But I do think that the arguments that persuade Quine to reject Carnap’s doctrine put considerable pressure on the view currently being attributed to him. Suppose, to revert one last time to the example from section 2, that we have accepted I1. If that provides us with the possibility of saying things that are in violation of the facts (see n. 61), then why not admit that those things are false? And if those things are false, then why not be done with and admit that I1 is true?63

62 Carnap (1956), p. 214. 63 I should like to thank Bill Brewer, Daniel Isaacson, Philip Percival, two anonymous referees for Erkenntnis, and especially Alexander George for their extremely helpful comments on earlier drafts of this essay.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

3 What are these Familiar Words Doing Here? Abstract This essay is concerned with six linguistic moves that we commonly make, each of which is considered in turn. These are: stating rules of representation; representing things categorically; mentioning expressions; saying truly or falsely how things are; saying vaguely how things are; and stating rules of rules of representation. A common-sense view is defended of what is involved in our doing each of these six things against a much more sceptical view emanating from the idea that linguistic behaviour is fundamentally messy. Both the fifth move and the sixth move involve vague concepts, and much of the essay is concerned with developing an approach to various problems and puzzles that attach to such concepts, most notably the sorites paradoxes.

My title is a quotation from Davidson’s essay ‘On Saying That’.1 And although my concerns are at some remove from his, they do connect at one significant point. We (non-philosophers as well as philosophers) find ourselves under the continual pressure of theory to deny that ordinary familiar semantic features of ordinary familiar words equip them to serve certain ordinary familiar functions. One of Davidson’s aims is to resist that pressure as far as the function of reporting indirect speech is concerned.2 In similar vein I want to look at some common things that we do with words and show how we can hold fast to a simple common-sense view of what we are doing despite the doubts to which reflection is apt to give rise. In fact I want to look at six things that we do with words, six linguistic moves that we make.3 These six moves are related in a number of important ways. Even so, they are really the subjects of six separate essays (six separate sketchy essays at that), and I am well aware that treating them together in the way that I shall be doing—worse still, trying thereby to make some headway with solving one or two associated philosophical problems, as I shall also be doing—will mean that in each case I can at best produce something highly programmatic. The theoretical pressure that particularly concerns me comes from a certain compelling picture of linguistic behaviour, which I will call the Governing Picture. The Governing Picture: 1 Davidson (1984c), p. 94. 2 See esp. ibid. p. 108. 3 This way of putting it is an allusion to Wittgenstein (1967), Pt. I, §§7 and 22.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

40

Language, World, and Limits

Linguistic behaviour is messy. It is a vast baroque labyrinthine edifice, constituted by millions upon millions of diverse but connected episodes, sometimes differing crucially but imperceptibly from one another, each dependent for its significance in all sorts of subtle and indeterminate ways on context, between them putting language to an unsurveyable variety of uses, and between them subject to constant processes of essentially unpredictable evolution which combine to give them a corporate life of their own, beyond the reckoning of any individual speaker.4 This picture is pretty much undeniable, certainly undenied, even by those who entertain a rigidly formalistic conception of the mechanisms that underlie all these messy surface phenomena. (Remember Wittgenstein’s claim in the Tractatus that ‘everyday language is part of the human organism and is no less complicated than it.’5) But there are those who use the Governing Picture to argue as follows. The meaning of a word is never, or hardly ever, something precisely defined that regulates applications of the word. Rather it is something that develops along with applications of the word, as they in turn both exploit and extend the possibilities that it affords. It has an open-ended dynamic. But this means that talk of ‘linguistic moves’, with all its connotations of bounded rule-governed games, is quite inappropriate. There is not one piece of linguistic behaviour, except in highly stylized contexts such as the actual playing of a game, that has its own delineated significance in isolation from the sprawling spawning socio-linguistic history to which it belongs; not one that should be thought of on the model of bidding Three No Trumps, or castling.6 To many who argue in this way I think Murdoch’s admonitory reply is apt. She writes: Here truism, half-truth, and shameless metaphysics join to deceive us. Yes, of course language is a huge transcendent structure, stretching infinitely far away out of our sight, and yes, when we reflect, we realize that often we cannot say quite what we mean or do not quite know what we mean. Common-sense does not usually take the trouble to reflect as far as this, or if it has done so realizes that nothing is really being changed and meaning and truth are what they have always seemed.7

We do ordinarily think of ourselves as making various linguistic moves: as describing, requesting, commanding, greeting, thanking, and the like. True, the Governing Picture should serve to remind us that making these moves is subject to all sorts of indeterminacy; that it is not always rigidly bound by rules; that language is versatile enough for sentences with the same surface grammar to be used to make quite different moves, indeed, very often, for the self-same sentence to be used to make quite different moves; and that the stock of moves we can make is not something clearly circumscribed, but is changing and expanding all the time. But we need not feel any pressure from the Governing Picture to renounce talk of linguistic moves altogether. 4 For endorsement of the Governing Picture, along with many interesting references, see Standish (1992), Ch. 2: see esp. the quotation from Taylor (1985), p. 231, which Standish gives on p. 74. 5 Wittgenstein (1961), 4.002. Cf. Davidson (1980), esp. p. 123 and §H; and Davidson (1984j). 6 There is something of this line of thought in Derrida: see e.g. Derrida (1982). 7 Murdoch (1993), pp. 188–189.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

41

In saying these things I take myself to be following a more or less Wittgensteinian line.8 Yet some of those who argue in the way I indicated above take themselves to be doing the same. They see Wittgenstein as likewise recoiling from (what they see as) the hopeless artificiality of dividing our use of sentences into different linguistic moves, and, at least as far as our use of declarative sentences is concerned, as acquiescing rather in its ‘homogeneity’: always simply to say how things are.9 In fact the matter is complex. Wittgenstein says things that fit with both readings.10 The mistake is to try to extract some single pithy thesis from what he says. Wittgenstein is not offering us a thesis. He is trying to clear away confusions that he has discerned. There are certainly elements in what he says that should make us wary of comparing saying hello with bidding Three No Trumps. He has warnings against various models of our linguistic practices that we adopt. But the point is not to spurn the models, still less of course to spurn them in favour of equally flawed alternatives. The point is simply to heed the warnings. There is nothing in Wittgenstein, it seems to me, to suggest that we cannot realistically think of ourselves as making all sorts of linguistic moves. Nor is there anything in the Governing Picture to suggest this. I am certain of that. In fact, however, that is not what I want to argue in this essay. My concern is not with the idea that we cannot realistically think of ourselves as making all sorts of linguistic moves. My concern is rather with the idea that there are some particular linguistic moves, the six to which I have alluded, such that we cannot realistically think of ourselves as making them. What I want to argue is that there is nothing in the Governing Picture to suggest that this is the case.

1. Stating rules of representation The appearance of ambivalence in Wittgenstein that I have just been talking about brings us nicely to the first of these six moves, which is this. We sometimes state rules of representation. For instance, we say, ‘Aunts are female’, meaning thereby that we are not to count somebody as an aunt unless we also count that person as female. Aunts have to be female. Why do I associate this particular move with the appearance of ambivalence in Wittgenstein? Well, there are many places in which Wittgenstein himself alludes to this move, emphasizing the ways in which it differs from making an empirical claim about how things are, for instance when he distinguishes between giving the criteria for a disease and giving its symptoms.11 On the other hand the distinction between making this move and making a true empirical claim about how things are is very 8 See e.g. Wittgenstein (1967), Pt. I, §§23–24, and p. 224, third and fourth paragraphs from the bottom. Concerning the point about sentences with the same surface grammar being used to make quite different moves, cf. the volte-face between the first and second editions of Hacker’s Insight and Illusion, respectively Hacker (1972) and Hacker (1986); see in particular p. ix of the latter. 9 See e.g. Lovibond (1983), §§6 ff. For a backlash, somewhat more in keeping with my reading, see Blackburn (1990). A third possible position, of course, is to see in this divergence of interpretation fuel for cynicism, either about Wittgensteinian exegesis or indeed about Wittgenstein. 10 In support of the opposed reading see e.g. Wittgenstein (1967), Pt I, §402—cited by Lovibond in Lovibond (1983), p. 26—and Wittgenstein (1969a), p. 25. 11 E.g. Wittgenstein (1969a), pp. 24–25.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

42

Language, World, and Limits

reminiscent of the distinction between uttering an analytic truth and uttering a synthetic truth, and there is much in Quine’s famous assault on the latter12 that is in turn very reminiscent of Wittgenstein. Thus Wittgenstein is every bit as suspicious as Quine is of the idea of Platonically conceived meanings, attaching to our words by dint of our linguistic behaviour and determining, by themselves, that certain assertions are true—‘true in virtue of meaning’—while others need the corroboration of experience for their truth.13 Furthermore he shares Quine’s sense of the constant erosion of our use of words over time, an erosion brought about by a variety of pragmatic forces working away against forces of conservatism, whereby the very sentences that we use at one time to say how things must be we may later find ourselves using to say how they are not.14 For now, I simply register this apparent tension in Wittgenstein’s thinking. In due course I hope it will be clear that it is only apparent. Quine, meanwhile, thinks that we do best to treat all declarative sentences, or more strictly all utterances of declarative sentences, as homogeneous assertions about how things are: insofar as there is anything like an analytic/synthetic distinction to be drawn, then it is simply a matter of how likely we are to retract our assertions when we subsequently discover that things are not how, in making those assertions, or better in making that whole body of assertions, we took them to be. Quine seems precisely to be using the Governing Picture, or at least one part of it—that language use evolves in fluid and unpredictable ways—to challenge the idea that we ever state rules of representation. We say, ‘Aunts are female’, simply because that currently looks like a correct thing to say. We are not thereby legislating for what to say in the future, irrespective of how our view of things may change. There is no question that aunts have to be female. Thus Quine. But now consider. Is Quine saying simply that there is no analytic/ synthetic distinction, or is he denying its very coherence? Dummett, commenting on Quine’s essay ‘Two Dogmas of Empiricism’,15 insists on the former. He writes: In the last third of the article, Quine employs notions in terms of which it is quite straightforward to define ‘analytic’ and ‘synthetic’: in these terms, an analytic sentence is one such that no recalcitrant experience would lead us to withdraw our assignment to it of the value true, while a synthetic one is one such that any adequate revision prompted by certain recalcitrant experiences would involve our withdrawing an assignment to it of the value true. The position arrived at at the conclusion of the article is not in the least that there would be anything incorrect about such a characterization of the notions of an analytic and a synthetic sentence, but simply, that these notions have no application: as thus defined, there are no analytic sentences, and there are no synthetic ones.16

12 The locus classicus is Quine (1961b). 13 See e.g. Wittgenstein (1974), p. 54. 14 See e.g. Wittgenstein (1969b), §§96–99. Indeed such thoughts are present in the very place where he draws the distinction between giving criteria and giving symptoms: see again Wittgenstein (1969a), p. 25. 15 Quine (1961b). 16 Dummett (1978e), p. 375. Note that, because Dummett couches his discussion in terms of sentences rather than in terms of their utterances, we must presume that attention is being restricted to sentences that can be classified as true or false without reference to individual utterances of them—sentences of the kind that Quine elsewhere calls ‘eternal’ (Quine (1960), §40).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

43

Yes and no.17 Obviously Dummett, in this passage, is not construing sentences purely phonemically. If he were, there would need to be some explicit caveat to discount change of meaning. Otherwise it would be entirely trivial that ‘analytic’ and ‘synthetic’, thus defined, had no application. Dummett must therefore be construing sentences as having a semantic component, as being identified, in part, by their meaning. Now on any construal that would be acceptable to Quine, Dummett is quite right: Quine’s position is that there are no analytic sentences and no synthetic ones, as defined. But this is not the end of the matter. For, on any such construal, the proposed definitions are not faithful to the distinction as traditionally conceived. As traditionally conceived, the distinction presupposes a much more robust conception of meaning than would be acceptable to Quine. On the traditional conception, given any analytic sentence, recalcitrant experience can certainly lead to its rejection in the way that Quine envisages; it is just that the sentence will thereby have undergone a change of meaning.18 Very well; suppose that the proposed definitions are suitably reconstrued, so as to bring them into line with the traditional conception. Now Quine’s position is that they are incoherent.19 What Quine is really attacking, then, is the more robust conception of meaning. This is a conception whereby meanings are clearly discriminable monadic entities that stand to words in something like the relation of exhibits to labels,20 the Platonic conception to which I alluded earlier and to which Wittgenstein is equally hostile. It seems to me that Quine is absolutely right to attack this conception. Certainly there is much in the Governing Picture to challenge it. But what does it have to do with the idea that we sometimes state rules of representation? There is simply no obvious connection.21 Can we not adopt a rule whereby it is incorrect to apply one word, ‘aunt’, and at the same time to deny application of another, ‘female’, without this in any sense requiring the existence of meanings as clearly discriminable monadic entities, without its rendering anything we say ‘true in virtue of meaning’, indeed without its even precluding a natural evolution in our language use whereby we later 17 Some of what I am about to say is anticipated by Dummett (1978e), later in his discussion: see pp. 411 ff. Cf. also Carnap (1990). 18 So too, on the traditional conception, given any synthetic sentence that is held true, and given any recalcitrant experience, the sentence can continue to be held true vis-à-vis that experience, in the way that Quine envisages; but again, sometimes, only by undergoing a change of meaning. If meaning is kept fixed, then there will never be any choice about whether or not a sentence that is held true should be rejected vis-à-vis any possible recalcitrant experience: if the sentence is analytic, it never should be; if the sentence is synthetic, it sometimes should be. This is why, on the traditional conception, the distinction is both exclusive and exhaustive, a fact that Dummett’s discussion somewhat obscures. 19 I have borrowed material in this paragraph from Essay 2 in this volume. Note that in his (1961b), Quine does describe the word ‘analytic’ as ‘un-understood’ (p. 34) and later relates the idea of analyticity to another idea which he describes as ‘nonsense, and the root of much nonsense,’ (p. 42): I am grateful to Alexander George for drawing my attention to these two passages. 20 Cf. Quine (1969a), p. 27. 21 Cf. Quine’s concession in his (1961b), pp. 25–26, that we sometimes stipulatively define novel terms, and that, whenever we do, ‘we have a really transparent case of synonymy’: I am grateful to Timothy Williamson for drawing my attention to this passage. (Not that I want to claim that Quine would be happy with everything I say in this section. Stating rules of representation covers far more than stipulatively defining novel terms. When I talk about ‘what Quine is really attacking’, I do not mean this exegetically. I am making a point about the force of his arguments.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

44

Language, World, and Limits

allow talk of ‘non-female aunts’? (If we do later allow talk of ‘non-female aunts’, this will simply mean that we are no longer abiding by the same rules.22)23

2. Representing things categorically The second move we sometimes make is to represent things categorically, by which I mean in a way that involves neither systematic context-dependence nor implicit relativization of any kind.24 (These, where they occur, indicate that the representation is from a particular point of view.) It is an aspiration of physicists, I believe, to make this move as extensively as they can.25 For it is a working presupposition of physicists that the most fundamental physical laws look the same from every point of view. Thus consider Newton’s first law of mechanics: a body continues in a state of rest or of uniform motion in a straight line unless it is compelled to change that state by forces acting upon it.26 This statement of the law holds only relative to an inertial frame. The earth, for example, is not an inertial frame. With the earth as frame, the sun describes something approximating to a colossal circle once every twenty-four hours even though there are no relevant forces acting upon it. It was in large part because of his dissatisfaction with the suppressed relativization in this statement of the law, and the partiality which this in turn implied, that Einstein was impelled to look for something more universal, and eventually to formulate his general theory of relativity.27 However, the Governing Picture seems to preclude our representing things categorically—in pretty much the same way as it seemed to preclude our stating rules of representation. (This is quite apart from the fact that it expressly includes the idea that episodes of linguistic behaviour depend for their significance in all sorts of subtle ways on context, which some people would say should already give pause.) The point is this. Granted the Governing Picture, it seems impossible for the meaning of a word to be anything apart from its continued usage, so that any word has, at any stage in its history, different possibilities of further meaning-preserving use woven into it. The use of a word can always be continued in different ways, for different purposes, to different effects. It has what was called earlier ‘an open-ended dynamic’. Thus, for example, there is no legislating in advance for the success of metaphors, which may be contrived to describe situations completely unlike anything that anyone has ever 22 Will it also mean that we no longer have our current concept of an aunt? Or will it mean that our current concept of an aunt has undergone a change?—When, some time in the fifteenth century, the Pawn was first allowed to move forward two squares in chess, did this create a new game, what we now call chess? Or did that very game undergo a change?—It is relatively clear what is going on in these cases. Say what you will as long as you do nothing to threaten such clarity. 23 Cf. again Wittgenstein (1969b), §§96–99—esp. §98, where he denies that his famous river-bed analogy makes logic ‘an empirical science’. And cf. Wittgenstein (1967), Pt. I, §354. For some extremely helpful material on the ideas in this section see various essays in Arrington and Glock (1996), esp. Hacker (1996b) and Hookway (1996), the former of which is an abbreviated version of the even more helpful Hacker (1996a), Ch. 7. 24 Cf. n. 16: precisely what an ‘eternal’ sentence is is a sentence that equips us to make this move. 25 Cf. Wittgenstein (1967), Pt. I, §410. 26 Newton (1947), Bk. I, Law I. 27 See Einstein (1960), pp. 11, 61, 71–72, and 99. For more on the aim to transcend perspective (of this and any other kind), and for an argument that this aim is achievable, see my (1997), esp. Chs. 1–4.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

45

encountered before and which may then give way to new, previously unimagined, literal uses: consider, for instance, the smooth adaptation of the word ‘hear’ to cover what we do to somebody’s voice over the telephone. But it seems to follow that there is no representing things except in a way that depends at the very least on temporal context. For no application of a word can be understood except as occurring at a particular stage in its development. Consider the following sentence: (1) Earshot of somebody is the distance within which it is possible to hear his or her voice. This was once more or less definitional. In the terms of the last section, it would once have been used to state a rule of representation. If someone were to utter the sentence now, on the other hand, then they would be saying something at best false. Yet it seems unsatisfactory, in accounting for this, just to say that (1), or more particularly the word ‘hear’, has undergone a change of meaning. That would be far too simplistic. As intimated above, it seems better to say that the meaning of the word ‘hear’ has evolved to accommodate its current usage. Well, yes; it is certainly unsatisfactory just to say that the word ‘hear’ has undergone a change of meaning.28 But some change has occurred. To pick up the theme of the last section: we are no longer abiding by the same rules. And this means, in particular, that we cannot draw conclusions about the categoricity of any of our representations by comparing current uses of the word ‘hear’ with erstwhile uses of it. To say that someone has represented things categorically is not to deny that the sentence they have used to do so might also be used, in another context, metaphorically perhaps, to say something quite different. Nor is it to deny that various processes of attrition and accretion might eventually ensure that the sentence can no longer effectively be used except to say something quite different. The point is only that their sentence is free of the sort of systematic context-dependence that attends the use of, say, the word ‘now’; the sort of context-dependence that would need to appear in an account of the semantics of the word. They have represented things in a way that is, at least in these narrowly semantic terms, from no point of view.

3. Mentioning expressions The third move we sometimes make is to employ expressions in such a way as to make those very expressions our subject matter. In the standard terminology, we sometimes mention expressions. This is in contrast to the more usual way of employing expressions, which, again in the standard terminology, is to use them: when we use expressions, in this quasi-technical sense, our subject matter (if we have one) is not them, but something determined by their semantics. There are a number of conventional devices for mentioning expressions. The commonest of these is the use of inverted commas. Thus, whereas cats have four legs, ‘cats’—note the singular verb coming up—has four letters. 28 Cf. n. 22: even if allowing the Pawn to move forward two squares did create a new game, it would be unsatisfactory just to say that chess was invented in the fifteenth century.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

46

Language, World, and Limits

What I have just proffered is basically a grammatical characterization of the distinction between using expressions and mentioning them. A more pragmatic characterization, it seems, would be this. Using expressions involves employing them in a way that exploits whatever meaning they have, so as to draw attention to something determined, in part, by that meaning; mentioning expressions involves employing them in a way that waives whatever meaning they have, so as to draw attention to the expressions themselves. Now one issue is how far it is appropriate to talk of ‘linguistic moves’ in this connection. Such talk grates a little. Neither mentioning an expression nor using an expression seems relevantly like stating a rule of representation, or representing things categorically, or thanking someone. Each seems to be something that one can do in the course of doing one of these other things. Thus one can thank someone and thereby mention an expression (‘Thank you for calling me “Sir”’). Or one can thank someone and thereby do nothing but use expressions (‘Thank you for not calling me by name’). Little of substance hangs on this, however. Our focus is the distinction itself— between using expressions and mentioning them—be the propriety of talk of ‘linguistic moves’ in this connection as it may. That said, I do not myself find such talk unduly dissonant. It is like mating in chess. One can castle, or move one’s bishop, and thereby mate one’s opponent. Or one can castle or move one’s bishop without so much as checking one’s opponent. Yet it does not seem inappropriate to say that mating is a move that one can make in chess—precisely in making some other move. The far more urgent issue, in this context, is this. The Governing Picture casts doubt on whether the distinction between using expressions and mentioning them can even be drawn. For given the Governing Picture, the grammatical characterization and the pragmatic characterization signally fail to accord with each other. According to the former, the distinction is a clear-cut one with clearly recognizable grammatical criteria of application. According to the latter—at least if the Governing Picture is correct—the question whether one is using an expression or mentioning it on any given occasion (as with any other question about what one is drawing attention to by means of an expression on any given occasion) is a complex, vague, and unruly matter that depends in all manner of ways on the particular circumstances: it certainly does not depend in any straightforward way on one’s use of devices such as inverted commas. Thus consider the following two sentences, and the underlined expressions that occur in them: (2) Christopher can never remember his nine-times table; he always says that eight nines are seventy-four (3) The only word for this is ‘preposterous’. On the grammatical characterization, the underlined expression in a (typical) use of (2) would be used, and that in a (typical) use of (3) mentioned. But on the pragmatic characterization, the reverse would be true. Again, consider this sentence: (4) Albert, who remembers virtually nothing of the physics he once knew, does remember that electrons have negative charge.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

47

On the grammatical characterization, ‘negative’ would be used in a (typical) use of (4). The pragmatic characterization, on the other hand, yields no clear verdict at all. On the pragmatic characterization, (4) illustrates how using an expression and mentioning it can merge imperceptibly into each other—the distinction is one of degree— whereas on the grammatical characterization, the distinction is one of kind.29 Does the Governing Picture entail the disintegration of the distinction then, and with it the illegitimacy of the idea that we ever mention expressions? Not at all. What it entails is that the two characterizations should be kept apart; and that, if we are going to talk about mentioning expressions, then we should be clear about which of the two characterizations we are operating with.30 But both characterizations give clear content to the idea that we sometimes mention expressions. (And both characterizations, come to that, give important content to this idea. The grammatical characterization, for example, allows us to say that one of the greatest intellectual achievements of the last century, namely Gödel’s proof that arithmetic cannot be consistently and completely axiomatized, would not have been possible without due appreciation of the distinction between using expressions and mentioning them.31)32

4. Saying truly or falsely how things are The fourth move we sometimes make is quite simply to say, truly or falsely, how things are (where the truth or falsity of what we say is taken to be one of the defining characteristics of our making this move). It may seem fantastic to include this move on my list. Could anybody really think that the Governing Picture poses any threat to the idea of our doing anything as basic as this? Well, if truth and falsity are understood in even a moderately ambitious way, then our philosophical heritage, extending back to Plato and beyond, is in fact replete with challenges to this idea, based on features of linguistic behaviour highlighted in the Governing Picture.33 However, these challenges raise metaphysical issues that are not really my current concern. I have in 29 Another notable example would be a use of the sentence ‘Aunts are female’ to state a rule of representation. On the grammatical characterization, this would involve using both ‘aunts’ and ‘female’. On the pragmatic characterization, it would involve mentioning them both (roughly—though it is a very nice question how roughly (cf. Baker and Hacker (1985), Ch. VI, §3(ii))). 30 This is a main theme of Essay 1 in this volume, which contains a fuller discussion of these issues. 31 Gödel (1967), e.g. p. 601. See also Quine (1987b), p. 84. Indeed I think the grammatical characterization allows us to go further and say that one of the great achievements, in turn, of analytic philosophy is to have made due appreciation of the distinction possible. Even the most rigorous writings in mathematics often flout it: cf. Quine (1987d), p. 232. (For an indication of the significance of the distinction on the pragmatic characterization, see Essay 1 in this volume.) 32 It is interesting at this point to consider Derrida again. His attitude to the distinction between using expressions and mentioning them is a curiously ambivalent one. He often seems to make play with words, and indeed to make philosophical points, precisely by flouting the distinction: see e.g. Derrida (1982), pp. 320–321. (This is a complaint that Searle levels against him in his commentary on this essay, Searle (1977), p. 203.) Furthermore there are places where Derrida seems to be overtly hostile to the distinction. Cf. his comment, ‘I try to place myself at a certain point at which . . . the thing signified is no longer easily separable from the signifier’, quoted in Wood and Bernasconi (1988), p. 88. Cf. also Derrida (1987), pp. 97 ff. However, in his reply to Searle—Derrida (1988)—he writes (p. 81), ‘I agree that [the confusion of “use” and “mention”] might very well be [a radical evil].’ 33 For a tiny sample, see: Plato (1961b), 183a4–b8; Nietzsche (1967), §616; and Derrida (1981), p. 168.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

48

Language, World, and Limits

mind much less heady worries about the idea of our making this move, worries based on the countless ways, again highlighted by the Governing Picture, in which, when we use a declarative sentence that equips us to make this move, the conditions might nevertheless not be right for us to do so: we end up not saying anything true or false at all. Examples include cases of reference failure, as when someone says, ‘That dagger is covered in blood’, and is hallucinating. They also include cases of what Travis calls natural isostheneia,34 as when—this is Austin’s delicious example35—someone says, ‘He is not at home’, and the person referred to is lying upstairs dead. Anyone who has anti-realist qualms about the law of the excluded middle36 might also want to include all those cases in which we have no procedure for telling, even in principle and even with some margin of error, whether what has been said is true or false. (Anti-realism challenges the idea that something can be true or false though we have no procedure for telling which.37) One such case might be someone’s saying, ‘Descartes would have loved Marmite.’ But of course, the mere fact that there are countless ways in which we can use a declarative sentence and fail to say something true or false does not, on its own, show that there are not also ways in which we can use a declarative sentence to succeed in saying something true or false. So long as there is no reason to think that known impediments to our making this move are somehow symptomatic of unknown impediments to our doing so, there does not yet seem to be any threat to the idea that we sometimes—indeed, often—say truly or falsely how things are.38 What then are we to say about someone who utters a declarative sentence in an attempt to make this move, though the circumstances are in fact unsuitable? Just that: that this is someone who has uttered a declarative sentence in an attempt to make the move, but, because the circumstances are unsuitable, their attempt is a failed attempt. It is not that they have made the move in a way that is somehow deficient. Rather they have not made the move at all.39

34 Travis (1998), which includes an extensive (and superb) discussion of such cases. 35 Austin (1970), p. 128. 36 See e.g. Dummett (1978a); and Dummett (1993). 37 I do not say that anti-realism challenges the idea that something can be true though we have no procedure for telling that it is. That way of putting it (though often found in the writings of anti-realists themselves) saddles anti-realism with the paradoxical consequence that there can be no unknown truths: see Fitch (1963). The version in the main text does not: see Melia (1991). (Our having a procedure for telling whether x is true or false is to be understood as allowing for the following possibility: that, although x is in fact true, the actual carrying out of the procedure would render it false (and would accordingly put us in a position to tell that it was false).) 38 For different but related reservations about the idea of our making this move (in fact of our making any move), based on suspicion of the very contrast between situations in which there are impediments to our doing so and situations in which there are not, see Derrida (1982). It may be that Derrida is insufficiently open to the possibility that one and the same sentence can be used in one situation to make one move and in another situation, inimical to the making of that move, to make another. 39 Is this idea perhaps negotiable? Could we eliminate the category of failed attempts to make this move, by simply extending our notion of falsity and assimilating all such cases to cases in which something false has been said (cf. Dummett (1978a))? For reasons why we may not be able to do this see again Travis (1998).—Note: there is also the question, on which Strawson is sometimes alleged to have equivocated in Strawson (1993), whether someone who utters a declarative sentence in an unsuccessful attempt to make this move thereby makes a statement that is neither true nor false or fails to make a statement at all: see e.g. Nerlich (1965). That seems to me to be an unimportant point of terminology.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

49

Herein, I think, is a clue as to how to respect anti-realist qualms about the law of the excluded middle without surrendering to them. What we can do is to hold fast to the law, as part of what constitutes the very making of this particular linguistic move;40 then, when someone utters a declarative sentence where we have no procedure for telling whether they have said something true or false, to say simply that they have failed to make the move. For if they have failed to make the move, then questions about the law of the excluded middle, in respect of what they have done, do not so much as arise—any more than such questions arise in respect of orders given or oaths expressed. In particular, what this person has done need not incline us either to abandon the law of the excluded middle or even to have reservations about re-affirming it. I realize, of course, that there is far more to be said about this. Some of it I have tried to say elsewhere.41 For now, I am content merely to advert to this way of holding fast to the law of the excluded middle even while insisting that nothing can be true or false without our having some procedure for telling which. I think it has the potential to defuse a number of reactionary worries about anti-realism.42 But does it not also entail that we cannot always tell whether someone has made this move? For surely there are times when, even though we do not know of any procedure for telling whether someone has said something true or false, we cannot rule out the possibility that there is one. This question betrays a misunderstanding. By our having a procedure for telling whether someone has said something true or false, I mean our knowing of such a procedure. Very well; suppose someone says, ‘Descartes would have loved Marmite.’ And suppose that archaeological-cum-technological advances eventually put us in a position, currently beyond our ken, to have a decent stab at ascertaining whether or not Descartes would have loved Marmite. Suppose, finally, that someone then utters the sentence anew. (This is, of course, the sort of possibility that the Governing Picture puts us in mind of.) Are we really to say that the first of these utterances is a failed attempt to say something true or false, but the second a successful attempt? Surely the advances in question give the first utterance (retrospectively) as much title to the claim of being a successful attempt as the second. Or if we are to distinguish between the two utterances in this way, are we also to accept that what is said on the second occasion depends on the nature of the advances and may yet differ from what is said on some third occasion when further advances allow us to address the question even more efficiently? That seems very counterintuitive.43 These concerns are reminiscent of concerns expressed above in sections 1 and 2. And my response is effectively as it was before. Certainly it is unsatisfactory just to say 40 Cf. Wittgenstein (1967), Pt. I, §136. 41 Moore (1997), Ch. 10, §4: see in particular the discussion of what I call ‘partial realism’, pp. 245–249. For a fascinating discussion that in effect raises problems for this position, see Williamson (1994b). Williamson ends his essay by asking whether these problems constitute a reductio ad absurdum of the position with which he has been concerned in the essay, namely intuitionism (p. 143). We can also ask whether they constitute a reductio ad absurdum of partial realism: I think not. 42 For a superb discussion of related issues in connection with Wittgenstein, see Hacker (1986), Ch. XI, §4. 43 For discussion of such questions see Diamond (1999).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

50

Language, World, and Limits

that the different utterances of this sentence differ in these ways, especially granted the evident constancy in the sentence that makes the differences possible. But this is not to say that there are not these differences. It is to say only that the full semantic story does not begin and end with them.

5. Saying vaguely how things are The fifth move is a special case of the fourth (where I mean ‘special’ only in the sense of species and genus: instances of the fifth move all but exhaust instances of the fourth). We sometimes say, truly or falsely and vaguely, how things are. That is, we sometimes make the fourth move by using a sentence which, in some contexts, cannot be used to say anything clearly true or clearly false—or clearly neither. Call such a sentence a vague sentence. And call an utterance of a vague sentence, in such a context, a vague utterance. Examples are: the sentence, ‘You are a child’; and an utterance of it addressed to a fourteen-year-old, in the absence of anything serving to hone its sense (such as a context-specific stipulation).44 How is the Governing Picture relevant here? Principally in serving to remind us that there are vague sentences, and that there are vague utterances of them. For unless we can give a coherent account of these, then we risk having to admit that there is no coherent account of non-vague utterances of vague sentences (say, an utterance of ‘You are a child’, addressed to a five-year-old). This is partly because of the infamous sorites paradoxes,45 which I shall discuss in the next section, and partly because of the fact that what counts as a vague utterance of a vague sentence is itself, of course, vague. The very idea of our saying how things are by using vague sentences is under threat. Now it seems to me that there are all sorts of things that might reasonably be said about vague utterances, each more or less appropriate in any given case. One of these is that the truth of the utterance is secured by the very fact of its being made. This allows for the possibility that a subsequent utterance of the negation of the same sentence, in the same context, or in a relevantly similar context, should also count as true. This is not incoherent. The idea is that there is a degree of freedom in the use of vague sentences whereby vague utterances of them can function somewhat like performatives, or somewhat like jurors’ verdicts: they can be true, in a way, by virtue of being made.46 (‘You could say she’s a child; you could say she’s not a child. It’s up to you.’47) 44 The phenomenon of vagueness has spawned a massive literature. Pre-eminent is Williamson (1994a). An excellent collection is Keefe and Smith (1997). Each of these contains extensive bibliographies. 45 The sorites paradoxes are a family of paradoxes modelled on the following, from which their name derives (the Greek adjective ‘sorites’ corresponds to the noun ‘soros’, meaning ‘heap’). One grain of sand does not make a heap; for any number n, if n grains of sand do not make a heap, then n + 1 grains of sand do not make a heap; therefore, there is no number of grains of sand that make a heap. 46 Austin, in his (1975), introduces the notion of a ‘performative’ by means of examples: these include an utterance of ‘I do’ in the course of a marriage ceremony, and an utterance of ‘I bet you sixpence it will rain tomorrow’ (pp. 5–6). He expressly denies that performatives are either true or false (p. 6). Others have subsequently adopted his notion, but have amended it so as to grant performatives truth: on the amended account, performatives make themselves true (see e.g. Quine (1981e), esp. p. 90). In invoking performatives here, I am obviously presupposing this amended account. 47 Cf. Sainsbury (1997), §6.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

51

The Governing Picture, in reminding us of the huge variety of ways in which sentences can work, opens us, or should open us, to the possibility of this sort of latitude. Another thing that might be said about a vague utterance, and that might indeed be said when the ‘performative’ tag just suggested seems appropriate but the utterer’s endorsement cannot be said to carry its usual authority, is that the utterance is neither true nor false; that it is a failed attempt to say something true or false, of the sort considered in the previous section. (‘Look, you haven’t said a single true thing about her! I agree, you could say she’s a child. But you said that only because you thought she was seven. In fact she’s fourteen. So that doesn’t count.’) The same thing might also be said when the ‘performative’ tag seems inappropriate and what matters is simply to register that the utterance is vague. (‘You predicted that the next person to come into the room would be a child. But actually this is a borderline case. Your prediction wasn’t really correct, and it wasn’t really incorrect either.’) Would this be coherent? A familiar argument due to Williamson purports to establish that it would not be.48 Williamson’s argument, if successful, shows that calling a vague utterance neither true nor false would commit one to a contradiction. (In the example given, the contradiction would be that the person in question was neither a child nor not a child—or perhaps, if it were a case where the ‘performative’ tag seemed appropriate, that the person in question was neither what the utterer would then and there call a child nor not what the utterer would then and there call a child.) I shall not rehearse Williamson’s argument in full here. All that matters for my purposes is that it rests on the assumption that the utterance concerned ‘says something’. Can this assumption be resisted? Consider a case of reference failure. Think again about the case in which an hallucination victim says, ‘That dagger is covered in blood.’ There is a sense in which their utterance says something. There is a sense in which it says that some dagger is covered in blood. But there is also a sense in which it does not. In particular, of course, granted that the utterance is neither true nor false, it does not say anything in any sense that requires us to regard in propria persona mimicry of it as true or false: that is a platitude. Thus when I say that the utterance is neither true nor false, and in particular that it is not true, I do not thereby commit myself to the claim that the dagger is not covered in blood. There is no dagger. There is no such claim. Can the same sort of assessment be given of a vague utterance? To be sure, given that what counts as a vague utterance is itself vague, the same sort of assessment cannot be given of a vague utterance unless what counts as saying something is also vague, that is unless the sentence ‘This utterance says something’ is a vague sentence. But it surely is. That is one consequence of the Governing Picture that we surely have to accept. Williamson himself has three arguments against the view that a vague utterance says nothing (in the relevant sense).49 One is that, had circumstances been different in such a way that the utterance had been non-vague, it would have said something; and more to the point, it would have said the same thing. Another is that there is no obstacle to understanding the utterance parallel to the obstacle to understanding an 48 Williamson (1994a), §7.2.

49 Williamson (1994a), pp. 195–197.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

52

Language, World, and Limits

utterance involving reference failure. Both of these strike me as question-begging. By far the most powerful of the three arguments turns on the apparent contribution that the content of the vague utterance can make to more complex utterances. Thus consider again an utterance of ‘You are a child’, addressed to a fourteen-year-old without any suitable sharpening. Whatever we are to say about this, it seems that we must accept the truth of an utterance of ‘If you are a child, then any younger sibling of yours is also a child’, addressed to the same fourteen-year-old.50 But, Williamson would insist, ‘the conditional says something only because its antecedent and consequent also do.’51 However, I think there is a corrective to this thought, which is to look beneath the forms of the sentences involved to the moves made with them.52 It is no accident that the truth of the conditional is itself no accident. The person making this utterance is stating a rule of representation: a prohibition against counting the addressee as a child without also counting any younger sibling of theirs as a child.53 But accepting this rule is quite compatible with giving each of various individual verdicts on the matter. In particular, I cannot see how it precludes refusing to count the addressee as a child and refusing to count the addressee as not a child. The truth of the conditional (that is, the holding of the rule) allows for, among other things, the lack of truth or falsity of an isolated appearance of its antecedent.54 In sum then: that a vague utterance is neither true nor false is another of the many things that might reasonably be said about it. And this carries no threat to the idea that we are forever using vague sentences to say, truly or falsely, how things are.

6. Stating rules of rules of representation This reference to rules of representation brings my essay full circle. It is a variation on the idea of a rule of representation that I want to consider in this final section, but still in connection with vagueness. I want to suggest an approach to the sorites paradoxes. (Nothing I said in the previous section really touches on these.) We sometimes—this is the sixth move—rule out cut-off points in connection with vague concepts. (By vague concepts I mean the concepts, like that of a child, that make vague sentences vague.) Thus consider Ellen, who is ten years old. We are prepared to endorse the following: 50 Let us take for granted that childhood is age-determined, at least to this extent. In fact, this is something of an idealization. (See further n. 56.) 51 Williamson (1994a), p. 196. 52 Cf. again Hacker’s volte-face, referred to above in n. 8. 53 If the person making the utterance knows that the addressee has a younger sibling, Stephen say, then he or she could also state what might be called an applied rule of representation, by saying, ‘If you are a child, then Stephen is a child.’ While the truth of the original utterance is necessary, the truth of this utterance would enjoy a sort of conditional necessity—conditional on Stephen’s being the addressee’s younger sibling. What I am about to say in the main text about the original rule would hold of this applied rule too. 54 It might be said, in defence of Williamson, that he expressly forestalls this counter-argument by talking about ‘material’ conditionals (Williamson (1994a), p. 196), where a material conditional, unlike a conditional used to state a rule of representation, is precisely one whose truth or falsity is determined by the truth or falsity of its antecedent and consequent. But in that case it is question-begging to suppose that there are any relevant conditionals that are both material and true (or material and false).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here? (5)

53

No particular day will mark the end of Ellen’s childhood.

(Being prepared to endorse (5) is part of what it is to have a full grasp of the concept of a child.) The Governing Picture itself suggests that we do well to have such rules, to safeguard the flexibility and the connection with casual observation that help to give these concepts their point.55 But of course, Ellen will stop being a child. And this creates a paradox. For her passage from childhood into adulthood seems, on reflection, to be impossible without a cut-off point. It seems that she cannot stop being a child unless there is a last day on which she is a child.56 This is a classic sorites paradox. Let us look a little more closely at how the paradox arises. (5) seems to be equivalent to the following: (6) For any day on which Ellen counts as a child, she will still count as a child the following day. But (6), combined with the premise that Ellen is a child today, yields the conclusion that she will remain a child ever after: her childhood will pass unchecked from one day to the next. So if Ellen is to stop being a child, then it seems that we must reject (6)—and thus (5). In fact, however, this does not follow. Certainly we must reject (6). What does not follow is that we must reject (5) as well. In order to see how (5) and (6) come apart, consider this: in rejecting (6), do we commit ourselves to endorsing (7)? (7) There will come a day on which Ellen counts as a child, even though she will no longer count as a child the following day. No. Think about what it is to endorse or to reject a sentence like (6) or (7). On their most natural interpretation, if such sentences are true, then they are necessarily true.57 To endorse such a sentence is to accept a rule of representation. To reject it is to decline to accept that rule. There is no reason whatsoever why we should not reject both (6) and (7). To insist otherwise would be a little bit like insisting that either ‘The opening move shall be a Pawn move’ or ‘The opening move shall not be a Pawn move’ must be a rule of chess. We can decline to have a rule of representation whereby we are not to count Ellen a child on any given day unless we also count her a child on the following day—on pain of having to admit that she will never stop being a child. But we can also, quite consistently with that, decline to have a rule of representation that 55 Cf. Dummett (1997), p. 109. 56 Not that the age at which Ellen stops being a child need be the same as the age at which any other person stops being a child. Cf. n. 50: it is something of an idealization to think of childhood as age-determined at all, even to an extent that precludes children having younger siblings who are not children, or adolescents reverting to childhood for that matter; but it would be a far greater idealization, with which we need have no truck, to think that, for any two people of the same age, one is a child if and only if the other is a child. Childhood is more contextual than that. 57 Cf. what I said about ‘If you are a child, then any younger sibling of yours is a child’ in the previous section. But cf. also n. 53: it is more accurate to speak here of conditional necessity than of necessity simpliciter. Ellen might never have been born. She might (God forbid) not survive her childhood. (The latter possibility is one from which I am prescinding throughout this section.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

54

Language, World, and Limits

forces us to acknowledge that there will come a last day on which she is a child—on pain of violating the very vagueness of our concept of a child.58 Ellen will gradually stop being a child.59 But to reject (7)—to decline to have that rule—is itself to have a rule. It is to have a second-order rule that we are not to adopt that first-order rule, or any other that commits us to it (for instance, the first-order rule that a person stops counting as a child on his or her fourteenth birthday). This second-order rule might be called a rule of rule of representation. It does not preclude our imposing precision on the concept of a child for certain specific and restricted purposes, say in legal contexts. We would not be violating the rule if we said that, for such and such purposes, we were going to count anyone under the age of fourteen as a child and anyone else as not a child.60 But we would be violating the rule if we saw this as binding on all subsequent uses of the concept of a child. The second-order rule precisely safeguards our entitlement to impose such precision, but with a different cut-off point, on some later occasion.61 We can accept and state this rule of rule of representation then. And that, I submit, is what we are doing when we endorse (5). (This is why I said at the beginning of this section that I wanted to consider a ‘variation’ on the idea of a rule of representation. The sixth move is like the first. Nevertheless, it is importantly distinct.) To think that (5) is equivalent to (6), or rather that endorsing (5) is equivalent to endorsing (6), is to confuse levels. There is no obstacle to our endorsing (5) and rejecting (6). Nor does our endorsing (5) lead to paradox: this is precisely because it does not involve our endorsing (6).

58 Cf. Williams (1995b), p. 217. Cf. also Wittgenstein (1978), Pt. V, §13. 59 In the previous section I tried to give some indication of the free play that our concept of a child allows for, alluding in particular to the way in which vague utterances that use this concept can function like performatives. Part of what it means to say that Ellen will gradually stop being a child, in semantic terms, is that this free play, applied over the course of her life, will first gradually increase, then gradually decrease. (Part of what this in turn means is that the sentence ‘Ellen is old enough for an utterance of “Ellen is a child” to function like a performative’ is every bit as vague as the sentence ‘Ellen is a child.’ This seems to me to rebut an argument in Williamson (1994a), §7.3, in favour of the view that the concept of a child has some sort of cut-off point. (Williamson actually couches the argument in terms of the concept of a heap.) This argument can be broached by considering the two following scenarios. In the first, there are two cooperative omniscient speakers who are asked to say whether Ellen is a child at various future dates. They diverge in when they stop calling her a child. In the second, there are again two co-operative omniscient speakers who are asked to say whether Ellen is a child at various future dates, but they are also instructed to use any discretion they are allowed as conservatively as possible (that is, roughly, they are told to stop calling her a child as soon as they can). They too diverge in when they stop calling her a child. The argument turns on the thought that, whatever sense we can make of the first of these scenarios, the second is unintelligible. For if it is, then there must after all be some sort of cut-off point for the concept of a child. But in fact, granted what I said about the vagueness of the sentence ‘Ellen is old enough for an utterance of “Ellen is a child” to function like a performative’, the second scenario is no more unintelligible than the first. Or if it is, then this is because the instruction given to the two speakers—to use any discretion they are allowed as conservatively as possible—is meant to apply even to itself, in which case I think we have reason to locate the unintelligibility of the scenario, not in the fact that there is some sort of cut-off point for the concept of a child, but in the unintelligibility of the instruction.) 60 My 1993 edition of The Chambers Dictionary defines ‘child’ as ‘a very young person (up to the age of sixteen for the purpose of some acts of parliament, under fourteen in criminal law)’. 61 This is another example of the free play that our concept of a child allows for (see n. 59). Cf. once again Sainsbury (1997), §6.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

What are these Familiar Words Doing Here?

55

An objector might say, ‘This is all very well, but suppose we consider uses of (5), (6), and (7), not to make the moves that you have been talking about, but to make simple empirical claims about how things are. What then?’ Well what then? It is not obvious from the sentences themselves what empirical claims are being envisaged. If what I have been urging about the flexibility of our concept of a child is right—if some applications of the word ‘child’ to Ellen can count as correct just by virtue of being made, while others count as neither correct nor incorrect—then it is of no avail here just to appeal to our understanding of the words in the sentences and then to try somehow to invoke compositional semantics. That simply does not deliver any suitable interpretations. The fact is that these sentences cannot be used to make empirical claims about how things are without a certain amount of supplementary gloss (and a certain amount of artificiality). Some utterances of them, on some suitable glosses, will be unproblematically true; others will be unproblematically false; and others again, granted what I said in the previous section, will be neither true nor false. For instance, an utterance of (6) will be unproblematically false if it means that, for any day on which Ellen counts as a child, she will undergo no ageing process during the ensuing twenty-four hours. An utterance of (5) will be unproblematically true if it means that we shall not, as a matter of fact, have a special day to mark the end of Ellen’s childhood. A corresponding utterance of (7) will then be unproblematically false. Conversely, an utterance of (7) may be unproblematically true if it means that there will come a last day on which everyone in a certain group (or a majority of people in that group, or at least one person in that group, or the utterer himself) is prepared to say, without hesitation, that Ellen is a child. And a corresponding utterance of (5) will then be unproblematically false. Or it may be that neither of these utterances will be true or false, given that whether someone is prepared to say, without hesitation, that Ellen is a child is itself a vague matter. None of this, so far as I can see, threatens paradox. A principal lesson of this section, then, as of all the others, is that there is elucidation to be gained from diverting attention away from the forms of sentences to the moves made with them. And despite the discouraging messiness of linguistic behaviour, the moves are there to be discerned. In particular, we sometimes state rules of rules of representation—just as we sometimes make each of the other five moves that I have focused on in this essay. The Governing Picture does not gainsay this. On the contrary, due appreciation of the Governing Picture can enhance our understanding of all six moves, and can help us to see our way round some familiar philosophical conundrums.62

62 I am very grateful to Uri Henig for a number of helpful discussions, and to Peter Hacker, Oswald Hanfling, and Timothy Williamson for comments on an earlier draft. I have also profited from Henig (unpublished).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

4 The Bounds of Nonsense Abstract This essay is about the distinction between sense and nonsense, or more strictly the distinction between truth-valued propositions and nonsensical pseudo-propositions, in Wittgenstein’s Tractatus. Two questions that are raised are: whether ‘truth-valued’ in ‘truth-valued propositions’ is pleonastic; and whether ‘nonsensical’ in ‘nonsensical pseudo-propositions’ is pleonastic. Neither question, it is conceded, has much exegetical or philosophical significance. But there is an associated question that does: namely, whether we have any understanding of what it is for something to be a pseudo-proposition without a truth-value independently of what it is for something to be a proposition with one. It is urged that, for Wittgenstein, we do not: a pseudo-proposition without a truth-value is an item that appears, falsely, to be a proposition with one. In an appendix the question is raised whether Kant would have done well to say something similar about empty thoughts.

There is a tripartite classification that is standardly attributed to the early Wittgenstein. According to this attribution, Wittgenstein acknowledges the following three mutually disjoint categories in the Tractatus:1 • thoughts; • tautologies and contradictions; • nonsensical pseudo-propositions. Thoughts are propositions with a sense, in other words propositions that are bipolar, or in yet other words propositions that are not only true or false but also such that, if true, they could nevertheless have been false, while, if false, they could nevertheless have been true.2 Tautologies and contradictions are propositions without a sense. They lack a sense because they lack the bipolarity that thoughts have. Although they are true or false, the true ones, namely the tautologies, are unconditionally true, while the false ones, namely the contradictions, are unconditionally false. Their lacking a sense in this very distinctive way is registered by saying that they are senseless, but not nonsensical.3 Finally, there are pseudo-propositions. These are concatenations of signs that do not belong to either of the first two categories.

1 Wittgenstein (1961). All unaccompanied references to Wittgenstein will be to this. 2 See e.g. 2.2–3, 4, and 4.2. 3 See e.g. 4.461–4.4611.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

57

They are neither true nor false. And, in contrast to tautologies and contradictions, they are nonsensical.4 Let us call attribution of this schematism to the early Wittgenstein the Standard Account. Michael Kremer and Cora Diamond have contested it.5 They have a rival account that differs with respect to the third category. Nowhere in the Tractatus, they urge, does Wittgenstein commit himself to the view that, just because a concatenation of signs is to be classified as a pseudo-proposition that is neither true nor false, it must also be classified as nonsensical. The most that he does is to highlight some pseudopropositions that he takes to be nonsensical.6 This allows Kremer and Diamond to credit Wittgenstein with a much more circumscribed and much less draconian conception of nonsense. Among the pseudo-propositions that Kremer and Diamond think Wittgenstein would decline to classify as nonsensical, perhaps the most interesting and the most compelling examples are those of mathematics. These count as pseudo-propositions for Wittgenstein because he takes them to be equations and therefore, on the account of equations that he proffers in the Tractatus, representational devices that are neither true nor false.7 He is nevertheless quite happy to acknowledge that they have a clear, codifiable, and important use.8 It follows that they have a kind of meaning, at least on a suitably generous conception of meaning.9 Given this, and given that Wittgenstein nowhere explicitly classifies these pseudo-propositions as nonsensical, it seems to Kremer and Diamond a needless exegetical affront to insist that this is what he would do. They have a point. Kremer’s and Diamond’s alternative account certainly has considerable appeal. But it does also have at least one significant pitfall that I should like to flag. If we adopt their account, we are in danger of making a mystery, or perhaps of compounding the already existing mystery, of Wittgenstein’s own avowal that the material in the Tractatus itself, or most of it anyway, is to be classified as nonsensical.10 For that material too has a use. It is a peculiar use, admittedly, and perhaps not one that is happily described—as I did describe the use that mathematical pseudopropositions have—as clear and codifiable. In fact one of the main exegetical challenges confronting any student of this text is to determine how the use to which the material in it is to be put should be described. But the use in question is a use all the same. Wittgenstein has written this book with, as he indicates in his penultimate remark, a quite particular intention concerning what his readers are supposed to do with it and what they are supposed to glean from it.11 And if we think that he not only intends his readers to profit from the book in this way, but intends them to do so by means of the recognition of this very intention—which is not at all implausible—then 4 See e.g. 4.1272, 5.4733, and 6.53. 5 See Kremer (2002) and Diamond (2011), pp. 242, 246 f., and 253, esp. n. 7. 6 4.1272. 7 4.241 ff., 5.53 ff., and 6.2. 8 6.211 ff. 9 In fact Kremer mounts a very persuasive case for this being the Tractatus’ own conception of meaning: see Kremer (2002), pp. 283 ff. Cf. also Wittgenstein (1978), Pt. V, §2, where, in keeping with 6.211 of the Tractatus, Wittgenstein says that what makes a sign-game into mathematics is its use outside mathematics, and then adds that this use constitutes the meaning of the signs in question. Remark 3.328 in the Tractatus is also relevant here. 10 6.54. 11 6.54.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

58

Language, World, and Limits

we do not have to be all that Gricean12 to conclude that the material in the book just does, ipso facto, have a kind of meaning. But why then the contrast, as far as any classification as nonsense is concerned, between this material and what we find in mathematics? I am far from suggesting that this is an insuperable objection to Kremer’s and Diamond’s account. There may well be a compelling Tractarian story to be told about how the many differences between the two cases can be marshalled to justify classifying the former, but not the latter, as nonsensical.13 But it would certainly make for a much easier exegetical life if, in accord with the Standard Account, we did not have to worry about telling any such story, but could simply accept those differences in their own terms and acquiesce in the idea that all this material, mathematics included, counts as nonsensical. Which account is to be preferred, then? Perhaps neither. Perhaps the text simply fails to settle the matter.14 And perhaps a principle of charity affords no help either. A principle of charity would afford help only if there were more at stake here than a boring matter of terminology, and there are grounds for thinking that, at one level, there is not. On the Standard Account, the term ‘nonsensical’ serves as a convenient sweepup term that acts as the adjectival counterpart to the noun ‘pseudo-proposition’—at least relative to a suitable domain.15 On Kremer’s and Diamond’s alternative account, the term ‘nonsensical’ does additional work that allows for finer discrimination. But it would be easy enough, in the former case, to use ‘pseudo-propositional’ instead; and it would be easy enough, in the latter case, to devise another term to do the additional work in question. And neither use of the term, let it be noted, is directly appropriated from ordinary language. Both accounts cast the term, at least to some extent, as a term of art—which indeed it is. This is sometimes denied. It is sometimes said that Wittgenstein uses ‘nonsensical’ in the Tractatus in none other than the way in which it is ordinarily used. But insofar as there is any such thing as ‘the’ way in which it is ordinarily used (and the same applies to ‘unsinnig’, the term that appears in

12 Grice (1967). 13 Diamond in fact anticipates the objection and tries to tell just such a story. In Diamond (2019), p. 185, she writes, ‘We need to distinguish cases like that of equations, which are [pseudo-propositions], which may look as if they are about things named in them, and which have a usefulness which is not dependent on taking them to be about those things, from cases of [pseudo-propositions] which look as if they are about things named in them (and are such that, taken in that way, they are nonsensical because they contain some sign or signs with no meaning), and which have a usefulness dependent upon both their capacity to mislead us (through their apparent aboutness) and our ultimate capacity to see through the deception. Propositions of mathematics . . . have a usefulness tied in in various ways with the functioning of . . . propositions [with a sense], a usefulness which is in no way dependent upon taking them to be a kind of failed . . . proposition [with a sense], whereas there are other propositions which are useful in particular contexts precisely through the recognition of such failure.’ Cf. also ibid. pp. 199–200. 14 That is my own view. The remark in the Tractatus that seems to me to come closest to settling it is 5.5303, which is at first sight very puzzling but which can certainly be made to appear less puzzling by a reading of the kind that Kremer and Diamond adopt: see esp. Kremer (2002), pp. 294–297. Even so, I do not take this remark to be decisive. (For one thing, as Wim Vanrie has pointed out to me, it is not obvious that 5.5303 has anything to do with equations, which it needs to have for a reading of the kind that Kremer and Diamond adopt to be relevant to it in any way that is itself relevant.) 15 This qualification anticipates one of the main issues of this essay.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

59

the original German text) this is simply false—if only because it would sound entirely natural, to an untrained ear, to call a blatant contradiction ‘nonsensical’. Nor, come to that, does ordinary language afford us any interesting relevant distinction between ‘nonsensical’ and ‘senseless’ (or between ‘unsinnig’ and ‘sinnlos’). This reminds us that there is an issue about Wittgenstein’s use of the term ‘senseless’ too. Here the Standard Account can go one of two ways. Wittgenstein may be said to reserve ‘senseless’ for tautologies and contradictions; or he may be said to apply it to any concatenation of signs that lacks sense, including any that is to be classified as nonsensical. Kremer’s and Diamond’s account, for all I have said so far, can likewise go one of two ways. They too may say that Wittgenstein uses ‘senseless’ in the former, narrower way; or they may say that he uses it in the latter, broader way. In fact they favour the second of these.16 As I have already indicated, there are grounds for thinking that not much of exegetical substance hangs on this. There are grounds, more specifically, for thinking that nothing in the Tractatus settles the matter, that nothing of significance in the book turns on the matter, and that the matter itself concerns nothing more than which of various equally serviceable, relatively technical uses Wittgenstein assigns to a couple of terms. This is not to deny that there are fascinating and important exegetical issues in the vicinity. (I hope that this essay will itself bear witness to that.) It is not even to deny that the formulation of these issues is sensitive to the matter in hand.17 It is just that, if so, this is not what makes them fascinating and important. Do similar remarks apply to Wittgenstein’s use of the term ‘pseudo-proposition’? Is there similar leeway in the interpretation of this term? Perhaps, on some readings of it, it denotes some concatenations of signs that lack a truth-value, on others, others.18 Perhaps, on some readings, it extends to concatenations of signs that have a truthvalue, in particular to tautologies and contradictions19—something that is certainly the case in Wittgenstein’s pre-Tractatus notebooks.20 It might be said, in response to this last suggestion, that here at last the Tractatus is decisive, since by the time of the book itself Wittgenstein has undeniably changed his mind and settled against the view that tautologies and contradictions are pseudopropositions. But in response to this response, while it is clear that by the time of the Tractatus Wittgenstein classifies tautologies and contradictions as propositions,21 and while it is also clear that this represents a change of mind—since at one point in his notebooks he explicitly denies that tautologies and contradictions are propositions22— what is not clear is that the change of mind is a change of mind about the status of tautologies and contradictions as pseudo-propositions. For we cannot simply take for granted that being a pseudo-proposition, on Wittgenstein’s understanding, is 16 See e.g. Kremer (2002), p. 300 and Diamond (2011), n. 5. 17 I take both Kremer (2002) and Diamond (2011) to illustrate this point. 18 But there is no reasonable reading on which it denotes all concatenations of signs that lack a truthvalue: cf. n. 15 and see further below. 19 I have chosen the word ‘concatenation’, incidentally, rather than ‘combination’, as a way of trying to circumvent the exegetical minefield that is 4.466–4.4661—part of whose message seems to be that we had better not unthinkingly refer to tautologies and contradictions as combinations of signs. (But see also 6.124.) I hope that ‘concatenation’ begs no questions. 20 Wittgenstein (1979), pp. 12 and 58. 21 4.46 ff. 22 Wittgenstein (1979), p. 58.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

60

Language, World, and Limits

incompatible with being a proposition. Perhaps, in the Tractatus, tautologies and contradictions are both. This is not an outrageous suggestion. To the extent that it is out of keeping with the way in which ‘pseudo-’ (or ‘Schein’ in the original German) normally functions, so be it: we can, as with ‘nonsensical’, treat ‘pseudo-proposition’ as a more or less technical term. Admittedly, there are contexts in the Tractatus where it is hard to hear Wittgenstein’s use of ‘pseudo-’, either juxtaposed with ‘proposition’ or juxtaposed with some other noun, as anything other than privative.23 But this may be a simple matter of implicature. (There are contexts in which it is hard to hear uses of ‘attempted’ as anything other than privative either, yet this does not gainsay the fact that being an attempted overhead bicycle kick is compatible with being an overhead bicycle kick.) Moreover, there is mathematics to be considered again. When Wittgenstein tells us that what we find in mathematics are pseudo-propositions, he does not put this by saying that the concatenations of signs that we find in mathematics are pseudopropositions; he puts it by saying that the propositions we find there are pseudopropositions.24 Unless his use of ‘propositions’ here is simply sloppy, or unless it involves some rhetorical device,25 then it indicates not only his preparedness to countenance the possibility of propositions that are also pseudo-propositions, but his outright commitment to the existence of such things.26 Once we have taken seriously the suggestion that, for Wittgenstein, some pseudopropositions are propositions, it is a comparatively small step to the more radical suggestion that, for Wittgenstein, all pseudo-propositions are propositions, even the ones that he would uncontentiously classify as nonsensical. There is certainly nothing on the Standard Account to preclude saying this. In fact there is some reason to say it. For Wittgenstein does talk about nonsensical propositions.27 To be sure, we must once again allow for the possibility that this is either sloppiness or rhetoric on his part. But is there any special reason to do so? Why should the term ‘proposition’ not be used, non-sloppily and non-rhetorically, in a broad enough way to embrace both the pseudo-propositional and (if this is different) the nonsensical? Perhaps what is required of a concatenation of signs for it to be a pseudo-proposition is, not that it should appear—that is, merely appear—to be a proposition, but that it should appear— that is, merely appear—to have a sense,28 or (if tautologies and contradictions do not count as pseudo-propositions) that it should appear—that is, merely appear—to have a truth-value. 23 For instance at 4.1272, 5.461, and 5.534. 24 6.2. 25 These possibilities are certainly not to be excluded. A fake pearl is not a pearl: but it would be both natural and perfectly intelligible to say something like, ‘Only one of these two pearls is real; the other is fake.’ 26 If the pseudo-propositions of mathematics are indeed propositions, then can we say that ‘pseudo-’, in this context, functions like ‘malformed’, in the phrase ‘malformed finger’—a malformed finger being none the less a finger? That is still not quite right. A better analogy, I think, is the use of ‘sham’, in the phrase ‘sham marriage’. A sham marriage, though none the less a marriage, appears to be something that it is not. See further below. 27 4.003 and 6.54. 28 Cf. the previous note. And cf. Wittgenstein (1979), pp. 9, 12, and 16. Here we find intimations of the further idea, which is not in the Tractatus, that the reason why (some) pseudo-propositions appear to have a sense is that they succeed in showing something: the very thing that appears to be their sense.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

61

Be that as it may, there is an interesting and substantive issue that has hitherto been in the background but that is now squarely in the foreground. It is an issue about the very subject matter of Kremer’s and Diamond’s quarrel with the Standard Account. What, harking back to the opening sentence of this essay, was that tripartite classification supposed to be a classification of?29 When I characterized the first two categories, I was explicit that each consisted of propositions of some kind. But I was deliberately non-committal about whether, between them, they exhausted the propositions. When I characterized the third category, I suggested that the overall classification was a classification of concatenations of signs. But ‘suggested’ is the operative word. And even then there was no suggestion that it was a classification of all concatenations of signs. I deliberately allowed for the possibility that there are, on the Standard Account, concatenations of signs of various kinds that do not belong to any of the three categories, hence that do not even count as pseudo-propositions—the most obvious candidates being, on the one hand, perfectly meaningful concatenations of signs to which this whole discussion appears to have no application, such as questions and commands, and, on the other hand, word salads such as ‘Interstellar backlash carrot’.30 Subsequently, I have been cagier still, for instance when talking about the ‘material’ in the Tractatus. In the light of our most recent reflections, however, the caginess appears both unnecessary and easily surmountable. There is a simple answer to the question what the tripartite classification is a classification of, and hence what Kremer’s and Diamond’s quarrel with the Standard Account is a quarrel about, namely: propositions. That certainly rules out everything that we want to rule out. However, it is not unique in doing so. An equally natural alternative would be to reserve the term ‘proposition’ for items in the first two categories, then to say that the domain as a whole consists of apparent propositions, with those in the third category being distinguished by the fact that their propositionhood is merely apparent. This seems to be Kremer’s and Diamond’s approach. At one point Kremer describes the pseudo-propositions of mathematics as ‘apparent propositions’.31 Diamond, for her part, speaks in a similar connection of ‘proposition-like constructions’.32 Elsewhere she speaks of ‘sentence-constructions’.33 Very well, which of these two uses of the term ‘proposition’ (if either) do we find in the Tractatus?34 Call the view that we find the first, broader use the Broad Interpretation; and call the view that we find the second, narrower use the Narrow Interpretation. We have seen reasons for adopting the Broad Interpretation. But we have also seen reasons for being suspicious of these reasons. Clearly there is much more to be said about the matter. And there is much more that has been said. 29 Cf. again n. 15. 30 Not that the candidature of any of these is entirely straightforward. Concerning questions and commands, there is an issue about whether they are non-propositional. See e.g. Davidson (1984d), where Donald Davidson presents a propositional account of them. Cf. also Wittgenstein’s own comment, in one of his pre-Tractatus notes, that ‘judgement, command and question all stand on the same level; but all have in common the propositional form’, (Wittgenstein (1979), Appendix I, ‘Notes on Logic: 1913’, p. 107). This reminds us that propositions themselves are not to be thought of as already assertoric. Concerning word salads, there is an issue about whether they are concatenations of signs. Perhaps nothing counts as a sign save in the context of a proposition: see e.g. 3.31 and 3.32. I shall not pursue any of these complications here. 31 Kremer (2002), p. 294. 32 Diamond (2011), p. 247. 33 Ibid. p. 263. 34 This is related to the issue that Peter Sullivan flags in Sullivan (2003), n. 25.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

62

Language, World, and Limits

Interesting and compelling arguments have been advanced both for the Broad Interpretation and for the Narrow Interpretation.35 My own view, as I have already intimated, is that we find at least some instances of the broad use in the Tractatus. If I am right, then the Narrow Interpretation is wrong. (I am construing each of the two interpretations in such a way that it requires uniformity of use throughout the book.) But I shall not argue for that view here. For I also believe that there is, in the Narrow Interpretation, and in the arguments for it, an insight whose exegetical and philosophical significance far exceeds that of its strict incorrectness. The choice between the two interpretations, or between both of them and some third interpretation whereby neither the broad use nor the narrow use is uniform throughout the book, appears to me to be another example of a relatively boring terminological matter which is not where the real exegetical or philosophical action is. To see what I have in mind, note first that, whatever our view about this matter, we cannot deny that Wittgenstein recognizes a clear and fundamental distinction between items in the first two categories, that is to say truth-valued propositions, and items in the third category (afforced, if the Standard Account is incorrect, by such non-nonsensical items as mathematical equations36), that is to say truth-valueless pseudo-propositions; nor that this distinction is of critical importance to him. Call this distinction the Principal Distinction. Part of the reason why the Principal Distinction is of critical importance to Wittgenstein, as he (all but) says in the preface to his book, is that there is nothing incoherent in the idea of our drawing it. Why does this matter? It matters because drawing this distinction, with due qualifications concerning the special status of tautologies and contradictions, can serve as an unproblematical surrogate for trying to do something the idea of which is incoherent, though it is also of peculiar and central relevance to Wittgenstein’s project, namely drawing a limit to what can be thought. This idea is incoherent because, in order to draw a limit to what can be thought, ‘we should have to find both sides of the limit thinkable (i.e. we should have to be able to think what cannot be thought)’.37 It is nevertheless of peculiar and central relevance to Wittgenstein’s project because it can all too easily appear to be the aim of the project, and, partly for that very reason, the attempt to do 35 They have been advanced for the Broad Interpretation by (e.g.) Colin Johnston, in Johnston (2007), and by Michael Morris, in Morris (2008), §4C. Both Johnston and Morris see a priority of syntax over semantics in the Tractatus whereby propositions, in the relevant broad sense, have a syntactic articulation even when they are nonsensical: in particular, they contain names. Interesting and compelling arguments for the Narrow Interpretation have been advanced by (e.g.) Colin Johnston again, this time in his later essay Johnston (2017), esp. §7, and by Wim Vanrie, in Vanrie (unpublished). (Johnston, in his later essay, does not address the issue explicitly in these terms, nor does he say that he has changed his mind. But he has confirmed this in private correspondence with me.) Among the most compelling pieces of evidence for the Broad Interpretation are: 3.13, 3.33, 4.003, 5.4733, 6.2, 6.53, and 6.54. Among the most compelling pieces of evidence for the Narrow Interpretation are various remarks in the 3.3s, esp. 3.326, 4.5–5, 5.5422, and 6–6.01. Both Johnston, in his earlier essay, and Morris acknowledge 3.326 as a threat to the Broad Interpretation. There Wittgenstein appears to claim that we cannot recognize a sign as having a syntactic use, nor therefore presumably as being propositional, until we have observed it being used ‘with a sense’. Johnston explains this by construing ‘with a sense’ in this context (a rendering of ‘sinnvollen’) as itself an allusion to syntax (see Johnston (2007), §3.2); Morris explains it by arguing that Wittgenstein is talking about our most usual or most convenient way of recognizing the syntactic use of a sign (see Morris (2008), pp. 165–166). 36 From now on I shall take this parenthetical qualification for granted. 37 P. 3.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

63

this thing provides rungs on the ladder that Wittgenstein wants us to climb and eventually to throw away.38 Throwing the ladder away involves acknowledging that the truth-valueless pseudo-propositions on the ‘far’ side of the distinction that we can draw, and do draw, namely the Principal Distinction, do not convey anything that cannot be thought: they do not convey anything at all. This in turn helps us to see that the very idea of that which cannot be thought is itself incoherent. Here, then, is the acceptable alternative to the unacceptable project in which Wittgenstein can appear to be engaged, and on which, without benefit of the lessons he passes on to us, either he or we might otherwise have embarked.39 The insight that I believe can be found in the Narrow Interpretation is an insight into the very nature of the Principal Distinction. This is perhaps unsurprising, since it is the Principal Distinction which, according to the Narrow Interpretation, Wittgenstein marks with the term ‘proposition’. Not that the matter is as straightforward as that. For the insight in question is also an insight into the very nature of the distinction that the Broad Interpretation has Wittgenstein marking with the term ‘proposition’: the distinction between items that belong to any of the three categories and items that belong to none of them. In order not to beg questions, let us henceforth use the neutral term ‘sentence’ to denote items that belong to any of the three categories. We can then accordingly use: ‘non-sentence’ to denote items that belong to none of them; ‘truth-valued sentence’ to denote items that belong to either of the first two; and ‘truth-valueless sentence’ to denote items that belong to the third. To see more clearly the insight that I am trying to tease out, let us begin with the sentence/non-sentence distinction, the distinction between items that lie within the domain of this discussion and items that lie outside it. What, to put it bluntly, are we talking about? What are sentences?40 Sentences, I submit, are those items to which truth-operations apply.41 There is an immediate and obvious objection to this. The objection is that truth-operations apply only to items that are truth-valued; hence that this answer marks, not the sentence/ non-sentence distinction, but the Principal Distinction, the distinction, within sentences, between those that are truth-valued and those that are not. But everything depends—and there is a recurring theme here—on whether ‘truth-operation’ is to be construed in a broad way or in a narrow way. The objection holds if ‘truth-operation’ is to be construed in a suitably narrow way. The most that can then be said of sentences in general is that they are those items to which truth-operations appear to apply. But ‘truth-operation’ can also be construed in a broad way, whereby (roughly speaking) appearances are all that count.42 And that, in a way, is the point. Sentences 38 6.54. 39 For further discussion of the issues raised in this paragraph, with particular reference to Kant, see Essay 6 in this volume. 40 That is: what, for Wittgenstein, are sentences? What follows is exegesis. Only very indirectly shall I address the question whether his views are correct or not. 41 Cf. 5.2 ff. and 6 ff. Cf. also Wittgenstein (1967), Pt. I, §136; and Wittgenstein (1978), Pt. I, App. III, §2. 42 This enables us to characterize the sentence ‘7 + 5 ≠ 13,’ for instance, as the negation of the sentence ‘7 + 5 = 13.’ Of course, it follows that, if we simply say, ‘Sentences are those items to which truth-operations apply’, without any further gloss, then we shall have said something that, even if correct, is unhelpful—just as, if we simply said, ‘Propositions are those items to which truth-operations apply’, without any further gloss, then we should have said something that, even if correct, did not thereby clarify whether we were

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

64

Language, World, and Limits

in general are characterized by their appearing to be the very thing that truth-valued sentences in particular are: truth-valued. It follows that, even if the proposed answer to the question ‘What are sentences?’ is correct, that answer is, in a critical sense, parasitic on the answer that we should have to give to the more specific question, concerning the Principal Distinction, ‘When is a sentence truth-valued?’ It is only in application to truth-valued sentences that truthoperations are what they appear to be. When terms like ‘proposition’ and ‘truth-operation’ are construed in the broad way, they are construed as denoting items that have the appearance of those that they denote when they are construed in the narrow way.43 The broad way of construing such expressions is a matter of psychology, the narrow way of construing them a matter of logic. And it is the narrow way of construing them that is primary. What it is for an item to appear truth-valued, and for the corresponding operations to appear to apply to it, is to be understood in terms of what it is for an item to be truth-valued, and for those operations actually to apply to it; not vice versa. This is the insight in the Narrow Interpretation. One way to think of this is by analogy with the debate in the philosophy of perception about the relation between a veridical perception and a subjectively indistinguishable hallucination. According to what is sometimes called the ‘highest common factor’ conception, the same experience is had in each of the two cases: the difference between them consists in the fact that a certain connection obtains between the experience and the world beyond it in the first case but not in the second. According to what is sometimes called the ‘disjunctivist’ conception, there is nothing relevant in common between the two cases: all that can be said of the experience that is had in the second case, and all that needs to be said of it, is that it has the false appearance of being the same as that which is had in the first case. And if we describe a case in a way that leaves open which of the two of it is (for instance, if we say, ‘He thinks he can see a dagger in front of him’), then we mean nothing more than that either it is the one case or it is the other.44 The insight in the Narrow Interpretation is that something like the disjunctivist conception applies to the Principal Distinction. If we classify an item as a sentence—a proposition in the broad sense—then we mean nothing more than that either it is a truth-valued sentence—a proposition in the narrow sense—or it is an item that has the false appearance of being such. Sentencehood has no independent essence of its own.45 using ‘proposition’ in the broad way or in the narrow way. Consider in this connection Frege’s famous characterization of what he calls a ‘thought’ as ‘that to which the question “Is it true?” is in principle applicable’ (Frege (1997b), p. 362). As it stands, without some further gloss, this characterization allows for more or less broad interpretations of Frege’s use of the term ‘thought’. (I should add that I intend no criticism of Frege in saying this. In particular, I do not mean to suggest that he does nothing, elsewhere, to indicate what his intended interpretation is. That said, it is noteworthy that we find exegetical disagreement precisely on the question whether Frege recognizes truth-valueless thoughts. Michael Dummett thinks he does; Gareth Evans thinks he does not. See, respectively, Dummett (1981a), Ch. 6, §4 and Evans (1982), Ch. 1, §6.) Cf. also in this connection both 5.5351 and Wittgenstein (1974), Pt. VI, §79. 43 Cf. Diamond (2011), §4. 44 See further the essays in Haddock and Macpherson (2008). 45 There is an issue about whether the disjunctivist conception of perception is incompatible with phenomenology. For phenomenology enjoins the assimilation of the two cases considered in the main text. However, anyone who thinks that there is an incompatibility here is surely forgetting the fundamental methodological tactic of phenomenology, which is precisely to bracket our normal way of understanding the world in order to focus on how things appear: see e.g. Husserl (1962), Ch. 3. Seen in these terms, the

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

65

A caveat before I proceed. If the concept of a truth-valued sentence is what Wittgenstein calls a formal concept,46 and if the concept of a truth-valueless sentence is not—both of which strike me as plausible47—then, although this lends further support to a disjunctivist conception of the Principal Distinction, it also means that my characterization of these issues has been misleading, even modulo the Tractatus’ own commitment to the idea that any characterization of these issues, inasmuch as it will contain nonsense masquerading as sense, must be misleading. In particular, by framing the concern with the Principal Distinction as a concern with the question, ‘When is a sentence truth-valued?’, I have suggested that the task at hand is to say, concerning items that satisfy one condition, namely that of being a sentence, what is required of them to satisfy a second condition, namely that of having a truth-value. But such misleadingness can, I believe, be overcome. The question ‘When is a sentence truthvalued?’ can be recast as follows: ‘If, in a given situation, either signs are being used to constitute a truth or a falsehood or it falsely appears that signs are being used to constitute a truth or a falsehood, what can be said of the situation to indicate that the former is the case?’ Similarly, were we to say, ‘Both history and mathematics involve sentences, but only history involves truth-valued sentences’, our claim could be recast as follows: ‘In both history and mathematics either signs are used to constitute truths and falsehoods or it falsely appears that signs are used to constitute truths and falsehoods, but only in history is the former the case.’ With that caveat in place, I shall revert to my previous way of putting things. Now, even without that caveat in place, it is plain that the concept of a sentence is  not a tidy homogeneous concept. Apart from anything else, what we might call its  ‘second disjunct’—the concept of a truth-valueless sentence—is not a tidy homogeneous concept. There are all sorts of ways in which an item can falsely appear to be a truth-valued sentence. The appearance can be more or less superficial, affecting a more or less extended group of people, with more or less tendency to deceive. Truth-valueless sentences may include: rubbish that two people utter to each other in an effort to fool a third person into thinking that they are exchanging metaphysical profundities; mathematical equations; the philosophical nonsense that we produce as a result of ‘our failure to understand the logic of our language’;48 and much else besides. What Wittgenstein perhaps underappreciates, in ways that he will later rectify,49 is the extent to which even the ‘first disjunct’—the concept of a truth-valued sentence—is neither tidy nor homogeneous. This criticism, if fair, cuts deep. For the untidiness and heterogeneity of the one concept, combined with the untidiness and heterogeneity of the other, surely pose a threat to the supposed clarity and fundamentality of the distinction between them: that is, the Principal Distinction. (How compelling can the appearances be while still counting as mere appearances?) Be that as it may, there will certainly be situations in which it is very difficult to tell whether a given sentence is truth-valued or truth-valueless. There are deep illusions of meaning. This is something that Wittgenstein does appreciate. Indeed, as I suggested Tractatus is arguably involved in a kind of phenomenology—at least at certain points, and at least if the Narrow Interpretation is wrong. 46 4.126 ff. 47 See e.g. 4.5 ff. and 5.473 ff., respectively. 48 4.003. 49 Cf. Wittgenstein (1967), Pt. I, §23 and §§134–137; and Wittgenstein (1978), Pt. I, App. III, §§1–4.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

66

Language, World, and Limits

earlier, it is part of the reason why the Principal Distinction is of such importance to him. It is part of the very rationale for his writing the book. Imagine people in such a situation. Suppose that they have produced a sentence in an attempt to say how things are. And suppose that, for very subtle reasons, their attempt is a failure. Careful scrutiny of their sentence will eventually expose it as truth-valueless, although they do not (yet) realize this and they think that it is truthvalued. Does the fact that they think this mean that at any rate their sentence cannot be complete gibberish; that some of the signs in it must have meanings, the very meanings that they have in other, truth-valued sentences? No. Here we need to recall Wittgenstein’s appropriation of Frege’s context principle, whereby signs have meanings only in truth-valued sentences.50 Indeed signs serve as symbols only in truthvalued sentences, where a symbol is a sign together with a particular logico-syntactic use.51 Or at least, this is so provided that ‘meaning’, ‘symbol’, and ‘logico-syntactic use’ are themselves construed in a suitably narrow way. Again we must allow for appearances; and again, I believe, there is a broad way of construing such terms that accommodates the appearances. But the crucial point is that any relation between the use of the signs in this truth-valueless sentence and the use of those very same signs in truth-valued sentences, while it may have great psychological significance, has no logical or semantic significance whatsoever. To suppose otherwise—to suppose that the connection between signs and their meanings, or between signs and their logicosyntactic use, can be extricated from the contribution that those signs make to the saying of how things are—is to imagine an independence of the constituents of language from the saying of how things are, and indeed an independence of the constituents of reality from things’ being how they are, that are quite foreign to the Tractatus.52 This too is an insight of the Narrow Interpretation. In fact it is the same insight. For those connections would enjoy an independence of the sort indicated if the signs’ ‘meanings’ and their ‘logico-syntactic use’ were to be originally construed in the broad way. And this would mean that it was possible to understand what it is for an item to be a symbol, in the broad sense, without understanding what it is for an item to be a symbol, in the narrow sense. In particular, it would be possible to understand what it is for an item to be a sentence without understanding what it is for an item to be a truth-valued sentence. Still, there are such things as meanings, symbols, and the logico-syntactic use of signs, in their respective broad senses. Thus one constituent of a truth-valueless sentence may be identifiable as its subject, in the broad sense, another as its predicate, in the broad sense. One sign in the sentence may have speed as its designatum, in the broad sense, another motion. (Here it is worth remembering Wittgenstein’s comment.53 ‘The reason why “Socrates is identical” says nothing,’ Wittgenstein tells us, ‘is that we have not given any adjectival meaning to the word “identical”’ (emphasis in original). This comment would have no rationale unless the word ‘identical’ had some claim to being an adjective in this sentence—the same claim that the word ‘Socrates’ has to being a proper name there, with the individual Socrates as its 50 3.3. I take this to be the purport of 3.3. But in view of the potential ambiguity of the term ‘proposition’, I concede that this is not straightforward and merits further discussion, which I shall not enter into here. 51 3.32 ff. 52 201 ff. 53 At 5.4733.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

67

designatum.) Such features of signs are admittedly superficial features of them, grounded in the psychology of those who engage with the signs. They are a matter of how the signs appear; nothing more. Even so, they are there to be acknowledged. But can the way in which signs appear bear so much weight? Well, the appearances can be both extensive and elaborate. Truth-valueless sentences do not stand in isolation. It is as true of truth-valueless sentences as it is of truth-valued sentences that we are able to recognize their constituents by recognizing systematic interconnections between them and other items of the same kind.54 If, in a given truth-valueless sentence, a given sign has speed as its designatum, in the broad sense, then there will be other sentences in which that sign likewise has speed as its designatum, in the broad sense. The Tractatus itself is a striking illustration of how extensive and how elaborate the appearances can be, and of how much weight they can bear,55 even when they are false appearances. Very well; but how far can we extend these ideas? There are some terms, notably ‘truth-valued’ and ‘sense’, to which I have not granted a broad use. Is there any reason why not? Could we say that the nonsensical sentences in the Tractatus are truthvalued, in a broad sense of ‘truth-valued’? Or that some of them have a sense, in a broad sense of ‘sense’? Hence that some of them express thoughts, in a broad sense of ‘thoughts’—a vindication, of sorts, for Wittgenstein’s reference in his preface to ‘the thoughts that are expressed in [this book]’?56 Well, as Wittgenstein himself reminds us in his later work, there is no harm in our saying what we like as long as it does not prevent us from seeing the facts.57 Even so, I think it is helpful to keep the terms ‘truth-valued’ and ‘sense’ in reserve to be used exclusively in the narrow way.58 For no matter how extensive and elaborate any false appearances in this area may be, and no matter how robust they may be, they can  always eventually be exposed.59 And it is convenient to have terms whose use in characterizing sentences is unambiguously of such a kind that it cannot be compromised by any such exposure in the way in which a broad use can be. As for which terms are best suited to this rôle, they are those whose use—or, more specifically, and more significantly, those whose narrow use—is pivotal in framing the tripartite classification that constituted our point of entry into this essay (or whatever modified version of that classification is required if the Standard Account is incorrect). These include ‘truth-valued’ and ‘sense’. They also include ‘bipolar’, ‘true’, ‘false’, and ‘pseudoproposition’. The three categories of sentences that make up that tripartite classification, and the various distinctions between them, including the Principal Distinction between items in the first two categories and items in the third, are, as I have already indicated, of 54 Cf. Johnston (2007), p. 388. 55 Enough for the Broad Interpretation to be correct? On that I pass no comment, except to say that that is a colossal weight. 56 P. 3. 57 Wittgenstein (1967), Pt. I, §79. 58 For the record, however, I also think it is interesting to compare a broad use of ‘sense’, should such a thing be indulged, with the use of ‘sense’ that we find in the work of Gilles Deleuze: see esp. Deleuze (1990b). I say a little more about this in Moore (2012), pp. 565–566. 59 If a sentence’s lack of truth-value could not be exposed, then everything would behave as if it had a truth-value, and if everything behaved as if it had a truth-value, then it would have a truth-value: cf. 3.3218.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

68

Language, World, and Limits

paramount importance to Wittgenstein in the Tractatus. There is therefore something pleasingly reassuring in the thought that, once the dust has settled, the familiar account of them with which we began this essay (with whatever amendments may be required in the case of the third category—none, I am still inclined to think), remains unambiguously intact.

Appendix In this appendix I want to compare and contrast the idea of a truth-valueless sentence with the idea, which we find in Kant, of an ‘empty thought’. First, a point of terminology. Kant’s use of the word ‘thought’ is broader than Wittgenstein’s in the Tractatus. To avoid confusion, therefore, I shall retain the original German word for Kant’s idea and call the items in question ‘empty Gedanken’. By a ‘Gedanke’ Kant means any judgement about how things are.60 Gedanken are involved in all our cognitions, where our ‘cognitions’ are those acts or states of mind through which we entertain particular objects. The reason why our cognitions always involve Gedanken is that they not only require intuitions, whereby objects are directly given to us; they also require concepts, whereby we judge, or think, how the objects, as thus given, are.61 Kant famously puts this second requirement by saying that intuitions without concepts are blind. He puts the first requirement by saying that Gedanken without content are empty.62 Gedanken without content—empty Gedanken—are thus exercises of concepts that are not related to intuitions. Although empty Gedanken cannot furnish cognitions, they are still judgments about how things are, albeit things we know not what. And they can still be true or false.63 In this respect they are precisely to be contrasted with the truth-valueless sentences of the Tractatus. It is nevertheless apt to compare the two because, just as truthvalueless sentences, though they have what we could think of as a second-class status with respect to truth-valued sentences, are none the less sentences, so too empty Gedanken, though they have what we could think of as a second-class status with respect to contentful Gedanken, are none the less Gedanken.64 This in turn means that, where there was a question in the Tractatus about the nature of sentences in general, there is a question in Kant about the nature of Gedanken in general. Are the answers to these questions of a piece? It seems not. We cannot say that Gedanken in general are items that appear to have content and that empty Gedanken in particular are distinguished by the fact that the appearances, in their case, are false. For there is no suggestion that empty Gedanken do always appear to have content. Admittedly there are all sorts of illusions 60 Thus Kantian Gedanken, unlike Wittgensteinian thoughts, include analyticities. See respectively Kant (1998), A6/B10—all unaccompanied references to Kant will be to this book—and 6.11. 61 A19/B33 and A50–51/B74–75. 62 A51/B75. 63 See e.g. Bxxv –xvii n., Bxxixff., B146–149, B166 n., A253–254/B309–310, A741ff./B769ff., and A769ff./ B797ff. 64 John McDowell has denied this. He writes, ‘For a [Gedanke] to be empty . . . would be for it not really to be a [Gedanke] at all, and that is surely Kant’s point; he is not, absurdly, drawing our attention to a special kind of [Gedanken], the empty ones’ (McDowell (1994), pp. 3–4). But, as the references given in the main text clearly show, that is exactly what Kant is doing—or at least what he takes himself to be doing.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Bounds of Nonsense

69

hereabouts.65 For example, there are illusions that involve merely apparent Gedanken, items that appear to be judgements about how things are but that do not even get as far as being empty such things. The most notable among these are muddled conjectures in which a priori concepts that have no application to objects given in intuition, such as the concept of the unconditioned, are confusedly amalgamated with concepts that have application only to objects given in intuition, such as the concept of the physical, the offending amalgam in this case being the concept of the physical world as an unconditioned whole.66 Come to that, there are cases of the very sort envisaged: empty Gedanken that appear, falsely, to have content. For instance, there are empty Gedanken which are grounded in the unity of consciousness—as all Gedanken are— but which so exploit that unity that they appear, falsely, to involve an intuition of the thinking subject as an object.67 The point, however, is that there are also empty Gedanken that appear to be just what they are. This suggests that Kant’s conception of Gedanken is a ‘highest common factor’ conception; that Gedanken, for Kant, have their own essence and can be divided into two categories according to whether or not they have, in addition, content. In fact, however, there is a very significant disjunctivist element in Kant’s conception. For Kant does not believe that we can make any real sense of Gedanken unless they have content.68 Our grasp of empty Gedanken, along with our grasp of what it is for them even to count as Gedanken, is extremely tenuous and parasitic on our grasp of contentful Gedanken. Empty Gedanken are to be thought of as abstracted from contentful Gedanken through the elimination of content and the retention of form.69 But we have no content-independent understanding of what the retention of form amounts to. That is, our understanding of what it is for Gedanken in general to have form is grounded in our understanding of what it is for contentful Gedanken in particular to have it. More specifically, it is grounded in our understanding of what it is for those a priori concepts that reflect the different logical forms of judgement to be applied to objects of intuition.70 When we refer to the form of a Gedanke, then, we are indicating either the application of such a priori concepts to objects of intuition or an exercise of such concepts that is of the same kind as this but without any relation to intuition. By the same token, when we classify an item as a Gedanke, we are saying that either it is a contentful Gedanke or it is an item that has been abstracted from such in the way described. Thus, just as in the Tractatus sentences in general are characterized by their having the appearance of truth-valued sentences in particular, so too in Kant Gedanken in general are characterized by their having the form of contentful Gedanken in particular. But this prompts a concern—I shall characterize it no more strongly than that, and I shall do no more in conclusion than raise it—about Kant’s very notion of a Gedanke.71 Given the analogy just indicated between empty Gedanken and truth-valueless 65 The ‘Transcendental Dialectic’ in Kant (1998) provides sustained testimony to this. 66 A307–308/B364–365, A418–419/B446–447, and A740–741/B768–769. 67 B421–422. 68 See e.g. B148–149, A242–243/B300–301, A678–679/B706–707, and A696/B724. 69 A253–254/B309 and A408–409/B435–436. 70 A70–83/B95–109. 71 A crude way to express the concern would be to ask whether an exercise of a priori concepts that is of the same kind as an application of them to objects of intuition, but without any relation to intuition, is not disconcertingly reminiscent of a meal that is like fish and chips, but without the fish and without the chips.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

70

Language, World, and Limits

sentences; given various further analogies that recommend themselves, such as that between the abstraction to which I have alluded in Kant and the attempt in the Tractatus to treat formal concepts as though they were concepts proper, an attempt that Wittgenstein expressly warns is destined to issue in nonsensical pseudopropositions;72 given how much weight the Tractarian appearances can bear; and given how tenuous Kant thinks our grasp of empty Gedanken is: we cannot but wonder whether the analogy between empty Gedanken and truth-valueless sentences should have been even closer. More specifically, we cannot but wonder whether Kant should have held that Gedanken in general are characterized by their appearing to be truth-valued judgments about how things are, while empty Gedanken in particular are distinguished by the fact that the appearances, in their case, are false.73

72 4.126–4.1272. 73 For very helpful comments on this essay I should like to thank Cora Diamond and Wim Vanrie. I finished the essay before I became aware of Bronzo (2017). I was both delighted and somewhat disconcerted to see the striking convergence of ideas. I have deliberately left my essay intact, making no attempt to modify it in the light of Bronzo’s essay. I hope that readers will find something of interest in our independent arrival at the same ideas. I should add that I have learned a great deal from Bronzo’s essay, which includes much that I had not in any way anticipated.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

5 Transcendental Idealism in Wittgenstein, and Theories of Meaning Abstract This essay involves exploration of certain repercussions of Bernard Williams’ view that there is, in Wittgenstein’s later work, a transcendental idealism akin to that found in the Tractatus—sharing with it the feature that it cannot be satisfactorily stated. It is  argued that, if Williams is right, then Wittgenstein’s later work precludes a philosophically substantial theory of meaning; for such a theory would force us to try to state the idealism. In a postscript written for the reprint of the essay, reasons are given for thinking that Williams is not right: Wittgenstein’s later work actually helps us to repudiate as ill-conceived all those questions whose answers invite us to embrace any such idealism. But the main thesis of the essay remains intact. Indeed the idea that Wittgenstein’s later work precludes a philosophically substantial theory of meaning is reinforced.

1. Transcendental idealism and the predicament that is inherent in it The new theory of meaning, like the old, points in the direction of a transcendental idealism, and shares also the problem of our being driven to state it in forms which are required to be understood, if at all, in the wrong way.

This is how Bernard Williams represents what he takes to be a fundamental continuity between the Tractatus and Wittgenstein’s later work.1 He argues that there is, in Wittgenstein’s progression of thought, a common element of transcendental idealism,2 which arises from each of the different conceptions of meaning. It is a species of idealism that has to be stated in terms of a limit of the world, which is not itself in the 1 Williams (1981c), p. 163. 2 The term ‘transcendental idealism’ was coined by Kant, and defined by him as a particular doctrine about space and time (Kant (1933), A369). The current trend, which I shall follow in this essay, is to use the term in an extended sense that is supposed to arise naturally out of Kant’s. I shall not attempt to redefine it. The essay itself ought to cast some light on how it is nowadays understood. But see Williams (1981c), pp. 148–149.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

72

Language, World, and Limits

world. In the Tractatus, this limit is the metaphysical subject, or transcendental self— what Wittgenstein means by ‘I’.3 In the later work, Williams urges, it is a plural descendant of that: the essential shift is from ‘I’ to ‘we’ (and from ‘my language’ to ‘our language’). But there is a predicament inherent in such idealism. Since ‘the world’ is to be understood as containing everything that can meaningfully be spoken of, there is no legitimate way of talking about a limit of it. If we do attempt to talk about such a limit, we shall produce either arrant nonsense or (at best?) irrelevant claims about items within the world, which, in purporting to concern something other than those items, will inevitably include downright falsehoods. We are told explicitly in the Tractatus that such attempts are attempts to say that which cannot be said but which makes itself manifest in (or is shown by) what can be said.4 On Williams’ conception, this predicament is implicit in the later work too. He might have added that it can be found throughout the history of philosophy. This is one of philosophy’s perennial problems: ‘how to put a supposed philosophical truth which, if it is uttered, must be taken to mean an empirical falsehood, or worse’.5 As a predicament issuing from transcendental idealism, it can already be found in Kant. Its impact for him can be made graphic through the question: do objects in space and time depend for their existence on our perceptions?6 It seems to be a crucial feature of Kant’s species of idealism that they do not. For, as he continually reminds us, his transcendental idealism is meant to coincide with empirical realism, a connection for which he argues in several places;7 and by ‘empirical realism’ is meant precisely the view that, through perception, we are aware of a (spatio-temporal) reality outside us which is independent of that perception. Yet there does not appear to be any way of stating Kant’s transcendental idealism which does not stand in direct conflict with this. Kant himself, when expounding his idealism, claims that objects in space necessarily presuppose perception and that they are nothing outside us.8 He resolves this apparent contradiction by appeal to a deep ambiguity in the use of expressions such as ‘outside us’.9 Such expressions may be taken either in an empirical sense or in a transcendental sense, that is either in such a way that we are ourselves construed as objects in the (spatio-temporal) world or in such a way that we are not.10 (In its empirical sense, ‘outside us’ implies ‘in space’.) The claim that objects in space and time are nothing outside us independent of our perceptions is then empirically false but transcendentally true.11 But this is why Kant faces the same predicament as 3 Wittgenstein (1961), 5.632 and 5.641. All unaccompanied references to Wittgenstein’s work will be to this. 4 5.61–5.62. 5 This phrase is quoted from Williams (1981c), p. 163. 6 Cf. Strawson (1966), p. 259. 7 E.g. in Kant (1933), A369 ff. It is interesting to compare this with 5.64 of the Tractatus. 8 Kant (1933), A372–376. 9 Ibid. A373. 10 We must obviously construe ‘empirical’ and ‘transcendental’, so used, as terms of art. Later in this essay, I shall take up the usage again, with a slightly different application. Given that the terms are being used in a special way, we must expect some strange consequences. One such, perhaps, is that the sentence ‘Even if we all believed otherwise, seven plus five would still equal twelve’ must be regarded as expressing an empirical truth. 11 In Kant (1950), his first published work after Kant (1933), Kant shows himself to be very sensitive about being interpreted as the wrong kind of idealist, a fate that he claims has already befallen him. (See the end of the First Part.) But if it is true that two diametrically opposed doctrines have to be stated in the same way, then it is only to be expected that this will generate confusion. Nor does Kant’s suggested remedy,

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

73

Wittgenstein. The transcendental interpretation of these terms is unavailable, a second crucial feature of Kant’s critical philosophy being that there can be no intelligible application of concepts beyond the world of possible experience, or, to bring out the connection of this principle with its Wittgensteinian echo, to anything not in the world.12 Thus Kant’s transcendental idealism shares the fate of the kind of transcendental idealism to be found in Wittgenstein: that it cannot be stated save in terms which, by its own lights, are unintelligible unless construed as embodying straightforward empirical falsehoods. A good case can be made for saying that this fate is also shared by certain antirealist views in the philosophy of language (views which gain much of their impetus from Wittgenstein’s later work). The motivating thought behind these views is that linguistic meaning depends ultimately upon what is manifest in our linguistic and non-linguistic practices, and that the meaning of a declarative sentence must (therefore) not be identified with conditions that may obtain independently of our being able to recognize that they do.13 But truth conditions, as classically conceived, are conditions of this kind. So the traditional identification of the meaning of a declarative sentence with its truth conditions should give way to an identification of its meaning with, say, the conditions that we recognize as justifying an assertion of the sentence. Different kinds of anti-realism are concerned with working out the implications that this has for different areas of discourse. The problem is that, given what it would be to identify the meanings of declarative sentences with their truth conditions in the traditional way (for example, uncritically to apply the law of the excluded middle to them), it is an integral part of our linguistic practice that we have a propensity to do just that. And this is more than superficially embarrassing for any anti-realist. Of course, it is not yet an insuperable difficulty. No anti-realist is, as such, committed to the doctrine that every aspect of our linguistic practice is sacrosanct. Here, none the less, is some indication of why (some) anti-realist views might fall victim to the fate we have been talking about. It may be that any philosophical statement of them is bound to conflict with certain basic claims that are integral to our linguistic practice and that they might be expected to ratify. There is evidence for this in Michael Dummett’s treatment of anti-realism concerning the past.14 He characterizes the dispute between the realist and the anti-realist in terms which, he explicitly acknowledges, ought to be rejected by both parties but especially (and more importantly, from our point of view) by the anti-realist. For, according to his characterization, the anti-realist maintains that ‘the past exists only in the traces it has left on the present’.15 But this is something that can be understood only in realist terms, as betrayed in our standard ways of speaking, and, so understood, it is false. Nevertheless, the characterization does, Dummett continues, ‘succeed in conveying something of the psychological effect of the two opinions’. Should it turn out that, ultimately, only a characterization that his own view should henceforth be known not as ‘transcendental idealism’ but as ‘critical idealism’, seem anything other than cosmetic. The problem is far too deep to be merely terminological. What is significant is that in his later writings, including the second edition of Kant (1933), he states his idealism in these bold but ambiguous terms much less frequently. It is as if he begins to sense the impossibility of shrugging off their empirical interpretation. 12 See e.g. Kant (1933), A696/B724. 13 Cf. Dummett (1978b), pp. 216–217. 14 Dummett (1978d). 15 Ibid. p. 370.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

74

Language, World, and Limits

of this unsatisfactory kind is possible, then the point would be established. Antirealist views would be among those whose fate is that they have to be stated in ways that are not right. In this essay, I hope to draw certain conclusions from Williams’ contention that the predicament is also latent in the later work of Wittgenstein. I shall offer no arguments for the contention, since I am principally interested in its consequences. It does however seem to me to be correct, subject to Williams’ own qualification that the element of idealism in the later work of Wittgenstein which gives rise to the predicament is ‘concealed, qualified, overlaid with other things’.16 The predicament is certainly no more than latent. It arises with none of the dramatic urgency that accompanies it in Kant, for example. Given Wittgenstein’s later conception of the correct way to do philosophy, this is hardly surprising. Philosophy, he tells us, consists of assembled reminders for particular purposes, not of theses.17 So we should not expect to find him directly propounding some philosophical theory, still less one that merits the title ‘transcendental idealism’ and of such a kind that the question of how precisely it ought to be stated should be such a delicate and bedevilling issue. What we do find are hints and suggestions, backed up by platitudes or rather mundane empirical observations, which are intended to remove the allure of what would otherwise be enticingly grand and ambitious philosophical theses, whose perversion of the terms within them is checked by Wittgenstein’s reminders as to how the terms are, and ought to be, used. This is indicative of the therapeutic rôle that philosophy is supposed to play. (It  is instructive to compare this with 6.53 of the Tractatus.) The insights that the later work affords are then bound to resist any direct statement. To try to state them directly would be to succumb to the very temptation that they ought to be counteracting. But this in turn is why the predicament explicitly faced in the Tractatus is also lurking in the later work: we are being induced into a certain way of thinking to which, by the very lights of that work, we cannot give voice. This is Williams’ contention. I think the contention would be impossible to justify, however, were it not for the fact that Wittgenstein himself sometimes flouts his own principles precisely by attempting to give voice to that way of thinking and by attempting to express certain general philosophical truths. The predicament is not entirely dormant. And when, rather than just uttering banalities, Wittgenstein attempts to put the implications of these banalities directly—to say, as it were, what they show—then we can view him as providing elucidations of the kind which, according to 6.54 of the Tractatus, that work consists of: pieces of nonsense that have to be transcended before the world can be seen aright. Perhaps the best examples are to be found in his work on mathematics. He suggests that, in some sense, mathematical truth is determined by what we believe.18 But what we would characteristically say, when engaging in mathematics and not when doing philosophy, is that mathematical truth is in no sense determined by what we believe.19 Again, his claim that ‘3 + 3 = 6’ is a rule as to the way in which we are going to talk,20 if it means anything at all, means something that it is in the 16 18 19 20

Williams (1981c), p. 147. 17 Wittgenstein (1967), Pt. I, §§127–128. E.g. Wittgenstein (1967), pp. 226–227. This point is borrowed from Williams (1981c), p. 163, and slightly adapted. This claim is quoted by G. E. Moore in his (1959), p. 279.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

75

spirit of mathematics to deny, surely not Wittgenstein’s intention.21 Here we do indeed see his theory of meaning, as applied to mathematics, being stated ‘in forms which are required to be understood, if at all, in the wrong way’.

2. Wittgenstein’s later work and its preclusion of a philosophically substantial theory of meaning My principal aim is to establish that there is one extremely important consequence of this: that Wittgenstein’s later work, if right, puts paid to an alluring pipe-dream that philosophers of language have long had. This is the dream of being able to state what the words and phrases in our language mean, so as to reveal, in some philosophically substantial way, what it is to understand them. When described like this, the dream sounds suitably heady. What exactly it amounts to ought to become clearer in due course. But already it is possible for me to sketch what I have in mind. In order to remain faithful to the spirit of Wittgenstein’s later philosophy, one can legitimately theorize about language only by commenting descriptively on various actual practices in which we engage, practices that are already permeated with meaning and that can be described only by one who already knows how to participate in them. That is, one can legitimately theorize about language only from within. As soon as one attempts to transcend that, by conveying the significance of these practices and by stating the meanings of expressions from without, then the predicament that I have been talking about comes to the fore. In terms borrowed from the Tractatus: the meanings of the words and phrases in our language are a limit to our world, revealed in our practices. These meanings (unlike the practices themselves) are not a part of the world, and any attempt to state or analyse them in the way suggested fails precisely by treating them as such. Meanings can only be shown, not stated.22 The matter can also be expressed in terms borrowed from Kant. Just as Kant distinguished between an empirical and a transcendental interpretation of expressions such as ‘outside us’, so too we can distinguish between an empirical and a transcendental interpretation of expressions such as ‘our language’, again according to whether or not we are ourselves being construed as objects in the world.23 In its empirical sense, our language is English (or German, or whatever); in its transcendental sense, our language is roughly our world-view, perhaps even our world.24 This ambiguity arises in connection with a host of expressions, for example ‘how we carry on’, but supremely in connection with the pronoun ‘we’. Obviously this pronoun can be used to pick out a particular group of human beings in the world, perhaps a group that shares some one (empirical) language. But it can also be used as the plural descendant of the ‘I’ of the Tractatus.25 To be sure, Wittgenstein himself would never have recognized 21 Cf. Wittgenstein (1967), Pt. I, §124. There is something to the idea that, when Wittgenstein makes these rather bold claims, he can regard them as meaningful and apposite precisely insofar as they further the aims of the rest of his work. But the idea contains an element of disingenuousness that Wittgenstein himself would not have found acceptable. 22 For links with the Tractatus conception of meaning, cf. 4.022 and 4.12 ff. 23 Cf. n. 10. 24 Cf. Wittgenstein (1967), Pt. I, §241. Cf. also Williams (1981c), p. 154. 25 For links with Kant, cf. Strawson (1966), p. 235.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

76

Language, World, and Limits

these distinctions. And even if they have a viable application when it comes to interpreting his work, we must be extremely wary of supposing that at any point in his later work it is possible to say, without betraying gross exegetical insensitivity, that he is using one of these expressions definitely in one sense rather than the other. The attribution to him of transcendental idealism is far too circumspect for that to be the case. Besides, it is arguable that he often makes capital precisely out of disregarding such ambiguities.26 It is nevertheless useful to have the relevant distinctions ready to hand. We can use them to provide another rather glib formulation of what it is that Wittgenstein’s later philosophy precludes, namely any account of how our empirical language engages with aspects of our transcendental language. These aspects manifest themselves; they cannot be spoken of. To revert to the earlier dictum: meanings can only be shown, not stated. This is still too glib, however. There is, after all, a perfectly harmless and commonplace sense in which nobody would deny that meanings can be stated. One says, for example, that ‘procrastinate’ means the same as ‘put things off ’, thereby stating the meaning of ‘procrastinate’. A German/English dictionary states the meanings of German words and phrases for anybody who knows English. To state the meaning of an expression in this sense is to produce another expression that has the same meaning. Qualms about synonymy aside, there can be nothing philosophically objectionable about the idea of stating meanings in this way. Something else must be at stake in this contrast between showing and stating. Clearly it has something to do with the idea of conveying meanings directly, which in turn, perhaps, calls to mind the idea of laying bare what exactly it is that one knows when one understands a given expression. But we are owed a more precise account than that. The allusion to knowledge helps. The thought that there is anything that one knows when one understands a given expression—I am presupposing that knowledge is here being construed as propositional knowledge—is, it seems to me, fundamentally out of keeping with Wittgenstein’s later conception of meaning and in particular with his conception of understanding as a practical capacity. What is not so clear, however, is the rôle that this thought need play within what I have dubbed the pipe-dream of philosophers of language, as we shall see shortly. It has been customary, in recent work in the philosophy of language, to draw a distinction between two senses of the phrase ‘theory of meaning’. In its wider sense (that used by Williams in the quotation with which I opened this essay), a theory of meaning is a general philosophical account of what it is for the expressions in a language to have meaning, an answer to the question: what is meaning? In its narrower sense (principally associated with the work of Donald Davidson), a theory of meaning is a theory of meaning for a particular language, a body of truths which, in a sense that needs to be made precise, systematically gives the meanings of all the expressions in the language and thereby acts as an answer to the question: what do these expressions mean? If what I am proposing is correct, then we have reason to doubt whether Wittgenstein’s later work could accommodate any theory of meaning in the wider sense. The quotation from Williams is thus not entirely felicitous, 26 Cf. Williams (1981c), pp. 147 and 160.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

77

though the infelicity is as much Wittgenstein’s as his. In order to get clearer on these issues, however, we should temporarily focus on theories of meaning in the narrow sense. The main reason why this will help is that they pose an immediate threat to my thesis, as is evidenced by the nature of much of the interest shown in them.27 For it appears that they can provide us with a way of realizing the dream that I have been talking about without containing anything that has the (on Wittgensteinian grounds) unacceptable consequences that I have suggested would accrue. A theory of meaning for a particular language was loosely characterized above as ‘giving’ the meanings of all the expressions in the language. Less loosely, such a theory is supposed to be a body of truths such that, if anyone had due knowledge of them all, then that person would ipso facto understand the language.28 They (theories of meaning) are certainly intended to give meanings in a more substantial sense than a German/English dictionary. (Possessing all the information in a German/English dictionary would not confer understanding of German on a monolingual speaker of French.) They are, to that extent, intended to do the right kind of thing. We must now broach the question of whether, by admitting the possibility of such theories, we are committed to anything that is repudiated in the later work of Wittgenstein. Given what has been said so far, it is not at all clear that we are. The claim that there could be a particular theory stating knowledge, due possession of which suffices for understanding a given language, does not obviously entail that due possession of such knowledge is a necessary condition of understanding the language, nor even that anybody who understands the language thereby knows anything at all. So we evade that particular Wittgensteinian scruple. Again, it has frequently been argued that such theories need only draw on very Spartan resources, resources that appear incapable of offending against any philosophical qualms about stating meanings or expressing transcendental truths. Typically, such theories are conceived as formal (finitely) axiomatizable theories, whose axioms and theorems treat only of items in the language and undisputed tracts of reality, relating these to one another by means of various semantic concepts founded on the idea of truth (or some kindred idea such as that of warranted assertibility), together with a certain amount of set-theoretical apparatus. No axiom or theorem concerning a given expression need even purport to be a statement of its meaning. It is not in that sense that such theories are intended to give meanings. Indeed, all the resources of such a theory are supposed to be purely extensional and hence capable of being understood and mastered without any prior grasp of the concept of meaning at all (otherwise these theories would lose much of their intended explanatory power). And the theorems of the theory need no more be thought of as expressing transcendental truths than are any of the sentences in the object-language. In short, such theories need contain nothing that violates any Wittgensteinian taboos against trying to express grand philosophical theses. Moreover, current orthodoxy has it that they are particularly apt devices for anybody who is convinced that meanings can only be shown, not stated. For once the smoke has cleared, we are told, the 27 See e.g. Davidson (1984a); Dummett (1975); and various essays in Evans and McDowell (1976a). 28 Cf. Davidson (1984f), p. 171; Evans and McDowell (1976b), p. ix; and McDowell (1977), p. 159.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

78

Language, World, and Limits

sense in which such a theory will give meanings is that its theorems will serve to show the meanings of the words and phrases in its object-language.29 We should beware of trying to read too much into the orthodoxy, however. It involves a use of the word ‘show’ that is not identical to that in the Tractatus. According to the latter, a sentence shows its own meaning,30 and this leaves no obvious room for the suggestion that a particular body of truths about the expressions in a language will be specially suited to showing their meanings. Similarly, the kind of thing we want to say on any interpretation of ‘show’ that can be appropriated to fit in with the later work, provided that it rests on a genuine analogy with the Tractatus, is that the meanings of our words are shown in how we use them, or that the meaning of a mathematical sentence, say, is shown in how it is proved.31 This does not issue very naturally in the idea of a theory of meaning, a particular body of truths designed to show the meanings of our words and sentences. The significant characterization of what theories of meaning should do remains the earlier one: they should state truths, due knowledge of which suffices for understanding their object-languages. When it is claimed that they should show meanings, what is being claimed is usually just this. But then in order to extract the kind of philosophical interest from them that would be relevant to what I have been calling the pipe-dream of philosophers of language (part of which, remember, was to reveal what it is to understand the expressions in our language), one would surely need to say something about what it would be to have due knowledge of these truths, and about what it would be for things to be the way that one would then know them to be.32 It remains an open question whether one could do this without falling foul of the scruples that emerge from Wittgenstein’s later work. However innocuous the content of the theorems themselves might appear to be, the claims made on behalf of them would come sufficiently close to attempted statements of meaning, in the relevantly substantial sense, for a Wittgensteinian sense of disquiet to be stimulated once again. This is not to suggest that the later Wittgenstein could not have taken such theories seriously (though my own view is that he could not). The point is rather that he could not have taken seriously a certain kind of interest in them. A more radical conclusion would have to be argued on independent grounds. For suppose we deliberately refrain from making such claims about these theories as will leave us open to the charge of trying to express transcendental truths. The possibility then remains open that they can continue to be of interest to us in what they reveal about various relations between language and reality; or about the formal properties of truth; or about logical form; or about structural validity; or about the composition of complex expressions and about how their semantic values are determined by the semantic values of their parts; or whatever. Somewhat more cynically, one might argue that if we hold back from making the relevant extravagant claims about these theories, then we are left with something that cuts so little philosophical ice that it could not possibly come into conflict with any plausible philosophical reflections, still less those in the later work of 29 Cf. Dummett (1981a), p. 227; Evans (1982), p. 26; and McGinn (1982), p. 223. 30 4.022. 31 Cf. Wittgenstein (1967), p. 220. 32 This is a point that Michael Dummett has often insisted upon, e.g. in Dummett (1975), p. 121 and Dummett (1976), pp. 70–71.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

79

Wittgenstein. The counter-argument would have to be, more or less, that there is still something of substance in the very composition of these theories, according to which language is made up of morphemes which stand, atemporally, in certain determinate semantic relations with aspects of the world, on the strength of which the semantic properties of combinations of these morphemes can be determined by fixed rules; and that this in itself would have been anathema to the later Wittgenstein, who sought to draw our attention to the flexibility, indeterminacy, and open-texturedness of meaning, to the creativity of language use, to the disparities between language and a rule-governed calculus, and to the inappropriateness of systematization and generality in theorizing about language.33 Be that as it may, any interest in these theories founded on the belief that, through them, we can realize the given philosophical dream, would ultimately have to be interest in certain powerful claims being made on behalf of them, and this would make the supposed innocuousness of their own resources an irrelevance. We are left with the thought that, given what we find in the later work of Wittgenstein, the dream cannot be realized. Nevertheless, this discussion of theories of meaning in the narrow sense may at least have given us a better idea of what the dream is. It is the dream of being able to construct a theory of meaning in the wide sense, which will enable us to specify, for any meaningful expression in our language, some feature of it (such as the intentions with which it is habitually or conventionally used, or the idea for which it stands in the minds of those who use it, or its contribution to the truth conditions of declarative sentences in which it occurs, or its contribution to the assertibility conditions of declarative sentences in which it occurs), together with some account of what it would be to know that it possessed this feature, which will jointly serve to give its meaning and to provide, if not a literal account, at least a model of how it is understood. (Specifying the feature may require that other expressions be taken into account. The dream is not meant to embrace atomism.) I shall call such a theory an m-theory, and the relevant feature of an expression its m-feature. It is my principal contention, then, that Wittgenstein’s later work, properly thought through, precludes any m-theory.

3. The idea that meaning is a matter of how we carry on Any m-theory compatible with Wittgenstein’s later work would have to incorporate the broad intuition that the meaning of an expression is a function of how it is actually used, or, to put the intuition even more loosely but in a form that we shall come back to, that meaning is a matter of how we carry on. I am going to present for consideration an argument to the effect that the m-feature of an expression would therefore have to be a feature such that no-one actually knew the expression to possess it; no-one’s understanding of the expression would in fact consist in knowing that it possessed that feature. Admittedly, the argument will trade on an uncompromising conception of knowledge involving incorrigibility. This makes it easier, and correlatively less interesting, to argue that a given individual does not have this or that item 33 Cf. Wright (1979), Ch. XV. See also e.g. Wittgenstein (1967), Pt. I, §81.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

80

Language, World, and Limits

of knowledge. But assessing the argument will provide a useful route to understanding (some of) the tension between Wittgenstein’s later work and any belief in the possibility of m-theories. (This is indeed all I shall claim on behalf of the argument. I do not myself wish to endorse it.) The argument is this. On a later Wittgensteinian view, any expression in our language means what it does, in the last analysis, because of how we carry on over time, because of our continuing linguistic and non-linguistic practices. The whole temporal spread is relevant. At any point in time, we may carry on in divergent ways, more than one of which could count as preserving the meaning of the expression. That the expression possesses its m-feature must therefore entail that it is subject to continuing practices of a certain kind. Thus, anybody who knew that it possessed its m-feature would ipso facto be capable of having knowledge of the future, which (on the uncompromising conception of knowledge being presupposed) nobody can have. Take the word ‘green’. Its meaning what it does entails that it stands in a certain relation to green things, which we may express, roughly, by saying that it applies to something at a given time if and only if that thing is then green, or, even more roughly, that it applies to green things. (What exactly this amounts to need not detain us now. Suffice it to remark that the notion of application that appears here will be some refinement of our intuitive notion of application, subject to the proviso that it is to be understood extensionally. Thus ‘green’ also applies to things whose colour can be obtained by mixing blue and yellow paint.) But it is purely contingent that this is so. Had we carried on differently, ‘green’ could have applied to blue things, or to nothing at all, or to ‘grue’ things, where a thing is ‘grue’ at a given time if and only if either the thing is then green and the time is prior to time t or the thing is then blue and the time is not prior to time t. (Here t is understood to be some time in the future.)34 But if it did apply to grue things, then nothing that has happened in the world so far need appear any different. So it is compatible with everything that anybody knows (incorrigibly) that ‘green’ does indeed apply to grue things. This is true even of people who have been speaking English successfully for many years, who will continue to do so for many years to come, perhaps beyond t, and who believe, quite rightly, that ‘green’ applies to green things. Not even these people know how the word ‘green’ will be used in the future. But a future use of the word ‘green’ in virtue of which it applied to grue things would be equally meaning-preserving, albeit by preserving a meaning other than that which ‘green’ actually has. So nothing that anybody now knows can serve to determine the meaning of the word ‘green’. Its m-feature must lie beyond anybody’s ken. As it stands, this argument is obviously far too quick. But by assessing it, in the light of various possible reactions and objections to it, we shall find ourselves being drawn to the heart of the matter—especially when it comes to adjudging the argument from  a later Wittgensteinian point of view. One natural reaction to the argument rests, I think, on a muddle. Consider (G), understood in the way outlined in the last paragraph:

34 This is an adaptation of the definition first proposed by Nelson Goodman, due to S. F. Barker and Peter Achinstein, in their (1960), p. 511.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning (G)

81

‘Green’ applies to green things.

If we distinguish between the sentence (G) and the claim that it is being used to make, then we find that, however non-trivial the claim may be, the sentence enjoys an indefeasibility which makes the charge of triviality an appropriate one. (It is almost as if (G) is an analytic contingency, although this is a needlessly paradoxical way of putting it.) Whatever ‘green’ applied to, (G) would still be true, provided that all other linguistic conventions were held constant. And by the same token (G*)

‘Green’ applies to grue things

would be false.35 If ‘green’ applied to other than green things, then (G) would express a different truth, and (G*) a different falsehood, but they would be true and false respectively none the less.36 The reaction to the argument that I have in mind is this. Since I clearly have no more insight into the future than anybody else, it follows that, if the comments made about ‘green’ towards the end of the argument are correct, then, for all I know, ‘green’ applies to grue things. Yet surely I cannot acknowledge this possibility without temporarily misusing English (for I can be confident that (G*) is not a true sentence of English, given the points just noted, just as I can be confident that (G) is a true sentence of English, which confidence has been betrayed in the very writing of this essay). But to admit that I cannot acknowledge a possibility without misusing the language in which I attempt to do so is tantamount to admitting that I cannot acknowledge it at all. It follows that either the background view of meaning is wrong or the argument itself is unsound. I think the muddle here can be teased out. Reconsider the possibility that ‘green’ applies to grue things. Suppose that this is the case. Then I for one have a false belief about ‘green’, and use it incorrectly. For I believe that it applies to green things, and this in part informs my usage, for example when I assert ‘Grass is green’ or even ‘I believe that “green” applies to green things’. If there is any fact of the matter as to what claims I take myself to be making when I assert these sentences, then certainly, on the given hypothesis, I am not asserting the right sentences in order to make them. Similarly, I am not using the right sentence even to express the hypothesis. I ought to be using (G) rather than (G*). (This reinforces the idea that (G) enjoys a certain indefeasability.) But none of this rules out the possibility that, for all I know, ‘green’ does apply to grue things. It is just that, if it does, then I have not been (indeed am not) using English correctly. It by no means follows that, just by saying ‘For all I know, “green” applies to grue things’, I must temporarily cease to use English correctly. 35 I am assuming that (G*) is indeed false, and therefore presupposing that it is not the case that something is green if and only if it is grue, which would be the case if, say, nothing were either green or blue at or beyond t. It is also worth pointing out that one of the conventions that has to be held constant, for (G*) to be guaranteed its falsehood, is that ‘green’ and ‘blue’ should not apply to any one monochromatic thing at the same time. 36 Cf. Dummett (1975), pp. 106–107. It is a closely related fact that (G) could never be used to inform somebody that the given relation holds between the word ‘green’ and green things. For somebody who was not already aware of this would not know which fact (G) was being used to state, precisely because the word ‘green’ occurs within it. And by the same token, anybody who did know which fact (G) was being used to state would already appreciate that it was true, which is indeed, rightly or wrongly, often taken to be a hallmark of triviality.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

82

Language, World, and Limits

When I do say that, I am alluding to the possibility that I am misusing the language, but only because the possibility is already there to allude to. I am not representing myself as a possible misuser of the language—as though I were somehow offering the sentence ‘For all I know “green” applies to grue things’ as one which, at least in my mouth, could be taken to express a truth, if not in English then in some idiolect of my own that I mistakenly take to be English. For a variety of reasons, many of them implicit in Wittgenstein’s later work on language, this idea is incoherent. Unless I succeed in saying something in English by uttering the sentence, which is after all what I am trying to do, then I do not succeed in saying anything at all. But conversely, there is no reason to suppose (on the strength of anything said above) that I cannot meet with such success, thereby making a true claim, in English, about the limitations of my knowledge. Let us address another possible reaction to the argument, which takes the form of an objection.37 To say that ‘green’ applies to green things is to make a semantic claim about it which is true in virtue of the meaning that it has now, even though some of the green things to which it applies may be green only when it has ceased to have that meaning. One way of putting the matter (not the only way) would be as follows. The relation of application can be thought of as holding atemporally between the members of certain ordered quadruplets of expressions, languages, times, and things, those, namely, such that the thing satisfies the expression relative to its meaning in the language at the time; and if the expression is ‘green’, the language English, and the time now, then the thing will itself be an ordered pair consisting of a physical object and a time, the physical object being one that is green at the time. (No doubt this should be made more precise, but it will serve our purposes.) Certainly, if ‘green’ now applied to grue things, this would be a difference that concerned future times, but the times would be those after t that occurred in the specified ordered pairs, and not the time occupying third position in the quadruplet, which remains by stipulation as now. It is unclear, then, why the different meaning that ‘green’ would have to have need depend in any way on what is going to happen at those future times. It is also unclear why the actual and current meaning of the word ‘green’ cannot reside in some feature of the word to which we now have access. The fact that nobody knows how this word will be used in the future is irrelevant. That may go to show only that nobody knows whether it will continue to possess this feature and thus retain its current meaning. This objection, like the original argument, is too quick as it stands but points in the direction of something important. It would certainly be decisive against the part of the argument that specifically concerns the word ‘green’, had that part stood alone. But it did not stand alone. It was only intended to illustrate the point that had already been made: granted that meaning itself is a matter of how we carry on, over time, then nothing that anybody now knows can serve to determine the meaning of any expression in our language (that is, no feature that any expression is known to possess can be its m-feature). Of course, one may not want to grant this. But that is beside the point, so long as one is prepared to grant that it is implicit in the later work of Wittgenstein. For the argument itself was designed to show merely what that work 37 I am grateful to Simon Blackburn for drawing my attention to this objection and its importance.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

83

will tolerate. (One problem with the objection is thus that it is formulated in terms that are not at all Wittgensteinian.) The objection nevertheless forces us to ask whether the comments specifically concerning the word ‘green’ do succeed in illustrating the core of the argument preceding them; and this question also turns out to be central to any appraisal of the argument from a later Wittgensteinian point of view, which means that we are getting to the heart of the matter. From that point of view, the core of the argument is essentially acceptable. It does, after all, tell against the view that our understanding is a species of propositional knowledge. But the subsequent comments about ‘green’ do not have a Wittgensteinian flavour. On the contrary, they elicit the following objection, which does (in motivation, if not in style). It would be impossible for ‘green’ to apply to grue things if everything else remained the way it is. There would have to be further differences, including differences in the language. For example, either ‘blue’ would have to cease applying to blue things, or the convention that ‘green’ and ‘blue’ should not apply to the same monochromatic thing at the same time would have to be relinquished. But, the objection runs, various apparently extralinguistic features of reality would have to be radically different too. If, after t, we found that reality did not conform at all well with those express prior expectations of ours that involved the word ‘green’, such as the expectation embodied in the sentence ‘Grass is green’, then, as certain familiar philosophical arguments testify, we would no longer be able to use the word for purposes of successful communication; the word would cease to have any meaning, and it would certainly not be true to say, at that stage, that it applied to grue things.38 But for reality to conform with these expectations, while ‘green’ applied to grue things, would involve marked differences from what is actually the case. For example, grass would need to be blue after t. But now the objection can be pushed a little further. Suppose that reality did conform with enough of our expectations for successful communication to be possible, so that ‘green’ still applied to grass after t but that up to the present everything remained the way it actually is. We have no firm grip on what it could mean to say that ‘green’ applied in such circumstances to other than green things. In these hypothesized circumstances, our continuing practices (how we carry on) would in all relevant and non-question-begging respects be the same as our actual continuing practices. But what it is for something to be green, and how we carry on, particularly with our use of the word ‘green’, are far too delicately interwoven for us to be able to say, without further ado, that these would be circumstances in which ‘green’ applied to things that were blue beyond a certain point in time. On the contrary, to imagine our practices starting out just as they actually have, and then continuing to issue in the same degree of successful communication but in such a way that the meaning of the word ‘green’ is not thought of as changing, just is to imagine, inter alia, that the word ‘green’ should continue to apply to green things. Admittedly, in saying that there is nothing I know that rules out the possibility that ‘green’ applies to grue things, I may be attempting to 38 This leaves open the possibility that t should in any case be some time after the word ‘green’ has ceased to be used with any meaning, say millions of years after the demise of all natural languages. But, for reasons that will eventually become clear, Wittgenstein’s later conception of meaning certainly precludes the thought that, for all anyone knows, ‘green’ applies to those grue things.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

84

Language, World, and Limits

express the perfectly correct thought that nothing I know rules out the possibility that I have a mistaken belief about the application of the word ‘green’ which may eventually come to light. But then, why not just say that? To say that, for all I know, ‘green’ applies to grue things is to say something altogether different. It is to say something absurd. Thus the objection. If it is correct, it has one immediate consequence: given Wittgenstein’s later conception of meaning, the mere fact that ‘green’ applies to green things cannot constitute its m-feature (for no feature of an expression that it is known to have can constitute its m-feature, whereas enough is known to guarantee that ‘green’ applies to green things); whatever exactly is involved in knowing that ‘green’ applies to green things, such knowledge could not possibly model, or act as any kind of surrogate for, understanding the word. It is revealing that the objection has this consequence. It highlights how it (the objection) stands with respect to what was pretty much an unargued assumption of the original argument, namely that it is a purely contingent fact that ‘green’ applies to green things. A distinctive kind of doubt has been cast on how this assumption was exploited. But we have to be very careful when elaborating what is involved here. Nothing has been said to suggest that the contingency claim is straightforwardly false. To be sure, when one looks at a sentence like (G) in a suitably Wittgensteinian frame of mind, one is tempted to see the sentence’s own triviality (whatever that might consist in) as necessitating some kind of triviality for any use to which it might be put. Why, after all, would one want to assert a thing like that? Not to impart substantial information.39 But it is a long way from noticing this to denying that the sentence expresses a contingent fact or even that it has perfectly legitimate uses. One might, for example, assert it simply as a way of emphasizing that the application of the word ‘green’ can only be satisfactorily specified by using ‘green’ itself in this essentially unilluminating way (that we do not have the linguistic resources to offer a weightier account of how ‘green’ relates to the world, that ‘green’ means what it does and not another thing). The contingency claim has not itself been denied outright, then. Indeed, denying it outright (insisting that ‘green’ necessarily applies to green things) would be tantamount to denying either the evident arbitrariness of how we use the word ‘green’ or the still more evident fact that being green has nothing to do with the English language. One could, however, deny that it is appropriate to bring the contingent/necessary distinction to bear on such a proposition in the first place, or at least to do so without qualification; and certainly, less dramatically, one could challenge what seems to be a corollary of the view that the proposition is not necessary, namely that any English-speaker can, in principle, form a conception of what it is for something to be green completely independently of considerations about how the word ‘green’ is used. It is something like this that has happened above. What has been denied is that the claim that ‘green’ applies to green things has any real bite. It is this that lends the objection its Wittgensteinian flavour. Before we assess more carefully what this amounts to, we can profitably look at its repercussions for any attempt to exploit, as the basis for an m-theory, formal theories of the kind discussed above, those conceived as theories of meaning in the narrow 39 Cf. n. 36.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

85

sense. I am thinking more specifically of theories that are expected to include, as axioms, formal counterparts of sentences such as (G). Relatedly, and more fundamentally, they are theories whose theorems will include suitably formalized embellishments (with due account taken of relativization to times, speakers, language, and so forth) of sentences such as the following: (T)

‘Grass is green’ is true ↔ grass is green.40

Although this biconditional appears banal to the untrained eye, it is, again, more or less a philosophical commonplace nowadays that it can be used to state a purely contingent fact. It is a fact that takes some discovering, that is capable of being forgotten, and that the majority of people, having no English, do not even believe.41 For the sentence ‘Grass is green’ to be true is, at least in part, for English-speakers to indulge in a certain kind of linguistic behaviour. For grass to be green, by contrast, is something wholly extralinguistic. The fact that the sentence and the stuff stand in this particular relation to each other is as non-trivial as the fact that the sentence ‘Snow is black’ stands in the same relation to blood. Or such is the orthodoxy. But even if the contingency claim itself is unassailable, we can now give voice to a strong Wittgensteinian reservation about this way of justifying it: it is not at all clear that for grass to be green is wholly extralinguistic. Another characteristically Wittgensteinian misgiving about the orthodoxy is that the predicate ‘. . . is true’ is most naturally taken as a purely disquotational device (so that calling a declarative sentence true is just an indirect surrogate for coming straight out with it). If that is so, then the left-hand side of any biconditional such as (T) will be too close in content to the right-hand side for the biconditional to pack any real punch. But the reservation expressed above (as it were, the reverse of that) cuts much deeper, and (unlike that) cannot be quelled by some such simple device as construing the predicate ‘. . . is true’ as a technical term belonging only to a metalanguage. Of course, as soon as one attempted to develop and defend the claim that due knowledge of a theory of this kind for English would suffice for understanding the language, one would ipso facto be involved in the construction of an m-theory. To that extent, we have yet to see where such an attempt would falter on a later Wittgensteinian conception of language. But it is already clear that something would be amiss. There would be a real problem in showing that due knowledge of such a theory could so much as approximate to understanding the language. Its theorems, to reclaim what I said above, would lack any real bite.

4. How a philosophically substantial theory of meaning would expose the predicament inherent in Wittgenstein’s transcendental idealism The motivating thought here, implicit in much of Wittgenstein’s later work, is that there is no clear demarcation of the extralinguistic.42 In particular, it is no easy matter 40 ‘↔’ here has its customary meaning as a truth functor standing for material equivalence. 41 Cf. Evans and McDowell (1976b), p. xi. 42 Cf. Wittgenstein (1967), Pt. I, the final paragraph of §136 and §381.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

86

Language, World, and Limits

to disentangle what is involved in something’s being green from what is involved in our using language in certain ways. There is a strong inclination to balk at this, to insist that what it is for something to be green is purely a matter of the world’s being a certain way irrespective of what we get up to, perhaps even that it can be known (only) by focusing attention on a green object and thinking: to be green is to be like that. But for the later Wittgenstein, this would not (and could not) give one insight into what it is for something to be green. Anybody who wants to gain such insight must observe us communicating with one another and exercising various discriminatory capacities that we possess. Not only that. Such a person must understand our acts of communication, see the point of our classifications—in short, become (or already be) one of us. How coherent is this? How coherent is the claim that this is required in order to know what it is for something to be green? It would be perfectly coherent, indeed boring, if, although we seemed to be using the word ‘green’ straightforwardly, we were in fact talking about it and saying what was required in order to understand it. (Perhaps some such illusion is involved in the sentence ‘By 1857, she had become George Eliot’ or, to take a more closely related example, ‘He believes that a carp is a kind of bird’.) If (G) too involved such an illusion, we might even be able to acquiesce in the view that it expressed a necessary and utterly trivial truth. But there is no reason to suppose that we have been subject to any such illusion. More importantly, if the idea of an m-theory were to be defended within the framework of Wittgenstein’s later work, then similar claims would be made in contexts where such an illusion would have to be ruled out. For at some stage in specifying the m-feature of ‘green’ and in saying what it would be to know that it possessed its m-feature, we should need to disclose how it stood in relation to being green. Whatever conclusions we had come to about the status of (G), and whether or not we chose even to make use of it, we should ultimately need to reveal something about the relation of ‘green’ to green things which required that they should be thought of as green things and not just as satisfying the English word ‘green’. M-theories thus bring to the fore precisely such questions as: what is it for something to be green? A Wittgensteinian answer is that it is, at least in part, for us to carry on in a certain way, or for our language to be a certain way; and this answer raises essentially intractable problems. But what are these problems? It is not that whenever we want to say how ‘green’ stands semantically in relation to extralinguistic features of reality, we somehow find that we cannot relate it to anything other than itself, or at any rate to anything not linguistic. We do not find that, and nothing has been said to suggest that we do. After all, we can always use (G). Nor is there a problem to the effect that the only claims we can make about how ‘green’ stands semantically in relation to extralinguistic features of reality are such that knowing them could not, for the later Wittgenstein, amount to anything. For all that has been said in this essay, knowing that ‘green’ applies to green things may, for the later Wittgenstein, amount to quite a lot. The problems lie much deeper. They concern the very coherence of some of the claims that have been made about what it is for something to be green. We need have no qualms about saying that ‘green’ applies to green things. We may even have very good reason to say this— indeed, as I suggested earlier, as a way of emphasizing what else can not be said. But as soon as we attempt to state the meaning of ‘green’, in the sense that has come to

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

87

preoccupy us, or as soon as we attempt to say what it is for something to be green, we encounter (literally) transcendental problems. ‘Transcendental’: that was one of the definienda in the distinction drawn several pages back but subsequently little heeded. Those expressions involving the firstperson plural that were described as ambiguous (‘our language’, ‘how we carry on’, ‘we’) have featured prominently in this recent exegesis of Wittgenstein, and the time has come to broach the question of their interpretation. If the exegesis is not to be regarded as totally insensate, then this question will not admit of any simple answer. One thing, certainly, that has been hinted at several times is that anything as parochial as a straightforward empirical interpretation must in many cases be ruled out. If there is any substance at all to the claim that for something to be green is, even in part, for our language to be a certain way, then our language had better not be thought of as just English. It had better be thought of, to some extent, transcendentally. To discuss what being green consists in, the kind of thing one has to discuss when constructing an m-theory, and to discuss it within the framework of Wittgenstein’s later philosophy, necessarily involves talking about aspects of our transcendental language. But we cannot talk about these aspects. They manifest themselves in our empirical practices. When we try to talk about them, we make claims that are either incoherent or empirically false. That is the predicament inherent in transcendental idealism, and it has now finally been brought to the surface. It has taken the following form. The question of what being green consists in has had to be construed as the question of what being green consists in, construed in some metaphysically ambitious way. It has not, therefore, been possible to construe it as a question in semantics about the meaning of the English word ‘green’, to be answered by citing a dictionary definition or stating its application; nor as a question in anthropology about the classifications we effect in describing our environment, to be answered in terms of human needs and interests (or perhaps the needs and interests of a particular group); nor yet as a question in natural science about the physical phenomena in virtue of which a thing is green, to be answered in terms of pigment, wavelength, or retinas. The question of what being green consists in has had to be construed as a question about the way the world is for us, a question about our world-view. It is not a question about anything we come across; it is a question about our way of coming across things. But then the answer born of the transcendental idealism in Wittgenstein’s later work is that something’s being green consists, partly, in our carrying on in a certain way, and more particularly in our using language in a certain way. It follows that if we did not use language in that way, grass (say) could not be green. But we have no way of understanding this, except as an empirical falsehood. What does the colour of grass (real grass, that stuff out there) have to do with language? By striving to capture, in Wittgensteinian terms, the essence of a reality that is empirically independent of us, we lapse into unintelligibility or crude error. It is not that Wittgenstein’s later work makes philosophical reflection on language completely pointless. Quite the contrary. There remains a wide range of permissible activities that would merit that label: drawing certain distinctions that are not usually recognized, and imposing a certain systematization on our ordinary understanding of words, in order to meet various new linguistic needs; describing particular linguistic practices, in order to gain a clear view of how certain expressions are used, in the

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

88

Language, World, and Limits

hope that this will lead us away from those alluring misuses of them that issue in philosophical perplexity; feeling our way around inside our world-view, in order to gain a sense for where the limits of comprehensibility lie; reflecting on why certain things count as conceptually impossible, for example that 3 + 3 = 7, and thereby coming to a better understanding of the relevant underlying linguistic practices, those involved in counting and doing arithmetic, perhaps.43 But the later work will not tolerate the idea of an m-theory, that is to say a theory of meaning that enables us to state how the words and phrases in our language engage with our world-view. This is not to deny that Wittgenstein himself, in his later work, produces comments that have the flavour of belonging to such a theory. For, as I have already intimated, he sometimes seems prepared, in the spirit of the Tractatus, to say things which, by his own principles, ought not to be said, but which will serve as elucidations—the rungs of a ladder to be thrown away after it has been climbed.44 Still, if what he produces is elucidatory nonsense, it is none the less nonsense. To conclude: it is of the essence of the later work of Wittgenstein to display the limits of our language as the limits of our world.45 Any theory that allowed for a philosophically substantial statement of what the expressions in a language mean would have to make provision for acknowledging this, if it were to be compatible with that work. But no theory could make provision for acknowledging this, for it is possible to acknowledge only what it is possible to say, and it is possible to say only how things are in the world. Similarly, nothing in the world (no fact) encapsulates what is shown to somebody who understands the expressions in a language. The meanings of those expressions cannot be stated, though the expressions can, of course, be used with their meanings. Let us turn for the last word to Wittgenstein:46 The limit of language is shown by its being impossible to describe the fact which corresponds to . . . a sentence, without simply repeating the sentence. (This has to do with the Kantian solution of the problem of philosophy.)47

Postscript for the reprint Although this essay has been reprinted with only a few minor amendments and corrections of typographical errors, there is much in it from which I would now distance myself. In particular, I would now adopt a different stance on the very question of whether there is an element of transcendental idealism in the later Wittgenstein. My attribution of transcendental idealism to him in this essay is both circumspect and qualified (see especially pp. 74 and 75–76). Now I am disinclined to go even that far.48 43 Cf. Wittgenstein (1967), Pt. I, §§122 and 132. 44 Ibid. 6.54. 45 Cf. Wittgenstein (1967), 5.62 and Williams (1981c), pp. 150–151. 46 Wittgenstein (1980), p. 10. 47 Many people made helpful comments on earlier drafts of this essay. I am especially indebted to Simon Blackburn, John McDowell, Leslie Stevenson, Garrett Thomson, and Philip Turetzky. 48 Thus in my more recent (1997), in which I deal with many of the same issues, the very first claim that I make about Wittgenstein, excepting a couple of footnoted remarks, is that he was not a transcendental idealist (see p. 126). (The subsequent context makes clear that I mean the later Wittgenstein. I have not changed my stance on the question of whether there is transcendental idealism in the early Wittgenstein. See my (2013).)

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Wittgenstein and Theories of Meaning

89

What I do still hold is that transcendental idealism has an extremely important part in Wittgensteinian exegesis. This is because, for precisely the reasons that Williams draws to our attention, there is an inducement in Wittgenstein’s later work to embrace transcendental idealism. The point, however, is that the inducement can be resisted; it can be resisted, moreover, in a thoroughly Wittgensteinian way. It arises because there are certain questions that we find it impossible to answer in a Wittgensteinian spirit without embracing transcendental idealism: such questions as, ‘What does being green consist in?’, where this is construed, not as a semantic or scientific question, but as a metaphysical question.49 And it can be resisted by repudiating these questions as ill-conceived.50 This leaves me free to stand by the main thesis of my essay: namely that, insofar as there is an element of transcendental idealism in the later Wittgenstein, his work subverts certain pretensions that philosophers of language have. In fact it leaves me free to go further and say that his work does subvert those pretensions. For what the pretensions are, in effect, are pretensions to answer just such questions. So when I say that there is much in this essay from which I would now distance myself, I have in mind incidental material about where the later Wittgenstein himself stands in relation to such questions, and where he stands in relation to transcendental idealism, rather than any of the main conclusions of the essay. It may help if I give a particular example. At the beginning of section 4 I refer to what I call ‘the motivating thought’ in the immediately preceding material, a thought that I describe as ‘implicit in much of Wittgenstein’s later work’, namely ‘that there is no clear demarcation of the extralinguistic’. In the accompanying footnote I cite Philosophical Investigations, §381—that being the section in which Wittgenstein asks, ‘How do I know that this colour is red?’ and replies, ‘It would be an answer to say: “I have learnt English”.’51 At the time I saw Wittgenstein as taking seriously his own question, understood with a certain philosophical intent. The question can of course be understood in various other ways. On one way of understanding it, it is equivalent to, ‘How do I know that this colour is called “red”?’52 And that, as I now see it, accounts for Wittgenstein’s reply: ‘It would be an answer to say, “I have learnt English”.’ He knows full well that, when the question is posed with the relevant philosophical intent, alluding to this other way of understanding it will not satisfy whoever has posed it. But then his reply serves as an invitation to whoever has posed it to say more about what would satisfy him or her. It is a familiar rhetorical device. It is not a commitment to transcendental idealism.

49 It is the main burden of section 4 of my essay to argue that this is so (i.e. that we find it impossible to answer these questions in a Wittgensteinian spirit without embracing transcendental idealism). I still subscribe to that argument. 50 See further Moore (1997), Ch. 6, §3. 51 Here I follow Williams (see his (1981c), p. 162). 52 Cf. the material on p. 86 of my essay concerning how words are often used to make claims about themselves.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

6 The Bounds of Sense Abstract This essay was written for a special issue of Philosophical Topics on the links between Kant and analytic philosophy. It explores these links through consideration of: Wittgenstein’s Tractatus; the logical positivism endorsed by Ayer; and the (very different) variation on that theme endorsed by Quine. It is argued that in all three cases we see analytic philosophers trying to attain and express a general philosophical understanding of why the bounds of sense should be drawn where they should—but thereby confronting the aporia, which Kant too confronted, that it is impossible to do this without transgressing those very bounds. It is suggested in conclusion that the problem is in trying to attain and express such an understanding; in other words, that such an understanding is attainable, but not expressible.

1. Introduction My title is, of course, a direct echo of the title that P.F. Strawson gave to his famous study of Kant’s Critique of Pure Reason.1 Strawson’s title, as he pointed out in his preface, was in turn a partial echo of a title that Kant himself had considered for the Critique, or at any rate for an embryonic version of the Critique.2 In a letter to Marcus Herz, written in 1771, Kant had told Herz that he was busy on a work which he called The Bounds of Sensibility and Reason.3 Not that this was Strawson’s only reason for his choice of title. The title also, he claimed, ‘[alluded] compendiously to the three main strands in [Kant’s] thought’.4 As he went on to amplify this claim, he indicated how, among other things, his title played on the ambiguity of the word ‘sense’, with its connotations of both experience and meaning. He wrote: In two ways [Kant] draws the bounds of sense, and in a third he traverses them. He argues, on the one hand, that a certain minimal structure is essential to any conception of experience which we can make truly intelligible to ourselves; on the other, that the attempt to extend beyond the limits of experience the use of structural concepts, or of any other concepts, leads 1 Strawson (1966), a study of Kant (1933). 2 Strawson (1966), p. 11. 3 Kant (1999), 10:123, p. 127. Cf. another letter to Herz, written the following year, Kant (1999), 10:129, p. 132, where he said that he had been ‘making plans for a work that might perhaps have the title, The Limits of Sensibility and Reason’. Although Zweig translates the title differently in the two cases, the original German is the same: the word that is rendered first ‘Bounds’ and then ‘Limits’ is ‘Grenzen’. 4 Strawson (1966), p. 11.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

91

only to claims empty of meaning . . . [But Kant] seeks to draw the bounds of sense from a point outside them, a point which, if they are rightly drawn, cannot exist.5

That Strawson managed to accomplish so much in the mere naming of his book prompted one reviewer to write, ‘The title itself is a roguish stroke of genius’.6 My aim in this essay is to explore some of the resonances of that book’s wonderful Kantian title, and to relate these to some key moments in twentieth-century analytic philosophy, one of whose defining features has been the aspiration to draw the bounds of sense. Both experience and meaning have been of fundamental concern to twentieth-century analytic philosophy, but the latter quintessentially so, and my chief concern in this essay will be with the bounds, in particular, of meaning. Henceforth I shall accordingly use ‘sense’ only in that sense. This aspiration of twentieth-century philosophy to draw the bounds of sense has in turn been dogged by the very threat of self-stultification to which Strawson saw Kant fall prey. That threat, at a highly schematic level, is clear. Any attempt to draw the bounds of sense by dividing some metaphorical space into two, that whereof one can make sense and that whereof one cannot, looks as if it must fall foul of the fact that one cannot make sense of the divide itself unless one can make sense of both sides. As Wittgenstein famously says, in his preface to the book whose contribution to this dialectic has arguably been more significant than any other and whose own echoes will have been clear in what I have said so far, ‘in order to be able to draw a limit to thought, we should have to be able to find both sides of the limit thinkable (i.e. we should have to be able to think what cannot be thought)’.7 There are various episodes in the history of analytic philosophy in the last century that have been, in effect, more or less self-conscious attempts to work around this aporia, either by drawing the bounds of sense in some entirely different way, or by drawing them in this way and denying that doing so is self-stultifying, or by drawing them in this way, acknowledging that doing so is self-stultifying, and learning to live with the self-stultification.

2. The Tractatus Let us begin with Wittgenstein’s Tractatus. Among the countless exegetical problems posed by that extraordinary and perplexing book, perhaps the deepest are problems about where it stands in relation to this aporia. All three of the ways just mentioned of trying to work around the aporia have been read into the book. The two extremes, whereby on the one hand Wittgenstein is read as attempting some quite different way of demarcating sense, and whereby on the other hand he is read as attempting to live with the self-stultification, are emblematic of what I shall call ‘new’ readings of the book and ‘traditional’ readings of it.8 New readings of the Tractatus hold the following: 5 Strawson (1966), pp. 11–12. 6 Cerf (1972), p. 601. 7 Wittgenstein (1961), p. 3. 8 I use these labels because they usefully signal two exegetical tendencies. I do not mean to suggest that there is a simple polarization in the secondary literature. Cf. in this connection Sullivan (2003), n. 2 and pp. 214–215. For a traditional reading, see Hacker (1986), esp.Ch. 1; and for new readings, see Diamond (1991) and Conant (1989). For an intermediate reading, see McGinn (1999). And for a quite different reading, see Moyal-Sharrock (2007). I myself try to provide something that is not readily classifiable as either a traditional

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

92

Language, World, and Limits The limit to be drawn is the straightforward limit that separates those signs to which, as a matter of brute historical fact, meanings have been assigned from those signs to which, as a matter of brute historical fact, no meanings have been assigned; and there is not the least self-stultification in characterizing what lies on the latter side of the limit in precisely that way. What gives the project the interest it has, as a philosophical rather than a merely lexicographical exercise, is in part its generality. But its chief interest lies in the fact that there are temptations of a distinctively philosophical kind to see meaning where it is lacking. Wittgenstein’s aim in the Tractatus is to eliminate such temptations. And the way in which he tries to achieve this aim is by producing signs to which such temptations attach; then indulging the temptations to such an extent that they eventually become unsustainable and disappear. Again, there is not the least self-stultification in this. Traditional readings of the Tractatus, by contrast, hold the following: There are, in Wittgenstein’s view, things that cannot be put into words. And what divides these things from the things that can be put into words is something that itself cannot be put into words. Any attempt to say what divides them must therefore issue in nonsense. If this means that any attempt to say what divides them must be self-stultifying, so be it. It does not follow that any such attempt must be a failure. For there may be a special kind of nonsense that is able to serve the very function required here. And indeed Wittgenstein thinks there is. Such nonsense is precisely what he takes the bulk of the Tractatus to consist in. He thinks that the nonsense that he himself produces can help us to apprehend some of the things that cannot be put into words. In particular, it can help us to apprehend what divides the things that cannot be put into words from the things that can. The self-stultification is benign.

It is not my purpose here to arbitrate in this exegetical debate. But I do want to highlight one advantage that new readings certainly have over their more traditional rivals, namely that they chime well with that section in the preface from which I quoted earlier. Significantly, my quotation, though relevant to the structure of the aporia that I was using it to illustrate, was concerned with drawing a limit to thought, not with drawing a limit to meaning. But it is the latter that is currently of concern to us. And precisely Wittgenstein’s point, in the larger context from which I quoted, is that, although the former is impossible—for the very reason given in the quotation— the latter, where attention turns from the object of thought to the expression of thought, and where the limit to be drawn is the limit that separates those signs that are used to express thoughts from those signs that are not, is perfectly possible. For  there is no suggestion in this latter case that the limit to be drawn is a limit between what has one kind of subject matter and what has another. It is a limit, rather, between what has any kind of subject matter and what has none. And we can certainly say what constitutes this limit—without having to say anything about a subject matter about which, in the nature of the case, nothing can be said. Whatever lies on the ‘wrong’ side of this limit ‘will simply be’, as Wittgenstein says, ‘nonsense’.9 reading or a new reading (or perhaps rather that is classifiable, with suitable qualifications, as both): see Essay 18 in this volume. 9 Wittgenstein (1961), p. 3. Cf. Sullivan (2003), esp. pp. 209–211, to which I am much indebted.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

93

Here is another way of approaching these issues—in terms of different kinds of possibility. Intuitively, some kinds of possibility are strictly subsumed by others. Thus whatever is technologically possible is physically possible, but not vice versa; whatever is physically possible is mathematically possible, but not vice versa. The natural picture here is that of a series of concentric circles, in which larger circles include possibilities that smaller circles exclude. To be sure, there are some kinds of possibility, notably those that have a cognitive element, that may not fit this picture. These kinds of possibility may cut across the others. For instance, a given physical possibility may qualify as such for reasons that are too complex or subtle for us fully to grasp and may thereby count as an impossibility of another, cognitive kind that in other cases extends beyond the physical: say, the imaginable.10 Furthermore, it may be that none of these kinds of possibility admits of a precise characterization and that what separates any two is always contestable. But still, the picture of a series of concentric circles, applicable in at least some central cases, is an intuitively compelling one. And it is often the means we use to indicate what a given kind of possibility excludes, or at least some of what it excludes: we say that certain things are not possibilities of that kind, by first identifying them as possibilities of some more inclusive kind. Thus a politician may say, adverting to what is technologically possible, ‘There are some ways of improving the safety of our railways that are unaffordable.’ A botanist may say, adverting to what is physically possible, ‘There are some temperatures below which plant life is unsustainable.’ (The politician is not vindicated by the technological impossibility of a completely failsafe automated signalling system; nor the botanist by the physical impossibility of any temperature below absolute zero.) Now it is natural to suppose, further, that there is one kind of possibility that subsumes all the rest. We might call this ‘ultimate’ possibility, or ‘logical’ possibility. True, there are various reasons why this idea of an all-inclusive kind of possibility is not as straightforward as it appears, even when complications concerning possibilities of a more or less cognitive kind are set aside. Thus consider the fact that the impossibility of a point in the visual field’s being both green and red, though logical in a loose sense, does not depend on the meanings of what would standardly be recognized as logical constants.11 But, to the extent that we are entitled to think in terms of one allinclusive kind of possibility, we have a further compelling illustration of the aporia that afflicts the drawing of the bounds of sense. For we obviously cannot say, except as a kind of joke, that what this all-inclusive kind of possibility excludes are possibilities of such and such another kind, as it may be the ‘illogical’ possibility that grass both is and is not green. This all-inclusive kind of possibility is not just another circle in the space we have been considering. It is the space we have been considering. To delimit it requires something of an altogether different kind. It requires an ascent, however indirect, to the metalanguage, whereby certain combinations of words can be said not to represent possibilities at all. They can be said, in one good sense of the phrase, not to make sense. This accords with Wittgenstein’s shift of attention in the preface to the 10 Kripke (1980) is a classic discussion of a variation on this theme. See also Edgington (2004): but see her n. 12, where she says that, although the metaphysically possible and the epistemically possible cut across each other, ‘one has to search hard for examples, which are rather contrived and on the whole not very important or interesting, of the metaphysically possible which is not epistemically possible’. 11 Contra Wittgenstein: see Wittgenstein (1961), 6.3751.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

94

Language, World, and Limits

Tractatus away from thought to the expression of thought. It also accords with various remarks in his later work. At one point, in Philosophical Grammar, commenting on the infinitude of the sequence of cardinal numbers, Wittgenstein insists that we should not say, ‘There is no last cardinal number’—as though we were excluding some possibility—but should rather say, ‘The expression “last cardinal number” makes no sense.’12 And in Philosophical Investigations, having remarked that ‘essence is expressed by grammar’, he says of a puzzle that he is wrestling with there, ‘The great difficulty here is not to represent the matter as if there were something one couldn’t do.’13 This shift of attention from the object of thought to the expression of thought, or from possibilities to the representation of possibilities, does seem to provide a nonself-stultifying way of drawing the bounds of sense. (Of course, there remains the exegetical puzzle of why Wittgenstein should nevertheless have stepped beyond those bounds when trying to execute this project in the Tractatus; but we have already seen how new readings of the Tractatus address this puzzle.) The basic procedure of drawing the bounds of sense by distinguishing, not between that of which sense can be made and that of which it cannot, but between that which has sense and that which does not, appears unimpeachable. But is it? I alluded earlier to the fact that the project is supposed to be a philosophical project, not a merely lexicographical one. It will of course count as a philosophical project insofar as it provides for a general account of what it is to have sense and does not merely, and literally, issue in a combined lexicon and grammar for some particular language. Nevertheless, if it is to have the kind of philosophical point that drawing the bounds of sense had for Kant, it must do more than that. For one thing, it must combat various illusions of sense. According to new readings of the Tractatus, that is precisely what the project, as executed there, does do; or at least, that is precisely what Wittgenstein intends it to do. The way in which Wittgenstein tries to realize this intention, according to new readings, is by so presenting various illusions of sense that they eventually disappear. But is there not more to it than that? Should there not be more to it than that? Surely our philosophical aspirations are not going to be satisfied except insofar as we have some sort of diagnosis—some sort of explanation for why various assignments of meanings to signs appear to confer sense where they do not—and, more generally and more significantly, except insofar as we have some general philosophical understanding of what assignments of meanings to signs can achieve and what, despite appearances, they cannot.14 How clear is it that we can attain and express such an understanding, 12 Wittgenstein (1974), p. 465. 13 Wittgenstein (1967), Pt. I, §§371 and 374, first and third emphasis his, second emphasis mine. Cf. in this connection Wittgenstein (1961), 4.113–4.116 and Williams (1981c), pp. 159–160—where in each case there is an image of working outward from within the space of sense toward its ‘edge’, which, if it means anything at all, surely means (something like) producing combinations of words that make sense with a view to registering the point at which similar combinations of words fail to make sense. Note: the failures to make sense that are of concern here may quite properly be distinguished from other failures to make sense. Thus, even in Wittgenstein (1961), where Wittgenstein is impatient with various attempts to discriminate between ways of failing to make sense (see 5.473 and 5.4733), he himself distinguishes between lacking sense and being nonsensical (see 4.461–4.4611). This observation may in turn help Wittgenstein to treat as an ally someone who is prima facie a foe: I am thinking of Deleuze and his comments in his (1990b), p. 35. 14 Cf. Sullivan (2003), pp. 211–212.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

95

itself a project in drawing the bounds of sense, without self-stultification? How clear is it, for instance, that we can attain and express such an understanding without identifying various things that cannot be expressed, and that cannot therefore be suitably identified, no matter what assignments of meanings to signs we make? How clear is it, for that matter, that this is not Wittgenstein’s project in the Tractatus?15 At the very beginning of his book Wittgenstein writes, ‘The world is all that is the case. The world is the totality of facts, not of things.’16 This is connected to his subsequent remark that ‘propositions can only say how things are, not what they are’.17 Thus although we can name things, and thereby speak about them, we cannot put them into words, or express them by means of propositions, in the way we can facts. These remarks help to combat the temptation, real enough in my view, to express what one understands in knowing the meaning of a word by casting it in propositional form, as though what one knew were that something is the case. Such remarks also have the kind of diagnostic generality to which I was referring in the previous paragraph. Yet the claim that the world consists of facts, not of things, stands in direct violation of something that Wittgenstein says later in the book, echoing his admonishment in the preface against drawing the limits of thought: [We] cannot say in logic, ‘The world has this in it, and this, but not that.’ For that would appear to presuppose that we were excluding certain possibilities, and this cannot be the case, since it would require that logic should go beyond the limits of the world; for only in that way could it view those limits from the other side as well.18

So Wittgenstein has found himself falling prey to the very threat of self-stultification that he highlighted in the preface.19 Nor is it clear that he has any other way of conveying his insights into the nature of language. True, he might urge upon us a distinction between effable states of knowledge, such as someone’s knowledge that grass is green, and ineffable states of knowledge, such as someone’s knowledge of the meaning of the word ‘green’; and he might insist, without any obvious self-stultification, that only knowledge of the former kind can be expressed by what has sense, or in other words can be expressed at all. But if he did thereby manage to avoid self-stultification, then he would do so only at the price of vacuity. He would still not have done justice to his own insights into what an abortive attempt to express knowledge of the latter kind would be an abortive attempt to do—insights, that is, into what would motivate the attempt. It remains unclear whether he has any way of conveying these insights 15 This paragraph is a summary allusion to ideas that I have expressed elsewhere. See Moore (2003), esp. 189–190, where similar questions are posed. 16 Wittgenstein (1961), 1 and 1.1. 17 Ibid. 3.221, his emphasis. 18 Ibid. 5.61. An obvious reply on Wittgenstein’s behalf is that he does not in fact say anything in 1.1 that he proscribes saying in 5.61, since he uses the word ‘world’ differently in the two cases: in the former, to refer to the realm of the actual; and in the latter, to refer to the realm of the possible. I incline to the view that he uses it to refer to the realm of the actual throughout the Tractatus; and that what enables him to refer to the realm of the possible in 5.61 is his use of other words and phrases, notably ‘limits’ and ‘in logic’. But even if I am wrong about that—even if Wittgenstein’s use of ‘world’ is ambiguous in the way proposed— what he says in 1.1, with its clearly implied application to any other possible world, is still surely offensive to the spirit of what he says in 5.61. 19 And, for reasons that I shall sketch in section 6, he has found himself endorsing a species of transcendental idealism to boot. The links with Kant are profound.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

96

Language, World, and Limits

without himself making that same abortive attempt. The aporia remains stubbornly in his way, still to be negotiated.

3. Kant Let us return to the very idea of demarcating that of which one can make sense by dividing some metaphorical territory into that of which one can make sense and that of which one cannot. In taking for granted that there is something incoherent about this, we are taking for granted a relatively undemanding conception of what it is to make sense of something. For the thought, presumably, is that (1)

in order to effect such a divide, one must make sense of it,

and (2) in order to make sense of such a divide, one must make sense of what lies on both sides of it. But there are more demanding conceptions of what it is to make sense of something that allow us to resist either (1) or (2). Thus on a conception whereby one does not make sense of something unless one’s understanding of that thing has a suitable grounding in experience, we can resist (1): one might be able to effect such a divide by arriving at an understanding of it that is not in any relevant sense experiential. Again, on a conception whereby one does not make sense of something unless one attains a significant amount of knowledge about that thing, we can resist (2): one might be able to attain a significant amount of knowledge about such a divide without knowing anything, or anything relevantly substantial, about what lies on its ‘far’ side. We need to consider ways of working around the aporia that are variations on one or other of these two themes. It is instructive, first, to ask whether Kant himself might be exonerated in these terms. Consider the opening section of John McDowell’s Mind and World.20 In that section McDowell refers to the famous passage in the Critique where Kant argues that we need both intuitions and concepts in order to know anything and insists that the former in the absence of the latter are ‘blind’ while the latter in the absence of the former issue in thoughts that are ‘empty’.21 ‘For a thought to be empty,’ McDowell comments, ‘. . . would be for it not really to be a thought at all, and that is surely Kant’s point; he is not, absurdly, drawing our attention to a special kind of thoughts, the empty ones.’22 But it seems to me that that is precisely what Kant is doing, or at least what he takes himself to be doing. This is why, elsewhere in the Critique, he insists on the distinction between what we can think and what we can know. What we can think outstrips what we can know precisely because it includes what we can think without intuitions: it includes our ‘empty’ thoughts.23 Granted this distinction, the distinction between what we can think and what we can know, it is entirely reasonable to equate the bounds of sense that Kant wishes to draw with the bounds of what we can know. It is entirely reasonable, in other words, to accredit Kant with a relatively demanding 20 McDowell (1994). 21 Kant (1933), A50–51/B74–75. 22 McDowell (1994), pp. 3–4. 23 E.g. Kant (1933), Bxxvi, note; B146; B166, note; and A771–772/B799–800.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

97

conception of what it is to make sense of things, whereby we can make sense only of what we can be given in intuition.24 This in turn allows him to work around the aporia in the first of the two ways suggested in the previous paragraph. Kant can freely admit that the distinction he has drawn between that of which we can make sense and that of which we cannot—between how things appear to us and how they are in themselves, basically—is not itself something of which we can make sense but, along with other things of which we cannot make sense, is something that we can quite legitimately think. Thus Kant writes: That we can . . . have no knowledge of any object as thing in itself, but only in so far as it is an object of sensible intuition, that is, an appearance . . . is proved in the analytical part of the Critique. Thus it does indeed follow that all possible speculative knowledge of reason is limited to mere objects of experience. But our further contention must also be duly borne in mind, namely, that though we cannot know these objects as things in themselves, we must yet be in a position at least to think them as things in themselves; otherwise we should be landed in the absurd conclusion that there can be appearance without anything that appears.25

Does this show that Kant does after all have a satisfactory way of drawing the bounds of sense? Well, no; that conclusion would be precipitate, not least because there is reason to doubt his distinction between what we can think and what we can know.26 And even if we grant Kant that distinction, there is reason to doubt whether his handling of it is as careful as it should be. For example, he claims to have ‘proved’ that we can have no knowledge of things in themselves. But would such a proof not issue in knowledge of the very sort it is meant to preclude?27 However that may be, we are not yet in a position to say that Kant has a satisfactory way of drawing the bounds of sense. But we can say that he avoids the immediate structural threat of selfstultification that constitutes the aporia with which we have been concerned.

4. Logical positivism Let us now return to twentieth-century analytic philosophy, where there is a notable attempt to draw the bounds of sense which looks as though it might retain this very advantage. Indeed it looks as though it might retain the advantages of both the principal attempts to draw the bounds of sense that we have considered so far while avoiding the defects of either. It shares with what we found in Wittgenstein a metalinguistic focus on the distinction between that which has sense and that which does not. But it 24 Cf. Kant (1933), B148–149; A239–240/B298–299; A247/B304; A696/B724. 25 Kant (1933), Bxxvi, his emphasis. Note that I am quoting from Kemp Smith’s translation. Guyer and Wood, in Kant (1998), have ‘cognition’ in place of ‘knowledge’ (the original German being ‘Erkenntnis’). This might be thought to forestall the criticism that I am about to level against Kant—on the grounds that, once a suitable distinction has been drawn between cognition and knowledge, some of what I am about to say about the latter does not apply to the former. While I have considerable sympathy for this view, my own view, which I shall here simply state and not defend, is that what I am about to say about knowledge does apply to what Kant means by ‘Erkenntnis’. 26 Cf. Moore (1997), pp. 139–140; see also pp. 250–251. 27 If Williamson (1996) is right, then there is even a question about whether Kant is entitled to make assertions about what we can know nothing about: Williamson argues that, in asserting something, one represents oneself as knowing it.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

98

Language, World, and Limits

also ventures a general philosophical account of why this distinction needs to be drawn where it does, and here it shares with what we have just found in Kant a relatively demanding conception of what it is to make sense of something. Each of these can, in ways that we have seen, serve to keep self-stultification safely at arm’s length. The position I am thinking of is the logical positivism that finds popular and forthright expression in A. J. Ayer’s Language, Truth and Logic.28 The principle of verification enunciated in that book, whereby a statement has ‘literal meaning’ if and only if it is either analytic or, in some suitably refined sense, empirically verifiable,29 is a way of distinguishing between that which has a kind of sense and that which lacks it— which provides one sort of protection against the threat of self-stultification. But the principle allows for all sorts of meaningful statements that lack this kind of sense. Most notably, it allows for statements of value, whose meaning is of an altogether different kind, namely to express feelings and/or to prescribe courses of action.30 To this extent such positivism works with a relatively demanding conception of what it is to make sense of something—which provides a different sort of protection against the threat of self-stultification. But how exactly is this latter sort of protection to be implemented? We saw two models in the previous section: drawing the boundary around sense in a way that does not involve making sense of that boundary, contra (1); and drawing the boundary around sense in a way that does involve making sense of that boundary but does not involve making sense of what is on the ‘wrong’ side of it, contra (2). Which of these do advocates of such positivism profess to be doing? This choice is, in effect, a choice about how to answer the question, ‘What status does the principle of verification itself have?’, a question that is often posed as a challenge to such positivism. Both alternatives are reckoned to be unattractive. But actually both alternatives are (prima facie) attractive; and advocates of such positivism, if they experience any embarrassment with the question at all, are liable to experience the embarrassment of riches. The first alternative, whereby drawing the boundary around sense does not involve making sense of that boundary, is to regard the principle as a literally meaningless prescription about how to use the expression ‘literal meaning’. The second alternative, whereby drawing the boundary around sense does involve making sense of it, but in a way that is innocuous, is to regard the principle as an analytic truth validated by how the expression ‘literal meaning’ is already used, at least by philosophers party to the relevant disputes. In Language, Truth and Logic it is unclear which of these tactics Ayer takes himself to be adopting. He calls the principle a ‘definition’, but insists that ‘it is not supposed to be arbitrary’.31 Later he makes clear that he took himself to be adopting the first alternative. He says that he was never tempted to regard the principle as either empirically verifiable or analytic, then continues: Happily not everything that the verification principle failed to license was cast by me on the pyre of metaphysics. In my treatment of ethics, I made provision for prescriptive statements . . . . Accordingly, in . . . Language, Truth and Logic, I treated the verification principle as a prescriptive definition.32

28 Ayer (1971). 32 Ibid. p. 149.

29 Ibid. Introduction.

30 Ibid. Ch. 6.

31 Ibid. pp. 20–21.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

99

So far, one might think, so good. However, as Ayer also goes on to remark, there remains the question of why the prescription should be obeyed. ‘I evaded this awkward question,’ he writes, ‘by defying my critics to come up with anything better.’33 But why is the question awkward? More particularly, why is it awkward for Ayer? Not because he has nothing to say in answer to it. There is plenty that he might say, and that would be consonant with his overall view of these matters, most pertinently that only when a statement has what he calls ‘literal meaning’ is there any such thing as determining its truth or falsity; indeed, only then that it is either true or false. What makes the question awkward for him is something that Michael Dummett forcefully argues in the essay to which Ayer is replying when he makes these remarks: namely, that no answer he gives will be fully satisfactory unless and until it is placed in the context of some general semantic theory that is of just the kind that he wants to cast ‘on the pyre of metaphysics’.34 Any such theory must include a philosophical account of what, if anything, enables the truth or falsity of statements of various kinds to be determined, and of how this in turn relates to whether statements of those kinds are true or false. If, for instance, the presence of an evaluative element in a statement prevents its truth or falsity from being determined in any way, and thereby ultimately prevents it from being true or false, then the theory must indicate why—which is as much as to say that it must engage with the ‘metaphysics’ of value. This is enough to constitute a significant ad hominem point against Ayer. More importantly, when combined with the worries expressed above in section 2 about the project of attaining and expressing a general philosophical understanding of what any assignment of meanings to signs can achieve, it indicates that there is still a threat of self-stultification to be negotiated. Logical positivism may not provide us with the best of both the Wittgensteinian world and the Kantian world after all.

5. Quine In section 3 I voiced a worry about Kant’s own way of drawing the bounds of sense. This is a worry about his distinction between what we can think and what we can know. The ‘empty’ thoughts that Kant sanctions seem to me (just as they seemed to Strawson35) to be too ‘empty’ to do the work that he requires of them. Still, at least in drawing such a distinction Kant indicated one way to avoid the immediate structural threat of self-stultification that afflicted his project. It was something of this same general sort, specifically a distinction of meaning between statements of different kinds, which, momentarily at least, seemed to provide logical positivism with protection against its own equivalent threat. In the work of Quine there is a descendant of logical positivism which is as hostile as its forebear to the excesses of metaphysics and as deeply committed to the links between sense, verification, and experience, but which is also utterly impatient with any such distinctions of meaning. On Quine’s view, if a statement has meaning at all,

33 Ibid. 34 Dummett (1992), esp. pp. 133–134. Ayer partially concedes this point in Ayer (1992), p. 150. 35 Strawson (1966), e.g. pp. 264–265.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

100

Language, World, and Limits

then it is either true or false, and there is such a thing as determining its truth or falsity. ‘Suppose,’ Quine writes, we think of truth in terms of Tarski’s paradigm. The paradigm works for evaluations . . . as well as for statements of fact. And it works equally well for performatives. ‘Slander is evil’ is true if and only if slander is evil, and ‘I bid you good morning’ is true of us on a given occasion if and only if, on that occasion, I bid you good morning . . . . There are good reasons for contrasting and comparing performatives and statements of fact, but an animus against the true/false fetish is not one of them.36

Furthermore, whatever else might distinguish determining that one statement is true from determining that another is, there is a holistic interdependence between such things which means that there is never any answer to the question, ‘What empirical evidence is required to verify just this statement?’37 In particular, there is no statement for which the answer is ‘None’—no statement which can be verified irrespective of what empirical evidence there is. That is, there is no such thing as an analytic statement. Hence even the distinction of meaning that Quine’s positivist predecessors wanted to draw within the range of statements that have sense is uncongenial to him. Indeed, it is his hostility to this distinction that is as emblematic as anything of his own brand of positivism.38 Prima facie, then, Quine is in trouble. He is espousing a kind of positivism which involves him in drawing the bounds of sense, but which lacks the very resource that looked as though it might enable someone in his position to draw those bounds nonself-stultifyingly. In fact, however, as I indicated in the previous section, what really carries the threat of self-stultification is the attempt to attain and express some general philosophical understanding of why the bounds of sense should be drawn where they should. But Quine’s refusal to recognize various distinctions of meaning between statements goes hand in glove with a refusal to recognize various distinctions of aim and methodology between explanatory projects: in particular, the distinction between trying to attain and express such a general philosophical understanding and trying to account, in broadly scientific terms, for how, as a result of interactions between us and our environment, some things come to have sense while others do not. As long as understanding why the bounds of sense should be drawn where they should is seen as part of this scientific enterprise, and not as some philosophical propaedeutic to it, then it is not at all clear that it carries any threat of self-stultification. And as long as it is not seen in this way, then it is not at all clear either that Quine will want anything to do with it or that he should. To be sure, Quine may find it harder than he supposes to keep some of his predecessors’ distinctions at bay. There is a revealing section in Pursuit of Truth39 where Quine addresses the question whether the empiricism that underpins his semantic views is itself empirical. Unsurprisingly, he insists that it is. He writes that ‘it is a finding of natural science itself, however fallible, that our information about the world comes only through impacts on our sensory receptors’.40 And he later adds, ‘It would take some extraordinary evidence to [testify to either telepathy or clairvoyance] . . . , but, 36 Quine (1981e), p. 90. 37 Or hardly ever. Quine does acknowledge some rare and artificial exceptions: see Quine (1986e), p. 620. 38 The locus classicus is Quine (1961b). 39 Quine (1992), §8. 40 Ibid. p. 19.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

101

if that were to happen, then empiricism itself . . . would go by the board.’41 Yet he also seems to acknowledge, as an issue quite different in kind from the empirical issue of what elicits or is used to justify any given scientific statement, the issue of why the statement counts as a scientific statement (and thereby counts as having sense). He writes: When I cite predictions [that is, predictions of sensory input] as the checkpoints of science, . . . I see [that] as defining a particular language game, in Wittgenstein’s phrase: the game of science . . . A [statement’s] claim to scientific status rests on what it contributes to a theory whose checkpoints are in prediction.42

Given the context, this last statement can readily be heard as a prescription, such as Ayer took the verification principle to be; or, worse still for Quine, as an analytic truth.43 The important point, however, is that Quine’s drawing of the bounds of sense is to be seen as part of a scientific enterprise and, seen as such, it does not appear to be under any special threat of self-stultification. His semantics is informed by his general world-view and is proffered from a point of immersion within that world-view. Thus just as he couches his empiricism in terms of impacts on sensory receptors, ocular irradiation, and suchlike, so too he couches his positivist conception of meaning in those same terms.44 There is much here to give pause however. I shall mention two worries in particular that Quine’s critics have had. First, there is a worry to which McDowell has given celebrated expression, a worry in which we can hear muffled echoes of Wittgenstein’s insistence that the world is the totality of facts, not of things.45 This is the worry that, by construing our evidence for our world-view as a matter of impacts on our sensory receptors, ocular irradiation, and suchlike, Quine is casting entities that are external to our world-view in a rôle that ought to be filled by entities that are already part of our world-view, namely experiences we have of things being thus and so; and, by the same token, he is representing what ought to be a logical or rational relation, namely the relation between our evidence and the rest of our world-view, as a merely causal relation.46 41 Ibid. p. 21. 42 Ibid. p. 20. There are three things to note here. First, my gloss on ‘predictions’ in the first pair of square brackets is taken from ibid. p. 21. Second, my replacement of ‘sentence’ by ‘statement’ in the third pair of square brackets is simply to bring the quotation into line with my usage elsewhere in this essay: I hope it does not do violence to Quine’s intentions. Third, in the second ellipsis Quine writes, ‘in contrast to other good language games such as fiction and poetry’, which indicates that he is not impatient with all distinctions of usage between statements. 43 But perhaps Quine manages to stop it from sounding like the latter when he subsequently observes that, given evidence for the falsity of empiricism, ‘it might indeed be well to modify the game itself ’ (ibid.). For the significance of this observation, and its relevance to whether his own earlier statement is analytic, see Essay 3, section 1, in this volume. 44 It is largely on this basis that he famously draws the conclusion that, if a question about meaning cannot be answered by adducing empirical evidence, then there is no fact of the matter concerning what its answer is: e.g. Quine (1960), Ch. 2 and Quine (1992), Ch. 3. 45 For McDowell’s own explicit reference to Wittgenstein in this connection, see McDowell (1994), p. 27. 46 See esp. ibid. Afterword, Pt. I, §3. It is interesting to note in this connection how evasive much of Quine’s language is. In Quine (1992), p. 41, he speaks of ‘the flow of evidence from the triggering of the

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

102

Language, World, and Limits

The second worry suggests, conversely as it may appear (but see further below), that Quine has represented the relation between our evidence and our world-view as something too intimate. This is a worry that John Campbell has expressed very forcibly.47 It is the worry that, by construing our evidence for our world-view in terms that depend so heavily on that very world-view, Quine has negated an important principle whose importance, indeed, he himself would be the first to emphasize, namely that our worldview is underdetermined by our evidence. ‘[Given that] the patterns of ocular irradiation have to be described in terms of the physics of the day,’ writes Campbell, ‘how . . . could they be consistent with some rival to the physics of the day?’48 Now one might think that Quine has a perfectly satisfactory riposte to Campbell’s rhetorical question. What matters, one might think, is not how the patterns of ocular irradiation are to be described, but what their content is. Here is an analogy. Imagine a brain in a vat, in a classical sceptical scenario,49 whose subject thinks that he is living the life of a medieval monk. And suppose we have to draw on various principles of computerized neurotechnology to describe what is happening to the brain. It simply does not follow that what is happening to the brain is enough to refute the subject’s impression of what kind of life he is living. Again, suppose we have to use some realist theory about middle-sized dry goods to describe the impact of Samuel Johnson’s foot on a stone. It simply does not follow that this impact is enough to refute Berkeleian idealism.50 Similarly, if we have to use the physics of the day to describe certain patterns of ocular irradiation, it simply does not follow that these patterns are enough to refute each and every rival to that physics. If there were people who, on broadly the same evidence as ours, accepted such a rival, then they could not acknowledge any such phenomenon as ocular irradiation (which of course would be a deficiency by our lights). They would have to tell their own rival story about what empirical evidence they had for their theory. Yet, for all that, their evidence would in fact (by our lights) involve ocular irradiation. There is nothing incoherent in this. Such is how the underdetermination of theory by evidence is bound to be described from a point of immersion in one of the underdetermined theories—the only kind of point, in Quine’s view, from which it can be described.51 Nor does this mean that various pragmatic forces cannot eventually bring us to a point of immersion in one of the rivals to the physics of the day. Campbell suggests that, without an Archimedean point, Quine’s view leads to an unacceptable conservatism.52 But the image of Neurath to which Quine famously appeals is precisely meant to show that this is not so: we can entirely rebuild our boat even while staying afloat in it, provided that we rebuild it plank by plank.53 Is this a legitimate reply on Quine’s behalf to Campbell’s rhetorical question? Only on one absolutely fundamental assumption: that it makes sense in Quine’s terms to talk about the ‘content’ of our evidence. If it does not, the question whether our evidence refutes this or that theory cannot so much as arise for Quine. But this now brings us back to McDowell’s worry. For McDowell’s worry is precisely that it does not senses to the pronouncements of science’. The word ‘flow’ here nicely straddles the very divide that McDowell is trying to get Quine to recognize. 47 Campbell (2002), Ch. 11, §5. 48 Ibid. p. 233. 49 See Putnam (1981), pp. 5–6. 50 Johnson famously thought it was enough: see Boswell (1887), vol. 1, p. 471. 51 See e.g. Quine (1960), Ch. 1. 52 Campbell (2002), p. 234. 53 Quine (1960), p. 3.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

The Bounds of Sense

103

make sense, in Quine’s terms, to talk about the content of our evidence; that Quine has construed evidence as a matter of events and episodes that can enter into causal relations but not into logical relations. This is why McDowell thinks that Quine needs a fundamentally new and more commonsensical conception of our evidence whereby our evidence is a matter of how we experience things as being. But really that is Campbell’s point too. ‘Scientific theorizing,’ Campbell writes, ‘can never let go of the idea that it is ultimately our experiences [as of macroscopic physical objects] that have to be explained.’54 The two worries, despite an initial impression of disparity, are of a piece. And they cut deep. For they suggest that there is after all room for some kind of philosophical propaedeutic to science. They suggest that we can legitimately seek a general philosophical understanding of what science is answerable to. Moreover, when this idea is combined with the broadly positivist conception of meaning that Quine favours, with its perceived link between what science is answerable to and what makes sense, then it leads down the very path that I have already identified as the main route to self-stultification: the path of trying to attain and express a general philosophical understanding of why the bounds of sense should be drawn where they should. So it seems that Quine has become yet another example of how not to evade the threat of self-stultification.

6. Conclusion I have suggested more than once that, when drawing the bounds of sense is construed as a philosophical enterprise, or in other words when the aim of the enterprise is to attain and express some general philosophical understanding of why the bounds are to be drawn where they are, then self-stultification looms. There is evidence for this in the Tractatus. And we have seen evidence in the work of various positivists for how hard it is, even with the firmest of resolves, to keep a suitable distance from just such a philosophical enterprise. That there are links here with Kant should be evident to anyone familiar with the accusation of self-stultification levelled against him by Strawson (see section 1). But the links are more profound than that. It is not just that attempts by twentieth-century analytic philosophers to draw the bounds of sense share certain structural defects with Kant’s attempts to do something analogous. They actually lead in the direction of transcendental idealism. When Wittgenstein declares in the Tractatus that the world is the totality of facts, not of things, he is setting the limits of the world as the limits of what can be thought or said.55 That grass is green is part of the world, because it is possible to think and to say that grass is green; but neither grass nor greenness is part of the world, because there is no such thing as either thinking or saying either grass or greenness.56 This is a kind of transcendental idealism. 54 Campbell (2002), p. 234. Cf. also Stroud (1984), Ch. 6, esp. pp. 250–254. 55 I do not mean to deny that he is doing the converse as well. There may be reciprocal dependence here. 56 This is connected with the fact, as I see it, that there is no such thing as saying what it is (in essence) for something to be green (cf. Moore (1997), pp. 134–135, 163–164, and 184). But surely there is such a thing as knowing what it is for something to be green? Perhaps there is. But see in this connection Wittgenstein (1967), Pt. I, §78. (There is also a connection with the remarks made toward the end of section 2 about the ineffability of knowing the meaning of the word ‘green’.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

104

Language, World, and Limits

The worry, of course, is that, inasmuch as a construction such as ‘thinks greenness’ is nonsense, then so too is a sentence such as, ‘There is no such thing as thinking greenness.’ Or, to put the worry somewhat less accurately but with greater rhetorical force, if there is no such thing as either thinking or saying something, then neither is there any such thing as either thinking or saying that there is no such thing as either thinking or saying that thing. Transcendental idealism itself is nonsense.57 If these suggestions are even broadly correct, then it is impossible to attain and express any general philosophical understanding of why the bounds of sense should be drawn where they should. It is impossible, on that ambitious construal of drawing the bounds of sense, to draw the bounds of sense; and any attempt to do so will issue in nonsense. But what follows? Each time that I have referred to what it is impossible to do, I have quite deliberately used the phrase ‘attain and express’: it is impossible to attain and express such a general philosophical understanding. What follows, then— or at least, one thing that follows—is that either it is impossible to attain such an understanding or it is possible to do that, but it is not possible at the same time to express the understanding, in other words the understanding is ineffable. My own view is that twentieth-century analytic philosophy provides the resources to accommodate the latter alternative, and indeed to accede to it. But that is a story for another occasion.58

57 Cf. Wittgenstein (1961), 5.6 ff. The links between Wittgenstein’s early work and transcendental idealism, and other related links, including the link between his later work and transcendental idealism, have been a preoccupation of mine for some time: see e.g. Moore (1997), esp. Chs. 6–9; Moore (2013); and Essays 5 and 18 in this volume, the last of which deals in particular with the threat that accompanies any attempt to categorize something as nonsense of uttering further nonsense (see esp. section 8). The inspiration for much of this is Williams (1981c). 58 See again the material cited in the previous note.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

PA RT I I

The World and Our Representations of it

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

7 A Note on Kant’s First Antinomy Abstract An interpretation of Kant’s first antinomy is defended whereby both its thesis and its antithesis depend on a common basic principle that Kant endorses, namely that there cannot be an ‘infinite contingency’, by which is meant a contingent fact about how an infinite region of space or time is occupied. The greatest problem with this interpretation is that Kant explicitly declines to apply counterparts of the temporal arguments in the antinomy to the world’s future, even though, if the interpretation is correct, such arguments are clearly there to be applied. This problem, it is argued, is surmountable.

There is a way of interpreting the arguments in Kant’s first antinomy1 whereby both its thesis and its antithesis depend on a common basic principle. (I mean that they depend on a common basic principle that Kant endorses. It is uncontroversial that they share a presupposition that he rejects, namely that the physical world exists as an unconditioned whole.2) After specifying the principle—which I will call ‘P’—I shall outline the interpretation in question, attempt to give it some support, and touch on one or two interesting corollaries.3 First, then, what is P? Let us call a contingent fact about how a region of space or time is occupied a ‘contingency’, and a contingent fact about how an infinite region of space or time is occupied an ‘infinite contingency’. A contingency may entail that parts of the region are not occupied at all; it may also be ‘incomplete’ in the sense that there may be parts of the region about whose occupancy it entails nothing. But it must, if it is to count as infinite, have implications beyond merely how some finite sub-region is occupied. P is the principle that there cannot be an infinite contingency.4 Before I proceed to outline the interpretation, I need to highlight certain features of Kant’s conception of space and time, of which P is an essential part. I shall call this 1 Kant (1933), A426–433/B454–461. 2 Kant (1933), A504–505/B532–533). 3 I claim no special originality for the interpretation. See in particular Strawson (1966), Pt. Three, Ch. III, §4, esp. the comment at the top of p. 193. But Strawson is reluctant to press the interpretation because of what seems to me an over-cautious resistance to reading question-begging idealism into Kant’s arguments (e.g. p. 175). I have more to say about this exegetical point below. 4 Cf. Wittgenstein (1975), p. 163; and, for reservations about infinite contingencies of a somewhat different kind, Wittgenstein (1974), p. 456. (Cf. also the intuitionistic recoil in the philosophy of mathematics from the idea of an ‘infinite co-incidence’, that is, the idea that every member of an infinite domain might have some mathematical property though no general reason could be given as to why.) I shall be returning several times to similarities between Kant and Wittgenstein.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

108

Language, World, and Limits

‘the Kantian conception’. The features in question are in fact well-articulated by Wittgenstein in his Philosophical Remarks5 where he endorses a strikingly similar conception. It is worth beginning with three pertinent quotations from there: We all . . . know what it means to say that there is an infinite possibility and a finite reality, since we say space and time are infinite but we can always only see or live through finite bits of them . . . . Experience as experience of the facts gives me the finite; the objects contain the infinite . . . . [No] experience could be too large for [space] or exhaust it: not of course because we are acquainted with the dimensions of every experience and know space to be larger, but because we understand this as belonging to the essence of space. – We recognize the essential infinity of space in its smallest part.6 [What] I see presupposes the possibility of seeing further. . . . I could correctly represent what I see only by an infinite form.7 We are . . . in the same position with time as with space. The actual time we are acquainted with is limited (finite). Infinity is an internal quality of the form of time. . . . Doesn’t it come to this: the facts are finite, the infinite possibility of facts lies in the objects.8

What these quotations draw out so effectively is how the infinitude of space and time can be part of the form of each contingency, even though each contingency is itself confined to a region of space or time that is finite. This broadly characterizes the Kantian conception too. Central to that conception (if not to Wittgenstein’s) is the principle that there cannot be a contingency that is not capable of being experienced, either directly or through successive synthesis. That is, for any contingent fact about how things are in space or time, there must be some possible experience, or some possible accumulation of experiences, of that fact.9 To say that a region of space or time is infinite, however, is to say precisely that there is no possible experience, nor any possible accumulation of experiences, that suffices to survey it exhaustively. It follows immediately that there cannot be an infinite contingency (principle P). But it does not follow that the infinite has no grounding in experience. This is the main thrust of the quotations from Wittgenstein. The point is this: any experience, and any accumulation of experiences, has written into its very form the possibility of further experience, of how things are elsewhere and elsewhen.10 It is in that sense that space and time, the a priori forms of what can be experienced, are themselves infinite, and are known a priori to be such. It is as if each contingency is located in an infinite framework of possibilities, every one of which is implicated in it. But no contingency is itself infinite. To assume the existence of the (physical) world as an unconditioned whole, in other words, to make the assumption that generates the first antinomy, is to grant an infinite contingency—in direct violation of P. It is to grant that there is a contingent

5 Wittgenstein (1975), §XII. 6 Ibid. p. 157, his emphasis. 7 Ibid. p. 160. 8 Ibid. p. 164. 9 Kant (1933), A491/B520 ff. 10 So Kant thinks that whatever we encounter in space or time is conditioned, not just because of what we can encounter beyond it, within it, before it, and after it, but also because the question of what we can encounter beyond it, within it, before it, and after it must so much as arise (e.g. Kant (1933), A487/B515).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

A Note on Kant’s Antimony

109

fact about how the whole of space and time is occupied.11 On the interpretation of the antinomy that I am canvassing, both the thesis and the antithesis are derived by (in effect) exposing a commitment to that contingency in the opposite position, then rejecting the opposite position in accordance with the Kantian conception, in particular, principle P.12 There is an immediate and obvious objection to any such interpretation, which ought to be addressed before we proceed any further. The Kantian conception, especially inasmuch as it supports P, includes controversial elements of his transcendental idealism, most notably, the phenomenalism, or quasi-phenomenalism, which disallows experience-transcendent facts. Yet Kant thinks that the arguments in the antinomy will convince anyone, regardless of prior metaphysical commitments. In fact, he offers them in indirect support of transcendental idealism.13 This objection can be countered as follows. Although Kant does himself endorse the arguments—excepting the original assumption that the world exists as an unconditioned whole—he is not offering them in a spirit of persuasion. He is offering them in a spirit of descriptive rational psychology. He takes them (rightly or wrongly) to be arguments that already force themselves upon us as soon as we think about these issues.14 Furthermore, he thinks that he can explain why this is so; and that the history of philosophy bears witness to the fact that it is so.15 There is no question here of question begging. All that we, as exegetes, need to ask is how likely it is that Kant should think that transcendental-idealist principles play a rôle in our thinking, even before we have recognized them as such. It is extremely likely.16 The arguments for the thesis can certainly be understood in accord with this interpretation. Both an infinitely old world and an infinitely big world would involve an infinite contingency. Admittedly, this should give us some pause. For if, as Kant believes, it is a priori that every event has a cause, then it must also be a priori that history is infinite (in the sense that the process of exploring the chain of causes leading up to any given event must be capable of extending indefinitely far backwards): it must be a non-contingent fact about the form of events, like the fact that time itself is infinite.17 But the point is this: even if it is a priori that history is infinite, the assumption under attack—that there is an infinitely old world, conceived realistically as an unconditioned whole—still involves an infinite contingency. For it entails not only that history is infinite, but also that there is a contingent fact about the course that this history has taken. What the argument for the temporal half of the thesis does is to 11 In a phrase borrowed from Wittgenstein (1979), p. 83, it means viewing the world sub specie aeternitatis, not in space and time but together with space and time. Cf. Kant’s argument for the second half of the antithesis (Kant (1933), A427/B455). 12 But there are, of course, explicable variations of emphasis from one argument to the next. 13 Kant (1933), Bxx and A506–507/B534–535. 14 See e.g. ibid. A339/B397 and A464/B490. 15 Cf. Moore (1988), p. 214. Al-Azm (1972) argues that what Kant principally has in mind are arguments put forward in debates between Newtonians and Leibnizians. 16 If he does, the fact that he also thinks that transcendental-realist principles play a rôle in our thinking merely indicates the confusion to which he takes us to be prone. But it is then not as straightforward as he suggests to adduce these arguments in support of transcendental idealism. We need some reason for rejecting the realist principles rather than the idealist ones. The asymmetry, I take it, is supposed to derive from explaining how we are led astray in the first place. See further Moore (1988), p. 215. 17 See e.g. Kant (1933), A497/B525 ff.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

110

Language, World, and Limits

draw out the incoherence in this contingency—on the Kantian conception.18 Similarly for its spatial counterpart. The arguments for the antithesis can also be understood in accord with the interpretation.19 Prima facie, a finitely old world or a finitely big world would involve only a finite contingency. However, there would be an infinite contingency in the fact that the whole of preceding time, or the whole of the rest of space, was unoccupied. As Kant says:20 Even if we suppose the whole of nature to be spread out before us, . . . yet still through no experience could . . . [the unconditioned whole] be known by us in concreto. For that purpose, . . . we should require . . . a completed synthesis and the consciousness of its absolute totality.21

The temporal argument is concerned primarily with establishing that this contingency would indeed be a contingency: given the homogeneity of time, there could be no (a priori) reason for the world’s beginning at any given time. (And what else could guarantee that nothing had happened before the world began?) Similarly, nothing could secure the world a position in otherwise empty space.22 Most of the effort in the spatial argument, however, is devoted to confirming the inaccessibility to experience of the resultant infinite contingency—and hence, on the Kantian conception, its incoherence. This incoherence is less obvious than in the temporal case, and less obvious than in the case of an infinitely big world. Could one not look out from the edge of a finite world and have a direct experience of infinite empty space? Kant explains why one could not. Space is the form of what is experienced, not its matter. There may indeed be experience of a vacuum, but only as a formal feature of a fact about how some finite region of space is occupied. (And even then, experience could not prove the existence of the vacuum. That would require infinite powers of discrimination.23 Kant indulges vacua only to a very limited extent.24) Nothing could count as experience of the fact that there was infinite empty space. Its postulation is incoherent.25 The fact that this interpretation helps to make sense of the arguments for the antithesis is one of its primary virtues: these arguments have proved particularly stubborn in resisting efforts to understand them. Another virtue of the interpretation is that it fits well with Kant’s own summaries of the arguments and his subsequent glosses on them.26 But its chief virtue, I submit, is that it involves Kant in a fundamentally 18 The argument has frequently been criticized on the following grounds. Kant says that the world’s history cannot be infinite, because it is now over. He appears to overlook the possibility that it can still be infinite in the following sense: there is (or was) no possibility of surveying it in a successive synthesis of experience, however far back that experience may reach (or might have reached). The point, however, is that nothing that was infinite in that sense could be the world’s history: the world’s history, like any other contingency, must be capable of being experienced, either directly or through successive synthesis. It is difficult to cast the argument in a way that does not make it look question-begging because, on the Kantian conception, the principal inference is such an immediate one. 19 Of course, if it is a priori that history is infinite, then there is another (quite different) argument for the temporal half of the antithesis. But this is not the argument that Kant thinks most immediately forces itself upon us (perhaps because it so readily calls to mind the argument for its contrary). So it is not his concern here. 20 Cf. Wittgenstein (1961), 1.11. 21 Kant (1933), A482–483/B510–511, my emphasis. 22 Ibid. A521/B549. 23 Ibid. A172–173/B214. 24 See ibid., A433/B461 note. 25 It is an interesting question how well this squares with what Kant says in Kant (1933), A24/B38–39 (just as it is whether Kant should indulge vacua even to the extent that he does). 26 See e.g. Kant (1933), A486–487/B514–515 and ‘The Antinomy of Pure Reason’, Section 9, §1; and Kant (1950), §52c.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

A Note on Kant’s Antimony

111

correct understanding of the infinite. The concept of the infinite is properly used to characterize the range of possibilities that finite things afford, both collectively and individually; it cannot coherently be used to characterize any one of those possibilities, except insofar as, through their form, the things that occur in that possibility point beyond themselves to the entire range. Kant applies this general scheme to possibilities that concern the occupation of space and time. (I shall not now try to defend this conception of the infinite. It receives possibly its clearest and most forceful expression in that section in Wittgenstein’s Philosophical Remarks from which the three quotations above were taken.) The greatest problem with the interpretation, on the other hand, is that Kant explicitly declines to apply counterparts of the temporal arguments to the world’s future.27 Yet both the supposition that the world will carry on forever and the supposition that it will have an end involve an infinite contingency—which means that, if this interpretation of the temporal arguments were correct, they could readily be converted into arguments against each of those suppositions. I think they can be. But the problem is surmountable. On closer inspection, we see that Kant does not, in fact, deny that such counterpart arguments can be constructed; he merely declines to construct them. There is no reason to suppose that he would not recognize such arguments as being every bit as cogent as those he does construct. In saying this, I am in sharp exegetical disagreement with a number of commentators— with some explicitly, and with others, whose interpretations of Kant’s temporal arguments do preclude the construction of future counterparts, implicitly.28 But I think that the text both supports me and, at the same time, helps to explain why Kant confines his attention in this way. The point is one that I emphasized earlier: Kant is concerned with arguments that he thinks irresistibly impress themselves upon us as soon as we think about these issues. The arguments concerning the world’s past do so, because of our urge to locate something unconditioned in the field of experience grounding what is conditioned. A thing’s temporal conditions necessarily precede it; so our thoughts are ineluctably led backwards. But it is ‘a matter of entire indifference’ to us29 what infinite contingency lies ahead of us. The idea of the unconditioned whole does not, in that case, exercise the same power over our thinking. We are content rather to regard the series of conditioned elements leading on from the present as ‘not . . . given, but only allowing of being given’.30 (Our thinking about the future is thus healthier in this respect than our thinking about the past.) The counterpart problem concerning the world’s future therefore does not arise for us. It is ‘gratuitous and unnecessary’.31 The fact remains that we could, if we wished, supply arguments directly analogous to those in the first antinomy, to show that it is false that the world will carry on forever and false that it will come to an end.32 27 Ibid. A410/B437; cf. 1933, A336–337/B393–394. 28 A commentator of the first sort is Bennett (see Bennett (1982), pp. 180–181)—though I should add that I think his essay casts a great deal of interesting light on asymmetries that exist between the past and the future. Commentators of the second sort are Strawson, in his discussion of the thesis (Strawson (1966), p, 176); and Guyer, in his discussion of the antithesis (Guyer (1987), p. 408). 29 Kant (1933), A410/B437. 30 Ibid. A411/B437–438. 31 Ibid. A411/B438. 32 I have been greatly helped by the comments of an anonymous referee.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

8 Bird on Kant’s Mathematical Antinomies Abstract This essay is concerned with Graham Bird’s treatment, in The Revolutionary Kant, of Kant’s mathematical antinomies. On Bird’s interpretation, our error in these antinomies is to think that we can settle certain issues about the limits of physical reality by pure reason whereas in fact we cannot settle them at all. On the rival interpretation advocated in this essay, it is not true that we cannot settle these issues. Our error is to presuppose that the concept of the unconditioned has application to physical reality. Once this presupposition has been abandoned, we can retrieve sound arguments from the antinomies, not indeed to demonstrate that the views originally being defended are correct, but to demonstrate that the views originally being attacked are incorrect. The essay concludes with some comments concerning how this disagreement relates to a broader disagreement about the best way to understand Kant.

My aim in this essay is to take issue with Graham Bird’s treatment of Kant’s mathematical antinomies in his recent commentary on the Critique of Pure Reason.1 It is an imposing and magisterial commentary, running to over eight hundred pages, every one of which contains significant insights and displays admirable scholarship. My disagreement, in such a context, is minor. That said, at the end of the essay I shall suggest, albeit very inchoately, a way in which this disagreement connects with some reservations that I have about Bird’s fundamental project, which is to repudiate what he calls ‘traditionalist’ interpretations of Kant in favour of what he calls a ‘revolutionary’ interpretation. I subscribe to what (I am fairly sure) Bird would call a traditionalist interpretation2—though I admit to being altogether less clear about the contrast than he is.3 1 Bird (2006): all unaccompanied references to what Bird says will be to this book. All unaccompanied references to what Kant says will be to Kant (1998). 2 See my (2012), Ch. 5. 3 Part of my reason for being unclear about the contrast is that Bird takes P.F. Strawson to be an archexponent of a traditionalist interpretation. But I think that there are some critical respects in which Bird is as unfair in his reading of Strawson as he takes Strawson to be in his reading of Kant. One central issue is how close Kant is to Berkeley: nowhere near as close, in Bird’s view, as Strawson makes him out to be (see e.g. p. 8). It is certainly true that, for Strawson, ‘Kant . . . is closer to Berkeley than he acknowledges’, (Strawson (1966), p. 22). But Strawson never denies the really crucial differences between them: that Kant,

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Bird on Kant’s Mathematical Antinomies

113

My focus, as I have already indicated, is Bird’s treatment of the mathematical antinomies, which constitutes Chapter 26 of his book. I have no special quarrel with his treatment of the details of the antinomic arguments. My concern is rather with his treatment, in §3 of the chapter, of what he calls ‘the character of the conflicts’. As I understand it, Bird’s conception of Kant’s conception is roughly as follows. We are part of a physical reality which is largely independent of us. We can obtain knowledge of this reality, and of what pertains to it, such as our own psychological states, although we cannot obtain knowledge of anything else. All the knowledge that we do obtain relates to experience. But it also involves certain a priori concepts. And we can put these concepts to a further, independent use: to entertain the idea of an experience-transcendent reality, the reality of ‘things in themselves’ such as God. We cannot however obtain any knowledge of such a reality. We cannot even know whether things in themselves exist. But, once we have entertained the idea, it is irresistible for us to speculate about such things, and indeed to try to obtain knowledge of them, by pure reason. Moreover, in certain cases, we are subject to a powerful illusion of having succeeded. In the specific case of the cosmological questions with which the mathematical antinomies are concerned, our mistake consists in: • starting with perfectly legitimate scientific questions about the structure of physical reality; • using certain a priori limit concepts to address these questions; • wrenching those concepts from that use; • using them instead to raise questions about the experience-transcendent limits of physical reality itself; • addressing these questions by pure reason; and • fancying that we have answered the questions.4 There are two points in particular that Bird emphasizes as part of this conception: (1) It is a mistake to think that we can settle issues about the character of the physical universe by pure reason.5 (2) The particular issues about the character of the physical universe that we do think we can settle by pure reason we cannot settle at all: our attempts to settle them constitute ‘an erroneous and undecidable’ pseudo-discipline.6 My own view is that there is a problem in ascribing (1) to Kant, and that, although he takes (2) to be true of the issues that arise in the dynamical antinomies, he does not take it to be true of the issues that arise in the mathematical antinomies. (This indeed marks one of the most important contrasts between the two sets of antinomies.) Let us begin with (1). Before we can properly consider Kant’s attitude to (1), we need to clarify what is meant by ‘pure reason’ in this context. If the term is used in the way in which Kant sometimes uses it—whereby to settle an issue by pure reason is unlike Berkeley, acknowledges material substance; that Berkeley, unlike Kant, acknowledges spiritual substances; and that Kant, unlike Berkeley, insists on the atemporality of things in themselves. 4 See esp. §3.1.

5 P. 671 and, a little more cautiously, p. 674.

6 Pp. 674 and 680.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

114

Language, World, and Limits

simply to settle it a priori7—then there is no question but that he would deny (1). For he would count what he achieves in the three analogies of experience8 as a counterexample. Come to that, he would count what can be achieved in pure mathematics as a counterexample.9 More often, however, and certainly in the passages that are most directly relevant to the antinomies, Kant uses the term ‘pure reason’ in such a way that not only must pure reason be free of any appeal to experience, it must be free of any appeal to intuition as well.10 Then no doubt Kant would accept (1): he would refuse to count anything that can be known by appeal to concepts alone, including anything that can be known by analysis of the concept of the physical universe, as determining the character of the physical universe itself.11, 12 Bird cites, in this connection, A409/B436. This is the passage in which Kant specifies a natural principle of pure reason: that if the conditioned is given, then the unconditioned must be given also. This passage certainly highlights a mistake that Kant thinks we naturally make when we try to address issues about the character of the physical universe by pure reason, perhaps indeed the chief mistake: the mistake, namely, of taking this principle to be a truth about the physical universe. But I am not persuaded that the passage is relevant in the way in which Bird takes it to be. For it does not suggest that the mistake in question is ineradicable; nor, in particular, that the mistake cannot be eradicated by pure reason; nor, crucially, that once the mistake has been eradicated, pure reason lacks the resources to address the original issues.13 It is quite compatible with what Kant says in this passage—even if he would deny it on other grounds—that we can settle issues about the character of the physical universe by pure reason. Be that as it may, my concern hereafter will be with (2), which states that we have no way of settling such issues. That there is something awry with Bird’s attribution of (2) to Kant is reflected in a crucial asymmetry between what Kant wants to say about the truth or falsity of the conclusions in each of the mathematical antinomies and what he wants to say about the truth or falsity of the conclusions in each of the dynamical antinomies. I do not mean the familiar point that he accedes to the falsity of both conclusions in the former case and to the truth of both conclusions in the latter case. I mean the less familiar point that he recognizes a modal contrast too. He says that the conclusions in each of the mathematical antinomies must both be regarded as false; and that the conclusions in each of the dynamical antinomies may both be regarded as true.14 The ‘must’ indicates that he does indeed think we can settle

7 B20; A712/B740; Kant (2002), §6; and Kant (2002), 5: 167, the opening sentence of the preface. 8 A176/B218 ff. 9 A157/B196. 10 E.g. A306/B363 and A750–751/B778–791. 11 Cf. A258–259/B314. I should emphasize that my concern here is only with the exegetical issue. My own view is that Kant should accept (1) even on the first and less attenuated conception of pure reason. Thus contemporary cosmologists tell us that the infinitude of space and time themselves, never mind the infinitude of the physical universe in space and time, is, in part, an empirical matter—something that Kant would strenuously deny. (In fact many contemporary cosmologists take space and time to be finite.) 12 In view of what I say in this paragraph, I take it to be a slip on Bird’s part when he refers to ‘the questionable assumption that we can decide issues about the character of the physical universe by a priori reasoning alone’ (p. 674). He surely means: ‘. . . . by pure reason alone’. 13 Cf. in this connection A425/B453. 14 E.g. A531–532/B559–560; see also Kant (2002), §§52c and 53.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Bird on Kant’s Mathematical Antinomies

115

what to say in the former case: this is reflected in the pithy summary that he gives15 of the argument for transcendental idealism from the reasoning in the first antinomy.16 Bird, significantly, says merely that ‘in the two mathematical antinomies both sides can be wrong’;17 and later, that ‘Kant . . . allows the conclusions to the mathematical antinomies to be both false’.18 This second quotation is not, admittedly, in propria persona; but neither is there any indication that Bird would demur. What grounds this modal distinction? The answer is: another distinction which Kant draws in the ‘Transcendental Doctrine of Method’.19 There he divides the misguided efforts of metaphysicians into two broad classes, according to whether their questions are ill-conceived or well-conceived.20 Kant’s idea is this. These metaphysical questions always involve some idea of reason, that is some a priori concept of the understanding freed of whatever conditions allow it to be applied empirically. Such a question is ill-conceived if it involves a confused amalgam of an idea of reason with some concept that can be applied only to objects of possible experience. It is wellconceived if it involves ideas of reason without any such distortion. In the former case, the question has no answer. Or rather, it has no answer as intended. Thus if the question is which of two apparent contradictories holds, each involving the confused concept, then the answer is neither.21 In the latter case the question has an answer, even as intended, but only at the level of things in themselves, which means that we lack the resources to ascertain what that answer is. The questions addressed in the mathematical antinomies are of the former kind: they are ill-conceived. The questions addressed in the dynamical antinomies are of the latter kind: they are well-conceived. Although my principal concern in this paper is with the mathematical antinomies, I want to take a brief digression to discuss the dynamical antinomies. There is an issue about what exactly Kant takes the arguments in the dynamical antinomies to establish, once the illegitimate inference to how things are in themselves is separated off. Consider in particular the third antinomy. The question there is whether there is any such thing as freedom, and Kant takes the arguments to leave open the possibility that there is, at the level of things in themselves, even though there is not in physical reality. The argument for the antithesis is in this respect fairly straightforward. Kant takes that argument to establish just what it purports to establish: that there is no freedom in physical reality because ‘everything [there] happens in accordance with laws of nature’.22 But what of the argument for the thesis? What Kant himself says, in his remark on it—and unlike Bird23 I take Kant to be speaking in his own voice in his various remarks on these arguments—is this: We have really established [the] necessity of a first beginning of a series of appearances from freedom only to the extent that this is required to make comprehensible an origin of the world . . . But because the faculty of beginning a series in time entirely on its own is thereby

15 A506–507/B534–535. 16 I note also that this argument for transcendental idealism seems to me altogether less secure on Bird’s interpretation, though I shall not pursue that here. 17 P. 674, emphasis added. 19 A740–741/B768–769. 18 P. 681, emphasis added. 20 This terminology is mine, not Kant’s. 21 A503–505/B531–533. 22 A445/B473. 23 P. 693.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

116

Language, World, and Limits

proved (though no insight into it is achieved), now we are permitted also . . . to ascribe to the substances in [different series in the course of the world] the faculty of acting from freedom.24

What Kant takes this argument to show, it seems to me, is that there must be something uncaused grounding whatever happens in accordance with laws of nature. This relates to something else that I think Kant thinks we know: that there must ultimately be something unconditioned corresponding to whatever is conditioned.25 This is of course to be distinguished from the assumption cited above, which Kant certainly holds to be illusory: that there must be something unconditioned and physical corresponding to whatever is conditioned and physical. It is also, in my view, evidence of another error on Bird’s part: the error of holding that Kant denies us any knowledge concerning things in themselves, even knowledge of their existence.26 To return to the mathematical antinomies: the questions addressed in these antinomies are, to repeat, ill-conceived. The question addressed in the first antinomy, for example, involves the concept of the physical universe as a whole. But this is a confused amalgam of the idea of unconditionedness and the concept of physical reality.27 The physical universe as a whole, if there were such a thing, would have to be both physical and all-encompassing. But the only physical things that can exist are objects of possible experience.28 And no object of possible experience can be all-encompassing. That is, no such object can encompass the whole of physical reality.29 The source of our mistake, when we conflate these concepts in this way, is a genuine item of knowledge that we have, the very item to which I just referred: that there must ultimately be something unconditioned corresponding to whatever is conditioned, and in particular corresponding to whatever is conditioned and physical. What we fail to appreciate, however, is that such unconditionedness must reside in things in themselves, which physical things are not. We naturally assume that some physical thing must be unconditioned, the very assumption that we have already seen Kant reject; or, to put it another way, that there must be such a thing as the physical universe as a whole, finite or infinite as the case may be. Once we drop this assumption, we can acquiesce in the conclusion that every physical thing is part of some other physical thing that is older and bigger—as the earth, for instance, is part of the solar system—though there is no one physical thing of which every physical thing is part.30 24 A448–450/B476–478, emphasis added. 25 See Bxx; A498–499/B526–527; and A696–697/B724–725. 26 Not that it is by any means the only evidence: see e.g. Bxxvi and Kant (2000), 5: 196. I should emphasize in this connection that by knowledge of the existence of things in themselves I mean something extremely tenuous: knowledge simply that there is a way things really are. And I should also emphasize that I think Kant thinks we have such knowledge, not because I take him explicitly to say so, but rather because I take him to endorse the idea that there is a way things really are without any of the caveats that would be required if he were endorsing something that he nevertheless took to be unknown, e.g. because he took it to be an article of faith. 27 A422/B450. Cf. also the passage from A420/B447–448, with which I think Bird struggles on p. 676. 28 Here it is important to remember that for Kant ‘experience’ is to be understood very broadly, as equivalent to ‘empirical cognition’: see B147. 29 Kant (2002), §52c. 30 ‘Transcendental Dialectic’, Bk II, esp. §§IV ff. This signals an interesting asymmetry between the thesis and the antithesis, at least in their temporal aspects. Although Kant holds both the thesis and the antithesis to be false, the latter is so to speak closer to the truth than the former. After all, Kant takes himself to have established in the second analogy that every event has a prior cause; and this means that there is a

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Bird on Kant’s Mathematical Antinomies

117

How exactly does this account set me apart from Bird? There is of course the point that I, unlike Bird, allow us, on Kant’s behalf, a minimal item of knowledge concerning things in themselves. But there are differences even apart from that. Bird again and again insists that in the mathematical antinomies we are, in Kant’s view, addressing questions that are undecidable for us. I disagree. Kant stands by the correctness of the arguments in these antinomies, except of course for what he sees as their erroneous shared presuppositions;31 and he takes them to show exactly how to decide the questions addressed. We have first to reject those shared presuppositions,32 then apply the arguments. Thus the correct verdict in the case of the first antinomy is that it is false that the physical universe as a whole is finite, and false that the physical universe as a whole is infinite, because it is false that the physical universe as a whole is.33 In the section beginning at A476/B504 Kant is explicit that we must not treat our failure to reach a satisfactory verdict on such matters as an excuse to say that they are beyond us. Bird refers to this passage. He mentions in particular Kant’s objection to our pleading ‘unavoidable ignorance’.34 ‘What [Kant] objects to,’ Bird comments, ‘is not the fact of our ignorance but the failure to explain it’.35 Yes, in the case of the dynamical antinomies. No, in the case of the mathematical antinomies.36 sense in which history is infinite. The point, however, is that it is a sense having to do with the a priori form of events, rather like the sense in which time itself is infinite. It does not yield a sense in which history— conceived as the contingent course of events hitherto—exists as an infinite unconditioned whole. (Here I think I am in agreement with remarks that Bird makes at pp. 692–693 about the relation between the second analogy and the antithesis of the third antinomy. I likewise think that Bird is right, on p. 675, to reject a suggestion that he finds in Jonathan Bennett about Kant’s own position being indistinguishable from the antithesis. That said, I am not sure that this is Bennett’s suggestion, or not in the form in which Bird finds it (see Bennett (1974), §§46 and 88); and, even if it is, it is not the preposterous suggestion that Bird makes it out to be, precisely because of the asymmetry.) 31 Kant (2002), §52a. Kant stands by these arguments. He does not however offer them in a spirit of persuasion. He offers them in a spirit of descriptive rational psychology. He takes them (rightly or wrongly) to be arguments that ineluctably force themselves upon us as soon as we think about these issues (e.g. A339/B397 and A464/B490). This gives us licence, incidentally, to acknowledge idealist elements in the arguments (e.g. in the argument for the spatial part of the antithesis of the first antinomy, at A427–429/ B455–457) even though Kant uses the arguments in indirect support of his idealism. We need not worry that, in so doing, we are thereby imputing question begging to him. All that we, as exegetes, need to ask is how likely it is that Kant thinks that there are idealist elements in our thinking even before we have recognized them as such. It is extremely likely. 32 A481–482/B509–510. 33 A501–502/B529–530 and A506–507/B534–535. Does this conclusion not hold of what is ahead of us in time, as well as of what is behind us in time?—Certainly it does.—Why then does Kant explicitly decline to apply counterparts of the temporal arguments to the world’s future (A410/B437; cf. A336–337/B393–394)?— Well, he does not deny that such arguments can be constructed. He merely declines to construct them. This relates back to the point that I made in the previous note. Kant is concerned with arguments that he thinks ineluctably force themselves upon us as soon as we think about these issues. The arguments concerning the world’s past do so, because of our urge to locate some unconditioned object of possible experience grounding what is conditioned. A thing’s temporal conditions necessarily precede it; so our thoughts are ineluctably led backwards. But ‘it is a matter of complete indifference’ to us (A410/B437) what would happen if our thoughts took a similar route forwards. For further discussion see Essay 7 in this volume. 34 A477/B505. 35 P. 676. 36 Notice that if Bird were right—if Kant held both that we are unavoidably ignorant in the case of the mathematical antinomies and that the thesis and the antithesis in each of them may both be false—then he (Kant) would have to allow for an incoherence in the concept of the physical universe as a whole which is beyond our ken. Not that that is in itself an objection to what Bird says. Kant does indeed think that we can

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

118

Language, World, and Limits

I want finally to fulfil the promise I made earlier, by relating this discussion to Bird’s fundamental project. I shall say very little, however. The matter is far too complex for me to go into detail here. It would be nice for me if I could claim that only an interpretation of the sort I favour—a sort that Bird would call traditionalist—can accommodate Kant’s insistence that there is no such thing as the physical universe as a whole. I do think that such an interpretation is particularly well-placed to accommodate that insistence, because I think that it is particularly well-placed to explain Kant’s repudiation of experience-transcendent facts about physical reality, of which any fact about the physical universe as a whole would be an instance.37 It is particularly well-placed to explain Kant’s repudiation of such facts because it entails, in a fairly direct way, that there is, for Kant, nothing to physical reality beyond what is capable of being given in experience. But that is as much as I want to claim. I do not want to claim anything stronger because an interpretation of the sort Bird favours can also, I believe, explain Kant’s repudiation of experience-transcendent facts about physical reality, albeit more subtly. Indeed I find the position in question—the combination of the broad view that Bird finds in Kant with a repudiation of experience-transcendent facts about physical reality—extremely attractive. The problem is that, great as my admiration for Kant is, not even a principle of charity allows me to convert my sympathy for the position into sympathy for the corresponding exegesis.38

be unavoidably ignorant even about what is possible (e.g. Bxxvi, note). It does however add to the mystery of what sort of grasp Bird thinks that Kant thinks we have of the concept. 37 Cf. A503–505/B532–533. 38 I am extremely grateful to Graham Bird for his comments on an earlier version of this essay. These comments saved me from a number of errors, though no doubt from fewer than he would have hoped. A slightly different version of this essay appeared in a special issue of Kantian Review devoted to his commentary, where he has a reply: Bird (2011), §4.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

9 Solipsism and Subjectivity Abstract This essay is concerned with solipsism, understood as the extreme sceptical view that I have no knowledge except of my subjective state. A less rough formulation of the view is mooted, inspired by a Quinean combination of naturalism and empiricism. An objection to the resultant position is then considered, based on Putnam’s argument that we are not brains in vats. This objection is first outlined, then pitted against a series of counterobjections. Eventually it is endorsed, but only at the price of exposing the formulation of solipsism in question as not, after all, a satisfactory formulation. This leads to further speculation about the status of solipsism itself. Two of the possibilities that are considered are, first, that it is incoherent and, secondly, that it is inexpressible.

1. Solipsism and its formulation Of the many ways in which the term ‘solipsism’ can be used, one is to stand for a familiar and extreme form of scepticism.1 A standard formulation of this scepticism would be that I have no knowledge except of how things seem to me; another, that my knowledge extends no further than my experiences. Both these formulations are rough. They may not even be equivalent. But a Quinean combination of naturalism and empiricism (however antithetical to the spirit of solipsism it may be) suggests the following recipe for providing an improved formulation. First, define a world as any distribution of microphysical states across the whole of space-time.2 Then single out some part of the actual world that has privileged evidential status for me. Huge empirical and philosophical questions will be involved in determining what might count as privileged evidential status here, and thus what should be singled out. Arbitrary decisions may also have to be taken. Quine, when discussing evidence, typically talks in terms of the stimulation of sensory receptors, or impacts on sensory surfaces.3 Neurophysiology suggests that we can focus on something spatially more confined. The details do not matter. What matters is that what is singled out should include a good deal less than we ordinarily take to be fixed 1 Insofar as this is an unusual way of using the term, it is worth emphasizing that solipsism, on this construal, is a doctrine about knowledge. It is not a doctrine about the constitution of reality. It does not, for instance, conflict in any (obvious) way with a physicalist conception of what the world is like. 2 Cf. Quine (1969c), pp. 147 ff. 3 See e.g. Quine (1969b), p. 75; Quine (1975), p. 68; and Quine (1990a), p. 1.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

120

Language, World, and Limits

by what I know. For the sake of argument let us suppose that what is singled out is the biography of my brain, on some reasonable interpretation of that phrase. Then the formulation in question—the formulation of solipsism that results from following this recipe—is the following: (S)

Any world that includes the biography of my brain is, for all I know, actual.

These worlds vary enormously. At one extreme are all those worlds that are compatible with everything I believe. Provided that I do not have incompatible beliefs, there are very many of these. At another extreme is that world in which the whole of the rest of space-time, by which I mean the complement of the region involved in the biography of my brain, is empty: there is just my brain, popping into existence, ticking over for a few decades, then popping back out of existence. In between are worlds representing possibilities that have become familiar through philosophical discussions of scepticism. For instance, there are those worlds in which I am a brain in a vat. (S) is the claim that I know nothing that distinguishes between the actual world and any of these rivals. (S) is by no means the strongest claim formulable in these terms that might be thought to do justice to the original sceptical position. In particular, (S) makes no provision for variations in the topology of space-time itself. (Not much scepticism is required to think that the topology of space-time is left undetermined by what I know.) Nor does (S) make any provision for structure-preserving variations in the biography of my brain. For instance, it leaves out of account worlds in which my brain suddenly doubles in size half way through my life, or in which my brain suddenly stops working on my eighth birthday and starts up again only after a billion years of total inactivity, or in which, as a result of suitable logarithmic stretching, my brain’s biography is temporally infinite (so that I always have existed and always shall exist). There are natural ways of strengthening (S) to allow for all of these. But for current purposes (S) itself will be a suitable focus for discussion. There are countless familiar ways of trying to refute (S), or, if not to refute it, then to fend it off by some other means, say by showing that it is meaningless or by showing that, even if it is true, it is innocuous. Quine himself is impatient with questions about what individuals do or do not know.4 Scepticism is not an issue for him. He acquiesces in the world-view that we have arrived at by ordinary scientific endeavour, and is not disposed to doubt that world-view except at the dictate of further scientific endeavour. It is only by so acquiescing, he would say, that we can see (S) for the sceptical claim that it is in the first place.5 This illustrates the naturalism that is part of the framework within which (S) was formulated. Others, not I,6 would say that this in turn should make us doubt the suitability of that framework for formulating any such scepticism. True philosophical scepticism, they would say, undermines our very right to work within this kind of framework. However that may be, there are certainly ways of casting doubt on the truth of (S) that are at the same time ways of casting doubt on its entitlement to represent the original sceptical position, as roughly formulated at the beginning of this essay. One of these is to deny that keeping fixed the biography of my brain suffices for keeping fixed 4 E.g. Quine (1987c).

5 E.g. Quine (1974), pp. 1–3.

6 See again n. 1.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Solipsism and Subjectivity

121

either how things seem to me or what my experiences are like. (Henceforth I shall elide any difference between these by talking in general terms about my subjective state.) Among the many philosophers who do deny this, some go further: they deny that keeping fixed the biography of my brain suffices for keeping fixed the fact that I have any subjective state at all. Certainly, the claim that keeping fixed the biography of my brain does suffice for this, and more specifically that it suffices for keeping fixed my actual subjective state, is far from obvious. Let us refer to this claim as the Sufficiency Claim: The Sufficiency Claim: Keeping fixed the biography of my brain suffices for keeping fixed my subjective state. If ‘suffice’ is understood constitutively, then the Sufficiency Claim is an extraordinarily ambitious piece of philosophy. If ‘suffice’ is understood empirically, that is to say with reference to the laws of nature, then the Sufficiency Claim is under threat from the fact that, in most of the worlds in which the biography of my brain remains fixed, the laws of nature do not. (Of course, this last point would be irrelevant if we restricted attention to those worlds in which the laws of nature do remain fixed. The result might still be an interesting form of scepticism. For instance, among the worlds in which the laws of nature do remain fixed might be those, or some of those, in which I am a brain in a vat.) To reject the Sufficiency Claim is, as I said, to cast doubt not only on the truth of (S), but also on (S)’s entitlement to represent the original solipsism. If the Sufficiency Claim is false, then this formulation of solipsism is at best too strong and at worst irrelevant. Indeed, if the Sufficiency Claim is false, then the original rough formulation of the position in question—namely, that I have no knowledge except of my subjective state—may itself be unsatisfactory, though this time by being too weak. For if the Sufficiency Claim is false, then my subjective state might already place significant restrictions on how things are; so much so, that to deny me any further knowledge is to urge a comparatively mild philosophical circumspection with respect to the distinction between appearance and reality. (I shall come back to this point at the end of the essay.) I myself am persuaded that the Sufficiency Claim is false. One of my aims in this essay is in fact to cast doubt on it. But my procedure will be a circuitous one. I shall devote most of the essay to an objection to (S) that does not involve denying the Sufficiency Claim. According to this objection, even if keeping fixed the biography of my brain does suffice for keeping fixed my subjective state, still it does not suffice for keeping fixed what thoughts I have.7 My strategy will be as follows. I shall first outline this objection to (S). I shall then defend the objection against various counter-objections. These counter-objections 7 The idea is familiar. My subjective state is understood in terms of what is phenomenologically indistinguishable for me. What thoughts I have is understood in an externalist way, as depending, in part, on what is going on outside me. For defence of this externalist conception of thought, see e.g. the essays in Pettit and McDowell (1986), whose introduction contains references to a good deal of further material. A particularly interesting and strong form of the externalist conception would have it that I do not think anything at all unless I am in communicative interaction with other people: cf. Davidson (1991) and Davidson (1992).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

122

Language, World, and Limits

will cut successively deeper. To fend them off I shall need to show that the objection itself is able to cut correspondingly deep. Eventually I shall need to show that it can cut deep enough to challenge the Sufficiency Claim, even though it does not trade on this fact. The whole exercise, I hope, will serve to cast some further light, and some further doubt, on solipsism, in the sense in which I am taking it.

2. The Putnam objection to (S) First, then, what is the objection to (S)? I shall refer to it as the Putnam objection. This is because I shall be trading on Putnam’s celebrated proof that we are not brains in vats.8 ‘Trading on’ is the operative phrase though. I shall not be directly rehearsing anything in Putnam. In calling the objection the Putnam objection I only mean to be affiliating it to his work. I do not mean to be making any imputation. The Putnam objection runs as follows. What thoughts I can have is constrained by what I am in causal contact with. For example, if I never come into suitable causal contact with trees, then I cannot have any thoughts about trees. But I do have thoughts about trees, and I know that I do, witness this very argument. So, given what I know, there must be trees in my environment. This is enough to refute (S), since many worlds that include the biography of my brain are tree-free. Not that there is anything very special about trees. These remarks about trees apply to much else besides. The Putnam objection therefore places a significant restriction on which worlds are compatible with what I know. It is not effective, and is not meant to be effective, against milder forms of scepticism. For instance, it is compatible with my not being able to rule out the following possibility: that my brain was surgically removed last night, during my sleep, and then put into a vat that it has been in ever since. But the objection does tell against (S). Why accept the premise that what thoughts I can have is constrained by what I am in causal contact with? Or, in terms more in keeping with Putnam’s own discussion, why accept that reference is causally constrained? (Putnam himself devotes much of his discussion to linguistic reference, which raises further complications. We can, however, ignore these.) Prima facie it is odd that Putnam should accept any such premise in this context. His own argument occurs in the first chapter of a book whose second and third chapters consist largely of a defence of a Quinean indeterminacy thesis, according to which there is no one privileged reference relation that is compatible with all ‘the appropriate operational and theoretical constraints’.9 Furthermore, before outlining his vat argument, Putnam tells us that it ‘first occurred to [him] when [he] was thinking about a theorem in modern logic, the “Skolem-Löwenheim Theorem”’,10 which is the very theorem that underpins the later indeterminacy arguments.11 This is an odd juxtaposition, because precisely what those indeterminacy arguments seem to show is that reference is not causally constrained; that I can have thoughts about anything whatsoever, irrespective of what I am in causal contact with;

8 Putnam (1981), Ch. 1. 9 This phrase occurs in Putnam (1981), p. 44. 11 E.g. ibid. p. 44. Cf. Putnam (1983).

10 Ibid. p. 7.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Solipsism and Subjectivity

123

that there are perfectly good ways of making sense of me whereby all my thoughts are about, say, ordinals between ℵ2 and ℵ3.12 However, the tension is not as great as it appears. What the later indeterminacy arguments show is that reference is not determined by content. In other words, given the content of what I think, or of what I say, there are endless different reference relations in terms of which it can be specified. What the premise of the vat argument is concerned with is what determines content in the first place. The sense in which I cannot have thoughts about trees without coming into suitable causal contact with trees is the sense in which my thoughts cannot have arboreal content unless my having those thoughts is sensitive to what trees are like. This premise—that what thoughts I can have is constrained by what I am in causal contact with—illustrates a general principle to the effect that a subject cannot have any thought of a certain basic empirical kind unless the subject’s having that thought is suitably sensitive to whether things are how they must be for the thought to be true, or, as I shall put it, unless that thought is suitably sensitive to its own content. Both ‘of a certain basic empirical kind’ and ‘suitably sensitive’, of course, have a colossal amount of work to do here, and each needs extensive discussion of its own. But for current purposes this highly schematic formulation of the principle—with the relationship between trees and thoughts about trees serving as a kind of paradigm13— will suffice. Let us refer to this principle as the Sensitivity Constraint: The Sensitivity Constraint: A subject cannot have any thought of a certain basic empirical kind that is not suitably sensitive to its content. One thing that I think follows from the Sensitivity Constraint is that most of a given subject’s beliefs are true. And I think an argument to show that it follows is implicit in things that Putnam himself says when he talks about the content of the beliefs that a brain in a vat would have, content, roughly, concerning what was going on in and around the vat.14 Nevertheless, I also think it would be misleading to give such an argument. A full account of these matters, I believe, would indicate that the order of derivation is rather the reverse: that the Sensitivity Constraint is required to ensure that most of a subject’s beliefs are true, which must in turn be assumed if the subject is to be credited with any beliefs at all. That some such exercise of charity is integral to making sense of a subject is a celebrated thesis of Davidson, and one for which he has provided a number of arguments.15 It is given further support, of a rather different kind, elsewhere in Putnam’s own book.16 There is obviously much more to be said about it. But let us at any rate grant the premise of the Putnam objection, namely that what thoughts I can have is constrained by what I am in causal contact with.17 12 This apparent tension is noted by Brueckner in Brueckner (1986), n. 2. 13 Cf. Nagel (1986), p. 71. 14 Putnam (1981), p. 14. 15 See e.g. Davidson (1984e), p. 168; Davidson (1984h), pp. 199–201; and Davidson (1986), p. 317. 16 Putnam (1981), pp. 39–40. 17 For reservations about how much this premise yields, see Sacks (1989), esp. §3.2. Note: some of what I say in this paragraph may engender puzzlement as follows. Viewed in the abstract, a Davidsonian exercise of charity looks as if it could pull in the opposite direction to the Sensitivity Constraint. Reconsider the brain in the vat. And suppose that it has, or seems to have, all sorts of thoughts about the solar system (say). Does not a Davidsonian exercise of charity—in forcing us to interpret these thoughts as mostly true—

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

124

Language, World, and Limits

3. The natural counter-objection The Putnam objection is met by a natural counter-objection, which, for convenience, I shall refer to simply as the natural counter-objection. (The literature is replete with instances of people responding to Putnam in this kind of way.18) The natural counter-objection does not aim to defend (S) against the Putnam objection. It aims rather to lessen the force of the Putnam objection by showing that, if the Putnam objection holds, then all that ultimately follows is that (S) does not after all do justice to the original solipsism. The argument is as follows. Even if the Putnam objection enables me to conclude that there are worlds including the biography of my brain that are not compatible with all I know, still, in those worlds, I may be able to undergo seemingly identical thought processes to reach a seemingly identical assurance. For example, in a treeless world, I may be able to reach a conclusion that carries precisely the same significance for me as is carried, in this world, by the conclusion that there must be trees in my environment. Moreover, if the Putnam objection holds, I am right to reach that conclusion in that world. For what I am then concluding is, so to speak, that there must be ‘pseudo-trees’ in my ‘pseudoenvironment’—which indeed there must. Granted this, it follows that the Putnam objection gives me an assurance of the falsity of (S) only at the price of showing that I do not fully understand it. In particular, I cannot rule out that the worlds with which it deals are, from some external standpoint, mere constructs of my own phenomenal bubble and not real variations on how things are. In sum, I do not know what (S) comes to at the level of things in themselves. Furthermore, this leaves room for, and indeed encourages, the following genuinely solipsistic thought: (S*) For all I know, any ‘noumenal’ possibility that includes whatever sustains my subjective state is, for all I know, actual. (The reason for the ugly repetition of ‘for all I know’ in (S*) is that these ‘noumenal’ possibilities may in fact be worlds as construed in this essay, and what sustains my subjective state may in fact be the biography of my brain. So without the first occurrence of ‘for all I know’, there is no guarantee that (S*) is not simply a re-formulation of (S), which the Putnam objection shows to be false.) I cannot be any more specific about these ‘noumenal’ possibilities. If I try, say by talking about brains in vats, then I am dealing once again with worlds, as construed in this essay—‘phenomenal’ possibilities—and my scepticism is vulnerable to the Putnam objection. But I do not need to be more specific. All I need to do, in order to sustain this genuinely solipsistic involve denying that they are limited in the way that the Sensitivity Constraint demands? (The brain’s thinking cannot be suitably sensitive to what the solar system is like.) Well, no. What the Davidsonian exercise of charity involves denying is that these thoughts are really thoughts about the solar system—just as it involves denying that any of the brain’s thoughts are really thoughts about its stomach. The brain has mostly true thoughts about goings-on in and around the vat, which it conceptualizes in its own highly idiosyncratic way. To interpret a subject as having mostly true thoughts is not to ignore what the subject’s thinking can be suitably sensitive to. On the contrary, it is to use what the subject’s thinking can be suitably sensitive to—under a suitable construal of suitable sensitivity—as a basis for interpretation. (I am grateful to David Bell and Mark Sacks for forcing me to clarify my thinking on this point.) 18 See e.g. Brueckner (1986); Sacks (1989), pp. 76–79; Brueckner (1990); Wright (1994); Falvey and Owens (1994); and Forbes (1995). Cf. also Craig (1990).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Solipsism and Subjectivity

125

thought, is to transcend the naturalistic framework within which (S) was formulated and entertain these ‘noumenal’ possibilities, in the abstract.19

4. The Putnam reply Here my essay begins to resemble a tennis rally: the natural counter-objection is met by what I shall call the Putnam reply.20 The Putnam reply begins with the following strengthened version of the Sensitivity Constraint: a subject cannot have any thought that is not suitably sensitive to its content. Let us refer to this as the Generalized Sensitivity Constraint: The Generalized Sensitivity Constraint: A subject cannot have any thought that is not suitably sensitive to its content. The Generalized Sensitivity Constraint applies, in particular, to thoughts about what is possible, about what is real, and the like. Hence no matter what nuance I may introduce into my thinking, no matter how carefully I may qualify what I say with references to the noumenal or with cautionary scare-quotes, no matter how hard I may thump the table, I cannot have these thoughts (that is to say, thoughts about what is possible, what is real, and the like) unless my thinking is suitably sensitive to what is possible, to what is real, and the like. If the best conception that I can form of what the possibilities are is a conception in terms of worlds, then I have no alternative but to assume that worlds do after all constitute real, ultimate possibilities. What I cannot help but conceive of as the realm of the possible I cannot help but conceive of as the realm of the possible and is indeed the realm of the possible. Scepticism couched in terms of unspecifiable ‘noumenal’ possibilities falls prey to the original objection in precisely the same way as scepticism couched in more specific terms (in terms of brains in vats, say).21

5. The reinforced counter-objection The Putnam reply is met by a reinforced counter-objection. The reinforcement is simply that some thoughts are available to a subject irrespective of what the subject’s thinking can be relevantly sensitive to (in other words, the Generalized Sensitivity Constraint is false). 19 For the distinction between ‘noumenal’ possibilities and ‘phenomenal’ possibilities, cf. Kant (1933), Bxxvi, note. 20 Cf. Putnam’s own reply to Wright (1994) in Putnam (1994), §8; and Genova’s reply to Craig (1990) in Genova (1991). 21 Prime targets for the Putnam reply are the four italicized expressions in the following quotation from Wright (1994), §VI (emphasis altered by me): ‘[T]he real spectre to be exorcised concerns the idea of a thought standing behind our thought that we are not brains-in-a-vat, in just the way that our thought that they are mere brains-in-a-vat would stand behind the thought . . . of actual brains-in-a-vat that “We are not brains-in-a-vat”. The spectre is that of a thought whose truth would make a mockery of mankind and its place in nature, just as our true thought that they are merely brains-in-a-vat makes a mockery of the “cognitive” activity of the envatted brains. What we would really like would be an assurance . . . not just that most of what we think is actually true . . . but that we are on to the right categories in terms of which to depict the most general features of the world and our place in it.’

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

126

Language, World, and Limits

There is more than one model that can be used to develop this reinforcement. One is a Kantian model. On this model, my concepts of what is possible, of what is real, and so forth are a priori. I have them irrespective of how and where I am situated: I do not glean them from anything with which I am in empirical contact. And although I cannot use them, or use anything else for that matter, to gain substantial knowledge of that which is transcendent, I can use them to think about that which is transcendent.22 A brain in a vat could share these concepts with me. It could think about things in themselves and entertain thoughts about ‘noumenal’ possibilities. More specifically, it could entertain the thought that, for all it knew, any ‘noumenal’ possibility that included whatever sustained its subjective state was, for all it knew, actual—even though it could not be more specific about what these ‘noumenal’ possibilities might be like. Similarly, I can entertain my own counterpart of this solipsistic thought: namely, (S*). A different model that can be used to develop the reinforcement is a Nagelian one. At times Nagel seems to embrace a realism robust enough to entail that no thought need be relevantly sensitive to its content.23 At any rate Nagel argues that we can use extrapolation from actual cases of limited cognitive ability—he mentions the congenitally blind or deaf—to entertain thoughts about what might be beyond our own cognitive abilities, or our own powers of specification. Similarly, he imagines a species whose intellectual limit is that of a normal nine-year old human. Just as there are features of reality that they would never be able to grasp, so too, Nagel says, we can entertain the thought that there are features of reality that we shall never be able to grasp.24 It seems clear that Nagel might have used the fantasy of the brain in the vat to make the same point. And as with the Kantian model, this in turn seems enough to show that I can entertain the solipsistic thought (S*). I said these models were different. On the Kantian model, I can have thoughts that are not relevantly sensitive to their content by exercising a priori concepts. On the Nagelian model, I can have thoughts that are not relevantly sensitive to their content by extrapolating from what I am in empirical contact with. None the less, the models are very closely related—much more closely related than these remarks suggest. The latter is indeed a variant of the former. For on the Nagelian model, I need to be able to exercise concepts that reach beyond what I am in empirical contact with in just the way in which my a priori concepts of possibility, reality, and so forth do on the Kantian model—a prime example being the concept of analogy, as expressed by the tetradic predicate ‘x1 is to x2 as x3 is to x4’.25 Whatever form the reinforcement to the counter-objection takes, sheer modesty seems to call for it. To see why, imagine a brain in a vat being manipulated in such a way as to simulate precisely the activity of your own brain. And imagine yourself observing it. As you are doing so, it must be thinking that it too is observing a brain in a vat, thinking that it is observing a brain in a vat, thinking that it is observing a brain in a vat, and so on, in ‘Russian doll’ fashion, ad infinitum. Now, if the Generalized 22 23 24 25

Cf. Kant (1933), A64/B89 ff., B146, B306 ff., and A320/B377. See e.g. the comments on trees in Nagel (1986), p. 72. Ibid. pp. 95 ff. (The whole of Ch. VI is relevant here.) Cf. Wright’s use of the concept of analogy in the quotation in n. 21.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Solipsism and Subjectivity

127

Sensitivity Constraint holds, then each thinker in this nested structure, with the exception of you, must have a limited conception of reality, or a limited conception of what the possibilities are, as revealed from the vantage point of its predecessor. Not you, however! You have a kind of a priori guarantee that your conceptions of these things fit them perfectly. For you, there is not so much as the possibility of mismatch. Surely modesty indicates that there is something wrong with this picture; that there is something wrong with the idea that an infinite series of phenomenologically indistinguishable thoughts should suddenly culminate in something with an a priori assurance that the series does not, and cannot, extend any further.

6. Whither solipsism? Well, yes; there is something wrong with this picture. But the problem lies not with the Generalized Sensitivity Constraint. It lies rather with how the story has been told. There is not an infinite hierarchy of thinkers in this scenario, any more than there is an infinite hierarchy of brains. At most, there are two: you and the brain. (I say ‘at most’, because the brain will soon be giving pause.) Apart from you two, and again at most, there is a ‘pseudo-thinker’ having ‘pseudo-thoughts’, a ‘pseudo-pseudo-thinker’ having ‘pseudo-pseudo-thoughts’, and so on. Once the story is told in these terms, modesty scarcely seems to come into it. For you to say, of a brain bobbing about in a vat, that it has at best a limited conception of reality, while denying that the same is true of yourself (where this of course does not mean denying that you are both mistaken and ignorant about all sorts of things), is not an arrogation, whatever else might be said against it. For my own part, I think the Generalized Sensitivity Constraint must hold. Admittedly I have stated it in such an anodyne way as to leave little room for dispute about it. But I think I have said enough to indicate how it rules out the ascription of certain thoughts to certain subjects: how, more particularly, it does this in the case of thoughts about what is possible, about what is real, and the like; and how this in turn prevents the reinforced counter-objection from sustaining (S*) as a deeper expression of solipsism than (S). I shall not now try to argue for the Generalized Sensitivity Constraint however. My interest, for now, is to see how things stand in this debate if it holds. Prima facie they stand much as they stood when the Putnam objection was first formulated. That is, (S) seems both to have been refuted and to do justice, of sorts, to the original solipsism. But we can perhaps say more now than we said before about worlds that are ruled out by what I know. Not only must there be trees in my environment—or at least, not only must there be as much in my environment as there must be to ensure that I know as much as I do know about my own thinking on such matters (there is still room for debate about how much this is)—I must also have whatever access to my environment I need to be able to conceive of the full range of worlds. The world in which there is nothing but the biography of my brain, for instance, is ruled out on the grounds that, in it, I would have no conception of real space. In that world, nothing in my thinking could target the centre of gravity of my left hemisphere, for example, or not under that description. Nor could I conceive of the full

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

128

Language, World, and Limits

range of worlds, construed as distributions of microphysical states across the whole of space-time. But I can conceive of that range, and I know that I can. So my knowledge rules out that world’s being actual.26 But now consider: I have been making claims about what ‘could’ be the case in one very austere world. Are these to be understood as claims about what is the case in some other world? They had better not be. For some of what I deny ‘could’ be the case in that world I am prepared to say is the case in the actual world. (For instance, I am prepared to say that I have thoughts about the centre of gravity of the left hemisphere of my brain.) No, they are claims about capacities and dispositions, the sorts of capacities and dispositions that are at work in the ‘suitable sensitivity’ of the Generalized Sensitivity Constraint. The ‘could’ here is more restrictive than a mere existential quantifier ranging over worlds. Something which, in this sense, ‘could’ not be the case in one world, but is the case in another, betokens capacities and dispositions that are absent from the former world but present in the latter. Compare: I cannot lift a ton weight, though there are worlds in which I do. I have belaboured this relatively simple point because I think it is important to be clear about what the implications of the Generalized Sensitivity Constraint are. That I can conceive the full range of worlds, for example, and that I can recognize the actual world as one among them, is not just due to the fact that the biography of my brain is part of the actual world, nor yet to the fact that it is part of that world and of others. It is due to the fact that I have thoughts that are suitably linked, through the range of capacities and dispositions that I have in virtue of possessing a normal body, to different parts of space-time. This cuts deeper than the fact that I have thoughts that are suitably sensitive to what trees are like. If my thinking lacked this connection with space-time, then it would be out of touch with its own most fundamental nature; I would lack a proper self-conscious awareness of what my thinking came to, not because I could not see what the content of my thoughts was, but because I could not see what it was for them to have that content. What I actually have, and what I would thereby lack, is the capacity to think ‘objectively’ about my own thoughts, and thus to arrive at a self-conscious awareness of their standing in the world, including their standing with respect to other potential thoughts about that same world. I have, and would lack, self-conscious awareness of my point of view on the world. But what is that, if not an aspect of my subjectivity? At long last the Sufficiency Claim—the claim that keeping fixed the biography of my brain suffices for keeping fixed my subjective state—seems to be coming under threat. To make this point more graphic, imagine a trans-temporal counterpart of some of the trans-world variations that we have been considering. Imagine a world that includes the biography of my brain and in which the following is true: up until my thirtieth birthday, say, my brain is in a vat being manipulated by scientists; it is then taken out of the vat and implanted in a normal body; and things so develop that, from my thirty-second birthday onwards, they are just the same as in the actual world. On the current reckoning, I can, in that world, begin to conceive of the full range of worlds and come to a proper self-conscious awareness of my own thinking some time after my thirty-second birthday, not before. But if the Sufficiency Claim is true, then 26 Cf. Evans (1982), §7.6.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Solipsism and Subjectivity

129

this world is subjectively indistinguishable to me from the actual world. And this makes a complete mystery of my actual subjectivity. It means that my subjective state is a kind of epiphenomenon, accidentally relevant, at best, to whatever awareness I can achieve of my own point of view—which is absurd. What my subjective state is, at least in part, is a representation of things from my own point of view of such a kind that, granted self-consciousness, I can see it as a representation of things from my own point of view. Quite how much this requires, and quite what room it leaves for error on my part (such as the massive error involved in my failing to realize that my brain has been in a vat since last night) are huge philosophical questions that I cannot address now. But certainly it requires more than that my brain be ticking over. It requires, among other things, that I have a proper conception of space-time. If, in worlds where I lack this conception, I have any subjective state at all (something that there is increasing reason to doubt), then the most that can be said of that state is that I think I can see it as a representation of things from my own point of view. (Perhaps I can see it as a ‘pseudo-representation’ of things from my ‘pseudo-point of view’.) At any rate the Sufficiency Claim has at last been discredited—on the assumption that the Generalized Sensitivity Constraint holds.27 There may have been reason to doubt the Sufficiency Claim anyway, reason which these reflections have served only to highlight. For granted that my subjective state controls my thinking in the way in which it does, the Sufficiency Claim renders obscure how my thinking can ever be about anything beyond the biography of my brain—whatever correlations there may be between activity within my brain and activity without. Given the Sufficiency Claim, the only sense in which I can ever experience that things are thus and so, other than within the biography of my brain, is by being in a certain subjective state while things independently are thus and so, which in turn means that my grasp of the world beyond my brain is radically indirect, indirect enough to render deeply problematical how I can so much as think about that world. But if the Sufficiency Claim threatens my capacity even to think about what is beyond my brain, then we are within sight of a transcendental argument against it. (Here I am trading on ideas that have become familiar through the work of McDowell.28) However that may be, my (actual) subjective state is grounded in a point of view that I have, not just in the world, but on the world. It affords me a kind of ‘possession’ of the world. This relates back to something that I remarked upon near the beginning of this essay, namely that, if the Sufficiency Claim is false, then even the rough formulation of solipsism with which we began—that I have no knowledge except of my subjective state—may be too weak to do it justice.29 For, as it now appears, my subjective state already determines a good deal about the world beyond me. Or at any rate, it does unless we are forced by these considerations into a radical re-thinking of what counts as ‘me’, in which case solipsism, as I have been construing it, may cease even to be an issue. 27 I am indebted in this paragraph to Eilan (1994–95) esp. §IV, and to Brewer (1995), esp. p. 250, though I doubt that either author would want to endorse everything I say in the paragraph. 28 See e.g. McDowell (1986a), esp. §8; and McDowell (1994), p. 17, n. 14, Lecture VI, §3, and Appendix I, §5. 29 Cf. McDowell (1986)a, pp. 147 ff. See also Burnyeat (1982).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

130

Language, World, and Limits

What then is the upshot of the Putnam objection for solipsism? Well, pressed hard enough, the Putnam objection challenges the Sufficiency Claim, as we have seen. This means that it is not just an argument against (S). It is an argument against (S)’s entitlement to represent the original solipsistic position. However, granted the point just made, it is also an argument against the entitlement of the original rough formulation to do that. This cannot but leave us wondering whether there is any such thing as ‘the original solipsistic position’. Perhaps what we are witnessing here are the exposure and the disintegration of a merely apparent position. That, certainly, is my own view. I think that the Putnam objection, ultimately, feeds into a critique of the very coherence of solipsism. But there is nothing yet to force this conclusion. Two other possibilities—at least—remain to be considered. One of these is that solipsism is a perfectly genuine position for which we simply have not (yet) found a suitable formulation. The other, altogether headier, possibility is that solipsism is a perfectly genuine position that has no suitable formulation. (This of course puts us in mind of Wittgenstein’s famous remark, in the Tractatus, that ‘what the solipsist means is quite correct; only it cannot be said, but makes itself manifest.’30) The first of these is much likelier to recommend itself to initial sober reflection. And yet what could the formulation in question be? Not (S*). Given the Generalized Sensitivity Constraint, (S*), if it means anything at all, means much the same as (S). The Generalized Sensitivity Constraint nullifies any attempt on our part to reckon with ‘deeper’ possibilities, which is what (S*) involves. On the other hand, if we resist the temptation to try to delve deeper, then it is not clear how we can avoid producing a formulation that is hopelessly weak, something that in effect just says that I know nothing beyond what I know; or that the actual world is, for all I know, the actual world. The fact of the matter is that, once we have got this far, unless we are prepared to deny that there is any genuinely solipsistic position, then it is very tempting to embrace the headier alternative of thinking that it is a position that is immune to proper formulation. It is tempting to think that brains in vats, for instance, can share something with us, a kind of ‘ultra-subjectivity’, which enables them to entertain phenomenologically identical thoughts to those which we are now entertaining; then to think that the only proper way to formulate solipsism is in terms of how this ‘ultrasubjectivity’ stands with respect to ‘deeper’ possibilities; and then, finally, having seen how this must fail, to conclude that solipsism is inexpressible. I myself am convinced that this temptation must be resisted, and that, if inexpressibility does come into the reckoning, then it comes in differently. That is something I have tried to argue elsewhere.31 For the time being, I am content to have given some indication of the challenge that the Putnam objection poses for solipsism. One final and very brief observation. The Putnam objection, or at least its reinforced version, depends on various assumptions that I have made no attempt to defend. The most blatant of these is the Generalized Sensitivity Constraint itself. But another is the assumption that I know my own thoughts (for example, I know that I have thoughts about trees). A large and fascinating literature has arisen concerning 30 Wittgenstein (1961), 5.62, emphasis removed.

31 Moore (1997), esp. Chs. 7 and 8.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Solipsism and Subjectivity

131

whether I am entitled to this assumption in view of the underlying externalism.32 It is worth noting, however, that there is a much more humdrum reason why I may need to be cautious about this assumption. If knowledge requires confidence, and if one effect of indulging in reflection of this kind is that I lose my confidence, however irrationally, then it may be that, while I am reflecting in this way, I do not know my own thoughts. This would be an interesting variant on Williams’ contention that reflection can destroy knowledge.33 It is a further question, of course, whether this in turn would be enough to salvage the counterpart of (S); whether, in other words, it would be enough to ensure that any world that included the biography of my brain was, for all I knew while I was reflecting in this way, actual.34

32 For a small sample of this literature, see e.g. Brueckner (1986); Davidson (1987); Heil (1988); Brueckner (1990); Falvey and Owens (1994); Burge (1995); and Peacocke (1995). 33 Williams (1985), pp. 148 and 167–169. 34 I should like to thank Mark Sacks for various suggestions that helped me to improve this paper.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

10 One or Two Dogmas of Objectivism Abstract This essay first appeared as a critical notice of Thomas Nagel’s The Last Word. Though the essay evidences broad sympathy with the spirit of Nagel’s book, its main burden is to query the letter of the book. Nagel’s defence of the view that there are certain beliefs and ways of thinking that are not from any point of view, or that are ‘objective’ in his own terms, is criticized on the grounds that it is too facile. It is also criticized for not being pitted against a critique of beliefs and ways of thinking that are from some point of view, or that are ‘subjective’ in his terms.

1. Introduction Thomas Nagel needs no plaudits from me. His crisp, forceful, rigorous treatment of matters that are both deep and important has long been a familiar feature of analytic philosophy. His latest book, The Last Word,1 is sure to add to his well-deserved reputation. It is an engaging onslaught on a certain highly pernicious abnegation of critical thinking which, under the cover of various unlovely ‘isms’, Nagel takes to be ‘epidemic in the weaker regions of our culture’.2 Nagel himself tends to designate his target by some of its less barbarous titles, if not perhaps the ones by which it is most likely to style itself: ‘relativism’, ‘scepticism’, ‘subjectivism’, ‘irrationalism’. We know the kind of thing he has in mind. Perhaps we must share Nagel’s pessimism about whether such a book will make what he is opposing any the less fashionable. But, like work directed at other forms of philosophical scepticism, it can at least add to a selfconscious grasp of various basic principles on which we rely in our thinking about the world. Nagel’s fundamental idea is that it is impossible to question the objectivity of certain ways of thinking without, sooner or later, betraying a commitment to that very objectivity. Thus his opponents, who claim that these ways of thinking are not objective, because of anthropocentrism or because of some hidden cultural bias, say, find that they have no suitable vantage point from which to press their claim. 1 Nagel (1997)—of which this essay first appeared as a critical notice. All unaccompanied references will be to this book. 2 P. 4.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

One or Two Dogmas of Objectivism

133

I find myself broadly in sympathy with the spirit of Nagel’s book, much less so with the letter of it. Most of this review will take the form of criticism. But I trust that I have already said enough to forestall any misunderstanding. I see this book as a welcome recoil from an all too seductive and all too pervasive misology.

2. Six things that may happen when we reflect critically on our beliefs Suppose we reflect critically on some of our beliefs. Then all sorts of things may happen. We may modify the beliefs. We may come to accept some of their consequences. We may achieve insights into our own psychology. It will depend on the nature of the beliefs, the nature of the reflection, the nature of the enquiry, and much else besides. Even if the reflection is of a distinctively philosophical character, there are many things that may happen. Here is a sample. First, we may find that we surrender the beliefs, or at least that we cast doubt on them. The methods by which we have arrived at the beliefs may come to seem unreliable; or we may notice a conflict between these beliefs and others we have. Secondly, and conversely, we may find that the beliefs are reinforced, say because the best explanation of why we have them is that they are true, or because there is reason to think that, in having them, we make them true. Thirdly, we may find that we stop even thinking in those terms. This is importantly different from surrendering the beliefs. Theists, reflecting on their beliefs about God, and in particular on their belief that God exists, may end up deciding that after all God does not exist. If they do, then this will mean their surrendering one of their beliefs—their core belief—about God. But the kind of thing that I am envisaging now would be exemplified by their surrendering something else: their very concept of God. Arguably, this is the kind of thing that is going on in certain anti-realist retreats from classical logic. To be sure, there are anti-realists who think that classical logic is just wrong. But there are also anti-realists who regard classical logic as a perfectly serviceable piece of machinery which, however, for various reasons, we do well to put aside in favour of a no less serviceable alternative.3 The three things that I have mentioned so far have a bearing on whether or not we retain our beliefs. But critical reflection may also lead us to draw conclusions about our beliefs that have no such bearing. Thus, for example—and this is the fourth thing that may happen—we may draw conclusions about whether or not certain of our beliefs are contingent. In particular, we may change our minds about this. Philosophers have a special knack for highlighting contingency in what seems necessary and necessity in what seems contingent. There is, however, something else that they sometimes try to do, something more subtle and more radical. They try to highlight contingency in what seems necessary whilst continuing to acknowledge the necessity. Typically they are motivated by a desire to alleviate the apparent mystery of our having epistemic access to all possibilities. They hope to find some unproblematically accessible feature of how things actually are that serves to explain why they must be the way they must 3 Cf. George (1993), pp. 71–72, and Williams (1985), p. 167.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

134

Language, World, and Limits

be. Patently, if this can be done, then it can be done only with great care. A blatant example of how not to do it is provided by a popular reading of Descartes, whereby he attributes the necessity of twice four’s being eight to our (humans’) being unable to grasp any of the other possibilities. This is clearly self-stultifying. What other possibilities?4 The fifth thing that may happen—the one that is Nagel’s chief concern—is that we draw conclusions about whether certain of our beliefs are objective or subjective. By an ‘objective’ belief, Nagel means a belief that is universal and detached, one that is not from any point of view. A ‘subjective’ belief is the opposite. Alternatively, a ‘subjective’ belief is one that contains, however deep down and however well concealed, an ‘I’ or a ‘we’. It ought to be almost as obvious that recognizing a belief as subjective, in this sense, is no immediate impediment to retaining it as it is that recognizing a belief as contingent is no immediate impediment to retaining it. The subjectivity of a belief does not, in itself, impugn its truth. (There are familiar arguments to the effect that even our belief that grass is green is subjective.) Nevertheless, it is a bold view and, in its own way, a sceptical view that all our beliefs are subjective. And Nagel’s target obviously includes anyone who thinks that. At one point he makes the standard but to my mind facile objection to this view: either the claim that all our beliefs are subjective is objective, in which case it is self-refuting, or it is subjective, in which case we have no reason to accept it because it does not rule out any objective claim, including the claim that it is objectively false.5 But the second limb of this dilemma contains a confusion. If the claim is subjective, there can still be reason to accept it, if only subjective reason. It does not need to rule out any objective claim. At least, it does not need to rule out any objective claim if there are no such things. It need only rule out other subjective claims, which it certainly does: it rules out the claim, from the same point of view, that some of our beliefs are objective. As regards what reason there is to accept it, that may be a matter, in part, of our coming to recognize the point of view that it is from as ‘ours’, something over which we may have no more control, at least while we are thinking about these issues, than we have over our position in time—our temporal point of view. But still, Nagel might say, what about alternative points of view from which some of our beliefs are objective? What can advocates of this view say about these? There are many things they can say. They might insist that there are no such points of view, since what makes their own claim subjective is the fact that the very concepts of objectivity and subjectivity are only available to be exercised from that particular point of view. The sixth thing that may happen is that we turn our attention to the genesis of our beliefs and to the conditions which, either as a matter of necessity or as a matter of contingency, make or have made it possible for us to form and sustain them. At this point philosophy, though it still has its own distinctive contribution to make, is liable to shade off into anthropology, sociology, history, or psychology. Clearly these six things, which by no means constitute an exhaustive list, have all sorts of bearing on one another. Nevertheless, we must be careful not to exaggerate 4 I am not, however, persuaded that this reading of Descartes is correct. See the excellent Bennett (1994). For Nagel’s discussion of Descartes on modality (itself, I think, based on a suspect reading), see Ch. 4, §II. 5 Pp. 14–15.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

One or Two Dogmas of Objectivism

135

the bearing. There are very few direct routes from any one of them to any other. They raise quite different issues, and the issues must not be conflated.

3. Nagel’s fundamental idea and two difficulties that he faces Is my charge that Nagel is guilty of conflating them? Not exactly, though we shall see later that there is cause for concern on this score. I have listed these six things in order to highlight some problematic features of Nagel’s fundamental idea. That idea, to repeat, is that it is impossible to question the objectivity of certain ways of thinking without eventually betraying a commitment to that very objectivity. But what form is this betrayal supposed to take? Is the idea simply that, in addressing the question whether certain ways of thinking are objective, we cannot help assuming that they are? No, not directly. The idea is rather that, in addressing the question whether certain ways of thinking are objective, we cannot help adopting those very ways of thinking. Thus if what is at issue is the objectivity of our belief that p, the proposition that Nagel says we cannot help assuming is not that our belief that p is objective, but simply that p. This idea can also be put in terms of the eponymous metaphor of the last word. Nagel nowhere explicitly says what he intends by this metaphor, but it is clear enough. In particular, we can take for granted that when the last word on an issue is that p, then two things at least must be true: first, we cannot help but think that p if the question arises; and second, the question does arise. (These two things are independent. The second indicates why the last word is always the last word on an issue. Something that is the last word on one issue might be totally irrelevant to another.) Nagel’s idea is that the last word on whether certain beliefs are objective is given over to those very beliefs. The fact that Nagel takes this somewhat oblique approach means that he faces two special difficulties. The first difficulty is simply that of showing that any given belief is the last word on the issue of its own objectivity. This is not just a matter of how irresistible the belief is. It is also a matter, as I put it just now, of whether ‘the question arises’. Given that the belief and the issue are not aligned, it may be non-trivial to show that the question does arise. The second difficulty is that of showing why, in exercising the belief, we are committed to its objectivity. By drawing attention to some of the things that may happen when we reflect on our beliefs, I hope to have made it easier to assess these two difficulties. It is very important, in evaluating Nagel’s idea, to retain a suitably variegated conception of what reflection can do and of how its effects can depend on what questions, and what range of subsidiary questions, are being addressed.

4. One way in which Nagel might try to face the first difficulty To begin with the first difficulty, then: how is Nagel able to show that any given belief is the last word on the issue of its own objectivity? In particular, how is he able to show that ‘the question arises’? Are we not always at liberty, when considering whether or

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

136

Language, World, and Limits

not a given belief of ours is objective, to ‘bracket’ the first-order issue with which the belief itself is concerned? True, if the belief is irresistible, then we cannot go any further than that. We cannot accede, even provisionally, to its negation, nor to any proposition that is manifestly incompatible with it. But we do not need to. We just need to be circumspect about how we proceed. To take a trivial example: if the belief in question is our belief that modus tollens is valid, what is to stop us from carefully ensuring that we come to a view on the matter without exercising any conditional thoughts? There are many things that Nagel is liable to say in response to this challenge, but one of the first is that a surreptitious and damaging slide has taken place, in the previous section and the present one, from talk about ways of thinking to talk about individual beliefs. He does think that there are individual beliefs that are immune to the challenge, including, as it happens, our belief that modus tollens is valid.6 But he also takes this to be of secondary importance. His fundamental idea has rather to do with whole frameworks of belief: methods of reasoning, modes of argument, and the like. Thus even if it were possible to consider the objectivity of our belief that modus tollens is valid without actually implementing modus tollens, it would not be possible to consider this issue, or indeed any other issue, without using logic.7 But there is a serious problem with this response, concerning the breadth of the notion of a ‘way of thinking’. This notion has to be (a) broad enough to meet the challenge, but (b) not so broad as to exacerbate the second difficulty that Nagel faces (the difficulty of showing why, in exercising a given way of thinking, we are committed to its objectivity). Let us consider (b) first. If a ‘way of thinking’ is guaranteed to include whatever we use in addressing a given issue, then it may be utterly vacuous to say that that way of thinking has the last word on that issue. Certainly it may be too vacuous to guarantee the objectivity of anything. For on such a broad conception, two instances of the same way of thinking need not even involve exercise of the same concepts—in which case a way of thinking is not even a candidate for objectivity. The candidates are rather the instances.8 The mere fact that we have to operate with some instance in thinking about an issue does nothing to foreclose the prospect that all the instances are subjective. This is a prospect that is serious enough in the case of logic. Its seriousness is greatly magnified in the case of ethics. Maybe, in Nagel’s words, ‘moral reasoning is . . . fundamental and inescapable’.9 But this manifestly allows for the possibility that what count as objective or as subjective are the particular forms that moral reasoning takes, and that these are, without exception, subjective. Turning next to (a): the problem this time is to make the notion of a ‘way of thinking’ broad enough to meet the original challenge. The danger is that, even given some relatively large battery of conceptual apparatus, we might, with sufficient ingenuity, 6 P. 56. 7 P. 69. 8 Nagel seems to miss this distinction in his discussion of tea-leaf reading on p. 24, where there is more than a hint of the danger that I am talking about. He contrasts tea-leaf reading with reasoning, on the grounds that, whereas a challenge to tea-leaf reading does not itself imply the authority of tea-leaf reading, a challenge to reasoning does imply the authority of reasoning. But this is an unfair comparison. Tea-leaf reading and reasoning belong to different categories. Tea-leaf reading is an instance of reasoning, albeit a poor one. Challenges to other (better) instances need not imply their authority either. 9 P. 101.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

One or Two Dogmas of Objectivism

137

find ways of addressing an issue without using that apparatus. It was with this in mind that I drew attention to the third thing that may happen when we reflect critically on our beliefs, or on a particular range of our beliefs: we may stop thinking in those terms. Thus suppose that what I said in section 2 about anti-realism is correct. Then even unregenerate classical logicians could, in principle, use non-classical resources to discuss the objectivity of their favoured logic. And to echo a point that I made in the previous paragraph, if something like this could happen in the case of logical reasoning, then it could certainly happen in the case of moral reasoning.

5. A second way in which Nagel might try to face the first difficulty There is another obvious approach for Nagel to take to the first difficulty—the difficulty of showing that a given belief is the last word on the issue of its own objectivity—which is this. In raising the question whether a given belief of ours is objective, we are raising a question about what would make the belief true. In particular, we are asking what our own involvement in this would be. In the case of an objective belief, we would not be involved at all, except insofar as the belief overtly concerns us. But in the case of a subjective belief, whose truth would depend on how things stood at the point of view that the belief was from, we would be involved. For instance, consider our belief that a certain gesture is insulting. The truth of this would depend on a complex web of social practices that we ourselves had spun, a web that served to define the point of view that our belief was from. There might even be a case for saying that having beliefs of that kind was part of spinning the web, so that what we believed was true, in part, because we believed it. But there could never be a case for saying anything like this with respect to an objective belief. Indeed, in the case of an irresistible belief of the sort that Nagel is concerned with, the order of explanation must rather be the reverse. In other words, when a belief is both irresistible and objective, the best explanation of our having it must conform, in outline, to the schema: ‘We believe that p because p’. That is why the question whether or not p arises.10 There are several hints of this line of thought in Nagel.11 There are also several problems with it. Among the objective beliefs that Nagel is concerned with are certain necessary beliefs, such as our most rudimentary beliefs in logic and arithmetic.12 But there would be formidable and familiar obstacles to extending this line of thought to them. Take our belief that twice four is eight. It is not immediately clear what might be meant by ‘the best explanation’ of our having this belief. (It is something of a philosophical artefact to talk about ‘our having this belief ’ at all.) But whatever the explanatory project might be, only confusion would accrue from thinking that there was any serious contribution to be made by the fact that twice four is eight. As soon as any appeal to this fact was understood as any more than an endorsement of something already implicit in the explicandum, we would be involved in all the 10 Cf. the superb discussion of these issues in Wright (1992), Ch. 5. 12 See e.g. Ch. 4.

11 E.g. pp. 56–57.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

138

Language, World, and Limits

epistemological and metaphysical problems of Platonism. Nagel, to be sure, thinks that we are involved in these problems. In a striking final chapter he argues that only a naturalistic fear of religion, to which he confesses his own strong susceptibility, keeps us from acknowledging that such problems arise. But we do not need to follow Nagel in this. Again, whatever it is that we are trying to explain, there is no need for us to appeal to anything other than ordinary secular facts about ourselves: our physiology, our techniques of teaching and inculcation, the ways in which we exploit our methods of calculation when describing our environment, and so forth. It was with this in mind that I drew attention to the sixth thing that may happen when we reflect critically on our beliefs: we may turn our attention to their genesis and to the conditions of our being able to form and sustain them. Where an elementary arithmetical belief such as our belief that twice four is eight is concerned, the demand for an explanation for our having the belief cannot be heard as anything other than a demand to indulge in just such reflection and to locate the contingencies that have made it possible for us to arrive at this stage in our intellectual development. But this, despite worries that Nagel voices,13 poses absolutely no threat to the necessity of the belief. We must be careful, as I warned in section 2, not to presuppose any simple direct links between the different things that may happen when we reflect on our beliefs—the fourth of these being precisely that we may draw conclusions about whether our beliefs are necessary or contingent. Our belief that twice four is eight is indeed necessary. It is necessary because, in ‘having this belief ’, we are effectively laying down a rule of representation: nothing is to count as a disjoint pair of quartets unless it collectively counts as an octet. But this is not in any conflict with the observation that there are all sorts of identifiable contingencies that make it possible for us to have the rule, nor therefore with the observation that we might not have had the rule. If we had not, twice four would not have failed to be eight. Rather, the question of what twice four is would not so much as have arisen for us. We would not have thought in those terms. Twice four would not have failed to be eight, because it must be eight. This ‘must’ is as hard as it either can or need be.14 We have a way of highlighting contingency in necessity, then, that helps us to achieve a naturalistic understanding of our knowledge of the latter. (This is the project I mentioned in connection with the fourth thing that may happen when we reflect on our beliefs.) But Nagel is uncomfortable with this idea, at least where certain fundamental concepts are concerned. He thinks, where these are concerned, that the position I have been defending does flout the necessity I claim to be respecting. At one point he writes, ‘No “language” in which modus ponens was not a valid inference . . . could be used to express thoughts at all’.15 As it stands, this slurs the distinction between a language in which there is no such inference as modus ponens, because the relevant concepts do not occur, and a language in which there is such an inference but it is not valid. On my view, the second of these is indeed impossible. It is part of our understanding of modus ponens that nothing is to count as such a language. But Nagel thinks the first is impossible too, at least if the language is to be capable of

13 E.g. pp. 55 ff. 15 P. 39.

14 Cf. Wittgenstein (1978), Pt. VI, §49 and McDowell (1993), pp. 282 ff.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

One or Two Dogmas of Objectivism

139

expressing rational thought.16 And with this I disagree. In fact, on the standard way of construing modus ponens, as an inference concerning material implication, I am inclined to think that English is a counter-example. At any rate, it seems to me that we have, in the necessity of the validity of the inference from a conditional plus its antecedent to its consequent, all the necessity we need. And I do not think that I have said anything that stands in violation of it. Of course, necessity can be divorced from objectivity. It is open to someone broadly sympathetic to Nagel to concede what I have been arguing in the case of necessity but to try to resist it in the case of objectivity. But I think a good deal of what I have been arguing still applies. Whatever features of our point of view may have been involved in the formation of one of our beliefs and may still be involved in sustaining it, whatever concerns, interests, sensibilities, or social practices may be needed for anyone to have the concepts necessary to share this belief, the belief itself may still be objective. We may be able to say as much as we need to say, in explaining why we have the belief, without putting any explanatory weight on what makes the belief true, if it is. Admittedly, as Nagel insists, time and again, we shall not thereby have given the whole story concerning our belief. In particular, we may not have done enough to indicate whether it is true. But that is precisely my point.17 To be sure, the best explanation of why we have the belief may have to indicate that it is true, on some reasonable interpretation of that still undefined phrase ‘best explanation’. It may have to conform, in outline, to the schema ‘We believe that p because p’. If so, this helps Nagel to address the first difficulty—in particular, it helps him to show that the question whether p arises—which is all that was strictly being maintained in the line of thought above. However, unless it marks a peculiarity of objective beliefs, it simply gives further weight to the second difficulty, the difficulty of showing that, in exercising the belief, we are committed to its objectivity. For maybe the best explanation of why we have some of our subjective beliefs must also conform, in outline, to the schema ‘We believe that p because p’. The subjectivity of the beliefs does not preclude this, since the explanation, in any given case, can be from the same point of view as the belief itself. Or at least, it can unless the notion of a best explanation itself precludes this, say by requiring maximum possible objectivity. This presages issues that we shall be addressing in section 6. Note, however, that if ‘maximum possible objectivity’ just means ‘maximum possible objectivity attainable by us in this context’, then even this does not prevent subjective beliefs from satisfying the condition. It allows for the possibility that there is a point of view that we are forced to adopt simply by engaging in this kind of explanatory project. Imagine, for instance, that the very concept of a belief can only be exercised from a certain point of view, say from a certain interpretative point of view. It is then once again open to us to envisage a subjective belief of ours, such that the best explanation of our having it, understood now as requiring maximum possible objectivity, conforms, in outline, to the schema: ‘We believe that p because p’. 16 Cf. p. 38. 17 I am not denying—indeed I think it is of the first importance—that we are under a persistent temptation, when discussing either the necessity or the objectivity of our beliefs, to go further than we are licensed to go and to say things that not only fail to entail the beliefs but are in tension with them. This requires diagnosis. It should not hinder us from saying what we are entitled to say. See Moore (1997), esp. Chs. 6–9.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

140

Language, World, and Limits

6. A third way in which Nagel might try to face the first difficulty, and his way of facing the second There seems to be a third (and obvious) way for Nagel to address the first difficulty, namely in tandem with the second. For if he can show that simply exercising a given belief commits us to its objectivity, then will he not have done enough to show that the belief is relevant to the issue of its own objectivity; or more precisely, that, by denying the objectivity of the belief, we are committed to surrendering the belief itself? Strictly speaking, no: he will have shown only that, by denying the objectivity of the belief, we prevent ourselves from exercising it. It does, however, follow that the belief cannot be both subjective and true unless there is a kind of blindspot associated with it, a blindspot that makes it impossible for us to acknowledge the belief ’s subjectivity without somehow detaching ourselves from it. This possibility may seem remote enough for solving the second difficulty directly to be tantamount to solving the first.18 Certainly Nagel gives the impression, on numerous occasions throughout the book, that he recognizes only a single problematic here. Again and again he claims that there are beliefs which are irresistible, which we cannot exercise without at the same time regarding them as objective, and whose objectivity we therefore cannot deny.19 But I confess that Nagel’s repeated direct assaults on the second difficulty leave me unmoved. No matter how often he insists that certain beliefs are a presupposition of coherent thought, or that they dominate anything we might want to say about them, or that we have no choice but to think them ‘straight’,20 or any of the other countless variations on this theme, I cannot hear his arguments as anything other than arguments for the unassailable truth of these beliefs: the question of their objectivity seems to me to remain completely untouched. Admittedly, if the arguments are successful, then they answer not only sceptics who doubt whether these beliefs really are true, but also sceptics who doubt whether these beliefs really are beliefs ‘about how things are’. But again, that has nothing to do with the objectivity of the beliefs. It has to do with whether they are beliefs at all. For any belief is a belief ‘about how things are’. When Nagel says, as he frequently does, that we cannot think of the beliefs as mere dispositions of ours, that too is a variation on the same theme. We cannot think of any belief as a mere disposition of ours, if this is meant to foreclose the question of whether the belief is true. But this does not prevent us from thinking that the belief is a disposition of ours, nor that its truth, if it is true, has something specially to do with the ground of the disposition. (My belief that today is Friday is a disposition of mine: its truth, or its falsity as the case may be, has to do with the day on which I have it.) There is still scope for regarding the belief as subjective. A last resort, to which Nagel occasionally seems to be drawn, is to say that certain beliefs are self-evidently objective. But really that is hopeless. The objectivity or subjectivity of a belief is a recondite matter that can be settled only after hard work. Anyone claiming that a belief is 18 Bernard Williams does however suggest that something like this possibility holds in the case of various ethical beliefs: see Williams (1985), pp. 199–200. 19 E.g. pp. 64–66 and 125. 20 P. 19.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

One or Two Dogmas of Objectivism

141

self-evidently objective is straightway vulnerable to the standard objections—of parochialism, lack of imagination, historical insensitivity, and the like—beloved of Nagel’s opponents. These considerations combine to show, I suggest, that Nagel’s best strategy, and the one that he should on the whole be interpreted as adopting, is not to address the second difficulty directly (thereby trying to address the first at the same time), but to address the second difficulty in the light of an antecedent solution to the first. That is, he must begin by showing that the beliefs he is interested in are relevant to the question of their own objectivity, then show why exercising the beliefs, in the course of addressing that question, commits us to their objectivity. This takes us back to ideas adumbrated in the previous section. Nagel will have done this if, in solving the first difficulty, he manages to establish that the question whether or not any given belief is objective can only be properly addressed in objective terms. Solving the second difficulty will then be a simple corollary. Insofar as this is what Nagel is about,21 then, patently, he is on to something of considerable philosophical importance. There is, however, an obvious problem of circularity. How do we tell that we are properly addressing one of these questions—that we are thinking about the issue in objective terms? In fact the circularity is even more alarming than this suggests. For if we began by asking whether a particular belief was objective, then what we shall have been driven to ask is whether a particular account of it that includes that very same belief is objective. At first blush, this circularity seems to constitute a fundamental theoretical objection to the strategy. But I do not think it does. The second question arises in a context of enquiry that did not exist at the outset. This leaves open the possibility that there are, in the new context, resources to assess the objectivity of our belief that were not previously available. On the other hand, neither is the circularity just the circularity that is destined to afflict any anti-sceptical argument, the circularity that means that the most hardened sceptic will always remain unsatisfied. Rather it signals a genuine practical problem: we need to be acutely sensitive, and we need to know that we are being acutely sensitive, both to what exactly is at issue, and to what exactly is demanded of us, at any given stage in our enquiry. Quite simply, to determine that a given belief, or a given account of something, or a given way of thinking, does not contain some tacit reference to ‘me’ or to ‘us’ is never easy. Even if Nagel is right in most of the conclusions he draws, he too often draws them with a facility that leaves the reader feeling a need for greater reassurance. That brings me to what is in many ways my most serious complaint about the book. The obvious way to have supplied such reassurance would have been by pitting the strategy against beliefs and ways of thinking to which it ultimately could not be applied, thereby removing any sense that Nagel was indulging in a simple mechanical exercise. What the book cries out for is an additional detailed critique of beliefs and ways of thinking that are not objective, for instance in the area of humour or chromatics, and that cannot be seen as providing the last word on the question of their own objectivity. Nagel might protest that he has already done enough to remove any sense that he is indulging in a simple mechanical exercise, by emphasizing that there 21 See e.g. p. 16.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

142

Language, World, and Limits

is no single way of applying his strategy even in cases where it can be applied.22 But without the contrast that would be created by failed attempts to apply the strategy, we have no real safeguard against a threat that is liable to worry many of Nagel’s opponents: namely, that his arguments can too readily be adapted and marshalled in support of an unacceptable bigotry, defending the objectivity of beliefs that are in fact subjective and whose false dignification as the last word may even sometimes be a matter, quite literally, of life and death. Nagel’s own quasi-religious defence of his position, in the last chapter, adds obvious piquancy to this threat. The fact is, his position is a kind of dogmatism. Like other kinds of dogmatism, and in full spite of the book’s title, it leaves us anxious to hear more.23

22 See pp. 26–27. 23 I discuss these issues in greater depth in Moore (1997), the first four chapters of which are devoted to arguing that objective beliefs, or more generally what I call ‘absolute representations’, are possible. But ‘possible’ is the operative word. Unlike Nagel, I am not willing to commit myself to their being actual, still less to their being recognizable.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

11 Apperception and the Unreality of Tense Abstract The aim of this essay is to characterize the issue whether tense is real. Roughly, this is the issue whether, given any tensed representation, its tense corresponds in some suitably direct way to some feature of reality. The task is to make this less rough. Eight characterizations of the issue are considered and rejected, before one is endorsed. On this characterization, the unreality of tense is equivalent to the unity of temporal reality. The issue whether tense is real, so characterized, is then related to Kant’s deduction of the categories in his Critique of Pure Reason. It is argued that Kant’s deduction does not provide the argument for the unreality of tense that it may appear to. The conclusion drawn at the end of the essay is that the unreality of tense cannot be argued for—not because tense is real, but because, even if it is unreal, its unreality is basic.

My concern is with the issue whether tense is real. Broadly speaking, what this means is that I shall be investigating the metaphysics of tensed thought and talk. The sort of question on which I hope to cast light is this: can one and the same fact account both for the truth of a memory I have today, about how things were yesterday, and for the truth of a thought that I had yesterday, about how things were then? Eventually I shall be concerned with how such questions relate to various aspects of what Kant calls ‘the deduction of the categories’ in his Critique of Pure Reason.1 More specifically, I hope to see how the answers to such questions might depend on the answers to questions about (to use Kant’s phrase) the unity of apperception.2 I begin with three disclaimers. First, although I talk about ‘the’ issue whether tense is real, and shall continue to do so throughout the essay, this is in full acknowledgement of the fact that I am discussing just one of many issues that could reasonably be referred to in this way. All sorts of debates have been conducted under this title. I make no pretence to be focusing on any one thing that has been in dispute in every case.3 Secondly, I am not so much interested in settling this issue as in characterizing it. This means that, granted the first disclaimer, I must beware that what I characterize is worthy of attention. In particular, I must try to respect what I shall call the Interest 1 Kant (1933), A84–92/B116–124. The deduction is given in A95–130 and, differently, in B129–169. 2 This phrase occurs e.g. in ibid. B135. 3 For a very helpful survey see Le Poidevin (1998). One of the issues that I shall not be discussing is whether a being that was completely outside time could have any conception of what the past, the present, or the future were: cf. Dummett (1978c).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

144

Language, World, and Limits

Constraint: to characterize the issue in such a way that neither the view that tense is real nor the view that tense is unreal is a straw man, in the sense of being easily refuted or completely without appeal.4 The third disclaimer is that I shall not be doing any serious Kantian exegesis. Most of the claims that I shall make about Kant’s deduction will, I regret, be unsubstantiated. But those who dispute my attributions might still find something of interest in the relationship between the issue on which I shall eventually fasten and the ideas that I shall claim to find in Kant. Now I have said that there is more than one issue that could reasonably be called ‘the’ issue whether tense is real. Various points of controversy cluster together here, and they need to be distinguished. But they are not always distinguished. As a result, many of the debates that have been called debates about whether tense is real are little more than a tissue of confusion and failure of communication, so that some philosophers have reacted with scepticism—perfectly justified scepticism—about whether anything at all merits the label. To show that I take such scepticism seriously, I shall devote the first section of this essay to highlighting some of the issues that could not reasonably be what is meant by the issue whether tense is real, at least not granted the Interest Constraint, though they look at first blush as though they could.5

1. Eight issues that could not reasonably be what is meant by the issue whether tense is real Tense is a feature of certain representations. That is, loosely, it is a feature of certain beliefs, judgements, thoughts, claims, and assertions.6 It is that feature which indicates 4 There is another constraint that I might be expected to try to respect in this connection: namely, to characterize the issue in such a way that it is not terminological, i.e. in such a way that those who profess a belief in the reality of tense and those who profess a belief in the unreality of tense do not have at their disposal some (mutually acceptable) scheme of translation whereby they can recognize that they are not in dispute after all. But I am much less confident that the characterization I shall provide conforms with this second constraint than I am that it conforms with the Interest Constraint. For one thing, the two constraints are in some tension with each other: the clearer it is that neither of the opposing views in any given philosophical debate is a straw man, the less clear it is that what divides them is more than terminological. But I am also much less certain that respecting this second constraint is necessary for ensuring that the issue characterized is worthy of attention. The expression ‘terminological’ has all sorts of pejorative overtones. But I see no reason why an issue cannot be terminological, in the sense defined, and also of considerable philosophical moment—if, for instance, it is an issue about which of two conceptual schemes is better equipped to meet certain theoretical and/or practical needs. (This is related to David Cockburn’s project in his extremely interesting (1997).) Whether the issue on which I shall be focusing is an issue of this kind is a question that I find enormously difficult. Nothing I say in this essay will help to settle the question. But neither will anything I say depend on how it ought to be settled. 5 I shall be relying heavily on Cockburn (1997), which, in effect, gives voice to such scepticism. At the end of one particularly pertinent chapter (Ch. 6), Cockburn says that he has been involved in ‘an attempt to dissolve the dispute between tensed and tenseless theories [i.e. theories that do and theories that do not accept the reality of tense] as this is widely understood’ (p. 127). ‘As this is widely understood’ indicates that he has in mind something metaphysical. (He allows room for a related ethical dispute.) Granted that my own concerns are broadly metaphysical, Cockburn’s chapter challenges my aspiration to identify even one dispute in this territory while respecting the Interest Constraint. 6 This is as much as I am going to say about how I shall be using the term ‘representation’, other than to add here that, on my conception, a representation is something that is true or false, without relativization. There is much more that could be said, obviously. In particular, there is much that could be said in response to the worry that, by assuming that there are things that are true or false without relativization, I may, in

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

145

that they are from a temporal point of view. Thus if I say, ‘It was humid yesterday’, both the grammatical form of the verb I use and my use of the word ‘yesterday’ are part of the tense of the representation I produce. Now very roughly, the issue whether tense is real—any issue that could reasonably be referred to in that way—turns on whether this feature of representations corresponds in some suitably direct way to a feature of what is represented. The task at hand is to make this less rough. First suggestion: The issue whether tense is real is the issue whether the truth conditions of a tensed representation (a representation from a temporal point of view) can be given by means of a representation that is itself tensed.7 In due course I shall argue that the issue is a variation on this theme. But as it stands this suggestion is clearly wrong. It makes the view that tense is unreal a straw man, in violation of the Interest Constraint. For patently the truth conditions of a tensed representation can be given by means of a representation that is itself tensed. If I say, today, ‘It was humid yesterday’, then what I say is true if and only if it was humid yesterday. Second suggestion: The issue whether tense is real is the issue whether it can be of material significance whether an event has occurred, is occurring, or is yet to occur. That will not do either. That, too, makes the view that tense is unreal a straw man. Of course it can be of material significance whether an event has occurred, is occurring, or is yet to occur. My attitude to an excruciating pain that I recall feeling yesterday is very different from my attitude to that same pain two days ago, while I was still anticipating it. But it remains to be shown that what is of significance here is something that directly answers to the tense of my various representations of the pain, and not just the tense of the representations themselves.8 the context of this essay, be begging some crucial questions. E.g. Richard Swinburne says that the unreality of tense follows from the assumption that truth is timeless (Swinburne (1990), p. 120). However, I think I can allay this worry. For whatever relativization may be involved in ascriptions of truth and falsity—such as the obvious relativization of a sentence-type to a time of utterance, and perhaps also, the relativization of an utterance to a time of evaluation (cf. Wright (1987), §7 and Campbell (1994), pp. 228–230)—I hereby simply stipulate that all the relevant relata are to be thought of as already bound up in the notion of a representation, thus guaranteeing that any given representation can be regarded as true or false without further relativization. Cf. Percival (1994), p. 205. 7 By the truth conditions of a representation I mean the conditions that must be satisfied—the way things must be—in order for the representation to be true. (This notion, it must be said straight away, needs all sorts of clarification. In fact, as we shall see, it needs clarification that is highly pertinent to the issue on which I shall eventually fasten. But because there is more than one acceptable way to clarify it, and because I think that the choice between these may itself be part of settling the issue—this relates to what I said in n. 4 about the possibility of the issue’s being terminological—I am deliberately going to leave the notion in this unclarified form.) Note: the canonical way of giving the truth conditions of a representation is to say that it is true if and only if things are a certain way. And sometimes it is possible to do this by simply reusing the sentence that was used in producing the representation in the first place. Thus I can give the truth conditions of an utterance of the sentence ‘Water contains oxygen’, by saying, ‘This utterance is true if and only if water contains oxygen.’ But sometimes greater subtlety is called for. I cannot give the truth conditions of an utterance by you of the sentence ‘I am thirsty’ by saying, ‘This utterance is true if and only if I am thirsty.’ 8 The issues that arise here were classically brought into focus in Prior (1959). They are pursued in MacBeath (1983) and Mellor (1983).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

146

Language, World, and Limits

Third suggestion: The issue whether tense is real is the issue whether properties such as being in the past and being in the future are properties that events have ‘intrinsically’ (that is, not from any temporal point of view). The problem with this suggestion is that it is likely to make the opposite view, the view that tense is real, a straw man, again in violation of the Interest Constraint. I say ‘it is likely to’, rather than ‘it will’, because there is a way of construing the claim that events have these properties ‘intrinsically’ whereby (pending various metaphysical concerns about whether an event can still have a property even when evidence for its having that property is lost, or concerns about whether the future is open9) it is fair to say that they do.10 For instance, I take it to be a simple fact that the first moon landing has already taken place. I do not say that the first moon landing has already taken place from this temporal point of view, as though it might not have already taken place from some other temporal point of view, any more than I say that I am English from my point of view, as though I might not be English from yours.11 But still, when I say, ‘The first moon landing has already taken place’, I am producing a representation from a particular temporal point of view. This is illustrated by the fact that somebody could once have used the sentence ‘The first moon landing has already taken place’ to say something false (just as a German could use the sentence ‘I am English’ to say something false). The straw man is the view that the first moon landing has the property of having already taken place, construed in such a way that any utterance of the sentence ‘The first moon landing has already taken place’ would have succeeded, or would still succeed, in ascribing that property to it. For insofar as the first moon landing has any such property, then it equally has the property (similarly construed) of being yet to take place; and insofar as there is any making sense of either of these properties, then no event can have them both.12 A defender of this view might try to escape this contradiction by saying that the properties in question have been underspecified; that what are really at stake are such properties as having already taken place now and (what is quite compatible with that) being yet to take place in the past. But the response to this is familiar. If this is meant as a defence of the view on anything like the lines originally envisaged, then the original objections can themselves be reformulated: insofar as the first moon landing has the property of having already taken place now, then it equally has the property of being yet to take place now.13 Fourth suggestion: The issue whether tense is real is the issue whether two tensed representations of the same type but from different temporal points of view (for instance, two utterances on successive days of the sentence ‘It was humid yesterday’14) have, modulo any other indexical features they share,15 the same content. 9 See Wright (1987) and Tooley (1997) respectively. See further section 3. 10 On this way of construing the claim, I think there is an issue here that could reasonably be what is meant by the issue whether tense is real—precisely because of the metaphysical concerns mentioned in parenthesis. However, it is not the issue on which I wish to focus. 11 Cf. Lowe (1987) and Lowe (1992). 12 This is reminiscent of one of the key steps in J.M.E. McTaggart’s celebrated argument for the unreality of time: see McTaggart (1993), pp. 31–32. 13 Cf. ibid. pp. 32–33. 14 For more on this notion of a representation’s type, see Moore (1997), pp. 9–11. 15 The indexical features of a representation are those features of it that indicate that it is from a certain point of view, in the way that tense does. Thus, if I say, ‘It is icy over there’, my use of the phrase ‘over there’

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

147

The problem with this is that there is not really an issue here. In one sense the two representations have the same content. In another sense they do not. Nothing of substance hangs on which sense we choose to adopt.16 Fifth suggestion: The issue whether tense is real is the issue whether physics needs to involve tensed representations. I am probably more sympathetic than most to the idea that physics plays a unique and privileged rôle in informing our conception of reality. Indeed I am prepared to say that the only facts are physical facts, facts about how things are physically.17 But even if this much is granted, this fifth suggestion fails. For consider: physics does not need to involve representations that are explicitly about dogs. Yet not even the most strident physicalism can combine with that to yield the conclusion that dogs are unreal. Sixth suggestion: The issue whether tense is real is the issue whether, if there had never been any tensed representations (say, because there had never been any creatures capable of producing them), events would still have had such properties as being in the past and being in the future. There is an issue here, certainly. But it does not seem to me to be a good candidate for the issue whether tense is real. Someone who denies the reality of tense can quite reasonably accept that, had things turned out in such a way that no creatures capable of producing tensed representations had ever existed, the Big Bang, say, would still have occurred some fifteen thousand million years ago. There is no reason why we should not use tensed representations to describe a world without tensed representations, even while denying that anything in that world corresponds in any suitably direct way to the tense of the representations we thereby use. It is rather as if someone who denied the reality of colour, because of its secondary-quality status, nevertheless insisted that, even if there had never been any creatures with visual apparatus, grass, say, would still have been green.18 Seventh suggestion: The issue whether tense is real is the issue whether a description of reality consisting only of tenseless representations (representations from no temporal point of view) would, no matter how exhaustive it was, be incomplete.19 is an indexical feature of the representation I produce: it indicates that the representation is from a certain spatial point of view. 16 There are similar choices to be made with respect to ‘proposition’, ‘belief ’, and other related terms; there is e.g. the choice about whether to say that the two representations express the same proposition. See Swinburne (1990), §3 and Mellor (1998), §2 for more on these choices. (I do not think that any of the choices generates a real issue. But it is a serious question why not—granted what I said above in n. 4 about the possibility of an issue’s being both terminological, in the sense defined there, and ‘of considerable philosophical moment’. I have nothing to offer in response to this question but an appeal to judgement. All the crucial distinctions that need to be drawn here, such as the distinction between a choice on which something of substance hangs and a choice on which nothing of substance hangs, are distinctions of degree, not of kind. My own judgement is that not enough hangs on these choices for there to be any serious philosophical point in debating them. This is what I mean by saying that none of the choices generates a real issue. I shall leave it to others to decide how far this distances me from either Swinburne or Mellor.) 17 See further Moore (1997), pp. 75–76. 18 Cf. the discussion of ‘rigidification’ in Wright (1992), pp. 113–114. 19 Cf. Dummett (1978c), pp. 356–357.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

148

Language, World, and Limits

I take it that for a description of reality to be complete is for it to be such that, given any fact, it (the description of reality) entails that that fact obtains.20 But then the very idea of a complete description of reality might be incoherent, for reasons having nothing to do with the reality of tense. (Perhaps some variation on Heisenberg’s uncertainty principle, or on Gödel’s theorem, might scupper the idea.21) If so, then the necessary incompleteness of a description of reality consisting only of tenseless representations would have no implications for the reality of tense. This suggestion also fails.22 Eighth suggestion: The issue whether tense is real is the issue whether assimilation of a description of reality consisting only of tenseless representations would, no matter how exhaustive the description was, leave room for ignorance. This suggestion is like the seventh, only worse. It is worse because, even if a complete description of reality were possible, and even if some being had assimilated such a description, there might still be things of which the being was ignorant. This is admittedly paradoxical; and the paradox deserves a fuller discussion than I can offer here. But the point is relatively simple. Granted that a complete description of reality is a description of reality such that, given any fact, it entails that that fact obtains, then there are all sorts of ways in which a being that had assimilated such a description might nevertheless lack knowledge. For one thing, the being’s knowledge might not be closed under entailment (that is to say, the being’s knowledge might not include everything it entails). But even if we put that consideration to one side, the being might lack non-propositional knowledge such as knowledge of how to write beautiful music. More pertinently, the being might lack tensed knowledge such as knowledge of what the date is. Or at least, some argument is required to show that this is not the case. Pending such an argument, the possibility remains open that the being’s lacking this knowledge would be nothing but a failure on its part to know some of the facts from a temporal point of view—not a failure on its part to know these facts at all. And if that were so, then nothing would follow about the reality of tense. Thus those who insist that tense is real because tenseless knowledge, no matter how exhaustive it was, would leave room for ignorance are guilty, it seems to me, of a straightforward non sequitur.23

2. One issue that could reasonably be what is meant by the issue whether tense is real I said in response to the first suggestion that it was a variation on the suggestion that I would eventually endorse. Indeed I think that if, in place of the word ‘can’, it had 20 Michael Dummett, in his (1978c), p. 356, glosses ‘complete’ as ‘observer-independent’. But he is presumably stating an entailment, not offering a definition. 21 For discussion of each of these, see Penrose (1989), pp. 248–250 and 105–108 respectively. 22 Can it not be turned into a good suggestion by a suitable insertion of the word ‘thereby’? Perhaps. But something needs to be said about what work is being done by the word, and I suspect that, if the amended suggestion really is a good one, then it will turn out to be the same as the suggestion that I shall eventually endorse. 23 Examples are Swinburne (1990), §§5 and 6, and Lucas (1998). For a fuller discussion of these issues see Moore (1997), pp. 53–58 and 171–172. Cf. Lewis (1983) and Perry (1993). Cf. also Butterfield (1985), which connects with my essay at several points.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

149

contained the word ‘must’, the suggestion would have been acceptable. I therefore propose the following: The issue whether tense is real is the issue whether the truth conditions of a tensed representation must be given by means of a representation that is itself tensed.24 The point is this. Tense is first and foremost a feature of certain representations. The fact that something in reality admits of being represented in this way is not enough to license the claim that the feature is ‘real’; that there is something in reality corresponding in some suitably direct way to the feature. But the fact that something in reality demands to be represented in this way, provided it were a fact, would be enough. Objection: Would it not be more than enough? Suppose that the truth conditions of a tensed representation do not have to be given by means of a representation that is tensed. Suppose, in other words, that they can be given by means of a representation that is tenseless. Does it follow that tense is unreal? After all, to echo a point that arose in connection with the fifth suggestion, it may be possible to give the truth conditions of a representation that is explicitly about a dog without oneself making any reference to the dog but rather by referring to its several parts and saying how they must stand in relation to one another in order for the representation to be true. If so, it certainly does not follow that nothing in reality directly corresponds to that feature of the representation which indicates that it is about a dog, nor that the dog is in any sense unreal. Reply: There is a crucial difference between the reality of tense and the reality of a dog. The latter, unlike the former, does not have to be understood in terms of some antecedent grasp of a certain feature of representations. Or so I claim. If I am wrong about that—if there are decisive arguments for some sort of idealism whereby the reality of the dog does have to be understood in this way—then so be it: the dog is unreal. Second objection: According to the proposed characterization, if the truth conditions of a tensed representation can be given by means of a representation that is tenseless, then tense is unreal. But might it not also be the case that the truth conditions of a tenseless representation can be given by means of a representation that is tensed? And if it is the case, should not parity of reasoning force us to conclude that tenselessness is unreal too?25 Reply: Yes, but there is no contradiction in this. Both tense and tenselessness are features of representations. It may be that neither corresponds in any suitably direct way to any feature of what is represented. Third objection: Suppose that tenseless representations are not possible at all. Or, less dramatically, suppose that it is not possible to identify a temporal point of view tenselessly. Then must not the truth conditions of a tensed representation be given by means of a representation that is itself tensed, but not for reasons that have anything to do with the reality of tense? Reply: It is true that I am taking for granted the possibility both of tenseless representations and of tenseless identifications of temporal points of view.26 The characterization 24 Cf. Priest (1986), pp. 162–163. Cf. also Mellor (1993), p. 59, with a slightly different but closely related gloss on the issue. 25 Cf. (Priest) 1986. 26 Arguments for and against the possibility of tenseless representations can be found in Le Poidevin (1998–99) and Teichmann (1998) respectively.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

150

Language, World, and Limits

above fails if either of these things is not possible. However, I have an alternative characterization of the issue for which I claim the following: if both of these things are possible, it is equivalent to the characterization above, and, if they are not, it succeeds anyway. According to the characterization above, if tense is real, then there is no giving the truth conditions of a representation from a given temporal point of view except by producing a representation that is itself from some temporal point of view. But this can only be because there is no giving the truth conditions of a representation from a given temporal point of view except by producing a representation that is from that very same temporal point of view. For suppose the latter is not the case. Suppose, for example, that I can give the truth conditions of an utterance I made yesterday of the sentence ‘It is humid today’ by saying, ‘My utterance yesterday was true if and only if it was humid yesterday.’ Then there is nothing to prevent me from giving the truth conditions of my utterance tenselessly, by tenselessly identifying the temporal point of view I had yesterday—say, by specifying yesterday’s date. Or, at least, there is nothing to prevent me from doing this unless tenseless representations are not possible at all, or unless tenseless identifications of temporal points of view are not possible—say, because there is no tenselessly saying what is meant by ‘anno domini’. What this shows is that the heart of the view that tense is real is that something in reality corresponds in a specially direct way to the actual tense of any given tensed representation, not just to its being tensed. Thus, whether or not tenseless representations are possible, and whether or not tenseless identifications of temporal points of view are possible, the following will do as an alternative characterization of the issue: The issue whether tense is real is the issue whether the truth conditions of a representation from a temporal point of view must be given by means of a representation that is from that very same temporal point of view.27

3. Why it is an open question whether tense is real on this construal The question I must now address is whether this characterization of the issue violates the Interest Constraint. More specifically, does it make the view that tense is real a straw man?

27 Counter-objection: Is there not an intermediate position, based on a belief in the openness of the future, which undermines what I say in this paragraph: namely, the position whereby there are some tensed representations whose truth conditions can be given by means of representations that are from the same temporal points of view or later ones, though they cannot be given by means of representations that are from earlier temporal points of view because the truth conditions come into existence at a certain time and are simply not available to be given before then?—Reply: This position certainly muddies the waters. Unless it can be ruled out (which, ultimately, I think it can) then what I say in the main text needs modification— but not, I think, in a way that affects the main point that I am making. For even if the position is correct, there remains a sense in which it is not possible to give the truth conditions of a representation from a temporal point of view except by producing a representation that is from that very same temporal point of view: the sense, namely, in which a temporal point of view is a (possibly infinite) period of time, not a point in time. For a very thorough discussion of the issues and complications that arise here, with its own distinctive angle on them, see Tooley (1997).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

151

Note first that the view does not impugn my right to say of yesterday’s utterance of ‘It is humid today’ that it was true if and only if it was humid yesterday. For even if I am right to say this, it remains an open question whether I thereby give the truth conditions of yesterday’s utterance. On the view that tense is real, so characterized, I do not.28 On that view, the truth conditions of yesterday’s utterance are not that it was humid yesterday. At most they correspond in some specially intimate way to the conditions that it was humid yesterday: the obtaining of either at most affords some sort of guarantee of the obtaining of the other. But they are not the same, because the truth conditions of yesterday’s utterance contain something that directly corresponds to the tense of the utterance, something that precludes my giving those truth conditions except by reproducing that tense, which I can no longer do. This in turn means that, if yesterday’s utterance was true, then the fact that made it true can no longer be so much as expressed.29 Whatever there is to be said against this view, I do not believe that it is a straw man. For instance, on this view, representations from temporal points of view other than the  present can still be acknowledged, now, as representations. They can still be acknowledged as either true or false. More particularly, they can still be acknowledged as having truth conditions. And we can still talk about these truth conditions, just as we can talk about the facts that make the representations true or false. In what terms? In precisely those terms—referring to them as we would to things of any other kind. (I can talk about ‘the fact’ that made something I said yesterday true, just as I can talk about ‘the tree’ that I was sitting under when I said it.) This much is necessary if the view is to have any intelligibility at all. What we cannot do, if the view is correct, is to express the fact that makes any one of these representations true or false. We cannot produce a representation, either tensed or tenseless, that is itself made true by the same fact. The picture that supports this view—the only picture, so far as I can see, that can support it—is that reality fractures into different worlds, where a world is constituted by a set of facts. Each temporal point of view carries its own world with it. Some facts, say the fact that e = mc2, may constitute more than one world.30 But for each world, there are also facts that peculiarly constitute it, in the sense that they constitute no other world. These facts can be expressed only from the corresponding temporal point of view. For any one of these facts, there may be facts constituting other worlds

28 I said in n. 7 that the canonical way of giving the truth conditions of a representation is to say that it is true if and only if things are a certain way. It does not follow that saying that a representation is true if and only if things are a certain way always constitutes giving its truth conditions. It may constitute giving other, correlated conditions. (See further below, in the main text.) 29 Cf. Dummett (1978d), p. 373 and Perry (1993), pp. 45–46. And cf. Frege (1967), pp. 24–26, for a similar view concerning personal points of view. Note, however, that Michael Dummett, in the first of these references, is talking about a particular philosophical thesis (a kind of anti-realism about the past, whereby the meanings of past-tense representations are given in terms of the conditions that we can now recognize as establishing their truth or falsity) which is such that, if any statement of it is true, then what makes that statement true cannot be expressed at a later time—though the thesis in question accepts the unreality of tense, as characterized here (see e.g. Dummett (1978d), pp. 363–364). It follows that the issue whether tense is real, as characterized here, is not the same as the issue whether we should endorse the philosophical thesis in question, even though that thesis has importantly similar consequences. It may also follow, ultimately, that that thesis is internally incoherent. 30 To say that a fact constitutes a world is an elliptical way of saying that it belongs to the set of facts that constitute that world.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

152

Language, World, and Limits

that correspond in some specially intimate way to it (in the sense alluded to above). But such correspondence falls short of identity. This picture does not seem to me to admit of any simple refutation.31 To be sure, if there is a correspondence of the kind just referred to, then there are awkward questions that remain to be addressed about its nature, and about just what sort of guarantee there is that, if one of these facts obtains, then all the corresponding facts obtain too. But those who adopt this picture are not, per se, under any obligation to answer these questions. For they are not, per se, committed to there being any such correspondence. Indeed they may expressly deny that there is. Thus suppose that the following entry occurs in a forty-year-old diary of mine: ‘Running a temperature today.’ They may challenge my right to say of this entry that it was true if and only if I was running a temperature on the day in question,32 in order to accommodate a certain kind of anti-realism about the past. (I have in mind a view that allows for the possibility that, although I am entitled to say of the diary entry that it was true, say because there is evidence that I was always impeccably accurate in my diary entries forty years ago, I am not entitled to say that I was running a temperature on the day in question, because there is no longer any evidence available that I was—a possibility that rests on a distinction between evidence that is available today for my running a temperature then and evidence that is available today for evidence that was available then for it.33) Less radically, they may challenge my right to say the same sort of thing about an utterance I made yesterday in the future tense. ‘Less radically’ because the corresponding view about the future arguably constitutes a kind of common sense. Thus suppose I said yesterday, ‘I shall be in London tomorrow.’ And suppose I am now in London. Even so, according to the view I have in mind, the future is open in such a way that my utterance yesterday was not (yet) true: it was not (yet) the case that I 31 Two comments are in order about how this squares with what I say in Moore (1997). First, in Ch. 3 of that book, I seem to regard the view that tense is real as incoherent. I do, but only because I am in effect taking for granted the unity of reality (see e.g. p. 49). Granted the unity of reality, the view that tense is real degenerates into an irremediable muddle about whether the facts that make tensed representations true or false obtain only relative to a temporal point of view or not. Secondly, the view that tense is real, as characterized here, seems to be nothing but the view that tensed representations enjoy what I call in the book ‘inherent perspective’ (p. 15). Yet I am perfectly happy, in the book, to acknowledge inherent perspective; I do not seem to think that it poses any threat to the unity of reality (see e.g. pp. 50–51). Have I changed my mind about this? No. The threat to the unity of reality comes not just from the inherent perspective in tensed representations, but from that together with the fact that, for most tensed representations, there are obvious candidates from other temporal points of view to answer to the same facts. What someone who believes in the reality of tense has to explain is not just why it is impossible to give the truth conditions of a tensed representation from another temporal point of view, but why it is impossible to do so even by means of one of these candidates; why, for instance, I cannot give the truth conditions of yesterday’s utterance of ‘It is humid today’ by saying that it was true if and only if it was humid yesterday. I can see no other explanation but that there are different worlds associated with different temporal points of view. (This means, in the terms of the book, that if tense is real—on this conception—then not only is it impossible to endorse a tensed representation from another temporal point of view, it is impossible indirectly to endorse a tensed representation from another temporal point of view: see pp. 15–16.) 32 Although the view that tense is real does not impugn my right to say this, neither does it secure my right to say it. 33 This is a variant of the thesis mentioned in n. 29. Cf. Wright (1987), §VI and Campbell (1994), Ch. 7. (In §5 of the latter Campbell suggests that the view is unstable. But on p. 248 he admits that he is prescinding from the possibility that evidence that there was evidence for something is to be distinguished from evidence for that thing.)

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

153

would be in London today; there was no fact obtaining then that corresponds to the fact obtaining now that licenses my saying today, ‘I am in London.’34 There are some large and fascinating philosophical questions, incidentally, about how these worlds compare with the possible worlds acknowledged by modal realists, or with the worlds acknowledged by those who accept the ‘many worlds’ interpretation of quantum mechanics.35 Discussion of these questions would take us too far afield. I shall simply make the following three observations, which I think highlight important analogies and disanalogies between these worlds and other kinds of world. First, each of these worlds is part of reality. (The view is after all that tense is real.) Secondly, each of them is part of reality at the expense of something else. In other words, there are other worlds that reality might have contained: there are other ways that the facts might have been. Thirdly, an object that appears in one of these worlds, in the sense that it figures in a fact that constitutes one of these worlds, can also appear in another.36 In particular, a representation produced from one temporal point of view can be referred to from another temporal point of view. Nevertheless no tensed representation answers to any facts except those that constitute the world associated with the point of view from which it is produced.37

4. A failed argument that tense is unreal on this construal, based on Kant’s deduction of the categories The view that tense is real abnegates the unity of reality, then. The question now arises whether this can be turned into an objection to the view. Is the unity of reality something that can be argued for? One might think that Kant’s deduction of the categories provides such an argument. In particular, one might try to extract the following argument from the deduction. Given any judgement I make, it must be possible for the ‘I think’ to accompany that judgement. It must be possible for me to think, not only that this is how things are, but that I think this is how things are. Otherwise the judgement would not be one of mine. Alternatively, if something is to count as a judgement, then its elements must be combined together in a certain way. They must be held together in a certain unity. And such unity must be capable of being acknowledged by the judge—by me, if the judgement is one of mine. But such unity is what I myself supply in making the judgement. So in acknowledging the unity I am in effect acknowledging the judgement as one that I have made. I am acknowledging that I think this is how things are. And the 34 If this is right, or if there is a failure of correspondence of any other kind, then not only does each of the worlds fail to contain all the facts, it fails even to ‘mirror’ all the facts in Leibnizian fashion—unless, perhaps, there is a last moment of time. See Leibniz (1973), §§56 and 61. Cf. Dummett (1981a), pp. 391 ff.; and cf. n. 27. 35 The locus classicus for the first of these is Lewis (1986), and for the second Everett (1957). 36 Indeed an object’s thus appearing in more than one world is a necessary condition of its persisting through time. However, it is not a sufficient condition of the object’s persisting through time. There can be facts about objects that no longer exist. Those who adopt this picture may even find it natural to deny, and are at liberty to deny, that objects ever do, strictly speaking, persist through time. 37 The vast literature on these issues includes Evans (1985b); Percival (1992); Butterfield (1995); and Fleming (1995).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

154

Language, World, and Limits

unity is the same for any judgement I make. It is what constitutes the judgement as one of mine. It is the unity of my own apperception. But now let f1 and f2 be any two facts. Assuming that it is in principle possible for me to judge that each obtains, then it must be in principle possible for me not only to make each of the two relevant judgements but to acknowledge the unity in which each of them is held together, the same in each case, the unity of my apperception. But then it must be in principle possible for me to acknowledge both judgements as ones that I have made. Hence it must be in principle possible for me to make a single judgement that embraces both. To telescope the argument: making separate judgements that f1 obtains and that f2 obtains enables me to make separate judgements that I think that f1 obtains and that I think that f2 obtains, which enables me to make a single judgement that I think that both f1 and f2 obtain, which enables me to make a single judgement that both f1 and f2 obtain. This means that f1 and f2 are part of the same world. They are part of my world. The unity of my apperception signals the unity of reality. To quote Kant: ‘A judgement is nothing but the manner in which given modes of knowledge are brought to the objective unity of apperception. This is what is intended by the copula “is”.’38 This argument fails, however. More particularly, there are Kantian reasons why it fails. Perhaps the most obvious objection to it is that it rests on the unargued assumption that, for any two facts, it is in principle possible for me to judge that each obtains. But this is not the objection I wish to focus on, for two reasons. First, this assumption is not one that an advocate of the reality of tense is liable to query, provided that I can make judgements from different temporal points of view, and pending worries about times at which I do not exist. (I shall say some more about this proviso below.) Secondly, the assumption is not one that prevents the argument from succeeding in Kantian terms. Suitably qualified and suitably understood, it is an assumption that Kant himself would have been prepared to grant.39 The objection I wish to focus on is rather this. Even if it is possible for me to acknowledge each of two separate judgements I make as held together in the unity of my apperception, it does not follow that I can simultaneously acknowledge them both as held together in that unity, or at least not in such a way that I can endorse them both. If the two judgements are made from two different temporal points of view, then it is straightforwardly question-begging to suppose that I can do more, whether from one of those points of view or not, than see them as two judgements answering to how things are in two worlds. It is straightforwardly question-begging, in other words, to suppose that I can integrate them into a single judgement. Even though, in making the two judgements, I bring each of them into the unity of my apperception, there remains the possibility that I do so only as two essentially separate cognitive acts, the later one of which precludes any repetition of the earlier. To exclude this possibility, one would need to appeal to something about the character of the judgements themselves. One would need to show that it is in the nature of the judgements themselves to allow for such repetition.40 38 Kant (1933), B141. 39 This is for reasons that I hope will soon be clear. (Note: the assumption is interestingly related to, though importantly different from and independent of, Kant’s famous claim in his ibid., A493/B521, that ‘[to say] that there may be inhabitants in the moon, although no one has ever perceived them, . . . only means that in the possible advance of experience we may encounter them.’) 40 Cf. Allison (1983), p. 162.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

155

But perhaps the very fact that each judgement is one of mine allows for this? Perhaps the two judgements cannot count as two judgements by the same person unless, in the passage from the point of view from which one judgement is made to the point of view from which the other is made, there is provision for a kind of ‘constancy’ of judgement, or a kind of ‘retention’ of judgement—the sort of thing that allows me at one time to see that things are precisely as I earlier predicted they would be, or to remember that things are precisely as I earlier saw they were. If I cannot now make any judgement that answers to the same fact as any of the tensed judgements I have made in the past, then we lose our grip on the idea that it really was I—not merely some counterpart of me—who made all those judgements.41 This suggestion, I think, cuts very deep. But it is ineffectual against an intransigent advocate of the reality of tense who is prepared to concede that it was indeed not strictly speaking I, but merely some counterpart of me, who made all those judgements. Such a person might argue that the unity of apperception, which enables any judgement to be made, cannot carry over from one world to another; that this would be contrary to its very unity; that the makers of judgements are essentially ephemeral.42

5. Why there is no arguing that tense is unreal on this construal I said that the argument given for the unity of reality fails even in Kantian terms. This is significant vis-à-vis Kant’s project in the deduction. Kant’s project in the deduction is not to defend the unity of reality. It is to defend our right to apply the categories as we do.43 And, as Dieter Henrich has famously argued, he undertakes to provide this defence, at least in the version of the deduction that appears in the second edition of his Critique of Pure Reason, in two steps.44 In the first step, he tries to show that we are justified in applying the categories to what is united in a certain way. In the second step, he tries to show that what we in fact apply the categories to—that which is in space and time—is indeed united in that way.45 In the second step he appeals to the fact that ‘space and time are represented a priori not merely as forms of sensible intuition, but as themselves intuitions which contain a manifold . . . and therefore are represented with the determination of the unity of this manifold’.46 That is, he appeals to the fact, to which he has already appealed and 41 Cf. Campbell (1994), Ch. 7. Cf. also what Gareth Evans calls ‘keeping track of the passage of time’ in Evans (1985a), pp. 309–310. But note that what Evans has in mind here does not just involve producing a representation at one time that answers to the same fact as a representation produced at an earlier time; it involves in some sense reproducing the very same representation. (This means that there is an issue about whether, and how, the representation counts as being from a temporal point of view at all.) 42 Cf. n. 36. 43 Kant (1933), A84 ff./B116 ff. The ‘categories’ are twelve fundamental a priori concepts of ours that Kant claims to have identified. A prime example is the concept of causation. 44 Henrich (1982). For what follows, see esp. §11. Cf. also Turetzky (1998), pp. 92–93. 45 The first step is completed by the end of §20. The second step is given, after some incidental observations, in §26. 46 Kant (1933), B160, his emphasis.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

156

Language, World, and Limits

attached much importance in ‘The Transcendental Aesthetic’,47 that space and time, and therewith that which is in space and time, are given to us as having the requisite unity. ‘Fact’ is the operative word here. Kant thinks it conceivable, in the abstract, that things should have been different. ‘Appearances might very well be so constituted,’ he writes, ‘that understanding should not find them to be in accordance with the conditions of its unity.’48 That is, they might be so constituted that they should lack the unity necessary for us to be justified in applying the categories to them. Had they been made up of different worlds associated with different temporal points of view, this would have been the case. We should have had no justification, for example, in applying the category of causation across such worlds. This is reflected in the fact that we could not have used the schema ‘p because q’ to adjudge that one of the facts that (peculiarly) constituted one world was causally dependent on one of the facts that (peculiarly) constituted another. We could not have used this schema because the two facts would have had to be represented from two incompatible points of view.49 We can now see better why the argument given in the previous section fails even in Kantian terms. It relies on certain highly abstract considerations about judgement and unity. But it is clear from the way in which Kant proceeds in the deduction that he does not think that the unity of temporal reality can be argued for in any such way, certainly not in any such abstract way. For Kant, this unity is something more like a brute fact, a fact about how time is given to us. If it is possible for someone to integrate two judgements from different temporal points of view into a single judgement, this is in part because of the quintessence of time (the quintessence of the ‘space’ in which each of the two different points of view is located). As for the way in which time is given to us, on Kant’s conception, this is through the ‘transcendental synthesis of imagination’, where ‘transcendental’ synthesis is defined as synthesis that ‘not merely . . . [takes] place a priori, but also . . . [conditions] the possibility of other a priori knowledge’, and ‘imagination’ is defined as ‘the faculty of representing in intuition an object that is not itself present’.50 It follows that time is given to us as having parts, as something that is partly past and partly future: imagination is the faculty that provides for the possibility of memory and expectation. Now this in turn may seem to conflict with the idea that time is given to us as having unity. But it does not. Past and future are indeed different parts of time, but they are parts that are (transcendentally) synthesized into a whole. They are features of a single temporal world. (This is why time is not only given to us as having unity, it is given to us as having the kind of unity that justifies our applying the categories to it and to its contents.51) Through memory we are able to recall, and thereby to represent, the very facts that we once witnessed as present. Through expectation we are able to anticipate, and again thereby to represent, the very facts that we shall later witness as present.

47 Ibid. A24-5/B39 and A31-2/B47. 48 Ibid. A90/B123. 49 To say that two points of view are incompatible is to say that no representation could be from both. 50 Kant (1933), B151, his emphasis. See also B151 ff. 51 See esp. §26 of the 2nd edn version of the deduction in ibid. Cf. Allison (1983), pp. 160–164, and Turetzky (1998), p. 93. Cf. also what Kant himself says later, at (1933), A581–582/B609–610, which, together with the footnote at A572/B600, in turn merits a striking comparison with Wittgenstein (1961), 2.0124 and 5.524.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Apperception and the Unreality of Tense

157

If this is on even broadly the right lines—and I am sure it is—then tense is unreal. The unreality of tense, or equivalently the unity of temporal reality, is of a piece with the unity of apperception. We can achieve insight into it through a kind of self-conscious reflection. But we cannot argue for it. Thus if we consider our own tensed representations and reflect on how we conceive them and what they answer to, then we can come to see them as representations, from different positions within a single world, of that single world—just as, if I consider my own representations (tensed or tenseless) and reflect on how I conceive them and my capacity to produce them, I can come to see them as representations produced by a single self-conscious subject occupying different positions within a single world. Even so, if someone resolutely refuses to acknowledge the unity of temporal reality—if someone conceives of tense as a feature of reality—then there is no more basic principle we can adduce to force a change of mind. The unity of temporal reality is indeed, to that extent, akin to a brute fact. There is a great deal to be said about it. But there is nothing, or at least nothing with any suasive power, to be said in favour of it.52, 53

52 This echoes a recurring theme of Moore (1997). I there make frequent use of two principles that are different expressions of the unity of reality: what I call the Fundamental Principle (pp. 21–22) and what I call the Basic Assumption (p. 74). I claim that such principles cannot be established by any arguments: see e.g. pp. 188–189. Cf. in this connection Michael Dummett’s remarks about what he calls ‘our prejudice’ that there must be a complete description of reality (Dummett (1978c), pp. 356–357), and John Campbell’s comments about the ineluctability of a realist view of the past (Campbell (1994), p. 4). And see again the remarks in n. 4. 53 I am extremely grateful to John Bigelow, Jeremy Butterfield, John Campbell, David Cockburn, Michael Dummett, Naomi Eilan, Christoph Hoerl, Teresa McCormack, Helen Steward, and especially Philip Percival for comments on earlier drafts of this essay.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

12 The Metaphysics of Perspective: Tense and Colour Abstract This essay first appeared as a contribution to a symposium on Barry Stroud’s book The Quest for Reality. Stroud is deeply sceptical in that book about whether we can successfully undertake what he calls ‘the philosophical quest for reality’. This essay is an attempt to defend the idea that we can successfully undertake such a quest. One way of doing this, it is proposed, would be by arguing that any fact can be represented from no point of view. The essay exploits various analogies between tense and colour—among other resources—to rebut reservations that Stroud might be expected to have about this proposal.

1. One way of successfully undertaking the philosophical quest for reality Is tense real? It is as difficult to know how to interpret this question as it is to know, on any reasonable interpretation, which side to take. But I am persuaded that there are reasonable interpretations, by which I mean interpretations whereby neither the view that tense is real nor the view that tense is unreal is crazy. On one such interpretation, to affirm that tense is real is to be committed to the following, and to affirm that tense is unreal is to be committed to its denial. Given any true tensed representation, in other words given any true judgement, thought, assertion or suchlike from a particular temporal point of view, there is no conveying what makes it true except from that same point of view. Thus suppose I know that it is humid today. Then what makes my knowledge true is the fact that it is humid today. But this is a fact that can be conveyed only today. If I say tomorrow, ‘It was humid yesterday,’ that will not convey the same fact. At best it will convey some intrinsically related fact, about (as it were) hesternal humidity, which can itself be conveyed only tomorrow. Reality fractures into different temporal worlds. The facts that peculiarly constitute one of these worlds can be conveyed only from the corresponding temporal point of view.1 1 Why might anybody think this? One reason would be to accommodate an intuition which many people have, that the future is open, which is to say, roughly, that nothing is the case at any given time about what is contingently the case at later times. Suppose the future is open. And suppose I said yesterday, ‘It will be

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Metaphysics of Perspective

159

I myself would deny that tense is real on this interpretation. But I am not concerned to defend that position now. I am concerned to signal the fact that there is a live metaphysical issue here about the character of reality; and that it is an issue with obvious potential for generalization. The potential arises because the issue is an issue about points of view. Granted that talk of temporal points of view is already metaphorical, there is the prospect of analogous issues concerning points of view of other kinds, captured by some further metaphorical extension of the notion. Thus we can ask: is perspective of any other kind real? Are there points of view of any other kind such that, given a true representation from one of these points of view, there is no way of conveying what makes the representation true except from the same point of view? Or is there perhaps some general reason, having to do with the very unity of reality, for denying that such a thing is ever possible, however broadly the notion of a point of view is construed? Notice that these questions have no direct connection with the question whether there can ever be representations that are not from any point of view. The possibility of such representations does not immediately preclude the reality of some kind of perspective: perhaps there are limits to what can be achieved with representations from no point of view. Nor, conversely, does the impossibility of such representations immediately guarantee the reality of some kind of perspective: perhaps, in order to convey what makes a true representation from a point of view true, it is necessary to do so from some point of view or other, even if not from that one. Nevertheless, although there are no direct connections between these questions, there are indirect connections. The view that there can be representations that are not from any point of view, or absolute representations as I shall call them, does go naturally with the view that perspective of any given kind is unreal. And anyone who holds both these views—as I do—is but one step away from the following ambitious conclusion. (A)

Any fact can be conveyed from no point of view.

I cannot even begin to rehearse arguments for (A) here.2 But I do want to insist that (A) is an interesting and defensible claim, with far-reaching consequences, both about the character of reality and about the metaphysics of perspective. Arguing for (A) would be one way, I believe, of successfully undertaking what Barry Stroud calls ‘the philosophical quest for reality’ in his fascinating and thought-provoking book The Quest for Reality.3 This belief will play a sufficiently prominent rôle in what follows to merit a label of its own. Thus: humid tomorrow.’ Then not only did this fail to convey the fact that verifies my knowledge that it is humid today. It was not even true. For it was not (then) the case that it would be humid today. By far the most natural way to capture this idea, if not the only way, is by appeal to different temporal worlds; more specifically, by appeal to a sequence of temporal worlds such that those later in the sequence contain details corresponding to gaps in those earlier in the sequence. (For further related discussion see Le Poidevin and Macbeath (1993), Introduction and Pt. 1.) 2 I try to do so in my (1997). 3 Stroud (2000)—the subject of a symposium in which this essay first appeared—p. x. All unaccompanied references will be to this book.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

160

Language, World, and Limits

(A) To argue for (A) would be one way of successfully undertaking the philosophical quest for reality. Now Stroud is sceptical about whether the philosophical quest for reality can be successfully undertaken. He does not argue outright that it cannot. He is aware, amongst other things, of the threat of self-stultification in doing so.4 But, by probing and dissecting various manifestations of the urge to undertake the quest, he makes clear the sorts of reservations he has. And, although he does not address (A) head-on, there is plenty in his book to indicate where, in (A), those reservations would find targets. What I should like to do in this essay is to deflect some of these reservations.

2. Two categories of reservation about (A) In general, reservations about (A) will fall into two categories: those that concern the truth of (A); and those that concern the capacity of (A), even if true, to provide what is required. The second category will include reservations about how far the notion of a point of view can be extended. Many people will deny that it can be extended far enough to give (A) any metaphysical bite. Although the two categories are not merely separate, but in some tension with each other, sceptics about (A) are liable, in practice, to voice reservations of both sorts. This is not least because defendants of (A) are liable, in practice, to respond to successive attacks on their position by alternately weakening and strengthening (A), for instance by alternately contracting and expanding the notion of a point of view. There is enough in Stroud’s book to indicate how he would react if he found his target moving in this way. I shall accordingly cull reservations of both sorts from his book, though obviously I hope that my own conception of (A) is sufficiently stable that, were I to try to present Stroud with a defence of it, his scepticism would eventually settle—in a way that would make some of these reservations look, to him as well as to me, irrelevant.

3. Two preliminary observations Before I proceed I want to make two observations about the particular focus that Stroud gives his book. First, although he begins at the same high level of abstraction at which we now are, by the end of Chapter 2 he has narrowed his discussion down to a discussion of colour. It is not that he is unconcerned with the more general metaphysical issues at stake. It is rather that he doubts whether we can get far with these unless we have a sharper focus. Indeed that doubt is itself part of his overall scepticism about the philosophical quest for reality. When, in his final chapter, he warns of the dangers of generalizing from what he has been arguing about colour, there is a sense in which this adds to the force of his arguments rather than detracting from their force. Although I too shall focus to some extent on colour, I have given my discussion an additional focus on tense, because I think that this will assist me in my project of deflecting reservations about (A). But of course, the most obvious reservation about (A), as far as its application to colour is concerned—a reservation that belongs to the 4 P. xi.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Metaphysics of Perspective

161

second category—is precisely that no extension of the notion of a point of view entitles us to assimilate tense and colour in this way; that colour is not a matter of perspective at all. Indeed Stroud spends much of his book—in effect—developing this very reservation. Obviously I do not share it. It seems to me that what makes tense a  matter of perspective has something fundamental to do with the following fact: if I say truly, on 25 June, ‘It is humid’, and if I say truly, the following 25 December, ‘It is snowing’, then it is not possible to convey what makes my two representations true simply by saying, ‘It is humid and it is snowing.’ This impossibility, it seems to me, is importantly reflected in the impossibility of conveying, simply by conjoining representations of the two relevant types, what makes true both my claim that lemons are yellow and a true representation produced by some creature, even some imaginary creature, whose sensory apparatus is incompatible with normal human vision. Not that I expect these highly schematic remarks to satisfy Stroud. But I hope they do enough to prevent what follows from being void of interest even before it starts. At any rate I shall assume henceforth that colour is a matter of perspective. My second observation concerns the fact that Stroud several times adverts to physics. The concepts of physics might be expected to provide flesh for the skeletal idea of an absolute representation. And it is in this sort of connection that Stroud mentions them. But here too he finds a natural target for his scepticism. For, as he complains5— again, in effect—unless ‘physics’ is understood in a question-beggingly indeterminate way, then there is no good reason to expect its concepts to provide flesh for these bones. I agree. Even so, I do think that there is good reason to expect absolute representations to be couched in physical terms. This is simply because I do not understand this expectation in a relevantly meaty way. I understand it as a defining characteristic of physics, or better, perhaps, as a directive for physicists.

4. Six reservations about (A) and their rebuttal I shall now list various reservations about (A) that I think can be found in Stroud’s book. (Of course, since Stroud never explicitly addresses (A), I have, in most cases, had to extrapolate from what he says. It is for Stroud to tell me if I have extrapolated so far that the reservations are no longer his.) First Reservation: (A) is not itself absolute. This reservation belongs to the first category. It derives from Stroud’s observation that ‘if we assert [all the sentences in the vocabulary of current physical science that are taken to be true] . . . we will not so far have said that the world we believe in is a physical world; . . . [we] will have said things in a certain vocabulary about the world, but we will not have said anything about that vocabulary’.6 Stroud’s point is that physicalism is not itself the stuff of physics. On my non-meaty understanding of ‘physical’ and its cognates this is pretty much equivalent to the claim that (A) is not itself absolute. Indeed it is not. (For one thing, talk of ‘conveying facts’ has to be from some interpretative point of view.) But this does not prevent (A) from being true. There would 5 On p. 53.

6 P. 52, his emphasis.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

162

Language, World, and Limits

of course be a threat to the truth of (A) if, granted its truth, it conveyed a fact that could only be conveyed in that way. But there is no reason to suppose that this is the case. Indeed if the truth of (A) is conceptual, as I take it to be, then there is good reason to construe the notion of a fact in such a way that (A) does not convey any fact at all—or in such a way that it conveys a fact that could just as well be conveyed, from no point of view, by saying, ‘0 = 0’. Second Reservation: Many facts are facts about particulars, but it is impossible to refer to particulars from no point of view. This reservation likewise belongs to the first category. It derives from scepticism that Stroud voices7 about the possibility of expressing the fact that the earth is 93 million miles from the sun, in suitably general terms. But is this what (A) requires? What (A) requires is the possibility of ‘conveying’ the fact that the earth is 93 million miles from the sun in suitably general terms. (In suitably general terms, because it is indeed impossible to refer to particulars from no point of view.) Admittedly, I have so far made free use of this notion of ‘conveyance’ without any explanation. But all that matters for these purposes is that conveying a fact can involve conveying not just that fact, but more besides: conveyance is a sort of implication. Stroud anticipates such a response. He extends his scepticism8 to the possibility of even implying that the earth is 93 million miles from the sun in suitably general terms. In particular, he points out, this cannot be done by saying that there is a planet of a certain kind which is 93 million miles from a star of a certain kind. True; but can it not be done by saying that any planet of a certain kind is 93 million miles from a star of a certain kind, where the earth and the sun are understood to be, or at least can be understood to be, this planet and this star of those respective kinds? Third Reservation: Many facts are facts about people’s psychology, but no amount of knowledge from no point of view can suffice for knowing anything about anyone’s psychology. This is another reservation belonging to the first category. In fact it is a variation on the second reservation. It forces us to consider once again how much ‘conveyance’ requires. There is an implicit statement of this reservation in Stroud’s claim9 that ‘if we restrict ourselves to [a full account of what goes on physically in a certain area during a certain period of time], we will not know whether anything psychological occurs during that period’. Perhaps we will not. But likewise, if we restrict ourselves to a full tenseless history of the world, we will not know what is happening now. Yet surely there is a good sense of conveyance in which such a history will convey what is happening now. It remains to be shown that it is not also good enough to deflect this third reservation.10 Fourth Reservation: Even if facts about the colours of things could be conveyed from no point of view, without the use of colour concepts, it would not follow that colour was unreal. 7 Pp. 55 ff. 8 Pp. 57–58. 9 P. 89. 10 This is a summary allusion to an extremely familiar and large debate. For two important contributions to the debate see Lewis (1990) and Nemirow (1990).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

The Metaphysics of Perspective

163

This reservation belongs to the second category. It finds expression towards the end of Chapter 3.11 And it invites, in response, use of that most familiar of philosophical tactics: it all depends on what you mean when you say that colour is unreal. Reconsider tense. I began this essay by specifying a debate about the reality of tense which, if (A) is true, must be settled negatively. In that sense (A) entails both the unreality of tense and the unreality of colour. But (A) does not entail the unreality of tense in any sense that means that we are wrong to think of ourselves as now living in the third millennium AD. Nor, by the same token, does it entail the unreality of colour in any sense that means that we are wrong to think of lemons as yellow. In this respect, much of Stroud’s extremely interesting and powerful critique of what he calls an ‘unmasking’ explanation of our belief in colour, that is to say an explanation that ‘explains [our] belief in [colour] without having to suppose that that belief is true’,12 is beside the point. (A) does not entail the falsehood of any of our beliefs about colour, save the beliefs of those few metaphysicians who have pondered whether, in the offending sense, colour is real and who have decided that it is. (A) even allows that no explanation of ‘our belief in colour’, in any sense in which we do by and large have that belief, could get anywhere without the supposition that the belief is true. As for whether this deprives (A) of its metaphysical bite, surely there is bite enough in the claim that colour is unreal, in the only sense in which (A) entails that it is. Fifth Reservation: It is not clear that we can draw any reasonable and relevant distinction between a sense in which colour is real and a sense in which it is not. This reservation obviously reinforces the fourth, in light of the response just given. Stroud voices it in terms of what he calls a ‘felt conflict . . . between two opposed conceptions of . . . what the world is like’, a conflict that he says arises from ‘two apparently incompatible ways’ of answering the question ‘whether objects are coloured’.13 But once again I urge the comparison with tense. The view that tense is unreal, as identified at the beginning of this essay, is not in any conflict with the view that some events have occurred, others are occurring, and others are yet to occur. Sixth Reservation: No expansion of the notion of a point of view capable of sustaining this comparison between colour and tense would leave us ‘with [any] determinate beliefs about the world at all’.14 This reservation reinforces the fifth, and thus, indirectly, the fourth.15 The underlying worry is that the notion of a point of view is being extended so far that, for any given belief, the conceptual apparatus needed to frame that belief is itself a point of view that the belief is from. This in turn threatens a vicious regress: in order fully to specify the content of any given belief, it is necessary to make explicit mention of this conceptual apparatus; but then it is necessary also to make explicit mention of the conceptual apparatus needed to frame this fuller specification; and so on ad infinitum. This seems to preclude a full specification of the content of the original belief, which in turn suggests that the belief has no determinate content at all. There is much to say in response. First, of course, once the notion of a point of view has been extended that far, then there is an altogether more urgent concern 11 E.g. p. 61.

12 P. 75.

13 P. 186.

14 P. 187.

15 It finds expression on pp. 186–187.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

164

Language, World, and Limits

about (A): namely, that (A) is false. This brings us back to the question of whether a balance can be struck between maintaining the truth of (A) and maintaining its capacity to provide what is required of it. I still see no reason why not. But in any case I think this reservation betrays certain misunderstandings about the very idea of a point of view. Let b be a belief from a point of view, however narrowly or broadly construed. Then it is not clear in what sense of ‘full specification’ a simple expression of b does not already constitute a full specification of its content. (If I believe that it is humid today, then the content of my belief is that it is humid today, neither more nor less.) Suppose, however, that some such sense of ‘full specification’ can be given. Suppose, in particular, that in order fully to specify b’s content, it is necessary to make explicit mention of any point of view that b is from and of any point of view that one’s own explicit mention of any point of view is from. Suppose, finally, that the resultant regress cannot be blocked by fully specifying b’s content from no point of view. (This may be true even if (A) is true: the notion of a point of view may itself be unfit to appear in an absolute representation.) Then it is not clear why, in order for b’s content to be determinate, it has to admit of full specification.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

13 Realism and the Absolute Conception Abstract This essay is concerned with Bernard Williams’ argument for the possibility of an absolute conception of reality, that is to say a conception of reality that is not from any point of view. The primary aim of the essay is to see what we can learn from this argument about its underlying realism, which Williams expresses as follows: ‘knowledge is of what is there anyway’. To accept such realism—it is claimed in the essay—is to adopt a conceptual structure that involves, among other things, a commitment to the unity of reality. This helps Williams to express and to motivate his opposition to realism about ethics, where realism about ethics, despite its name, is quite different from realism tout court: realism about ethics is the view that the best reflective explanation of our having the ethical knowledge we have serves as a direct vindication of that knowledge.

1. Realism, science, and ethics It is often said that Bernard Williams opposes ethical realism. And so he does.1 But what does this mean? The term ‘realism’ has a notorious and bewildering variety of uses. What does Williams oppose? The first and most basic thing that needs to be emphasized is that what he opposes is just what its name implies: realism about ethics. This highlights something that is becoming increasingly standard in philosophical uses of the term ‘realism,’ namely, its relativization to a subject matter. Granted such relativization, a realist about history may or may not be a realist about mathematics, say. Indeed, we shall see in due course that Williams’ opposition to realism about ethics is to be understood precisely in contrast with his commitment to realism about science. But here already there is a complication. For the term ‘realism’ is also sometimes used without relativization. We sometimes hear it said of a given philosopher that he or she is a realist tout court. More to the point, we sometimes hear it said of Williams. Moreover, I think this is an appropriate thing to say of him, properly understood.

1 For an early indication of this opposition, see Williams (1973). For later dissatisfaction with the early way of putting it, see Williams (1996), p. 19.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

166

Language, World, and Limits

I also think it is an appropriate point of leverage in the attempt to understand his position. Williams’ realism—tout court—receives famous and memorable expression in his book on Descartes, where he writes, ‘Knowledge is of what is there anyway.’2 This is his summary way of putting what he describes in the previous sentence as ‘a very basic thought,’ namely that if knowledge is what it claims to be, then it is knowledge of a reality which exists independently of that knowledge, and indeed (except in the special case where the reality known happens itself to be some psychological item) independently of any thought or experience.3

This is a basic realism which is not itself tied to any particular subject matter.4 Grafted on to this unqualified realism is the distinction that most concerns Williams, a distinction between different ways of explaining how we come by the knowledge we have. It is this that underlies the contrast he wants to draw between science and ethics. The idea is not that we do not have ethical knowledge.5 Nor is the idea that the ethical knowledge we do have is not ‘what it claims to be’ and so lies outside the ambit of his unqualified realism.6 The idea is rather that the best reflective explanation of our having the ethical knowledge we have, unlike the best reflective explanation of our having the scientific knowledge we have, cannot directly vindicate that knowledge: it cannot directly reveal us as having got anything right.7 The position that motivates this idea is roughly as follows. We (human beings) not only inhabit a reality that is there anyway. We also inhabit different social worlds that we have created for ourselves. Part of what it is to inhabit a particular social world is to operate with a particular set of what Williams calls ‘thick’ ethical concepts. By a ‘thick’ ethical concept Williams means a concept whose applicability is both ‘actionguiding’ and ‘world-guided’. Examples are the concepts of infidelity, blasphemy, and racism. To apply a thick ethical concept in a given situation, for example to accuse someone of infidelity, is, in part, to evaluate the situation, which characteristically means providing reasons for doing certain things; but it is also to make a judgement that is subject to correction if the situation turns out not to be a certain way, for 2 Williams (1978), p. 64, his emphasis. 3 Ibid. 4 Of course, it immediately suggests at least one thing that could reasonably be meant by realism about any given subject matter, namely, the view that that subject matter admits of knowledge. But, in itself, Williams’ realism is neutral with respect to any such view. This may make it seem rather anodyne. However, it is by no means so anodyne that no philosopher has seen fit to reject it. Many notable philosophers have marshalled many notable arguments against any such realism, in some cases with a view simply to denying it, in other cases with a view, more radically, to repudiating the very concepts in whose terms it is couched. I shall present an example of the latter in section 4. (For further examples, and for further discussion, see Moore (1997), Ch. 5, §8 and Ch. 6.) For my own part, I think Williams’ realism is no more than the intuitive deliverance of reflective common sense. I shall have more to say about this too. 5 See n. 4: the denial that we have ethical knowledge is certainly one thing that could be intended by the rejection of ethical realism, particularly when it takes the form of a denial that talk of ethical knowledge so much as makes sense. But that is not what Williams intends. 6 Or at least—as I have tried to argue in Moore (2003), pp. 347–348—the idea had better not be that. That had better not be part of what he is getting at in his repeated insistence that ‘ethical thought has no chance of being everything it seems’ (e.g. Williams (1985), pp. 135 and 199). If that were part of what he is getting at, then other doctrines of his, including doctrines that we shall be examining later, would be severely compromised. 7 See esp. Williams (1985), Ch. 8. See also Williams (1995a) and Williams (1995c), pp. 205–210.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

167

example, if it turns out that the person who has been accused of infidelity did not in fact go back on any relevant agreement. In favourable circumstances, a judgement involving a thick ethical concept can be immune to any such correction and can count as an item of ethical knowledge.8 Now the social worlds that we inhabit admit of incompatible rivals in which quite different thick ethical concepts are exercised. Although we need to inhabit some social world, there is no one social world that we need to inhabit. A good reflective explanation for someone’s having a given item of ethical knowledge must therefore include an account of their inhabiting a social world that allows them to have it. This explanation may draw elements from history, psychology, sociology, and/or anthropology. But it cannot itself make use of any of the thick ethical concepts exercised in the knowledge, because it must be from a vantage point of reflection outside their social world. This means that it cannot directly vindicate the knowledge. This contrasts with the case of scientific knowledge. A good reflective explanation for someone’s having a given item of scientific knowledge can make use of the very concepts exercised in the knowledge, and so can straightforwardly and directly vindicate the knowledge, by revealing that the person has come by the knowledge as a result of being suitably sensitive to how things are. Thus Williams’ realism about science, but not about ethics. Here is another way to characterize the position. Inhabiting a social world means having a certain point of view. Ethical knowledge is knowledge from such a point of view. What prevents a good reflective explanation of someone’s having such knowledge from directly vindicating it is the fact that the explanation must include an account of how they have the relevant point of view (where this does not itself consist in their knowing anything). By contrast, there can be scientific knowledge that is not from any point of view. A good reflective explanation of someone’s having such scientific knowledge need not involve the same kind of indirection. This position invites countless questions, of course. For instance, what are the criteria for a ‘good’ reflective explanation? Or for a ‘direct’ vindication of an item of knowledge? But one question that has troubled critics as much as any concerns the science side of Williams’ ethics/science contrast. What reason is there for thinking that there can be scientific knowledge that is not from any point of view? Williams’ own reason for thinking this, familiarly, is grounded in the unqualified realism that forms the basis of his position.9 Taking that realism as a premise, he argues for the possibility of what he calls ‘the absolute conception’ or ‘the absolute conception of reality’ where what this is is, precisely, a conception of reality that both constitutes scientific knowledge and is not from any point of view.10

8 Williams (1985), pp. 140–148. 9 We shall see later (section 3) that ‘basis’ is a somewhat inappropriate metaphor here. For now, we can let it pass. 10 See esp. Williams (1978), pp. 64–65. For further discussion see Williams (1978), pp. 65–68, 211–212, 239, 245–249, and 300–303; Williams (1985), pp. 138–140; Blackburn (1994); Dancy (1993), Ch. 9, §2; Heal (1989), §7.2; Hookway (1995); Jardine (1980); Jardine (1995); Putnam (1992), Ch. 5; and Strawson (1989), Appendix B.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

168

Language, World, and Limits

I have tried to defend Williams’ argument elsewhere.11 In this essay, I am more interested in understanding Williams’ position than in motivating it. In particular, I want to see what the conclusion of his argument can teach us about its premise, the underlying realism.

2. The absolute conception If the conclusion of Williams’ argument is to teach us anything, we need to be clear about what that conclusion is. When I defended Williams’ argument, I prefaced my defence with, in effect, a list of twenty-two things that it is not.12 My list was meant as a safeguard against various possible misconstruals of Williams’ position, many of which I take to be actual. I shall not rehearse that list in full now. But I do want to draw attention to one item on the list that is especially pertinent to this discussion. Williams’ conclusion is not that there are some uniquely privileged God-given concepts waiting to be discovered—as it were, the ‘one true eternal’ stock of concepts that equip us to represent things from no point of view.13 Talk of ‘the’ absolute conception encourages this idea. But there is nothing in Williams to preclude the thought that, if we are to represent things from no point of view, then we shall be involved in decisions between various incompatible but equally legitimate conceptualizations; that these decisions may be highly parochial, in that they may be tailored to certain context-specific needs and interests of ours; that they may be hard-earned, in that they may involve us in intensive conceptual and empirical endeavour; and that it may take long-term active participation and commitment on our part both to sustain these decisions vis-à-vis their rivals and to implement them in the joint process of representing how things are and justifying our representations. McDowell, writing about the absolute conception, caricatures it as involving a picture of ‘science as a mode of inquiry in which the facts can directly imprint themselves on our minds, without need of mediation by anything as historically conditioned and open to dispute as canons of good and bad scientific argument’.14 That is simply unfair. (It is unfair even apart from the point I am making about rival conceptualizations. Williams nowhere denies the need for mediation of the sort McDowell describes in discovering what the facts are, that is in applying whatever conceptual apparatus is in play. It is not clear, in fact, that even if Williams had been committed to there being uniquely privileged God-given concepts, he would have had to deny the need for mediation of the sort McDowell describes in discovering what they are.) 11 Moore (1997), Ch.  4, §3. I may, however, attach less substance than Williams does to the relation between a conception of reality that is not from any point of view and science. I take it to be more or less a defining characteristic of science that, if a conception of reality that is not from any point of view can be couched at all, then it can be couched in scientific terms: see ibid. pp. 75–76. 12 Ibid. Ch. 4, §1. I say ‘in effect’ because I was arguing for a conclusion that is a slight variation on Williams’ conclusion; but I think the differences are inessential. (I was not concerned with completeness. Contrast Williams’ definition of the absolute conception in Williams (1978), p. 65 with what I say in my (1997), p. 64.) 13 See Moore (1997), p. 64. Cf. also ibid. pp. 95–96. (There is a hint that this is Williams’ conclusion in Korsgaard (1996), pp. 68–69. But it is only a hint. What Korsgaard goes on to say seems to me to show great exegetical sensitivity.) 14 McDowell (1986b), p. 380.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

169

Even more unfair, it seems to me, is the related but further idea, all but embraced by McDowell, that the possibility of the absolute conception entails what Davidson calls ‘a dualism of scheme and content’15—a dualism that Davidson, McDowell, and others have done so much to discredit.16 Scheme, according to this dualism, is constituted by concepts; content is that extraconceptual element in reality which we seek to capture, by an imposition of our concepts on it, whenever we represent things to be a certain way. Content is something that we passively receive. Concepts, by contrast, are things that we actively exercise.17 The reason why the possibility of the absolute conception is thought to entail this dualism is, precisely, that it is thought to require uniquely privileged God-given concepts, where part of what uniquely privileges these concepts is in turn thought to be that they constitute a scheme that is, in McDowell’s words, ‘peculiarly transparent, so that content comes through undistorted’.18 But we need not accept that the possibility of the absolute conception requires uniquely privileged God-given concepts. And even if we did accept this, we need not accept that what uniquely privileges the concepts has to be characterized in terms of scheme and content—still less, in terms of ‘transparent’ scheme and ‘undistorted’ content.19 I see no reason, then, to think that Williams’ conclusion entails any scheme/content dualism. A different worry, which is worth pausing to address, is that his premise entails such a dualism. Does not the idea that knowledge is of a reality that exists independently of that knowledge entail that it is of something extraconceptual, something on which we impose our concepts whenever we know anything to be the case? No. Williams’ premise is that knowledge is of a reality that exists independently of being known, not independently of being knowable.20 It does nothing to foreclose the possibility that what is known is essentially conceptual. In fact, it is really nothing but a kind of schematic summary of such commonplaces as this: even if no-one had known that e = mc2, it would still have been the case that e = mc2.21 This commonplace certainly allows for the fact that e = mc2 to be, in McDowell’s words again, ‘essentially capable of being embraced in thought in exercises of spontaneity [that is, in exercises of conceptual capacities]’.22 (Indeed—although this is not really to the point as far as Williams’ premise is concerned—it allows for this without in any way prejudicing the thought that our knowledge that e = mc2 may be part of the absolute conception.23) I have suggested that representing things from no point of view can still leave room for decisions between rival conceptualizations. What sort of thing do I have in mind? I have in mind the sort of thing that Quine has in mind when he suggests that a pair of scientific theories might be ‘empirically equivalent’, in the sense that ‘whatever observation would be counted for or against the one theory counts equally for or 15 Davidson (1984g). See esp. pp. 187 and 189. 16 See e.g. Davidson (1984g); McDowell (1994); and Rorty (1980), esp. Ch. VI, §5. See also Rorty (1991b), pp. 138–139. 17 See again McDowell (1994). See also Child (1994). 18 McDowell (1986b), p. 381. 19 Cf. Williams (1995c), p. 209. 20 For the importance of this distinction, cf. McDowell (1994), p. 28. 21 ‘Commonplace’, as I suggested in n. 4, does not preclude opposition. For an especially stark example of opposition to just this sort of idea (that even if no one had known that p, it would still have been the case that p), see Heidegger (1962a), §44(c). 22 McDowell (1994), p. 28. 23 Cf. Child (1994), pp. 61–62.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

170

Language, World, and Limits

against the other’, yet such that each involves ‘theoretical terms not reducible to’ the other’s.24 He later has a splendid analogy to illustrate this. He writes: [Irresolubly rival systems of the world] describe one and the same world. Limited to our human terms and devices, we grasp the world variously. I think of the disparate ways of getting at the diameter of an impenetrable sphere: we may pinion the sphere in calipers or we may girdle it with a tape measure and divide by pi, but there is no getting inside.25

Suppose now that we have our own system of the world but are also aware of such a rival. (This may be because our choices between conceptualizations have been quite conscious.) Quine raises the question of what we are to say about the rival. He distinguishes two attitudes that we can take. The sectarian attitude, as he calls it, is to repudiate the alien concepts and to regard the rival system as empirically warranted nonsense. (For Quine, this is not the oxymoron it sounds. ‘Empirically warranted nonsense’ is, very roughly, nonsense which, if it did count as sense, would also have the right sort of connection with experience to count as true.) The ecumenical attitude is to acknowledge the alien concepts and to regard the rival system as simply true.26 Two very powerful forces in Quine’s philosophy have made him vacillate over the years between these alternatives. His naturalism has inclined him toward sectarianism. His empiricism has inclined him toward ecumenism. By his naturalism, I mean his conviction that there is no higher authority, when it comes to deciding what is true, than whatever has in fact led us to adopt our own system of the world. By his empiricism, I mean his conviction that there is no other evidence for the truth of a system than its empirical warrant: systems answer to nothing but experience.27 He has eventually settled for sectarianism.28 This is surely the right alternative for Quine. After all, in the case in which we are aware of an empirically equivalent rival system to our own, whose concepts are not incommensurable with ours, he is committed to regarding the rival as, however warranted, false.29 His sectarianism 24 Quine (1990a), §§41–42. The quoted material occurs on pp. 96–97. 25 Ibid. p. 101. (This analogy, incidentally, is curiously equivocal as far as the dualism of scheme and content is concerned. It can be construed in such a way as to provide further ammunition against the dualism. But it can also be construed in such a way as to provide support for it. Quine himself, as it happens, is not hostile to the dualism: see Quine (1981c). For criticisms of Quine on this matter see McDowell (1994), Afterword, Pt. I.) 26 Quine (1990a), §42 and Quine (1986a), pp. 156–157. (Note: on p. 156 of the latter he characterizes sectarianism as the view that the rival system is false rather than nonsense. But this is an aberration. It is subverted on the very next page.) Taking the ecumenical attitude would not commit us ever to exercising the alien concepts. If we chose not to, this would be a little like regarding empirically warranted French sentences as true but choosing only to speak in English. Taking the sectarian attitude would be a little like regarding English as the only real language. 27 For an example of a swing to sectarianism, see Quine (1981a), pp. 21–22. For an example of a swing to ecumenism, see the first edition of Quine (1981b), p. 29. (This is corrected in later editions. The earlier version is quoted in Gibson (1986), p. 153, n. 2.) 28 Quine (1986a), p. 157 and Quine (1990a), p. 100. (This explains the correction referred to in n. 27.) Cf. Rorty (1991a), §2. 29 The possibility of empirically warranted false systems is an immediate corollary of his thesis that truth is underdetermined by evidence. See Quine (1969d), pp. 302–303, in which he also distinguishes between mere underdetermined truth and indeterminacy. For further discussion, see Essay 2 in this volume. (Note: Davidson is surely wrong to claim, as he does in Davidson (2001), p. 76, n. 4, that ‘Quine has changed his mind on the issue [whether there can be empirically equivalent, but incompatible, theories]

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

171

nevertheless leaves him uncomfortable. He recognizes the invidiousness of regarding one system as true and another as nonsense, even though there is no cosmically telling between them and even though it is nothing but a kind of historical accident that one of these systems has our allegiance rather than the other. So he is keen to remind us that we can change our allegiance. The sectarian, he tells us, is as free as the ecumenist to oscillate between the two [systems]. . . . In his sectarian way he does deem the one [system] true and the alien terms of the other meaningless, but only so long as he is entertaining the one [system] rather than the other. He can readily shift the shoe to the other foot.30

This is not to concede, along with the ecumenist, that both systems should be regarded as true. It is not even to concede that both systems can be regarded as true. But it is to concede that each system can be regarded as true. And, as Quine himself admits, to concede this is but one terminological step away from conceding ecumenism. After all, ecumenists and sectarians alike are agreed that, whichever system has our allegiance, we must pay the rival system every compliment we can, short of giving it too our allegiance. Does anything of substance hang on whether this includes calling the rival system ‘true’? But then, come to that, does anything of substance hang on which system has our allegiance? It now looks melodramatic to suggest, as I did earlier, that, when we have decided between two rival conceptualizations, long-term active participation and commitment on our part may be required to sustain our decision vis-à-vis its alternative. It even looks melodramatic to describe the two conceptualizations as ‘incompatible’. In what sense are they incompatible? Well, they are incompatible in the sense that the concepts involved must lead their own separate and independent lives. Or, a little more prosaically, they are incompatible in the sense that it is impossible to exercise concepts in accord with one conceptualization except at the expense of doing so in accord with the other.31 What may require long-term active participation and commitment is, not upholding the selected conceptualization in a way that downplays the other, which is something we have no reason to do, but upholding the selected conceptualization in a way that prevents interference from the other.32 To select, maintain, and implement a conceptualization requires keeping any rivals clearly in focus as rivals. This can take hard work. And it is this that constitutes giving allegiance to the conceptualization, or to any system that uses it. So yes; much of substance hangs on which system has our allegiance; and we had better be clear about which does. The problem now is that operating with one conceptualization rather than another, in the scientific case that we have been considering, is beginning to look very much like operating with one set of thick ethical concepts rather than another. Is not the more than once’. The issue on which he has changed his mind is not that, but what the best construal of such theories is. Cf. again Essay 2, n. 9, in this volume.) 30 Quine (1990a), p. 100. 31 This does not rule out the possibility of combining the concepts by brute aggregation—that is, by first producing a representation in accord with one conceptualization, then conjoining a representation in accord with the other—although sectarians, of course, will deny even that possibility. 32 It is as if we were French purists who had nothing against English but wanted to banish Franglais.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

172

Language, World, and Limits

indulgence that Quine says we should show to an empirically warranted rival system of the world, and that I have agreed we should show, very reminiscent of the indulgence that Williams says we can show to judgements involving thick ethical concepts that we do not ourselves share?33 How then can we say that neither of the two rival scientific systems is from a point of view? Admittedly, there is one obvious and important difference between the scientific case and the ethical case, reflected in Quine’s lax sectarianism. On Quine’s view, as we have seen, we are free to shift our allegiance back and forth between the two scientific systems. Indeed he cites a possible benefit in our doing so (although, disconcertingly for my purposes, he describes the benefit as ‘an enriched perspective on nature’34). The ethical analogue is much harder to envisage. Oscillations between social worlds may be possible, either for individuals or, very differently, for groups. They may occur as a result of a kind of restlessness, or a kind of unconfidence, or even a kind of ‘ethical experimentation’.35 But this sort of thing is necessarily more awkward, more disorderly, and altogether more demanding than its scientific counterpart, as well as having much less clearly defined criteria of success. I agree with Williams when he calls it a ‘wild exaggeration’ to assimilate adopting a scientific system to living in a social world. What makes two social worlds incompatible is far more radical than what makes two scientific conceptualizations incompatible, even when each world is, in Williams’ terms, a ‘real option’ for some group of people.36 But the problem remains. ‘More radical’, ‘harder to envisage’, ‘more demanding’: these all indicate differences of degree. But what is required is a difference of kind. We need some independent handle on the idea that social worlds do, and scientific conceptualizations do not, furnish different points of view.

3. What the argument for the possibility of the absolute conception requires It seems to me that the best handle on this is given by the very argument for the possibility of the absolute conception. That is, I think we best understand the content of Williams’ conclusion, and of the intended contrast between science and ethics that goes with it, if we look at them in the context of the argument that he gives for that conclusion.37 Understanding the argument in turn, of course, requires understanding its premise, the underlying realism. And I have already indicated that one of my aims in this essay is to see what we can learn about the premise from the conclusion. Am I therefore involved in a vicious circle? In a circle, yes; in a vicious circle, I think not.

33 Williams (1985), pp. 140 ff. (Note that Williams’ indulgence, unlike Quine’s, is ecumenical. In suitably favourable circumstances, Williams thinks, we can regard a judgement involving an alien thick ethical concept as true.) 34 Quine (1986a), p. 157. (I see no reason, incidentally, to think that the possibility of shifting our allegiance in this way detracts from the importance of keeping each system at bay while trying to maintain our allegiance to the other.) 35 Williams (1985), p. 157. 36 Ibid. pp. 160 ff. See also, in greater detail, Williams (1981b). 37 Cf. Moore (1997), pp. 82–83.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

173

What Williams is presenting us with, it seems to me, is a package of ideas that need to be understood together. This package is roughly as follows. All knowledge answers ultimately to a unified, substantial, autonomous reality which can, in principle, be conceived as such. To conceive it as such is to form a single conception of it such that, for any item of knowledge, the conception indicates what makes that item of knowledge true;38 more to the point, for any two items of knowledge, the conception indicates what makes both those items of knowledge true, in such a way that it can be used in an account of how they cohere. This means that the conception cannot be from the same point of view as any given item of knowledge. For if it were, it would not be able to indicate, with the detachment necessary to be used in this way, what makes both that item of knowledge and an item of knowledge from an incompatible point of view true. So the conception cannot itself be from any point of view. Science is able to provide this conception.39 We can now see why social worlds are thought to furnish different points of view in a way in which scientific conceptualizations do not. The idea is this. Given two scientific systems of the world, of the sort considered in the previous section, there is no impediment to using the conceptual resources of one to indicate (non-reductively) what makes the other true; nor to using this indication of what makes the other true in giving an account of how the two systems cohere. By contrast, given two incompatible social worlds, even if (improbably) it is possible to use the thick ethical concepts associated with one to indicate what makes an item of knowledge involving the thick ethical concepts associated with the other true, it is out of the question to use this indication of what makes the second item of knowledge true in giving an account of how the two items of knowledge cohere. To give an account of how the two items of knowledge cohere, and in particular to frame that part of the account that indicates what makes both the items of knowledge true, requires at the very least the sort of detachment from either social world that would be needed to indulge in some suitably reflective history, psychology, sociology, and/or anthropology. (I am not now trying to defend the position, just to clarify it.) This, of course, is highly reminiscent of the idea that initiated this discussion: the idea that the best reflective explanation of our having whatever ethical knowledge we have, unlike the best reflective explanation of our having whatever scientific knowledge we have, cannot directly vindicate it. But the two ideas are not the same. There was no reference in what I just said to explanation. Indicating what makes an item of knowledge true is different from explaining how a given individual or a given 38 By ‘indicates’ here, I do not mean ‘makes reference to’; I mean something more like ‘expresses’. Thus consider someone who knows that the earth orbits the sun. In order to indicate the fact that makes this item of knowledge true, the conception must actually incorporate the claim that the earth orbits the sun— or else a set of claims that entail that the earth orbits the sun. It cannot just incorporate some claim about the fact that makes this item of knowledge true, for instance the claim that the item of knowledge is made true by the fact which Copernicus famously established. For (part of) the significance of this distinction, see further later, esp. n. 56. 39 For amplification, see Moore (1997), esp. Ch. 4, §3. (Why think that science can provide the conception? See n. 11: for me, this is more or less a matter of definition; for Williams, it may be a more substantial matter.)

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

174

Language, World, and Limits

group has come by the knowledge. The former typically falls short of the latter.40 Williams’ focus on explanation introduces something not present in the original argument for the possibility of the absolute conception, something that, at least as far as this current discussion is concerned, is actually both a complication and an irrelevance. But is it perhaps more than that? Does it perhaps stand in some tension with the original argument? The original argument requires that, for any item of knowledge, it should be possible to indicate, from no point of view, what makes that item of knowledge true. The appeal to explanation suggests that there are some items of knowledge, involving thick ethical concepts, such that it is impossible to say anything, from no point of view, that directly vindicates them. Do these not tell against each other?41 I do not think so. The word ‘directly’ is vital. What is precluded, in the case of an item of knowledge involving thick ethical concepts, is exercise, from no point of view, of those very concepts. This is enough to prevent any direct indication, from no point of view, of what is known. But it leaves open the possibility of telling a story, from no point of view, whose consequences, in some non-trivial but suitably relaxed sense of ‘consequences’, include the fact that things are as they are thereby known to be.42 The original argument requires nothing, it seems to me, that is threatened by the appeal to explanation. Note, incidentally, that in order to give a full account of how two items of knowledge from incompatible points of view cohere, it is necessary to go beyond the resources of the absolute conception. The absolute conception can supply part of this account, certainly. In particular, it can indicate what makes the items of knowledge both true; that is of its very essence. But it cannot indicate, on its own, how either of them is made true (how, for instance, the point of view of either contributes to its having whatever content it has). To give a full account of that requires exercise of such concepts as the concept of knowledge and the concept of content, neither of which can be exercised except from some sort of psychosocial point of view. A fortiori the absolute conception cannot, on its own, explain how a given individual or a given group has come by either of the items of knowledge. It cannot, on its own, explain how a given individual or a given group has come by any item of knowledge. In particular, it cannot, on its own, explain how we have come by it. Another common misconstrual of Williams’ position is to think that he does demand 40 But as regards that part of the explanation that concerns how the individual or the group in question has actually acquired the belief—no matter that it constitutes knowledge—the latter typically falls short of the former. It would be setting absurdly high standards to expect a good reflective explanation of how I have come by my belief that water contains hydrogen, for instance, to extend any further back than the various reference books and other authorities that have led me to believe this. (But the best reflective explanation of how I have come by my knowledge that water contains hydrogen—and of how, in particular, it counts as knowledge—would have to extend all the way back to the fact that water contains hydrogen. Some critics of Williams perhaps miss this crucial distinction. See e.g. Quinn (1993), §11, esp. p. 140; and Rorty (1991a), §4, esp. pp. 57–58.) 41 This is in effect the criticism that I levelled against Williams in Moore (1991). I have tried to correct what I say there in Moore (2003): see esp. n. 20. 42 This is an allusion to the notion of ‘weak entailment’, and to the attendant contrast between endorsement and indirect endorsement, that I use in Moore (1997): see pp. 15–16, and cf. also pp. 35–36.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

175

this of the absolute conception; that he takes the absolute conception ‘to explain itself ’.43 Williams himself is partly to blame for this misconstrual, because of some occasional incautious formulations of a demand that he does make. At one point, for instance, he says that ‘the substance of the absolute conception . . . lies in the idea that it could nonvacuously explain how it itself . . . [is] possible’.44 For the most part, however, he is careful to insist that all the absolute conception need do is to include part of a good reflective explanation of how we have come by it: it must ‘help to explain . . . our capacity to grasp [it]’.45 The reason for this more circumspect demand is precisely the fact that a good reflective explanation of how we have come by any given knowledge must be from some sort of psychosocial point of view. Williams is well aware of this, despite the impression that his critics sometimes give.46 There is of course the further question, which part of the explanation the absolute conception must include. The answer, I take it, is that insofar as our grasp of the absolute conception involves our being sensitive to the fact that things are a certain way, it must include that part of the explanation which says that things are indeed that way— which it will trivially do. As far as the rest of the explanation is concerned, the most that we can demand of the absolute conception—although we can indeed demand this—is that it should stand in some intelligible relation, presumably some relation of consequence of the sort mentioned earlier, to the psychological and sociological elements in it that explain our actual sensitivity to how things are.47 However that may be, I repeat that the best handle on Williams’ notion of the absolute conception is given by the very argument for its possibility. What the absolute conception is is that which the argument for the possibility of the absolute conception is an argument for the possibility of. It is a conception, in other words, that can indicate what makes any given items of knowledge true, in such a way as to form part of an account of how they are made true, even when they are from incompatible points of view; a conception fit to sustain our sense of what is known as what is there anyway. And our sense of what is known as what is there anyway is that which the absolute conception is a conception fit to sustain. I do not pretend that such interdependencies are to be regarded with complete nonchalance. Close scrutiny of Williams’ argument, and of everything he says in connection with it, is required to ensure that these particular interdependencies constitute a useful and robust structure. But I am certain they do. My point here is simply that the structure has no right way up.

4. Realism and its variants The structure is a conceptual structure. To make use of it is to operate with certain concepts, most notably a certain concept of knowledge and a certain concept of real43 Cf. Moore (1997), p. 65. Culprits are Putnam, in Putnam (1992), p. 98; and Quinn, in Quinn (1993), p. 136. 44 Williams (1985), p. 139. 45 Ibid. p. 140, my emphasis. 46 See e.g. Williams (1978), pp. 301–303 and Williams (1985), p. 140. One critic who suggests that he is not aware of this is Putnam: see Putnam (1992), p. 100. 47 See further Williams (1978), pp. 245–246.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

176

Language, World, and Limits

ity. Someone might elect not to operate with these concepts. This would involve, among other things, rejecting the fundamental propositions that hold the structure together. In particular, it would mean rejecting the realism. That is certainly something that someone might do. However, although the realism can be rejected in this way, there is an important sense in which it cannot be denied. To deny it would be, not to repudiate the concepts in whose terms it is couched, but rather (on the contrary) to appropriate those concepts and to repudiate the realism itself—to count the realism as false. And that is not something that anyone who fully understands the realism can do. To appropriate those concepts is, among other things, to acknowledge the truth of the realism. If someone appears to deny that knowledge is of a reality that exists independently of being known, then this only goes to show that he or she is not really—not properly—operating with the concepts of knowledge and reality with which Williams is operating. We can compare this with the case of someone who appears to deny a principle of Euclidean geometry, say the principle that between any two points there is at most one straight line. This only goes to show that he or she is not really—not properly—operating with Euclidean concepts. To be sure, there may be good reason for not doing so. There may be good reason for operating with variant concepts instead. Thus, as is well known, there are non-Euclidean geometries that entitle us to say, ‘There can be two points between which there is more than one straight line.’ Moreover, it is to these variants of Euclidean geometry that we should turn, rather than to Euclidean geometry itself, in order to find the best tools for describing physical space. I am not ruling out the possibility that there is some similar variant of Williams’ conceptual structure that provides the best tools for undertaking some metaphysical task. But if this variant entitles us to say, ‘Knowledge is of a reality that depends on being known’, then it cannot involve exactly the same concepts of knowledge and reality as those which Williams is using.48 What alternatives are there? (There are alternatives that involve repudiating any concepts that so much as resemble these of course. Some particularly radical philosophers do advocate such alternatives. But I mean: what alternatives are there that are sufficiently close to Williams’ structure to count as variants of it, in the way in which non-Euclidean geometries count as variants of Euclidean geometry?) The most familiar, perhaps, are various species of idealism, in which the term ‘reality’ stands for something that does depend for its existence, or for some of its essential characteristics, on being known. But there are also non-idealistic variants in which what is overturned is not the knowledge-independence of that which is termed ‘reality’ but its unity. To see the sort of thing that these latter variants involve, we need to look more closely at the rôle that unity plays in Williams’ realism. And it is here especially that the conclusion of his argument—that the absolute conception is possible— can help us. 48 This is related to what I mean in Moore (1997) by my repeated insistence that Williams’ realism, encapsulated in what I there call the Basic Assumption, cannot be justified (e.g. pp. 107, 109, and 188–189). It cannot be justified because there is no issue about whether it is true. See further Moore (1999). (Note: when I suggested in n. 4 that Williams’ realism is no more than the intuitive deliverance of reflective common sense, I did not mean that such variants are impossible. My point was rather that the concepts that Williams is using are concepts that we habitually and naturally use; and that, granted these concepts, Williams’ realism emerges as a basic conceptual truth.)

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

177

What the conclusion reveals is the following fundamental principle at the heart of Williams’ realism. Given any two items of knowledge, or more generally, given any two true representations of reality, it is possible to indicate what makes them both true.49 This is implicit in the following extract from the original argument: [If A and B each has knowledge, and their representations differ,] then it seems to follow that there must be some coherent way of understanding why these representations differ, and how they are related to one another. . . . [A] story can be told which explains how A’s and B’s can each be perspectives on the same reality.50

Williams does elsewhere insist on the ‘non-additive’ nature of knowledge. He writes, ‘Not all pieces of knowledge can be combined into a larger body of knowledge.’51 But this relates to what I said in the previous section about directness. Again I see no conflict. What his argument and its conclusion require is the possibility of telling some story that gives an indirect indication of what A and B between them know. What the ‘non-additive’ nature of knowledge prevents, or may prevent, is directly conjoining what A knows with what B knows, without dilation or manipulation of any kind.52 The basic idea, then, is that all knowledge ultimately answers to reality—the same reality in every case—and this is an idea that means nothing unless it means that, for any given items of knowledge, there is some single way of indicating, however indirectly, how reality is thereby known to be. What are the alternatives to this? Here is one. Where we have . . . [two true representations] that conflict . . . , their realms are . . . less aptly regarded as within one world than as two different worlds, and even—since the two refuse to unite peaceably—as worlds in conflict.

This is Goodman, summarizing the chief motif of his iconoclastic work Ways of Worldmaking.53 As we can see, he is using conceptual apparatus that differs from that of Williams in precisely the radical way envisaged. Not that Goodman would mind admitting that the resultant apparent disagreement between him and Williams is essentially terminological. On the contrary, to admit this would itself be to signal a striking example of the very pluralism that he is advocating. Earlier in the same book he writes: While I stress the multiplicity of world-versions, I by no means insist that there are many worlds—or indeed any; for . . . the question whether two versions are of the same world has as many good answers as there are good interpretations of the words ‘versions of the same world.’ The monist can always contend that two versions need only be right to be accounted versions of the same world. The pluralist can always reply by asking what the world is like apart from all versions.54 49 This is basically what I call the Fundamental Principle in Moore (1997): see pp. 21–22. 50 Williams (1978), p. 64. 51 Williams (1985), pp. 148–149. 52 See again n. 42. The kind of dilation or manipulation that I have in mind is illustrated in the case in which A knows that it is humid and B, in the very same place six months later, knows that it is snowing: in order to indicate what A and B between them know (which is not, of course, that it is both humid and snowing) we make explicit reference to the dates and times concerned. 53 Goodman (1978), p. 116. 54 Ibid. p. 96. (Actually, the pluralist had better do better than that. The monist can just as easily acknowledge the force of that rhetorical question.)

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

178

Language, World, and Limits

But if we do adopt the pluralism that Goodman advocates, then we shall not say, with Williams, that for any two items of knowledge, there is some single way of indicating what is thereby known. We shall say instead, with Goodman, that some items of knowledge answer to different worlds, worlds that cannot be combined into one, and that, in such a case, there is no single way of indicating what is thereby known.55 To understand better what is at issue here, let us look at two applications of these ideas. The first is to the question whether tense is real. On one way of construing that question, it is, precisely, the question which of these alternatives is the right one for tensed knowledge (knowledge from a temporal point of view). Those who claim that tense is real, on this construal of the question, adopt a Goodmanian pluralism. Their position is this. Given some tensed item of knowledge, there is no indicating what makes it true except from the same temporal point of view. This is because what makes it true contains something directly corresponding to its tense that does not obtain from any other temporal point of view. Thus, suppose I know that it is humid today. Then what makes my knowledge true is the fact that it is humid today. But this is a fact that I can indicate only today. If I say tomorrow, ‘It was humid yesterday’, that will not indicate the same fact. At best it will indicate some intrinsically related fact, a fact about (as it were) hesternal humidity which can itself be indicated only tomorrow.56 Reality fractures into different temporal worlds, then. Each temporal point of view carries its own world with it. The facts that peculiarly constitute one of these worlds can be indicated only from the corresponding temporal point of view. It immediately follows that there are some items of knowledge, namely items of knowledge from different temporal points of view, for which there is no single way of indicating how reality is thereby known to be—a consequence that is, of course, in direct violation of Williams’ realism.57 The second application of these ideas—an application more relevant to Williams’ concerns—is to the analogous question about social reality. The analogue of the view that tense is real is the view that reality fractures into different social worlds, and that the facts that peculiarly constitute one of these social worlds can be indicated only from the corresponding social point of view. Williams himself uses the expression ‘social worlds’, as I did in echo of him when outlining his view in section 1.58 And sometimes he makes claims that suggest that he adopts precisely this Goodmanian view about them. He writes: [An observer] cannot stand quite outside the evaluative interests of [a] community he is observing, and pick up [one of their thick ethical concepts] simply as a device for dividing up in a rather strange way certain neutral features of the world. . . . [Their ethical knowledge is] part of their way of living, a cultural artifact that they have come to inhabit.59

But of course, Williams does not adopt Goodmanian pluralism about social worlds. This quotation is another illustration of the distinction that arises within his realism between, on the one hand, directly expressing what members of the community 55 Cf. ibid. p. 115. 56 Is this not self-stultifying? Is not my very reference to this fact that I say can be indicated only tomorrow an indication of it today? No. Cf. earlier, n. 38: when I talk about ‘indicating’ a fact, I mean something like ‘expressing’ it, not making reference to it. 57 See Essay 11 in this volume for a fuller account of the material in this paragraph. 58 See e.g. Williams (1985), p. 150. 59 Ibid. pp. 142 and 147.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

179

know—that is, what they know through exercise of their thick ethical concepts—and, on the other hand, telling a story that indicates indirectly what they know. The observer cannot do the former because, as someone who lives outside their social world and who does not share the values that sustain exercise of their thick ethical concepts, he cannot himself make use of those concepts. Nor, crucially, does he have the wherewithal to construct neutral equivalents of those concepts: that is the point that Williams is making in the quotation.60 By contrast, he may be able to do the latter. He may be able to understand enough about the community, about their social world, and about its history to be able to see how their use of these concepts enables them to live in that world, and he may be able to say, in the light of that, how their circumstances warrant the exercise of the knowledgeable judgements they make using these concepts. Admittedly, if he does succeed in doing this, he might still not have carved out the same chunk of logical space as they do in making any of the relevant judgements. But carving out the same chunk of logical space must not be confused with carving out a chunk of the same logical space. He will have done the second of these, which is all that Williams requires. In particular, he will have said enough to entail what they know, in the relaxed sense of entailment that I alluded to in the previous section. If the choice between Williams’ realism and Goodman’s pluralism is essentially terminological, then can it matter which we adopt? It can. The word ‘terminological’ has pejorative overtones. But an issue can be both terminological and of considerable moment. It can be, in effect, an issue about which of two conceptual structures is better equipped to meet certain theoretical and/or practical needs. (Think again about the choice between Euclidean geometry and its non-Euclidean variants.) Very well, then; what is there to be said in favour of either of these alternatives? There are different things to be said in different contexts, but one thing that I think must be said quite generally in favour of Williams’ realism is that it is our natural starting point. Goodman’s pluralism is an unintuitive departure from the way we normally think and speak.61 Still, the way we normally think and speak is not sacrosanct. This advantage can be outweighed.62 Let us turn back to the question whether tense is real. Here the Goodmanian alternative might be thought to have an advantage in relation to the idea that the future is open. It is far beyond the scope of this essay to say fully what I have in mind. But here is a sketch. Suppose, what many people intuitively think, that the future is open. That is, roughly, suppose that nothing is the case at any given time about what is contingently the case at later times. And suppose I said yesterday, ‘It will be humid tomorrow.’ Then not only did this fail to indicate the fact that verifies my knowledge that it is humid today. It was not even true. For it was not (then) the case that it would be humid today. By far the most natural way to capture this idea, and arguably the only way to do so, is by appeal to different temporal worlds; more

60 See further ibid. pp. 141–142, and the references therein. See also McDowell (1981), pp. 144–145. 61 Cf. the parenthetical remark in n. 48. 62 Cf. in this connection Wittgenstein (1967), Pt. I, §402.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

180

Language, World, and Limits

specifically, by appeal to a sequence of temporal worlds such that those later in the sequence contain details corresponding to gaps in those earlier in the sequence.63 But even if Goodman’s pluralism has the whip hand in this context (which actually I doubt), there are other contexts in which Williams’ realism retains all sorts of advantages over it, and indeed over any of its rivals. In particular, it retains significant advantages in the context of moral philosophy. Paramount among these is something that will serve to bring my essay full circle: namely, the way in which Williams’ realism, by sustaining the argument for the possibility of the absolute conception, creates an ideal framework for expressing his opposition to realism about ethics. It is important here to be clear just what the relation between his realism tout court and his opposition to realism about ethics is. The latter does not of course follow from the former (in the way that the belief that the absolute conception is possible does). Rather, the former is part of a conceptual structure that facilitates expression of the latter. If we repudiated Williams’ realism in favour of Goodman’s pluralism, or even in favour of some species of idealism, this would not undermine any of his arguments against ethical realism; but it would make those arguments harder, if not perhaps impossible, to state. Williams’ realism gives us, I think, a richer understanding than its variants do of the nature of our ethical experience. By enabling us to see that our ethical knowledge is from a point of view that admits of equally legitimate and incompatible alternatives, in a way in which our scientific knowledge is not—and by enabling us to see how history, psychology, sociology, and/or anthropology are needed to explain why we have the ethical point of view we have, where this is not itself a matter of our knowing anything—Williams’ realism gives us a much firmer grip on what people are getting at when they make clumsy appeal to the fact/value distinction. Up to a point, their clumsy appeal is apposite. As Williams observed long ago, in the factual case, there is a possible thought. . . . ‘I am convinced that p, but it is possible for all that that not-p,’ . . . [which] registers the impersonal consideration that how things are is independent of my belief; however they are, they are, whatever I believe. . . . But . . . [there is] no parallel thought possible on the moral side: . . . there just is no content to ‘I am convinced that racial discrimination is intrinsically wrong, but it is possible for all that, that it is not,’ except things like ‘How convinced am I?’ or ‘I suppose somebody might make me change my mind.’64

However, as Williams’ critique of thick ethical concepts clearly shows, there is more to it than that. The concept of being intrinsically wrong is not itself a thick ethical concept. Its applicability is not ‘world-guided’ in the way in which, say, the concept of being a racist is. My conviction that racial discrimination is intrinsically wrong is not an item of knowledge. But—and this is the point—it does enable me to know such things as that Wagner was a racist. The clumsy appeal to the fact/value distinction obscures this. Williams’ more layered view makes it very clear. It also makes clear what kind of thing I need, or more generally what kind of thing we need, if we are to 63 For a thorough discussion of the issues and complications that arise here, a discussion that attempts in a quite different way to capture the belief that the future is open, see Tooley (1997). (Note: whether we are to regard the future as open or not may itself be a terminological issue of the very kind I have tried to identify. See further Cockburn (1997), esp. Ch. 9.) 64 Williams (1972), p. 49.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Realism and the Absolute Conception

181

maintain our point of view and continue to have such knowledge, from that point of view. We need confidence.65 Not that Williams’ realism itself gives us confidence. On the contrary, it contributes significantly to undermining our confidence—not least, by making us aware of our need of it. But that is the predicament that we must learn to face if we are to live in light of the truth, something we have every reason to do. It would be a serious mistake to think that we would be better off if we had never reflected in this way on our ethical experience; if we had never thereby known what it is to be, in a phrase borrowed from Wallace Stevens and quoted by Williams at the beginning of Ethics and the Limits of Philosophy,66 ‘shaken realists’.67

65 Williams (1985), pp. 170–171 and Williams (1995c), pp. 205–210. For further discussion, see Fricker (2001). 66 This phrase is taken from Stevens’ poem ‘Esthétique du Mal’ in Stevens (1954); Williams uses it as an epigraph for Williams (1985): see p. x. and see further Williams (1985), pp. 167 ff. 67 I am very grateful to Anita Avramides for many helpful discussions on these matters.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

14 One World Abstract This essay first appeared as a contribution to a special issue of European Journal of  Philosophy to mark the fiftieth anniversary of the publication of P.F. Strawson’s The Bounds of Sense. In that book Strawson asks whether we should agree with Kant’s claim, in his Critique of Pure Reason, that there can be only one world. What Kant means by this claim is that the four-dimensional realm that we inhabit must constitute the whole of empirical reality. Strawson gives reasons for challenging this claim. This essay raises the question whether, even if Strawson is right, we may nevertheless have reason to believe that there can be only one world on a broader understanding of ‘world’. The aim is as much to clarify the issue as to settle it, although an attempt is made to motivate the view that there can indeed be only one world on this broader understanding.

My starting point in this essay is a brief but fascinating section in The Bounds of Sense,1 in which Strawson discusses a likewise brief but fascinating passage in Kant’s Critique of Pure Reason.2 I am referring respectively to Part Two, Chapter III, §8 (which incidentally contains Strawson’s sole reference to Wittgenstein in his entire book—we shall see the significance of this reference in due course) and A216/B263. In the latter Kant writes that ‘all appearances lie in one nature, and must lie therein’. In the former Strawson asks why we should accept any such conclusion. Why, to appropriate the title of that section, only one objective world? It is important to be clear what Strawson, following Kant, is asking. Consider the fourdimensional realm that we inhabit. Call this ‘the Cosmos’. Then Strawson’s question, in effect, is why we should agree that the Cosmos constitutes the whole of empirical reality. The word ‘empirical’ is important here. It connects with Kant’s explicit reference to ‘appearances’.3 Kant himself allows for the possibility that the Cosmos does not exhaust reality in a broader sense of ‘reality’ (a point to which we shall return). But the Cosmos does, Kant insists, exhaust empirical reality. And Strawson asks why we should agree. It is a good question. Who knows but that physicists will one day convince us that the Cosmos does not exhaust empirical reality?4 Kant seems to be involved in an error akin to that of insisting that space must be Euclidean. In this case, as in that, it seems 1 Strawson (1966). All unaccompanied references in this essay to work by Strawson will be to this book. 2 Kant (1998). All unaccompanied references in this essay to work by Kant will be to this book. 3 Cf. A418–419/B446, note. 4 Consider e.g. some variation on the ‘many worlds’ interpretation of quantum mechanics or certain versions of string theory.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

One World

183

that there can be empirical evidence for a possibility that he disavows. Strawson considers different forms that such evidence might take. He does not appeal to anything that appears, or might be expected to appear, in physics. Rather he appeals to various relatively homespun thought experiments.5 In one of these there is a community whose members can successively occupy each of two spatially unrelated universes.6 Even if Kant is wrong to insist that the Cosmos constitutes the whole of empirical reality, however, an analogous question arises with respect to whatever does. Indeed— harking back to the point that I made in connection with Kant’s use of the term ‘appearances’—an analogous question arises with respect to whatever constitutes reality in a yet broader sense of ‘reality’ that may embrace the non-empirical or the supernatural. Is there any reason to subscribe to the unity of that, in some correspondingly extended sense of ‘unity’, or to believe that it consists of only one world, in some correspondingly extended sense of ‘world’? This question, though different from Strawson’s, is an extension of his. And it is on this broader question that I wish to focus. Part of my interest in the question lies in the fact that I have elsewhere nailed my own colours to the ‘unity’ mast. That is, I have elsewhere subscribed to the unity of reality, in the relevantly extended sense.7 And although I have also insisted on the unjustifiability of this stance,8 this has been according to high standards of justification. I have not denied that the stance has anything to be said in its favour. I hope that this essay will show that it has. But what is the issue? What would it be for there to be more than one world, in the relevantly extended sense? Such worlds would have to lack even whatever intermundane unity binds together the two universes that Strawson envisages; and it is difficult to see what could then secure for worlds other than our own the status of the actual as opposed to the merely possible. One way to make this difficulty graphic is to consider David Lewis’ modal realism. Lewis famously believes in the existence of infinitely many possible worlds, each a spatio-temporally unified universe that is both isolated from and causally independent of all the others. He also holds that for something to be possible is for it to be the case in one of these universes. It immediately follows that, for Lewis, it is not possible that two or more of these universes be coactualized. Lewis acknowledges this and says, with commendable candour, that he ‘would rather not’—though he takes it to be a price worth paying for a gain in theoretical simplicity.9 The question is: what exactly would he rather do instead? What exactly would constitute the co-actualization of two or more of these universes?

5 See also, in addition to the section already cited, pp. 142–143. 6 This incidentally precludes our characterizing the Cosmos à la David Lewis as all and only that which exists ‘at some distance and direction from here, or at some time before or after or simultaneous with now’ (Lewis (1986), p. 1). If there were spatially unrelated but temporally related universes, as in Strawson’s thought experiment, then, given the disjunctive nature of this characterization, it would capture all of them, not just the four-dimensional realm that we inhabit, or at any rate that we now inhabit. (Not that this is any indictment of Lewis. His characterization is not intended to capture what I mean by ‘the Cosmos’. It is intended to capture what he means by ‘the actual world’: see further below.) 7 See e.g. Moore (1997), Ch. 4, §4. The unity of reality is part of what I there call the Basic Assumption. 8 See Moore (1997), esp. pp. 113 and 188–189. See also Essay 11 in this volume, esp. p.157. 9 Lewis (1986), pp. 71–72.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

184

Language, World, and Limits

Had Lewis allowed for possible worlds other than our own to be co-actual with it, what exactly would it have taken for any of them to satisfy this condition?10 The very intractability of such questions is, I think, a reason to accept the unity of reality. Roughly: if reality is not unified, then it is obscure what makes it reality that is not unified, as opposed to reality plus items of unrealized possibility. The very unity of reality is part of what constitutes it as reality. To be real is to stand in some suitable relation of unification to this (or to us, or to here, or to now).11 Or so it seems to me. But even if I am right, there is of course more to be said. In particular, there is more to be said about this relation of unification—call it ‘the Unity Relation’.12 We can take for granted, as a minimum, that the Unity Relation is both reflexive and symmetric,13 and also that for anything in reality to be unified in the relevant sense (including, at the limit, for reality as a whole to be unified in the relevant sense) is for all of its parts to stand in the Unity Relation to one another. But until we have some further account of the Unity Relation, this is purely schematic and there is no telling why some other, weaker relation—for instance, and most notably, the ancestral of the Unity Relation—might not be robust enough to reach out from here to all and only what is real. Before we go any further, we need to settle on some further terminology. In particular, we need terms for: (i) the sum total of that which is co-actual with us and (ii) any maximally unified whole where by a ‘maximally’ unified whole I mean a unified whole that is not part of some bigger unified whole. The need to settle on such terminology is relatively urgent, because there are several crucial terms in this vicinity—‘nature’, ‘universe’, ‘world’, ‘reality’, et al—that not only get used (with suitable modification) in both the (i)-rôle and the (ii)-rôle, but that also get used in further incompatible ways, and the scope for confusion is vast. So far, I myself have played a little fast and loose with these terms, to allow myself latitude in characterizing the various different positions that I am interested in. Inasmuch as my usage has been systematic, then I have reserved the term ‘reality’ for the (i)-rôle and the term ‘world’ for the (ii)-rôle. Thus I have presented the principal issue as an issue about whether ‘reality’ is unified, or whether, on the contrary, it consists of more than one ‘world’. This means that my use of ‘reality’ has been broad. Even so, it has not been as broad as it might have been. For instance, Lewis’ modal realism is sometimes expressed as the thesis that ‘reality’ comprises infinitely many possible worlds in addition to the one that is actual: ‘reality’, on this 10 Kant raises similar questions at A230/B282 ff., partly with a view to dismissing them as bad questions. This is essentially what I want to do too, though for reasons that are more radical than Kant’s (albeit akin to his). My reasons are more radical than Kant’s because his criteria for being a good question are tied, once again, to the empirical. 11 Cf. the material cited in the previous note, esp. A231/B284 and A234/B286. 12 Kant gives an account of a related relation at A418–419/B446, note, as ‘thoroughgoing connection through an inner principle of causality.’ But again (cf. n. 10) this extends only to empirical reality. 13 But not transitive? Not transitive. The reasons for this will emerge in due course.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

One World

185

usage, embraces everything there is, even what is merely possible.14 To repeat: the scope for confusion is vast. I propose that we henceforth never use any of these terms in the broad way that encompasses even the merely possible.15 I further propose that we allow each of them to be used in either the (i)-rôle or the (ii)-rôle, registering the distinction by writing them in small capitals when they are used in the (i)-rôle and in ordinary lower case type when they are used in the (ii)-rôle. In these terms, then, the question is: can there be more than one world in the world, more than one reality constituting reality, more than one actual realm within the actual realm? Equivalently: can there be more to the world than the world?16 Now I have already characterized such questions as intractable. But I do not mean to imply that no-one has proffered interesting answers to them. John Leslie has. He proposes, as part of an elaborate theodicy, that something is actual if and only if God contemplates it; and this is expressly meant to accommodate universes other than our own.17 Let us not worry about his reasons for this proposal. The question is: can we really make sense of it? Michael Dummett, in a different context, suggests that at any rate we cannot make sense of a universe’s being actual if it contains no sentient creatures. And he does so in such a way as to suggest that God’s contemplating such a universe would precisely fail to compensate for its lack of sentience; would precisely fail, in other words, to elevate it from the status of the merely possible. ‘What would be the difference,’ Dummett asks, ‘between God’s creating such a universe and his merely conceiving of such a universe without bringing it into existence? It seems to me that the existence of a universe from which sentience was perpetually absent is an unintelligible fantasy’.18 He continues: ‘What exists is what can be known to exist. What is true is what can be known to be true.’ Dummett is tapping a deep intuition here: that whatever is actual must be within epistemic reach of some actual finite being; that finite knowledge is, to that extent, capable of playing a fundamental metaphysical rôle that infinite knowledge (or infinite contemplation, or infinite conception) cannot play.19 This has a significance to which we shall return. 14 Cf. Lewis (1986), pp. 100–101, n. 1. Cf. also Moore (2012), p. 414, where I contrast such usage, along with Lewis’ own use of ‘actual’, with that which is found in Bergson. 15 So I shall not allow the fact that we can entertain the possibility of flying pigs to warrant saying that there are, in reality, flying pigs. Note, however, that some maximally unified wholes are merely possible (this is a quite separate matter—it concerns their own status, not what they encompass) and I shall sometimes use terms in a way that reflects this. 16 By ‘the’ world I mean whichever world includes the Cosmos. I should acknowledge, however, that it is not entirely obvious that there is such a thing. For it is not entirely obvious that there are any worlds at all; that there are any maximally unified wholes. This raises some exceedingly important and exceedingly difficult metaphysical issues concerning what counts as a whole, issues that I shall ignore in this essay. I shall assume that the world exists. (For the record, I shall also assume, as I have been doing up to now, that the world exists. That too is not entirely obvious. For it is not entirely obvious that there is anything that deserves to be called ‘the sum total’ of what is co-actual with us. Note that, if the world exists but the world does not, say because no worlds do, then not only is there not more to the world than the world, there is not even that much.) 17 Leslie (2001), esp. Ch. 1. 18 Dummett (2004), p. 92. 19 I say that the intuition is deep; I do not say that it is unassailable. I have argued elsewhere that it may involve an unacceptable idealism: see Moore (2012), esp. Ch. 14, §4. A further issue is that, if there cannot be a universe devoid of sentient creatures, then it is unclear in what sense such a universe can so much as

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

186

Language, World, and Limits

In the current context, however, the intuition, however deep, and however defensible,20 helps at most by yielding a non-trivial necessary condition for something’s being actual. There is no corresponding non-trivial sufficient condition. Although it is true that whatever is within epistemic reach of some actual finite being is actual, it is trivially true. It merely shifts our attention to what it is for a finite being to be actual, as opposed to merely possible, in the first place. And if we now ask whether Leslie is right that being contemplated by God is enough to make the difference, then it once again seems to me that we are asking something intractable. What difference? To further motivate my scepticism about such questions, I need to say more about the Unity Relation. Here it is helpful to consider matters as it were from below rather than from above. So far I have been taking for granted the unity of the Cosmos. But this is not beyond dispute. Consider the following view. Tensism: Suppose that t is a true tensed thought about the Cosmos. That is, suppose that t is a true thought about the Cosmos from some temporal point of view π. Then only from π is it possible to express the fact that makes t true. This is because the fact that makes t true ipso facto contains something that corresponds to t’s tense, something that is itself accessible only from π. Thus suppose I think, truly, that it is humid today. Then what makes my thought true is a fact involving ‘hodiernal’ humidity. If I say tomorrow, ‘It was humid yesterday,’ that will not express the same fact. At best it will express some intrinsically related fact involving ‘hesternal’ humidity, which can itself be expressed only tomorrow. Each temporal point of view is accompanied by its own world, and the facts that peculiarly constitute one of these worlds are expressible only from the corresponding point of view.21 It follows that the Cosmos is not itself a world. It is a multiplicity of worlds—more specifically, temporal worlds.22 Now tensism seems to me no less problematical than Leslie’s view. But I mention it because I think it helps to give us a grip on the Unity Relation. First, it suggests what the relata of the relation are, namely facts.23 Second, it suggests the following principle. be conceived (even by God). A third issue is that we appear to be able to show, by a modification of a famous argument due to Fitch, that if whatever is actual is within epistemic reach of some actual finite being, then whatever is actual is in fact epistemically reached by some actual finite being, which is as much as to say that whatever is true is known to be true by some actual finite being, which is (surely) not the case: see Fitch (1963). 20 See again the previous note. 21 The word ‘peculiarly’ in this sentence is important. Some facts e.g. the fact that e = mc2, may constitute more than one world. This in turn explains why we cannot take for granted that the Unity Relation is transitive (see n. 13). 22 For further discussion of tensism see Essay 11 in this volume. See also Essay 13 in this volume, esp. section 4. In the latter I consider other views that entail the disunity of the Cosmos: the view that there is a multiplicity of ‘world-versions’, as defended by Nelson Goodman in his (1975); and the view that there is a multiplicity of social worlds, a view that can be extracted from, though it is not endorsed in, the work of Bernard Williams (see e.g. Williams (1985), esp. Chs. 8 and 9, though note also Ch. 4, n. 19). Cf. also the view that Carol Rovane dubs ‘multimundialism’ in her (2009). 23 This means that worlds have facts as their parts. Cf. Wittgenstein (1961), 1–1.1.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

One World

187

The Unity Principle: Two facts, one of which is expressible from some point of view, stand in the Unity Relation to each other if and only if they are both expressible from that point of view.24, 25 For such co-expressibility is precisely what may be lacking if tensism is correct. If tensism is correct, then the fact that it is humid today cannot be expressed tomorrow, though facts about how things are then can be. Not that the Unity Principle can serve as an analysis of the Unity Relation. For one thing, it has no application in lifeless worlds where there is no question of any fact’s being expressed from any point of view (assuming, contra Dummett, that there are such worlds). More significantly, it is a substantive principle. And it is a substantive principle for which I lack any compelling argument. It nevertheless has some plausibility, and I shall adopt it as a working assumption in what follows. I shall also continue to assume, in opposition to tensism and other such views, that the Cosmos is unified. Now one immediate corollary of the Unity Principle is that, for any fact f1 about the Cosmos, and for any point of view π from which f1 is expressible,26 any other fact f2 stands in the Unity Relation to f1 if and only if f2 is likewise expressible from π.27 This seems to me to put additional pressure on the idea that reality is not unified. For it means that, if reality is not unified, then—given that the Cosmos is unified—there are facts that are inexpressible from ‘here’, where ‘here’ is just any point of view from which some fact about the Cosmos (say, the fact that Jupiter is larger than Saturn) is expressible. And that, it seems to me, is a deeply problematical idea.28 24 This is related to what I have elsewhere called the Fundamental Principle: see Moore (1997), pp. 21–22. Cf. also Moore (1997), pp. 150–151, where I advert to the idea (which I believe can be found in the early Wittgenstein) that the world’s unity, or perhaps the world’s unity, is the possibility of its being represented from a single point of view. This bears on what I shall say later. 25 N.B. Expression needs to be understood in a comparatively weak way if the Unity Principle is not to fall foul of facts which, for reasons having nothing to do with the unity of reality, can be precisely captured only from one particular point of view. To express a fact f, let us say, is to produce a linguistic item whose content f0 contains f in a sense of containment weak enough to allow for f ’s supervenience on f0: cf. Moore (1997), pp. 15–16, and Essay 11, n. 31, in this volume. Expressibility likewise needs to be understood in a comparatively weak way. This is so that the Unity Principle does not fall foul of points of view whose occupants, as things stand, and again for reasons having nothing to do with the unity of reality, lack the sophistication to produce linguistic items of the relevant sort. For discussion that bears on what is required here see Moore (1997), Ch. 5, §1, esp. pp. 81–82. 26 If reality is not unified, then π can just be the world. If reality is unified, however, then the world exhausts reality and will not count as a point of view: see Moore (1997), p. 14. 27 I said earlier that Leslie’s view accommodates universes other than our own. We now have a better grip on why. None of the facts that peculiarly constitute any of those universes is expressible from π. This is because it is impossible, from π, to determine which part of reality any purported expression of any such fact is answerable to. This in turn is because it is impossible, from π, to single out any of those universes. And this is for reasons that Strawson himself famously explores in Strawson (1959), Pt. I. (I am assuming that it is impossible, from π, to single out any of those universes in a purely qualitative way—e.g. as the densest, or as the seventh most beautiful.) Note incidentally that, even if it had been possible to express such facts from π, this would not have been, as we might naturally suppose that it would, a threat to Dummett’s intuition. The reason why we might naturally suppose that it would have been a threat to Dummett’s intuition is that we might naturally think that some of the facts in question could have been about lifeless universes. In fact, however, their very expressibility from π means that they could at most have been about lifeless parts of our universe—which is of course not lifeless. 28 Cf. Essay 16, sections 1 and 2, in this volume.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

188

Language, World, and Limits

This all takes on additional significance when the point of view in question is mine. For whatever can be expressed from my point of view can now be seen to constitute a world: ‘my’ world; ‘the’ world.29 It is irresistible to relate this to ideas that we find in Kant’s ‘Transcendental Deduction’.30 It is also irresistible to relate it to ideas that we find in Wittgenstein’s Tractatus.31 Similar significance would accrue from taking the point of view in question to be ours—for various more or less inclusive values of ‘we’. It would be irresistible to relate this to ideas that we find in Wittgenstein’s later work.32 Many significant currents of philosophical thought are converging here. Collectively they signal how much is at stake in the claim that reality is not unified. To claim that reality is not unified is to claim that there are facts that are not part of ‘my’ world, or ‘our’ world. Not only is it obscure what this claim comes to. There is a venerable philosophical tradition to which it stands opposed. Nevertheless, it is a claim that Kant himself makes, or at least that he is prepared to  take seriously. In some parts of the Critique of Pure Reason, including the ‘Transcendental Deduction’, Kant appears to argue strenuously for the unity of reality, which in turn he appears to identify with the Cosmos.33 In fact, however, as I pointed out at the beginning of this essay, what Kant identifies with the Cosmos is empirical reality. He expressly allows for the possibility that there is more to reality than that. ‘An existence outside this field,’ he says, ‘cannot be declared absolutely impossible’.34 Partly he has in mind God.35 But he also allows for the possibility of beings whose sensibility is not ‘bound to the same conditions that limit our intuition’36 in other words beings whose own world is not spatio-temporal.37 If there were such beings, then there would assuredly be facts about their world that were not expressible from our point of view. For Kant, then, there is a distinction to be drawn between reality (‘our’ reality) and reality. Quite how this relates to that other distinction that he draws, between empirical reality and—and what?—here already there is an issue— the reality of things in themselves?—the reality of things in themselves?—that is a delicate exegetical question that need not detain us now. Suffice to remark that there is a familiar concern, famously pressed by Strawson himself in The Bounds of Sense),38 about whether Kant, or anyone else for that matter, has the resources to make sense of any such distinction. Strawson’s own reflections on the possibility of our engaging with other spaces certainly do not provide us with those resources. For suppose that Strawson is right in allowing for such a possibility. That does not yield a distinction between ‘our’ reality and reality: it yields a distinction between ‘our’ reality and the Cosmos. It shows 29 See again n. 16. 30 A95–130 and, differently in the second edition, B129–169. See also A401–402 and, for an outstanding discussion, Gomes (2011). In these terms, the second part of the second edition of the ‘Transcendental Deduction’ (§§22 ff.) can be viewed as an attempt to show that ‘my’ world extends to the whole of the Cosmos; Strawson’s original question can be viewed, at least in part, as the question why we should think that the Cosmos extends to the whole of ‘my’ world; and the issue with which I am primarily concerned can be viewed as the issue whether ‘my’ world extends to the whole of reality. 31 See esp. Wittgenstein (1961), 5.6 ff.; and cf. n. 26. 32 See e.g. Wittgenstein (1967), Pt. I, §§241–242. And see further below. 33 See also ‘The Postulates of Empirical Thinking in General’ at A218–235/B265–294, esp. A225/B272 ff. 34 A601/B629. 35 See e.g. A641/B669. 36 A27/B43. 37 See further A230/B282 ff. 38 Pt. Four.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

One World

189

that there may be more to ‘our’ reality than the Cosmos, not that there may be less to ‘our’ reality than reality. Moreover, it does so by extending Kantian principles that are intended by Kant himself to demonstrate the unity of the Cosmos, as Strawson makes clear at the end of the section with his appeal to ‘the connexion between objective reality and systematic unity’.39 In the scenario that Strawson envisages, there is, as he puts it, ‘a single concept of the objectively real’ extending to the different spaces.40 There is nothing here to threaten the unity of reality. Kant’s preparedness to allow for the possibility of beings with their own nonspatio-temporal world is something of an altogether more radical kind then. And note: the Kantian thought is not simply that there might have been such beings. The Kantian thought is that there may be such beings. May there? In what would their actuality consist? We have been catapulted back to precisely the sort of question that I earlier characterized as ‘intractable’. But was I too quick to dismiss a Lesliean response to such questions? Or, if not a Lesliean response, at any rate a response involving some appeal to God? Could we not defend Kant, and at the same time maintain the unity of reality, by declaring all facts to be expressible from God’s point of view? True, if some facts are not expressible from our point of view, this would mean abandoning the Unity Principle and trying to get some other grip on the Unity Relation.41 But that principle was only ever a working assumption, an assumption whose plausibility may now have been discredited. Invoking God in this way is problematical, however—for all sorts of reasons. First, and most obviously, it has no purchase if God does not exist. But it arguably has no purchase even if God does exist. For arguably God has no point of view.42 And even if God does have a point of view, we cannot simply take for granted that all facts are expressible from that point of view. (A theistic advocate of tensism will doubtless say that facts about temporal worlds are not expressible from God’s point of view because God’s point of view is atemporal.) Finally, even if all facts are expressible from God’s point of view, and even if this provides grounds for subscribing to the unity of reality, it does so only by trading one mystery—the mystery of a non-unified reality—for another—the mystery of Divine infinitude.43 This illustrates something that I take to be of the utmost importance. The claim that all facts are expressible from the point of view of a finite being, and hence that the whole of reality is, to that extent, accessible from somewhere within its midst, has a certain metaphysical bite which the claim 39 P. 152. 40 P. 152. 41 The following variation on the Unity Principle suggests itself: two facts, one of which is expressible from some point of view, stand in the Unity Relation to each other if and only if they are both expressible, if not from that point of view, then from a (common) point of view. 42 Cf. Moore (1997), p. 27. 43 Would there be grounds for subscribing to the unity of reality, but without incurring the same cost, in the existence, not of one privileged point of view from which all facts are expressible, but of an infinite sequence of points of view π1, π2, π3, . . . such that (i) for each πi in the sequence, any fact expressible from πi is expressible from πi+1 but not vice versa, and (ii) every fact is expressible from some πi in the sequence? No. That would incur the same cost. The mystery to which I have adverted lies not in the existence of one privileged point of view from which all facts are expressible, but in the existence of facts that are not expressible from ‘here’. The first stage in this sequence that involved such facts would also, already, involve the mystery.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

190

Language, World, and Limits

that all facts are expressible from the point of view of an infinite being lacks (an echo of the suggestion that I paraded earlier in connection with Dummett’s intuition).44 Now I have already urged that Strawson’s reflections on the possibility of our engaging with other spaces do nothing to support the disunity of reality. But do they support its unity? This is where Strawson’s one reference to Wittgenstein in The Bounds of Sense is relevant. To see how, let us begin by considering two initial concerns that we might have about this part of Strawson’s discussion. First, when he introduces Wittgenstein, he does so as a way of heralding ‘the social character of our concepts’,45 which he in turn takes to be relevant to the ratification of any given individual’s claim to have encountered something that, although it is beyond the Cosmos, is still part of ‘our’ world, and hence part of reality. This reference to the ‘social’ is unfortunate. Shortly afterwards he refers to the ‘public’ rather than the ‘social’, and he makes clear that by the ‘public’ he means the ‘objective’.46 This is much better. It is the public rather than the social, or the shareable rather than the shared, that Wittgenstein himself takes to pertain to our concepts.47 And it is the public rather than the social that reflects the fundamental character of ‘our’ world, indeed of ‘my’ world, indeed of any world.48 The second initial concern that we might have about this part of Strawson’s discussion relates to its whiff of verificationism. For Strawson seems to focus at least as much on what it would be for us to tell that something was part of reality as on what it would be for something to be part of reality. Insofar as his concerns are ontological, then they have less to do with what it takes for something to be part of reality than with what it takes for various kinds of relativization to be found within reality. Thus suppose it seems to me that I have visited a space in which pigs can fly. That (the fact that it seems to me that way) is part of reality. But what else is? Is there in fact a space in which pigs can fly? Or is it simply that, in these four dimensions, that is to say here in the Cosmos, I have imagined (or dreamt, or had my brain stimulated in such a way as to make it seem to me) that I was in such a space? The public ratifiability that Strawson considers is obviously related to the question of just what form of relativization we would think was appropriate in order to ‘place’, within reality, the appearance of volant pigs that I have experienced. Now we can readily give Strawson the benefit of the first of these doubts. This is simply because, as we have seen, he provides his own corrective. And once we have given him the benefit of that doubt, we can more easily see our way to giving him the benefit of the second doubt too. We can say that what is really at stake in Strawson’s discussion is not what would be involved in our collectively telling that something was part of reality, but what would make it possible for us to do so. True, this does 44 I do not mean to imply that nothing can remedy the lack. Consider e.g. Leibniz’ equation, which he takes to be fully visible only from God’s point of view, of what belongs to reality with what is for the best (see e.g. Leibniz (1973), §§53–60). And indeed we find essentially the same idea in Leslie. For although Leslie does not subscribe to the Leibnizian view that this is the best of all possible worlds, he does subscribe to the Leibnizian view that—if I may put it this way—this is the best of all possible worlds. 45 P. 151, emphasis in original. 46 Ibid. 47 Cf. Wittgenstein (1967), Pt. I, §243; and cf. McGinn (1984), pp. 79–80. 48 Even Dummett’s intuition does not require that there be more than one sentient creature in any world, at least not in its rawest form.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

One World

191

not make the whiff of verificationism go away, but it does arguably make that whiff less malodorous. And it gives us reason to think that, in order for some fact to be part of reality, there has to be shareable access of sorts to it from ‘our’ point of view. Furthermore, although the devil of what exactly this comes to lies in the detail of shareable access, it is hard to see how such access can fail to include expressibility. So Strawson’s reflections do indeed support the unity of reality. This in turn connects with what I said earlier about the finite/infinite distinction. For ‘we’ are finite. The accessibility of reality at issue here is the accessibility of reality to one of its own finite constituents: it is, as I put it earlier, the accessibility of reality from some particular point within its midst. If reality is unified, then that accessibility in a way constitutes its unity. Given any two facts, it ensures that they stand in the Unity Relation to each other, by suitably anchoring them both at the point in question. These various connections are, moreover, chief among those that Kant himself establishes in his ‘Transcendental Deduction’—albeit not, in his case, with a view to extending them beyond the Cosmos to reality as a whole. To say that the unity of reality is constituted by its accessibility from some point within its midst is to connect that unity with the possibility of a single finite representation of any arbitrary selection of its component facts: a representation that can be accompanied by the very same ‘I think’, or ‘we think’, that can accompany a representation, from there, of each of those facts severally.49 ‘If reality is unified,’ I said in the previous paragraph. I must repeat something that I have already been at pains to emphasize, that nothing in this essay is conclusive as far as that goes. In particular, where my support for the unity of reality has depended either on the claim that the Cosmos is unified or on the Unity Principle, I have been assuming each of these without argument. But I do believe that this essay, by indicating some of the challenges confronting the claim that reality is not unified, helps to show how extraordinary a claim it is. More than that, perhaps, the essay serves as a reminder of something of which Strawson’s work too is a reminder: the remarkable extent to which, and the remarkable depth at which, Kant’s work engages with our finitude.50

49 Cf. B131–132 and 138–139. 50 I am extremely grateful for comments from the audience at a meeting of the Post-Kantian Seminar in Oxford at which I presented an earlier version of this essay. Special thanks are due to Anil Gomes and Yuuki Ohta.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

PA RT I I I

Ineffability

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

15 Being, Univocity, and Logical Syntax Abstract The focus of this essay is the idea of the univocity of being, championed by Duns Scotus and given prominence more recently by Gilles Deleuze. Although one concern of the essay is with how this idea can be established, its primary concern is with something more basic: how the idea can even be properly thought. This issue is explored with the aid of some ideas about logical syntax borrowed from Wittgenstein’s Tractatus. One incidental benefit of the exercise is to indicate how there can be dialogue between analytic philosophers and those of other traditions.

1. The univocity of being There are three strikingly different groups of philosophers with whom an essay title such as this, in which mention of being or existence is conjoined with mention of the broadly semantic or the broadly grammatical, is liable to resonate especially loudly: analytic metaphysicians; historians of medieval philosophy; and students of poststructuralism, or more specifically students of the work of Deleuze. I shall say a little in connection with all three. But I shall say most in connection with the first and third, whose concerns I hope to illuminate by discussing them in tandem with one another. Prima facie their concerns are quite different. Analytic metaphysicians have long been exercised by questions about what exists. Do mathematical entities exist, for example?— and if not, how come there are infinitely many primes? Do holes exist?—or is it just that, among the physical things that exist, some are perforated or porous or some such? The more analytic metaphysicians have grappled with these questions, the more self-conscious they have become about what they are up to. There is accordingly a thriving branch of analytic metaphysics, sometimes referred to as metametaphysics, whose aim is to clarify what is at stake in addressing such questions. How far is it a matter of ascertaining the human-independent constitution of reality and how far a matter of settling on a way of speaking, perhaps even settling on an interpretation of the verb ‘exist’?1 The 1 In the predicate calculus, assigning a domain to a formal language with an existential quantifier is part of interpreting the language. But that is not what is meant here by settling on an interpretation of the verb ‘exist’. What is meant here would cut across that purely formal exercise by sanctioning the use of the verb ‘exist’ for only some of the things that could be part of any such domain. (Indeed some philosophers, of whom Quine is the most notable example, would take exception to it on precisely those grounds: see Quine (1960), §49.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

196

Language, World, and Limits

concerns within analytic metaphysics of this self-conscious kind are primarily semantic, epistemological, or methodological.2 Not so Deleuze’s concerns. Deleuze resurrects a medieval debate that arises even given an inventory of all the things that exist. Suppose we have such an inventory, never mind for the time being what might have been required to compile it. There is a question, on which the medieval debate turns, about the nature of the things in this inventory. One way to put the question is as follows: are any of these things so different in kind from ordinary things such as apples and snakes that our very talk of the ‘being’ of the former has to be understood differently from our talk of the ‘being’ of the latter? This is really a debate about transcendence. Another way to put the question is this: do any of these things enjoy an absolute transcendence? The concerns that motivate this question are more characteristic of mainstream traditional ontology than those that motivate the analytic metaphysicians’ questions about how to determine what exists. They have their source in Aristotle, who recognized different senses of being corresponding to his different categories.3 And the tools that the medievals used in pursuing their debate were largely Aristotelian tools. But they had a very particular focus which gave their debate its distinctively medieval stamp. They were interested in the nature of God. Their debate was about whether God’s transcendence, relative to His creation, was so radical that no language, or at least no language of a certain basic kind that includes the language of being, could be used with the same meaning with respect to both; and, if that were the case, whether it followed that any use of such language with respect to both must always involve brute equivocation or whether there was room for a more nuanced position whereby the literal use of such language with respect to God’s creatures could be extended to an analogical use of such language with respect to God Himself. Deleuze’s interest is a variant of this. He is interested in what it takes to make sense of everything as part of one and the same purely immanent reality—a reality that is free of any such absolute transcendence. Of course, there is no sharp division between Deleuze’s quest and the analytic metaphysicians’ quest. It is all very well my urging us to prescind temporarily from what would be required to compile an inventory of all the things that exist. But if this inventory had to include things that were absolutely transcendent, then there would be an issue about what it would mean even to say that such things ‘exist’, and that issue would lie squarely in the analytic metaphysicians’ territory. Even so, prima facie at least, there are two broadly different sets of concerns here. What animates Deleuze’s concerns? All manner of things. But he has two projects in particular that deserve special mention. One of these is to extend Heidegger’s work on being. Heidegger drew a distinction between being and the entities that have it.4 2 Two classic texts are Carnap (1956) and Quine (1961a). See also Chalmers, Manley, and Wasserman (2009), whose introduction by Manley provides very helpful orientation. Three recent contributions to the discussion are: Sider (2011), Ch. 9; Parfit (2011), App. J; and Turner (2010). An excellent discussion of how far a disagreement about what exists might be nothing but a disagreement about how to use the verb ‘exist’ is Williamson (1987), esp. §IV. 3 Aristotle (1941a), Ch. 4, and (1941b), Bk. I, Ch. 2, 185a20 ff. 4 In Heidegger (1962a) the translators register the former by using a capital ‘B’: see their n. 1 on p. 19. I  shall take the liberty of dropping the capitalization in all quotations from this book and other books where the same practice is followed.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

197

The entities that have it are all the things that we can be given or all the things of which we can make sense; their being is their very giveability or intelligibility.5 By drawing this distinction Heidegger enabled being itself to become a distinct focus of attention, a privilege which he argued traditional metaphysics had prevented it from enjoying. Not only that; he also contributed a great deal to our understanding of being. Deleuze wants to enlarge that understanding. He thinks that Heidegger has helped us towards a unified account of being. In particular, he thinks that Heidegger has helped us towards an account of being which, though it certainly acknowledges the many profound differences between entities, and indeed the many profound differences between their ways of being, does not cast any of them in the rôle of the absolutely transcendent and does not involve any Aristotelian polysemy. (Heidegger himself emphasized the magnitude of this task. At one point he put the pivotal question as follows: ‘Can there . . . be found any single unifying concept of being in general that would justify calling these different ways of being ways of being?’6) Nevertheless Deleuze thinks that Heidegger’s investigations are importantly incomplete. It is not enough, in Deleuze’s view, for an account of being to accommodate all these differences; it is not even enough for it to expose and articulate them all. If we are properly to understand what it is for everything to be immanent, then we must also understand how these differences themselves contribute to the fundamental character of being. We must acknowledge a kind of ontological priority that these differences enjoy over the entities between which they obtain, the entities whose being is under investigation, the things of which we can make sense. For this, Deleuze thinks, we need to look beyond Heidegger.7 The second project is somewhat more nebulous but no less significant. Appeals to absolute transcendence can be used to evade all sorts of problems, both theoretical and practical. Thus many people turn to their belief in God to help them acquiesce in suffering, and many of these in turn appeal to God’s absolute transcendence to help them address the question of how suffering can be part of God’s plan. Their answer is that we cannot really understand how, since all talk of God, including talk of God’s 5 Heidegger (1962a), §2 and p. 228. He says that entities include ‘everything we talk about, everything we have in view, everything towards which we comport ourselves in any way’ (ibid. p. 26). But in drawing the distinction between being and entities he also says that ‘the being of entities “is” not itself an entity’ (ibid. p. 26). At the very end of this essay I shall say a little about the apparent tension between these. (I have in mind the fact that precisely what seems to be required of us, for us to be engaged in reflection of this sort, is that we ‘talk about’, ‘have in view’, or ‘in some way comport ourselves towards’ being. A similar problem afflicts the characterization of entities that I ventured in the main text: ‘all the things of which we can make sense’. For Heidegger is intent on making sense of being. Suffice to say, for now, that my references to sensemaking in the main text may all need to stand under some suitable tacit restriction; more specifically, they may all need to be understood in such a way as to exclude certain sense-making that is characteristic of Heidegger’s project.) 6 Heidegger (1982), p. 176, emphasis in original. For a fascinating and instructive discussion see McManus (2013). 7 See Deleuze (1994), pp. 64–66; and cf. Smith (2001), pp. 169–170. (I am indebted to Smith’s excellent essay throughout this section.) Note: in a very interesting discussion of Heidegger on different ways of being, Kris McDaniel describes Heidegger as acknowledging ‘several senses of the word “being”’ (McDaniel (2009), p. 309). This suggests that McDaniel sees things very differently from Deleuze. The difference between them is not as great as it appears, however. In fact it is largely terminological. Certainly there is no suggestion in McDaniel’s discussion that Heidegger was an advocate of the radically transcendent.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

198

Language, World, and Limits

plan, is at most an analogical extension of talk that we can really understand. In Deleuze’s view, such appeals to absolute transcendence come too easily. Alluding to the famous Dostoevskian adage, he writes, ‘One must not say, “If God does not exist, everything is permitted.” It is just the opposite . . . It is with God that everything is permitted’.8 Learning how to live without recourse to absolute transcendence is in large part an ethical exercise for Deleuze. It means learning how to confront all that we are given in such a way as to make sense of it in its own terms, resisting the seductions of, as it may be, an inscrutable theodicy or an abstract teleology in whose transcendent terms all our afflictions are ultimately justified. It is a non-escapist lifeaffirming exercise. And the kernel of the second project is to think it through at an ontological level. In the medieval debate Duns Scotus repudiated absolute transcendence. He did not deny that some of our basic talk about God has to be interpreted non-literally.9 But he did deny that it all does. That, for Duns Scotus as for Deleuze, allows such talk too much free rein. It makes it merely formulaic and means that we cannot have any real grasp of what we are talking about when we indulge in it.10 In particular, this must apply to our very talk of God’s being. Duns Scotus accordingly championed ‘the univocity of being’.11 Deleuze seeks to do likewise. He considers two broad approaches to this problem, one that he finds in Spinoza and one that he finds in Nietzsche. The first approach involves viewing being as an entity in its own right. Any mention of the being of any other entity is to be understood as a reference to this entity. I say any mention of the being of any ‘other’ entity. 8 2005, p. 10. Note: the Dostoevskian adage, though often attributed to Dostoevsky in the form cited, does not in fact appear in that precise form anywhere in his corpus. The closest it comes to doing so is in the mouth of Mitya Karamazov, in Dostoevsky (1982), Pt. Four, Bk. 11, Ch. 4. There Karamazov says, ‘What’s to become of man then? Without God and without a future life? Why, in that case, everything is allowed. You can do anything you like!’ (p. 691). 9 This is important. It explains my continual use of the word ‘absolute’ to qualify the kind of transcendence that I am interested in. It is certainly possible for someone who repudiates absolute transcendence to accede to a relative transcendence whereby we have no way of making basic claims about a given entity in some of its aspects except by using language that demands to be understood non-literally. Spinoza may have been a case in point. We shall be considering Spinoza’s non-absolutely-transcendent conception of God in due course. But this conception did not, for instance, prevent Spinoza from both denying that God loves anyone in any strict sense (Spinoza (2002), Pt. V, Prop. 17, Corollary) and allowing for a non-strict sense in which God does love us (Spinoza (2002), Pt. V, Prop. 36, Corollary): the difference is that God’s ‘love’, unlike love properly so-called, is not ‘accompanied by the idea of an external cause’ (Spinoza (2002), Pt. III, ‘Definition of the Emotions’, §6). Whether this betokens relative transcendence or not depends partly on whether such claims about love are suitably basic and partly on whether we have some other, literal way of expressing what we mean when we say, ‘God loves us.’ I shall not pursue that question here. 10 It is instructive to consider Descartes in this connection. Descartes acceded, if not to the absolute transcendence of God, then at least to an extreme version of the relative transcendence of God: see e.g. Descartes (1984), Third Meditation, and Descartes (1985), Pt. I, §51. And in a way he would have been happy to concede that we lack any real grasp of what we are talking about when we talk about God: see e.g. Descartes (1985), Pt. I, §25, and Descartes (1991), p. 25. Indeed that helps to fuel his most celebrated argument for the existence of God: the fact that we lack any such grasp helps to show that our idea of God is an idea of something so great that only something that great can explain how we have come by it (Descartes (1984), Third Meditation). But of course, unless we have some grasp of what we are talking about when we talk about God, it is not obvious that we have an idea of God in the first place. For discussion see Williams (1978), pp. 143–145. And for discussion of some of the more general issues that arise here see Williams (2006), esp. pp. 14–15. 11 Duns Scotus (1987), pp. 19–20. For a helpful discussion see Tonner (2007).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

199

In fact, of course, any mention of the being of being itself had better be included. Thus in Spinoza, whose version of this approach Deleuze treats as more or less canonical, the rôle of being is assigned to an infinite substance in which everything that is, is; and that includes this very substance, which is in itself.12 Spinoza works hard, on Deleuze’s interpretation, to keep all intimations of the absolute transcendence of this substance at bay. He needs to work hard. The way in which he executes his project means that this substance is radically different in kind from any other entity. Its infinitude, if not its transcendence, is absolute13—which means, among other things, that its infinitude is unique to it.14 But Spinoza achieves his goal, at least on Deleuze’s interpretation, by not only allowing his substance to express itself through its attributes (notably through thought and extension, the two of its attributes of which we are aware), but by allowing it to do so in just the same sense as any other entity—that is, in just the same sense as any of its modes—each of which likewise expresses the essence of this substance through one of the substance’s attributes.15 Spinoza is not only able to say that his substance is extended, for example. He is able to say that it is extended in just the same sense in which Mount Everest is extended.16 There is no hint of absolute transcendence here. It is a further question, however, whether there is complete univocity of being, as Deleuze claims. The doctrine that any mention of the being of a thing is to be understood as a reference to one particular entity is already to be found in Aristotle. ‘All that “is”,’ Aristotle writes, ‘is related to one central point’17 and he goes on to insist that any mention of the being of a thing is to be understood as a reference to this ‘point’.18 Yet for Aristotle, this doctrine, so far from showing that there is a single sense of being, precisely corroborates his view that there are different senses of being. He likens it to the doctrine that any mention of the healthiness of a thing is to be understood as a reference to health, where that, he urges, betokens different senses of healthiness: a healthy diet is a diet that is conducive to health, a healthy complexion a complexion that is symptomatic of health, a healthy person a person who enjoys health, and so on. The common reference to health in each of these cases shows only that the differences of sense are not brute ambiguities. They remain differences for all that. In a distinction that Aristotle draws, the term ‘healthy’ applies in all these cases not ‘kath’ hen’, in virtue of one thing, but ‘pros hen’, with reference to one thing. Similarly, on Aristotle’s view, in the case of the term ‘being’.19 So what makes Aristotle and Spinoza, who both accept that any mention of the being of a thing is to be understood as a reference to one particular entity, arrive at such different conclusions? For Aristotle, it is pretty much axiomatic that certain differences between things are great enough, in themselves, to preclude a single sense of being in application to those things. Spinoza, by contrast—at least on Deleuze’s reading of him—is prepared to understand all differences between things, however great, as themselves constituting the character of being. (This is what I earlier said that Heidegger fails properly to do—again, on Deleuze’s reading of him.) The difference between Aristotle and 12 13 15 17 19

Spinoza (2002), Pt I, Def. 3, Ax. 1, and Prop. 15. This substance is what Spinoza calls ‘God’: see n. 9. Ibid. Pt. I, Def. 6. 14 Ibid. Pt. I, Props 14 and 15. Ibid. Pt. I, Defs. 3, 4, and 5, and Pt. II, Def. 1. 16 See Deleuze (1990a), pp. 27, 46 ff., and 59. Aristotle (1941c), Bk. IV, Ch. 2, 1003a32–33. 18 Ibid. 1003b6–7. Ibid. 1003a32 ff. See Shields (2012) for an interesting discussion.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

200

Language, World, and Limits

Spinoza is thus a fundamental difference of approach. On the Spinozist approach, it is not just that any mention of the being of a thing is to be understood as a reference to one particular entity; any mention of the multiplicity and diversity of things is to be understood as a reference to that entity, whose essence is expressed in the very differences between them.20 This is what Deleuze calls an ‘affirmation’ of difference. For x to be affirmed of y, in Deleuze’s sense, is—very roughly—for the essence of y, or for the sense of y, to be expressed through x; and for x to be affirmed tout court is for x to be affirmed of being.21 This affirmation of difference allows Spinoza to see being in the differences between things, rather as one sees a single image in the differences of hue, the differences of brightness, and the differences of location between the myriad different pixels that compose it. It allows him to make sense of things as part of one and the same immanent reality by making integrated sense of what differentiates them.22 But Deleuze thinks we can go further. For Spinoza still believes in one privileged unified entity that is prior to all multiplicity, prior to all diversity, prior to all difference. Deleuze thinks we can acknowledge the univocity of being without appeal to any such entity. (To this extent he is in line with Heidegger. Heidegger denied that being is itself an entity.23) If we deny that there is any such entity, we can understand any mention of the multiplicity and diversity of things entirely in its own terms and still make sense of things as part of one and the same purely immanent reality. This is not just an affirmation of difference. It is an affirmation of the affirmation of difference. For it allows difference itself to be its own affirmation. Such affirmation is no longer conceived as a reference to some other entity, some single substance in which everything is anchored. There is no such entity. Difference is itself the ultimate reality. Indeed it is itself ever different, like an ever changing image in which the only constancy is the change. Being can still be seen in the differences between things; but it no longer has any stable identity or overall unity of its own. It no longer has any independent status as an entity in its own right.24 This is the second of the two approaches that Deleuze considers, the one that he finds in Nietzsche.25 There is a famous sentence on the final page of Deleuze’s Difference and Repetition where he summarizes his thinking by saying, ‘All that Spinozism needed to do for the univocal to become an object of pure affirmation [i.e. affirmation that is affirmation of itself] was to make substance turn around the modes’, and then refers to Nietzsche’s doctrine of eternal return which he interprets as providing the wherewithal to do precisely that.26 It is the Nietzschean approach that Deleuze favours. But on what grounds? On what grounds, for that matter, does he favour Spinoza’s original preparedness to 20 These differences, which are differences of power, are what Deleuze characterizes as intensive differences: see Deleuze (1994), Ch. V, for an exploration of intensive difference in general and Deleuze (1990a), Pt. Three, for application of this notion to the case of Spinoza. 21 See e.g. Deleuze (1994), pp. 52 ff. 22 See Deleuze (1990a), Ch. XI. And cf. some of the metaphors on the very last page, p. 304, of Deleuze (1994). 23 See n. 5. 24 I shall say a little more about this in section 2. 25 Deleuze (2006), esp. Ch. 5, and more esp. §§10–13, but see also Deleuze (1994), pp. 35–42. 26 Deleuze (1994), p. 304. For Deleuze’s interpretation of the doctrine of eternal return see Deleuze (2006), esp. Ch. 2.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

201

understand all differences as constituting the character of being in defiance of Aristotle’s unpreparedness to do so? Nothing that I have said so far in this essay appears to forestall a simple stand-off between Spinoza and Aristotle; nor, perhaps, between Nietzsche and Spinoza. It is natural to look for arguments here. And relevant arguments are to be found.27 Nevertheless, they are not my primary concern in this essay. My primary concern is not with how the univocity of being can be established. It is with something more basic. It is with how the univocity of being can even be properly thought.28 I hope that some of what I have said so far has helped in this respect. In the second section of this essay I want to recapitulate this material, but in a different form, a form that relates it back to some of the concerns of analytic metaphysicians—and indeed of analytic philosophers more generally.29

2. How to think the univocity of being in logico-syntactic terms I begin with a matter that might initially appear quite unrelated. Wittgenstein, in his Tractatus Logico-Philosophicus, draws a distinction between what he calls ‘signs’ and what he calls ‘symbols’.30 Signs are the written marks or noises that we use to communicate. Symbols are signs together with their logico-syntactic use. Logical syntax is akin to ordinary grammar, but deeper. Thus ordinary grammar associates the use of the word ‘are’ in ‘Human beings are animals’ with the use of the word ‘eat’ in ‘Human beings eat animals.’ Logical syntax recognizes differences between these, reflected in the fact that it makes sense to add to the latter sentence, but not to the former, ‘including themselves’.31 Wittgenstein captures the relation between a sign and a symbol as follows: ‘A sign is what can be perceived of a symbol’.32 And he points out that ‘one and the same sign . . . can be common to two different symbols’.33 Thus the word ‘round’ is sometimes used as a noun to denote a slice of bread, sometimes as an adjective to indicate circularity: one sign, two symbols. Now I am going to assume the following: difference of logico-syntactic use is a difference of degree; in other words, one symbol can share more or less of its logico-syntactic use with another. I do not claim that this doctrine is in the Tractatus itself. Whether it is, or whether for that matter its negation is, and in either case how explicitly or implicitly, are large exegetical issues that I shall put to one side. My aim is to make use of Wittgenstein’s ideas, not to rehearse them. Adopting this doctrine seems to me the 27 See again the material cited earlier in the main text: Deleuze (2006), esp. Ch. 5, and Deleuze (1994), pp. 35–42. 28 Cf. Smith (2001), p. 168, where he quotes a remark that Deleuze himself made in a seminar: ‘[Univocity is] the strangest thought, the most difficult to think, if it has ever been thought.’ 29 For a helpful discussion of the material in this section see (in addition to Smith (2001), to which I have already made several references) de Beistegui (2004), pp. 225–241. 30 Wittgenstein (1961), 3.31 and 3.32 ff. 31 It is often said that symbols are signs together with their meaning, and not just their logico-syntactic use: see e.g. Black (1964), p. 130. That this is wrong is well argued by Johnston (2007). But see below for why the difference between these may not be as great as it seems. 32 Wittgenstein (1961), 3.32. 33 Ibid. 3.321.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

202

Language, World, and Limits

only plausible and interesting way of extending those ideas, or at least the only plausible and interesting way of extending them that subserves our current purposes. Here is an illustration of the doctrine. The word ‘round’, as well as having the two meanings already indicated, is also sometimes used as a noun to denote a complete series of holes in golf. This is yet another symbol, different from either of the other two. But it is less different from the other noun than it is from the adjective. This can be seen in the following way. There are many meaningful sentences involving the word ‘round’, such as ‘I had a round yesterday’, in which the meaning of the rest of the sentence prevents the word ‘round’ from functioning as an adjective but still allows it to function as either of the two nouns.34 In other meaningful sentences, including sentences that build on this one, such as ‘I had a round yesterday and I had it toasted’, the meaning of the rest of the sentence prevents the word ‘round’ from functioning as one of those two nouns but not as the other.35 However, there is no equivalent transverse ordering. There is no meaningful sentence in which the meaning of the rest of the sentence prevents the word ‘round’ from functioning as one of those two nouns but still allows it to function either as the other noun or as the adjective.36 Now, are there any ambiguities that do not involve any difference of logico-syntactic use, ambiguities that are, so to speak, simple differences of meaning? (In current terms such an ambiguity would involve one symbol, not just one sign.37) One’s first thought is that there surely are. But on reflection the matter seems less clear. Exact sameness of logico-syntactic use cuts very finely indeed.38 Very crudely speaking, if 34 Is this perhaps contestable? What if someone were to insist that the word ‘had’, in this example, is itself correspondingly ambiguous, meaning either ‘ate’ or ‘played’? So be it. The point that I am making then merely requires a different example, say with the word ‘enjoyed’ in place of the word ‘had’. If someone were to deny that there is any example that suits my purposes, then I suspect that they would simply be assuming the opposite of what I am assuming—that difference of logical syntax is a straightforward difference of kind—in which case their denial would lie outside the ambit of my discussion. 35 Is this perhaps contestable? What if someone were to insist that the word ‘round’, in this example, can function as either noun; it is just that, in one case, the sentence can only be used to say something false? I would deny that. I think an utterance of ‘I had a complete series of holes yesterday and I had it toasted’ would be meaningless, not false. But I agree that this is contentious. The contentiousness will be significant later. 36 If it is true that difference of logico-syntactic use is a difference of degree, then that seems to me to have a bearing on a number of other philosophical issues. In particular, I think it has a bearing on Frege’s problem about the semantics of predicates: see Frege (1997a). It is far beyond the scope of this essay to go into detail about what I have in mind. But I can give an indication of what I have in mind by quoting from Wright’s essay on this problem. Wright writes, ‘[We could] stipulate that “refers to” should have a use linking the name of a predicate to an expression—par excellence, the predicate itself—for its semantic value. In that case, “‘is a horse’ refers to is a horse” . . . would be well-formed, but—just for that reason—“refers to”, so used, would not speak of the relation that holds between a singular term and the object for which it stands’ (Wright (1998), p. 256, some emphasis removed). But why not, if difference of logico-syntactic use were a difference of degree? For then a single symbol could combine in the same way with other symbols, even though those other symbols had a different logico-syntactic use from each other. 37 See n. 31. The discussion that ensues in the main text explains the final sentence in that note. 38 E.g. consider the ambiguity of the word ‘billion’, used in American English to denote a thousand million and used in British English to denote a million million. That may look like an ambiguity that does not involve a difference of logico-syntactic use. But I think there is a good case for saying that, even in that case, a difference of logico-syntactic use is involved. For it seems to me that, in the sentence ‘The number of stars in the Andromeda nebula is less than a billion but more than a thousand million’, the meaning of the rest of the sentence prevents the word ‘billion’ from functioning as the American-English noun. Admittedly, this is a matter of contention—essentially the same contention as was noted in n. 35—but in this context

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

203

there were an ambiguity that involved no difference of logico-syntactic use, there would have to be two things such that whatever could be meaningfully said of one could be meaningfully said of the other. This is of course weaker than the requirement that there be two things such that whatever could be truly said of one could be truly said of the other. Even so, it is still a strong requirement. And who knows but that there are deep reasons of philosophical principle why it could never be met? Considerable work in some or all of metaphysics, philosophical logic, and philosophy of language would be needed to settle the issue. But here is a more tractable issue. Is it possible to expose an ambiguity without exposing any difference of logico-syntactic use? This time there seems little room for doubt. Clearly it is. For it is possible to show that a word is ambiguous by producing a sentence involving the word, the rest of whose meaning is presumed given, and then pointing out that a single utterance of the sentence can be interpreted as true or as false depending on how this particular word is construed. (For instance, imagine my saying, ‘I had a round yesterday’ on the morrow of a day on which I had a slice of toast but did not so much as set foot on a golf course.) The word is thereby shown to be ambiguous even though the question of whether any difference of logico-syntactic use is involved is left open. Let us now reconsider the univocity of being. And let us think of this issue as an issue about whether the word ‘being’ and its various cognates are relevantly ambiguous.39 To be sure, this may not be as innocent as it appears. It is not entirely obvious that the issue can be cast in this linguistic form without some loss. But even if it cannot, the new casting of the issue will at the very least be of (highly pertinent) interest. Anyone committed to the non-univocity of being would in these terms be committed to an ambiguity in the word ‘being’. But the commitment would, I claim, be unsustainable unless the ambiguity could be exposed in a way other than that just described. The ambiguity would have to be, and would have to be seen to be, an ambiguity involving a difference of logico-syntactic use. Why do I claim this? For two principal reasons. Or perhaps rather, for one principal reason that can be broached in two ways. First, to think that the word ‘being’ is ambiguous is to think that some things are so different in kind from others that talk about the ‘being’ of the former cannot be understood in the same way as talk about the ‘being’ of the latter. But unless the ‘cannot’ here means ‘cannot, as far as the meaningfulness of the talk is concerned’ as opposed to ‘cannot, as far as the truth of the talk is concerned’, then it is just not clear why anyone would think such a thing. Unless there are differences in kind between things that are so great that the very business of characterizing some of these things requires a different logical syntax from the business of characterizing others, what is to prevent the devising of vocabulary that can be truly applied to all of them? And if nothing is to prevent this, then what is to prevent the devising of a term that can be truly applied to whatever the word ‘being’ can that is as much as it needs to be. For I claim only that it is not obvious that there are ambiguities that do not involve a difference of logico-syntactic use. 39 I add ‘relevantly’ because the word ‘being’ itself functions as a non-count noun, as a count noun, and as a present participle, but these ambiguities cut right across the issue of the univocity of being. Henceforth, I shall take such a qualification for granted.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

204

Language, World, and Limits

be truly applied to, under each of its supposed interpretations? But given such a term, and given the work that it has to do (notably, enabling us to refer to the character of whatever exists), what is the rationale for thinking that the word ‘being’ is ambiguous in the first place? Why not accept that this new term is just a synonym for ‘being’, with (as we now see) its single generic meaning?40 This question leads naturally to the second way of broaching the matter. Unless the ambiguity in the word ‘being’ involved a difference of logico-syntactic use, what rationale would there be for acknowledging different senses of the word ‘being’ as opposed to acknowledging differences among the entities to which the unambiguous word ‘being’ can be truly applied? This is an old idea, famously and marvellously captured by Quine in §27 of his Word and Object. Quine is there concerned with a somewhat different issue: whether the terms ‘true’ and ‘exist’ are ambiguous. But his response to the claim that they are is essentially the same as the response that I am now recommending to the claim that ‘being’ is non-logico-syntactically ambiguous. Quine writes, specifically in connection with ‘true’: There are philosophers who stoutly maintain that ‘true’ said of logical or mathematical laws and ‘true’ said of weather predictions or suspects’ confessions are two usages of an ambiguous word ‘true’ . . . What mainly baffles me is the stoutness of their maintenance. What can they possibly count as evidence? Why not view ‘true’ as unambiguous but very general, and recognize the difference between true logical laws and true confessions as a difference between logical laws and confessions?41

Similarly, I submit, in the case of ‘being’. And note that acceding to a single sense of being in this way would not preclude, in fact would encourage, acknowledging different kinds of being corresponding to the various fundamental differences between entities.42 This is why Heidegger, who is certainly keen to acknowledge different kinds of being—for example, those kinds of being that are peculiarly enjoyed by ‘whos’ and those kinds of being that are peculiarly enjoyed by ‘whats’43—can nevertheless be considered a champion of the univocity of being.44 In linguistic terms, then, the doctrine of the non-univocity of being had better be construed as the doctrine that the word ‘being’ has more than one logico-syntactic use; that there is one sign here that is common to more than one symbol. A possible analogy would be the use of the existential quantifier ‘∃’ in formal languages to represent both first-order quantification and second-order quantification. There too, 40 These are rhetorical questions: but see Turner (2010) for discussion. 41 Quine (1960), p. 131. 42 Cf. Deleuze (1990b), p. 185. This relates to the point that I made about McDaniel in n. 7. 43 For the distinction between a ‘who’ and a ‘what’ see Heidegger (1962a), p. 71. For the recognition of different kinds of being see ibid. p. 26. And see ibid. §20, for admonishment of Descartes for not appreciating that different kinds of being do not equate with different senses of being. 44 Parfit presents reasons for thinking, contra Quine, that ‘exist’ is ambiguous (Parfit (2011), pp. 720 ff.). I do not myself find these reasons compelling. For an excellent defence of Quine’s view (which nevertheless involves a complete misunderstanding of what Heidegger means by ‘being’) see van Inwagen (2009). For further relevant material see Wittgenstein (1969a), p. 58 and Quine (1960), pp. 241–242.—Note: Ryle’s remarks, in Ryle (1949), Ch. I, §3, are often taken to epitomize the view under attack here, by Quine and van Inwagen themselves among others (see Quine (1960), p. 131, n. 2 and van Inwagen (2009), pp. 485 ff.). But in current terms Ryle is surely an exponent of the view that ‘exist’ is ambiguous in a way that involves a difference of logico-syntactic use: see Price (2009), p. 331.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

205

arguably, there is one sign that is common to more than one symbol.45, 46 Nor would the analogy stop there. Anyone who took this view of both ‘being’ and ‘∃’ would no doubt insist that, in the case of ‘being’, just as in the case of ‘∃’, the similarities between how the different symbols are used are sufficiently striking and sufficiently important for the commonality of the sign to be both natural and warranted.47 This was essentially Aristotle’s position. For Aristotle, the differences in reality to which different senses of being corresponded were categorial differences, deep ‘grammatical’ differences of just the sort envisaged here. But he also insisted that the use of a single word to embrace these different senses of being was both natural and warranted: this was what the comparison with healthiness was intended to show. To repeat: the advocate of the non-univocity of being had better think that different uses of the word ‘being’ differ in their logico-syntactic use. But this presents a challenge of its own: how to show that they do. The sheer fact that the word is used in linguistic contexts which themselves differ in their logico-syntactic use is not decisive. Consider these two contexts: ‘That person is . . . ’ and ‘That tree is . . . ’ These differ in their logico-syntactic use. But it does not follow that the phrase ‘exactly two metres in height’, which can be meaningfully inserted into both, is logico-syntactically ambiguous48—not granted the assumption that difference of logico-syntactic use is a difference of degree.49 Similarly, the fact that we can talk about the ‘being’ of that person and the ‘being’ of that tree does not show that ‘being’ is logico-syntactically ambiguous. It may be a necessary condition of the non-univocity of being that ‘being’ should have application to things that are so different that there is no single logico-syntactic way of making reference to all of them; but it is not a sufficient condition. The difficulty is exacerbated by the following fact. The simple way of exposing an ambiguity which we considered earlier—namely, producing a sentence involving the ambiguous word and pointing out that a single utterance of it can be interpreted as true or as false, leaving open whether the word is logico-syntactically ambiguous— has no counterpart when it comes to showing that a word is logico-syntactically ambiguous. It is of no avail to produce a sentence involving the logico-syntactically ambiguous word and then to point out that a single utterance of it can be interpreted as meaningful or as meaningless. Provided that interpreting an utterance as meaningless is not a contradiction in terms, then this is something that one can do to any utterance whatsoever. (One can always construe some word in the utterance as occurring 45 I add ‘arguably’ for at least two reasons. First, it is not entirely uncontroversial that ‘∃’ has both these uses: for some famous dissent see Quine (1970a), pp. 66 ff.; see also van Inwagen (2002). Secondly, even if ‘∃’ does have both these uses, it is not obvious that they are logico-syntactically different uses: some material later in the main text bears on this point. 46 I am taking for granted in this essay that if a sign is common to more than one symbol, then it is ambiguous. For a fascinating argument to the contrary see Williams (1996). But I wonder how deep the disagreement between Williams and me is: see e.g. Williams’ own reference to the possibility of settling this matter by stipulation (Williams (1996), pp. 59–60), and see David Wiggins’ reply to Williams in Wiggins (1996), §§35–40. 47 Wittgenstein (1961), 5.52 is relevant here, as is Turner (2010), §II. 48 This is the material that I had in mind at the end of n. 45: the sheer fact that ‘∃’ can occur alongside both individual variables and predicate variables, if it is a fact, is not decisive for the logico-syntactic ambiguity of ‘∃’. For pertinent discussion see Quine (1970a), Ch. 2; but note that Quine’s notion of grammar is considerably looser than the notion of logico-syntactic use at work in this essay. 49 A related point is that there is relative transcendence as well as absolute transcendence.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

206

Language, World, and Limits

without either its standard meaning or any other meaning.) It cuts no ice at all where ambiguity is concerned. Given an utterance of the sentence, ‘Her brooch is round’, for example, we can construe ‘round’ as occurring without either its standard adjectival meaning or any other meaning. But that is quite irrelevant to the use of ‘round’ as a noun. Nothing about this sentence is relevant to the use of ‘round’ as a noun, given the meaning of the rest of the sentence. The meaning of the rest of the sentence precisely precludes the use of ‘round’ as a noun here.50 I am not suggesting that there is no way of exposing a logico-syntactic ambiguity. In the case of the word ‘round’, it is perfectly acceptable simply to point out that the word has both a nominal use and an adjectival use. Or, if there are certain theoretical purposes at hand for which further detail is required, either in connection with the word ‘round’ or in connection with the difference between nouns and adjectives, then we can go into just such further detail. Even in cases where the difference of logicosyntactic use is less marked, as it would be in the case of the word ‘being’, and where the tools for characterizing the difference are not ready to hand, as they might not be in the case of the word ‘being’, we can do what we did where the two uses of ‘round’ as a noun were concerned: produce a sentence involving the ambiguous word (‘I had a round yesterday’) together with a context within which the sentence can be meaningfully embedded under one interpretation but not under the other (‘. . . and I had it toasted’). The problem, however, is that this would be a way of exposing the ambiguity only to those who were already disposed to see it. If there were genuine controversy about whether the word had more than one logico-syntactic use, as there is in the case of the word ‘being’, no such expedient would help to settle the matter. The denier of logico-syntactic ambiguity could simply deny that embedding the given sentence in the given context resulted in any relevant meaninglessness51—whilst also of course acknowledging the ever-present and uninteresting possibility noted in the previous paragraph, that the word in question be construed as occurring without any meaning whatsoever. The advocate of the non-univocity of being may now appear to be in trouble. I have been urging, on the one hand, that the relevant ambiguity in ‘being’ would have to be exposed as a logico-syntactic ambiguity while suggesting, on the other hand, that there would be no exposing it as such that did not essentially involve preaching to the converted. But actually the trouble is just as great for an advocate of the univocity of being. Insofar as there is a kind of surd in what one of them wants to assert, there is a kind of surd in what the other wants to deny. This is why there is an issue, not merely concerning how the univocity of being can be established, but concerning how it can even be properly thought. What can an advocate of the univocity of being do, to impress his doctrine on himself as well as on others, beyond blankly proclaiming that ‘being’ has just one meaning? Well, one option that he might take is to identify being as an entity in its own right and to insist that any talk of the ‘being’ of a thing is a reference to this entity. But as we saw in section 1, this would not be enough. An Aristotelian would insist that any talk of the ‘healthiness’ of a thing is a reference to health, but would deny that ‘healthiness’ 50 The considerations in this paragraph owe much to Wittgenstein (1961), 5.473–5.4733. 51 This is what I had in mind in n. 35.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

207

has just one logico-syntactic use. The advocate of the univocity of being would need to insist further that any talk of the ‘being’ of a thing is not just a reference to this entity, but a reference of one particular logico-syntactic kind to this entity. But now there would be another impasse of sorts. The Aristotelian would see the differences between the things to which the word ‘being’ can be truly applied as simply too great for that to be a viable option. The most effective way for the advocate of the univocity of being to evade this sort of Aristotelian response is by being pre-emptive. Given various things to which the word ‘being’ can be truly applied, and given the various differences between them, the advocate of the univocity of being can say that precisely one of the functions of the word is to signify these differences; that for these things to be is for them to differ in the ways they do; that difference is itself the fundamental character of being.52 There is then no question of two things’ differing to such an extent that the word ‘being’ has no single logico-syntactic use in relation to both. The very semantics of the word forestalls this.—But wait! Is it not question-begging to appeal to ‘the’ semantics of the word when what is at issue is whether the word even has one logico-syntactic use?— That is not what is at issue. What is at issue is how we can properly think of the word as having one logico-syntactic use. And the suggestion is: by first thinking in terms of its semantics; by thinking of its semantics as itself already encompassing all diversity in things. The word ‘being’ is to be understood in such a way that, whenever there are things that differ from one another, even if they differ so much that there is no single logico-syntactic way of making reference to all of them, this word can be truly applied to them, in a single sense, and therefore with a single logico-syntactic use. One way to regard the shift from thinking of being as an entity in its own right to thinking of difference as the character of being is as follows: it is a shift, in the attempt to think of ‘being’ as having just one logico-syntactic use, from modelling the word on a noun to modelling it on a verb.53 So long as the word was modelled on a noun— so long as the word was conceived as standing in an invariant semantic relation to one particular entity—there was an issue about whether talk of the ‘being’ of things could do suitable justice to all the ways in which things differ from one another. For even if such talk could secure the univocity of being, how could it do so except at the price of introducing fresh concerns, if not the same old concerns, about the univocity of participation in being? And if it did introduce such concerns, how could the obvious regress be blocked, except by locating the really important univocity in precisely that differing of things from one another which was giving pause? But the regress never even starts, nor are there any such concerns, if the word ‘being’ is modelled on a verb and is conceived as expressing such differing in the first place. Substance, the entity to which the noun ‘substance’ refers, is now thought of as turning, the activity to which the verb ‘turn’ refers, around the modes. Furthermore, to think of difference as the character of being54 is to allow for the affirmation of the affirmation of difference to which I referred in section 1. What I have in mind is this. The affirmation of difference is what enables being, univocal 52 Cf. Deleuze (1994), p. 222. 53 Cf. Deleuze (1990b), 26th Series. 54 A famous remark of Nietzsche’s is pertinent here: ‘To impose upon becoming the character of being— that is the supreme will to power’ (Nietzsche (1967), §617).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

208

Language, World, and Limits

being, to be seen in the differences between things. It is a making sense of difference, as the face of univocal being, where this is as much difference’s making sense as it is difference’s being made sense of. But if difference is the very character of being, then it is itself what enables univocal being to be seen in the differences between things. It is its own affirmation. The making sense of difference is simply things’ differing. The affirmation of difference is the affirmation of the affirmation of difference. What we have been witnessing—at least, what we have been witnessing if Deleuze’s reading of Nietzsche and Spinoza is correct—is the power of the Nietzschean verb over the Spinozist noun, or the power of Nietzschean differing over Spinozist substance. Spinozist substance is an entity that differs from other entities in various ways. Nietzschean differing is not an entity at all. It cannot be said to differ from other entities. It cannot be said to differ from anything in the way in which entities differ from one another. On the other hand, it can in a way be said to differ. For there is a sense in which, in the differing of entities from one another, differing itself is ever different. (If it were not, the differing of entities from one another would stand to it in something like the relation of instantiation to a universal, and it would count as an entity after all.) But how can it be said to differ if it cannot be said to differ from anything in the way in which entities differ from one another? One way is through a break with traditional grammar: it can be said to differ from itself.55 And if differing is said to differ from itself, this gives further fillip to the idea that the manifestation of being through differing is itself a differing of sorts, a differing through which being is manifest; in other words, that the affirmation of difference is the affirmation of the affirmation of difference. This is the line taken by Deleuze, in his exposition and defence of the univocity of being. As I have tried to make clear, I have not myself been concerned to defend the doctrine, still less to defend any of the associated exegesis; I have been concerned with what it takes to think the doctrine. But I have had a more particular aim too: to connect what it takes to think the doctrine with issues that exercise analytic philosophers. Not that this kind of linkage is likely to win any converts. Just the opposite in fact. It is likely to crystallize alternative ways of thinking in the minds of analytic philosophers. Thus many analytic philosophers will recoil from unadorned talk of anything’s differing from itself by demanding some kind of relativization, such as that which allows for talk of Ellen’s differing from herself by being both a child and an adult: a child then and an adult now. Others will recoil no less from relativized talk of something’s differing from itself and they will deny that there is strictly any identity between that girl and this woman.56 Again, many analytic philosophers will insist that the word ‘differing’, as it occurs in the sentence ‘Differing is not an entity’, functions as a singular term, and—an entity being nothing but what is picked out by a singular term—that the sentence is self-stultifying. They are then liable to conclude that, if differing is anything at all, then it is (perforce) an entity. Others will take a leaf out of each of Frege’s and the early Wittgenstein’s books: they will acknowledge the self-stultification 55 See Moore (2012), Ch. 7, §7, for a discussion of this sort of break with traditional grammar, and Moore (2012), Concl., §3(b), for application of the discussion to this very case. 56 See Lewis (1986), pp. 204–206.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Being, Univocity, and Logical Syntax

209

in the sentence ‘Differing is not an entity’ and they will conclude that there is an insight here to which the sentence is gesturing but which cannot strictly be expressed.57 Those who take this last option will of course be manifesting an element of conciliation. How close they will be to a convergence of view with any champion of the univocity of being is going to depend in part on how comfortable any champion of the univocity of being is with this kind of appeal to the inexpressible: in some cases, I submit, very comfortable.58 But that is not the point. The point is not about convergence of view. It is not even about rapprochement. The point is about dialogue. Some analytic philosophers might eventually feel at home with these ways of thinking; some might even eventually be persuaded by Deleuze to subscribe to the univocity of being. But first they have to be able to listen to what he is saying.59

57 See Frege (1997a), esp. p. 184; and Wittgenstein (1961), 4.12 ff. 58 I particularly have in mind Heidegger: see Heidegger (1999), §§27 and 265. This relates back to the apparent tension in Heidegger to which I referred in n. 5. I believe that the tension is real, and that Heidegger’s own appeal to the inexpressible serves a similar function to that of the analytic philosophers envisaged in the main text. 59 I am very grateful to Denis McManus and Philip Turetzky for comments that helped me to improve this essay.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

16 Ineffability and Religion Abstract It is argued that, although there are no ineffable truths, the concept of ineffability nevertheless does have application—to certain states of knowledge. Towards the end of the essay this idea is related to religion: it is argued that the language that results from attempting (unsuccessfully) to put ineffable knowledge into words is very often of a religious kind. An example of this is given at the very end of the essay. This example concerns the Euthyphro question: whether what is right is right because God wills it, or whether God wills it because it is right.

1. The question whether there are any ineffable truths: its clarification and one simple argument for answering it negatively Are there any ineffable truths? I am reminded, in addressing this question, of Donald Davidson’s reaction to a similar question that he poses in his celebrated essay on the idea of a conceptual scheme:1 could there be a language whose sentences were untranslatable into any of ours? ‘It is tempting,’ Davidson writes, to take a very short line indeed: nothing, it may be said, could count as evidence that some form of activity could not be interpreted in our language that was not at the same time evidence that that form of activity was not speech behaviour. If this were right, we probably ought to hold that a form of activity that cannot be interpreted as language in our language is not speech behaviour. Putting matters this way is unsatisfactory, however, for it comes to little more than making translatability into a familiar tongue a criterion of languagehood. As fiat, the thesis lacks the appeal of self-evidence; if it is a truth, as I think it is, it should emerge as the conclusion of an argument.2

Davidson’s objection is to the shortness of the short line, then, not (so to speak) to its direction. At a key moment later in the essay he finds himself addressing the following subsidiary question: ‘[H]ow well [do] we understand the notion of truth, as applied to language, independent of the notion of translation[?]’3 And his answer,

1 Davidson (1984g).

2 Ibid. pp. 185–186.

3 Ibid. p. 194.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

211

which he goes on to elaborate, is a virtual re-appropriation of the short line: ‘[W]e do not understand it independently at all’.4 In considering whether there are any ineffable truths, I too find myself tempted to take a short line which, on reflection, I think has only its shortness to be held against it. In what I take to be a Davidsonian vein I want to say that we do not understand the notion of truth (at least when it is construed in the most narrowly literal way) independent of what can be expressed linguistically; or rather, more cautiously, that we do not understand the notion of a truth independent of that. (The individuation involved in talking about a truth, by forcing the question of where one truth stops and another begins, clearly puts additional demands on talk in general about truth.) But I also concede that there is much more to be said.5 In due course I shall give a brief indication of what more I think can be said. But first I want to make clear that I take myself to be defending a negative answer to my opening question (whether there are any ineffable truths). That is, I am construing the effability of a truth as the possibility of its being expressed linguistically. That is already something of a dictate. (There are non-linguistic means of expression.) But it still leaves a great deal of room for further construal. Any claim about what can be expressed linguistically raises substantive questions about the interpretation of all three of ‘can’, ‘expressed’, and ‘linguistically’. For instance, can a truth be expressed linguistically only if there is a sentence in some existing language that has precisely that truth as its content? Or does it suffice that there be some (true) sentence in an existing language that has that truth as part of its content? If the latter, in what sense of ‘part’? Does it perhaps suffice that there be some sentence in an extension of an existing language that has that truth as its content? If so, then what counts as an extension of an existing language? Or is the appeal to existing languages too restrictive? Does it suffice that there be some (true) sentence in some possible language that has that truth as (part of) its content? Then what counts as a possible language?6 Many of these questions have divergent but equally legitimate answers. They are equally legitimate in the sense that any apparent conflict between them is purely terminological. They simply correspond to different definitions of effability. For my own part, when I deny that there are any ineffable truths, I do so by what seem to me to be the lowest reasonable standards. (I shall not now try to spell out in detail what these are. But, as a rough indicator, I take something to be effable if and only if, first, it has content in virtue of which it is either true or false—both my belief that grass is green and indeed the truth that grass is green would be clear cases in point—and, second, this content is entailed by the content of some possible representation in some possible language that is capable, in principle, of being understood by a finite being.7) This makes my denial as uncontroversial as I think it can be—though certainly not 4 Ibid. 5 For some relevant observations, which include further scepticism about the idea of a truth that cannot be expressed linguistically, see Horwich (1990), Ch. 2, n. 4; Quine (1992), pp. 77 ff.; and Wright (1992), pp. 72–73. 6 This last question, of course, takes us back to Davidson: see again Davidson (1984g). And see again the material from Horwich (1990) cited in the previous note. 7 Note that, since it is impossible by this criterion for something to be effable if it lacks content, and since, on some views, necessary truths lack content, then I must either reject these views or tacitly restrict my rejection of ineffable truths to contingencies. (And only if I do the former, of course, can I avoid having

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

212

Language, World, and Limits

completely uncontroversial. It also leaves me free to acknowledge the many important and interesting senses in which there are ineffable truths. Most of us have said something like, ‘Words cannot express how I feel at the moment’, and have known a certain anguish in saying it. I by no means want to challenge our right to say such things, nor to belittle the significance of our doing so. Here I am reminded of a wonderful sentence by Bernard Williams in which he refers to ‘that worthwhile kind of life which human beings lack unless they feel more than they can say, and grasp more than they can explain’.8 Furthermore, my overarching concern in this essay is with the relationship between ineffability and religion, and the sort of ineffability that I do allow has much, clearly, to do with the latter. Even so, I am keen to adopt the least demanding criteria that I can make sense of, because they are the ones that are most suitable for broaching the issues of principle that ultimately interest me.9

2. An outline of another argument for answering the question negatively Having to that extent clarified my rejection of ineffable truths, I now want to proceed to the brief indication promised of how I think this rejection can be justified. ‘Brief indication’ is the operative phrase. I shall just give the barest outline of an argument, partly because I have filled in some of the details elsewhere10 and partly because, as will become apparent, I think the interesting questions about the relationship between ineffability and religion lie elsewhere. The argument has two steps. The first step is to the intermediate conclusion that any truth must be expressible in some way or other. The second step is from there to the conclusion that any truth must be expressible linguistically. First, then: it makes sense to talk about a truth only where it makes sense to posit some corresponding identifiable entity that can be said to be true. Thus suppose we say that there is a truth in the concluding section of the chairman’s report. Then unless we mean simply that one of the sentences in the concluding section is true, we must at the very least be envisaging the possibility of someone’s making a claim that is somehow entailed by the chairman’s report and that is true. There cannot be a truth in the concluding section of the chairman’s report unless some such possibility obtains. Again, suppose there is a truth about human endurance that someone is about to learn. Then it must be possible for that person to come to manifest a belief about human endurance that is true. In general, given any truth, unless it is already of a kind to testify to its own expressibility (as in the case of the sentence in the to concede that, on my conception of the ineffable, there are truths that are both ineffable and capable of being put into words.) 8 Williams (1981a), p. 82. 9 This too I think is reminiscent of Davidson. He faces similar questions about his standards of translatability; and he adopts similarly low standards, low enough to facilitate his tackling the fundamental issues of principle about the coherence of the idea of a conceptual scheme that are his ultimate concern. When he says that there could not be a language whose sentences were untranslatable into any of ours, he is not denying the familiar fact that there are many sentences of French, say, that have no exact equivalents in English. 10 See Moore (1997), esp. Chs. 4 and 8.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

213

concluding section of the chairman’s report), then it must be possible for there to be some corresponding independent true thing. But that possibility just is the possibility of the truth’s being expressed. Any truth must be expressible in some way or other. That brings us to the second step in the argument. The expression of a truth, if not already linguistic, must at least be some kind of representation. That is, it must be something with content, which thereby answers to reality—correctly, of course, inasmuch as it is the expression of a truth. A very primitive organism interacting with its environment and thereby processing data, say a cockroach trying to find its way back to some scrambled eggs that it has recently been tasting,11 would be producing representations in this sense and thus, if it has got things right, would be expressing truths: truths about how its environment is.12 But given any true representation r1, it must be possible, however indirectly, to integrate it with any other true representation r2. That is, it must be possible to produce a third true representation r3 whose content either is or in some sense includes the product of r1’s and r2’s. For r1 and r2 will between them have content, and this content will furnish a truth which, like any other, must be expressible. Now in many cases there will be radical differences of kind and content between r1 and r2; for instance, r1 may have been produced by a cockroach looking for scrambled eggs, and r2 may be the manifestation of a recondite belief about human endurance. More particularly, in many cases there will be such radical differences of kind and content between r1 and r2 that r3 must express the content of at least one of them without the benefit of whatever immersion in that content, or engagement with it, or access to it, its producer enjoyed. (Otherwise r3 may not be able to express the content of the other representation at all. It is not possible, for instance, to express a truth about human endurance from the point of view of a cockroach.) But this in turn will require an abstraction and a sophistication that make it out of the question for r3 to be anything other than linguistic. In such a case, r3 will be, in the undemanding sense that I am adopting, a linguistic expression of the truth originally expressed by r1. But since r1 was just an arbitrary true representation, we can conclude that any truth must be expressible linguistically.

3. Ineffable states of knowledge, and reasons for believing in them I think we must give up the notion of an ineffable truth then. But I do not think we must give up the notion of ineffability per se. Despite the argument that I have just sketched, I am convinced that some things are ineffable. But what things? Clearly, if the word ‘things’ is not to stand proxy for ‘truths’ here— which it does on the most natural interpretation—then something needs to be said 11 This is an allusion to Thomas Nagel’s famous example: see Nagel (1986), p. 25. 12 Here I am gliding over worries about whether there would be enough purposiveness and cognition in such a case for it to be appropriate to talk about the production of representations. But I think I am entitled to do so. For suppose these worries are well-founded. Then, granted the first step of my argument—in other words, granted that where there is no possibility of a representation, there is no truth—we have one less reason to take seriously the idea of a truth that cannot be expressed linguistically. So I am not, by gliding over these worries, begging any questions in my own favour.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

214

Language, World, and Limits

about what it does denote. And this has to be done without reducing the idea that some things are ineffable to either nonsense or triviality. For example, it is either nonsensical or trivial to say that subatomic particles are ineffable. Subatomic particles can be described linguistically. But they cannot be expressed linguistically. There does however seem to me to be an acceptable answer to the question, ‘What things?’, namely, ‘States of knowledge’. The idea that some states of knowledge are ineffable is certainly not either nonsensical or trivial. But it is, in my view, true. In a while I shall try to substantiate this idea. But first I think it is worth signalling another advantage that it has over the idea that there are ineffable truths: one can give examples. One can specify a particular ineffable state of knowledge without in any sense belying its ineffability. Of course, one cannot do this by using the schema ‘A knows that x’, which would require that the knowledge be put into words.13 But one might be able to do it by using some other familiar schema, such as ‘A knows how to x’.14 And certainly one might be able to do it in a less familiar way. By contrast, suppose that one wanted to specify a particular ineffable truth. We might think that one could do this by identifying and describing some non-linguistic expression of that truth. But whatever one proffered, one would straightway be vulnerable to the objection that it would itself, if one’s attempt had been successful, either constitute or be convertible into a linguistic expression of the truth in question.15, 16 Very well, then; what can be said in favour of the view that there are ineffable states of knowledge? Well, there are states that have all the hallmarks of knowledge but that nevertheless lack content; they are not representations; they do not answer to how things are. Hence nothing counts as expressing them linguistically. When I say that these states have all the hallmarks of knowledge I mean basically three things. First, they are enabling states. That is, very roughly, they are states that dispose those who are in them to act on their goals, their aims, their projects, and their wants in ways that tend to lead to the promotion or satisfaction of those goals, aims, projects, and wants. Second, those who are in these states are enabled not only to realize specific 13 But cf. n. 7: I am here prescinding from the possibility that knowledge of necessities counts as knowledge without content. If it does, then there is in fact room for the view that an ineffable state of knowledge—for instance, my knowledge that it is either raining or not raining—can be specified by using the schema ‘A knows that x’. Cf. Wittgenstein (1961), 4.461. 14 This would require that not all states of ‘knowledge how’ are states of ‘knowledge that’: for an argument in favour of the opposite view, which I cannot address here, see Stanley and Williamson (2001). 15 How so? Suppose, for instance, that one attempted to specify an ineffable truth as the (complete) truth expressed by a given piece of music. How would that be convertible into a linguistic expression of the truth in question? Is the thought that, by framing some suitable linguistic description of the music and of the conditions that enable it to express this truth, one could oneself express the truth by saying something of the form, ‘Things are how such and such music, in such and such conditions, expresses them as being’? If so, then does that not simply beg the question of whether one could frame a suitable linguistic description of either the music or the conditions in the first place?—Well, the thought need not be that. The undemanding criteria of effability that I have adopted would allow one to express the truth by using demonstratives and saying, ‘Things are how this music, in these conditions, expresses them as being.’ (This raises myriad further questions, to be sure. But I shall say no more about the matter now. I am grateful to Mark Sacks for forcing me to say even as much as I have.) 16 Supplementary note for the reprint: I have subsequently become dissatisfied both with the previous note and with the paragraph in the main text to which it attaches. I now think that what I say there is undermined by some of the considerations in Essay 11 in this volume, section 3 (and no doubt this is what the late Mark Sacks was getting at). However, this has no bearing on the rest of this essay.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

215

possibilities but to negotiate possibilities and to adapt their activities to an indefinite range of goals, aims, projects, and wants—and other contingencies. Third, these states have a place in ‘the logical space of reasons’.17 The states I have in mind can nearly all be classified as states of understanding, the concept of understanding being a very broad and versatile one. All sorts of things can be objects of understanding: languages; specific linguistic items such as words or sentences; theories; data; works of art; people. But wherever it occurs and whatever its object, understanding helps us to organize our knowledge into something coherent, manageable, and unified. It is itself a kind of knowledge. It is knowledge of how to process knowledge. However, in the cases I have in mind, it is not knowledge that answers to anything. A state of understanding of the sort I have in mind is an enabling state, just like any other state of knowledge; and in particular, it enables whoever is in it to make sense of things. But that is not to say that it enables whoever is in it to make the ‘right’ sense of things. There is no question of right or wrong here.18 A simple example is my knowing how to exercise the concept of greenness: my knowing what it is for something to be green. This does not consist in my knowing that anything is the case. It is rather a matter of my having the wherewithal to know that various things are the case. For instance, it enables me to know that the leaf I am looking at is green. But it does not itself consist in my knowing that anything is the case because it does not answer to how the world is. Even if the world were different in some way, provided I were still in this state, it would count as a state of knowledge. I would still know what it is for something to be green. At worst, the knowledge would be of less use to me, say because I had little or no interaction with green things. (This was the sort of thing I had in mind when I said just now that there was no question of right or wrong here. It is neither right nor wrong for me to make sense of things in terms of whether or not they are green.) By contrast, my knowledge that the leaf I am looking at is green does answer to how the world is. If the leaf were brown, though I were for some reason still in this state,19 then my state would no longer be a state of knowledge. It would be a state of mistaken belief about what colour the leaf was. I would be wrong about the leaf. Another example of my ineffable knowledge is my knowing how to act out my autobiography: my being able to make narrative sense of my own life. Even if I do this in a way that renders my life desolate or ugly in various ways, I cannot thereby be said to have misrepresented anything. It is much like my knowing how to exercise concepts. Indeed a very important part of it is my knowing how to exercise concepts. Making narrative sense of my life is to a very significant extent a matter of determining what concepts I shall live with: whether or not, for example, to make use of the concept of chivalry, or that of blasphemy, or that of sin.20 17 This phrase is taken from Sellars (1963), p. 169, where he writes, ‘In characterizing . . . a state as that of knowing . . . we are placing it in the logical space of reasons’ (his emphasis). 18 The ideas in this paragraph are spelt out in much greater detail in Moore (1997), Ch. 7, §4 and Ch. 8. Cf. also Wittgenstein on the ‘autonomy of grammar’: see Hacker (1986), Ch. 7, §2. 19 States are here being construed as ‘narrowly’ dispositional. For discussion of some of the issues that this raises see Pettit and McDowell (1986), Introduction. 20 These are all examples of what Bernard Williams has famously called ‘thick’ ethical concepts: see e.g. Williams (1985), pp. 140 ff. Thick ethical concepts are both ‘action-guiding’ and ‘world-guided’. To live with

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

216

Language, World, and Limits

Not that this leads as directly to the link between ineffability and religion that concerns me as it seems to. It is certainly true that a central example of what I was just talking about is determining whether or not to make use of religious concepts, the concept of blasphemy and the concept of sin being themselves cases in point. But this link between ineffability and religion is not all that distinctive. There is little of note to be said about it, in this context, that cannot just as well be said about the link between ineffability and astrology, or ineffability and sport. The link between ineffability and religion that concerns me, on the other hand, is much more peculiarly a feature of religion. And it emerges, as I shall now try to indicate, in a much more oblique way.

4. The attempt to put some of these ineffable states of knowledge into words There are some ineffable states of knowledge such that we can say what would result from attempting (unsuccessfully, of course) to put them into words. This is a thesis that I have been at pains to defend elsewhere,21 and I shall not now attempt a full rehearsal of that defence. But some of its key features are relevant and worth reiterating. First, we can say what would result from attempting to put these ineffable states of knowledge into words only because there is a shared temptation to treat the states of knowledge as though they were effable. To attempt to put them into words is to succumb to the temptation, or at least to affect to succumb to it, and there is enough uniformity in the temptation to sustain standards of ‘correctness’ for doing so.22 Now the temptation to treat ineffable states of knowledge as though they were effable is of a piece with the urge that we humans have to transcend some of our most basic limitations. For the very fact that we have such ineffable knowledge, which enables us to make sense of things, is an indication of our radical finitude: it is an indication of the fact that there are things of which we need to make sense, things that are not of our own making. (An infinite being, for whom there existed nothing that was not of its own making—nothing whose every feature it had not itself determined—would have no need to make sense of things in order to know them. Its creating them would be its knowing them.23) This means that the attempt to put such knowledge into words will often involve a distinctive interplay between our awareness

one of them—or better, perhaps, to live by one of them—is to have a certain outlook, to value certain things, to invest certain things with importance, to have certain reasons for doing things, and to have certain reasons for refraining from doing things. Many of the thick ethical concepts by which we live we do so simply in virtue of our participation in a particular social world. But individuals are able to reflect on the thick ethical concepts by which they live and, in some cases, if they see fit, to rid themselves of those concepts in favour of others. 21 Moore (1997), Chs. 9–11. 22 It follows that if ever we are able to identify language of a certain kind as what results from attempting to put ineffable knowledge into words, then it will be language of a kind that people have an inclination to produce; and it may well be language of a kind that they very often do produce, perhaps under the illusion—perhaps not—that they are straightforwardly saying how things are. 23 Cf. Kant (1998), B72 and B138–139; and Kant (2000) §§76–77.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

217

of the basic limitations in question and a pretence either that they do not exist or that they are somehow other than they are. How, then, does all of this bear on the link between ineffability and religion that concerns me? Simply in that the language that results from attempting to put our ineffable knowledge into words is very often of a ‘religious’ kind (where the scare-quotes signal that I mean no more than that the language has certain resonances and associations that would naturally be classified as religious). ‘Very often’ is an important qualifier. I do not claim that attempts to put our ineffable knowledge into words result only in such language; nor, for that matter, that such language results only from attempts to put our ineffable knowledge into words. It would be a serious depreciation of both ineffability and religion to think that there was complete overlap here. But there is significant overlap, both in the sense that there is overlap that counts for much and in the sense that there is much of it. In the remainder of this essay I shall try to sketch why I think this is so.

5. Three areas of overlap between the language that results from such attempts and religious language There are three broad areas of overlap. First, because the attempt to put our ineffable knowledge into words involves prescinding from our limitations, there is an area in which we make play with images of unlimitedness and infinitude. These we are tempted to apply in the first instance to ourselves. We are tempted to regard ourselves as standing in some sort of relation of creativity to the objects of our knowledge. Where we are creative, in the formation and the free exercise of concepts, we are tempted to regard ourselves as creating what we know, not just creating the wherewithal to make sense of what we know. But then, with re-awakened self-consciousness about the very limitations from which we have been prescinding, we naturally recoil from describing ourselves in such terms. So we end up making play with those same images of unlimitedness and infinitude, but applying them now to a reality beyond ourselves. Our original temptation to say that we are thus and so becomes a temptation to say that there is that which is thus and so. Our original temptation to say that we create and sustain all the objects of our knowledge becomes a temptation to say that all the objects of our knowledge have a creator and a sustainer; that there is that which determines their every feature. Eventually it becomes a temptation to talk of God. Secondly, there is an area in which we talk of conceptual necessities as though they admit of alternatives. This is because attempting to put our ineffable knowledge into words means treating it as though it had content; which in turn means treating it as though there were something independent of it to which it answered; which in turn means treating our grasp of the sense that we make of things as though it were the grasp of what is there anyway, indeed as though it were the grasp of some cosmic contingency that we had come up against; which means, finally, treating the conceptual necessities that we acknowledge, in virtue of having this grasp, as though they themselves were contingencies—as though they admitted of alternatives. Aware,

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

218

Language, World, and Limits

however, that such conceptual necessities are not contingencies, we are inclined also to talk of these alternatives as ‘transcendent’ alternatives, in the hope that this will somehow preserve the necessity. An integral part of this entire process is that we fasten on the deep contingencies that do in fact underpin our ineffable knowledge (above all, the various relevant constancies of nature, including the various relevant constancies of human nature, that enable us to make sense of things in the way we do) and treat them as though they were part of the cosmic contingency of which our ineffable knowledge gives us a grasp. Thirdly, there is an area in which we endow those very same contingencies with a kind of necessity. This is for two reasons. The first reason is (once again) to respect the necessity of the conceptual necessities that those contingencies underpin, the conceptual necessities that we acknowledge in making sense of things in the way we do. The second reason is to safeguard our capacity to make sense of things at all. The second and third areas might almost be assimilated. In each case we are trying to transcend the framework within which we make sense of everything. And in each case we are aware of the need, at the same time, to be able to make sense of it. What result are two different aspects of the fact that we cannot ultimately combine these feats. We end up trying to make sense of the framework in incompatible ways: on the one hand, as a framework that we are in the process of transcending, a framework within which we merely happen to make sense of everything, a framework that admits of alternatives; and on the other hand, as a framework that we know we cannot transcend, a framework within which we do indeed make sense of everything, a framework that admits of no alternatives. In the second area we try to resolve this tension by construing the framework as a framework within which we make sense merely of everything that is non-transcendent. In the third area we try to resolve the tension by acquiescing in the thought that the framework is something that is beyond our horizon altogether, something for which the question of what alternatives it admits of is a question that we cannot so much as broach. This means that, although there is certainly a sense in which the second and third areas are assimilable, there is also a sense in which each is the obverse of the other. In the second area we treat what is necessary as though it were contingent. In the third area we treat what is contingent as though it were necessary. In both areas, had we not been involved in an attempt to transcend our framework—had we not been involved in an attempt to express the inexpressible— we could have unproblematically affirmed the necessity of what is necessary while at the same time acknowledging the contingencies that allow us to have the concepts to think in such terms. For example, we could have unproblematically affirmed that whatever is green must be coloured, while at the same time acknowledging that we would not have had the concept of greenness at all had we not had, say, the faculty of sight. The first area of overlap is the most self-explanatory. It is as if we have an aspiration to be God, which, under pressure of our self-conscious awareness of the absurdity of such an idea, survives as an aspiration to come together in union with God, and thus to acknowledge and to affirm God. Wittgenstein writes of ‘two godheads: the world and my independent I’.24 The aspiration is to see these two—God and the self—come 24 Wittgenstein (1979), p. 74.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

219

together in union. Attempting to put our ineffable knowledge into words involves talking as though they do. It involves talking as though God exists. There is clearly much more to be said about this. But I shall reserve further comment for the very end of my essay, in the light of discussion of the second and third areas of overlap. To return to the second area, then: let us reconsider the necessity that whatever is green is coloured. It is a shared feature of our concept of greenness and our concept of colour that the former contains the latter. This is reflected in the fact that we use the sentence ‘Whatever is green is coloured’ to express a rule. The rule is that we are not to count something as green unless we also count that thing as coloured. That we have such a rule—that we have the concept of greenness at all—is a contingency, sustained by all sorts of other contingencies. As I commented above, we would not have had the concept of greenness had we not had the faculty of sight. When we reflect on our ineffable knowledge of what it is for something to be green, and attempt to put that knowledge into words, we are liable to see just such contingency in what the knowledge answers to. We are liable to see the link between being green and being coloured as a component in some brute structure that has somehow been imposed on the world, and that might just as easily not have been imposed on it. This in turn will suggest other possibilities: possibilities in which this link does not obtain; possibilities in which there are colourless green things. But these cannot be ‘real’ possibilities. It is after all necessary that whatever is green is coloured. And this is why, as I intimated above, they will appear as ‘transcendent’ possibilities. That is, they will appear as possibilities that go beyond all that we can make sense of. And all that we can make sense of will in turn appear to have the structure it has because of whatever, at that ‘transcendent’ level, prevents these other possibilities from being realized—whether we cast this as Divine fiat,25 our own fiat somehow devolved to us from on high, or something else entirely.26 Attempting to put our ineffable knowledge into words involves talking as though there is a supernatural reality that both surpasses and shapes all that we can make sense of then. What follows? Are we also liable to say that, for all we know, we shall some day discard this link between being green and being coloured, in acknowledgement of colourless green things? No. Precisely what ‘transcendent’ possibilities are supposed to be are possibilities that we can never reckon with; and, as far as ‘real’ possibilities are concerned, there is no question of our ever having to acknowledge colourless green things. Even so, we are liable to seek insurance against those contingencies which, though they do not allow for the ‘real’ possibility of our some day discarding the link between being green and being coloured because it no longer obtains, do allow for the ‘real’ possibility of our some day discarding the link between being green and being coloured because we are no longer able even to think in such terms. For it is indeed a ‘real’ possibility, familiar to anyone who has considered the problem of induction, that hitherto recurring patterns of nature should break down in such a way that, even if we are still here, and even if we are still able in some sense to function, we shall have 25 This is what Descartes is often reckoned to have done: but for an important corrective, itself pertinent to the current discussion, see Bennett (1994). 26 Cf. Moore (1997), Ch. 9, §4, esp. p. 212.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

220

Language, World, and Limits

completely lost whatever it takes to know how to exercise the concept of greenness. Naturally we hope that this will not happen. But this is only a hope. And even if this ‘real’ possibility, the possibility that we shall cease to be able to make chromatic sense of our environment, is of purely theoretical interest, the possibility that we shall cease to be able to make good narrative sense of our lives is one to which we are constantly and disturbingly exposed. Some of the concepts that we use to tell our stories rest on frail historical and social contingencies that we shall certainly outlast. Our hope is that we shall not also outlast all the contingencies that allow us to replace them. But it is possible that we shall. In attempting to put our ineffable knowledge into words, and in reckoning at the same time with a distinction between ‘transcendent’ possibilities and ‘real’ possibilities, we are liable to banish those threatening possibilities to the realm of the ‘transcendent’. We are liable to take comfort in some ‘transcendent’ safeguard (as it may be, Divine providence) that we shall always be able to find point in what we do and what befalls us. And this in turn brings us back to the third area of overlap. In seeking security, not only against the break-down of the conceptual necessities that we acknowledge in telling our stories, but also against the break-down of our capacity to tell any stories at all, we are liable to talk as though the contingencies that enable us to carry on doing so are in some sense necessary, just as the contingencies that enable the earth to carry on revolving on its axis are in some sense necessary: utterly steadfast, utterly to be relied upon. There is another aspect of this. These contingencies include our valuing certain things and our regarding certain things as important.27 One way in which we are liable to cast these particular contingencies as necessary is by counting ourselves as having come to recognize and to value what is valuable, and as having come to recognize and to regard as important what is important, where being valuable and being important are themselves part of the fabric of the world. In particular, we are liable to count infinitude as intrinsically valuable. For infinitude, in the form of freedom from various limitations, is basic among the things that we value. In fact it is precisely because we value it that we are tempted to treat our ineffable states of knowledge as though they were effable in the first place. To succumb to that temptation is to be motivated by the lure of infinitude. There is a sense, then, in which we are doubly liable to treat infinitude as something more than what we (merely, effably) value; as something that is, simply, valuable.28 Clearly, at work in all of this, are certain Gestalt switches. What we see now under the aspect of necessity we see now under the aspect of contingency; what we see now under the aspect of unlimitedness we see now under the aspect of limitation. When we are trying to put into words what cannot be put into words, the world shimmers. Our efforts will be judged, in part, by how well our use of language enables us to reproduce that effect. If we can make play with words in a way that reproduces it very well—if we can frame a sentence that thereby counts as the result of attempting (unsuccessfully) to put some ineffable state of knowledge into words—still, of course, this will differ in all sorts of important ways from framing a sentence that counts as the result of attempting (successfully) to put some effable state of knowledge into words. In particular, in the former case, it is possible—more than possible—that the 27 Cf. the discussion of thick ethical concepts in n. 20.

28 Cf. Moore (1997), p. 270.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

221

negation of the sentence would have served just as well. For whatever features of the original sentence allow it to be put to this effect (the images it conjures up, the associations it has, the lines of thought it suggests), negating it may very well leave them intact. Thus consider the following famous sentence from Wittgenstein’s Tractatus: The world of the happy man is a different one from that of the unhappy man.29

Robert Fogelin, commenting sceptically on this sentence, remarks, ‘A competing sage might say that the world of the happy man is no different from that of the unhappy man (and this too has a ring of profundity)’.30 In fact, however, if Wittgenstein’s sentence is the result of an attempt to express the inexpressible, then any unclarity about what makes it better suited to this rôle than its denial, indeed any intimation that it is no better suited to this rôle than its denial, need occasion neither suspicion nor surprise. Who knows but that both Wittgenstein’s sentence and Fogelin’s reversal of it, each in a suitable context, may be apt to achieve the same broad effect; or at least, that they may be apt to achieve complementary effects?31 Certainly they have something important and relevant in common which neither shares with, say, ‘The world of the happy man is a funnier one than that of the unhappy man’ or ‘The brain of the happy man is a different one from that of the unhappy man’.32 One thing that this serves very forcibly to remind us is that, when we say of some sentence that it counts as the result of attempting to put some ineffable state of knowledge into words, we are not thereby saying anything about how that sentence should be assessed if proffered as a simple expression of the truth. It is clear that we are not saying anything to vindicate such a use: we are not giving any reason to think that the sentence is true. But it is equally clear, and it is equally important, that we are not saying anything to discredit such a use: we are not giving any reason to doubt that the sentence is true. This is relevant to the question of how deflationary my project in this essay is. While the project is certainly compatible with the view that standard uses of religious language, in assertoric mode, are either false or nonsensical, it is a further question whether it encourages such a view. I insisted earlier that religious language does not result only from attempts to put ineffable knowledge into words. Much depends, obviously, on what other impulses to the use of religious language there are, and how far these can be assimilated to that which I have been describing. But that is by no means the only relevant issue. There is also the issue of whether various uses of religious language could survive the self-conscious recognition that they were nothing but unsuccessful attempts to express the inexpressible. It is not at all clear to me that they could not. In general, I would want to downplay any suggestion that I was involved in a ‘debunking’ exercise.

29 Wittgenstein (1961), 6.43. 30 Fogelin (1987), p. 103. 31 Fogelin himself subsequently says of the two sentences, ‘As denials of each other, their meaninglessness should show the same thing’ (Fogelin (1987), p. 103). (He intends this as a criticism, however. He goes on to complain: ‘Wittgenstein’s preference for . . . [one over the other] is wholly arbitrary and has no place in the Tractatus’ (Fogelin (1987), p. 103.) 32 Consider in this connection the creative use of contradiction in mystical and religious writing. Examples abound. They can be found in the writings of Plato, the Psalmists, Lao Tze, Nicholas of Cusa, Kierkegaard, and countless others.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

222

Language, World, and Limits

Let us return to the idea that attempting to put ineffable states of knowledge into words can involve making play with contradictory ideas. I want to close with what seems to me a particularly significant example of this, an example that illustrates all three areas of overlap. I have in mind the Euthyphro question, raised by Socrates in Plato’s dialogue of that name: whether what is holy is holy because the gods approve of it, or whether the gods approve of it because it is holy;33 or, in a version with greater resonance to Christian ears, whether what is right is right because God wills it, or whether God wills it because it is right. As I have already observed, there is a temptation, in the third area of overlap, to treat some of what we contingently value as intrinsically valuable; as being the right thing for us to value. A Gestalt switch then takes us back to the second area of overlap, where we acknowledge ‘transcendent’ alternatives: ‘transcendent’ possibilities in which different, incompatible things are the right things for us to value. Running these together, we are liable to see God’s ‘transcendent’ will as what prevents these ‘transcendent’ possibilities from being realized. What is right, we are liable to say, is right because God wills it. But we shall also want to do justice to the ‘transcendence’ of the alternatives. Our original urge was to say that ‘really’ nothing else is to be willed. So we are also liable to say that God wills what is right because it is right. It is like an antinomy. Both sides of the debate alternately urge themselves upon us. Moreover, whichever view about the Euthyphro question we are inclined to adopt at any given moment, there remains the original link between what we are prone to regard as intrinsically valuable and what we ourselves happen to value. Thus whether we are more inclined to say that what is right is right because God wills it, or to say that God wills it because it is right, it will still be the case that what is right—by our own reckoning—impinges on us primarily as what we will. What is right; what God wills; what we will: these we shall regard as one and the same. This takes us back finally to the first area of overlap, in which we see ourselves as one with God. But of course, everything that we are liable to say here is tempered by our selfconscious awareness of the many opposing things that can, quite properly, be said: most glaringly, and most pertinently, that we are not one with God. Ultimately all this talk is destined to crumble and disintegrate. Of course it is. How can there be ultimate satisfaction where attempting to do the impossible is concerned? To say that some bit of language results from the attempt to express some inexpressible state of knowledge is by no means to say that producing that bit of language does justice to everything that fuels the attempt. It cannot. To be engaged in the attempt is to be engaged in a restless and hopeless endeavour in which not even inconsistency can bring peace. I think of Kierkegaard’s desperate and moving struggle to come to terms with the Euthyphro question, itself a struggle (I would argue) to come to terms with something ineffable. Kierkegaard wrestles hard to understand how Abraham can be willing to sacrifice Isaac at God’s command. He writes: It is supposed to be difficult to understand Hegel, but to understand Abraham is a trifle. To go beyond Hegel is a miracle, but to get beyond Abraham is the easiest thing of all. I for my part have devoted a good deal of time to the understanding of the Hegelian philosophy, I believe also that I understand it tolerably well, but when in spite of the trouble I have taken there are 33 Plato (1961a), 10a.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Ineffability and Religion

223

certain passages I cannot understand, I am foolhardy enough to think that he himself has not been quite clear. All this I do easily and naturally, my head does not suffer from it. But on the other hand when I have to think of Abraham, I am as though annihilated.34

The use of ‘religious’ language to attempt to put into words what cannot be put into words relates to that sense of annihilation. But it neither refers to it nor even expresses it. It compounds it.35, 36

34 Kierkegaard (1954), pp. 43–44. 35 Since writing this essay I have come across McPherson (1955). McPherson’s essay connects with my own in a number of interesting ways. I am sceptical, however, about his distinction, on which he places so much emphasis, between different kinds of nonsense (e.g. pp. 133 and 141): see further Moore (1997), pp. 198–200. 36 I am extremely grateful to Mark Sacks for comments that helped me to improve this essay.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

17 On Saying and Showing Abstract This essay constitutes an attempt to probe the very idea of a saying/showing distinction of the kind that Wittgenstein advances in the Tractatus—to say what such a distinction consists in, to say what philosophical work it has to do, and to say how we might be justified in drawing such a distinction. Towards the end of the essay the discussion is related to Wittgenstein’s later work. It is argued that we can profitably see this work in such a way that a saying/showing distinction arises there too. In particular, in the final sub-section of the essay, it is suggested that we can see in Wittgenstein’s later work an inducement to say what we are shown. There is not, and may there never be, any treatise by me . . . on these things, for the subject is not communicable in words, as other sciences are. Rather is it that, after long association in the business itself and a shared life, a light is lit in the soul, kindled, as it were, by a leaping flame, and thenceforward feeds itself (Plato).

1. The saying/showing distinction in the Tractatus. Two examples A linchpin of Wittgenstein’s Tractatus1 is his distinction between what can be said and what can be shown. The distinction first seriously emerges in the early 4s in connection with the idea that a proposition shows, but does not say, what its sense is.2 What can be shown, we are told, cannot be said.3 By the end of the work the catalogue of what can be shown has become large and multifarious. It includes: the logical form of reality;4 the logical relations between propositions;5 the limit of empirical reality;6 the truth in solipsism, or what the solipsist means;7 and the mystical.8 Recent work bearing on the distinction, and sympathetic to it, has pointed to an even broader conception of what can be shown. For example, it has been suggested that one lesson of Wittgenstein’s later work, where the terrain is very different from that of the Tractatus and where the distinction is certainly not explicitly retained, is none the less that there are certain things which, though they cannot properly be put into words, can be 1 Wittgenstein (1961). All unaccompanied references will be to this. 2 4.022ff. But the first allusion to the distinction is in 2.171–2.172. 3 4.1212. 5 4.1211, 6.1201, and 6.1221. 6 5.5561. 7 5.62. 8 6.522.

4 4.121.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

225

shown.9 My aim in this essay is to probe the very idea of a saying/showing distinction along Tractarian lines. I shall attempt to say what it consists in, what philosophical work it has to do, and how one might be justified in drawing such a distinction. Consider the two following examples. The first is inspired by the Tractatus itself.10 It concerns somebody’s visual field, where this is to be thought of as a three-dimensional portion of public space. A more or less complete description is produced of what is in this field, from the point of view of the person, though with no explicit reference to anything outside the field. However complete the description may be, it cannot represent the fact that everything of which it treats is seen from a particular point at the edge of the field. For this is a fact that cannot be represented without explicit reference to something outside the field. Yet there is a sense in which this fact will be manifest in the form that the description takes. (The description will use terms like ‘left’ and ‘right’.) So there is something that cannot be said here but that is shown by what can be said. The second example11 concerns a limited language with a co-ordinate system by means of which things are located on the surface of the earth. Every sentence from the language is of the form ‘X is at 〈i, j〉’. It is clear that this language will be fully comprehensible only to somebody who can identify some point on the grid. Yet none of its sentences can be used to say where any point on the grid is. At most they can be used to say what is at any given point. What cannot be said here must again be in some sense manifest to, or shown to, anybody with a full understanding of what can be said.12 Is this the kind of thing to which Wittgenstein is alluding in the Tractatus? In at least one crucial respect, it is not. Wittgenstein wants to claim that there are things that can be shown though they cannot be said in any possible language, whereas each of the examples furnishes us with a notion of unsayability that is relative to certain limited linguistic resources. In each case what is shown can also be said, given appropriate enriching of those resources. The examples nevertheless furnish a useful model of Wittgenstein’s distinction, and they provide some clues as to what unifying theme might link his inventory of what can be shown. In each of the examples what is showable is some feature of a framework, and what is sayable is how that framework is filled. The frameworks are the outer limits of the visual field and the co-ordinate system respectively. Each of these determines a certain bounded whole, a space, and it is the parts of this space which are described whenever anything is said (within the prescribed linguistic limits). But nothing that can be said (within those limits) serves to describe the whole itself, as a whole—as part of some larger space. By contrast, what can be shown does concern the whole as a whole, related to something outside it, namely a subject. In the first instance it is shown that the contents of the whole are

9 See e.g. Williams (1981c), p. 163 and Lear (1984), p. 242. I try to develop this line of thought in Essay 5 in this volume. 10 See 5.633–5.6331. 11 This example is derived from an unpublished lecture course given by Ron Williams at Colorado State University. 12 I believe that there are also some useful examples to be gleaned from mathematics: see e.g. Moore (1985).

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

226

Language, World, and Limits

seen in a certain way by the subject. In the second instance it is shown that the contents of the whole are to be located in a certain way by the subject. Now the contingencies of any relationship between the whole and what lies outside it are irrelevant to the description of its parts. For the purpose of such a description, the whole, or rather its framework, counts as rigid and fixed. So, relative to that description, what can be said is how things are within the framework and what can be shown, bearing as it does on the framework itself, explains how they must be. The distinction between what can be said and what can be shown therefore relates in a fundamental way to the familiar philosophical distinction between what is actually the case and what is necessarily the case. This is indeed reflected in the Tractatus, where it is claimed that the only necessity is logical necessity.13 For elsewhere it is clear that Wittgenstein intends an intimate connection between logic (logical form and structure) and what can be shown.14 In general, this model does give us some sense of how the various doctrines advanced in the Tractatus concerning the saying/showing distinction might be held together. The relevant whole in the Tractatus is the world, all that is the case. Its shape and its form, when it is conceived as a whole (a limited whole), are what can be shown. And through the ethical and mystical elements of the work these are related to the metaphysical subject, which lies outside the world, as its limit.

2. Some preliminaries If these ideas are to be sharpened, however, something still needs to be said about the special way in which the word ‘show’ is being used and, more particularly, about the grammar of the saying/showing distinction. Otherwise the claim that certain things can be shown though they cannot be said will be too easily understood as a triviality that falls far short of Wittgenstein’s thesis. (I can show you a chair but I cannot say it.) It is not just that ‘things’ is too imprecise. The surface grammar is misleading. Any narrower sortal that does not issue in an interpretation that makes the claim trivial in this way will issue in an interpretation that would make it unacceptable to Wittgenstein—for example, that some of the ways in which things stand in relation to one another can be shown but not said, or, more generally, that certain truths can be shown but not said.15 Concomitant with the problems that arise here is a question that has received surprisingly little attention and with respect to which the Tractatus offers next to no help: what is the genus of which saying and showing are species? Of course, there is a sense in which this is a bad question, born precisely of an overly crude construal of the claim that certain things can be shown but not said. We are less inclined to pose the question when we return to the text. For although Wittgenstein speaks of propositions as both saying and showing certain things, he also suggests that the real contrast is between what we (language-users) say by means of propositions and what shows itself, or makes itself manifest.16 Still, to the extent that a division is being made here, we are bound to want some sense of the territory that is being divided. 13 6.37. 14 E.g., 4.121. 5.634 is also relevant here. 15 See, e.g., 4.063. 16 E.g., 4.121 and 6.124. Max Black makes this point in Black (1964), p. 190.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

227

In the light of this, it is appropriate (though not particularly Tractarian) to focus on individual states of enlightenment—to see the distinction as being, primarily, between knowing something that can be said (as it were, having the answer to some question17) and being shown something, where this is construed as being in a state of enlightenment that is not of that kind. The claim that certain things can be shown though they cannot be said is then tantamount to the claim that we can be in a state of enlightenment though we are unable to give voice to it (we are unable to put it into words). If we view matters in this way, then we are no longer likely to be vexed by the question of what the common object of saying and showing is or by the question of what their genus is. We nevertheless do have something to say about what are being distinguished, namely states of enlightenment. Moreover, by turning our attention away from what is shown to those who are shown it, we help to allay the natural worry that much of what we shall want to say in this connection is precisely what, according to the doctrine being discussed, cannot be said. For whatever we can or cannot say about what is shown, we are certainly able to say plenty about people and about what it is for them to be in states of enlightenment. Thus we may be able to say a lot about showing without prejudicing any issues, and in particular without abrogating the very concept under discussion. Were this not the case, there would be legitimate room for suspicion about whether it was a bona fide concept at all.

3. Basic justification for drawing such a distinction Let us focus, then, on ourselves as knowing subjects. The following are some theses about us in that capacity that would be regarded by many as axiomatic: we are cast into a world that is not of our own making (a world that is independent of us); it is this world of which we have knowledge (if we have any knowledge at all); and, because our epistemic grasp thereby depends upon, and is limited by, our particular standing in the world, we are, in a profound epistemological sense, finite. I shall argue that these theses provide crucial motivation for the drawing of a saying/showing distinction. (My aim is not to defend them but to see where they lead.) One repercussion that they are often thought to have is that we can have no knowledge unless some part of the world into which we are cast, that part which we are to know, is given to us in some way, or (less metaphorically) affects us in some way. This is a distinctively Kantian thought. It occurs at the beginning of his Critique of Pure Reason18 and has a primordial rôle to play throughout his critical philosophy.19 Heidegger, commenting on this, recasts the thought in his own idiosyncratic way: ‘because our Dasein is finite—existing in the midst of the essent which already is and to which our Dasein is abandoned— . . . it [our Dasein] must of necessity receive the essent’.20 Since what we receive thereby becomes ours, a corollary of this thought is 17 See 6.5. 18 Kant (1933), A19/B33. 19 Cf. also Wittgenstein (1979), p. 74, where the idea that I am given the world is expressed explicitly and linked in an interesting way to the will. See McGinn (1983), pp. 101–104, for one attempt to expound the idea. 20 Heidegger (1962b), p. 31, emphasis added.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

228

Language, World, and Limits

that what we know is thereby ours. As such, it stands in an essential relation to us. It is ‘our world’. This idea receives one of its clearest formulations in Schopenhauer, working very much within the Kantian tradition. It is likewise introduced at the beginning of his most important book, as a fundamental principle standing in need of no proof. He writes, ‘Everything that exists for knowledge, and hence the whole of this world, is only object in relation to the subject, perception of the perceiver, in a word, representation’.21 Anything of such elemental importance to Schopenhauer is bound to be relevant to this essay, given the influence that Schopenhauer clearly exercises on Wittgenstein in the writing of the Tractatus.22 Indeed a Tractarian theme is already present here: ‘the world is my world’.23 But the important idea at which this train of thought has arrived is that our knowledge must always somehow relate back to us, via our reception of what we are given. Thus, still in the Kantian tradition though this time in Fichte, we find the bold claim that ‘with respect to . . . [external] things, we know only what is produced through our consciousness itself ’.24 Now given the point of departure for this train of thought, it might appear that a reductio ad absurdum has been effected. Indeed we do find Fichte arguing on this basis that our knowledge is of a reality posited by us, a conclusion which, at least on its most narrowly literal interpretation, stands in direct conflict with the original premise that our knowledge is of an independent reality not of our own making. But there is a simple way of resisting the reductio. Let us distinguish between what we are given and our reception of it. Our knowledge is grounded, immediately, in the latter, which is clearly dependent on us. But it also thereby relates, mediately, to the former, which is not dependent on us. It is this lack of dependence that was originally being maintained. Does this involve a dualism between content and scheme, or between what is given and how it is received, or again between uninterpreted data and their interpretation? This is an urgent question, given the severe censure to which dualisms of this kind have recently been subjected.25 On one construal some such dualism certainly is involved—but not on the construal that would make it objectionable. What is objectionable, and what has been objected to, is the thought that, within our reception of what we are given, we can recognize a sharp dichotomy between a passive element, whereby we first let the world impinge on us, and an active element, whereby we go on to make sense of what we have already thereby received. Donald Davidson has given voice to one popular version of this thought: ‘Conceptual schemes . . . are ways of organizing experience; they are systems of categories that give form to the data of sensation; they are points of view from which individuals, cultures, or periods survey the passing scene’.26 ‘Experience’; ‘the data of sensation’; ‘the passing scene’: these all already incorporate the idea of reception. By contrast, the more fundamental dualism involved in the picture above is precisely between our reception and what we receive (as it is in itself). But as well as being more fundamental, it is also more innocuous. One can accept this dualism without thinking that the world has to be understood apart from some 21 Schopenhauer (1969), p. 3. 22 See Janik and Toulmin (1973) and Malcolm (1958), p. 5. 23 Witness 5.62. 24 Fichte (1956), p. 74. 25 See e.g. Davidson (1984g). 26 Ibid. p. 183.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

229

way of receiving it. ‘The world’ here simply plays the same kind of schematic rôle as the phrase ‘how things are’. Richard Rorty has complained that ‘the notion of “the world” as used in a phrase like “different conceptual schemes carve up the world differently” must be the notion of something completely unspecified and unspecifiable’.27 But why must it? Surely we can moot the possibility that different conceptual schemes carve up the world differently and thereby mean that they differ with respect to how they treat physical objects, people, events, and the like. We can even moot the possibility that they do this without using the concepts of a physical object or a person or an event. Any distinction between scheme and content in the picture above is thus an anodyne one. But it does mean that there are two facets to the knowledge that we have. On the one hand there is that in our knowledge which is determined by what we are given. On the other hand there is that in our knowledge which is determined by our own receptive capacities, which enable us to take whatever we are given. To the extent that we can appropriately focus attention on these severally, we can acquire different kinds of enlightenment. To acquire the first kind of enlightenment we must exercise suitable sensitivity to what lies outside us. To acquire the second kind of enlightenment we must indulge in a kind of introspective, a priori reflection of our own ability to exercise such sensitivity. In the first case we actually receive. In the second case we are self-consciously receptive. This is certainly the kind of distinction being sought. For one thing it dovetails neatly with the idea that emerged from the two examples considered in section 1 of a framework and its filling, together with the attendant idea of a subject vis-à-vis a whole. The whole in this case is ‘the subject’s world’, that which he or she receives. But still, it might be objected, what justification is there for relating this back to the saying/showing distinction? In Kant, where there are these two kinds of enlightenment, is there not also the possibility of giving voice to both? One can say, for example, that 7+5=12 with no less right than that Jesus had twelve apostles. This objection mislocates the distinction in Kant. Certainly one can say that 7+5=12, and much else about how the phenomenal world must be in contrast to how it actually is. To say such things is to express a priori knowledge. But it is not yet to say anything about us or our receptive capacities as things in themselves. The distinction between a priori knowledge and a posteriori knowledge is not the relevant distinction. There is, in Kant, a deeper contrast between two kinds of enlightenment, one of which does take the form of a sort of introspective awareness of our existence as things in themselves—though not, of course, knowledge of ourselves as we are in ourselves. Kant writes, ‘In the synthetic original unity of apperception, I am conscious of myself, not as I appear to myself, nor as I am in myself, but only that I am. This representation is a thought, not an intuition.’28 Whether in Kant’s view there is any legitimate way of giving voice to this thought (this kind of enlightenment) is not clear, and it would be rash to claim that this takes us directly back to the saying/showing distinction. For one thing, Kant has already tried to give voice to it in this quotation. In general when he distinguishes between thought and knowledge, he allows for the possibility of 27 Rorty (1982), p. 14, emphasis in original. 28 Kant (1933), B157, emphasis in original. Cf. also Kant (1956), p. 101.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

230

Language, World, and Limits

applying concepts in expressing the former.29 Nevertheless, when set against the enlightenment we have when intuitions and concepts are both at play, this furnishes us with exactly the kind of contrast that can substantiate a saying/showing distinction. It is again a contrast between knowing how things are (actually receiving) and, as a knowing subject, having an awareness of one’s own existence at the limit of what one can know (being self-consciously receptive). This contrast can substantiate a saying/showing distinction in the following (Tractatus-inspired) way. To put a state of enlightenment into words is to offer to a potential receiver (to anyone capable of understanding the words) a model, or picture, of what one has received. One’s offering mirrors one’s reception. But clearly, only states of enlightenment that involve some reception can be treated in this way. States of enlightenment that consist merely in being self-consciously receptive are not susceptible to being put into words. They indicate the limits of what can be put into words. To be in such a state of enlightenment is to be shown something.30

4. A third example. The Tractatus reconsidered. Its ethical and mystical aspects We can now see how the two examples considered earlier capture the structure of this. What the subject is shown in each case concerns how he or she stands with respect to something given, conceived as a whole. (In the first case I am assuming that the person who assimilates the description is the same as the person from whose point of view it is produced.) The nature of the reception of what is given determines both the extent of what is given (the extent of the whole) and also the way in which the subject knows about it; and in that respect the reception imposes a framework on the whole as the subject knows it. How the framework is filled is a function of what is actually given, and thus of what is received and known, and this can be put into words. This is what can be said. But, as I have remarked, the examples provide a model that is imperfect in the following respect. In each case what the subject is shown can also be put into words. This is because the whole is part of some larger whole which includes the subject and which can also be given (indeed, which he or she can be given). By contrast, for a subject to be shown something in an unmitigated Tractarian sense, there must be no way of being given the relevant larger whole. What can be shown cannot be said. A better example, from this point of view, would have been the following. It amalgamates features of the two previous examples and treats of somebody’s epistemic standing with respect to the whole of space. It concerns a person who, like the rest of us, comes to know about the objects in space by herself being one such and interacting with them causally. She receives the objects in space perceptually, as an embodied agent with various senses. What she comes to know about them can be put into words. But her knowledge is also essentially of such a kind that she can act on it. (This is related to the fact that it is at root egocentric or perspectival, relating back to her and her own spatial location. Fundamentally she knows such things as that the table 29 See Kant (1933), B165.

30 Cf. 4.12–4.121.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

231

is immediately to the left of her.) Suppose she self-consciously reflects on how she is able to exploit her knowledge—how, for example, knowing that the table is immediately to the left of her enables her to put something on it. This will in turn entail selfconscious reflection on how she receives the spatial objects around her: her receiving them perceptually essentially involves her knowing how to act on them.31 But there is no way of saying what she comes to see, or in other words what she is shown, as she self-consciously reflects in this way. A great deal can be said about how things are arranged in space, including how they are arranged with respect to her, and about what is required if she is going to manipulate them in various ways. But nothing that can be put into words will do justice to the non-inferential connection that she perceives between the way in which spatial arrangements are presented to her and her own agency. That she is apprised of such a connection is part and parcel of her simply perceiving and acting. One might even speak here of a transcendental connection between her agency and spatiality. Certainly there is a neo-Kantian flavour to this. It also has a Tractarian flavour. The connection between agency and showing is pivotal to the ethical elements in the Tractatus, and in the Notebooks (just as it is pivotal to much of what we find in Kant, if my way of representing Kant has been at all correct32). In accordance with the ideas that we have been exploring, the metaphysical subject in Wittgenstein’s early work receives the world in such a way as to be a limit of it.33 And the subject’s will, the free exercise of which is what is essentially good or evil,34 determines how the subject limits the world, that is it causes the world as a whole to wax or wane—or so we should like to say.35 But we ought not strictly to say this. For we are alluding here to what the subject is shown in adopting different attitudes towards the world. In freely adopting these attitudes and acting accordingly the subject is shown what is mystical; the subject is shown value and meaning. For the world to wax or wane is for the world to gain or lose meaning.36 A good exercise of the will involves adopting the right attitude, the attitude of Glück, or ‘happiness’.37 However much this may puzzle us, however many questions it may leave open, and however great may be the tension between what we are licensed to say and what looks as if it needs saying, we have at least been given a way of viewing the connection between the subject and the world which makes both it and the attendant concept of showing relatively accessible. For even if we do not know what is involved in adopting an attitude of Glück towards the world, or being glücklich (Wittgenstein’s glosses on this as being in agreement with the world or as doing the will of God38 do not make it any the less schematic) we can at any rate think of it as something that the subject does as an entity within the world, something that can legitimately be described and that makes a sayable difference to how the world is. The glücklich man and the unglücklich man are effably different. There are two worries that we are liable to have about this, but they can both be quashed.

31 Cf. Taylor (1979), p. 154; Evans (1982), pp. 161–162; and Evans (1985c), esp. pp. 371 and 396–399. 32 See e.g. Kant (1933), B428–432 and Kant (1956), p. 101. For more on the connection, see Holiday (1985). 33 5.632. 34 See Wittgenstein (1979), pp. 79–80. 35 See 6.43. 36 Wittgenstein (1979), p. 73. 37 6.43 and Wittgenstein (1979), pp. 78 and 87. 38 Wittgenstein (1979), p. 75.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

232

Language, World, and Limits

The first worry is that these remarks stand in direct conflict with the Tractarian doctrine that what can be shown cannot be said. This worry can be quashed by appeal to the distinction between what is shown to somebody and what it is for that person to be shown it. Even if the former cannot be put into words, it does not follow that the latter cannot be. Saying what it is for somebody to be glücklich is not the same as saying what the world of the glücklich man is like and does not involve putting into words anything that is shown.39 (Note, however, that I do not claim to be doing straight exegesis here. In the Notebooks Wittgenstein says that there can be no describing the mark of the glücklich life.40) The second worry is that the remarks above contradict the Tractarian doctrine that the metaphysical subject cannot be found anywhere in the world. This worry can be quashed by appeal to the connection between the metaphysical subject and the empirical subject. There is no need to view these as two separate entities. There is one subject capable of being assessed from two points of view. From an external point of view the subject is a particular human being in the world, the empirical subject. From his or her own point of view, as recipient of the world and set against it as a limited whole, the same subject becomes that limiting factor outside the world to which it belongs, the metaphysical subject.41 A human being, through acting in a certain way and self-consciously reflecting, can be shown aspects of himself or herself in this latter guise. This sounds much headier than it really is. After all, Kant has already taught us to look upon ordinary self-consciousness as not being consciousness of anything (any thing) in the world.42 And the perspectival differences that arise as a result of the phenomenon of self-consciousness, however imperfectly understood, have become a philosophical commonplace.43

5. Unity in the Tractarian distinction What can be shown according to the Tractatus is not confined to the ethical and mystical aspects of how the subject stands with respect to the world. It also includes a great deal about logical form, structure, and representation. How do these relate to one another? In several ways. The most fundamental common factor is the idea of the world’s overall shape and form, its framework. This determines how the facts, which constitute the world,44 are held together, but it does not determine what the facts actually are. Related to this is the idea of the world’s simply existing, regardless of what it is actually like. What can be shown is always connected not with the world’s being how it is but with its being however it is, not with its being how it can be truly represented

39 For an interesting echo of this idea, see Dummett (1981b), p. 136. 40 Wittgenstein (1979), p. 78. 41 See 5.6 ff. In Pears (1971), pp. 89–90, this is related back interestingly to Wittgenstein’s ethics. Cf. also Mounce (1981), p. 91, where he claims that the truth behind the solipsistic remarks is that I (the subject) have a neighbourless point of view on the world. 42 See esp. Kant (1933), A341–405 and, differently in the second edition, B399–432. 43 Cf. Nagel (1980), Pt. I, Nagel (1983), and Nagel (1986), Ch. IV. 44 See 1 and 2.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

233

as being but with its being such as to admit of true representation at all.45 Thus what can be shown concerns how things must be; what can be said is how they are. Another common factor is meaning. It is no mere pun to say that the ethical, mystical, and logical elements in the Tractatus all concern meaning. It highlights something fundamental about the subject’s relation to the world. Correlative with the notion of meaning is that of understanding, and understanding lies at the root of all states of enlightenment, whether they are instances of knowing something sayable or of being shown something. (More on this in section 7.) A third common factor bringing unity to the Tractarian remarks on showing is unity itself. The subject, in self-consciously reflecting on the world as his or hers and adopting an attitude towards it as a whole, must see it as a unified whole essentially given as such. But this is possible only to the extent that its parts are held together in unity (are united), and this requires a unifying form—logical form. In order to represent the world, propositions must share this form.46 If there were no such form, that is a general form of any proposition, there would be no such unity.47 This form can be given as simply, ‘This is how things stand’.48 These last remarks are reminiscent of the Transcendental Deduction in Kant’s Critique of Pure Reason.49 In each of its versions, it traces out essentially the same link between the unity of what can be given for consciousness and the logical forms of judgments.50 Kant’s concern is with the unity that is involved in objective articulable knowledge of the world.51 ‘A judgment,’ he tells us, ‘is nothing but the manner in which given modes of knowledge are brought to the objective unity of apperception’.52 He goes on to say, ‘This is what is intended by the copula “is”’. So, like Wittgenstein, he indicates the most general form of a judgment (or proposition) by appeal to the simple idea of asserting things to be so. And in Kant, just as in Wittgenstein, this issues in a kind of transcendental solipsism.53 For the unity involved here is the unity of apperception, the unity of what is held together in one consciousness. In the Prolegomena54 Kant writes, ‘The logical functions of all judgments are but various modes of uniting representations in consciousness’),55 and in the first edition version of the Deduction he writes, ‘All objects with which we can occupy ourselves, are one and all in me, that is, are determinations of my identical self . . . [This] is only another way of saying that there must be a complete unity of them in one and the same apperception. But this unity of possible consciousness also constitutes the form of all knowledge of objects; through it the manifold is thought as belonging to a single object’.56 This is echoed in Wittgenstein’s claim that ‘the world is my world: this is

45 See e.g. 6.44. Brian McGuiness develops this idea in an extremely helpful way in his (1966). 46 2.18 and 4.12. 47 Wittgenstein (1979), p. 75. 48 4.5. Even in his later work, where Wittgenstein recoils from the importance that he here attaches to these thoughts, he still connects the notion of the general form of a proposition with a ‘frame through which we look’ (Wittgenstein (1967), Pt. I, §114—see also §§134–136). 49 Kant (1933), A84–130 and, differently in the second edition, B116–169. 50 See e.g. §20 of the second edition version. 51 See Kant (1933), A89–90/B122. 52 Ibid. B141. 53 The phrase ‘transcendental solipsism’ is borrowed from Hacker (1972), e.g. Ch. III. 54 Kant (1950). 55 §22. 56 Kant (1933), A 129.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

234

Language, World, and Limits

manifest in the fact that the limits of language (of that language which alone I understand) mean the limits of my world’.57 We are also reminded of Frege’s discussion of the unity of sentences, which echoes Kant’s claim that the copula ‘is’ indicates some kind of unity.58 Frege claims, in effect, that there is no legitimate way of talking about how the copula, or predicates in which it occurs, relates to reality. It is as if we find ourselves unintentionally talking about objects in the world whenever we try to talk about the unifying form that holds them together, that to which the copula relates. Here it is tempting to say precisely that we are shown how the copula relates to reality though we cannot put it into words. And no doubt Wittgenstein has Frege’s problem in mind at various points in the Tractatus.59

6. Relativization in the concept of showing Showing is grounded in the subject’s reception of what he or she is given. This suggests that there might be a deep relativization in the concept of showing, much deeper than the relativization that I noted in connection with the two examples in section 1, which was relativization to limited linguistic resources. For if the reception in question is itself capable of being given to some other subject, a subject with receptive capacities of some radically different kind, then cannot this subject say what the original subject is shown? (We might even envisage one subject, God say, creatively and intellectually intuiting a second subject’s reception in Kant’s sense, without itself being given anything or being shown anything.60 But if we do, we should note that such a subject might also be incapable of saying anything, insofar as saying something involves exercising receptive capacities.) For one subject to be able to say what another subject is shown in this way, the two subjects would have to occupy different points of view in a very profound sense. It is not possible to say what the Tractarian glücklich man is shown, for example, simply by being somebody else. Any other person receives the same world, in the same way, and in that sense occupies the same point of view, as he does. To say what he is shown we would have to station ourselves outside his world and thus outside our own world, the world.61 The impossibility of our doing this is reflected in our ill-begotten attempts to say what he is shown. We talk of his world’s waxing, of its becoming a different world from that of the unglücklich man.62 But this makes sense only if his world is part of another to which we have access. And it is not. There is nothing we can cite that does not already belong to his world. So we cannot say what the glücklich man is shown. And this suggests that in fact, ultimately, there must be a saying/showing distinction, such as we find in the Tractatus, that is free of relativization. For what we cannot say 57 5.62, emphasis in original. An alternative approach to the question of why there should be this connection between (so-called) transcendental solipsism and logical form is to be found in Hintikka (1958), where Jaakko Hintikka argues that Wittgenstein ‘is interested only in what can be said to be mine necessarily; for otherwise he would be doing empirical psychology. But the only necessity there is . . . is the empty tautological necessity of logic’ (p. 89, emphasis in original). Hintikka claims that the metaphysical subject must therefore be identified with language: they have the same limits. 58 See e.g. Frege (1997a). Cf. also in this connection Plato (1961d), 261c6–262e2 and Plutarch (1976), X, 1011c. 59 See esp. the 4.12s. 60 See Kant (1933), B72. 61 Cf. 4.12. 62 6.43.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

235

cannot be said. It cannot be said in any possible language. The ‘we’, to use Jonathan Lear’s phrase, disappears.63 There is, however, an obvious objection to this. The ‘we’ disappears only in the sense that it becomes implicit. It can easily be brought back into view. For ‘cannot be said’ is simply elliptical in the above account for ‘cannot be said by us (here, from this point of view)’, and ‘any possible language’ for ‘any possible language that we (or perhaps even I) could understand’.64 This indicates, surely, that even the Tractarian distinction involves hidden relativization; and this in turn reinforces the thought that some kind of relativism is a natural concomitant of the very idea of a saying/showing distinction. In response to this objection: making the ‘we’ explicit in this way indicates genuine relativization only insofar as an alternative can be recognized, which, in the nature of the case, it cannot. We cannot recognize a standpoint from which that which cannot be said by us nevertheless can be said; and we can get no proper grip on the idea of such a standpoint’s existing even though we cannot recognize it. Still, the objection may run, even if we cannot get a proper grip on the idea, will not self-conscious reflection convince us that it is at least pointing to something? To this the correct response is: yes. The saying/showing distinction is here turned on itself. For it is through self-conscious reflection that we can be shown things, and talk of unrecognizable standpoints certainly gestures, however clumsily, in the direction of something that we can be shown. That our own point of view is not privileged; that what we are shown may be sayable elsewhere; that any saying/showing distinction is bound to involve relativization; that what we take to be an absolute saying/showing distinction we do so only because ‘absolute’ is itself a relative term: these are the kinds of claims that we should like to make when giving voice to what we are shown, though of course we cannot give voice to it, and each of these claims, to the extent that it is a genuine claim at all, is false. What we are shown we must be content to pass over in silence.65

7. Understanding How does understanding connect with this? The concept of understanding is extremely broad. Among the many things that I might naturally be said to understand (or not) are: English; some particular word or phrase; an innuendo; general relativity; Kant’s third Critique; why somebody is behaving in a certain way; modern art; Beethoven’s fourteenth string quartet; life; a close friend; death. And if the solipsist really does mean something,66 then what the solipsist means can be added to this list. Understanding is one of the most general cognitive states by which we can be related to the world. For one thing, whenever somebody knows that such and such is the case, this will be because there is something (a situation, a theory, a piece of information, a piece of evidence, . . .) that he or she understands. Understanding something is always associated with having a certain receptive capacity, and states of enlightenment of whatever kind, including instances of being 63 Lear (1984), p. 238.

64 Cf. 5.62.

65 Cf. 7.

66 5.62.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

236

Language, World, and Limits

shown something, are always grounded in some form of understanding. But our paradigm of understanding is linguistic understanding. So too it is linguistic understanding that supplies our paradigms of showing. Two facts about the Tractatus testify to this (as well as to the more general link between showing and understanding). First, there is the fact that the concept of showing first seriously emerges in the early 4s in the context of a discussion of the sense of a proposition, a discussion that focuses very much on linguistic understanding. A proposition shows its sense, we are told; that is, it shows how things stand if it is true, and to understand it is to know how things stand if it is true.67 The second relevant fact is that the solipsistic remarks, which make their first appearance in the 5.6s somewhat out of the blue when the discussion has hitherto been of a very impersonal, abstract, logical character, are immediately set in the context of considerations about linguistic understanding. As soon as we are told that what the solipsist means is correct and that it makes itself manifest, we are given this elucidation: ‘the world is my world: this is manifest in the fact that the limits of language (of that language which alone I understand) mean the limits of my world’.68 This may indeed provide us with a clue as to why the solipsist is tempted to express what he means in the way in which he does (using the pronoun ‘I’ rather than ‘we’ or ‘it’ or ‘God’). For what he is shown concerns his standing with respect to the world as mediated by linguistic understanding; and insofar as he is inclined to see such understanding as essentially and non-derivatively his, he will be tempted to express what he is shown accordingly. Three special advantages accrue from taking linguistic understanding as a prime instance of being shown something, or more strictly of what can ground being shown something. First, it casts a great deal of reciprocal light on to the philosophy of language. If we view understanding in this way, then we are led to the rather radical thesis that it is possible to show but not to state what expressions mean. This thesis has recently received much support.69 If it is correct, then it must have serious implications concerning the aims and aspirations that we should have when practising the philosophy of language. In particular, it should curtail a natural ambition that we might have to elucidate meaning by outlining what would be involved in stating the meanings of the expressions in a given language.70 But the thesis that it is impossible to say what any expression means is perfectly compatible with the thesis that it is possible to say a great deal about what it is to understand the expression, and this will be significant when it comes to deciding what we may aspire to in practising the philosophy of language.71 It also serves to emphasize that, however mysterious and esoteric the concept of showing initially threatens to be, there is no reason why instances of being shown something should not be mundane, familiar, and susceptible of description. Understanding a language is certainly each of these things, and this is the second reason why it is advantageous to accord it prominence as an example of being shown something: we are reminded of how much we can say about showing. 67 4.002, 4.024, and neighbouring sections. 68 5.62, emphasis in original. 69 Cf. Dummett (1973), p. 227; Dummett (1981b), p. 129; Evans (1982), p. 26; McGinn (1982), p. 223; and, in a somewhat different way, my Essay 5 in this volume, pp. 75–76. 70 Cf. Essay 5 in this volume, where I try to explore these ideas in the light of Wittgenstein’s later work. 71 Cf. again the passage from Michael Dummett cited in n. 39.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

237

The third advantage is that we are provided with a helpful gloss on what it can be like not to be shown something, in the sense of not being sensitive to what one is capable of being shown—being prevented from focusing attention in an appropriate way. I am not thinking in terms of anything as straightforward as not understanding a particular language or misunderstanding a particular word. I have in mind the case where somebody certainly knows the meaning of a word or phrase, but (perhaps because of a misleading analogy with some other expression) she misconstrues its logic and thus attempts to use it in a way for which it is unsuited. She produces nonsense, often in the guise of unanswerable questions, and when she self-consciously reflects on her own linguistic understanding, she finds her attention drawn to these ‘questions’, which obscure what is there to be shown. This, or something like it, is, according to the Tractatus, the general diagnosis for most philosophical perplexity.72 And the philosopher’s task is obviously not to answer such ‘questions’ but to prevent their arising, by reminding people of the logic of their words—of what exactly they are entitled to say and to think—so that they are then free to focus attention on what is there to be shown though it cannot be said.73 A valuable tool in carrying out this task will be the injunction to look at what expressions are actually used for, for this will highlight their logic.74 Philosophy, on this conception, is essentially an activity, whose aim is to clarify and elucidate our use of words; there are no distinctively philosophical propositions or theses in which it results.75 Perhaps the most striking feature of this conception of philosophy is its close resemblance to that found in the later work of Wittgenstein. Indeed the principal difference is purely terminological: Wittgenstein later speaks of ‘grammar’ where here he speaks of ‘logic’.76 To compare Wittgenstein’s later work with the Tractatus in these respects, and, more specifically, to compare them in such a way that a saying/showing distinction arises in the later work, is, I shall now suggest, very instructive.77

8. The later work of Wittgenstein 8.1 Wittgenstein’s later conception of philosophy. Its resemblance to that of the Tractatus In his later work, just as in the Tractatus, Wittgenstein urges us to view philosophical perplexity as the result of a natural impulse to misconstrue the grammar of our language. For example, there is an inclination to draw false analogies between different linguistic constructions, the similarity of whose surface grammar hides crucial differences of function. Again, we are often tempted to interpret perfectly good metaphors and idioms too literally and thus inappropriately to assimilate them to utterances where the same expressions are used more straightforwardly. The upshot is that language ‘goes on holiday’, good sense gives way to nonsense, and we find ourselves puzzled and confused by certain pseudo-questions that arise, posing as philosophical 72 4.003. 73 4.113–4.115 and 6.53. 74 6.211. 75 4.112. 76 See Wittgenstein (1967), Pt. I, §§109 and 122–128, and Wittgenstein (1969a), pp. 26–27. 77 Cf. again the works cited in n. 9.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

238

Language, World, and Limits

problems.78 A caricature of such a ‘question’ is, ‘Why is it that only I can know when ouch!?’—which in Wittgenstein’s view is what the more specious ‘Why is it that only I can know when I am in pain?’ pretty much amounts to. The philosopher’s task is a therapeutic one: it is to get us to command a clear view of our language so that we stop feeling the urge to answer such ‘questions’. To the extent that we can do this (command a clear view of our language), we can also feel our way around inside our world-view and gain a sense for where the limits of comprehensibility lie. That is, we can self-consciously reflect on how we make sense of things, or how we receive them; we can focus attention on our own receptive capacities without being distracted by ill-begotten philosophical conundrums. And, if the comparison with the Tractatus can indeed be sustained here, then this is tantamount to saying that we can be shown something. An individual who has been shown something in this way reveals this principally by using language properly and resisting the allure of ambitious philosophical theses with all their attendant bemusement. Such an individual has to that extent been cured. Here the following objection might be voiced: it is absurd to suppose that one can reveal such a state of enlightenment simply by carrying on in exactly the way in which one would have carried on if one had never fallen into temptation in the first place (to switch the metaphor); for there must surely be some further distinguishing mark that differentiates one from one’s unfallen but unenlightened fellows. This objection rests on various mistaken presuppositions. First, it is a mistake to suppose that the individual in question carries on in exactly the way in which he or she would have carried on had there never been any lapse into temptation. One difference is precisely that there has been such a lapse. The individual’s reaction against this lapse, which may be a continuing process, and which (especially if it is) will give him or her a strong feeling of empathy for those who still succumb to temptation, might be enough to justify talk of a special kind of enlightenment—just as we can say that somebody understands an argument better the livelier an appreciation he or she has of the counterarguments to it and of how they are to be rebutted.79 (On this point it is a notable feature of Wittgenstein’s later work that in it he exhibits a real sense of the allure of what he is trying to resist: the same cannot always be said of those purporting to follow him.) Secondly, there is in any case no clear warrant for thinking that not falling into temptation entails being unenlightened. Could one not reveal that one had been shown something just by (self-consciously) carrying on—as it were sinlessly? Of course, this question invites a strenuous exercise of the imagination. The metaphor of sinlessness embodies an idealization transcending anything that we ever encounter. The question should not be taken to suggest (and is a bad question insofar as it does suggest) that in his later work Wittgenstein distinguishes between two kinds of people: those who need therapeutic treatment or redemption in order (once again) to see the world aright, and those who do not, because they never become ill or they 78 See Wittgenstein (1967), Pt. I, §38. 79 Cf. Lear (1984), pp. 240–241, where Jonathan Lear writes, ‘Post-neurotic consciousness is fundamentally more complex than a healthy consciousness that has never suffered disease or cure’. (Another helpful paper by Lear that casts light on many of these issues is his (1982).)

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

239

never stray from the straight and narrow in the first place. In fact, such a suggestion is implicit in the original objection, and is the third and most seriously mistaken of the presuppositions on which it rests. For underlying the objection is the following picture of Wittgenstein’s later view of philosophy: the vast majority of people carry on exactly as they ought, but some, ‘philosophers’, are led astray, and it is incumbent upon a small sub-group of these, those who have perceived their fall into sin and who have been restored to some kind of innocence, to expend their energies trying to redeem their fellow philosophers. This picture is a travesty of what we find in Wittgenstein’s later work. The puzzles, confusions, and conundrums, indeed the deep intellectual malaise,80 that Wittgenstein probes throughout his later work ought to be viewed not as the foibles of some small, unfortunate group but as part of a human predicament. It is true that Wittgenstein speaks of philosophical perplexity as the ailment of the philosopher.81 But we are all, in that sense, philosophers. This aspect of the human predicament is for Wittgenstein, just as for Kant, a natural and universal one.82 And there is no such thing as never succumbing to temptation or reaching the point at last of being able to resist all temptation. There is rather a continuing, piecemeal, nonsystematic process of trying to resist the allure of particular temptations as and when they arise, and the philosopher’s task is to try to further this process.83

8.2 Inducement in the later work to say what we are shown Now there is an irony in the Tractatus which carries over to the later work too. Most of the examples of philosophical nonsense that we find in the early work do not arise as candidates for excision by the philosophical methods outlined there. On the contrary, they actually constitute the work, and they do so moreover to the virtual exclusion of proper sense as construed by the Tractatus itself.84 Of course, if it is true that philosophy is not a body of doctrine but an activity aimed at combating nonsense,85 then the fact that the Tractatus itself consists mostly of nonsense does not ipso facto prevent it from counting as good philosophy. For it may be that uttering nonsense of a certain kind is one way to get people to see that they themselves are uttering nonsense of that kind (it is not hard to imagine how this could be) with the overall effect that less nonsense is uttered. If so, then uttering nonsense of the relevant kind would be one good way to do philosophy.86 True, there are parts of the Tractatus that prevent this from counting as a full vindication of it in its own terms, for example the claim that ‘the [only] correct method in philosophy would . . . be . . . to say nothing except what can be said’87 and the claim in the Preface that, if the Tractatus has any value, it consists, in part, in the fact that (true) thoughts are expressed in it. Nevertheless, we have here a sense of how, even by broadly its own lights, the Tractatus can count as a successful tissue of nonsense. Indeed, can it not eventually be turned successfully on itself? For surely anybody who understands its purport will then want to discard it

80 82 85 86

See Wittgenstein (1967), Pt. I, §111. 81 E.g. in his (1969a), p. 59. See Kant (1933), A642/B670. 83 Compare Wittgenstein (1967), Pt. I, §127. 84 See 6.54. 4.112. One could speak here of elucidatory or illuminating nonsense. Cf. Hacker (1972), I.4. 87 6.53.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

240

Language, World, and Limits

along with other such nonsense: he or she will want to throw away the ladder after having climbed it.88 The irony mentioned above seems not to be a deep one. But this is too quick. The nonsense that we find in the Tractatus cannot in fact be regarded merely as part of some clarificatory exercise. It is a carrying out of that exercise in the wake of the kind of philosophical enlightenment that it is supposed to deliver. It is implicit in the Tractatus that, as philosophical clarification is achieved, we are shown things. But this, ironically (and this is the original irony, which we can now recognize as a very deep one) issues in a new impulse to utter nonsense. We want to say what we are shown. What the Tractatus itself consists of, for the most part, are illegitimate attempts to do just that. So the success of the philosophical enterprise does after all bring with it its own brand of philosophical nonsense which cannot be discarded by any simple process of self-application. Similarly in the later work, though in a much less marked way, the enlightenment that is supposed to result from philosophical therapy brings with it a new susceptibility to philosophical sickness. Self-consciously attending to our own receptive capacities, we come to view the world as dependent, in some way, on us. We are led towards a kind of transcendental idealism that is like the transcendental idealism of the Tractatus, except that the metaphysical subject of the earlier work is replaced by some sort of plural descendant: there is a shift from ‘I’ to ‘we’. We want to say: the limits of our language now mean the limits of our world.89 But this idealism, like its earlier version, is something we are shown and there is no legitimate way of giving voice to it. When we attempt to do so, we find ourselves making claims which, insofar as they are intelligible at all, have to be understood as straightforwardly false claims about how we (human beings, or some particular group of human beings) stand in relation to the rest of the world. The limits of our language (English? German? human discourse?) do not mean the limits of our world. They lie within it awaiting empirical investigation by, say, anthropologists, sociologists, or linguists.90 The rub is that Wittgenstein himself sometimes tries in his later work, if only via hints and suggestions, to express a kind of transcendental idealism. (This is what gives this view of his later work viability.) There are consequently times when he says—often, it is true, with some hesitation—precisely the sorts of things that by his own lights ought not to be said.91 The best examples of this are to be found in his work on mathematics. He suggests that in some sense mathematical truth is determined by what we believe.92 What mathematicians would characteristically say, on the other hand, at least when not self-consciously trying to develop a philosophy of mathematics, is that mathematical truth is in no sense determined by what we believe. And for 88 6.54. 89 Cf. 5.62. 90 Cf. the following quotation from an unpublished early draft of McDowell (1993): ‘No doubt there is something right in finding a kind of idealism in the [Wittgensteinian] thought that we are guardians or trustees of meaning, not its puppets or slaves—so that, to put it in . . . [a] dangerous way, we are involved on the right-hand sides of statements like “‘Diamonds are hard’ is true if and only if diamonds are hard”. But if there is a kind of idealism here . . . it is probably something we should do well not to try to state as a thesis at all; but perhaps we can say that it shows itself in the relation between language, or thought, and reality.’ 91 Consider in particular some indefinite ‘transcendent’-sounding uses that he makes of the pronoun ‘we’, as argued convincingly by Bernard Williams in Williams (1981c). Two examples might be Wittgenstein (1967), Pt. I, §200 and, better, Wittgenstein (1974), Pt. II, §42. 92 Wittgenstein (1967), pp. 226–227.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

241

Wittgenstein, what mathematicians would characteristically say must surely be respected. It supplies, presumably, the data for any philosopher of mathematics seeking to gain a clear view of mathematical discourse and it is not to be interfered with. The philosopher is not, as such, in a position to take issue with any particular mathematical tenet.93 To be sure, we have to distinguish between what mathematicians would characteristically say when practising mathematics and what they would characteristically say otherwise, even ‘when not self-consciously trying to develop a philosophy of mathematics’. It is only the first of these that the philosopher need in any sense treat as sacrosanct. Such a distinction arises because everyone, the mathematician included, is prone to confused philosophical meanderings, and there may well be a distinctive and typical way in which mathematicians confusedly wander off from their own discipline. Moreover, they may very well import their confusions back into their discipline. (Intuitionists will cite use of classical reasoning as an example of this, especially, perhaps, where false analogies are drawn between the finite and the infinite. Think also of early work on infinitesimals.94) The lines between mathematics and the philosophy of mathematics, or between mathematical work that is untainted by philosophical confusion and that which is not, or again between interfering with mathematics and trying to combat intellectual diseases to which mathematicians are especially prone, are not sharp. So it is always open to Wittgenstein to characterize the kind of Platonism that frequently permeates mathematical texts and that supposedly tells against his own ruminations on mathematics, not as part of mathematics, nor yet as part of a legitimate philosophy of mathematics, but as a symptom of an extramathematical sickness indicating a need for therapy. Indeed he writes in this vein that ‘what we “are tempted to say” . . . is . . . not philosophy; . . . it is its raw material. Thus, for example, what a mathematician is inclined to say about the objectivity and reality of mathematical facts, is not a philosophy of mathematics, but something for philosophical treatment’.95 There is nevertheless a serious problem in that being able to distinguish between what mathematicians would say because it is part of legitimate mathematical practice and what they would say because there are philosophical confusions to which they are particularly prone itself requires the kind of sensitivity that can be acquired only by studying legitimate, unadulterated mathematical practice. And even if there is no vicious circle here, the fact remains that by his own principles Wittgenstein is entitled to describe and comment on mathematical practice only in a way that is mathematically non-revisionary. It is not clear that he always succeeds in confining himself in this way. Consider his discussion of the diagonal proof that there are more real numbers than natural numbers.96 He expends what seems like a disproportionate amount of energy in trying to make us appreciate that the result of this proof, as of any other, contains no more than the proof itself. His worry is that the usual way of putting the result suggests that it somehow does contain more. Thus he writes, ‘The dangerous, deceptive thing about the idea: “The real numbers cannot be arranged in a series”, or again “The set [of real numbers] . . . is not denumerable” is that it makes the 93 See ibid. Pt. I, §124. 94 Cf. Kline(1972), p. 389 and Boyer (1949), p. 223. 95 Wittgenstein (1967), Pt. I, §254, emphasis in original. 96 Wittgenstein (1978), Pt. II, §§1–40.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

242

Language, World, and Limits

determination of a concept . . . look like a fact of nature’.97 An essentially related concern that he has is that the standard ways of couching the result, and of couching correlative results, are such as to deprive them of an anchoring in mathematical practice. ‘What,’ he asks, ‘can the concept “non-denumerable” be used for? . . . [The diagonal proof] . . . may lead us to say that 2À0 > À0 . . . But if we do say it—what are we to do next? In what practice is this proposition anchored?’.98 Earlier he complains, ‘One pretends to compare the “set” of real numbers in magnitude with that of . . . [natural] numbers’.99 But here one wants to reply that one pretends no such thing: one does it (and discovers that the first set is bigger). A great deal can be said about what the concept ‘non-denumerable’ can be used for and about what follows from the proposition that 2À0 > À0 , how it fits into mathematical practice, what questions it leaves open. It is perfectly acceptable to say that there are more real numbers than natural numbers. Wittgenstein’s worries about this are ill-founded, almost paranoiac. In striving to rid us of a confused and pernicious interpretation of certain mathematical results, Wittgenstein casts illegitimate doubt on the results themselves. He begins to challenge mathematical practice in precisely the way in which he himself would insist that it cannot be challenged. We now have, I think, an account of why this should be. He is searching for a statement of the ‘true’ meaning of the mathematical results as revealed by their proofs, the meaning we grasp as soon as we have a really clear view of those proofs and of how they fit into the surrounding mathematical practice, the meaning that we grasp when we ‘take a wider look round’.100 But we grasp this meaning only in the sense that we are shown it. It cannot be stated. As he searches for a statement of it, Wittgenstein begins to wrench the very mathematical discourse that is supposed to form his subject matter. The difficulty here, as he himself remarks elsewhere, is: to stop.101 Describing the relevant language-games is all that he, the philosopher, can do—and all he need do. The language-games can then speak for themselves. Again we may quote Wittgenstein: ‘Let the use of words teach you their meaning . . . let the proof teach you what was being proved’;102 ‘The correct method in philosophy . . . [is] to say nothing except what can be said’.103 There is, however, much that can be said—even by the philosopher. Some of it is in urgent need of being said. Philosophical confusion is a universal condition of humankind and to the extent that it is deep-rooted and exerts a damaging influence on people’s lives,104 then what the philosopher has to say may be of profound practical significance. It is often mistakenly assumed that Wittgenstein’s later work, if correct, signals the end of philosophy. On the contrary, it highlights just how much philosophical work still needs to be done. As long as there are people reflecting on their situation and trying to understand the world, then there will be philosophical confusions needing to be teased out.105 We are always attempting to surmount our particular, historically and culturally conditioned conceptual apparatus, because of the allure of apparent shortcuts to a panoramic understanding of the world: and we succeed 97 Ibid. §19. 98 Ibid. §§12 and 35. 99 Ibid. §22. 100 Ibid. Pt. II, §6. 101 Wittgenstein (1981), §314. 102 Wittgenstein (1967), p. 220, emphasis in original. 103 6.53. 104 Compare Wittgenstein (1967), Pt. I, §111. 105 Cf. Kant (1933), A298/B354–355.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

On Saying and Showing

243

only in mishandling the apparatus. But even when philosophy has been put to effective work in making us resist this allure, indeed because it has been put to effective work, there is still a danger, though now a new danger, of philosophical confusion. For we gain insight and are shown something that we cannot express, though we are tempted to do so. We have to learn to say only what can be said, to recognize the limits of our knowledge. I shall end this essay as I began it, with a quotation from Plato:106 The purifier of the soul is conscious that his patient will receive no benefit from the application of knowledge until he is refuted, and from refutation learns modesty; he must be . . . made to think that he knows only what he knows and no more.107

106 This quotation is from Plato (1961d), 230c–d. The opening quotation was from Plato (1961e), 7, 341c–d. 107 I should like to thank Philip Turetzky for many valuable conversations on these topics. Thanks are also due to Tim Crane, Naomi Eilan, Ross Harrison, Philip Percival, and the editor of Philosophy for very helpful comments on earlier versions of this essay.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

18 Ineffability and Nonsense Abstract Criteria of ineffability are presented which, it is claimed, preclude the possibility of truths that are ineffable, but not the possibility of other things that are ineffable—not even the possibility of other things that are non-trivially ineffable. Specifically, they do not preclude the possibility of states of understanding that are ineffable. This, it is argued, allows for a reappraisal of the dispute between those who adopt a traditional reading of Wittgenstein’s Tractatus and those who adopt the new reading recently championed by Cora Diamond, James Conant, and others. By maintaining that what the nonsense in the Tractatus is supposed to convey is ineffable understanding, rather than ineffable truth, we can do considerable justice to each of these readings. We can also do considerable justice to the Tractatus.

1. Lewis’ argument that there are inexpressible truths David Lewis holds that there are inexpressible truths. The following argument, which is a variant of an argument that he discusses, can be used to motivate this view.1 Suppose that S is an item that can express a proposition, in the minimal sense that, for some proposition p and for some possible world w, S expresses p in w. Say that S univocally expresses proposition p (in possible world w) if and only if S expresses p (in w) and S does not express any proposition other than p (in w). And let κ be the number of possible worlds. Now clearly the number of propositions that can be univocally expressed by S is no greater than κ. But the number of propositions altogether is at least as great as 2κ, the number of sets of possible worlds—perhaps greater, if propositions slice logical space more thinly than sets of possible worlds do. And 2κ is greater than κ. Hence there are some propositions that cannot be univocally expressed by S. How does this argument motivate the view that there are inexpressible truths? Well, on the assumption that any truth that can be expressed at all can be univocally expressed by S, the argument can readily be converted into an argument for that view.2 1 See Lewis (1986), §2.3. 2 I am assuming, for these purposes, that a truth is nothing but a true proposition. I am also assuming that there are at least three possible worlds. If there were only two, and if truths sliced logical space no more thinly than sets of possible worlds, then there would be only two truths.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

245

This does not settle the matter, however. The assumption that any truth that can be expressed at all can be univocally expressed by S is not as innocent as it may appear. It may appear completely innocent. Surely, we may say, if a truth can be expressed at all, then there is a possible world in which, by sheer stipulation perhaps, S expresses it, and in which circumstances elsewhere ensure that S expresses no other proposition, in some other language say. But a little probing of the original argument reminds us that there is at least one proposition that we have independent reason to deny can be univocally expressed by S, namely the proposition that S is false.3 And if S does in fact express a proposition, but a falsehood rather than a truth—for instance, if S is the sentence ‘Grass is pink’—then the proposition that S is false is itself a truth. So it is a truth which, though certainly expressible, cannot be univocally expressed by S.4 Lewis himself would accede to this. But he would still regard the original argument as motivating the view that there are inexpressible truths. The point, he would say, is not that there is one special proposition that cannot be univocally expressed by S. The point is not even that there is one special family of propositions which, because they involve the semantics of S, cannot be univocally expressed by S. The point is that there are many more propositions than there are possible worlds. Assuming that κ is infinite, only an infinitesimal minority of all propositions can be univocally expressed by S. And that, on Lewis’ view, has nothing specifically to do with S. It is because only an infinitesimal minority of all propositions can be expressed at all. Most are too ‘untidy’. In Lewis’ own words, they are ‘utterly unpatterned and miscellaneous’.5 Furthermore, of these ‘untidy’ propositions, half—if I may so put it—are truths.

2. Standards of expressibility. Expression and ineffability defined How might we respond either to the original argument or to its use in motivating the view that there are inexpressible truths? There are all sorts of ways of resisting each. We might reject the very idea of possible worlds as entities fit to be counted; likewise propositions. We might deny that every set of possible worlds determines a proposition. We might challenge the attempt to quantify over all truths.6 We might insist that, even apart from any semantic paradoxes, there are limits to what S can express which are not limits to what can be expressed. (The only assumption that has been made so far about S is that it is an item that can express a proposition. On some views this is scarcely an assumption at all. On these views suitable conventions can equip anything to express a proposition. But there are other views whereby nothing can express a proposition unless it has some suitable internal complexity. On these views, in order for an item to express any given proposition it needs, at some relevant level, to share the internal complexity of that proposition. Hence even if we stipulate that the letter ‘G’ shall abbreviate the sentence 3 See Moore (1984). 4 Thus it could not be the case both that ‘Grass’ was a name of the sentence ‘Grass is pink’—understood as univocally expressing some proposition—and that ‘. . . is pink’ denoted falsehoods. 5 Lewis (1986), p. 108. 6 Cf. Sullivan (1999–2000). Not that this, though it certainly helps us resist the original argument, is likely to help us resist the conclusion that there are inexpressible truths.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

246

Language, World, and Limits

‘Grass is green’, what then expresses the proposition that grass is green is not, so to speak, just the letter but some suitably structured precondition of the very possibility of our making that particular stipulation.7)8 Another way in which we might respond to the original argument and to the conclusion that there are inexpressible truths would be to accept the conclusion, for the reasons indicated in the argument, and by the standards of expressibility implicit in the argument, but then to try to remove some of the sting from the conclusion by pointing out how high those standards are. The argument presupposes that for S to express a proposition p is for S to have precisely p as its content. There are lower standards whereby S expresses any proposition that is in some sense ‘part’ of its content. For instance, there are standards of expressibility whereby the sentence ‘Grass is green and coal is black’ expresses, among other things, the proposition that grass is green, and indeed the proposition that grass is coloured. This serves to remind us of something that is basic to the very idea of inexpressibility. There are many different standards of expressibility. These different standards depend on a variety of factors such as: whether expression is restricted to linguistic expression; if so, whether it is restricted to expression in an existing language; or whether it can include expression in a possible language; if the latter, what counts as a possible language; whether, for example, a possible language must be something that a human being could in principle learn; whether it must be something that a human being could in practice learn; whether linguistic expression is confined to the literal; in what sense of ‘part’, if any, what is expressed can be ‘part’ of the content of what expresses it (the point just being canvassed); and many more. There is no right or wrong about which of these standards to adopt. Different standards are appropriate for different purposes.9 What rationale would there be for adopting standards lower than those implicit in the original argument? One rationale would be the following. If, by whatever good grace, there were a ‘complete description’ of the world, that is to say a set of true representations that narrowed down to one which possible world it was, still, by the original high standards, we might be forced to say that there were countless truths (consequences of this complete description) that remained unexpressed. That would be bizarre.10 Another rationale would be that the most interesting questions about inexpressibility, from a philosophical point of view, are not questions about the awk7 But would there then be an analogue of the original argument, in which the relation of expression was replaced by whatever more complex relation holds, in this example, between the proposition that grass is green and the letter ‘G’? 8 It is a nice question, incidentally, to which the discussion below adds some piquancy, how the argument and the conclusion would best be resisted within the framework of Wittgenstein’s Tractatus— Wittgenstein (1961), which will be the source for all unaccompanied references hereafter—according to which ‘Man possesses the ability to construct languages capable of expressing every sense’ (4.002; cf. 4.5). (In raising this question, I am, of course, prescinding from the nonsensicality of that framework.) At least two of the responses listed above suggest themselves. Thus it could be denied that every set of possible worlds determines a proposition, or, in more Tractarian terms, that every sub-space of logical space is a sense (see 3.4–3.411). Or it could be urged that only what has suitable complexity can express a given truth (see 3.1 ff., 3.332, 3.34 ff., and 4.5; also relevant is 4.0621). Note, however, that this latter response would fall prey to the analogue of the original argument mentioned in the previous note. 9 Cf. Alston (1956). 10 Lewis (1986), §3.2 is relevant here. See also the Tractatus, 4.26.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

247

ward carving up of logical space; they are questions about logical space itself. We are all familiar with various kinds of inexpressibility that relate to the former: kinds of inexpressibility whose associated standards of expressibility are relatively high. Most of us have at one time or another found that we cannot express how we feel about something. This may have been because of some deep non-negotiable lack of fit between private sentiment and the public manifestation of it. It may have been because of deficiencies in whatever languages we speak. It may have been because of deficiencies merely in our own vocabularies. It may have been because of practical, social, or psychological impediments to our using even the linguistic resources available to us. But it has not been because of the kind of inexpressibility by which philosophers are typically exercised, or at least by which some of them like to think that they are exercised: inexpressibility that has to do not with how things are but with their being any way at all; inexpressibility whose associated standards of expressibility are low enough to make it (the inexpressibility) a matter of the deepest principle.11 To repeat: there is no right or wrong about which standards to adopt. But it will be as well for me to settle on some standards and to give some indication of what they are before I proceed. Now roughly, I intend my discussion to be based on the lowest standards of expressibility consonant with its retaining the sort of philosophical interest to which I have just adverted. It is important, however, that these are not the lowest standards there are. In particular, I shall discount all but linguistic expression. For example, I shall discount expression by means of music. I shall also discount expression of what is not itself either true or false. For example, I shall discount expression of moods. A little more precisely, I shall define expression in the following way: x expresses y if and only if (i) x is a linguistic item with content that makes it either true or false, (ii) y is a non-linguistic item12 with content that makes it either true or false, and (iii) the content of x entails the content of y.13 I shall also allow for a derivative sense in which people can express things: a person A expresses an item y, in this derivative sense, if and only if A produces an item x that expresses y in the defined sense.14 I shall otherwise adopt the lowest standards of expressibility I can.15 And I shall reserve the label ‘effability’ for expressibility by these standards—as I shall, mutatis mutandis, its cognates ‘effable’, ‘ineffable’, and the rest.

11 Cf. Wittgenstein (1967), Pt. I, §610; Wittgenstein (1980), p. 16, fourth paragraph; and Cooper (1991). 12 This is not to preclude the possibility that, as in the case of certain complex beliefs, y depends for its existence on language. 13 Thus the sentence ‘Grass is green and coal is black’ expresses, among other things, my belief that grass is green. 14 Thus you can express my belief that grass is green by saying, ‘Grass is green and coal is black.’ 15 Note that, by these standards, it is no obstacle to the expressibility of a truth that, if it were expressed, it would be a falsehood. Thus the thesis that all truths are expressible (by these standards), whatever else might be said against it, does not yield the conclusion that all truths are expressed in the way in which the thesis that all truths are knowable notoriously yields the conclusion that all truths are known: see Fitch (1963).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

248

Language, World, and Limits

3. The question whether there are any ineffable truths Very well; are there, in these terms, any ineffable truths? I am convinced that there are not.16 But I shall not try to argue for that conclusion here.17 Actually, I am tempted to say that it does not need much in the way of argument. I find myself in the same sort of position vis-à-vis the question whether there are any ineffable truths as Donald Davidson does vis-à-vis a similar question that he poses in his celebrated essay on the idea of a conceptual scheme:18 whether there could be a language whose sentences were untranslatable into any of ours. ‘It is tempting,’ Davidson writes, ‘to take a very short line indeed: nothing, it may be said, could count as evidence that some form of activity could not be interpreted in our language that was not at the same time evidence that that form of activity was not speech behaviour.’19 He proceeds to make clear that the only thing wrong with this short line would be its shortness, not, so to speak, its direction. At a key moment later in the essay he addresses the following subsidiary question: ‘How well [do] we understand the notion of truth, as applied to language, independent of the notion of translation [?]’20 And his answer, which he goes on to elaborate, is a virtual re-appropriation of the short line: ‘We do not understand it independently at all.’21 Similarly, I am inclined to say that we do not understand the notion of truth, at least when it is taken in its strictest sense, independently of what can be expressed; or rather, more cautiously, that we do not understand the notion of a truth independently of that. (The individuation involved in talking about a truth, by forcing the question of where one truth stops and another begins, clearly puts additional demands on talk in general about truth.)22 But I concede that there is much more to be said.23 The reason why I shall not try to say any of it now is that my chief concerns lie elsewhere. And this in turn connects with the fact, whose relevance will be clear later, that the idea of an ineffable truth is quite foreign to Wittgenstein’s Tractatus.24

16 Or at least, with the possible exception of necessary truths, which may be said to lack content, I am convinced that there are not. The idea that necessary truths lack content is of a piece with the idea embraced in the Tractatus that necessary truths lack sense: see 4.46–4.461 and cf. 3.13. For current purposes I do not need to go into detail about how I understand ‘content’. I can accordingly leave this matter unsettled. (But note: if I were to accede to a conception of content that was sufficiently demanding for necessary truths to lack it, then I should likewise have to accede to the paradox of ineffable truths that can be put into words, for instance the truth that all aunts are female. This paradox would at worst, I think, reflect idiosyncrasy in my use of ‘ineffable’.) I shall ignore this complication throughout the main text. 17 For a sketch of an argument see Essay 16 in this volume, section 2. 18 Davidson (1984g). 19 Ibid. p. 185. 20 Ibid. p. 194. 21 Ibid. 22 Cf. Kremer (2001), p. 61. 23 For some relevant observations, which include further scepticism about the idea of an ineffable truth, see Horwich (1990), Ch. 2, n. 4; Quine (1992), pp. 77 ff.; and Wright (1992), pp. 72–73. 24 See n. 8. Cf. also 4.063. (Note: on a conception of content whereby the Tractatus is committed to the ineffability of necessary truths—see n. 16—then what is foreign to it is not, of course, the idea of an ineffable truth, but the idea of a truth that cannot be put into words.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

249

4. The question whether there is anything at all that is (non-trivially) ineffable Suppose we accept, then, that there are no ineffable truths; and likewise, no ineffable falsehoods. Does this mean, given my criteria of expression, that there can be no further interest in the question whether anything is ineffable? I do not think so. Most things, of course, since they clearly do not have content that makes them either true or false, are trivially ineffable. Ineffability, on my understanding, is not the same as indescribability. A sock, for example, is ineffable, even if we can give a complete description of it. So too, for that matter, is the opening bar of Beethoven’s fifth symphony—which gives the lie to Ramsey’s famous quip about our not being able to whistle what we cannot say (at least on one uncharitably literal interpretation of that quip).25 The issue is whether there are things whose ineffability is not trivial in this way. Are there perhaps things which, though ineffable, are of such a sort that we might have expected them to be effable, say because most things of that sort are? I believe so. I have urged elsewhere that certain states of knowledge fit the bill.26 What I had in mind included some (but by no means all) states of knowing how to do something: for instance, states of knowing how to exercise a given concept. However, Jason Stanley and Timothy Williamson have recently defended a view that constitutes a challenge to this idea. This is the view that ‘knowledge-how is simply a species of knowledge-that’.27 Consider a standard case: say, Ellen’s knowledge of how to play the violin. On Stanley’s and Williamson’s view, there is some suitable way for Ellen to play the violin such that what she knows is that this is a way for her to play the violin. And that makes her knowledge effable. One way to express it would be: first, to get Ellen to play the violin; then to draw attention to the way in which she does so; and then to say, ‘This is a way for Ellen to play the violin.’28 I agree with very much of what Stanley and Williamson say.29 Certainly I think they show their view to be, as they put it, ‘the default position’.30 I am not persuaded, 25 Ramsey’s quip is ‘What we can’t say we can’t say, and we can’t whistle it either’ (Ramsey (1931), p. 238). Ramsey is alluding to the Tractatus. He is also, as P.M.S. Hacker points out in his (2000), n. 5, alluding to Wittgenstein’s well-known expertise in whistling. 26 Moore (1997), Ch. 8. 27 Stanley and Williamson (2001), p. 411. 28 Admittedly, this would not display the precise modes of presentation involved in her knowledge. Thus suppose Ellen had not known how to play the violin, but had been shown someone else playing the violin in that way and had been told, ‘This is a way for you to play the violin.’ She would then have had knowledge—different knowledge—which could have been expressed in the same way: see Stanley and Williamson (2001), pp. 428–429. And the difference here is, precisely, a difference in mode of presentation. But this merely illustrates how low my standards of expressibility are. By these low standards, we should likewise have to say that, if Ellen woke up one day oblivious of who she was but aware that she was lying in hospital, her knowledge could be expressed by saying, ‘Ellen is lying in hospital’: cf. Moore (1997), p. 197. 29 I am gratified to find several echoes of their argument in my own argument for the view that nothing in the semantics of the phrase ‘knows how to’ provides ready proof of the existence of ineffable knowledge: see Moore (1997), Ch. 8, §1. This is why I cannot resist the following captious response to their claim, in Stanley and Williamson (2001), n. 47, that I use ‘the alleged distinction [between knowledge-how and knowledge-that] to argue that there is ineffable knowledge’. I would say rather that I argue (independently) that there is ineffable knowledge, some of which I identify as ‘knowledge-how’. 30 Stanley and Williamson (2001), p. 431.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

250

Language, World, and Limits

however, that there are no overriding considerations.31 And although I cannot mount a full defence of my own opposed view here, I shall try to do a little to motivate it. To this end, rather than give a direct reason for denying that ‘knowledge-how’ is a species of ‘knowledge-that’,32 I shall mention a related case, which I think makes similar trouble for Stanley’s and Williamson’s view. Consider my knowledge of what it is for an object to be green. On their view this is knowledge, concerning something, that that thing is what it is for an object to be green. But concerning what? A simple reply would be: ‘What it is for an object to be green.’ But what kind of thing is that? If I try to express my knowledge by indicating a green object and saying, ‘This is what it is for an object to be green’, what can I be referring to by ‘this’? There does not seem to me to be any good answer. Nothing short of an unacceptable Platonism, it seems to me, can subserve the extension of their account to this case.33 I do not think that my knowledge of what it is for an object to be green is knowledge that anything is the case. Nor, crucially, do I think that it is effable.34

5. Ineffable understanding and a reappraisal of the dispute between advocates of the traditional reading of the Tractatus and advocates of the new reading of it But suppose I am wrong about the existence of ineffable knowledge. Are there any other candidates for the title of non-trivial ineffability? I think there are. I have in mind certain states of understanding.35 It is clear that many states of understanding are effable; an example would be a solicitor’s understanding of some legal nicety that she has to explain to her clients. But it is far from clear that they all are. The case of understanding has a particular significance in view of how much attention has recently been devoted to Wittgenstein’s remark, in the penultimate section of the Tractatus:

31 For that matter, I am not persuaded that, even if their view is correct, it tells decisively against the possibility of ineffable knowledge. For even if ‘knowledge-how’ is a species of ‘knowledge-that’, some ‘knowledge-how’ may be ineffable because some ‘knowledge-that’ may be ineffable. (I was too quick to dismiss this possibility in Moore (1997): see Ch. 8, Appendix.) Here I am simply echoing the point that I made in n. 16, that some truths may be ineffable: necessary truths. Thus consider my knowledge of how to exercise the concept of greenness. And suppose, with Stanley and Williamson, that this is knowledge, concerning some way to exercise the concept of greenness, that it is a way to exercise the concept of greenness. Then it is knowledge (presumably) of what could not be otherwise. So it may be said to lack content, in which case, by my criteria of expression, nothing is to count as expressing it. (Cf. the Tractarian idea that ‘thoughts’ must have sense, which means that when, for example, I know that it is either raining or not raining—this example is borrowed from 4.461—my knowledge does not count as a thought: see 2.225–3, 3.13, 4, and 5.1362. This would accord very well with my own general conception of these matters, whereby knowledge is ineffable when it has nothing ‘to answer to’: see Moore (1997), Ch. 8, §3.) 32 This is done in Schiffer (2002) and Koethe (2002). 33 Cf. Moore (1997), pp. 134–135. 34 Cf. Wittgenstein (1967), Pt. I, §§75, 78, and 610. Note: this forces me to retract the claim I make about finite verbs in Moore (1997), p. 168. I would now modify that claim by adding ‘typically’. 35 In fact I think that states of understanding are states of knowledge. However, I would relinquish this if it could be shown that there was no room for ineffable knowledge. I would rather leave room for ineffable understanding. See Moore (1997), pp. 161 and 183 ff.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

251

Anyone who understands me eventually recognizes [my propositions] as nonsensical.36

This remark appears to be part of the ‘frame’ of the book, in which Wittgenstein intends to make a straightforwardly true claim about the nonsense that constitutes the bulk of the book inside that ‘frame’. He avoids paradox precisely by casting himself as the intended object of understanding, rather than the nonsense, which of course cannot be understood.37 He goes on to connect understanding him in this way with ‘seeing the world aright’.38 And this in turn suggests that, whatever else is going on in this deeply puzzling work, it is part of the design of the project that the reader who successfully grapples with it should be in some sense better off than he would have been had he not done so.39 The idea that the Tractatus has a ‘frame’, in which Wittgenstein at once dissociates himself from the rest of the book and proclaims the benefits of properly engaging with the rest of the book, finds particular favour with (and is particularly associated with) those who adopt the new reading of the Tractatus recently championed by Cora Diamond, James Conant, and others. This reading is pitted against what I shall call the traditional reading. On the traditional reading: The Tractatus consists mostly of nonsense because what Wittgenstein is trying to convey, about language and its limits, is, by its own lights, ineffable. The only way in which he can convey it—the only way in which he can get the reader to ‘see the world aright’—is by dint of a special kind of nonsense: what we might call ‘illuminating’ nonsense.40 On the new reading: There is nothing ineffable. There is only the temptation to see sense where it is lacking. Wittgenstein’s aim is therapeutic. The book consists mostly of nonsense because he is trying, by indulging this temptation, to eliminate it; by producing nonsense that appears to have sense, and then testing the appearance, to get the reader to acknowledge the illusion, so that the temptation disappears, and the reader is left realizing that the nonsense is precisely that: sheer lack of sense, which conveys nothing whatsoever.41 These readings look as if they could scarcely be further apart. But two things should be noted straight away. First, even on the traditional reading, there is no reason why we 36 6.54. Note that Wittgenstein’s use of the term ‘proposition’ is different from mine in sections 1 and 2 of this essay. I was using ‘proposition’ in something more like the way in which Wittgenstein uses ‘possible situation’—e.g. 2.202—or ‘sense’—e.g. 2.221. 37 See e.g. Conant (1991), p. 159; Conant (2000), p. 198; and Diamond (2000), pp. 150–151. 38 6.54. 39 But note that even this seemingly cautious formulation runs into some difficulties in view of the very first sentence of the preface to the book. (Those who see the book as having a ‘frame’ typically reckon the preface as part of the ‘frame’: see e.g. Conant (1991), p. 159 and Diamond (2000), p. 149.) The sentence in question reads, ‘Perhaps this book will be understood only by someone who has himself already had the thoughts that are expressed in it . . . ,’ (p. 3, my emphasis). (My concern here is with the word ‘already’. But of course, there is also a problem attaching to ‘the thoughts that are expressed in it’, which sits ill with the idea that the bulk of the book consists of nonsense.) For some comments relating to the tension between this sentence and what I say in the main text see Friedlander (2001), pp. 155 ff. See also Conant (1989), pp. 245–246. 40 This is Hacker’s term: see Hacker (1986), p. 18. 41 See e.g. Diamond (1991); Diamond (2000); Conant (1989); Conant (1991); Conant (2000); Witherspoon (2000); and Kremer (2001). For a powerful recoil in favour of the traditional reading see Hacker (2000). Poised somewhere in between are Reid (1998); McGinn (1999); Proops (2001); and Sullivan (2002).

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

252

Language, World, and Limits

should not be left realizing that the nonsense in the book is sheer lack of sense; in other words, that what makes it nonsense is simply that Wittgenstein has ‘failed to give a meaning to certain signs in his propositions’.42 (He could have stipulated, for instance, that he was using ‘object’ to mean ‘emerald’ and ‘simple’ to mean ‘green’. Had he done so, then 2.02—‘Objects are simple’—would not have been nonsense. It would have been true. As it stands, it is of a piece with ‘Omples are sibject.’) Secondly, even on the new reading, the nonsense in the book is supposed to serve a function. We are supposed to ‘see the world aright’ as a result of processing this nonsense. So it is not like the nonsense that a monkey at a typewriter might have produced. It has been carefully crafted to have certain effects on those who understand the language.43 (We do well to remember in this connection how exercised proponents of the new reading sometimes get, without any intimation of irony, about the best way to translate parts of this nonsense.44) Once we have taken due account of these two things, and once we have taken due account of the possibility of ineffable understanding, there are ways of construing the two readings whereby, to borrow a wonderful phrase of David Wiggins’ from a different context, ‘Suddenly it seems that what makes the difference between [them] has the width of a knife-edge.’45 The point is this. In overlooking the possibility that things other than truths are non-trivially ineffable,46 we are bound to see the two readings are irreconcilably differing about whether Wittgenstein believes there is anything (any truth) that is ineffable. But once we register that the state that we are supposed to get into, the state of understanding Wittgenstein, may itself be ineffable, then we can see the two readings quite differently. Where the traditional reading holds that, for 42 Here I am alluding to 6.53. The key relevant sections of the book, of which the traditional reading can take just as much note as the new reading, are 5.473 ff. 43 As Diamond says, Wittgenstein holds that the nonsense he has produced ‘may be useful or even for a time essential’: Diamond (1991), p. 181. Cf. Diamond (2000), pp. 158–159. (Hacker fastens on this in Hacker (2000), pp. 361–362, though, rather unfairly in my view, he ridicules the idea that the presence of such nonsense in a text can be adduced in support of any interpretation of that text.) 44 E.g. Witherspoon (2000), n. 13; cf. also Conant (2000), nn. 4 and 81. Not that this is a criticism. The idea of translating nonsense is not the absurdity it may sound, as witness the existence of French and German translations of Lewis Carroll’s famous nonsense-poem ‘Jabberwocky’ (see Hofstadter (1980), pp. 366–368 and 372–373 for references and discussion). The matter is further complicated by Conant’s suggestion that which parts of the Tractatus count as nonsense may depend partly on the reader: see Conant (2000), n. 102. 45 Wiggins (1995), pp. 327–328. 46 There is certainly a tendency, on both sides of the debate initiated by this new reading, to do precisely that. E.g. consider this: ‘[Wittgenstein] did think that one can mean something that cannot be said . . . [For example,] what the solipsist means is quite correct; only it cannot be said, but it makes itself manifest (Tractatus, 5.62) . . . So there are, according to the author of the Tractatus, ineffable truths that can be apprehended’: this is from Hacker (2000), p. 368, his emphasis. And consider this: ‘The proponents of the [traditional reading of the Tractatus] . . . see that the sentences they are attracted to are nonsense, but they still want to hold onto what (they imagine) the nonsense is trying to say. They conclude that [the Tractatus must convey] . . . . an “insight” into the truth of certain deep matters—even though, strictly speaking, this truth cannot be put into language. [But they are wrong.] Wittgenstein’s aim is to enable us to recognize that there is no ineffable “it” . . . ’: this is from Conant (1991), p. 160, his emphasis, some emphasis removed. (I have inserted the sentence ‘But they are wrong’ to make clear, what quoting this passage out of context would otherwise have made unclear, that the next sentence is asserted in propria persona.) For awareness of the possibility which I am saying tends to be overlooked see: Kremer (2001), §IX, in which I find a great deal of pleasing convergence with my own thinking about these matters; Proops (2001), pp. 378 ff., in which, however, the possibility is not always carefully enough distinguished for my liking; and Sullivan (2002), §2.4. See also McGinn (1999), about which I shall have more to say shortly.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

253

Wittgenstein, the nonsense in the Tractatus conveys ‘something’ ineffable, we can construe this occurrence of ‘something’ as embracing just such understanding. And where the new reading holds that, for Wittgenstein, the nonsense in the Tractatus conveys ‘nothing’, and ‘nothing’ is ineffable, we can construe these occurrences of ‘nothing’ as restricted in their range to truths. We can then look back at the sketches of the two readings proffered above and see each as entirely consonant with the other. I should emphasize, however, that I am not, in saying this, claiming to do full justice to the intentions of all those who advocate either of these readings. To claim that would be crazy. Perhaps what I have done is something more like what Marie McGinn does in a recent essay, where she proposes a third reading that is intended to retain the advantages of each of these two.47 She says that these two readings ‘offer us the unappealing alternative between reading Wittgenstein’s remarks as nonsense that conveys ineffable truths about the world and as nonsense that conveys nothing whatsoever’.48 She then advances her own reading, which is designed to enable us to escape this choice.49 But whether any such third reading is a genuine alternative to each of the original pair or something more like a charitable reconstruction of each of them or (least plausibly, I suppose) a compendium of what each of them actually comes to, I am certain that it deserves our serious attention.

6. An objection to the idea of ineffable understanding. The objection deflected Of course, this is no kind of vindication of Wittgenstein unless the idea of ineffable understanding itself deserves our serious attention. All that I have done so far is to advert to the possibility of such a thing. I cannot, in these confines, do much more. But in the three remaining sections of this essay I should like to deflect one natural objection to the idea, and then say a little more about how the idea relates to the Tractatus. The objection is an echo of an objection to the Lewisian idea of inexpressible truth with which we began. If, in fact, there is such a thing as ineffable understanding, the standards of effability that allow this fact to be so described make it (the fact) appear much more noteworthy than it really is. These standards, though relatively low, are still high enough for a state of understanding to count as ineffable just because it does not have content that makes it either true or false; or just because the best attempt to express it issues in something that does not have content that makes it either true or false. If these standards are relaxed, that is if the corresponding criteria of expression are relaxed, and if Wittgenstein’s aim in writing the Tractatus is realized, then we are at perfect liberty to say the following: that Wittgenstein’s understanding of language and its limits, and the reader’s eventual understanding of him, which is in effect the reader’s eventual shared understanding of language and its limits, are all expressed in the book. Moreover, if our standards of meaningfulness are similarly relaxed, we are also at liberty to say that the sentences in the book are, for that very reason, meaningful. 47 McGinn (1999): see esp. pp. 496–497. There is much in this essay with which I am sympathetic. There is even more that I admire. 48 Ibid. p. 498. 49 See the summary of her position in ibid. pp. 512–513.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

254

Language, World, and Limits

Indeed, if we think that Wittgenstein, in writing the Tractatus, intends his sentences to produce some particular effect in the reader by means of the recognition of this very intention, then we do not have to be all that Gricean50 to think that this just is the kind of thing that we ordinarily mean by (non-natural) meaning. And if the effect in question is partly that the reader should recognize Wittgenstein’s sentences as lacking meaning, then this gives an ironical twist to the accusation of self-stultification that is ordinarily levelled at Wittgenstein. The accusation of self-stultification that is ordinarily levelled at Wittgenstein is that his work subverts itself in such a way that he does not in the end mean anything by it. It now appears that, on the contrary, his work subverts itself in such a way that he does in the end mean something by it!51 I have two comments to make in response to this objection. First, concerning this last point, it is not clear that Wittgenstein does intend his sentences to produce some particular effect in the reader ‘by means of the recognition of this very intention’. Does he not rather intend his sentences to produce their effect by means of the reader’s ultimate failure, in trying to construe them as the network of truth-evaluable statements that they appear to be, to make anything of them? The second comment is that the objection can in any case be, at most, an objection to the terminology. Perhaps it is misleading to apply the word ‘ineffable’ to whatever does not have content that makes it either true or false.52 But the concept for which the word has been reserved is impervious to this objection. The concept is simply what it is. Nothing in the concept is intended, or would be able, to abnegate the myriad effects that a text can have on a reader, or the myriad uses to which words can be put, or the myriad ways in which a person can share his or her understanding with another person.

7. The understanding afforded by the Tractatus How then does the idea of ineffable understanding relate to the Tractatus? Is it an acceptable reading of the book to say that what it is supposed to subserve, namely understanding of Wittgenstein, is a case in point? Well, there is at least this much to be said for the reading. If there is such a thing as understanding Wittgenstein via the Tractatus, in the way that he intends, then any attempt to express this understanding (as opposed merely to talking about it) is liable to issue in the same kind of nonsense: ‘A picture cannot depict its pictorial form’, ‘Value lies outside the world’, et cetera.53, 54 And if it really is impossible to do any better 50 See Grice (1967). 51 As Wittgenstein himself observes, ‘If everything behaves as if a sign had meaning, then it does have meaning’ (3.328). 52 It would certainly be highly unorthodox to apply the word to my knowledge that aunts are female: see nn. 16 and 31. 53 Here it is worth remembering the context in the Tractatus in which Wittgenstein makes the following important claim: ‘In philosophy the question, “What do we actually use this word or this proposition for?” repeatedly leads to valuable insights.’ He does so when discussing our use of mathematical ‘propositions’, which in his view are really pseudo-propositions, nonsense just like the bulk of his own book (6.211). (For the idea that mathematics consists of nonsense see 6.2; see also 5.533–5.534.) A passage from Wittgenstein’s later work that is very pertinent here is Wittgenstein (1967), Pt. I, §§498–499. [Comment added for the reprint: See Essay 4 in this volume, written later than this essay, for some important qualifications to what I say in this note.] 54 These two examples are derived from 2.172 and 6.41 respectively.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

255

than this, when attempting to express the understanding, then it follows that the understanding cannot be expressed. But is this all that we can say about the understanding that is supposed to accrue from reading his book: that the closest we can get to expressing it is producing nonsense of the very kind we find in the book? If it is, that must make us feel uneasy. And we shall not feel much easier if all we can add to this is that such understanding includes a capacity to recognize the resultant nonsense as nonsense. This is too reminiscent of the sign that reads ‘Mind the plinth’ and that is on a plinth whose sole purpose is to support it. In fact, however, we can say more. If there is such a thing as understanding Wittgenstein via the Tractatus, in the way that he intends, then it includes a capacity to recognize as nonsense not only the nonsense in the Tractatus, but other, similar ‘transcendental twaddle’.55 It includes a capacity, ‘whenever someone . . . [wants] to say something metaphysical, to demonstrate to him that he [has] failed to give a meaning to certain signs in his propositions’.56 Nor should the significance of this be underestimated. For Wittgenstein, just as for Kant, our susceptibility to illusions of sense is as deep as our susceptibility to sense itself. It is almost as hard to imagine the latter without the former as it is to imagine the former without the latter.57 This means that our grasp of concepts comes to have two aspects. Alongside our knowledge of how to use them, there is our knowledge of how not to use them; alongside our command of what sense they can be used to make, there is our command of what ostensible sense they can be used to make. Each of these admits of degrees. Our command in the latter case can be more or less complete, just as our command in the former case can. Some illusions of sense are very superficial indeed (‘The square root of 2 is green’); some, a bit deeper (‘It is five o’clock on the sun’58); some, deep enough that a little reflection is needed to carry conviction that they are illusions of sense at all (‘It is five o’clock at the North Pole’); some, so deep that it can be a matter of unresolved controversy whether that is what they are (‘I have performed infinitely many tasks’59).60 The depth of some of these illusions means that making sense includes, for us, a continual struggle against merely seeming to make 55 This phrase is Wittgenstein’s own: see Hacker (1986), p. 104. 56 6.53; cf. 4.003. Cf. also McGinn (1999), pp. 502 ff. 57 Cf. 4.002. (Cf. also Wittgenstein (1967), Pt. I, §111 and the material from ‘Big Typescript’ to which Hacker refers in Hacker (1996a), pp. 112–113.) Kant would go further. He would say that we can no more free ourselves of such illusions than we can prevent the moon from appearing larger to us when it is nearer the horizon: see Kant (1998), A297–298/B354–355. This is because, on Kant’s view, we have an ineliminable urge to transcend the very limitations that enable us to make sense of things in the first place: see Kant (1998), Avii–viii and A642/B670. Cf. Cavell (1979), p. 109, where he writes, ‘Nothing is more human than the wish to deny one’s humanity,’ a theme adopted by Jonathan Weiss in his (2001). Cf. also what I call ‘our aspiration to be infinite’, in Moore (1997), Ch. 11. 59 See Moore (1989–90). 58 This example is taken from Wittgenstein (1967), Pt. I, §350.

60 Note: these examples, consisting as they do of solitary sentences, should not be allowed to obscure how context-dependent the making of sense is. For each of these sentences we can imagine a context in which, even without special stipulation, an application of that sentence would make sense. But I am thinking of applications of them in more standard contexts. (See further Mackie (1981) and Angene (1982): I side with Lyle Angene.) Note also: the contextual factors that determine whether an application of a sentence makes sense may include the applications of other sentences, some of which purport to be consequences of it and some of which purport to have it as a consequence. This means that, whenever a fairly complex text contains

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

256

Language, World, and Limits

sense and against the perplexity that merely seeming to make sense can cause. For instance, although each of us knows what it is for something to be green, this is partial knowledge, and, at least inasmuch as we are prone to philosophical reflection, it is constantly having to be re-earned against the lure of certain incoherent pictures of what it is for something to be green: most notably, against the lure of the kind of transcendental idealism that we find in the Tractatus, whereby what it is for something to be green is for the fact that that thing is green to be part of ‘my’ world—if not to be part of ‘me’.61 It follows that the capacity to recognize ‘transcendental twaddle’ for what it is, which is part of the understanding that is supposed to accrue from reading the Tractatus, is, for us, a precondition of the very capacity to make sense. But can we say some more about such understanding, and in particular about why it is ineffable?

8. Nonsense as the vehicle for affording such understanding Consider this. On the austere view of nonsense which is fostered by the Tractatus, and which, to the best of my knowledge, is endorsed by all those who adopt the new reading, nonsense is only ever sheer lack of sense.62 On this view, the judgement that something is nonsense is always a judgement about the actual history, to date, of some particular sign (‘No meaning has so far been given to this sign’). Such a judgement is always empirical, provisional, and metalinguistic. However, the discussion above suggests that the judgement that something is nonsense is sometimes none of these things. It suggests that the judgement is sometimes arrived at by reflection on concepts; and that it is therefore most appropriately expressed by using, not mentioning, the corresponding signs (‘There is no such thing as its being five o’clock at the North Pole’). Does this mean that the austere view must be rejected? I do not think so. I think the claim that something is nonsense is always empirical, provisional, and metalinguistic. The semantics and rules of a language can determine what makes sense. They cannot, except by default, determine what fails to make sense. It is true that the rules of a language can proscribe as well as prescribe. But the proscriptions are so to speak boundary marks for the prescriptions. They register where the prescriptions cease to have application. Thus there are rules determining what ‘o’clock’ it is at different points on the surface of the earth for each of the earth’s orientations with respect to the sun; and, as far as these go, no meaning attaches to the sentence ‘It is five o’clock at the North Pole.’ So there is a sense in which, as far as these rules go, we are not allowed to say that. Anyone intending to apply these rules who does say that is rather like someone intending to play chess who moves his rook diagonally. We can say that such a person has made an illegal move. But we can also say that such a person has failed some nonsense, there is liable to be a radical indeterminacy about precisely where the nonsense is to be located. In particular this is true of the Tractatus. See again Conant (2000), n. 102, cited above in n. 44. 61 See 1.1, 5.62, and 5.63—according to which, respectively, ‘The world is the totality of facts’, ‘The world is my world’, and ‘I am my world’ Wittgenstein’s emphasis. Cf. also Kant (1998), A129. I try to say some more about the lure of transcendental idealism in Moore (1997), Chs. 6–9. 62 See again 5.473 ff., to which I referred in n. 42.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

257

to make any move at all. An illegal move is not a special kind of move. It is as if he has poured coffee all over the board.63 And whether what he has done constitutes a move in any other actual game is a matter of brute fact. The judgement that it does not is an empirical, provisional judgement about what games there are. The point, however, is this. Whether what he has done constitutes a move in chess, which is what he intends, is not a matter of brute fact. That is determined by the rules of chess.64 Likewise, whether any meaning attaches to the sentence ‘It is five o’clock at the North Pole’, as far as our ordinary chronometry goes, is not a matter of brute fact. It is determined by the relevant concepts. And where illusions of sense are concerned, there are always relevant concepts: there are always relevant intentions about which concepts are being exercised. So reflection on those concepts is required to recognize the illusions as illusions. That is, reflection on those concepts is required to recognize that, at least as far as they go, no meaning has so far been given to these signs.65 Someone might say, ‘This is all very well. But you have still not explained how it can be appropriate to express such recognition—the recognition that something is a mere illusion of sense—by using, rather than mentioning, the relevant signs. Why is it not just as nonsensical to say, “There is no such thing as its being five o’clock at the North Pole” as it would be to say, “There is no such things as its being frumptiliously quirxaceous”?’ A number of (mutually incompatible) responses suggest themselves.66 First response: The reason why it is appropriate to express the recognition that something is a mere illusion of sense by using the relevant signs, rather than mentioning them, is that what is recognized is parasitic on the illusion. If I say, ‘There is no such thing as its being five o’clock at the North Pole’, then I am doing something like what Gareth Evans would say I am doing when I make a singular negative existential statement: I am denying that there is anything that we understand when we have a certain illusion of understanding, and I am identifying the illusion by re-creating it. You cannot understand what I am saying without ‘entering into’ the illusion.67  Second response: It is not appropriate to express the recognition that something is a mere illusion of sense by using, rather than mentioning, the relevant signs. But nor, contrary to appearances, do we do that. If I say, ‘There is no such thing as its being 63 Cf. in this connection Wittgenstein (1967), Pt. I, §500. 64 Cf. in this connection Percival (1994), p. 191. 65 Hacker (2000), pp. 365–367, is very pertinent here. I agree with much of what he says though I think that he is unfair to Diamond. See also Sorensen (2002), §7. This contains an account of ‘the understanding of nonsense’—as a representation of an absence of meaning—which appears straightforwardly incompatible with the austere view. In fact, however, Sorensen is presupposing a context that makes what he is talking about something like the recognition of the illusion of sense that I am talking about. There is no (obvious) incompatibility with the austere view. 66 Cf. in what follows Baker and Hacker (1980), p. 480. Cf. also Anscombe (1971), pp. 85–86. 67 Evans (1982), Ch. 10, esp. §10.4. (And see in particular pp. 344 ff. for the importance to Evans’ account of our using, not mentioning, the relevant singular term.) Jonathan Weiss, in his (2001), expressly rejects any such application of Evans’ ideas to this sort of case on the grounds that, where nonsense is concerned, there is no analogue of what Evans calls ‘the game-to-reality shift’: see Evans (1982), p. 362 and Weiss (2001), pp. 233 ff. But since, on Evans’ view, the simple use of a non-referring singular term itself issues in a kind of nonsense (Evans (1982), Ch. 9, e.g. p. 338), it seems to me that Weiss is in effect rejecting, not merely the application of Evans’ ideas to this sort of case, but Evans’ ideas themselves.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

258

Language, World, and Limits

five o’clock at the North Pole,’ then I am mentioning, not using, the expression ‘five o’clock at the North Pole’—just as, if I say, ‘By 1857 she had become George Eliot,’ then, despite the surface grammar of what I say, I am mentioning, not using, the name ‘George Eliot’.  Third response: It is indeed as nonsensical to say, ‘There is no such thing as its being five o’clock at the North Pole’ as it would be to say, ‘There is no such thing as its being frumptiliously quirxaceous.’ But this is still the best we can do when trying to express our recognition of that illusion—the point being that our recognition of that illusion, as of any other similar illusion, is incapable of being expressed. It is of a piece with the ineffable understanding afforded by the Tractatus. I shall not try to arbitrate between these responses. Nor shall I speculate on what others may be available. I shall simply record (what may come as a surprise) that I am uncomfortable with the third. I find the postulation of that link between ineffability and nonsense overly crude. True, there is pressure on us elsewhere to say something along the same lines. Imagine that Henry is in the grip of this particular illusion and says, ‘It is five o’clock at the North Pole.’ Then a very natural way for us (to try) to register Henry’s mistake is to say, ‘Henry thinks that it is five o’clock at the North Pole.’ Yet without some ancillary gloss, it would appear that the nonsensicality of Henry’s assertion infects ours.68 This suggests that the best we can do to express our understanding of Henry—I do not, of course, say our understanding of Henry’s assertion— is to produce nonsense, which in turn entails that our understanding of Henry is ineffable.69 However, I think it would be very rash to accede to this without further ado. It is not at all clear to me that we cannot make room for the straightforward truth of an assertion of ‘Henry thinks that it is five o’clock at the North Pole’, perhaps along the lines of the second response above. (It is not at all clear to me, for that matter, that we cannot make some room for the straightforward truth of an assertion, in simple exegetical mode, of ‘Wittgenstein holds that objects are simple.’ Indeed, it is not at all clear to me—though this raises additional complications—that we cannot make some room for the straightforward falsity of an assertion, in simple exegetical mode, of ‘Wittgenstein holds that facts are simple.’) I recoil from the third response, then. Nevertheless I do ultimately want to endorse something of that sort. The discussion above is focused on one particular, rather superficial illusion of sense. But suppose we turn our attention to an altogether more fundamental family of illusions to which we are subject. I have in mind illusions that manifest the urge we have, itself ill-conceived, to transcend our limitations: illusions that arise when we try to apply concepts that are adapted to these limitations as though they were not, indeed

68 Cf. 5.5422. 69 There is something very like this train of thought in Diamond (2000), §4 and p. 161, though Diamond does not herself talk in terms of ineffable understanding, and I am closer to her than the remarks that I am about to make in the main text suggest. Cf. also Sorensen (2002), p. 173, where he writes, ‘I can think you have a [particular] meaningless thought only if I think I have the same meaningless thought.’ (Sorensen, who is arguing for the possibility of meaningless thought, sees this conditional as apt for an application of modus ponens. Others might see it as apt for an application of modus tollens.)

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Ineffability and Nonsense

259

as though they were not adapted to any limitations at all. An example, I suggest, is the illusion that attaches to the sentence ‘The world exists as a limited whole.’70 Consider the understanding that someone might have of this general phenomenon, and the associated capacity to recognize the illusions in this family as illusions. Can that understanding be expressed? Will not the attempt to express it involve producing more of the very nonsense in question? Will it not involve trying to transcend the relevant limitations in an effort to stake off the territory that is home to the illusions; and perhaps also trying to implement some form of transcendental idealism whereby we cannot talk about anything that is not part of ‘our’ world, a world that is itself in some mysterious way bound by these limitations? In sum, will it not involve trying ‘to draw a limit to thought’—where this is something that cannot be done unless ‘both sides of the limit [are] thinkable’?71 I ask these questions rhetorically. I realize that much more would be required to demonstrate that they have the answers implicit in my asking them. But if they do, then surely they provide a fair characterization of the project of the Tractatus. The illusions of sense in question are both the target of the book and the content of the book, and we, the readers, are supposed to understand the project—are supposed to understand Wittgenstein—by discovering that we cannot in the end make sense of the book. It falls apart in our hands. The very process of watching it fall apart is what brings us to our understanding of how these illusions arise,72 an understanding which, if the closest we can come to expressing it is producing just such nonsense, is ineffable. To the extent that we achieve this understanding, we have, as I put it earlier, a command of the nonsense. We know how not to use the relevant concepts, as well as how to use them. And our grasp of those concepts is thereby fuller than it would have been had we not been tempted to extend them in this way and had we only ever had the latter knowledge.73 We are masters of a technique: the practice of doing what Wittgenstein characterizes as philosophy. This is an activity rather than a body of doctrine.74 And our mastery is ineffable. It seems to me, then, that we can say all of the following about the Tractatus: • that it is written to convey something ineffable; • that what it consists of, for the most part, is what would result from a (necessarily forlorn) attempt to express what it is written to convey; 70 Cf. n. 57. Cf. also, in connection with this particular illusion, Plato (1961c), 144e–145a; Kant (1998), A503–505/B531–533; 6.45; and Murdoch (1993), p. 1. 71 P. 3. 72 Cf. Crary (2000), p. 13. (It is tempting to say that, just as a proposition shows its sense—4.022—so too a piece of nonsense shows its lack of sense. Cf. Conant (1989), n. 39, where he writes, ‘The only thing [nonsensical] propositions can show . . . [is] their nonsensicality.’ In fact, however, this is just more nonsense.) 73 Cf. Jonathan Lear’s comment in his (1984), p. 240: ‘Post-neurotic consciousness is fundamentally more complex than a healthy consciousness that has never suffered disease or cure.’ There is also a Biblical echo here: see Genesis, Ch. II, vv. 16–17, and Ch. III, vv. 2–5. 74 4.111–4.116. (Wittgenstein talks of philosophy as aiming at the clarification of thoughts (4.112). I take it that the clarification of thoughts includes the elimination, where necessary by re-casting the thoughts, of associated impulses to nonsense (e.g. 4.115; cf. 3.323–3.325). This is all connected with the powerful conception of thought that the Tractatus engenders, whereby what can be thought at all can be thought with complete clarity (4.116).) The phrase ‘masters of a technique’ is a deliberate allusion to Wittgenstein (1967), Pt. I, §§150 and 199.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

260

Language, World, and Limits

hence • that what it consists of, for the most part, is nonsense; • that the reader is supposed to grasp what it conveys by processing this nonsense, and in particular by eventually recognizing it as nonsense; • that recognizing this nonsense as nonsense means recognizing it as sheer lack of sense which says nothing about how things are; and finally • that not only is grasping what the book conveys supposed to be quite compatible with, it is supposed to demand, utter impatience with the idea of ineffable thoughts, or ineffable truths. Whether this is more in line with the traditional reading or the new reading— whether, perhaps, it is more in line with something intermediate à la McGinn (see section 5)—I shall leave to others who may be interested in the question to judge.75

75 I am very grateful to Jane Heal, Julie Jack, Marie McGinn, Howard Robinson, and Peter Sullivan for comments that helped me to improve this essay.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography Al-Azm, Sadik  J. 1972. The Origins of Kant’s Arguments in the Antinomies. Oxford: Oxford University Press. Allison, H.E. 1983. Kant’s Transcendental Idealism: An Interpretation and Defense. New Haven: Yale University Press. Alston, William P. 1956. ‘Ineffability’. In The Philosophical Review, 65. Angene, Lyle E. 1982. ‘Five O’Clock Here’. In Analysis, 42. Anscombe, G.E.M. 1971. An Introduction to Wittgenstein’s Tractatus: Themes in the Philosophy of Wittgenstein. Bristol: Thoemmes Press. Aristotle, 1941a. Categories, trans. E.M. Edghill. In his The Basic Works of Aristotle, ed. Richard McKeon. New York: Random House. —— 1941b. Physics, trans. R.P. Hardie and R.K. Gaye. In his The Basic Works of Aristotle, ed. Richard McKeon. New York: Random House. —— 1941c. Metaphysics, trans. W.D.  Ross. In his The Basic Works of Aristotle, ed. Richard McKeon. New York: Random House. Arrington, Robert L. and Glock, Hans-Johann (eds) 1996. Wittgenstein and Quine. London: Routledge. Austin, J.L. 1970. ‘Truth’. Reprinted in his Philosophical Papers, ed. J.O.  Urmson and G.J. Warnock. Oxford: Oxford University Press. —— 1975. How To Do Things With Words, ed. J.O. Urmson and Marina Sbisà. Oxford: Oxford University Press. Ayer, A.J. 1971. Language, Truth and Logic, 2nd edn. Harmondsworth: Penguin. —— 1992. ‘Reply to Michael Dummett’. In L.E. Hahn (ed.), The Philosophy of A.J. Ayer. La Salle: Open Court. Baker, G.P. and Hacker, P.M.S. 1980. Wittgenstein: Understanding and Meaning, Volume 1 of An Analytical Commentary on the Philosophical Investigations. Oxford: Blackwell. —— 1985. Wittgenstein: Rules, Grammar and Necessity, Volume 2 of An Analytical Commentary on the Philosophical Investigations. Oxford: Blackwell. Barker, S.F. and Achinstein, Peter 1960. ‘On the New Riddle of Induction’. In The Philosophical Review, 69. Bennett, Jonathan 1974. Kant’s Dialectic. Cambridge: Cambridge University Press. —— 1982. ‘The Age and Size of the World’. Reprinted in Ralph C.S. Walker (ed.), Kant on Pure Reason. Oxford: Oxford University Press. —— 1994. ‘Descartes’ Theory of Modality’. In The Philosophical Review, 103. Bergström, Lars 1993. ‘Quine, Underdetermination, and Skepticism’. In The Journal of Philosophy, 90. Bird, Graham 2006. The Revolutionary Kant: A Commentary on the Critique of Pure Reason. La Salle: Open Court. —— 2011. ‘Replies to my Critics’. In Kantian Review, 16. Black, Max 1964. A Companion to Wittgenstein’s ‘Tractatus’. Ithaca: Cornell University Press. Blackburn, Simon 1990. ‘Wittgenstein’s Irrealism’. In Wittgenstein: Eine Neuerberwehrung, ed. Johannes L. Brandl and Rudolf Haller. Vienna: Holder- Richler-Temsky. —— 1994. ‘Enchanting Views’. In Peter Clark and Bob Hale (eds), Reading Putnam. Oxford: Blackwell. Boghossian, P.A. 1990. ‘The Status of Content’. In The Philosophical Review, 99.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

262

Bibliography

Boswell, James 1887. The Life of Samuel Johnson, LL.D. Oxford: Oxford University Press. Boyer, Carl B. 1949. The History of the Calculus and its Conceptual Development. New York: Dover. Brewer, Bill 1995. ‘Compulsion by Reason’. In Proceedings of the Aristotelian Society, Supp. Vol. 69. Bronzo, Silver 2017. ‘Wittgenstein, Theories of Meaning, and Linguistic Disjunctivism’. In European Journal of Philosophy, 25. Brueuckner, Anthony 1986. ‘Brains in a Vat’. In The Journal of Philosophy, 83. —— 1990. ‘Skepticism About Knowledge of Content’. in Mind, 99. Burge, Tyler 1995. ‘Our Entitlement to Self-Knowledge’. In Proceedings of the Aristotelian Society, Supp. Vol. 69 (printed in the ordinary Proceedings, 96). Burnyeat, M.F. 1982. ‘Idealism and Greek Philosophy: What Descartes Saw and Berkeley Missed’. In Godfrey Vesey (ed.), Idealism Past and Present. Cambridge: Cambridge University Press. Butterfield, Jeremy 1985. ‘Indexicals and Tense’. In Ian Hacking (ed.), Exercises in Analysis: Essays by Students of Casimir Lewy. Cambridge: Cambridge University Press. —— 1995. ‘Quantum Theory and the Mind: Worlds, Minds and Quanta’. In Proceedings of the Aristotelian Society, Supp. Vol. 69. Campbell, John 1994. Past, Space, and Self. Cambridge, Mass.: The MIT Press. —— 2002. Reference and Consciousness. Oxford: Oxford University Press. Carnap, Rudolf 1956. ‘Empiricism, Semantics, and Ontology’. Reprinted as Appendix A in his Meaning and Necessity: A Study in Semantics and Modal Logic, 2nd edn. Chicago: University of Chicago Press. —— 1990. ‘Quine on Analyticity’. In Richard Creath (ed.), Dear Carnap, Dear Van: The QuineCarnap Correspondence and Related Work. Berkeley and Los Angeles: The University of California Press. Cavell, Stanley 1979. The Claim of Reason. Oxford: Oxford University Press. Cerf, Walter 1972. ‘Review of The Bounds of Sense’. In Mind, 81. Chalmers, David J., Manley, David, and Wasserman, Ryan 2009. Metametaphysics: New Essays on the Foundations of Ontology. Oxford: Oxford University Press. Child, William 1994. ‘On the Dualism of Scheme and Content’. In Proceedings of the Aristotelian Society, 94. Cockburn, David 1997. Other Times: Philosophical Perspectives on Past, Present and Future. Cambridge: Cambridge University Press. Conant, James 1989. ‘Must We Show What We Cannot Say?’. In Richard Fleming and Michael Payne (eds), The Senses of Stanley Cavell. Lewisburg: Bucknell University Press. —— 1991. ‘The Search for Logically Alien Thought: Descartes, Kant, Frege, and the Tractatus’. In Philosophical Topics, 20. —— 2000. ‘Elucidation and Nonsense in Frege and Early Wittgenstein’. In Alice Crary and Rupert Read (eds), The New Wittgenstein. London: Routledge. Cooper, David E. 1991. ‘Ineffability’. In Proceedings of the Aristotelian Society, Supp. Vol. 65. Craig, Edward 1990. ‘Davidson and the Sceptic: The Thumbnail Version’. In Analysis, 50. Crary, Alice 2000. ‘Introduction’. In Alice Crary and Rupert Read (eds), The New Wittgenstein. London: Routledge. Dancy, Jonathan 1993. Moral Reasons. Oxford: Blackwell. Davidson, Donald 1980. ‘Criticism, Comment, and Defence’. Reprinted in his Essays on Actions and Events. Oxford: Oxford University Press. —— 1984a. ‘Truth and Meaning’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography

263

—— 1984b. ‘Quotation’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984c. ‘On Saying That’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984d. ‘Moods and Performatives’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984e. ‘Thought and Talk’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984f. ‘Reply to Foster’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984g. ‘On the Very Idea of a Conceptual Scheme’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984h. ‘The Method of Truth in Metaphysics’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984i. ‘The Inscrutability of Reference’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1984j. ‘What Metaphors Mean’. Reprinted in his Inquiries into Truth and Interpretation. Oxford: Oxford University Press. —— 1986. ‘A Coherence Theory of Truth and Knowledge’. Reprinted in Ernest LePore (ed.), Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson. Oxford: Blackwell. —— 1987. ‘Knowing One’s Own Mind’. In Proceedings and Addresses of the American Philosophical Association, 60. —— 1991. ‘Epistemology Externalized’. In Dialectica, 45. —— 1992. ‘The Second Person’. In Midwest Studies in Philosophy, 17. —— 2001. ‘Indeterminism and Antirealism’. Reprinted in his Subjective, Intersubjective, Objective. Oxford: Oxford University Press. de Beistegui, Miguel 2004. Truth and Genesis: Philosophy as Differential Ontology. Bloomington: Indiana University Press. Deleuze, Gilles 1990a. Expressionism in Philosophy: Spinoza, trans. Martin Joughin. New York: Zone Books. —— 1990b. The Logic of Sense, trans. Mark Lester and Charles Stivale and ed. Constantin V. Boundas. New York: Columbia University Press. —— 1994. Difference and Repetition, trans. Paul Patton. New York: Columbia University Press. —— 2005. Francis Bacon: The Logic of Sensation, trans. Daniel W. Smith. London: Continuum. —— 2006. Nietzsche and Philosophy, trans. Hugh Tomlinson. London: Continuum. Derrida, Jacques 1981. Dissemination, trans. B. Johnson. London: Athlone Press. —— 1982. ‘Signature, Event, Context’. In his Margins of Philosophy, trans. Alan Bass. Brighton: Harvester Press. —— 1987. The Post Card: From Socrates to Freud and Beyond, trans. Alan Bass. Chicago: The University of Chicago Press. —— 1988. ‘Limited Inc a b c . . . ’, trans. S.  Weber. Reprinted in his Limited Inc, ed. G.  Graff. Evanston: Northwestern University Press. Descartes, René 1984. Meditations on First Philosophy. In The Philosophical Writings of Descartes, Vol. 2, trans. John Cottingham, Robert Stoothoff, and Dugald Murdoch. Cambridge: Cambridge University Press. —— 1985. Principles of Philosophy. In The Philosophical Writings of Descartes, Vol. 1, trans. John Cottingham, Robert Stoothoff, and Dugald Murdoch. Cambridge: Cambridge University Press.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

264

Bibliography

—— 1991. ‘Letter to Mersenne, 26 May 1630’. In his The Philosophical Writings of Descartes, Vol. 3: The Correspondence, trans. John Cottingham, Robert Stoothoff, Dugald Murdoch, and Anthony Kenny. Cambridge: Cambridge University Press. Diamond, Cora 1991. ‘Throwing Away the Ladder: How to Read the Tractatus’. Reprinted in her The Realistic Spirit: Wittgenstein, Philosophy, and the Mind. Cambridge, Mass.: The MIT Press. —— 1999. ‘How Old Are These Bones? Putnam, Wittgenstein and Verification’. In Proceedings of the Aristotelian Society, Supp. Vol. 73. —— 2000. ‘Ethics, Imagination and the Method of Wittgenstein’s Tractatus’. Reprinted in Alice Crary and Rupert Read (eds), The New Wittgenstein. London: Routledge. —— 2011. ‘The Tractatus and the Limits of Sense’. In Oskari Kuusela and Marie McGinn (eds), The Oxford Handbook of Wittgenstein. Oxford: Oxford University Press. —— 2019. ‘Wittgenstein and What Can Only be True’. Reprinted with amendments in her Reading Wittgenstein with Anscombe, Going on to Ethics. Cambridge, Mass.: Harvard University Press. Dostoevsky, Fyodor 1982. The Brothers Karamazov, trans. David Magarshack. Harmondsworth: Penguin. Dummett, Michael 1973. Frege: Philosophy of Language. London: Duckworth. —— 1975. ‘What is a Theory of Meaning?’. In Samuel Guttenplan (ed.), Mind and Language. Oxford: Oxford University Press. —— 1976. ‘What is a Theory of Meaning? (II)’. In Gareth Evans and John McDowell (eds), Truth and Meaning: Essays in Semantics. Oxford: Oxford University Press. —— 1978a. ‘Truth’. Reprinted in his Truth and Other Enigmas. London: Duckworth. —— 1978b. ‘The Philosophical Basis of Intuitionistic Logic’. Reprinted in his Truth and Other Enigmas. London: Duckworth. —— 1978c. ‘A Defence of McTaggart’s Proof of the Unreality of Time’. Reprinted in his Truth and Other Enigmas. London: Duckworth. —— 1978d. ‘The Reality of the Past’. Reprinted in his Truth and Other Enigmas. London: Duckworth. —— 1978e. ‘The Significance of Quine’s Indeterminacy Thesis’. Reprinted in his Truth and Other Enigmas. London: Duckworth. —— 1981a. Frege: Philosophy of Language, 2nd edn. London: Duckworth. —— 1981b. The Interpretation of Frege’s Philosophy. London: Duckworth. —— 1992. ‘The Metaphysics of Verificationism’. In L.E. Hahn (ed.), The Philosophy of A.J. Ayer. La Salle: Open Court. —— 1993. ‘Realism and Anti-Realism’. Reprinted in his The Seas of Language. Oxford: Oxford University Press. —— 1997. ‘Wang’s Paradox’. Reprinted in Rosanna Keefe and Peter Smith (eds), Vagueness: A Reader. Cambridge, Mass.: The MIT Press. —— 2004. Truth and the Past. New York: Columbia University Press. Duns Scotus, John 1987. Philosophical Writings, trans. Allan Wolter. Indianapolis, Ind.: Hackett Publishing. Edgington, Dorothy 2004. ‘Two Kinds of Possibility’. In Proceedings of the Aristotelian Society, Supp. Vol. 78. Eilan, Naomi 1994–95. ‘The First Person Perspective’. In Proceedings of the Aristotelian Society, 95. Einstein, Albert 1960. Relativity: The Special and the General Theory, trans. Robert W. Lawson. London: Methuen. Evans, Gareth 1982. The Varieties of Reference, ed. John McDowell. Oxford: Oxford University Press.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography

265

—— 1985a. ‘Understanding Demonstratives’. Reprinted in his Collected Papers. Oxford: Oxford University Press. —— 1985b. ‘Does Tense Logic Rest Upon a Mistake?’. In his Collected Papers. Oxford: Oxford University Press. —— 1985c. ‘Molyneux’s Question’. In his Collected Papers. Oxford: Oxford University Press. Evans, Gareth and McDowell, John (eds) 1976a. Truth and Meaning: Essays in Semantics. Oxford: Oxford University Press. —— 1976b. ‘Introduction’. In their (eds), Truth and Meaning: Essays in Semantics. Oxford: Oxford University Press. Everett, Hugh 1956. ‘“Relative State” Formulation of Quantum Mechanics’. In Reviews of Modern Physics, 29. Falvey, Kevin and Owens, Joseph 1994. ‘Externalism, Self-Knowledge and Skepticism’. In The Philosophical Review, 103. Fichte, J.G. 1956. The Vocation of Man, ed. and trans. Roderick  M.  Chisholm. Indianapolis: Bobbs-Merrill. Field, Hartry 1972. ‘Tarski’s Theory of Truth’. In The Journal of Philosophy, 69. —— 1975. ‘Conventionalism and Instrumentalism in Semantics’. In Noûs, 9. Fitch, F.B. 1963. ‘A Logical Analysis of Some Value Concepts’. In Journal of Symbolic Logic, 28. Fleming, G.N. 1995. ‘Quantum Theory and the Mind’. In Proceedings of the Aristotelian Society, Supp. Vol. 69. Fogelin, Robert J. 1987. Wittgenstein. London: Routledge & Kegan Paul. Forbes, Graeme 1995. ‘Realism and Skepticism: Brains in a Vat Revisited’. In The Journal of Philosophy, 92. Frege, Gottlob 1967. ‘The Thought: A Logical Inquiry’, trans. A.M.  and Marcelle Quinton. Reprinted in P.F. Strawson (ed.), Philosophical Logic. Oxford: Oxford University Press. —— 1997a. ‘On Concept and Object’, trans. Peter Geach. Reprinted with revisions in his The Frege Reader, ed. Michael Beaney. Oxford: Blackwell. —— 1997b. ‘Notes for Ludwig Darmstaedter’, trans. Peter Long and Roger White. Reprinted in his The Frege Reader, ed. Michael Beaney. Oxford: Blackwell. Fricker, Miranda 2001. ‘Confidence and Irony’. In Edward Harcourt (ed.), Morality, Reflection, and Ideology. Oxford: Oxford University Press. Friedlander, Eli 2001. Signs of Sense: Reading Wittgenstein’s Tractatus. Cambridge, Mass.: Harvard University Press. Friedman, Michael 1975. ‘Physicalism and Indeterminacy of Translation’. In Noûs, 9. Genova, A.C. 1991. ‘Craig on Davidson: A Thumbnail Refutation’. In Analysis, 51. George, Alexander 1993. ‘How Not to Refute Realism’. In The Journal of Philosophy, 90. Gibson, R.F. 1986. ‘Translation, Physics and Facts of the Matter’. In L.E. Hahn and P.A. Schilpp (eds), The Philosophy of W.V. Quine. La Salle: Open Court. Goddard, L. and Routley, R. 1966. ‘Use, Mention and Quotation’. In The Australasian Journal of Philosophy, 44. Gödel, Kurt 1967. ‘On Formally Undecidable Propositions of Principia Mathematica and Related Systems I’, trans. Jean van Heijenoort. In Jean van Heijenoort (ed.), From Frege to Gödel: A Source Book in Mathematical Logic, 1879–1931. Cambridge, Mass.: Harvard University Press. Goldstein, Laurence 1984. ‘Quotation of Types and Other Types of Quotation’. In Analysis, 44. Gomes, Anil 2011. ‘Is Kant’s Transcendental Deduction of the Categories Fit for Purpose?’. In Kantian Review, 15. Goodman, Nelson 1975. ‘Words, Works, Worlds’. In Erkenntnis, 9. —— 1978. Ways of Worldmaking. Indianapolis, Ind.: Hackett Publishing.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

266

Bibliography

Grice, H.P. 1967. ‘Meaning’. Reprinted in P.F.  Strawson (ed.), Philosophical Logic. Oxford: Oxford University Press. Guyer, Paul 1987. Kant and the Claims of Knowledge. Cambridge: Cambridge University Press. Hacker, P.M.S. 1972. Insight and Illusion: Wittgenstein on Philosophy and the Metaphysics of Experience. Oxford: Oxford University Press. —— 1986. Insight and Illusion: Themes in the Philosophy of Wittgenstein. Oxford: Oxford University Press. —— 1996a. Wittgenstein’s Place in Twentieth-Century Analytic Philosophy. Oxford: Blackwell. —— 1996b. ‘Wittgenstein and Quine: Proximity at Great Distance’. In Robert L. Arrington and Hans-Johann Glock (eds), Wittgenstein and Quine. London: Routledge. —— 2000. ‘Was he Trying to Whistle it?’. In Alice Crary and Rupert Read (eds), The New Wittgenstein. London: Routledge. Haddock, Adrian and Macpherson, Fiona (eds) 2008. Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Heal, Jane 1989. Fact and Meaning: Quine and Wittgenstein on Philosophy of Language. Oxford: Blackwell. Heidegger, Martin 1962a. Being and Time, trans. John Macquarrie and Edward Robinson. Oxford: Blackwell. —— 1962b. Kant and the Problem of Metaphysics, trans. James  S.  Churchill. Bloomington: Indiana University Press. —— 1982. The Basic Problems of Phenomenology, trans. Albert Hofstadter. Bloomington: Indiana University Press. —— 1999. Contributions to Philosophy (From Enowning), trans. Parvis Emad and Kenneth Maly. Bloomington: Indiana University Press. Heil, John 1988. ‘Privileged Access’. In Mind, 97. Henig, Uri unpublished. ‘Meaning and Negation’. Oxford M.Litt. thesis. Henrich, Dieter 1982. ‘The Proof-Structure of Kant’s Transcendental Deduction’. Reprinted in Ralph C.S. Walker (ed.), Kant on Pure Reason. Oxford: Oxford University Press. Hilbert, David 1967. ‘On the Infinite’, trans. Stefan Bauer-Mengelberg. In Jean van Heijenoort (ed.), From Frege to Gödel: A Source Book in Mathematical Logic, 1879–1931. Cambridge, Mass.: Harvard University Press. Hintikka, Jaakko 1958. ‘On Wittgenstein’s “Solipsism”’. In Mind, 67. Hofstadter, Douglas R. 1980. Gödel, Escher, Bach: An Eternal Golden Braid. Harmondsworth: Penguin. Holiday, Anthony 1985. ‘Wittgenstein’s Silence: Philosophy, Ritual and the Limits of Language’. In Language and Communication, 5. Hookway, Christopher 1988. Quine: Language, Experience and Reality. Oxford: Polity Press. —— 1995. ‘Fallibilism and Objectivity: Science and Ethics’. In J.E.J. Altham and Ross Harrison (eds), World, Mind, and Ethics: Essays on the Ethical Philosophy of Bernard Williams. Cambridge: Cambridge University Press. —— 1996. ‘Perspicuous Representations’. In Robert L. Arrington and Hans-Johann Glock (eds), Wittgenstein and Quine. London: Routledge. Horwich, Paul 1990. Truth. Oxford: Blackwell. Husserl, Edmund 1962. Ideas: General Introduction to Pure Phenomenology, trans. W.R. Boyce Gibson. New York: Collier. Hylton, Peter 1982. ‘Analyticity and Indeterminacy of Translation’. In Synthese, 52. Janik, Allan and Toulmin, Stephen 1973. Wittgenstein’s Vienna. New York: Touchstone. Jardine, Nicholas 1980. ‘The Possibility of Absolutism’. In D.H. Mellor (ed.), Science, Belief, and Behaviour: Essays in Honour of R.B. Braithwaite Cambridge: Cambridge University Press.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography

267

—— 1995. ‘Science, Ethics, and Objectivity’. In J.E.J. Altham and Ross Harrison (eds), World, Mind, and Ethics: Essays on the Ethical Philosophy of Bernard Williams. Cambridge: Cambridge University Press. Johnston, Colin 2007. ‘Symbols in Wittgenstein’s Tractatus’. In European Journal of Philosophy, 15. —— 2017. ‘The Picture Theory’. In Hans-Johann Glock and John Hyman (eds), A Companion to Wittgenstein. Oxford: Wiley Blackwell. Kant, Immanuel 1933. Critique of Pure Reason, trans. Norman Kemp Smith. London: Macmillan. —— 1950. Prolegomena to Any Future Metaphysics, trans. Lewis White Beck. Indianapolis and New York: Bobbs-Merrill. —— 1956. Critique of Practical Reason, trans. Lewis White Beck. Indianapolis and New York: Bobbs-Merrill. —— 1998. Critique of Pure Reason, trans. and ed. Paul Guyer and Allen W. Wood. Cambridge: Cambridge University Press. —— 1999. Correspondence, trans. and ed. Arnulf Zweig. Cambridge: Cambridge University Press. —— 2000. Critique of the Power of Judgment, ed. Paul Guyer and trans. Paul Guyer and Eric Matthews. Cambridge: Cambridge University Press. —— 2002. Prolegomena to any Future Metaphysics that will be able to Come Forward as a Science, trans. Gary Hatfield. In his Theoretical Philosophy After 1781, ed. Henry Allison and Peter Heath. Cambridge: Cambridge University Press. Keefe, Rosanna and Smith, Peter (eds) 1997. Vagueness: A Reader. Cambridge, Mass.: The MIT Press. Kierkegaard, Søren 1954. Fear and Trembling, under the pseudonym Johannes de Silentio, trans. Walter Lowrie. Princeton: Princeton University Press. Kline, Morris 1972. Mathematical Thought from Ancient to Modern Times. Oxford: Oxford University Press. Koethe, John 2002. ‘Stanley and Williamson on Knowing How’. In The Journal of Philosophy, 99. Korsgaard, Christine M. 1996. The Sources of Normativity. Cambridge: Cambridge University Press. Kremer, Michael 2001. ‘The Purpose of Tractarian Nonsense’. In Noûs, 35. —— 2002. ‘Mathematics and Meaning in the Tractatus’. In Philosophical Investigations, 25. Kripke, Saul 1980. Naming and Necessity. Oxford: Blackwell. Lear, Jonathan 1982. ‘Leaving the World Alone’. In The Journal of Philosophy, 79. —— 1984. ‘The Disappearing “We”’. In Proceedings of the Aristotelian Society, Supp. Vol. 58. Leibniz, G.W. 1973. ‘Monadology’, trans. Mary Morris and G.H.R. Parkinson. In his Philosophical Writings, ed. G.H.R. Parkinson. London: Dent. Le Poidevin, Robin 1998. ‘The Past, Present, and Future of the Debate About Tense’. In Robin Le Poidevin (ed.), Questions of Time and Tense. Oxford: Oxford University Press. —— 1998–99. ‘Egocentric and Objective Time’. In Proceedings of the Aristotelian Society, 99. Le Poidevin, Robin and MacBeath, Murray (eds) 1993. The Philosophy of Time. Oxford: Oxford University Press. Leslie, John 2001. Infinite Minds: A Philosophical Cosmology. Oxford: Oxford University Press. Lewis, David 1983. ‘Attitudes De Dicto and De Se’. Reprinted in his Philosophical Papers, Vol. I. Oxford: Oxford University Press. —— 1986. On the Plurality of Worlds. Oxford: Blackwell. —— 1990. ‘What Experience Teaches’. In  W.G.  Lycan (ed.), Mind and Cognition. Oxford: Blackwell. Lovibond, Sabina 1983. Realism and Imagination in Ethics. Oxford: Blackwell.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

268

Bibliography

Lowe, E.J. 1987. ‘The Indexical Fallacy in McTaggart’s Proof of the Unreality of Time’. In Mind, 96. —— 1992. ‘McTaggart’s Paradox Revisited’. In Mind, 101. Lucas, J.R. 1998. ‘Transcendental Tense’. In Proceedings of the Aristotelian Society, Supp. Vol. 72. MacBeath, Murray 1983. ‘Mellor’s Emeritus Headache’. In Ratio, 25. McDaniel, Kris 2009. ‘Ways of Being’. In David  J.  Chalmers, David Manley, and Ryan Wasserman (eds), Metametaphysics: New Essays on the Foundations of Ontology. Oxford: Oxford University Press. McDowell, John 1977. ‘On the Sense and Reference of a Proper Name’. In Mind, 86. —— 1981. ‘Non-Cognitivism and Rule-Following’. In S.H. Holtzmann and C.M. Leich (eds), Wittgenstein: To Follow a Rule. London: Routledge & Kegan Paul. —— 1986a. ‘Singular Thought and the Extent of Inner Space’. In Philip Pettit and John McDowell (eds), Subject, Thought, and Context. Oxford: Oxford University Press. —— 1986b. ‘Critical Notice of Bernard Williams’ Ethics and the Limits of Philosophy’. In Mind, 95. —— 1993. ‘Wittgenstein on Following a Rule’. In  A.W.  Moore (ed.), Meaning and Reference. Oxford: Oxford University Press. —— 1994. Mind and World. Cambridge, Mass.: Harvard University Press. McGinn, Colin 1982. ‘The Structure of Content’. In Andrew Woodfield (ed.), Thought and Object. Oxford: Oxford University Press. —— 1983. The Subjective View: Secondary Qualities and Indexical Thoughts. Oxford: Oxford University Press. —— 1984. Wittgenstein on Meaning: An Interpretation and Evaluation. Oxford: Blackwell. McGinn, Marie 1999. ‘Between Metaphysics and Nonsense: Elucidation in Wittgenstein’s Tractatus’. In Philosophical Quarterly, 49. McGuiness, Brian 1966. ‘The Mysticism of the Tractatus’. In The Philosophical Review, 75. Mackie, J.L. 1981. ‘Five O’Clock on the Sun’. In Analysis, 41. McManus, Denis 2013. ‘Ontological Pluralism and the Being and Time Project’. In Journal of the History of Philosophy, 51. McPherson, Thomas 1955. ‘Religion as the Inexpressible’. In Antony Flew and Alasdair MacIntyre (eds), New Essays in Philosophical Theology. London: SCM Press. McTaggart, J.M.E. 1993. ‘The Unreality of Time’. Reprinted in Robin Le Poidevin and Murray MacBeath (eds), The Philosophy of Time. Oxford: Oxford University Press. Malcolm, Norman 1958. Ludwig Wittgenstein: A Memoir. Oxford: Oxford University Press. Melia, Joseph 1991. ‘Anti-Realism Untouched’. In Mind, 100. Mellor, D.H. 1983. ‘MacBeath’s Soluble Aspirin’. In Ratio, 25. —— 1993. ‘The Unreality of Tense’. Reprinted in Robin Le Poidevin and Murray MacBeath (eds), The Philosophy of Time. Oxford: Oxford University Press. —— 1998. ‘Transcendental Tense’. In Proceedings of the Aristotelian Society, Supp. Vol. 72. Moore, A.W. 1984. ‘Possible Worlds and Diagonalization’. In Analysis. 44. —— 1985. ‘Set Theory, Skolem’s Paradox and the Tractatus’. In Analysis, 45. —— 1988. ‘Aspects of the Infinite in Kant’. In Mind, 97. —— 1989–90. ‘A Problem for Intuitionism: The Apparent Possibility of Performing Infinitely Many Tasks in a Finite Time’. In Proceedings of the Aristotelian Society, 90. —— 1991. ‘Can Reflection Destroy Knowledge?’. In Ratio (New Series), 4. —— 1996. ‘Philosophy of Logic’. In Nicholas Bunnin and E.P. Tsui-James (eds), The Blackwell Companion to Philosophy. Oxford: Blackwell. —— 1997. Points of View. Oxford: Oxford University Press. —— 1999. ‘Misplaced Celebrations?: Reply to Mark Sacks’ Critical Notice’. In International Journal of Philosophical Studies, 7.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography

269

—— 2003. ‘Williams on Ethics, Knowledge, and Reflection’. In Philosophy, 78. —— 2012. The Evolution of Modern Metaphysics: Making Sense of Things. Cambridge: Cambridge University Press. —— 2013. ‘Was the Author of the Tractatus a Transcendental Idealist?’. In Peter Sullivan and Michael Potter (eds), Wittgenstein’s Tractatus: History and Interpretation. Oxford: Oxford University Press. Moore, G.E. 1959. ‘Wittgenstein’s Lectures in 1930–33’. In his Philosophical Papers. London: George Allen & Unwin. Morris, Michael 2008. Routledge Philosophy GuideBook to Wittgenstein and the Tractatus. London: Routledge. Mounce, H.O. 1981. Wittgenstein’s Tractatus: An Introduction. Oxford: Blackwell. Moyal-Sharrock, Danièle 2007. ‘The Good Sense of Nonsense: A Reading of Wittgenstein’s Tractatus as Non-Self-Repudiating’. In Philosophy, 82. Murdoch, Iris 1993. Metaphysics as a Guide to Morals. Harmondsworth: Penguin. Nagel, Thomas 1980. ‘The Limits of Objectivity’. In The Tanner Lectures on Human Values: 1980. Cambridge: Cambridge University Press. —— 1983. ‘The Objective Self ’. In Carl Ginet and Sydney Shoemaker (eds), Knowledge and Mind: Essays in Honour of Norman Malcolm. Oxford: Oxford University Press. —— 1986. The View from Nowhere. Oxford: Oxford University Press. —— 1997. The Last Word. Oxford: Oxford University Press. Nemirow, Laurence 1990. ‘Physicalism and the Cognitive Role of Acquaintance’. In W.G. Lycan (ed.), Mind and Cognition. Oxford: Blackwell. Nerlich, Graham 1965. ‘Presupposition and Entailment’. In American Philosophical Quarterly, 2. Newton, Isaac 1947. Mathematical Principles of Natural Philosophy, trans. Andrew Motte and Florian Cajori. Berkeley: The University of California Press. Nietzsche, Friedrich 1967. The Will to Power, trans. Walter Kaufmann and R.J. Hollingdale and ed. Walter Kaufmann. New York: Random House. Parfit, Derek 2011. On What Matters, Volume Two. Oxford: Oxford University Press. Peacocke, Christopher 1995. ‘Entitlement, Self-Knowledge and Conceptual Redeployment’. In Proceedings of the Aristotelian Society, Supp. Vol. 69 (printed in the ordinary Proceedings, 96). Pears, David 1971. Wittgenstein. London: Fontana. Penrose, Roger 1989. The Emperor’s New Mind: Concerning Computers, Minds, and the Laws of Physics. Oxford: Oxford University Press. Percival, Philip 1992. ‘Thank Goodness That’s Non-Actual’. In Philosophical Papers, 21. —— 1994. ‘Absolute Truth’. In Proceedings of the Aristotelian Society, 94. Perry, John 1993. ‘The Problem of the Essential Indexical’. Reprinted in his The Problem of the Essential Indexical and Other Essays. Oxford: Oxford University Press. Pettit, Philip and McDowell, John (eds) 1986. Subject, Thought, and Context. Oxford: Oxford University Press. Plato 1961a. Euthyphro, trans. Lane Cooper. In his The Collected Dialogues of Plato, Including the Letters, ed. Edith Hamilton and Huntington Cairns. Princeton: Princeton University Press. —— 1961b. Theaetetus, trans. F.M. Cornford. In his The Collected Dialogues of Plato, Including the Letters, ed. Edith Hamilton and Huntington Cairns. Princeton: Princeton University Press. —— 1961c. Parmenides, trans. F.M. Cornford. In his The Collected Dialogues of Plato, Including the Letters, ed. Edith Hamilton and Huntington Cairns. Princeton: Princeton University Press. —— 1961d. Sophist, trans. F.M. Cornford. In his The Collected Dialogues of Plato, Including the Letters, ed. Edith Hamilton and Huntington Cairns. Princeton: Princeton University Press.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

270

Bibliography

—— 1961e. Letters, trans. L.A. Post. In his The Collected Dialogues of Plato, Including the Letters, ed. Edith Hamilton and Huntington Cairns. Princeton: Princeton University Press. Plutarch 1976. Platonic Questions. In his Moralia, Vol. 13, Pt. I, trans. Harold Cherniss. Cambridge, Mass.: Harvard University Press. Price, Huw 2009. ‘Metaphysics After Carnap: The Ghost Who Walks?’ In David J. Chalmers, David Manley, and Ryan Wasserman (eds), Metametaphysics: New Essays on the Foundations of Ontology. Oxford: Oxford University Press. Priest, Graham 1986. ‘Tense and Truth Conditions’. In Analysis, 46. Prior, A.N. 1959. ‘Thank Goodness That’s Over’. In Philosophy, 34. Proops, Ian 2001. ‘The New Wittgenstein: A Critique’. In European Journal of Philosophy, 9. Putnam, Hilary 1981. Reason, Truth and History. Cambridge: Cambridge University Press. —— 1983. ‘Models and Reality’. Reprinted in Paul Benacerraf and Hilary Putnam (eds), Philosophy of Mathematics: Selected Readings. Cambridge: Cambridge University Press. —— 1992. Renewing Philosophy. Cambridge, Mass.: Harvard University Press. —— 1994. ‘Comments and Replies’. In Peter Clark and Bob Hale (eds), Reading Putnam. Oxford: Blackwell. Quine, W.V. 1960. Word and Object. Cambridge, Mass.: The MIT Press. —— 1961a. ‘On What There Is’. Reprinted in his From a Logical Point of View: LogicoPhilosophical Essays. Cambridge, Mass.: Harvard University Press. —— 1961b. ‘Two Dogmas of Empiricism’. Reprinted in his From a Logical Point of View: LogicoPhilosophical Essays. Cambridge, Mass.: Harvard University Press. —— 1969a. ‘Ontological Relativity’. Reprinted in his Ontological Relativity and Other Essays. New York: Columbia University Press. —— 1969b. ‘Epistemology Naturalized’. Reprinted in his Ontological Relativity and Other Essays. New York: Columbia University Press. —— 1969c. ‘Propositional Objects’. Reprinted in his Ontological Relativity and Other Essays. New York: Columbia University Press. —— 1969d. ‘Reply to Chomsky’. In Donald Davidson and Jaakko Hintikka (eds), Words and Objections: Essays on the Work of W.V. Quine. Dordrecht: Reidel. —— 1970a. Philosophy of Logic. Englewood Cliffs, N.J.: Prentice-Hall. —— 1970b. ‘On the Reasons for the Indeterminacy of Translation’. In The Journal of Philosophy, 67. —— 1974. The Roots of Reference. La Salle: Open Court. —— 1975. ‘The Nature of Natural Knowledge’. In Samuel Guttenplan (ed.), Mind and Language. Oxford: Oxford University Press. —— 1977. ‘Facts of the Matter’. In R.W. Shahan and K.R. Merrill (eds), American Philosophy From Edwards to Quine. Norman: University of Oklahoma Press. —— 1981a. ‘Things and Their Place in Theories’. In Theories and Things. Cambridge, Mass.: Harvard University Press. —— 1981b. ‘Empirical Content’. Reprinted in Theories and Things. Cambridge, Mass.: Harvard University Press. —— 1981c. ‘On the Very Idea of a Third Dogma’. In Theories and Things. Cambridge, Mass.: Harvard University Press. —— 1981d. ‘Five Milestones of Empiricism’. Reprinted in Theories and Things. Cambridge, Mass.: Harvard University Press. —— 1981e. ‘On Austin’s Method’. Reprinted in Theories and Things. Cambridge, Mass.: Harvard University Press. —— 1981f. ‘Goodman’s Ways of Worldmaking’. Reprinted in Theories and Things. Cambridge, Mass.: Harvard University Press. —— 1981g. ‘What Price Bivalence?’ In The Journal of Philosophy, 78.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography

271

—— 1986a. ‘Reply to Roger F. Gibson’. In L.E. Hahn and P.A. Schilpp (eds), The Philosophy of W.V. Quine. La Salle: Open Court. —— 1986b. ‘Reply to Gilbert H. Harman’. In L.E. Hahn and P.A. Schilpp (eds), The Philosophy of W.V. Quine. La Salle: Open Court. —— 1986c. ‘Reply to Hilary Putnam’. In L.E. Hahn and P.A. Schilpp (eds), The Philosophy of W.V. Quine. La Salle: Open Court. —— 1986d. ‘Reply to Paul  A.  Roth’. In L.E.  Hahn and P.A.  Schilpp (eds), The Philosophy of W.V. Quine. La Salle: Open Court. —— 1986e. ‘Reply to Jules Vuilleman’. In L.E. Hahn and P.A. Schilpp (eds), The Philosophy of W.V. Quine. La Salle: Open Court. —— 1987a. ‘Excluded Middle’. In his Quiddities: An Intermittently Philosophical Dictionary. Harmondsworth: Penguin. —— 1987b. ‘Gödel’s Theorem’. In his Quiddities: An Intermittently Philosophical Dictionary. Harmondsworth: Penguin. —— 1987c. ‘Knowledge’. In his Quiddities: An Intermittently Philosophical Dictionary. Harmondsworth: Penguin. —— 1987d. ‘Use Versus Mention’. In his Quiddities: An Intermittently Philosophical Dictionary. Harmondsworth: Penguin. —— 1990a. Pursuit of Truth. Cambridge, Mass.: Harvard University Press. —— 1990b. ‘Comment on Stroud’. In R.B. Barrett and R.F. Gibson (eds), Perspectives on Quine. Oxford: Blackwell. —— 1992. Pursuit of Truth, rev. edn. Cambridge, Mass.: Harvard University Press. —— 1995. From Stimulus to Science. Cambridge, Mass.: Harvard University Press. Quinn, Warren 1993. ‘Reflection and the Loss of Moral Knowledge: Williams on Objectivity’. Reprinted in his Morality and Action. Cambridge: Cambridge University Press. Ramsey, F.P. 1931. ‘General Propositions and Causality’. In R.B. Braithwaite (ed.), F.P. Ramsey: The Foundations of Mathematics. London: Routledge and Kegan Paul. Reid, Lynette 1998. ‘Wittgenstein’s Ladder: The Tractatus and Nonsense’. In Philosophical Investigations, 21. Rorty, Richard 1971–72. ‘Indeterminacy of Translation and of Truth’. In Synthese, 23. —— 1980. Philosophy and the Mirror of Nature. Oxford: Blackwell. —— 1982. ‘The World Well Lost’. Reprinted in his The Consequences of Pragmatism. Brighton: Harvester Press. —— 1991a. ‘Is Natural Science a Natural Kind?’. Reprinted in his Objectivity, Relativism and Truth: Philosophical Papers Volume I. Cambridge: Cambridge University Press. —— 1991b. ‘Pragmatism, Davidson and Truth’. Reprinted in his Objectivity, Relativism and Truth: Philosophical Papers Volume I. Cambridge: Cambridge University Press. Rovane, Carol 2009. ‘Did Williams Find the Truth in Relativism?’. In Daniel Callcut (ed.), Reading Bernard Williams. London: Routledge. Ryle, Gilbert 1949. The Concept of Mind. London: Hutchinson. Sacks, Mark 1989. The World We Found: The Limits of Ontological Talk. London: Duckworth. Sainsbury, R.M. 1997. ‘Concepts Without Boundaries’. Reprinted in Rosanna Keefe and Peter Smith (eds), Vagueness: A Reader. Cambridge, Mass.: The MIT Press. Schiffer, Stephen 2002. ‘Amazing Knowledge’. In The Journal of Philosophy, 99. Schopenhauer, Arthur 1969. The World as Will and Representation, trans. E.F.J.  Payne. New York: Dover. Searle, J.R. 1977. ‘Reiterating the Differences: A Reply to Derrida’. In Glyph, 1. Sellars, Wilfrid 1963. ‘Empiricism and the Philosophy of Mind’. Reprinted in his Perception and Reality. London: Routledge & Kegan Paul.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

272

Bibliography

Shields, Christopher 2012. ‘Being Qua Being’. In Christopher Shields (ed.), The Oxford Handbook of Aristotle. Oxford: Oxford University Press. Sider, Theodore 2011. Writing the Book of the World. Oxford: Oxford University Press. Smith, Daniel  W. 2001. ‘The Doctrine of Univocity: Deleuze’s Ontology of Immanence’. In Mary Bryden (ed.), Deleuze and Religion. London: Routledge. Sorensen, Roy A. 2002. ‘Meaningless Beliefs and Mates’s Problem’. In American Philosophical Quarterly, 39. Spinoza, Benedictus de 2002. Ethics. In his Spinoza: Complete Works, ed. Michael L. Morgan and trans. Samuel Shirley. Indianapolis, Ind.: Hackett Publishing. Standish, Paul 1992. Beyond the Self: Wittgenstein, Heidegger and the Limits of Language. Aldershot: Avebury. Stanley, Jason and Williamson, Timothy 2001. ‘Knowing How’. In The Journal of Philosophy, 98. Stevens, Wallace 1954. Collected Poems. London: Faber and Faber. Strawson, Galen 1989. The Secret Connexion: Causation, Realism, and David Hume. Oxford: Oxford University Press. Strawson, P.F. 1959. Individuals: An Essay in Descriptive Metaphysics. London: Methuen. —— 1966. The Bounds of Sense: An Essay on Kant’s Critique of Pure Reason. London: Methuen. —— 1993. ‘On Referring’. Reprinted in A.W.  Moore (ed.), Meaning and Reference. Oxford: Oxford University Press. Stroud, Barry 1984. The Significance of Philosophical Scepticism. Oxford: Oxford University Press. —— 2000. The Quest for Reality: Subjectivism and the Metaphysics of Colour. Oxford: Oxford University Press. Sullivan, Peter  M. 1999–2000. ‘The Totality of Facts’. In Proceedings of the Aristotelian Society, 100. —— 2002. ‘On Trying to be Resolute: A Response to Kremer on the Tractatus’. In European Journal of Philosophy, 10. —— 2003. ‘Ineffability and Nonsense’. In Proceedings of the Aristotelian Society, Supp. Vol. 77. Swinburne, Richard 1990. ‘Tensed Facts’. In American Philosophical Quarterly, 22. Taylor, Charles 1979. ‘The Validity of Transcendental Arguments’. In Proceedings of the Aristotelian Society, 79. —— 1985. Human Agency and Language. Cambridge: Cambridge University Press. Teichmann, Roger 1998. ‘Is a Tenseless Language Possible?’. In Philosophical Quarterly, 48. Tonner, Philip 2007. ‘Duns Scotus’ Concept of the Univocity of Being: Another Look’. In Pli, 18. Tooley, Michael 1997. Time, Tense, and Causation. Oxford: Oxford University Press. Travis, Charles 1998. ‘Sublunary Intuitionism’. In Grazer Philosophische Studien, 55, a special issue entitled ‘New Essays on the Philosophy of Michael Dummett’, ed. Johannes L. Brandl and Peter M. Sullivan. Turetzky, Philip 1998. Time. London: Routledge. Turner, Jason 2010. ‘Ontological Pluralism’. In The Journal of Philosophy, 107. van Inwagen, Peter 2002. ‘Generalizations of Homophonic Truth-Sentences’. In Richard Schantz (ed.), What is Truth? Berlin: de Gruyter. van Inwagen, Peter 2009. ‘Being, Existence, and Ontological Commitment’. In David J. Chalmers, David Manley, and Ryan Wasserman (eds), Metametaphysics: New Essays on the Foundations of Ontology. Oxford: Oxford University Press. Vanrie, Wim unpublished. ‘Logical Syntax and Nonsense in Wittgenstein’s Tractatus’. Oxford B. Phil. thesis. Voss, Stephen and Sayward, Charles 1976. ‘Eternal Sentences’. In The Australasian Journal of Philosophy, 54.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

Bibliography

273

Waismann, Friedrich 1965. The Principles of Linguistic Philosophy, ed. Rom Harré. London: Macmillan. Weiss, Jonathan 2001. ‘Illusions of Sense in the Tractatus: Wittgenstein and Imaginative Understanding’. In Philosophical Investigations, 24. Whiteley, C.H. 1957. ‘Names of Words: A Note on Analysis “Problem” No. 10’. In Analysis, 17. Wiggins, David 1995. ‘Categorical Requirements: Kant and Hume on the Idea of Duty’. In Rosalind Hursthouse, Gavin Lawrence, and Warren Quinn (eds), Virtues and Reasons: Philippa Foot and Moral Theory. Oxford: Oxford University Press. —— 1996. ‘Replies’. In Sabina Lovibond and S.G.  Williams (eds), Essays for David Wiggins: Identity, Truth and Value. Oxford: Blackwell. Williams, Bernard 1972. Morality: An Introduction to Ethics. Cambridge: Cambridge University Press. —— 1973. ‘Consistency and Realism’. Reprinted in his Problems of the Self: Philosophical Papers 1956–1972. Cambridge: Cambridge University Press. —— 1978. Descartes: The Project of Pure Enquiry. Harmondsworth: Penguin. —— 1981a. ‘Conflicts of Values’. Reprinted in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press. —— 1981b. ‘The Truth in Relativism’. Reprinted in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press. —— 1981c. ‘Wittgenstein and Idealism’. Reprinted in his Moral Luck: Philosophical Papers 1973–1980. Cambridge: Cambridge University Press. —— 1985. Ethics and the Limits of Philosophy. London: Routledge. —— 1995a. ‘Ethics and the Fabric of the World’. Reprinted in his Making Sense of Humanity and Other Philosophical Papers 1982–1993. Cambridge: Cambridge University Press. —— 1995b. ‘Which Slopes Are Slippery?’. Reprinted in his Making Sense of Humanity and Other Philosophical Papers: 1982–1993. Cambridge: Cambridge University Press. —— 1995c. ‘Replies’. In J.E.J. Altham and Ross Harrison (eds), World, Mind, and Ethics: Essays on the Ethical Philosophy of Bernard Williams. Cambridge: Cambridge University Press. —— 1996. ‘Truth in Ethics’. In Brad Hooker (ed.), Truth in Ethics. Oxford: Blackwell. —— 2006. ‘Tertullian’s Paradox’. Reprinted in his Philosophy as a Humanistic Discipline, ed. A.W. Moore. Princeton: Princeton University Press. Williams, S.G. 1996. ‘Ambiguity and Semantic Theory’. In Sabina Lovibond and S.G. Williams (eds), Essays for David Wiggins: Identity, Truth and Value. Oxford: Blackwell. Williamson, Timothy 1987. ‘Equivocation and Existence’. In Proceedings of the Aristotelian Society, 88. —— 1994a. Vagueness. London: Routledge. —— 1994b. ‘Never Say Never’. In Topoi, 13. —— 1996. ‘Knowing and Asserting’. In The Philosophical Review, 105. Witherspoon, Edward 2000. ‘Conceptions of Nonsense in Carnap and Wittgenstein’. In Alice Crary and Rupert Read (eds), The New Wittgenstein. London: Routledge. Wittgenstein, Ludwig 1961. Tractatus Logico-Philosophicus, trans. D.F. Pears and B.F. McGuiness. London: Routledge & Kegan Paul. —— 1967. Philosophical Investigations, trans. G.E.M. Anscombe, 3rd edn. Oxford: Blackwell. —— 1969a. The Blue Book. In his The Blue and Brown Books: Preliminary Studies for the ‘Philosophical Investigations’, 2nd edn. Oxford: Blackwell. —— 1969b. On Certainty, ed. G.E.M. Anscombe and G.H. von Wright and trans. Denis Paul and G.E.M. Anscombe. Oxford: Blackwell. —— 1974. Philosophical Grammar, ed. Rush Rhees and trans. Anthony Kenny. Oxford: Blackwell.

OUP CORRECTED PROOF – FINAL, 17/06/19, SPi

274

Bibliography

—— 1975. Philosophical Remarks, ed. Rush Rhees and trans. Raymond Hargreaves and Roger White. Oxford: Blackwell. —— 1978. Remarks on the Foundations of Mathematics, ed. G.H. von Wright, R. Rhees, and G.E.M. Anscombe and trans. G.E.M. Anscombe, 3rd edn. Oxford: Blackwell. —— 1979. Notebooks: 1914–1916, ed. G.H.  von Wright and G.E.M.  Anscombe and trans. G.E.M. Anscombe, 2nd edn. Oxford: Blackwell. —— 1980. Culture and Value, trans. Peter Winch. Oxford: Blackwell. —— 1981. Zettel, ed. G.E.M. Anscombe and G.H. von Wright and trans. G.E.M. Anscombe. Oxford: Blackwell. Wood, David and Bernasconi, Robert (eds) 1988. Derrida and ‘Différance’. Evanston: Northwestern University Press. Wright, Crispin 1979. Wittgenstein on the Foundations of Mathematics. London: Duckworth. —— 1987. ‘Anti-Realism, Timeless Truth and Nineteen Eighty-Four’. In his Realism, Meaning and Truth. Oxford: Blackwell. —— 1992. Truth and Objectivity. Cambridge, Mass.: Harvard University Press. —— 1994. ‘Putnam’s Proof that We are not Brains in a Vat’. In Peter Clark and Bob Hale (eds), Reading Putnam. Oxford: Blackwell. —— 1998. ‘Why Frege Did Not Deserve his Granum Salis: A Note on the Paradox of “The Concept Horse” and the Ascription of Bedeutungen to Predicates’. In Johannes L. Brandl and Peter Sullivan (eds), New Essays on the Philosophy of Michael Dummett, a special issue of Grazer Philosophische Studien, 55.

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Index absolute representations 1, 3–5, 45, 132, 158, 165 Abraham 222–3 affirmation 197–200, 207–8 analytic metaphysics 195–201; see also metaphysics analytic/synthetic distinction 17, 41–4, 68n.60, 81, 98, 100–1; see also a priori knowledge; necessity and possibility anti–realism 47–9, 73–4, 133, 136–7, 151n.29, 152–3, 241; see also bivalence and the law of the excluded middle; idealism; realism a priori knowledge 108–10, 110n.19, 113–14, 126–7, 229–30; see also analytic/ synthetic distinction; necessity and possibility Archimedean point 102 Archimedes, see Archimedean point Aristotle 196–7, 199–201, 206–7 Austin, J.L. 35, 47–8, 50n.46 Ayer, A.J. 97–9, 101 Beethoven, Ludwig 235–6, 249 Bennett, Jonathan 111n.28, 114n.13 Berkeley, George 102, 112n.3 Bird, Graham C8 bivalence and the law of the excluded middle 36, 47–9, 73–4, 115; see also anti–realism, truth–aptitude brains in vats 102, 119 Bronzo, Silver 63n.40 Campbell, John 102–3, 152n.33, 156n.50 Carnap, Rudolf 27, 37–8 Carroll, Lewis 252n.44 Cartesian souls 19–20, 22–3, 68–9, 112n.3 Cavell, Stanley 255n.57 Cerf, Walter 91 Christianity 222; see also religion Cockburn, David 144n.4, 144n.5 colour 4, 134, 141–2, 147, 160–4, 218–20; see also grue Conant, James 251, 252n.44, 252n.46, 259n.72 content, see scheme/content dualism Copernicus, Nicolaus 173n.38 Davidson, Donald 11, 15, 20, 39, 61n.30, 76–7, 123, 169, 170n.29, 210–12, 228, 248 Deleuze, Gilles 67n.58, 94n.13, 195–201, 208–9 Derrida, Jacques 47n.32, 48n.38

Descartes, René 47–9, 133–4, 166, 198n.10, 204n.43, 219n.25; see also Cartesian souls Diamond, Cora 57–61, 251, 258n.69 difference: as expressing the essence of being 199–209 see also univocity of being Dostoevsky, Fyodor 197–8 Dummett, Michael 20n.16, 21, 23–4, 42–3, 48n.39, 63n.42, 73–4, 99, 143n.3, 148n.20, 151n.29, 157n.52, 185, 187, 187n.27, 189–90, 190n.48 Duns Scotus, John 198 Edgington, Dorothy 93n.10 Einstein, Albert 44 Eliot, George 86, 257–8 empirical realism 72–3 ethics and evaluation 5, 35, 98–100, 136, 140n.18, 144n.5, 165–8, 171–5, 178–81, 189n.43, 197–8, 220, 222, 226, 231–5, 254–5; see also thick ethical concepts Euclid, see Euclidean geometry Euclidean geometry 175–6, 179, 182–3 Euthyphro question 222–3 evaluation, see ethics and evaluation Evans, Gareth 63n.42, 155n.41, 257 external questions, see internal versus external questions externalism about the mind 64, 121–4, 129–31; see also solipsism facts: as candidates for being perspectival 3–4, 143, 158–60, 226; see also tense as the constituents of reality 95, 101, 103, 151–3, 178, 186–7, 191 of the matter 14, 17–19, 22–5, 27, 33–4, 36, 81–2, 101n.44; see also indeterminacy/ truths as candidates for ineffability 1, 5–7, 130, 210–14, 226, 244–54, 250n.31, 260 Fichte, J.G. 227–8 finitude and infinitude 26n.35, 40, 93–4, 107, 113n.4, 116–17, 185, 189–91, 198–9, 216–17, 220–1, 227–8, 241–2, 245, 255–6 Fitch, F.B. 48n.37, 183n.9, 247n.15 Fogelin, Robert 221

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

276

Index

Frege, Gottlob 22–3, 63n.42, 66, 202n.36, 208–9, 234 future 80, 81, 82, 111, 117n.33, 150n.27, 152–3, 156, 158n.1, 179–80, 219–20; see also past; tense Gibson, Roger 35n.48 God 113, 133, 168–9, 185–6, 188–90, 196–8, 199n.12, 217–20, 222, 231, 234, 236; see also Euthyphro question; religion Gödel, Kurt, see Gödel’s theorem Gödel’s theorem 47, 148 Goodman, Nelson 177–80, 186n.22 Grice, H.P. 57–8, 253–4 grue 80–4 Guyer, Paul 111n.28 Hacker, P.M.S. 41n.8, 252n.43, 252n.46, 257n.65 Hegel, G.W.F. 222–3 Heidegger, Martin 169n.21, 196–7, 199–200, 204, 209n.58, 227–8 Heisenberg, Werner, see Heisenberg’s uncertainty principle Heisenberg’s uncertainty principle 148 Henrich, Dieter 155–6 Herz, Marcus 90 Hilbert, David 25n.33 Hintikka, Jaakko 234n.57 Hookway, Christopher 19n.13, 20n.11, 20n.12 Horwich, Paul 211n.5 Husserl, Edmund 62n.35 idealism 149, 176, 180, 185n.18 transcendental, see transcendental idealism see also anti–realism indeterminacy 13–14, 17, 39–40, 42, 122–3, 255n.60; see also facts of the matter; underdetermination/indeterminacy distinction ineffability: and its attempted expression 216–23, 235, 239–43, 254–5, 258–9 defined 211–12, 226–7, 245–8 facts/truths as candidates for, see facts/truths as candidates for ineffability states of knowledge/understanding as candidates for, see states of knowledge/ understanding as candidates for ineffability see also saying/showing distinction inexpressibility, see ineffability infinitude, see finitude and infinitude internal versus external questions 27, 37–8 Isaac 222 Johnson, Samuel 102 Johnston, Colin 62n.35, 201n.31

Kant, Immanuel 3–5, 13, 68, 75–6, 90–1, 94, 96–9, 103, 125n.19, 126, 182–3, 183n.9, 184n.11, 188–9, 191, 227–32, 234–6, 238–9, 255–6 his Transcendental Deduction 143–4, 153–7, 188, 191, 233–4 his mathematical antinomies 68–9, 107, 112 mathematics as construed in, see mathematics: as construed by Kant transcendental idealism in, see transcendental idealism in Kant unconditioned in, see unconditioned in Kant Kierkegaard, Søren 221n.32, 222–3 Kremer, Michael 57–61 knowledge 4–5, 75–87, 95–7, 99, 113, 116–17, 119, 138–9, 148, 162, 165, 185, 213–23, 227–34, 242–3, 247n.15, 250n.35 a priori, see a priori knowledge states of as candidates for ineffability, see states of knowledge/understanding as candidates for ineffability see also scepticism Korsgaard, Christine 168n.13 Lao Tze 221n.32 law of the excluded middle, see bivalence and the law of the excluded middle Lear, Jonatan 230n.30, 234–5, 238n.79, 259n.73 Leibniz, G.W.F. 153n.34, 189n.42; see also Leibnizians Leibnizians 109n.15 Leslie, John 185–6, 187n.27, 189, 189n.42 Lewis, David 183–5, 183n.6, 244–5, 253–4 logical positivism 24, 97–101, 103 logical syntax 62n.35, 66–7, 201–9; see also sign/symbol distinction in the Tractatus Löwenheim, Leopold, see Löwenheim–Skolem theorem Löwenheim–Skolem theorem 122–3 making sense 35, 96–8, 123, 196–200, 207–8, 215–20, 228–9, 237–8, 255–7, 255n.57, 259; see also sense: distinguished from nonsense; understanding mathematical antinomies, Kant’s, see Kant’s mathematical antinomies mathematics 14, 16, 18–19, 22–3, 25n.33, 36, 47, 65, 107n.4, 133–4, 137–8, 165, 195–6, 204 as construed by Kant 22–3, 113–14, 229–30 as construed by the later Wittgenstein 74–5, 78, 87–8, 93–4, 240–2 as construed in the Tractatus 57–8, 60–3, 63n.42, 254n.53 McDaniel, Kris 197n.7

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Index McDowell, John 68n.64, 96–7, 101–3, 129, 168–9, 240n.90 McGinn, Marie 253, 260 McTaggart, J.M.E. 146n.12 McPherson, Thomas 223n.35 meaning, theories of, see theories of meaning medieval philosophy 195–6, 198 Melia, Joseph 48n.37 Mellor, D.H. 147n.16 mention, see use/mention distinction metaphysics 40, 47–8, 87, 89, 98–100, 115, 137–8, 143, 144n.5, 159–60, 163, 175–6, 185, 185n.16, 189–90, 202–3, 255 analytic, see analytic metaphysics Morris, Michael 62n.35 Mounce, H.O. 232n.41 Murdoch, Iris 40 Nagel, Thomas 4, 126, 132 necessity and possibility 14, 52–4, 72–3, 80–4, 86–8, 93–5, 108, 110–11, 117n.36, 125–31, 125n.19, 133–4, 137–9, 153, 175–6, 183–6, 211n.7, 217–21, 226, 229–30, 232–3, 234n.57, 244–6, 248n.16, 248n.24, 250n.31; see also analytic/synthetic distinction; a priori knowledge Neurath, Otto 102 Newton, Isaac 44; see also Newtonians Newtonians 109n.15 Nicholas of Cusa 221n.32 Nietzsche, Friedrich 198–201, 207n.54, 208 nonsense: as the result of attempting to express the inexpressible, see ineffability: and its attempted expression distinguished from sense, see sense: distinguished from nonsense objective beliefs and ways of thinking, see absolute representations Parfit, Derek 204n.44 partial realism 49n.41 past 73–4, 111, 117n.33, 151n.29, 152–3, 155, 156, 157n.52; see also future; tense perspectival representations 1, 3–4, 44, 128–9, 132, 158, 225, 230–2 see also facts: as candidates for being perspectival; points of view physicalism 19–20, 25, 119n.1, 147, 161–2; see also physics physics 19–20, 22–5, 29, 33, 44, 147, 161–2, 182–3; see also physicalism Plato 47–8, 221n.32, 222, 224, 242–3; see also Platonism Platonism 41–3, 137–8, 241, 250

277

points of view 1, 3–5, 44–5, 128–9, 132, 134, 137, 139, 146n.13, 156n.47, 158, 167, 171–3, 175, 177–8, 180–1, 187–91, 213, 225, 228, 230, 232, 234–5 temporal, see temporal points of view see also facts: as candidates for being perspectival; perspectival representations positivism, see logical positivism possibility, see necessity and possibility Psalmists 221n.32 Putnam, Hilary 122–31, 175n.46 Quine, W.V. 12, 17, 41–4, 99–103, 119–20, 122–3, 169–72, 195n.1, 204, 205n.45, 205n.48 Quinn, Warren 174n.40 Ramsey, F.P. 249 realism 5, 73–4, 126, 156n.50, 165 empirical, see empirical realism partial, see partial realism see also anti–realism; idealism; Platonism religion 6, 137–8, 140–1, 211–12, 216–23; see also Christianity; God representations: absolute, see absolute representations perspectival, see perspectival representations see also rules of representation; truth–aptitude rules of representation 41–6, 47n.29, 52–4, 74–5, 137–8, 219 rules of, see rules of rules of representation rules of rules of representation 52–5 Rorty, Richard 24, 24n.27, 174n.40, 228–9 Rovane, Carol 186n.22 Ryle, Gilbert 204n.44 saying/showing distinction 60n.28, 71–2, 74–8, 87–8, 221n.31, 224; see also ineffability scepticism 4, 18–19, 26n.38, 31–2, 119–22, 132, 134, 140–1, 159–61 concerning the possibility that we are brains in vats, see brains in vats see also solipsism scheme/content dualism 169, 170n.25, 212n.9, 228–9 Schopenhauer, Arthur 227–8 Searle, John 47n.32 Sellars, Wilfrid 215n.17 sense: distinguished from nonsense 2–3, 35, 56, 90, 202–6, 202n.38, 253–60 making, see making sense see also truth–aptitude; univocity of being showing, see saying/showing distinction sign/symbol distinction in the Tractatus 66, 201–9; see also logical syntax

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

278

Index

Skolem, Thoralf, see Löwenheim–Skolem theorem Socrates 66–7, 222 solipsism 4, 119, 224–5, 227–8, 232n.41, 233–6; see also externalism about the mind, scepticism Sorensen, Roy 257n.65, 258n.69 sorites paradoxes 50, 52–5; see also vagueness space, see Euclidean geometry; spatio–temporal features of reality spatio–temporal features of reality 18–19, 71n.2, 72–3, 80, 82–4, 107, 112, 119–20, 127–9, 150n.27, 155–6, 175–6, 182–3, 188–9, 198–9, 225–6, 230–1; see also Euclidean geometry; future; past Spinoza, Benedictus de 198–201, 198n.9, 208 Stanley, Jason 214n.14, 249–50 states of knowledge/understanding as candidates for ineffability 1–3, 5–7, 103n.56, 208–9, 213–23, 226–7, 229–30, 249–60 Stevens, Wallace 180–1 Strawson, P.F. 48n.39, 90–1, 99, 103, 107n.3, 111n.28, 112n.3, 182–4, 187n.27, 188–91, 188n.30 Stroud, Barry 4, 158 subjective beliefs and ways of thinking, see perspectival representations Swinburne, Richard 147n.16 syntax, see logical syntax synthetic, see analytic/synthetic distinction Tarski, Alfred 35, 100 temporal points of view 4, 44–5, 134, 143, 158, 177n.52, 178, 186–7; see also future; past; spatio–temporal features of reality; tense tense 4–5, 143, 158–9, 160–1, 162, 162–4, 178, 179–80, 186–7, 189–90; see also future; past; temporal points of view theories of meaning 2–3, 75–80, 82–9, 94–5, 97–100, 103–4 thick ethical concepts 166–7, 171–4, 178–81, 215n.20, 220n.27 Thomas, Dylan 13 time, see future; past; spatio–temporal features of reality; tense Tooley, Michael 150n.27, 180n.63 Tractatus 2, 7, 40, 56, 71–6, 78, 88, 91–6, 103, 130, 188, 201–2, 220–1, 224, 246n.8, 248, 248n.16, 250–60, 250n.31 mathematics as construed in, see mathematics as construed in the Tractatus sign/symbol distinction in, see sign/symbol distinction in the Tractatus see also logical syntax; saying/showing distinction; Wittgenstein’s early work

transcendence 5–6, 40, 109, 113, 118, 126, 196–9, 205n.49, 217–19, 222; see also univocity of being Transcendental Deduction, see Kant’s Transcendental Deduction transcendental idealism 104, 259 in Kant 3–4, 71n.2, 72–4, 88, 95n.19, 103, 107n.3, 109, 112, 114–15, 117n.31, 118, 233–4 in the early Wittgenstein 71–5, 88n.48, 95n.19, 103, 104n.57, 104n.58, 233–4, 240, 255–6, 259 in the later Wittgenstein 2–3, 71–5, 88–9, 104n.57, 104n.58, 240–1 Travis, Charles 47–8, 48n.39 truth–aptitude 34, 47–52, 54n.59, 55–6, 99–100, 140–1, 147n.16, 151, 170–1, 172n.33, 214–15, 229–30, 253–4; see also bivalence and the law of the excluded middle; sense: distinguished from nonsense truths as candidates for ineffability, see facts/ truths as candidates for ineffability unconditioned in Kant 68–9, 107–11, 114, 116, 117n.33 underdetermination 17, 102, 169–72; see also underdetermination/indeterminacy distinction underdetermination/indeterminacy distinction 17, 170n.29; see also indeterminacy, underdetermination understanding 16, 75–85, 96, 104, 197–8, 215, 225, 228–30, 233, 235–7, 250–1, 250n.35, 254–6 states of as candidates for ineffability, see states of knowledge/understanding as candidates for ineffability see also making sense unity 4–7, 143, 152n.31, 153–7, 159, 173, 175–82, 200, 233–4; see also univocity of being univocity of being 195; see also difference: as expressing the essence of being; transcendence; unity use/mention distinction 11, 45–7, 81, 84–6, 89, 256–8 vagueness 46, 50–5; see also sorites paradoxes van Inwagen, Peter 204n.44 Vanrie, Wim 58n.14, 62n.35 Waismann, Friedrich 37 Weiss, Jonathan 257n.67 Wiggins, David 205n.46, 251–3 Williams, Bernard 2–3, 5, 71–7, 89, 94n.13, 130–1, 140n.18, 165, 186n.22, 211–12, 215n.20, 240n.91

OUP CORRECTED PROOF – FINAL, 18/06/19, SPi

Index Williams, S.G. 205n.46 Williamson, Timothy 49n.41, 51–2, 54n.59, 97n.27, 196n.2, 214n.14, 249–50 Wittgenstein, Ludwig: his early work 1, 6–7, 56, 97–8, 101, 186n.23, 208–9, 218–19; see also Tractatus; transcendental idealism: in the early Wittgenstein

279

his later work 2–3, 6–7, 37n.60, 41–2, 44n.23, 56n.3, 65, 67, 71, 93–4, 101, 107–8, 110–11, 182, 188, 190, 224–5, 233n.48, 237, 239–43 see also mathematics: as construed by the later Wittgenstein; transcendental idealism: in the later Wittgenstein Wright, Crispin 125n.21, 126n.25, 202n.36