Quantifier Variance and Realism This Page Intentionally Left Blank Quantifi Er Variance and Realism Essays in Metaontology

Quantifier Variance and Realism This page intentionally left blank Quantifi er Variance and Realism Essays in Metaontology

Eli Hirsch Oxford University Press, Inc., publishes works that further Oxford University’s objective of excellence in research, scholarship, and education.

Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto

With offi ces in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam

Published by Oxford University Press, Inc. 198 Madison Avenue, New York, New York 10016 www.oup.com

Oxford is a registered trademark of Oxford University Press 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, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of Oxford University Press.

Library of Congress Cataloging-in-Publication Data Hirsch, Eli Quantifi er variance and realism : essays in metaontology / Eli Hirsch. p. cm. ISBN 978-0-19-973211-1 (alk. paper) 1. Ontology. 2. Object (Philosophy) 3. Language and languages—Philosophy. 4. Metaphysics. I. Title. BD336.H57 2010 110—dc22 2009043573 ISBN-13: 9780199732111

1 3 5 7 9 8 6 4 2 1 Printed in the United States of America on acid-free paper For Pam This page intentionally left blank acknowledgments

The essays reprinted here have not been altered except for a few necessary adjustments of citations. I thank the original publishers for permission to reprint. 1. “A Sense of Unity,” The Journal of Philosophy 74, September 1978 2. “Basic Objects: A Reply to Xu,” Mind & Language 12, September 1997 3. “Objectivity Without Objects,” in R. Cob b-Stevens (ed.) World Congress of Philosophy, Volume 5: Epistemology, 1999 4. “The Vagueness of Identity,” Philosophical Topics 26, 2000 5. “Quantifi er Variance and Realism,” Philosophical Issues 12, 2002 6. “Against Revisionary Ontology,” Philosophical Topics 30, 2003 7. “Comments on Theodore Sider’s Four Dimensionalism,” Philosophy and Phenomenological Research 68, May 2004 8. “Sosa’s Existential Relativism,” in J. Greco (ed.), Ernest Sosa and His Critics, Blackwell, 2004 9. “Physical-Object Ontology, Verbal Disputes, and Common Sense,” Philosophy and Phenomenological Research 70, January 2005 10. “Ontological Arguments: Interpretive Charity and Quantifi er Variance”, in T. Sider, J. Hawthorne, and D. Zimmerman (eds.), Contemporary Debates in Metaphysics, Blackwell, 2008 11. “Language, Ontology, and Structure,” Nous 42, September 2008 12. “Ontology and Alternative Languages,” in D. Chalmers, D. Manley, and R. Wasserman (eds.), Metametaphysics, Oxford University Press, 2009 I thank Brandeis University for the award of a Norman Grant that helped with the preparation of the book.

vii This page intentionally left blank contents

Acknowledgments vii

Introduction xi

1. A Se nse of Unity 3 2. Basic Objects: A Reply to Xu 27 3. Obje ctivity Without Objects 36 4. The Vagueness of Identity 45 5. Quantifi er Variance and Realism 68 6. Against Revisionary Ontology 96 7. Comments on Theodore Sider’s Four Dimensionalism 124 8. Sosa’s Existential Relativism 132 9. Physical-Object Ontology, Verbal Disputes, and Common Sense 144 10. Ontological Arguments: Interpretive Charity and Quantifi er Variance 178 11. L angu age, Ontology, and Structure 197 12. Ontology and Alternative Languages 220

Index 251

ix This page intentionally left blank introduction c C

C This volume bears the name of its most central essay. All of the c essays here are in one way or another discussions of what is implied by the idea of quantifi er variance. Let me in this Introduction make a few comments about that idea. J. O. Urmson wrote: “If two sentences are equivalent to each other, then while the use of one rather than the other may be useful for some philosophical purposes, it is not the case that one will be nearer to reality than the other. . . . We can say a thing this way and we can say it that way, sometimes. . . . But it is no use asking which is the logically or metaphysically right way to say it.”1 By “equivalence” Urmson evidently meant some coarse-grained relation that I’ll assume might be defi ned as follows: two sentences are truth-conditionally equivalent if, relative to any (actual or possible) context of utterance, they are true with respect to the same possible worlds. (In other words, relative to any context of utterance, they express the same coarse-grained proposition.) Urmson’s dictum is that even if sentences diff er radically in logical form, so long as they are truth-conditionally equivalent, it makes no sense to say that one of them is metaphysically more right than the other. Urmson’s real target is the comparison of whole languages rather than isolated sentences. Let’s say that two languages are truth-conditionally equivalent if, for any sentence in one, there is a

1. Urmson (1956), p. 186.

xi xii introduction truth-conditionally equivalent sentence in the other. Urmson’s dictum, then, is that truth-conditionally equivalent languages are of equal metaphysical merit. This seems on the face of it plausible. If two languages are truth- conditionally equivalent then, in any context, speakers of either language can assert sentences (with respect to a possible world) that hold true just in case the sentences asserted by speakers of the other language hold true. It’s not clear what sense it could make for the speakers of one language to consider its true assertions to be closer to reality than the true assertions made in the other language. (Notions of fi ne- grained facts do not seem to negate this point, according to my argument in essay 11.2) The doctrine of quantifi er variance might be viewed as simply a corollary of Urmson’s dictum. When two philosophers X and Y are engaged in an ontological dispute it will often (perhaps not always) happen that we can conceive of two possible languages, the X-language and the Y-language, such that these languages are truth-conditionally equivalent and, in any context of utterance, speaker of the X-language can both reasonably and truthfully assert the same (phonetically individuated) sentences (with respect to a possible world) that the X-philosophers assert, whereas speaker of the Y-language can both reasonably and truthfully assert the same sentences that the Y-philosophers assert. For obvious reasons (that nothing profound ought to be read into) we can call these languages diff erent “ontological languages.” The meanings of such quantifi er expressions as “there exists something” (and many related expressions) vary from language to language. (The idea of “quantifi er variance” is not meant to imply that in diff erent ontological languages only the meanings of quantifi er expressions vary; obviously the truth conditions of singular sentences must vary from one ontological language to another.3) Urmson’s dictum implies that truth-conditionally equivalent ontological languages are of equal metaphysical merit. That is the doctrine of quantifi er variance. The doctrine says that there is no uniquely best ontological language with which to describe the world. Quantifer variantism has to be distinguished from two claims that I think follow from it. Some critics might accept quantifi er variantism but deny that the two claims really follow. (The two claims might also

2. Two highly relevant discussions of fi ne-grainedness are Hawthorne (2009) and McGrath (2008). 3. This point is apparently missed in the objection to quantifi er variance posed in Hale and Wright (2009), p. 184. introduction xiii be defended independently of quantifi er variantism.)4 The fi rst is that ordinary language is a perfectly good ontological language in which common sense judgments about the existence and identity of objects are strictly and literally true (see especially essays 6 and 9). The second is that many (perhaps not all) ontological disputes are merely verbal, with each ontological camp asserting truths in its own ontological language (see especially essays 9, 10, and 12). These claims, especially the second, have drawn a great deal of critical attention. Perhaps I can briefl y restate them here. If we are addressing two disputing revisionary ontological camps, say, mereological essentialists and four-dimensionalists, we can make the following speech to them: “Either you are both speaking plain English, in which case you are both asserting trivial falsehoods, or each of you is in eff ect asserting trivial truths in your own ontological language, in which case your dispute is verbal.” I certainly have no stake in pressing for the second disjunct over the fi rst. Once we understand that those are the only two options, the dispute has been thoroughly defl ated; there is nothing of substance left to it. (And shifting to “philosophy- room English” does not help, according to my argument in essay 11.) Sider, the most prominent opponent of quantifi er variantism, holds that (a) the world contains a natural quantifi cational structure, and (b) there is a uniquely best ontological language. (See the discussion of Sider especially in essays 7, 9, and 11.) Since he takes (a) and (b) to stand or fall together, he sometimes characterizes quantifi er variantism as denying (a) and sometimes as denying (b). Although I’m skeptical about (a), I want to defi ne quantifi er variantism as the denial of (b), not as the denial of (a). I need to keep (a) and (b) separate because (as I have argued throughout Hirsch 1993) I don’t see any obvious connection between saying that an objective phenomenon is natural and saying that there is some virtue in representing it with a particular form of language; for example, there is no obvious connection between saying that a property is natural and saying that there is some virtue in expressing it with a word rather than a complex term. I therefore defi ne quantifi er variance in terms of (b), not in terms of (a). The connection to (a) has to be argued out separately. There are two misunderstandings of quantifi er variance that keep cropping up in the literature. I’ve tried to address them especially in essay 5, but let me say a word about them here. The fi rst has to do with the issue of realism. In assessing quantifi er variantism it is essential to

4. In fact, some of my arguments for the two claims (see especially essay 11) suggest that a version of them might be defended by appealing only to the part of quantifi er variantism that rejects “necessity invariantism” (see below). xiv introduction pay careful attention to the diff erence between object-level and meta- level claims. Quantifi er variantism makes no claim at all about what the world is like, not even a claim about what the world is “really” like “in itself.” It claims that, whatever the world is like, there can be no uniquely best ontological language to describe it. One should not detect in this claim even a hint of an anti-realist view that what exists in the world somehow depends on language. A second misunderstanding is to suppose that quantifi er variantists must be committed to viewing the quantifi er expressions in the diff erent ontological languages as being similar in meaning in some deep and controversial way; otherwise, why even call them “quantifi er expressions”?5 Given the way that ontological languages are constructed, it immediately follows that in all of them the expression “there exists something” will share the formal-syntactic inferential roles familiar from formal logic. I need to emphasize that I attach no importance to claiming any similarity beyond this formal-syntactic similarity. Early on in essay 5, where I fi rst introduce the idea of quantifi er variance, I say: “I might characterize this change [in the use of ‘there exists something’] as ‘giving up the quantifi er’ and ‘giving up the notion of the existence of something’, but it seems more natural to characterize it rather as ‘acquiring a new kind of quantifi er’ and ‘ acquiring a new notion of the existence of something’.” (See my related remarks in essay 8.) I don’t consider anything of importance to depend on which characterization one chooses. If one chose the fi rst then maybe we should talk of quantifi er “elimination” rather than “variance,” but nothing of importance turns on this.6 What seems interesting and initially (or permanently) mind-boggling is the idea that the world could be adequately described without employing our quantifi er-meaning, without employing our notion of the existence of something. That key idea is there whether one puts it in terms of quantifi er “elimination” or quantifi er “variance.” This being said, I’m confi - dent that most people will fi nd the second formulation more natural. I said earlier that quantifi er variantism might be viewed as simply a corollary of Urmson’s dictum. That needs to be qualifi ed. Quantifi er variantism says that there is no uniquely best ontological language with which to describe the world. I take this to imply that (i) there are a number of possible truth-conditionally equivalent ontological languages, and (ii) these languages are of equal metaphysical merit. Sider’s invariantism grants (i) but denies (ii), thereby denying Urmson’s

5. This question is pressed, I think excessively, in Hale and Wright (2009). 6. See “Being Explained Away” in Burgess (2008). introduction xv

dictum. We might call this normative invariantism.7 A diff erent view, which might be called necessity invariantism, denies (i). According to this view there is only one (metaphysically) possible ontological language; it is impossible for any language to lack the quantifi er-meaning that we have in our language. Linguistic constructions that purport to yield alternative ontological languages merely produce some form of codes or secondary languages, but not possibly genuine (primary, natural) languages. By my own lights, necessity invariantism may be a more promising view than normative invariantism. I’m not aware, however, of any philosopher who has explicitly defended this view.8 Even quantifi er variantists might allow that there are some necessary constraints on the kinds of quantifi er variations that are possible. (See my remarks about this in essays 10 and 12.) Is it even possible for there to be a (primary) language in which the most unrestricted quantifi er ranges over all trees except for a certain one in Central Park? There may be some kind of necessary constraint here related to similarity, to “going on the same way,” though it seems extremely diffi cult to see how such a constraint might be defi ned.9 The question about constraints on metaphysically possible ontological languages needs to be distinguished from questions about whether there are empirical psychological constraints on the sorts of ontological languages that humans employ. This question is discussed in essays 1 and 2. (Essay 1 goes back to 1978, and it’s clear that I simply take quantifi er variantism for granted in it.)10

7. In the sense I have in mind, a language lacks “metaphysical merit” to the extent that, on purely metaphysical (rather than pragmatic) grounds, it is wrong or irrational to employ the language. Although Sider is certainly a normative invariantist in the sense defi ned, he may regard this aspect of his view as less central than his ideas about objective quantifi cational structure. There are, needless to say, substantive questions in the background of how I am framing these issues. 8. In some of my early discussions of Sider I took him to be a necessity invariantist, but it’s now clear that this is not his view. A version of such a view, based on ideas in Bealer (1982), is discussed in Hirsch (1993), chs. 6 and 7. Matti Eklund’s “maximalism” has sometimes suggested to me (though perhaps not to Eklund) a form of necessity invariantism (see the discussion of Eklund in essay 12). Various vaguely formulated complaints against quantifi er variance having to do with compositionality (see in essays 9 and 12) may be aiming towards a necessity invariantist position. If there are necessity invariantists out there, let’s see an explicit formulation and defense of the position. 9. Fine (2009) seems to suggest that quantifi er variations are possible only if they follow a certain specifi c mechanism (see at p. 164, note 2). Hale and Wright (2009) seem to suggest that quantifi er variations are possible only if they conform to certain “abstraction principles.” I myself fi nd neither of these constraints intuitively compelling. 10. In one of my earliest publications (Hirsch, 1976; republished in Hirsch 1982) I spoke of the possibility of describing the world in terms of diff erent “identity schemes.” That was essentially the doctrine of quantifi er variance, except that I would now guard more carefully than I did then against the danger of anti-realist misinterpretations. xvi introduction

I have emphasized that quantifi er variantism is compatible with realism. In fact, I would want to stipulate that only realists count as quantifi er variantists. Some post-modernists seem to say that language creates reality. They might then say that changes in quantifi er-meanings change what objects there are in the world. I certainly do not want to call such a view quantifi er variantism. What varies in quantifi er variantism is only the language; everything else remains the same. An anti-realist conception in which varying the language somehow changes all of reality is an entirely diff erent story. Quantifi er variantism is often said to be the descendant of Car- nap (see the discussion of Carnap especially in the beginning of essay 12). I have a problem, however, in calling Carnap a quantifi er variantist, insofar as he is often viewed as a verifi cationist anti-realist. The same holds for Putnam, though his connection to quantifer variantism runs, I think, much deeper and clearer than Carnap’s (see the discussion of Putnam especially in essays 3 and 5). Let’s not call any philosophers quantifi er variantists unless they are clearly committed to the idea that (most of) the things that exist are completely independent of language.11

References Bealer, G. (1982). Quality and Concept (Oxford University Press, N.Y.). Burgess, J.P. (2008). Mathematics, Models, and Modality (Cambridge University Press, N.Y.). Fine, K. (2009). “The Question of Ontology” in D. Chalmers, D. Manley, and R. Wasserman, eds., Metametaphysics (Oxford University Press, N.Y.). Hale, B. and Wright, C. (2009). “The Metaontology of Abstraction” in D. Chalmers, D. Manley, and R. Wasserman, eds., Metametaphysics (Oxford University Press, N.Y.). Hawthorne, J. (2009). “Superfi cialism in Ontology” in D. Chalmers, D. Manley, and R. Wasserman, eds., Metametaphysics (Oxford University Press, N.Y.). Hirsch, E. (1976). The Persistence of Objects (Philosophical Monographs, Phila- delphia, Pennsylvania). ———. (1982). The Concept of Identity (Oxford University Press, N.Y.). ———. (1993). Dividing Reality (Oxford University Press, N.Y.). McGrath, M. (2008). “Conciliatory Metaontology and the Vindication of Common Sense,” Nous, 42, pp. 482-508. Urmson, J. O. (1956). Philosophical Analysis (Oxford University Press, London).

11. My thanks to Matti Eklund and Ted Sider. Quantifier Variance and Realism This page intentionally left blank one c C

a sense of unity

C Philosophers have often raised questions about our concept c of the unity of a thing. Most typically what is sought is an analysis of what our concept of unity consists in. The answer to this question commonly takes the form of citing various conditions that seem to provide a defi nition of our judgments of unity. These conditions may be said to constitute our criteria of unity, our criteria of identity. The question I want to raise in the present paper is somewhat dif- ferent from this typical one. Suppose that we have already ascertained what our criteria of unity are. Then I want to ask why it is that we employ just those criteria rather than others. What determines us to base our judgments of unity on just those conditions?

I. Criteria of Unity

Let me present an example to illustrate the diff erence between these two kinds of questions: the one I am asking and the more typical one. Suppose that you have a tree in your backyard and that next to the tree there stands a table. Common sense would judge that the tree is a single object and the table is a single object. Each of these objects is of course composite; the tree, for example, is composed of a trunk, some branches, twigs, leaves, and so on. Now something that common sense would defi nitely not judge is that there is a single object that is

3 4 quantifier variance and realism: essays in metaontology

composed of the tree together with the table. If I am, say, touching the tree and you are touching the table, common sense would not say that there is some single object that you and I are both touching. But why not? Why should we not say that the tree and the table add up to a unitary object? At one level the answer to this question would consist in citing relevant criteria of unity through space, i.e., criteria that determine whether or not an aggregate of matter can properly be said to add up to a single thing. In the case under discussion two relevant criteria would seem to be spatial connectedness and dynamic cohesiveness. Gener- ally an object must be spatially connected, in the sense that any of its parts can be connected by a continuous curve whose points all touch the object. And generally an object must be cohesive, in the sense that all its parts tend to remain together under various pressures. I do not mean to suggest that these two conditions (connectedness and cohesiveness) are strictly necessary for an object’s unity in all imaginable circumstances; nor am I suggesting that they are suffi cient for unity. But these conditions are pretty likely to fi gure in any general analysis of an object’s unity through space, and with respect to our simple example they seem enough to rule out the tree-cum-table as a unitary object.1 At one level, then, we can explain by appealing to criteria why the tree and the table do not add up to one object. The question I want to raise, however, is why these criteria function the way they do. What is it that induces common sense to base a judgment of unity on the particular conditions of connectedness and cohesiveness? Why allow those conditions to dictate the matter? Why does common sense not choose some other criteria of unity, criteria that might allow for the judgment that the tree and the table compose a single object (where this object happens to be disconnected and noncohesive)? Let me extend this example a bit, so as to bring identity through time into play. The tree is not just spatially composite; it is also, in a sense, temporally composite. Insofar as the tree persists through time, it (or its history) can be thought of as comprising a succession of temporary stages, where these stages can be delimited in any number of ways. (For example, we can think of the tree as combining an early stage in which it is short followed by a later stage in which it is tall, or as combining stages in which it is in bloom with stages in which it is

1. For a more general discussion of our criteria of unity through space, see my monograph, The Persistence of Objects (Philadelphia: Philosophical Monographs, 1976), pp. 38–44 (reprinted in my The Concept of Identity [New York: Oxford University Press, 1982], pp. 105–112). a sense of unity 5 not in bloom.) And the same can be said for the table; this too is temporally composite. If, however, we were to combine in thought some early stage of the table with some later stage of the tree we would not, at least by the lights of common sense, arrive at a unitary persisting thing. In the case I am imagining, where a tree and a table are situated together in a normal way, common sense could not even take seriously the idea that some single persisting thing is fi rst a table and then a tree. But why not? Why should we not judge in this case that there is a single persisting object that combines a table-stage and a tree- stage? Again, the answer at one level consists in citing criteria, in this case criteria of unity through time. Two criteria that seem to suffi ce for the case (though they do not suffi ce for all cases) are qualitative continuity and spatiotemporal continuity.2 If we tried to think of there being a single object that is fi rst a table and then a tree we should have to say that this object changed discontinuously, as regards both its qualities and its location. Our criteria of unity through time do not (in general) allow us to say this. And again, my question is, what induces common sense to credit those particular criteria of unity through time? Why not choose other criteria which might accommodate the judgment that a table changed discontinuously into a tree? There are philosophers, notably W. V. Quine, who in fact recommend a revision in our common-sense notion of an object which would have precisely the eff ect of accommodating the judgments that I have just instanced as confl icting with our ordinary criteria of unity. In terms of Quine’s revised concept of an object we would indeed say, in the imagined example, that there is at a given moment some object that is composed of the table and the tree, and that there is over a period of time an object that is fi rst a table and then a tree. On Quine’s proposal an object “comprises simply the content, however heteroge- neous, of some portion of space-time, however disconnected and gerrymandered.”3 Any space-time portion of reality qualifi es as an object, in Quine’s terms. But this technical notion of an object is crucially dif- ferent from the ordinary notion, as Quine himself amply stresses. In terms of the ordinary notion only a select few space-time portions qualify as objects, namely, those which satisfy our criteria of unity. It is the ordinary notion that concerns me. Quine often marks off the ordinary notion from his technical one by using the word ‘body’ for the

2. I discuss these criteria at length op. cit., chs. 2–4, passim. 3. Word and Object (Cambridge, Mass.: MIT Press, 1960), p. 171. 6 quantifier variance and realism: essays in metaontology ordinary notion. Hence he says: “Man is a body-minded animal.”4 In these terms what I am asking for is an explanation of why common sense is body-minded. What is at stake in this question is not merely the use of two or three words (such as ‘body,’ ‘object,’ ‘thing’), but a whole way of thinking. Exactly how to characterize that way of thinking is itself an essential part of the philosophical diffi culty. But one can say, to begin with, that the category of a body seems to constitute for common sense the primary way of breaking up the world into units. And this category is defi ned in terms of various specifi c and complicated criteria of unity, for example, spatial connectedness, dynamic cohesiveness, spatiotemporal and qualitative continuity. The question, then, is why common sense should divide reality up in just that particular way.

II. Unity and Similarity

A possible answer to this question, which I want to consider and defend, is that we think of the world in terms of our criteria of bodily unity because we are innately disposed to think in this way. According to this hypothesis, a sense of bodily unity is part of our inborn constitution, and this is what determines us to interpret our experience in the way we do. This hypothesis has something in common with Kant’s view about the a priori category of substance. I want to stress, however, two diff erences between the hypothesis under consideration and Kant’s view. First of all, Kant had little, if anything, to say about specifi c criteria of bodily unity. In fact Kant’s category of substance is not equivalent to the common-sense idea of a body. Ordinary bodies, like trees and tables, are created and destroyed, but Kant’s idea of substance referred to the underlying matter that was supposed to persist forever. The second diff erence is that Kant maintained that his category of substance is a necessary ingredient of understanding. This necessity claim is no part of my hypothesis. My hypothesis claims only that, as a matter of contingent fact, human beings are innately disposed to interpret their experience in a certain way. This hypothesis is more closely related to some of Chomsky’s ideas about innate grammatical schemata. And it is even closer to the views expressed by gestalt psychologists like Wolfgang KÖhler, who have

4. The Roots of Reference (LaSalle, III.: Open Court, 1973), p. 54. a sense of unity 7 maintained that, as a matter of empirical fact, our sensory fi elds are “naturally” and “spontaneously” organized in terms of distinctive kinds of units.5 I want to broach this idea by way of an analogy. I want to compare the idea of an innate sense of unity to the more familiar philosophical idea of an innate sense of similarity. It has been persuasively argued by Quine, and also by Anthony Quinton, that our grasp of general concepts must ultimately be rooted in an innate tendency to classify objects in certain defi nite ways.6 In order for a child to acquire the use of a general term, he must be able to extrapolate from observed cases of the term’s application to new cases. This extrapolation evidently requires that the child have some basis for deciding which new cases go together with the observed cases. At least with respect to the most elementary vocabulary, the basis for this decision would apparently have to be innate. The idea here is not that our fully developed scheme of classifi - cations depends on nothing but our primitive classifi catory impulses. Perhaps the scheme is eventually aff ected by various practical and theoretical needs. At bottom, however, there must be the innate tendency to classify things in certain ways rather than others. Quine sometimes refers to this innate tendency as an innate “sense of similarity”; sometimes he refers to it as an innate “quality space.” It should be borne in mind that the fi rst expression (‘sense of similarity’) is not meant to imply any special views about the possibility of reducing properties to similarity relations; and the expression “quality space”applies not just to qualities properly speaking, since all properties, including relational properties, would have to be treated in the same manner. The general point is simply that we are innately disposed to classify in certain ways rather than others. Now for common sense the most basic thought about physical reality is the thought that some specifi ed body has some specifi ed property. The fi rst ingredient of this thought (the specifi cation of a body) is linked to our criteria of bodily unity, and the second ingredient (the specifi cation of a property) is linked to our principles for classifying bodies. We have just seen that, according to Quine and others, the classifi catory ingredient is rooted in the innate disposition to classify in

5. Gestalt Psychology (New York: Liveright, 1947), chs. 5 and 6; parenthetical page references to KÖhler will be to this book. 6. See Quine, “Natural Kinds,” in Ontological Relativity and Other Essays (New York: Columbia, 1969), p. 116 ff ; Word and Object, p. 83 ff ; The Roots of Reference, p. 19; and Quinton, The Nature of Things (Boston; Routledge & Kegan Paul, 1973), pp. 261–265. 8 quantifier variance and realism: essays in metaontology distinctive ways. One can scarcely resist the speculation that perhaps the other ingredient, that related to our criteria of unity, is likewise rooted in the innate disposition to adopt certain criteria of unity rather than others. The general scheme we then wind up with is this: As our innate sense of similarity stands to our principles for classifying bodies, so does our innate sense of unity stand to our criteria of unity for bodies. It will be instructive to try to make out what Quine’s reaction might be to this proposal. Some of his remarks may certainly seem to suggest that he too believes in an innate sense of unity. He says that “body-unifying considerations, though complex, are rooted in instinct” (Roots of Reference, 55), and he refers to our “instinctive body- mindedness” (ibid. 56). But this is puzzling, since the view standardly attributed to Quine is that, besides such obvious general faculties as perception, intelligence, and motor behavior, quality space is essentially the only innate endowment that can confi dently be related to the process of learning language. Not that Quine is at all adamant about this; he seems quite open to other possibilities, even perhaps to some of Chomsky’s suggestions.7 But it would certainly be extremely odd to attribute to Quine the unheralded position that working side by side with the language-learner’s innate quality space is the quite distinct disposition to adopt certain identity criteria. That is in fact the position that I want to maintain, but I have to doubt that this is Quine’s position. Actually if we look more closely at that section in The Roots of Ref- erence from which I previously quoted, we fi nd that when Quine refers to our “instinctive body-mindedness” he probably does not mean to introduce an innate disposition distinct from our sense of similarity. Rather he seems to be suggesting that our body-mindedness is itself the result of our innate sense of similarity.

Thanks to [the child’s] instinctive body-mindedness, he is an apt pupil when the general terms are terms for bodies. He is able to appreciate not only that the second-order similarity of a dog to a dog exceeds that of a dog to a rabbit, but also that the latter in turn exceeds that of a dog to an apple or buckle. . . . And then there is the yet slighter degree of second-order similarity, residing in just those very general body-unifying considerations that preserve the identity of each dog,

7. See Quine’s “Philosophical Progress In Language Theory,” in H. E. Kiefer and M. K. Munitz, ed., Language, Belief, and Metaphysics (Albany: suny Press, 1970), p. 6. a sense of unity 9

each rabbit, each apple, each buckle, in short each body. This would be a second-order similarity basis for the child’s ostensive learning of the general term “body” itself, or “thing,” to take the likelier word (56).

What Quine seems to be saying here is that our “instinctive body- mindedness” is actually nothing more than our disposition to appreciate the complicated similarity relations that obtain between those space-time portions of reality which we count as bodies. But there is something wrong here. To operate with the ordinary concept of bodily unity is not just a matter of appreciating various similarities between those portions of reality which qualify as bodies. Imagine someone who did not operate with the ordinary concept, but who operated instead with that technical notion of an object which, as I mentioned earlier, Quine ultimately favors over the ordinary notion. Someone who operated with this revised concept would be treating all portions of reality, whether disconnected or whatever, as units on a logically equal footing. But certainly he might very well appreciate the relevant similarities between those select portions of reality which common sense dignifi es as bodies. To be body-minded, in the way that common sense is body-minded, is to adopt an ontology that excludes all of those portions of reality which do not qualify as bodies. Common sense simply does not credit such portions of reality. Our common-sense adoption of this exclusionary ontology cannot be regarded as merely a corollary of our disposition to appreciate certain similarities. Perhaps someone will be tempted to suggest that we exclude portions of reality other than bodies because our sense of similarity provides no basis for comparing or contrasting such portions of reality, and hence we cannot classify them in any way. But this is wrong. If we did credit such portions of reality as units we certainly could classify them in various ways. If, say, there is a brown table and a brown tree in my backyard and there is a brown table and a brown tree in your backyard, then we could say that my table-cum-tree is similar to your table- cum-tree at least with respect to the property of being brown, or, even more obviously, with respect to the property of being a table-cum-tree (i.e., the property of being exhaustively composed of a table and a tree). In these respects both items could be said to contrast with any table- cum-tree that is not brown, or with any chair-cum-tree. Of course we do not ordinarily draw any such comparisons and contrasts. This is because we do not ordinarily credit any such unit as a table-cum-tree. But that fact is in no way explained by our sense of similarity. 10 quantifier variance and realism: essays in metaontology

It is unclear, then, what connection Quine intends to educe between our body-mindedness and our sense of similarity. I think that part of the trouble here is that Quine does not distinguish between two questions. One is a question about why our language is the way it is; the other is a question about how our language is learned. The fi rst question is: “Why does ordinary language contain just these particular criteria of unity?” The second question is: “How do children learn these criteria of unity?” It is the fi rst question that I raised at the outset of this paper. It was this question that I also expressed by asking why it is that common sense is body-minded. Now what we have just seen is that Quine certainly off ers no answer to this question. There is no way that a sense of similarity can be seen as delivering common sense into body-mindedness. On the other hand Quine may have provided a viable answer to the second question, about how a child acquires the criteria of unity implicit in our language. The child must learn to distinguish between those portions of reality which do, and those which do not, satisfy the criteria. This he may be able to do, so long as his classifi catory impulses are attuned to the complex considerations that enter into these criteria. That is, if the child’s sense of similarity reveals a general contrast between what qualifi es in our language as a unitary body and what does not, he may be in the position to imbibe the rule that only the bodies are referred to as units, the rest being excluded. Thus he may be able to pick up the body-minded way of talking. There are many complications here, of course, as Quine readily admits. But the general idea may seem workable. What cannot be explained along these lines, however, is why the language, which is being passed on to the child, contains just those criteria of unity.

III. Conventionalism

The answer to this question that I am advocating is that our language contains those criteria of unity because of our innate disposition to see the world in a certain way, where this disposition must be distinguished from our sense of similarity. Now one possible alternative to this answer would be to maintain that there is in fact no reason why our language had to contain just those criteria of unity, but that this is nothing more than an arbitrary convention of language. Our ordinary body- mindedness, according to this “conventionalist” position, is merely one scheme for conceptually dividing the world into units, and any number a sense of unity 11 of other schemes might have done just as well. The scheme that we have gets passed on from generation to generation, in the manner suggested by Quine. At the very outset of Köhler’s discussion of the topic of unity in his book Gestalt Psychology, he peremptorily dismisses this conventionalist alternative in the following words:

On the desk before me I fi nd quite a number of circumscribed units or things: a piece of paper, a pencil, an eraser, a cigarette, and so forth. The existence of these visual things involves two factors. What is included in a thing becomes a unit, and this unit is segregated from its surroundings. In order to satisfy myself that this is more than a verbal aff air, I may try to form other units in which parts of a visual thing and parts of its environment are put together. In some cases such an attempt will end in complete failure (137/8, my italics).

I think we may assume that a case of “complete failure” in Köhler’s terms would occur if we tried to see a tree, or some part of it, as forming a unit together with a nearby table, or some part of it. Köhler’s line of reasoning seems to be as follows. I cannot get myself to see the tree and the table as forming a unit, though I can of course easily utter the words “The tree and the table form a unit.” This shows that a judgment of unity is “more than a verbal aff air,” more than an arbitrary linguistic convention. Unity is something that we experience; it is, as Köhler says a few sentences later, a “visual fact.” The conventionalist is not likely to be convinced by this argument. The issue is not whether we experience unity; obviously we do. As I look around me I can see that some portions of the scene add up to a unitary object and some do not. This the conventionalist would not deny. His suggestion, however, is that the way that I experience unity is determined by the arbitrary conventions of my primary language, i.e., the language I habitually speak and in terms of which I think. Of course I cannot alter my experience merely by mouthing some strange sentence (e.g., “The tree and the table make up one thing”), because it is my primary language that matters. Hence Köhler’s stark dichot- omy between “visual facts” and “verbal aff airs” does not speak to the issue. Köhler’s failure to address the possible infl uence of language on our experience is a fl aw in his whole treatment of sensory organization. Some of his most impressive observations pertain to the way that we see things as forming groups or clusters, a phenomenon which he sometimes refers to as the formation of “group-units.” Though this 12 quantifier variance and realism: essays in metaontology

figure 1 phenomenon of “group unity” is not directly relevant to the topic of bodily unity under discussion, it may serve to highlight the problematical relationship between our language and the structure of our experience. Köhler points out that, when we look at Figure 1, we see two groups of dots, each group containing three dots. We do not see three groups of two dots each, or two groups divided in some other way. What we see very clearly are two groups of three dots each. Here the phenomenological datum is so striking that one is immediately inclined to share Köhler’s assumption that our language can have nothing to do with the matter.8 But is this really so clear? Consider that in order for us to describe our experience of Figure 1 we must make use of the English word ‘group’ (or some equivalent word like ‘cluster,’ ‘collection,’ ‘arrange- ment’). Two groups are what (we say) we experience. We must, then, have learned at some point what the criteria are for applying the English word ‘group,’ where these criteria presumably coincide with just those “principles of grouping” which gestalt psychologists try to elicit. Is it not possible that, in learning how to use the word ‘group,’ we learned these principles of grouping, and that, had we spoken a diff erent kind of language, we might have experienced Figure 1 quite diff erently? I am not saying that this is plausible; I doubt that it is. But the question needs to be focused on properly. I would like to apply some of these points to the topic of unity through time, a topic which Köhler essentially ignores. There is a comment that Sydney Shoemaker makes about unity through time which seems to parallel Köhler’s dismissal of conventionalism:

It is a striking fact that motion, though it involves the persistence through time of the moving object, is often directly observed rather than inferred. . . . And I think it is partly because there is an experience of motion that spatiotemporal continuity occupies the central role it

8. See KÖhler, op. cit., p. 142. a sense of unity 13

does as a criterion of identity. . . . It does not seem to be just a matter of convention that we use spatiotemporal continuity as a criterion of identity. On the contrary, when I see motion (as opposed to inferring it) there seems to be no way in which I could describe what I see except by saying “It (or: something) is moving,” and in saying this I imply the persistence of something through time.9

What does Shoemaker mean when he says that what I see could only be described in terms of the ordinary proposition “It (or: something) is moving?” Here the expression ‘what I see’ has the characteristic kind of ambiguity that Wittgenstein discussed under the heading of “seeing as.”l0 Of course if I am asked to describe the way I ordinarily experience a moving object I must employ just those ordinary notions which are constitutive of my ordinary experience. But the conventionalist would hold that, had we learned a diff erent kind of language, with diff erent identity criteria, we might have experienced a moving object in a radically diff erent way (under a radically diff erent “aspect”). Shoemaker’s comments merely dismiss this possibility. To fl esh out the conventionalist’s point we would have to sketch an alternative language, one which did not make use of our ordinary identity criteria, and explain how the world might have been experienced in terms of that language. Strange languages, of the general sort required, are of course not unknown to philosophical literature. An alternative language that is especially relevant to the present purpose is one that employs Quine’s technical notion of an object, a notion I have mentioned several times before. From the vantage point of that “space- time language,” as I will henceforth call it, any space-time portion of reality, however disconnected or noncohesive, qualifi es as a unitary object. The space-time language is, in an important sense, a language without criteria of unity. In the sense that a club that allows anyone to join has no criteria of membership, a conceptual scheme that allows any space-time portion to qualify as a unit can be said to have no criteria of unity. What would the world look like from the space-time standpoint? Well, in a sense everything remains the same. If there is a mouse moving across the fl oor, then, from the space-time vantage point, there is that particular spatially and temporally extended chunk of reality which corresponds to common sense’s moving mouse. But that chunk

9 . Self-Knowledge and Self-Identity (Ithaca, N.Y.: Cornell, 1963), pp. 203/4. 10. Philosophical Investigations (New York: Macmillan, 1953), IIxi. 14 quantifier variance and realism: essays in metaontology of reality would be seen as criss-crossing and overlapping a myriad of other “objects,” such as an object that consists of an early stage of the mouse together with a later stage of a nearby table, or an object that consists of an early stage of the mouse together with a later stage of the mouse’s head. Hence that space-time chunk of reality which corresponds to common sense’s moving mouse would not stand out as it does for common sense, because it would be seen as embedded in a reality swarming and whirling with an endless number and variety of objects corresponding to every shift in one’s attention. There would scarcely be, from this standpoint, a “moving object” in anything like the ordinary sense. This description of the space-time standpoint pretty quickly fades into impressionism. But that is to be expected on any account since we are trying to construct a language that goes against the deepest habits of our thought. What the conventionalist would insist is that, if we had learned that language in childhood as our primary language, we could have seen the world that way. Now my aim is not to defend conventionalism, but only to try to get somewhat clearer as to what the issues are. Actually, as I stated earlier, I favor the position that our criteria of unity are rooted in an innate disposition, what I called our “sense of unity.” This seems to be Köhler’s position, and it is also, I believe, implicit in Shoemaker’s remarks. I think that what is really behind Köhler’s and Shoemaker’s abrupt dismissal of the conventionalist idea is the intuition that this idea is too incredible to be taken seriously. And it does indeed seem to me intuitively incredible that, had I only been trained to speak diff erently, my experience of unity might have been completely diff erent. Still, it would be better not to have to rely entirely on this intuition, but to have some argument to lean on. One possible argument against conventionalism would be of an anthropological sort. If it could be established that every language known to us contains criteria of unity essentially like ours, then this would evidently be a problem for the conventionalist. Of course, if a philosopher holds, as Quine apparently does, that how we translate a foreign language (how we choose a “translation manual”) is itself (something like) a convention, then it seems that we could never hope to establish, as a factual matter, that our criteria of unity are universal. But I am assuming, as I think most philosophers would, that it is a factual question, and in principle an answerable one, whether other people operate with a concept of bodily unity like ours. If, then, an affi rmative answer to this question were forthcoming, this would seem a sense of unity 15 to be a strong objection to the view that our criteria of unity are merely arbitrary conventions.11 Another line of attack against the conventionalist, which I will take up presently, might draw on speculations about how children see the world before they learn a language. Before pursuing that point, however, I want to consider another possibility that is found in Köhler’s discussion.

IV. An “Empiricist” Explanation

Interestingly enough, although Köhler immediately dismisses the conventionalist approach, he addresses himself at length to the “empiricist” (or, as Köhler calls it, the “empirist”) explanation that our judgments of unity are based on previous learning.12 What “previous learning” must mean here, if this position is to be distinguished from conventionalism, is learning about the world rather than learning about language. The empiricist position, in the sense relevant to Köhler’s discussion, is that the criteria of unity in our language are neither innately determined nor arbitrarily conventional, but are rather the result of our having derived these criteria from something that we learned about the world. As against this, Köhler wants to argue that our criteria of unity are innate rather than learned and that, therefore, our ordinary experience of unity is elementary rather than derivative. The reference to previous learning can still be interpreted in two ways. It might mean that each human being must derive the criteria of unity from something that he learns about the world; or it may mean that prehistoric people over the millenia derived the criteria from what they learned about the world, and then passed these criteria down to us through our language. On either interpretation the essential diffi - culty is to explain how our criteria of unity could have been derived from anything learned about the world. A crude but not unfamiliar explanation is depicted by Köhler in the following words: “Since early childhood we have often observed that

11. But perhaps not a decisive objection: see Hilary Putnam’s suggestion that perhaps all languages have a common origin, in “The ‘Innateness Hypothesis’ and Explanatory Models in Linguis- tics,” Synthese, XVII, 1 (March 1967): 12–28, p.18. 12. For his (somewhat unclear) distinction between the two terms “empiricist” and “empirist,” see KÖhler, op. cit., p.113. I will use the more familiar word “empiricist” to signify the third alternative to the conventionalist and innateness positions that have been discussed, though obviously an “empiricist,” in some more general sense, could easily hold either of these latter two positions. (Cf. Quine’s remarks about “empiricism” in the passage cited in fn 7, above.) 16 quantifier variance and realism: essays in metaontology sets of sensations which have approximately the same color, and diff er in this respect from their environment, tend to behave as units, i.e., to move and be moved, to appear and disappear, at the same time. Such is the case with stones, with papers, with plates, with shoes, with many animals, with the leaves of plants . . . It is only an example of the well known generalizing power of memory if, as a result of such experiences, we treat all homogeneously colored areas as units” (141). Köhler’s own critique of this position is quite complicated and relies on a variety of somewhat specialized phenomenological data. But it seems that his discussion overlooks the obvious and decisive objection to the proff ered explanation. The explanation is premised on our having often observed that certain “sets of sensations . . . tend to behave as units.” Of course one would immediately like to know how the word ‘sensation’ is being used here. But, even more to the point, we need to ask how the idea of something “behaving as a unit” can possibly operate within this explanation. Something behaves as a unit if its parts move together, appear and disappear together, and, in general, are causally interdependent. This is essentially the condition that I earlier called “dynamic cohesiveness.” So the idea is that we treat certain conditions (such as color homogeneity) as criteria of unity because we have learned to associate those conditions with dynamic cohesiveness. But why, then, do we treat dynamic cohesiveness as a criterion of unity? How did we ever learn that? Evidently not the slightest gesture is being made to answer this question. Furthermore, to make matters much worse, the criterion of dynamic cohesiveness itself presupposes various other criteria of unity. To judge that the parts of a thing move together we must be able to trace those parts through time. For this we require various criteria of unity through time (e.g., qualitative and spatiotemporal continuity). Hence the proff ered explanation of how we derive our criteria of unity really presupposes many of these criteria from the start. When one refl ects on these diffi culties it becomes clear that there cannot be any remotely straightforward way of explaining how our criteria of unity might have resulted from previous learning. The sort of “empiricist” who believes in such an explanation must describe an elementary level of experience which does not already presuppose our ordinary criteria of unity. He must then go on to explain how at that level we learn something about the world which somehow gets us to adopt these ordinary criteria. It is far from clear how one could even begin to formulate such an account. a sense of unity 17

Traditionally, sense-data languages and the like were often taken to provide a level of experiential judgment more elementary (more “immediate”) than that of common sense. It is not clear, however, that sense-data descriptions can have any bearing on the present topic. For one thing, these descriptions require criteria of unity for sense data, and these criteria seem generally to be simply borrowed haphazardly from our ordinary criteria of bodily unity. Furthermore, the sense-data maneuver relates to the “problem of the external world,” which is not the problem we are discussing. When I originally introduced the case of the table and the tree, my question was not “How can I be sure that this is not an hallucination?,” or “How do I know that this sort of scene ever exists unperceived?,” but rather “How do I know what criteria to employ in dividing the scene into units?” This latter question pertains to the external world, and traditional sense-data dialectics about how we get to external reality seem quite irrelevant to answering the question. We might try to think of the space-time language as providing a level that is elementary in a sense relevant to the present discussion. For, as I explained earlier, the space-time language might be said to contain no criteria of unity, since at that level we would not exclude any portion of reality from qualifying as a unit. Perhaps we might then focus on the following question. Suppose that we (as infants or as prehistoric people) started out by experiencing the world from the space-time vantage point. Could we at that level have discovered something about the world which would have taught us to adopt our ordinary criteria of bodily unity? That is, could we have discovered something that would have induced us to convert to common-sense body- mindedness? It seems fairly clear that purely theoretical motives could not provide this inducement. Neither science nor metaphysics, the two repositories of good theory, have ever been much enthralled by our common-sense criteria of unity. This relates to the point that I made earlier vis-à-vis Kant: that the common-sense concept of a body cannot be equated with the scientist’s or metaphysician’s concept of underlying matter. A creature with purely theoretical needs, who started out without any criteria of unity, might possibly develop some concept of matter conservation, or he might plunge directly into the physics of fields and space-time manifolds. At any rate, he could scarcely have any reason to take a detour through the spe- cific conceptual concoction that constitutes common-sense body- mindedness. 18 quantifier variance and realism: essays in metaontology

A more likely suggestion would be that practical rather than purely theoretical needs might motivate the adoption of our ordinary criteria of bodily unity. Such a suggestion seems rather common in pragmatist literature. For example, William James says: “But what are things? Nothing . . . but special groups of sensible qualities, which happen practically or esthetically to interest us, to which we therefore give substantive names, and which we exalt to this exclusive status of independence and dignity.”l3 The hint of phenomenalism in James’s remark is not relevant to the present issue. What is relevant is James’s suggestion that our practical (and aesthetic) interests are what induce us to “dignify” certain portions of reality as unitary things. Translated into the model I am considering, James’s idea is that, if we had originally started out with the space-time standpoint (i.e., without any criteria of unity), our practical concerns would eventually convert us to body- mindedness I do not think that James’s suggestion can withstand careful refl ection. Our concerns and interests are almost never directed toward objects taken one at a time, but are directed rather toward a multiplicity of objects related in various complicated ways. A typical concern, as ordinarily conceived, would be to alter the relations between objects, say, to bring them closer together or further apart, to attach or detach them, to replace one with the other, and so on. But any such concern might be said to embrace that whole portion of reality which includes those objects or their relevant stages. If, for example, I want to bring a chair nearer to a table, the target of my concern might be described as that space-time chunk of reality which includes the table and the chair (or perhaps their stages during the period when their increased prox- imity is of concern to me). One might say roughly that my concern here is to replace a wider or less compact table-cum-chair with a nar- rower or more compact table-cum-chair. Of course that is not how I would ordinarily describe my concern. When I ordinarily describe my concern I do so by conceptually dividing reality in the ordinary way; I talk about a unitary chair and a unitary table, and not about any such thing as a table-cum-chair. This is because the ordinary concept of unity is already given, and I describe the target of my concern in terms of that concept. But there is apparently nothing in the concern as such which could explain why I operate with that concept of unity, why I divide reality up in just that way. And this same point could be made for virtually any example one cares to consider. Contrary to James’s

13. The Principles of Psychology (New York: Henry Holt, 1890), I, p. 285. a sense of unity 19 suggestion, there appears to be no clearcut connection between our interests or purposes and our concept of unity. It is perhaps true that we tend to classify objects, especially artifacts, in terms of their aptness to fulfi ll human purposes. But this is a comment about our classifying tendencies, not about our unifying tendencies. The units are already given, and we classify them from a practical standpoint. There are all kinds of space-time portions which, if we only treated them as units, we could easily classify in practical terms as well. James’s idea is that an object is a portion of reality which is especially spotlighted by our needs and interests. But the fact is that our human concerns striate space-time in a manner that crisscrosses and overlaps the space-time paths of objects in every imaginable way. There seems to be nothing in our practical standpoint which can account for the specifi c concept of unity we have. It seems to me that there is no way that the “empiricist” position under discussion can overcome these diffi culties. If we imagine ourselves as having started out without our ordinary criteria of unity, there is nothing that we might have learned about the world, or about our practical relationship to the world, which could have yielded those particular criteria. In criticizing the empiricist position, I have been using the word “learning” in a rather loose and intuitive way. Certainly the prospects for the position would be immediately diminished if one limited learning to the sort of stimulus-response model that Quine sometimes seems to favor. But my criticism of the position does not rely on any particular analysis of what learning consists in. Taking the concept of learning in what seems to be the broadest sense relevant to this discussion, as roughly equivalent to “inference” or “rational derivation,” the crucial point remains just this: If we had started out by describing the world without our ordinary criteria of unity, then there seems to be no inference of any sort whatever that could have led to our adopting those criteria. My argument against the empiricist view has taken the rather extreme form of denying not only that we do in fact arrive at our criteria of unity by some kind of inference; I deny that we could even in principle have arrived at these criteria in any such manner. According to my version of the innateness hypothesis, we are innately disposed to adopt the criteria of unity that we have; but, apart from this specifi c and complicated disposition, there is nothing about the world which could have taught us to adopt our criteria of unity, since there are no considerations, theoretical or practical, which mark off just those criteria as being especially right or reasonable. It might seem initially tempting to 20 quantifier variance and realism: essays in metaontology maintain an innateness view somewhat diff erent from mine. It might be maintained, contrary to my view, that our criteria of unity are rationally derivable in principle, that they are the peculiarly right criteria for us to have, and that precisely for that reason we have evolved the innate propensity to operate with those criteria.14 But, if my arguments in this section against the empiricist position have been successful, then they show that our criteria of unity are not even derivable in principle.

V. Focusing on Objects

I have discussed three general approaches to our concept of bodily unity. According to one approach, our criteria of unity are essentially arbitrary conventions which could easily have been otherwise. This idea seemed intuitively implausible, though I have yet to present a def- inite argument against it. A second possibility was that these criteria are somehow derived from some facts about the world, facts that one could describe without presupposing the criteria. I tried to argue in the last section that this view is untenable. So we are left with the third possibility, which is that our criteria of unity are neither arbitrary conventions nor learned, but are determined by our innate disposition to experience the world in terms of just those criteria. It should be understood that the issue posed by these alternatives relates most directly to the bare foundations of our concept of bodily unity. To believe in an innate sense of unity is to suppose that our most basic and general criteria of unity result from an innate disposition. This does not preclude the possibility that these criteria may be enriched and elaborated in various ways, indeed in ways that are likely to include an element of convention as well as an element of practical and theoretical reasoning. The conventionalist and the “empiricist” are not saying merely that various nuances of our fully developed concept of unity may result from convention or inference. They are denying an innate status to any of our criteria of unity, even such seemingly fundamental criteria as cohesiveness and continuity. The position I have been arguing for is that these most fundamental criteria must be innate. Earlier, when discussing the conventionalist position, I suggested that one might try to clarify some of these issues by refl ecting on how

14. This seems in fact to be KÖhler’s position; see Kohler, op. cit., pp. 162–164. My own position might be compared to Chomsky’s view that “there is no a priori ‘naturalness’ to such a system [of innate grammatical principles], any more than there is to the detailed structure of the visual cortex.” See Noam Chomsky, Language and Mind (New York: Harcourt Brace Jovanovich, 1968), p. 88. a sense of unity 21 children are likely to experience unity before the learning of language. In this connection one naturally thinks of Jean Piaget’s discussions of the child’s development of the object concept.15 According to Piaget’s scheme, the child’s concept of an object invariably passes through a succession of stages. The principle of transition from one stage to the next is characterized by Piaget in a somewhat elusive manner. In some cases the transition seems to be essentially nothing more than matura- tional development, but the more fundamental idea seems to be that each successive stage resolves with increasing success various conceptual confl icts and tensions that arise at earlier stages. We are invited to compare this process to the development in theoretical science of increasingly more adequate ways of thinking about the world. All of this would obviously repay close examination. My concern at present, however, is less with Piaget’s account of these transitions, than with his depiction of the very earliest stages. Would Piaget agree with my contention that our experience is from the very start governed by our most basic common-sense criteria of unity? In the initial stages, according to Piaget, the child’s orientation is essentially solipsistic; the infant does not initially appreciate that objects can persist when they are not perceived. Piaget bases this interpretation of the infant’s experience on his (somewhat controversial) fi ndings that infants at an early stage do not engage in any “search behavior.” If an infant at this stage “is reaching for an object that is interesting to him and we suddenly put a screen between the object and him, he will act as if the object not only has disappeared but also is no longer accessible.”16 The infant, as here depicted, seems to treat the object in rather the way in which we would treat an after-image, as something whose esse is percipi. Now the most striking feature of this account, from my present standpoint, is the way that Piaget unabashedly describes the infant as reaching for an object, as being interested in the object, and as, apparently, noticing the object’s disappearance from his fi eld of view. In fact, Piaget consistently describes infants in very early pre-linguistic stages as focusing on objects and following them as they move. These descriptions seem perfectly natural, even inevitable. But it is important to see that these descriptions imply that the infant’s experience is directed toward units approximating to our ordinary things, units that he can focus upon,

15. The Construction of Reality in the Child (New York: Balantine Books, 1954), ch. 1; Genetic Epis- temology (New York: Columbia, 1970), pp. 43/4, 52–57; On the Development of Memory and Identity (Worcester, Mass.: Clark UP, 1968), pp. 17–37. 16. Genetic Epistemology, op. cit., p. 43. 22 quantifier variance and realism: essays in metaontology reach for, follow with his eyes, lose sight of, and so on. In the initial stages, at least according to Piaget, these units are treated as having a status akin to after-images (or akin to the philosopher’s sense data), and, like after-images, their unity seems to be defi ned in terms of such familiar criteria as cohesiveness and continuity. Hence Piaget apparently would agree that our experience is at the very earliest stages—certainly before the acquisition of language—governed by our most basic common- sense criteria of unity. Indeed it may not even be completely clear that the child’s drift away from solipsism, as characterized by Piaget, deserves to be counted as a development in the child’s concept (defi nition) of unity, rather than merely an alteration in the child’s beliefs about the unperceived persistence of the units he has picked out. It may be somewhat incautious of me to try to relate Piaget’s highly complex thesis to the present discussion. I can, however, more confi - dently cite the recent work of T. G. R. Bower, in which a modifi ed version of Piaget’s developmental scheme is expounded and impres- sively argued. The point that presently concerns me is Bower’s depiction of the infant’s experience in early pre-linguistic stages.

The evidence that infants do segregate their environments into units is clear. A large number of studies on the eye-fi xation behavior of infants has shown that infants will fi x on the external contours of objects in their visual fi eld. If the objects are moved, the infants will track them. If after moving together, the contours of an object break and begin to move independently, very young infants will display massive surprise. This indicates that the common motion (common fate) had specifi ed for them a single unit.17

The condition that Bower calls “common motion” (or “common fate”) is essentially what I earlier called “dynamic cohesiveness.” As I pointed out, this condition presupposes various criteria of unity through time, such as qualitative and spatiotemporal continuity. So Bower is in eff ect saying that the infant’s experience is pre-linguistically organized around units that are defi ned in terms of such fundamental common-sense criteria as cohesiveness and continuity.18

17. Development in Infancy (San Francisco: W. H. Friedmand, 1974), p. 102. 18. For the purposes of this discussion I am deliberately leaving it quite vague just what our “fundamental criteria” are, except to suggest that they would undoubtedly involve some appeal to continuity and cohesiveness. I attempt a more thorough presentation of these fundamental (or basic) criteria in ch. 4 of The Persistence of Objects, op. cit, esp. pp. 42–44. For some weird complications, however, see Bower, op. cit., pp. 189–192. a sense of unity 23

To speculate about the experience of infants may seem a rather dubious undertaking for a philosopher. But I would submit that the facts I am here rehearsing are, for the most part, so commonplace, and seem so central to our intuitive grasp of what human nature consists in, as almost to invite the designation “transcendental.” Try to imagine a person whose eyes and hands do not fi xate upon objects in an essentially ordinary way, but whose attention meanders about without ever settling on (what we ordinarily regard as) a single object. Can we imagine what it would be like to initiate such a creature into our ordinary thought- world? Where could we begin? I am now taking back my earlier tentative concession to Quine that perhaps the child’s quality space is essentially all that is required to explain how the child can learn the body-minded way of talking. Something else that is required is the child’s instinct to focus on objects in the ordinary way. This instinct is just another form of what I have been calling our “innate sense of unity.” In order for the child to learn the ordinary way of talking, he must already be focusing on objects in a manner that exhibits his disposition to adopt our basic criteria of unity. The easiest case to refl ect upon is one in which the infant is tracking a moving object. Here the characteristic alterations in the infant’s eyes, face, and body vividly display the eff ort to keep an object in focus. For the infant, as indeed for us, a moving object evidently stands out as something to be focused upon. When a moving object passes through our fi eld of view, we, even as adults, experience the unmistakable impulse to fi x our gaze on it and follow it as it moves. Our various purposes, expectations, concerns—our whole “mental set”—will eventually determine where we look, and for how long. But the primitive power of an object, especially a moving object, to fi x our attention is unmistakable. A point that I especially want to stress, in the light of Quine’s approach to this topic, is that our quality space (our sense of similarity) cannot account for our disposition to focus on objects in the way we do. The infant who is tracking a moving object is not merely register- ing passively some complex similarities between the presented scene and various other space-time portions of the world which, as he later learns, are called “moving bodies.” He is exhibiting the quite irreducible instinct to direct his attention in a distinctive way, by correlating the position of his eyes (or hands) with the position of the object he is tracking. It is, I am inclined to say, strictly a logically contingent fact that the focus of the infant’s (and our) attention tends (however briefl y) to follow the path of an object, rather than to meander through space- 24 quantifier variance and realism: essays in metaontology time in any number of other imaginable ways. However, as I remarked a moment ago, this fact seems completely central to our way of experiencing the world. What it means for us to focus (our attention) on an object can be explained partially by reference to the manifest correlations between the movements of our eyes and hands and the movements of the objects in our surroundings. At a deeper, and somewhat more nebulous, level our focusing propensities can be seen as providing us with a general epistemic orientation toward the world, with a general schema for learning. Consider, for example, how readily we learn about the shape of an object, though an object’s shape is equivalent to a complicated fact about the interrelations between the object’s parts. Compare this with the relatively slower and more arduous process of learning about the interrelations between things that do not add up to an object, as ordinarily conceived. Or consider how readily we discern the various changes (e.g., the patterns of movement) of a single object, though these changes are equivalent to complicated interrelations between the object’s temporal stages. It is generally far more diffi cult for us to size up the interrelations between object-stages that do not add up to what we ordinarily regard as a single persisting object. We can say, in general, that the qualities and internal constitution of a unitary object, as ordinarily conceived, are far more readily discerned than the qualities and internal constitution of any other kind of space-time portion. (In behavioral terms we can say that we are far more likely to “respond” to the former properties as “stimulus,” than to the latter.) Our heightened readiness to focus upon, and hence to learn about, an object, rather than any other kind of space-time portion, is evidently central to what it means for us to regard a portion of the world as a unitary object. Some of the points that I have just been discussing can be illustrated by reference to Figures 2 and 3 (which are adapted from KÖhler). The reader will readily discern that there is a central portion of Figure 2 shaped like the letter H. Let me call that portion “α.” On the other hand, one is not apt to notice that, embedded within the center of Figure 3, there is also an H-shaped portion. I will call that portion of Figure 3 “β.” The evident fact, then, is that we readily tend to focus on α, but not on β. This fact about our focusing propensity, to recast an earlier point, cannot be accounted for by reference to our sense of similarity. What is peculiar about β is not that we fail to see it as similar to this or that, but rather that, in a sense, we fail to see it, period. Once β is pointed out to a sense of unity 25

figure 2

figure 3 us, however, we readily see it as being similar to various things (if nothing else, at least to corresponding portions of replicas of Figure 3). This distinction between α and β can be expressed in terms of our disposition to learn. If presented with Figure 2, we would be apt to learn various things about α, for example, that it is shaped like an H, or that it contains three line segments, or that it is not round. By contrast, we are not likely to learn anything at all about β when we are presented with Figure 3. (In behavioral terms, the idea would be that α’s properties are far more likely than β’s to elicit a response from us.) Now a space-time portion of the world which does not correspond to what we ordinarily count as a unitary object can be compared to β: we do not readily focus on (or learn about) such a portion of the world. One might almost say that, primitively, a unit is just this: a portion of reality that we naturally focus on. Or, to put this from another angle, one might almost say that the root impulse behind our body-minded ontology is to exclude those portions of reality which “we do not see,” in the peculiar sense in which we do not see β. I have been suggesting that there is a close connection between the following two ideas: (1) “experiencing (and thinking about) bodily unity in an essentially ordinary way” and (2) “exhibiting essentially ordinary focusing and tracking behavior in the presence of bodies.” I 26 quantifier variance and realism: essays in metaontology am uncertain just how close we ought to say this connection is. (This seems of a piece with the uncertainty one feels about the closeness of the connection between, for example, anger and anger-behavior.) Even most modestly construed, the connection seems amply to warrant the conclusion that before the acquisition of language very young infants experience bodily unity in a manner akin to adults. Most modestly construed, the connection is explanatory: We explain people’s focusing and tracking behavior by reference to their experience of unity. Put in these terms, my point about infants is this: A seemingly plausible (indeed a seemingly compelling) explanation of the infant’s focusing and tracking behavior, and the similarity of that behavior to our own, is that the infant experiences the world as broken up into units essentially the way that we do. Someone who wants to reject this explanation certainly has the burden of suggesting an alternative.

VI. Conclusion

My argument against the empiricist position consisted in maintaining that there is nothing about the world which could have taught us to adopt our ordinary criteria of unity, had we started out without those criteria. My argument against the conventionalist view was basically to suggest that children experience unity in an essentially ordinary way before they acquire language. Perhaps this latter argument also works to some extent against the empiricist view, for it may seem immediately implausible to suppose that very young infants have already arrived at their experience of unity by way of learning. Our concept of bodily unity, or at least the basic core of that concept, is rooted in our primitive, pre-conventional experience of unity. And it seems that only our innate constitution can plausibly account for the specifi c and complicated conditions that a portion of the world has to satisfy if it is to be experienced primitively as a unit. As far as defi ning what enters into this element of innateness, one point I have repeatedly stressed is that our quality space, our sense of similarity, cannot explain why we experience unity in the way we do; nor can it account for our correlative focusing and tracking behavior. The conclusion suggested by this whole train of argument is that an innate sense of unity is a quite irreducible feature of our experience of the world. two c C

basic objects: a reply to xu

C In “From Lot’s Wife” Fei Xu defends the thesis that adults as well c as infants have the concept of a category of physical objects whose identities do not depend on such relatively specifi c sortals as “dog” and “cup.” (All page references are to Xu, 1997.) Let me call such objects “basic objects.” Whereas I agree with the essentials of Xu’s position on basic objects I have doubts about some of its details, and especially about the experimental evidence off ered on its behalf. Following Spelke, Xu suggests that a basic object is a “bounded, coherent, three-dimensional physical object that moves as a whole” (p. 365), and she takes that defi nition to imply that “part of an object is not an object so long as that part does not fall off and start to move independently on its own” (p. 387). She infers, in other words, that a basic object must be a maximal object, in the sense of not being part of another basic object. Some of her own examples seem to show, however, that this inference is mistaken. The rotating wheel of a (stationary) car moves on its own while remaining part of the car; the same holds for a windshield wiper. Countless such examples seem to show that, on Xu’s defi nition, basic objects need not be maximal. Xu may well be right in suggesting (p. 384) that when we are asked to “count the objects” in a place we often take “object” to mean “maximal object,” but I think “object” in this sense cannot be equated with “basic object.” Another problem with Xu’s characterization of basic objects is that it seems to exclude essentially stationary objects, such as trees, fences

27 28 quantifier variance and realism: essays in metaontology and fi re hydrants. I think it is highly doubtful that any ordinary concept of an object—certainly none employed in adulthood— systematically excludes such objects. A more promising approach, it seems to me, is to recognize degrees of articulation as a function of various familiar gestalt conditions, including boundary contrast, common motion (fate), qualitative homogeneity, internal coherence, and good form (Hirsch, 1982). A basic object is constituted by a continuous stretch of matter that is suffi ciently articulated. An immobile object such as a fi re hydrant may be seen as a basic object despite lacking the condition of common motion because it satisfi es enough of the other conditions to a high degree. Of course, if we are talking about degrees of articulation then we are also talking about degrees of object-basicness, and the idea must be that some stretches of matter are more easily or naturally seen as enduring objects than others. Xu would like basic objects to be maximal in order to avoid a certain problem (p. 387). The problem is that if one basic object is part of another it may be impossible to trace the careers of these objects by appealing to continuity considerations. Imagine a tree that consists of a trunk and a minuscule branch. It may be that both the tree and the trunk are suffi ciently articulated to be naturally seen as basic objects. A space-time path that jumps from the whole tree to the trunk (with the little branch remaining intact) will be highly continuous, and will therefore qualify as a continuous path traced under the concept “basic object,” but this path is decidedly not the path of any basic object—it is not the path of the tree, nor the path of the trunk, nor the path of anything else that one could naturally see as an enduring object. Such examples show that continuity under “basic object” is not suffi cient for identity. To deal with this kind of problem I suggested in my earlier work that we need to appeal to a condition of minimizing change (Hirsch, 1982). Jumping from the whole tree to the trunk would introduce changes of size and shape that could be avoided by sticking to the tree. The rule for tracing the identity through time of a basic object is to follow a continuous and change-minimizing succession of highly articulated bits of matter. Xu seems to accept the principle that if “F ” is a sortal then any continuous succession of F-stages are stages of the same enduring F-thing. In my own work this principle was taken as defi nitive of what it means for a term to qualify as a sortal. It is clear from the points I have been making that “basic object” is not a sortal in this sense. I don’t basic objects: a reply to xu 29 doubt that it might be called a “sortal” in some other sense. In what follows I will talk of “basic objects” and “sortal-dependent objects,” meaning by the latter expression objects whose identities depend on specifi c sortals like “cup.” A key idea in Wiggins’s writings is that we can have no clear grasp of what the identity of an object consists in except by appealing to specifi c sortals (Wiggins, 1980). I accept that insight, and regard it as compatible with what I have been saying about the notion of a basic object. That notion is enormously vague, amorphous, and even in a sense “subjective.” Very often there will be no reasonable way of deciding whether a stretch of matter is “suffi ciently articulated” or whether a space-time path “minimizes change.” Although even our sortal- dependent identity concepts are surely infected with a great deal of vagueness, they are far clearer and more determinate than the notion of a basic object. Nevertheless, I think it is plausible to maintain that the notion of a basic object, however vague and amorphous, has a genuine place in our overall conceptual scheme, and probably plays a central role in the infant’s view of the world. My main reason for thinking that basic objects fi gure in our ordinary adult conceptual scheme derives from a thought experiment (mentioned by Xu on p. 387) in which we imagine someone con- fronting unfamiliar kinds of objects and being unable to apply any specifi c sortals to them. It seems obvious that the person will still be able to pick out “this object” and “that object” and trace the identities of the objects in some manner. When we carefully think about this kind of case I think we realize that what the person is doing is picking out articulated bits of matter and following continuous and change- minimizing successions. In this imagined context of sortal ignorance what is meant by “this object” and “that object” are basic objects in something like the sense I have explained. For adults the world is most saliently broken down into sortal- dependent objects, with the basic objects vaguely lurking in the background. That is Xu’s position, and also mine. How the world might be viewed by infants is, of course, much more diffi cult to assess. Although Xu’s discussion of this problem is among the most acute I have seen, I think she still falls short of fully appreciating just how diffi cult the problem is. Let me fi rst consider what Xu says about the experiment presented in Xu and Carey (1996). In this experiment babies are shown one kind of object, say a ball, emerge briefl y from behind a screen and return, and a little later are shown another kind of object, say a bottle, emerge 30 quantifier variance and realism: essays in metaontology briefl y from behind the screen and return (Xu’s fi gure 5). From an adult’s perspective the “expected outcome” when the screen is removed is to fi nd (at least) those two objects; the “unexpected outcome” is to fi nd only one of the objects. The fi ndings were that infants are not more surprised by the one-object outcome until about 12 months. Moreover, the younger infants’ lack of surprise was not due to their not noticing the property diff erences (e.g. of shape and behaviour) of balls and bottles. Habituation experiments revealed that 10-month-old babies do notice the relevant property diff erences, but still do not exhibit more surprise at the one-object outcome. The conclusion drawn from these experiments by Xu and Carey was that 10-month- old infants do not, and 12-month-old infants do, view the experimental situation in terms of sortal-dependent objects. There seems, however, to be a serious problem with this argument. It seems in fact that Xu and Carey face questions from two directions. On the one hand, 10-month-old infants may very well operate with sortal-dependent identity concepts, and perceive the experimental situation as involving a ball and a bottle, but simply have no inkling that there is anything unexpected about a ball going out of existence and being replaced by a bottle. Just as adults are not especially surprised by a car going out of existence and being replaced by a pile of scrap, or an ice cube going out of existence and being replaced by an expanse of water, or a dog going out of existence and being replaced by a corpse—and there are countless such examples—the 10-month-old infant may have no reason to be especially surprised by a ball going out of existence and being replaced by a bottle. On the other hand, the 12-month-old infants may continue to operate with the sortal-independent basic identity concept of early infancy, but they may already know that there is something very surprising about an object persisting through a change from ball-properties to bottle- properties. Oddly, Xu is keenly aware of the second question, but seems to ignore the fi rst. She rightly observes that “success in this task [of inferring that two objects will be found behind the screen] may or may not show that the infants represent sortals such as ball or bottle (p. 377)”— thus indicating her awareness of the second question—but she goes on to assert that “the failure [to make the inference] certainly suggests that they do not represent sortals ball or bottle (p. 377),” thereby ignoring the fi rst question. This is puzzling. My objection is especially obvious given Xu’s (and my) assumption that the concept of a basic object continues to operate into adulthood basic objects: a reply to xu 31 side by side with the sortal-dependent concept. This means that when we, as adults, are confronted with the unexpected outcome in the Xu– Carey experimental setup, we can legitimately describe it in either of the following two ways:

It’s as if a ball has gone out of existence and been replaced by a bottle. or: It’s as if a (basic) object has persisted while changing from ball- properties to bottle-properties. It seems evidently wrong to suggest that our surprise as adults will be greater if we describe the situation the fi rst way rather than the second. Why, then, should it be supposed that the infants’ surprise will be in any way aff ected by whether they perceive the situation in the rstfi way (in terms of the sortal-dependent concept) or the second way (in terms of the basic concept)? The issue that I am raising with respect to the Xu–Carey experiment is one instance of a more general problem that I think pervades the recent infant cognition literature on the genesis of our object concept. A number of discussions have attempted to derive theories of infant ontology from data about infants’ surprise reactions. These discussions seem to ignore a fundamental diffi culty that can be outlined as follows. If some situation S elicits surprise reactions in infants this indicates that the infants have expectations incompatible with S. Ontology deals with ways of breaking S into units, such as, stuff s, objects, processes, momentary states and events, and clusters. In any interesting example, there will be several coherent ontological schemes that can be applied to S. Each of these ontologies will be capable of supporting expectations that are incompatible with S and that would cause S to elicit surprise. Hence, the infants’ surprise reactions to S cannot apparently reveal which ontology enters into the infants’ beliefs. In order for the infants’ surprise reactions to provide evidence with respect to the infants’ ontology some principles are needed that say that certain ontological schemes are more likely than others to support certain kinds of expectations. Such principles, however, do not seem to have been formulated or defended in the literature. With respect to the Xu–Carey experiment no principles are presented that make it clear why infants are more likely to be surprised if they are operating with an ontological scheme of sortal-dependent objects rather than a scheme of basic objects. The same problem applies to the study of Spelke et al. (1995)—I will call this the SKSW experiment—that purports to show, in Xu’s words 32 quantifier variance and realism: essays in metaontology

(p. 372), that “infants as young as 4 or 5 months use . . . spatiotemporal generalizations to individuate objects and trace identity, just like adults.” In this experiment (Xu’s fi gure 1) the infants were found to be surprised by a situation that might be described in ordinary terms as follows: “It’s as if an object moved from the back of one screen to the back of another screen without passing through the space between the screens.” (Alter- natively: “It’s as if one object vanished in the back of a screen and a second very similar object continued to exist in the back of another screen.”) If the experiment is supposed to demonstrate that the babies operate with a concept of an enduring physical object, we ought to have before us some alternative hypothesis against which to test this demonstration. Consider, then, the hypothesis that babies are natural Quineans, that is, they view any space-time portion of reality, however discontinuous or gerrymandered, as an object on an ontologically equal footing (Quine, 1960). The Quinean ontology, as Quine himself empha- sizes, is strikingly diff erent from the ordinary ontology of enduring objects that move continuously through space. Does the SKSW experiment undermine this hypothesis? How? If the babies are Quineans then they fi nd themselves in the experimental setup confronted with (what they perceive as) a certain kind of discontinuous space-time portion of reality. The fact that they are surprised shows that they do not expect to be confronted with that kind of discontinuity. Nothing is shown at all about their believing in an ordinary ontology of enduring physical objects. It must be understood that from a Quinean perspective the kind of situation that seems to be confronted in the experimental setup—that is, the kind of situation that we might describe in ordinary terms as an object seeming to jump discontinuously from one place to another—is highly unusual and anomalous, just as it is from the ordinary perspective. Quinean babies would do well, therefore, to fi nd such discontinuities surprising. The hypothesis of Quinean babies is of course just one of an indef- inite number that could be considered. There is the hypothesis that babies are Humeans who believe only in momentary events and their interrelations (Hume, 1739); or that they are Butlerians who believe only in merelogically intact substances that cannot survive the loss of any parts (Butler, 1736); or that they are Strawsonian feature-placers (Strawson, 1959). We can even consider the hypothesis that infants operate with one of my deliberately strange ontologies of in-out objects or contacti-objects (Hirsch, 1982, 1993). The SKSW experiment seems no more successful in rebutting any of these various basic objects: a reply to xu 33 hypotheses than in rebutting the Quinean one. If Spelke’s answer is that she has no interest in rebutting any such intuitively unmotivated hypotheses then what was supposed to be the point of the SKSW experiment? In reading and rereading Spelke’s enormously stimulating work (e.g. Spelke, 1982, 1984) I fi nd myself worrying that, as a philosopher reading a psychologist, I am not quite getting it. Often it seems clear to me that she is addressing the question that most interests me philosophically, namely: Do infants see the world in terms of enduring physical objects? But at other times it seems to me that she has slipped off that question and is rather asking the following quite diff erent question: Given that infants see the world in terms of enduring physical objects, what expectations do infants have about the behaviour of these objects? Xu, however, is quite clear in addressing the fi rst question (Xu and Carey, 1996 is perhaps even clearer in this regard), and she evidently thinks that the SKSW experiment answers this question. Unfortu- nately, she fails to explain how the results of the experiment serve to clarify which ontological scheme infants employ. I am afraid that the same complaint applies to every experiment she mentions in her paper that involves infant surprise reactions. What is absent is any explanation of how those surprise reactions tend to show that infants, rather than being Quineans or Humeans or Butlerians or various other alternatives, divide the world up into (basic) objects in essentially the way adults do. My argument may be pushing me towards a position of extreme scepticism about infant ontology, but I would like to stop short of that. I take as our starting point the need to formulate a theory about how infant ontology develops into adult ontology. Assuming that we know what the basic structure of adult ontology is, we need to fi ll in the story in two ways: fi rst, by describing infant ontology, and, second, by explaining the processes that lead from that ontology to the adult ontology. The basic constraint on how we develop this story is that we want it to be as simple as possible. It seems evident that, other things being equal, the simplest story will minimize as far as possible the cognitive distance between infants and adults. Hence it seems virtually certain that the simplest story will ascribe to infants an ontology of enduring physical objects rather than some unfamiliar ontology such as Quine’s. Moreover, in early infancy, prior to the time that infants give any indication of being able to distinguish between such sortal properties as “bottle” and “ball,” the simplest story will say that physical objects are traced in essentially the way that adults trace basic objects when in 34 quantifier variance and realism: essays in metaontology a state of sortal ignorance, that is, by following a spatiotemporally and change-minimizing path of articulated bits of matter. Within the context of such reasoning to the best and simplest explanation some of the experiments on infant surprise may play a modest role. The SKSW experiment, for instance, shows us that the infant’s cognitive reaction to continuities and discontinuities is in some ways strikingly like that of adults. This experimental result surely does not directly refute the hypothesis of Quinean babies (we do not have in that sense a “crucial experiment”), but it does somewhat enhance the simplifying argument from analogy between infants and adults, and makes it even easier to believe that the ontology of infants is probably in many ways very similar to that of ordinary adults. It must be noted, however, that even this minimalist interpretation is not obviously available for many of the experiments on infant surprise. In particular, it is not obvious how the results of the Xu–Carey experiment enhance the simplicity of the hypothesis Xu and Carey advance as to when infants represent sortal-dependent objects. I would conclude, then, that on grounds of simplicity we can perhaps plausibly conjecture that very early in life infants begin to trace the identities of basic physical objects along continuous change-minimizing paths. I am sceptical that much more than that can be said about infant ontology. Theorists of infant cognition, such as Spelke and Xu, have tried to say more on the basis of experiments on infant surprise reactions, but they owe us a clearer formulation and defence of the principles connecting infant ontology to infant surprise.

References Butler, J. 1736: Of Personal Identity, Appendix 1 of The Analogy of Religion. Hirsch, E. 1982: The Concept of Identity. Oxford University Press. Hirsch, E. 1993: Dividing Reality. Oxford University Press. Hume, D. 1739: A Treatise of Human Nature. Quine, W.V. 1960: Word and Object. Cambridge, MA.: MIT Press. Spelke, E.S. 1982: Perceptual Knowledge of Objects in Infancy. In J. Mehler, E. Walker and M. Garrett (eds), Perspectives on Mental Representation. Hillside, NJ: Lawrence Erlbaum Associates. Spelke, E.S. 1984: Perception of Unity, Persistence, and Identity: Thoughts on Infants’ Conceptions of Objects. In J. Mehler and R. Fox (eds), Neonate Cognition: Beyond the Buzzing Blooming Confusion. Hillside, NJ: Lawrence Erlbaum Associates. basic objects: a reply to xu 35

Spelke, E.S., Kestenbaum, R., Simons, D. and Wein, D. 1995: Spatiotemporal Continuity, Smoothness of Motion and Object Identity in Infancy. British ]ournal of Developmental Psychology, 13, 113–42. Strawson, P.F. 1959: Individuals. London: Methuen. Wiggins, D. 1980: Sameness and Substance. Cambridge, MA.: Harvard University Press. Xu, F. 1997: From Lot’s Wife to a Pillar of Salt. Mind and Language, 12, 365–92. Xu, F. and Carey, S. 1996: Infants’ Metaphysics: The Case of Numerical Identity. Cognitive Psychology, 30,111–53. three c C

objectivity without objects

C The position I want to present here is, I think, closely related to c some things Hilary Putnam has been saying—and associating with the philosophy of Wittgenstein—in his most recent work, especially his 1993 paper “The Question of Realism.”1 My formulation, however, may strike you as quite diff erent from Putnam’s, and I’ll have to try to explain what the connection is. My most central point is that we can imagine and understand languages whose words do not refer to anything but whose sentences express truths equivalent to those expressed in English. For such languages we cannot formulate any standard referential semantics. I’ll give a rough sketch of one such language. In this language the word “car” is replaced by the two words “incar” and “outcar” that function as follows: With respect to any situation in which a car leaves a garage a true sentence in this language is “There is an incar that shrinks and vanishes at the door of the garage and simultaneously there is an outcar that appears and grows and then moves down the street.” An incar corresponds in a way to the succession of car-parts left in the garage as the car departs, but whereas successions (or other set-theoretical items) do not, I assume, move or shrink, in this other language the sentence “An incar moves and shrinks” is true with respect to the situation in which a car leaves a garage.2

1. Reprinted in Hilary Putnam, Words and Life (Harvard University Press, Cambridge, Mass., 1994). 2. I have discussed this language in The Concept of Identity (Oxford University Press, N.Y., 1982), and Dividing Reality (Oxford University Press, N.Y., 1993).

36 objectivity without objects 37

Given my limited time, I cannot spell out the details of this language, or the various directions that it might be fi lled out. The general directive I’m laying down is that you should imagine the language developed in ways that make it as close to English as is allowed by the constraints on truth-conditions that I’ve specifi ed. I hope that these few remarks are enough to indicate the general kind of language I am imagining and why—for this is the essential point—I am saying that in this kind of language the words don’t refer to anything. I take it to be an obvious tenet of common sense sanity that when a car leaves a garage no object whatsoever shrinks and vanishes. Of course, given the nature of our business, many philosophers in fact deny this—as I call it—tenet of common sense sanity. I’ll say something more about those philosophers presently. But given the common sense assumption it immediately follows that the word “incar” in the imagined language does not refer to an object that shrinks and vanishes when a car leaves a garage. What could the word “incar” refer to, then, that makes the sentence “An incar shrinks and vanishes” come out true when a car leaves a garage? Supposing, as I do, that the English sentence “No successions shrink or vanish” is true, and imagining the other language as close to English as possible, that sentence is also to be imagined as true in the other language, and so, therefore, is the sentence “No successions are (identical with) incars.” This seems to make it impossible to hold— at least in any obvious way—that the word “incar” refers to certain successions. You can try to play around with some more complicated possibilities but I think it becomes clear pretty quickly that none of them work. We have to conclude that in this other language sentences containing the word “incar” have truth-conditions that are in some way dependent on the occurrence of that word, but the word doesn’t refer to anything. Moreover, if the truth-conditions of the sentence “The incar shrinks” do not depend on the reference of the word “incar,” these truth-conditions evidently cannot depend on the reference of the word “shrinks.” Generalizing the example we can imagine that every count (or individuative) noun in that language operates on the model of strange words like “incar” and “outcar,” so we have “onbooks” and “off - books” depending on whether a book is on a shelf, “inpigs” and “out- pigs” depending on whether a pig is in a pen, and so on. If none of the count nouns of the language refer to anything then it seems to follow that neither do any other words. (If in a grammatically basic sentence of the form “The N is A” the noun “N” does not refer to anything then the truth-conditions of the sentence cannot be determined by the 38 quantifier variance and realism: essays in metaontology

reference of the adjective “A”; and if “A” does not have a referential role in such basic sentences it is hard to see how it could have any such role in other kinds of sentences.) I hope it is clear that I am not off ering up another example of inscrutability of reference, which is what results when we have an embarrassment of too many adequate reference schemes for a language. In my example there are no adequate reference schemes; there is no way of mapping words onto objects which will yield the truth- conditions of the sentences. What we have in my example is scrutably the absence of reference. Suppose we consider the somewhat more modest example in which words like “incar” and “outcar” are simply added to (but do not replace) ordinary words like “car.” Since in this language a sentence like “There is a car here” will have the same truth-conditions as in English we may be justifi ed in saying that the word “car” refers to cars in that language. But if we say this we need to recognize the limited role played by the reference of the word “car” in explaining the truth conditions of sentences containing that word. For example, we can’t explain the truth- conditions of the sentence “That car is smaller than this incar” by appealing, as in the standard referential semantics, to the fact that the term “smaller than” refers to any ordered pair whose fi rst member is referred to by “car” and whose second member is referred to by “incar”—we can’t give this explanation because nothing is referred to by “incar.” For the same reason we can’t give the standard referential semantic explanation for the truth-conditions of the sentence “That car and this incar are both white.” So we can say, if we wish, that in the hybrid language—the one containing both “car” and “incar”—the ordinary word continues to refer in the ordinary way, but this remark will be directly relevant only to certain isolated stretches of the language (such as the sentence “There is a car here”). The fact that remains—and this is the fact I’m interested in emphasizing—is that even for the hybrid language we have no reference scheme: there is no way of mapping words onto objects which will yield the truth- conditions of the sentences of the language. Perhaps the point I am making will become clearer—and less intuitively jarring—if we relate it to some of Putnam’s most recent formulations of his “conceptual relativism.” In “The Question of Realism” Putnam says that “all situations have many diff erent descriptions, and that even descriptions that, taken holistically, convey the same information may diff er in what they take to be ‘objects’ ”. Hence, “there isn’t one single privileged sense of the word ‘object’ ”. Rather, “there are objectivity without objects 39 many usable extensions of the notion of an object.”3 My example of the Incar-language is an illustration of one way in which our notion of an object could be extended. When I say that in that language the word “incar” does not refer to anything I am employing the ordinary unex- tended notion of an object, of something to be referred to. But I’m imagining that in the other language they do correctly assert the sentence “The word ‘incar’ refers to certain objects.” For, as Putnam explains, to imagine a revision in our concept of an object is necessarily to also imagine a correlative revision in a whole family of concepts, including the concept of what it is to refer to something. He says, “Accepting the ubiquity of conceptual relativity does not require us to deny that truth genuinely depends on the behavior of things distant from the speaker, but the nature of the dependence changes as the kinds of language games we invent change.”4 Given what we ordinarily mean by “something” and what we ordinarily mean by “reference,” we can’t correctly assert the sentence “In the other language the word ‘incar’ refers to something,” but the change in the meaning of “something” and the meaning of “reference” that is entailed by speaking the other language would make it correct to say “The word ‘incar’ refers to something.” The hard thing to realize—and I think that Putnam himself has had trouble realizing this—is that the position of “conceptual relativism” just outlined does not imply any kind of linguistic idealism. In his earlier work, for instance, in Reason, Truth, and History, Putnam was prone to say things like this: “‘Objects’ do not exist independently of conceptual schemes.” “. . . ‘objects’ themselves are as much made as discovered. . . .”5 We should note that in these remarks Putnam has “objects” in quotes. I don’t think he was ever really happy with the blatantly idealist position that objects do not exist independently of conceptual schemes. I think he was always aiming for the idea that the concept of an object—“objects” in quotes—depends on a choice of some particular conceptual scheme, a choice that is not settled by how the world is, because the world can be adequately described with many diff erent concepts of an object. But there is no doubt that some of Putnam’s earlier formulations led many people—and probably at times him too—to idealist absurdities. Happily he has now decisively

3. Putnam, “The Question of Realism,” op. cit., 304f. 4. Ibid., 309. 5. Hilary Putnam, Reason, Truth, and History (Cambridge University Press, Cambridge, 1981). The fi rst quote is on 52, the second on 54. 40 quantifier variance and realism: essays in metaontology distanced himself from his earlier idealist formulations. In his 1993 paper he says (in the course of contrasting his view with that of Rorty), “I have, especially in recent years, tried not to state my own doctrine as a doctrine of the dependence of the way things are on the way we talk. . . . In any sense of ‘independent’ I can understand, whether the sky is blue is independent of the way we talk.”6 The relativist non-idealist Putnam would say the following about my example: When a car leaves a garage nothing whatsoever shrinks. That is an objective fact not in any way dependent on language. Therefore, in this other language the word “incar” does not refer to anything. But people could have spoken that language, and, in terms of the concepts they would have then employed, it would have been correct to assert the sentence “Something—something referred to by the word ‘incar’—shrinks and vanishes.” In saying that speakers of the Incar language would have another concept of an object—rather than simply saying that they would have no concept of an object (since they don’t have ours)—we are tacitly drawing an analogy between how the word “object,” as well as closely related words and expressions, function in their language and ours. The analogy is partly formal or syntactic, for we are imagining that the formal logic of quantifi cation theory holds in their language the same as in ours. But there seems to be something more than a merely formal analogy here. When we imagine one of those people pointing in a certain direction and saying “That incar is shrinking,” it strikes us that they are doing something like referring to an object and saying that it’s shrinking. Now, in terms of our concept of an object there isn’t any shrinking object there to be referred to, but it’s in some sense as if there’s a shrinking object there for them. In saying this, I’m marking a linguistic or conceptual analogy which you can’t help seeing. It’s because you see this analogy that you so easily learn how the other language operates. I taught you the language ostensively, by giving you a few examples, on the basis of which you’re able to “go on.” In eff ect, I ostensively taught you a new concept of an object. The fact that you can so easily grasp how this new concept operates in the other language, because you tacitly grasp the relevant analogy between that language and ours, is what Putnam means by saying that there are many usable extensions of our concept of an object. Critics in the literature have often attacked him from both sides, posing the following dilemma for him: “Either Putnam is espousing a

6. Putnam, “The Question of Realism,” op. cit., 301. objectivity without objects 41 kind of lunatic linguistic idealism or he is making the utterly trivial claim, never denied by anyone, that we can use words in any way we like.” This criticism, however, is blind to the deep issues Putnam has been addressing. Let’s put aside the idealist horn of the dilemma. I am assuming that all of us, including the current Putnam, want a position that does not carry even a whiff of idealism (though we can’t be responsible for olfactory hallucinations). It’s the other horn that should interest us. Who ever denied that we can use language in any way we like? But what has to be seen here is that Putnam is not making the claim that the noise or symbol “object” could have been used in accordance with a convention that makes it mean “It is raining outside.” That would indeed be completely trivial. What he is claiming, however, is that the conventions of language might have been such as to alter the meaning of our whole referential apparatus, that family of interrelated expressions including “object,” “something,” “there exists,” “such that it,” “set,” and “refers.” To appreciate how nontrivial this claim is one has to recognize how often it is implicitly denied in some of the most infl uential literature of the day. If you look into some of the most prominent contemporary discussions in the philosophy of language—in particular, the work of Lewis, Davidson, and Kaplan—you are likely to fi nd an approach to semantics that starts out roughly as follows: “We have on one side some arbitrary language L and on the other side the (actual and possible) things that exist in the world. The semantic task is to specify a mapping of words in L onto things in the world whereby some Tarski-like compositional rules will generate the right truth-conditions for the sentences of L.” This formulation sounds so familiar, and so innocuous, that it is hard to see that it implicitly contains one of the most fundamental errors in metaphysics—the error that Putnam has for years been struggling with. The error is in assuming that what we mean by “things in the world” and by the rest of our referential apparatus must match the concepts employed by speakers of an arbitrary language L. If L is the Incar-language the assumed referential match simply does not occur. In terms of our ordinary concept of “things in the world” there may simply be no relevant mapping of the words of L onto things in the world. Of course it’s precisely at this point that we feel pressured to fall off the edge of linguistic idealism, and we have to struggle hard to resist that pressure. “Oh, so you’re really saying that there aren’t things in the world, that this is just a matter of what language we speak.” No, that’s not what we’re saying! In general, what things there are in the world has 42 quantifier variance and realism: essays in metaontology nothing to do with language. We can’t, in general, choose what things there are in the world. But in saying what I have just said I had to mean something by “things in the world,” I had to employ a certain concept of “things in the world.” And the “conceptual relativist” claim is that this concept might not be employed by speakers of another language. The standard referential semantics—represented in the literature by Lewis, Davidson, and so many others—implicitly denies this claim. In lodging this criticism of referential semantics my assumption has been that when authors like Lewis and Davidson publish articles in American journals their intention must be to use the English language, that is, the language that predominates in North America. But there is another possibility. Perhaps these philosophers have adopted a technical language, including a technical sense of “things” and the rest of the referential apparatus, for the specifi c purpose of doing semantics. Suppose that in their technical language it is correct to assert “Every portion of space-time, however gerrymandered, is an object.” Given this new concept of an object, and the correlative concept of referring to an object, it would be correct to assert “The word ‘incar’ refers to certain objects.” Along these lines it may indeed be feasible to present a referential-style semantics for any language intelligible to us. I don’t know what many referential semanticists would say to this point—I rather doubt that any of them would be happy to allow that they have all along been using a highly technical language without announcing this fact—but I do know for sure what one of the most eminent of these philosophers would say. I’m referring to David Lewis. Lewis is one of the most infl uential deniers of what I earlier called “common sense sanity.” Lewis claims to be using the ordinary concept of an object and to be correctly asserting—in the dominant language of North America—that every portion of space-time, however gerrymandered, is an object. There are in fact things that shrink and vanish when cars leave garages; this is a truth in ordinary language, though ordinary folk may be prone to deny it because of their tendency to use a “restricted quantifi er.”7 I think the short and decisive rebuttal of Lewis’s view is that if you try to tell ordinary folk about the gerrymandering objects which, on Lewis’s account, they are supposedly committed to, they will look for the nearest place to commit you to. But this is not an occasion on which I can seriously examine common sense philosophy and its enemies. I do, however, want to look briefl y at an argument Lewis gives against the common sense philosophical

7. David Lewis, On Plurality of Worlds (Basil Blackwell, N.Y., 1986), 213. objectivity without objects 43 position—as far as I know his only argument.8 This is an argument from vagueness, and we will see that it presupposes Lewis’s rejection of conceptual relativism. Lewis points out that, on the face of it, our ordinary concept of something seems to be vague. There are many sentences of the form “Something is composed of (the succession) of such-and-such bits of matter” which are indeterminate in truth-value, apparently because we are vague about what we are going to count as “something.” But that can’t really be, insists Lewis. It makes no sense to say that the quantifi er is vague, though our common sense judgments, taken at face value, seem to imply this. What must really be going on here is that we are often vague about how we want to restrict the quantifi er. It must be that every portion of space-time, however gerrymandered, really counts as an object, so that, strictly speaking, something surely is composed of any specifi ed (succession of) bits of matter. What may be indeterminate, however, is whether the composed thing satisfi es some vaguely intended contextual restrictions. It is not the concept “something” that is vague, but the contextual distinction between “unrestricted something” and “suitably restricted something.” What seems to be glaringly missing in this argument is any explanation from Lewis as to why he thinks that the quantifi er can’t be vague (as it seems on face value to be). But I think I can fi ll in Lewis’s argument for him. A concept is vague if it can be precisifi ed in more than one way. If the concept “something” is vague then there must be diff erent precisifi ed concepts of “something.” But that implies the basic relativist idea, which I think Lewis is fi rmly against, that there are diff erent equally legitimate concepts of “something.” If one believes that our concept of “something” is extendible in diff erent directions (as Putnam puts it) then one must surely also allow that the concept might be precisifi able in diff erent directions. Lewis will have none of this relativist stuff , and that forces him to deny that our common sense ontology can be taken at face value. Putnam, especially in his most recent writings, staunchly defends a common sense ontological position, while at the same time maintaining the doctrine of conceptual relativism. One fi nds that there is a very deep connection between these two strands in his thinking. You can’t be a common sense philosopher without being a conceptual relativist. The argument about ontological vagueness is an important part of this point, but let me state the connection a bit more broadly. Our ordinary

8. Ibid., 211–13. 44 quantifier variance and realism: essays in metaontology referential apparatus is not just vague, it is incredibly messy in various ways. If you try to divide up the totality of (actual and possible) successions of bits of matter into those which common sense does, and those which common sense does not, count as constituting a “unitary object,” you don’t seem to fi nd any kind of neat set of necessary and suffi cient conditions on which this division is based. Rather you seem to fi nd something more on the order of a “complicated network of family resemblances” vaguely marking off the “unitary objects.” Witt- genstein’s famous characterization of our concept of a game applies at least as well to our concept of an object, the latter concept being at least as vague and messy as the former. Now the point I’m getting at is this: No one could possibly look at the vague mess of our concept of an object and say: “That is the uniquely correct concept of an object.” If you think there is one correct concept of an object it’s surely not going to be our common sense concept, at least not if the latter is taken at face value. The only way to accept our common sense ontological judgments at face value—to accept them for the humble, vague, messy, but strictly and objectively true judgments that they are—is to give up the “absolutist” idea that there is somehow a uniquely correct referential apparatus that must show up in any language capable of stating the objective truth. We see that the journey “from the familiar to the familiar”—John Wisdom’s striking characterization of the nature of philosophy—must necessarily pass through conceptual relativism.9

9. John Wisdom, “Metaphysics and Verifi cation.” Mind 67:188 (October 1938) (cited in Putnam, “The Question of Realism,” op. cit., 300). four c C

the vagueness of identity

C The Evans-Salmon position on vague identity has deservedly c elicited a large response in the literature. I think it is in fact among the most provocative metaphysical ideas to appear in recent years. I will try to show in this paper, however, that the position is vul- nerable to a fundamental criticism that seems to have been virtually ignored in the many discussions of it. I take the Evans-Salmon position to consist of the following two theses:

Thesis I. There cannot be objects x and y such that it is indeterminate whether x is (identical with) y. Thesis II. The only way for an identity sentence to be indeterminate in truth-value is if one of the expressions fl anking the identity symbol is referentially ambiguous.1

The argument for Thesis I is essentially as follows. We are assuming that the sense of identity under discussion satisfi es the standard formal logic of identity including Leibniz’s Law. Suppose, now, that it is indeterminate

1. The brief formulation given by Gareth Evans fails to distinguish between Theses I and II; see his “Can There Be Vague Objects?” Analysis 38 (1978): 208. Nathan Salmon independently developed a clearer formulation in Reference and Essence (Oxford: Oxford University Press, 1982), 243–46, and “Modal Paradox: Parts and Counterparts, Points and Counterpoints,” Midwest Studies in Philosophy 11 (1986): 110ff . The distinction between the two theses is carefully explained in Timothy Williamson, Vagueness (London: Routledge, 1994), 253–54.

45 46 quantifier variance and realism: essays in metaontology whether x is y. Since it is determinate that x is x, x diff ers from y with respect to the property of being determinately x, from which it follows by Leibniz’s Law that x is not y. Since the supposition that it is indeterminate whether x is y leads to the conclusion that x is not y, this supposition is incoherent. Thesis II, it is argued, follows immediately from Thesis I. A singular term is referentially ambiguous in the intended sense if there is a selection of things such that it is indeterminate to which one of them the term refers. If we have the identity sentence “a is b” and neither “a” nor “b” is referentially ambiguous, then there are things x and y such that “a” determinately refers to x and “b” determinately refers to y, whence, by Thesis I, the truth-value of “a is b” must be determinate. In assessing this argument for Thesis II, we are to ignore standard examples of reference failure leading (on some views) to truth-value gaps. Let it indeed be stipulated that “a is b” counts as an identity sentence in this discussion only if the sentence “Something is uniquely a and something is uniquely b” is determinately true.2 The typical lineup in the literature is that critics of the Evans- Salmon argument deny that vagueness can be treated with the standard two-valued logic that seems to be presupposed in the argument; believers in a two-valued logic of vagueness typically accept the argument.3 My main aim is to show that the argument is incorrect even from the standpoint of two-valued logic. Since the most infl uential two-valued treatment of vagueness is supervaluationism, I am going to adopt that perspective in most of what follows. Specifi cally I am going to maintain that, on supervaluationist grounds, Thesis I is dubious—in fact, it is true in one sense but false in another—and, even if true, Thesis I does not entail Thesis II.4

2. We can imagine an example in which, because of the vagueness of “bald,” it is indeterminate whether there is uniquely someone who is the bald person in the room. The sentence “That guy is the bald person in the room” may therefore be indeterminate in truth-value without either “that guy” or “the bald person in the room” being referentially ambiguous in the sense explained. This will not qualify as a counterexample to Thesis II. 3. An atypical discussion, I think, is Robert Stalnaker. “Vague Identity.” in Philosophical Analysis: A Defense by Examples, ed. D. F. Austin (Kluwer Academic Publishers, Dordrecht, 1988), which seems to challenge the argument without abandoning standard logic (see n. 11 below). Although Salmon, Reference and Essence, 244, and Salmon, “Modal Paradox,” 110–11, seem to insist that the argument must be accepted even by believers in a multi-valued logic of vagueness, it seems clear, as Williamson observes, that “a variety of non-standard systems can be specifi ed in which the Evans-Salmon argument is formally incorrect” (Williamson, Vagueness, 255). 4. Richmond Thomason. “Identity and Vagueness,” Philosophical Studies 43 (1982): 329–32, uses supervaluationist reasoning to show that if “a” or “b” is referentially ambiguous, “a is b” can be the vagueness of identity 47

I want to begin by examining Thesis II. It is widely acknowledged that there are prima facie counterexamples to Thesis II, and a major source of general philosophical interest in the Evans-Salmon position lies in its ontological consequences with respect to these examples. The following example is representative. A ship constructed of a thousand wooden planks stands alone in a certain harbor on a Monday morning, and later that day a number of planks are all at once—not gradually—replaced by qualitatively similar planks. If the number of replaced planks is close to one then, by the lights of common sense, we still have the same ship, but if the number is close to a thousand we have a diff erent ship. There will evidently be a number somewhere between one and a thousand such that if that number of planks were replaced we would be undecided whether to say that we have the same ship or a diff erent ship. In this example the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday” seems to be indeterminate in truth-value. Therefore, Thesis II would imply that either the expression “the ship in the harbor on Monday” or the expression “the ship in the harbor on Tuesday” must be referentially ambiguous. The problem is that neither of these expressions seems to be referentially ambiguous. It does not seem that there is a selection of diff erent things in the harbor on Monday (Tuesday) competing for the role of referent of the expression “the ship in the harbor on Monday (Tuesday).”5

indeterminate in truth-value (a result that he interprets as confl icting with Evans, “Can There Be Vague Objects?”), but I am not aware of any attempt in the literature to criticize Theses I or II on supervaluationist grounds. A less familiar two-valued treatment of vagueness is the epistemic view defended in Williamson, Vagueness. Much of the discussion that follows could be adapted to that view. Whereas supervaluationists speak of there being diff erent (admissible) precisifi cations of a vague expression, the epistemic view speaks of there being diff erent epistemically possible interpretations of a vague expression, i.e., interpretations such that speakers cannot know which of them is the correct one (see Williamson, Vagueness, esp. 164, 237, 257–58). The supervaluationist reconstruction that I later give of the Evans-Salmon argument and my supervaluationist rebuttal of the argument remain virtually unaff ected if one substitutes “epistemically possible interpretation” for “precisifi cation.” See, further, n. 32. 5. In addition to the sort of example I give of vagueness of identity over time, problematical examples can also be constructed for vagueness of identity through possible worlds and vagueness of identity through space. On the former, see, e.g., Salmon, Reference and Essence, 240, and Timothy Williamson, Identity and Discrimination (Oxford: Oxford University Press, 1990). 128–35: on the latter, see, e.g., Sydney Shoemaker, “On What There Are,” Philosophical Topics 16 (1988): 201–23, at 208. 48 quantifier variance and realism: essays in metaontology

Defenders of Thesis II must somehow construe these expressions as being referentially ambiguous.6 This problem may be relatively easy to deal with for those philosophers who have, on independent grounds, already adopted a “four-dimensional” view of objects.7 On a common version of this view, any succession of ship-stages is an object, though only some of these objects are referred to by the word “ship.” (Note that I follow the practice of saying that a general term refers to the objects in its extension.) It may then be suggested that insofar as the word “ship” is somewhat vague, it is sometimes indeterminate whether a given four-dimensional object comprised of a succession of ship-stages qualifi es as a ship. Hence, in our example the vagueness of the identity statement might be explained by saying that it is indeterminate which one of the countless ship-like four-dimensional objects present in the harbor on Monday (Tuesday) we are referring to as “the ship in the harbor on Monday (Tuesday).” The example does, then, involve referentially ambiguous expressions, just as Thesis II requires. The example does seem to present a serious problem, however, for those philosophers who reject a four-dimensional view of objects. Without entering into the general controversy surrounding the four- dimensional view, the question that immediately concerns me is whether the existence of certain vague identity sentences—such as the one in our example—is suffi cient to show that the view is correct. It is instructive to look at an answer to this question suggested by Sydney Shoemaker in his paper “On What There Are.” Shoemaker distinguishes between what he calls “permissive” and “restrictive” views. The former is essentially the sort of four-dimensional view I mentioned earlier.8 The latter view, which seems to Shoemaker, and also to me, much closer to common sense, denies that every way of cutting up space-time yields an object. This view restricts objects roughly to the referents of such ordinary sortals as “ship.” Shoemaker wants to defend restrictivism, but he concedes—I suppose, reluctantly— that, for this view to be defensible, it must at least be suffi ciently

6. I ignore the implausible suggestion sometimes made that such sentences as “The ship in the harbor on Monday is the ship in the harbor on Tuesday” are not really statements of identity. I am assuming throughout this discussion that such sentences are indeed vague and that Thesis II must not force us to deny this (which is, apparently, Salmon’s response to the problem in “Modal Paradox,” 113). 7. Cf. Stalnaker, “Vague Identity,” 351–52. 8. It is, more accurately, the four-dimensional view generalized so as to treat possible worlds as in eff ect additional dimensions that can be cut up into objects in every which way. the vagueness of identity 49 permissive to deal with vague identity sentences in accordance with Thesis II.9 So we can deny the permissivist’s claim that every succession of ship-stages is an object, but we have to at least countenance enough ship-like objects in the harbor to explain the referential ambiguity in the earlier example of the expression “the ship in the harbor on Monday (Tuesday).” Russell said somewhere that even if it’s hard to pinpoint where the ontological proof goes wrong, the proof is clearly trying to pull a rabbit out of a hat. I think we should respond in somewhat the same way to Shoemaker’s proof of the existence of the ship-like objects in the harbor. I assume—as I suspect Shoemaker does—that vagueness is a matter of semantic indecision; expressions or sentences are vague insofar as we (the linguistic community) haven’t pinned down precisely how we want to use them. Shoemaker’s argument implies that our indecision about precisely what truth conditions we want to attach to the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday” requires us, contrary to what seems to be the natural intuition of common sense, to believe in a lot of ship-like objects occupying the same place as a ship. In other words, our being undecided on precisely how to talk about the identity of ships forces us, against our will, it seems, to decide to talk about the ship-like non-ships. That, it seems to me, cannot be right. But where, then, does the argument go wrong? Let me concentrate fi rst on the derivation of Thesis II from Thesis I. If we examine that derivation it is clear that an assumption being made is this:

(P) For any expression “a,” if the sentence “Something is uniquely a” is determinately true, then either (i) something is such that it is determinately referred to by “a,” or (ii) the expression “a” is referentially ambiguous.

(P) implies that in the example under discussion the only way for the expression “the ship in the harbor on Monday” to be referentially indeterminate is for it to be referentially ambiguous. That is, unless there are at least two distinct objects competing for the role of referent of the expression, there must be something such that the expression

9. Shoemaker, “On What There Are,” 217. One of Shoemaker’s examples is the ship of Theseus case, assuming that this yields a vague identity sentence (208). A related argument against the restrictive view, stemming from the nonvagueness of existence, is given in David Lewis, On the Plurality of Worlds (New York: Basil Blackwell, 1986), 212–13. 50 quantifier variance and realism: essays in metaontology determinately refers to it.10 Should we believe this? If we are inclined to be uncompromising common-sense restrictivists, and are therefore disinclined to believe that there are a number of ship-like objects competing for the role of referent of “the ship in the harbor on Monday,” then we might conclude that what the example shows is that (P) is false. An expression that is not referentially ambiguous may nevertheless be referentially indeterminate by virtue of the fact that the identity conditions associated with it are indeterminate. This does not seem obviously unintelligible. Since the identity conditions associated with the expression “the ship in the harbor on Monday” are indeterminate it may be in a sense indeterminate to which thing the expression refers, even though there do not exist diff erent things to which the expression might refer. If this is correct—that is, if (P) is false— then Thesis II does not follow from Thesis I.11 It is, I think, diffi cult to confi dently assess (P)—and to assess the derivation of Thesis II from Thesis I—without saying something more about the notion of “determinateness.” One of the basic ideas of the supervaluationist approach is that “vagueness” and “determinateness” can be clarifi ed in terms of admissible precisifi cations.An expression or sentence is vague if it can be made precise in more than one reasonable way.12 It will prove to be highly instructive to restate the derivation of Thesis II from Thesis I in terms of precisifi cations, as follows. Thesis I says in eff ect that “is (identical with)” cannot be vague. Rendered in terms of precisifi cations we have:

1. (Thesis I). There cannot be objects x and y, and two precisifi cations of the expression “is (identical with),” such that on one precisifi cation

10. Cf. Shoemaker, “On What There Are,” 208: “If there is to be this sort of indeterminacy of reference [i.e., the sort that would allow an identity sentence to be indeterminate in truth-value], then there must be in the world all of the entities that are the potential referents [of the expressions fl anking the identity symbol].” 11. The rejection of (P) is, I think, implied in Stalnaker, “Vague Identity.” Stalnaker seems to imply that “we can distinguish subtly diff erent potential referents of indeterminate terms” (354) in cases where there do not exist in the world diff erent things that are the potential referents. 12. As David Lewis puts it, “If vagueness is semantic indeterminacy, then wherever we have vague statements, we have alternative precisifi cations of the vague language involved, all with equal claim to being ‘intended’ ” (“Vague Identity: Evans Misunderstood,” Analysis 48 [1988]: 128–30, 128–29). Kit Fine compares vagueness to an “unfi nished picture,” with each precisifi cation being a way of completing the picture (“Vagueness, Truth, and Logic,” Synthese 30 [1975]: 265–330, at 283). (In fact. Fine uses the word “specifi cation” rather than “precisifi cation.” Other words have also occurred in the literature, but “precisifi cation” seems to be the most widely used.) the vagueness of identity 51

the expression refers to (the pair consisting of) x and y and on the other precisifi cation the expression does not refer to x and y.

Our standing assumption is that we are dealing with an identity sentence “a is b” such that the sentence “Something is uniquely a and something is uniquely b” is determinately true. This assumption can be rendered in terms of precisifi cations as:

2. Assume an example of “a is b” such that the sentence “Some- thing is uniquely a and something is uniquely b” is true on every precisifi cation.

From 1 and 2 we are supposed to derive Thesis II. Thesis II says that if the sentence “a is b” is indeterminate in truth-value then “a” (or “b”) is referentially ambiguous. In terms of precisifi cations:

3. (Thesis II). If the sentence “a is b” is true on one precisifi cation and false on another, then there are distinct objects x and y, and two precisifi cations of the expression “a” (or the expression “b”), such that on one precisifi cation the expression refers to x and not to y and on the other precisifi cation the expression refers to y and not to x.13

I want to eventually come back to consider whether 1 is true, but the immediate question is whether 3 follows from 1 and 2. It may seem that an affi rmative answer to this question is clearly implied by the fact that with respect to any expression there are only three options (not necessarily exclusive):

Option A. There is something x such that on any precisifi cation the expression refers to x. Option B. There are distinct objects x and y, and two precisifi cations of the expression, such that on one precisifi cation the expression refers to x and not to y and on the other precisifi cation the expression refers to y and not to x. Option C. On some precisifi cation the expression does not refer to anything.14

13. Strictly speaking, Thesis II is the proposition that if the assumption in 2 holds then 3 holds. 14. Assuming that we can safely ignore certain special kinds of infi nitary structures, that A, B, and C are the only three options is virtually a truth of logic. Compare: “Either there is someone x whom Jones loves at every time, or there are distinct people x and y such that Jones loves x and not y at some time and Jones loves y and not x at some time, or there is a time at which Jones does not love anyone.” 52 quantifier variance and realism: essays in metaontology

With respect to the sentence “a is b” in the inference of 3 from 1 and 2, it may seem obvious that 2 rules out Option C as applied to either of the expressions “a” and “b,” leaving us only with Options A and B. So we have the result that if Option B does not apply to these expressions then Option A must apply. But it seems clear that if Option A applies, then it follows from 1 (i.e., Thesis I) that the sentence “a is b” has the same truth-value on every precisifi cation. From the last two sentences it follows that if Option B does not apply then the sentence “a is b” has the same truth-value on every precisifi cation. Contraposing the last sentence, we get 3 (i.e., Thesis II). So it may seem clear that Thesis II does indeed follow from Thesis I. There is, however, a deep mistake here, or so I will attempt to show. The mistake is in the assumption that 2 rules out Option C. That assumption implies with respect to our earlier example that, since the sentence “Something is uniquely the ship in the harbor on Monday” is true on every precisifi cation, the following sentence must be false: (Q) On some precisifi cation of the expression “the ship in the harbor on Monday,” the expression does not refer to anything.l5 I will try to show, however, that, contrary to what seems initially obvious, (Q) is true. It is the truth of (Q) (and other sentences like it) that blocks the derivation of Thesis II and that is the source of a kind of vagueness of identity that is ignored in the Evans-Salmon position. Since my defense of (Q) needs to be a bit complicated, perhaps I should try to foreshadow what the main underlying idea is in this defense. To describe the world vaguely is in a sense to straddle diff erent choices for describing the world precisely (the choices are the precisifi cations). On one choice we assert, “The ship in the harbor on Monday is the ship in the harbor on Tuesday”; on another choice we assert the negation of that sentence. A crucial point is that each of these choices aff ects how the choices themselves are to be described. It will emerge that on each choice what we must say is: “On the other choice, there is nothing in the world that is referred to by the expression ‘the ship in the harbor on Monday (Tuesday).’ ” The Evans-Salmon argument evidently presupposes the notion of “there being something that an expression refers to,” but that notion is itself critically vague. The reason why (Q) is true—and the reason why the Evans-Salmon argument fails—is that diff erent ways of making

15. I always take for granted a relevant context of utterance. In (Q) we are imagining a context of utterance appropriate to the described example. the vagueness of identity 53 precise what we mean by such expressions as “the ship in the harbor on Monday” correspond to diff erent ways of making precise what we mean by the expression “there being something that an expression refers to.” But let me now develop this idea in some detail. In the discussion that follows I am assuming a common-sense restrictive view, by which I mean a restrictive view that, unlike Shoe- maker’s, does not compromise common-sense intuitions about what exists in order to satisfy Thesis II. Hence, I am not strictly trying to prove that (Q) is true; rather, I am trying to rebut the derivation of Thesis II by showing that a perfectly coherent alternative to Thesis II is (Q) in conjunction with common-sense restrictivism. Since I am trying to argue that (Q) is compatible with the fact that the sentence “Something is uniquely the ship in the harbor on Monday” is true on every precisifi cation of the sentence, I will certainly not acquiesce to paraphrasing (Q) as “It is indeterminate whether something is the ship in the harbor on Monday.” The argument that I am embarking on requires that one carefully keep track of which bit of language is being subjected to precisifi cations. I am maintaining that a sentence of the form “Something is uniquely a” may be true on every precisifi cation, while the sentence “On some precisifi cation of the expression ‘a,’ the expression does not refer to anything” is also true on every precisifi cation. Let me here stress an essential point of terminology. I always use “precisifi cation” as a meta-level operator, never as an object-level operator. Hence, I say, “On some precisifi cation of the sentence ‘The ship in the harbor on Monday is the ship in the harbor on Tuesday,’ the sentence is true”; I do not say, “On some precisifi cation (it is true that) the ship in the harbor on Monday is the ship in the harbor on Tuesday.” I am aware that some authors use “precisifi cation” as an object-level operator, but the meta-level terminology that I am stipulating is both coherent and natural, and the reader must bear it in mind if there is any hope of following my argument. A related point is that when precisifying a bit of language we do not precisify any part of it that is in quotes, i.e., that is merely being mentioned. (For example, in precisifying the sentence “The expression ‘free will’ is almost never used by ordinary people,” we do not precisify “free will.”) It may be helpful to compare (Q) to the statement “On some precisifi cation of the expression ‘the ship in the harbor on Monday,’ the expression does not refer to the ship in the harbor on Monday.” Although it may initially seem that this statement is obviously false, Shoemaker’s position implies that (on supervaluationist grounds) it is true. This is because, according to Shoemaker’s position, the expression 54 quantifier variance and realism: essays in metaontology

“the ship in the harbor on Monday” refers to a certain ship-like object on one precisifi cation i and refers to another ship-like object on another precisifi cation j. Hence, the statement “On precisifi cation j of the expression ‘the ship in the harbor on Monday,’ the expression does not refer to the ship in the harbor on Monday” is true when the latter occurrence of “the ship in the harbor on Monday” is subjected to precisifi cation i; and, generalizing from this, we fi nd that the statement “On some precisifi cation of the expression ‘the ship in the harbor on Monday,’ the expression does not refer to the ship in the harbor on Monday” is true (on every precisifi cation of the statement). Now, on my defense of (Q), it will turn out that not only is it surprisingly true that, on some precisifi cation of the expression “the ship in the harbor on Monday,” the expression does not refer to the ship in the harbor on Monday, but, even more surprisingly, on some precisifi cation of that expression, the expression refers to nothing.16 We can get a better sense of what (Q) is saying if we focus on two specifi c precisifi cations. Let us consider a range of precisifi cations of

16. A note on disquotation: The statement On some precisifi cation of the expression “the ship in the harbor on Monday,” the expression does not refer to the ship in the harbor on Monday does not entail the statement On some precisifi cation of the sentence “The expression ‘the ship in the harbor on Monday’ does not refer to the ship in the harbor on Monday,” the sentence is true. Supervaluationists may reject the latter statement by holding that “refers” is vague and is systematically connected to other vague expressions in such a manner that any admissible precisifi cation must sustain disquotation for reference (cf. Williamson, Vagueness, 304n. 12). Disquotation, however, is certainly not violated by the former statement, which uses the vague expression “the ship in the harbor on Monday” to say something about a precisifi ed version of that expression. Because of the diff erence between these two statements, there is a crucial ambiguity in the sentence “It is not determinate that ‘the ship in the harbor on Monday’ refers to the ship in the harbor on Monday”; associating the sentence with the fi rst statement yields a “de re” reading on which it is true, but associating it with the second statement yields a “de dicto” reading on which it is arguably false. By the same reasoning, (Q) does not entail the statement

On some precisifi cation of the sentence “The expression ‘the ship in the harbor on Monday’ does not refer to anything,” the sentence is true. Hence, I cannot accept as a paraphrase of (Q), “It is not determinate that the expression ‘the ship in the harbor on Monday’ refers to something,” since that can easily be taken to imply that the sentence “The expression ‘the ship in the harbor on Monday’ does not refer to anything” is true on some precisifi cation of the sentence. (I will make some general comments about the relationship between precisifi cations and determinateness at the end of this paper.) In short, my argument in this paper leaves it open whether disquotation for reference holds with respect to vague expressions. the vagueness of identity 55 sentences of the form “The ship at place one at time one is the ship at place two at time two,” with each precisifi cation requiring that a ship not precipitously change by more than n percent of its matter, for some specifi c number n. Let precisifi cations 1 and 2 specify, respectively, 30 percent and 40 percent, and imagine that the ship in the harbor had 35 percent of its matter replaced. Then the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday” is false on precisifi cation 1 and true on precisifi cation 2. Let us recall an important point about precisifi cations that has been stressed in supervaluationist literature.17 The precisifi cation of an expression or sentence must not be treated in isolation, but must be coherently extended to related expressions and sentences. Since a precisifi cation must preserve all clear-cut truths, we cannot coherently conjoin a precisifi cation of “wealthy” with a precisifi cation of “rich” that makes the clear-cut true sentence “All wealthy people are rich” come out false. It is often helpful, indeed, to think of a precisifi cation as extending over the whole language. Precisifi cation 1 represents one coherent way of making the English language precise, where this includes the precise truth-conditions specifi ed for sentences of ship-identity. The truth- conditions of these sentences on precisifi cation 1 must aff ect many other expressions and sentences, including obviously the word “ship” and expressions like “the ship in the harbor on Monday,” but including also—and this is very important—sentences that do not contain the word “ship.” In our standing example, we can imagine circumstances on the harbor such that the following is a clear-cut truth: “If the ship in the harbor on Monday is not the ship in the harbor on Tuesday, then some physical body in the harbor on Monday went out of existence before Tuesday.” Since the sentence “The ship in the harbor on Monday is not the ship in the harbor on Tuesday” is true on precisifi cation 1, the sentence “Some physical body in the harbor on Monday went out of existence before Tuesday” must also be true on precisifi cation 1. In what follows I will sometimes talk explicitly about precisifi cations of the whole language, but often I will speak of the precisifi cation of some particular expression or sentence in order to emphasize that it is that expression or sentence that is being evaluated on the (whole- language) precisifi cation.18

17. See the discussion of “penumbral connections” in Fine, “Vagueness, Truth, and Logic,” 270, and Williamson, Vagueness, 145–47. 18. By talking of the “whole language,” I skirt some technical questions about semantic levels, but I doubt that they seriously aff ect what I say. 56 quantifier variance and realism: essays in metaontology

Now consider the sentence: (Q∗) On precisifi cation 2 of the expression “the ship in the harbor on Monday,” the expression does not refer to anything. I want to show that (Q∗) is true on precisifi cation 1. (The sentence (Q∗) says something about precisifi cation 2, and I am evaluating that sentence on precisifi cation 1.) It seems clear that if (Q∗) is true on precisifi cation 1, then (Q) must be true on every precisifi cation. This follows from the fact that precisifi cations 1 and 2 are arbitrarily chosen precisifi cations yielding diff erent truth-values for the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday.” So the task of defending (Q) is considerably simplifi ed by considering (Q∗).19 It may be helpful to picture precisifi cations 1 and 2 as two precise languages. My claim is that the sentence (Q∗), as understood in the fi rst language, says something true about the second language. The diagram that follows may give one an initial sense of what I mean by claiming that (Q∗) is true on precisifi cation 1. I am claiming that Ernie’s fi nal comment is true (and so, of course, is Bert’s corresponding fi nal comment). The reader may be helped to understand the argument that follows by construing it as Ernie’s defense of his fi nal comment. If (Q∗) is not true on precisifi cation 1 then the most obvious idea would be that the following sentence is true on precisifi - cation 1: (a) On precisifi cation 2 of the expression “the ship in the harbor on Monday,” the expression refers to the ship in the harbor on Monday. But I think one can easily see that it is completely implausible to hold that (a) is true on precisifi cation 1. For if (a) is true on precisifi - cation 1 then it would seem absurdly arbitrary to deny that so is the sentence (b) On precisifi cation 2 of the expression “the ship in the harbor on Tuesday,” the expression refers to the ship in the harbor on Tuesday.

19. Throughout this discussion I adopt the typical supervaluationist assumption that an existential sentence such as (Q) is true if it is true on every precisifi cation. In fact, I only require the much weaker assumption that if (Q) is true on every precisifi cation then (Q) is not determinately false. If I can show that (Q) is not determinately false, this suffi ces to block the derivation of Thesis II. (Indeed, it is likely that the only supervaluationist assumption strictly required for the ensuing argument is the basic idea that every case of vagueness corresponds to a range of admissible precisifi cations.) the vagueness of identity 57

It‛s sort of It‛s sort of the same ship the same ship and sort of not and sort of not the same ship. the same ship.

I choose the I choose the description: description: “It‛s not the “It is the same ship.” same ship.”

The way Bert uses The way Ernie uses the expression “the the expression “the ship” it does not ship” it does not refer to the ship, or refer to the ship, or to anything else. to anything else.

ERNIE BERT Precisification 1 Precisification 2 58 quantifier variance and realism: essays in metaontology

But it seems that if (a) and (b) are true on precisifi cation 1, then the following sentence must also be true on precisifi cation 1:

(c) On precisifi cation 2 of the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday,” the sentence is true if and only if the ship in the harbor on Monday is the ship in the harbor on Tuesday. But (c) is obviously not true on precisifi cation 1, for that would require, contrary to our assumption, that the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday” has the same truth-value on precisifi cations 1 and 2. It might be suggested that the truth on precisifi cation 1 of (a) and (b) does not imply the truth on precisifi cation 1 of (c) unless one assumes the truth on precisifi cation 1 of the sentence (d) On precisifi cation 2 of the expression “is (identical with),” the expression (as it functions in the sentence under consideration) refers to any pair of things that are identical with each other. But if (d) is not true on precisifi cation 1, what could be the truth on precisifi cation 1 about the precisifi cation 2 of “is”? I think it is evidently hopeless to look for an answer to this question that will sustain the judgment that (a) is true on precisifi cation 1.20 If (a) is not true on precisifi cation 1, then I think it is suffi ciently clear (given the common-sense restrictive view I am assuming) that there is no truth on precisifi cation 1 of the form “On precisifi cation 2 of the expression ‘the ship in the harbor on Monday,’ the expression refers to——.” There is no sensible way to fi ll in this blank which will yield a sentence true on precisifi cation 1.21 Hence, (Q∗) is true on precisifi cation 1. Of course we know from the problem of inscrutability that there are countless “reference schemes” for any language that can

20. In claiming that, on precisifi cation 1, the conjunction of (a), (b), and the denial of (d) is untenable as an explanation of the truth on precisifi cation 2 of the sentence “The ship in the harbor on Monday is the ship in the harbor on Tuesday,” I am in eff ect appealing to the following consequence (which must be true on precisifi cation 1) of Thesis I: If x is not y then there is no precisifi cation of the expression “is (identical with)” on which the expression refers to x and y. I am entitled to assume Thesis I if what I am trying to show is that Thesis II does not follow from it. In fact, I will later defend a version of Thesis I that has the mentioned consequence. 21. Consider, for instance, the suggestion that the truth on precisifi cation I is “On precisifi cation 2 of the expression ‘the ship in the harbor on Monday,’ the expression refers to the succession consisting of the original ship and the ship that replaces it.” This suggestion would have to be reconciled with the fact that, at least on the common-sense view being assumed, a clear-cut truth, hence, a truth the vagueness of identity 59 apparently generate all the correct truth-conditions for the sentences of the language. I don’t take that general problem to have a special bearing on the issue I am discussing. (Q∗) is true on precisifi cation 1 in the sense that there is no remotely plausible or nonarbitrary way of formulating a truth on precisifi cation 1 of the form “On precisifi cation 2 of the expression ‘the ship in the harbor on Monday,’ the expression refers to——.”22 If we want to make a speech about how to evaluate English on precisifi cation 2, where this speech is designed to be true on precisifi cation 1, it would have to be along the following lines: “On precisifi cation 2 of the English language, sentences of the form ‘The ship at place one is the ship at place 2’ have certain defi nite truth-conditions, but the parts of the sentences fl anking the word ‘is’ function ‘syncategorematically,’ or rather like Russell’s ‘incomplete symbols’; that is, although these expressions do contribute systematically to the truth-conditions of the sentences—certainly the expressions are not in any sense vacuous— they do not refer to anything. Since they do not refer to anything, neither indeed does the word ‘is’ as it functions in these sentences.” A speech very much to the same eff ect about precisifi cation 1 would be true on precisifi cation 2. Is all of this too mysterious? Much of the mystery is dispelled, I think, when one realizes that what is really going on here is that the precisifi cations diff er in their referential apparatus. That is, each precisifi cation assigns a slightly diff erent semantic content to the family of expressions including “something,” “everything,” “thing,” “(there) exists,” “(such that) it,” “is (identical with),” “set,” and “refers.” The referential apparatus of English is vague. That is the fundamental point underlying the truth of (Q) and the falsehood of Thesis II. Consider the sentence “Something constructed of a thousand planks went out of existence,” or, equivalently, “Something constructed of a thousand planks that existed at one time is not identical with anything that existed at a later time.” With respect to the imagined situation on

on every precisifi cation including precisifi cation 2 is “No ship is identical with a succession of diff erent physical bodies.” So we would need to search for some exotic truth on precisifi cation 1 of the form “On precisifi cation 2 of the expression ‘succession (of diff erent physical bodies),’ the expression refers to———.” There is no sensible way to make this work. 22. At most the problem of inscrutability might induce us to regard (Q∗) as indeterminate in truth-value on precisifi cation 1. That would suffi ce to block the derivation of Thesis II. 60 quantifier variance and realism: essays in metaontology the harbor (or some version of it), these sentences are true on precisifi cation 1 and false on precisifi cation 2. It seems obviously implausible (at least from the standpoint of common-sense restrictivism) to suggest that the vagueness of these sentences derives from some vagueness in the expression “constructed of a thousand planks.” Presumably, then, the sentences contain other words that are vague. It would seem sense- less to suggest that what this shows is that “something” is vague, but “existence” and “identical” are precise, or perhaps the opposite, that “existence” and “identical” are vague but “something” remains precise.23 Since these words seem intuitively to be closely related, and to contribute in combination equally to the indeterminacy of truth-values of sentences in which they appear, the only reasonable judgment, it seems, is that they are all vague. By the same reasoning, the vagueness of “the diff erence between a thing (other than a set) and a set” is revealed by the indeterminacy of truth-value of the sentence “Over the period from Monday to Tuesday there was a set (or succession) of two diff erent things constructed of a thousand planks, rather than a single thing.” And the vagueness of “refers” is revealed by the indeterminacy of truth- value of the sentence “On precisifi cation 2 of the expression ‘the ship in the harbor on Monday,’ the expression does not refer to anything.” Expressions like “thing,” “exists,” “identical,” “refers”—what I called our referential apparatus—constitute in some sense the linchpin of the connection between language and the world. Vagueness—that is, semantic indecision—occurs at that level. This is not surprising. On the contrary, it would be surprising if this level of language enjoyed a kind of “crystalline purity” absent from the rest of language.24 One has to puzzle over why Evans, Salmon, and many other philosophers seem to ignore or deny the vagueness of our referential apparatus.25 It’s as if these philosophers would like to believe that the semantic content of words like “thing” and “exist” arrives from on high in some

23. Note that throughout this discussion I draw no distinction between “existence” and “being,” and I take no account of “restricted quantifi ers.” I mean “existence” in the broadest and most unrestricted sense available in ordinary language. 24. See Ludwig Wittgenstein, Philosophical Investigations (Oxford: Oxford University Press, 1953), §108. 25. Consider the following from David Lewis: “Vagueness is semantic indecision. But not all of language is vague. . . . The words for identity and diff erence [are not]. . . . Nor are the idioms of quantifi cation. . . . How could any of these be vague? What would be the alternatives between which we haven’t chosen?” (Lewis, On the Plurality of Worlds, 212). One alternative between which we haven’t chosen is precisifi cation 1 or 2. Where is the force of Lewis’s question? (Lewis’s question, it should be noted, is an essential part of his attack against the common-sense restrictive view.) the vagueness of identity 61 manner that precludes semantic indecision.26 A grotesque argument for this idea would be the following: “If the semantic content of ‘thing’ and ‘exist’ is a matter of human decision then it follows, absurdly, that it is a matter of human decision what things exist.” This argument deserves no comment. Nevertheless, I do think that there is a serious connection between the issues I have been discussing and the perennial question whether in some sense “things do not precede language.” The statement that language creates objects, while obviously preposterous in any literal sense, may be a good example, I think, of what John Wisdom described as the use of paradox in philosophy to convey metaphorically an insight hard to express in purely prosaic terms.27 I have been describing a complicated confi guration involving two precisifi cations of the expression “the ship in the harbor on Monday” such that when we adopt the standpoint of either precisifi cation, in the sense of trying to assert sentences that are true on that precisifi cation, we wind up saying, “On the other precisifi cation, the expression does not refer to anything.” It is as if each precisifi cation correctly claims that by its own standards the expression refers to something. So it is as if each precisifi cation has its own domain of objects to refer to, objects that do not exist relative to the other precisifi cation. I think there need be nothing wrong with saying this. It is in some sense “as if ” each precisifi cation has its own domain of objects. This seems to be a very natural image or fi gure by which to grasp certain complicated philosophical facts. As long as it remains clear that we are dealing with metaphor rather than cold sober description, there is no harm. Certainly it would be misguided to deny the vagueness of our referential apparatus on the grounds that such vagueness implies that in some literal sense the existence of things depends upon language. Nothing of the sort is implied.28 Perhaps some philosophers are prone to ignore the vagueness of the referential apparatus simply because they are misled by an overly narrow paradigm of vagueness, centered on such adjectives as “bald” and “red,” with respect to which the things to be referred to are, so to

26. And that also precludes, perhaps, semantic variation from language to language. I cannot enter into the issue of the “conventionalism” of identity, which runs in some ways parallel to the issue of the vagueness of identity. Shoemaker claims that just as the vagueness of identity statements requires a permissive relaxation of common-sense restrictivism, so does the element of convention in such statements (see Shoemaker, “On What There Are,” 217). I would want to deny both claims. 27. John Wisdom, Philosophy and Psycho-Analysis (Berkeley: University of California Press, 1953), 36–50. 28. The sentence “The existence of things depends upon language,” taken literally, is absurdly false on every precisifi cation of our referential apparatus, hence, absurdly false, period. 62 quantifier variance and realism: essays in metaontology speak, already established, and the only semantic indecision has to do with which things go with which words. The vagueness of the referential apparatus, by contrast, involves semantic indecision as to what we are to mean by “there being a thing to refer to.” The discussion of (Q∗) reveals that this more subtle and complicated kind of vagueness also applies to such “sortal” or “individuative” words as “ship,” words that are related in an especially intimate way to the identifi cation and reidentifi cation of things.29 This completes my argument that (Q∗) is true on precisifi cation 1; hence, (Q) is true (on every precisifi cation).30 The truth of (Q) (and other sentences like it) blocks the derivation of Thesis II in the earlier argument from 1 and 2 to 3. Prior to reconstructing the derivation in terms of precisifi cations, it was clear that the derivation presupposes the proposition (P), which implies that the expression “the ship in the harbor on Monday” can be referentially indeterminate only if it is referentially ambiguous. The burden of proof was on a defender of the derivation to explain why (P) should be believed. That burden might have been discharged through an analysis in terms of precisifi cations, but in fact the analysis turns out to have the opposite eff ect: The truth of (Q) implies the falsity of (P). (Q) shows how “the ship in the harbor on Monday” can be referentially indeterminate without being referentially ambiguous. The kind of referential indeterminacy implied by (Q) has a complex structure that eludes the Evans-Salmon argument. Since the sentence “Something is the ship in the harbor on Monday” is determinately true, we are not dealing with an example of standard reference failure, but since on some precisifi cation of the expression “the ship in the harbor on Monday” the expression does not refer to anything, there is no object to which the expression can be said determinately to refer.31

29. Might this more subtle kind of vagueness turn out to confl ict with some constraints implicit in a standard referential or model-theoretic semantics? I doubt that there is any relevant problem here, but perhaps this question needs to be examined further. 30. I emphasize that there is no way of intuitively broaching the sentence (Q)—of getting any intuitive feeling of what the sentence is saying—except from the standpoint of some particular precisifi cation (e.g., precisifi cation 1), from which standpoint one can then generalize. The same remark holds for other sentences about precisifi cations that I later formulate. 31. As stated earlier (see n. 16) I am defending the de re claim “For anything x, it is not determinate that the expression ‘the ship in the harbor on Monday’ refers to x.” I am not defending the de dicto claim “It is not determinate that, for something x, the expression ‘the ship in the harbor on Monday’ refers to x.” (For further clarifi cation of the relevant de re sense of determinate reference, see n. 35.) the vagueness of identity 63

I conclude that the derivation of Thesis II has been refuted. The logical coherence of common-sense restrictivism is therefore not threatened by the vagueness of identity. My comments on Thesis I can be more brief. A reconstruction of the argument for Thesis I in terms of precisifi cations might go like this. The basic premise is that there cannot be something x such that on some precisifi cation of the expression “is (identical with)” the expression does not refer to the pair consisting of x and x. Now suppose that something x and something y are such that on some precisifi cation i the expression “is (identical with)” refers to x and y, and on some other precisifi cation j the expression does not refer to x and y. It follows from the basic premise that x diff ers from y with respect to the property of being such that the pair consisting of it and x is referred to by “is (identical with)” on precisifi cation j; hence, by Leibniz’s Law, x is not y. Since the inference to “x is not y” is valid on every precisifi cation, including precisifi cation i, the stated supposition is incoherent. (In other words, if we adopt the standpoint 0f any particular precisifi cation and try to suppose that, although something x is identical with something y, on some other precisifi cation “is (identical with)” does not refer to x and y, we contravene Leibniz’s Law.)32 This argument may be essentially on the right track, but, as it stands, it contains an important mistake in the basic premise. The basic premise is false because the following sentence is true: (R) The ship in the harbor on Monday is something x such that on some precisifi cation of the English language no expression refers to x. If x is such that on some precisifi cation no expression refers to x, then obviously on some precisifi cation the expression “is (identical with)” does not refer to x (or to the pair consisting of x and x). To show that (R) is true (on every precisifi cation), it will suffi ce to show that the following is true on precisifi cation 1: (R∗) The ship in the harbor on Monday is something x such that on precisifi cation 2 of the English language no expression refers to x. Since (Q∗) is true on precisifi cation 1 it immediately follows that a truth on precisifi cation 1 is “The ship in the harbor on Monday is

32. The argument in this paragraph seems to lose none of its force if, in accordance with the epistemic view of vagueness, one replaces “precisifi cation” with “epistemically possible interpretation” (cf. n. 4). The same remark applies to the earlier attempt to derive Thesis II in the argument from 1 and 2 to 3. Williamson does not seem to address this kind of epistemic version of the Evans-Salmon position when he discusses related matters in Vagueness, 256–66. 64 quantifier variance and realism: essays in metaontology something x such that on precisifi cation 2 the expression ‘the ship in the harbor on Monday’ does not refer to x.” This does not yet establish— although it certainly suggests the possibility—that (R∗) is true on precisifi cation 1. I will assume that if (R∗) is not true on precisifi cation 1 then the following sentence must be true on precisifi cation 1: (e) The ship in the harbor on Monday is something x such that on precisifi cation 2 the expression “physical body” refers to x. Now a truth on precisifi cation 1 is “The ship in the harbor on Monday is something x such that x exists on Monday and x does not exist on Tuesday.” Hence, if (e) is held to be true on precisifi cation 1, it would seem to be completely arbitrary to deny that both of the following sentences are also true on precisifi cation 1:

(f ) The ship in the harbor on Monday is something x such that on precisifi cation 2 the expression “exists on Monday” refers to x. (g) The ship in the harbor on Monday is something x such that on precisifi cation 2 the expression “does not exist on Tuesday” refers to x. But (e), (f ), and (g) cannot all be true on precisifi cation 1, since this would imply that a truth on precisifi cation 1 is:

(h) On precisifi cation 2 the sentence “Some physical body exists on Monday and does not exist on Tuesday” is true with respect to the situation on the harbor. But we are imagining an example (or one version of it) in which it is given as part of our characterization of precisifi cation 2 that (h) is false. I don’t doubt that more complicated moves can be made here, but I think they could only serve to postpone the eventual acceptance of (R∗) as true on precisifi cation 1.33 As I have already mentioned I don’t

33. One suggestion might be that what is true on precisifi cation 1 is not (e) but: (i) The ship in the harbor on Monday is something x such that on precisifi cation 2 the expression “stage” refers to x. From the standpoint of precisifi cation 1, in other words, the truth is, “A physical body is sometimes called a ‘stage’ on precisifi cation 2.” But this suggestion is really quite hopeless. A clear-cut truth (at least for common-sense restrictivism), hence, a truth on every precisifi cation including precisifi cation 2 is “Any stage must be the stage (in the history) of something.” Hence, if (i) is true on precisifi cation 1, there ought to be a truth on precisifi cation 1 of the form “The ship in the harbor on Monday is something x such that on precisifi cation 2 the expression ‘stage (in the history) of———’ refers tox .” How can this blank be fi lled in? Not with the expression “the ship in the harbor on Monday,” if, in accordance with the earlier argument, a truth on precisifi cation 1 is “On precisifi cation 2 the expression ‘the ship in the harbor on Monday’ does not refer to anything.” the vagueness of identity 65 think we have to be concerned about concoctions of arbitrary reference schemes. The sense in which (R∗) is true on precisifi cation 1 is that we cannot formulate a story remotely plausible on precisifi cation 1 of the form “On precisifi cation 2 such-and-such expressions refer to such-and-such things,” where a consequence of this story would be a truth on precisifi cation 1 of the form “The ship in the harbor on Mon- day is something x such that on precisifi cation 2 the expression ‘——’ refers to x.” Since (R∗) is true on precisifi cation 1, (R) is true (on every precisifi cation). (R) is really a corollary of points we have already gone through in connection with (Q). The essential underlying fact in all of this is the vagueness of our referential apparatus. Metaphorically speaking, the idea is that the ship that exists in the harbor relative to any given precisifi cation does not exist relative to some other precisifi cations. That, more soberly put, is what (R) says. The false premise in the argument constructed earlier for Thesis I is that there cannot be something x such that on some precisifi cation of the expression “is (identical with)” the expression does not refer to the pair consisting of x and x. The following, however, remains true: There cannot be some thing x such that on some precisifi cation of the expression “is not (identical with)” the expression refers to the pair consisting of x and x. From that premise we can derive a version of Thesis I that is correct:

Thesis Ia. There cannot be something x and something y such that on one precisifi cation of the expression “is (identical with)” the expression refers to x and y, and on another precisifi cation the negation of the expression refers to x and y.

The incorrect version of Thesis I that comes out of the false premise is:

Thesis Ib. There cannot be something x and something y such that on one precisifi cation of the expression “is (identical with)” the expression refers to x and y, and on another precisifi cation the expression does not refer to x and y.

When I tried to construct a derivation of Thesis II from Thesis I, I construed the latter as Thesis Ib, but even if I had used Thesis Ia the derivation is blocked by (Q).34

34. Given the generally critical tone of this discussion, perhaps I need to emphasize that Thesis Ia represents an important philosophical discovery implicit in the work of Evans and Salmon. 66 quantifier variance and realism: essays in metaontology

One general lesson that comes out of this whole discussion is that there are complicated facts about precisifi cations that cannot be straightforwardly or univocally rendered in terms of the standard notions of “determinate” and “borderline.” Presumably, if the sentence “P” is true on every precisifi cation we will want to say “It is determinate that P. ” This holds, I think, even where “P” is a singular sentence of the form “a is F. ” If the singular sentence is true on every precisifi cation we will count “It is determinate that a is F” as true at least on one reading, the de dicto reading. For the case of a singular sentence, however, there will also be the following de re reading: “a is something x such that it is determinate that x is F. ” But what shall we mean by “it is determinate that x is F ”? There are two possible defi nitions: the fi rst is “on every precisifi cation ‘F ’ refers to x”; the second is “on every precisifi cation, if any expression refers to x then ‘F ’ refers to x.” The second defi nition, but not the fi rst, would allow us to say that the ship in the harbor on Monday is something x such that x is determinately brown, even though on some precisifi cation no expression including “brown” refers to x. On this defi nition (but not the other) it would also follow from the correct Thesis Ia that it cannot be indeterminate whether x is y, that is, either it is determinate that x is y or it is determinate that x is not y. But if we adopt this defi nition how shall we express the important idea that x is such that on every precisifi cation some expression refers to x? Perhaps we could reserve the expression “it is determinate that x exists” to express this idea. If so, however, we would have to tolerate saying, “It is determinate that there exists something x (e.g., the ship in the harbor on Monday) such that it is not determinate that x exists.” I suppose it matters little which formulations we adopt, as long as we understand what they mean and are prepared to carefully follow out the logic of precisifi cations.35 Let me end by remarking that although it seems often to be assumed in the literature that to deny the Evans-Salmon position is to believe in some notion of “vagueness in the world,” I don’t discern any signifi cant

35. On the fi rst defi nition it is clearly false to say, “Something x is such that ‘the ship in the harbor on Monday’ determinately refers to x.” Even on the second defi nition the sentence is false if construed as follows: “Where y is the expression ‘the ship in the harbor on Monday,’ something x is such that, on every precisifi cation, if any expression refers to y then ‘refers’ refers to the pair consisting of y and x.” With regard to the Evans-Salmon argument for Thesis II, the crucial point is that when we restate the argument in the more careful terms of precisifi cations, as I did earlier, we fi nd that the assumption of determinate reference does not hold in the sense required for the argument. the vagueness of identity 67 connection between that notion and anything I have been trying to defend in this paper.36 I’m not sure if I understand the question whether vagueness is in the world, but I want to stress that, on my view, the vagueness of “existence,” “identity,” and the rest of the referential apparatus is purely a matter of people being undecided about how to use their language.37

36. On the general topic of vagueness in the world, see Williamson, Vagueness, chap. 9. 37. For a number of essential criticisms of this paper, I am indebted to Tamar SzabÓ Gendler, Timothy Williamson, and Palle Yourgrau. Thanks also to Alan Sidelle and Jennifer Whiting. five c C

quantifier variance and realism

C A pervasive theme in Hilary Putnam’s writings for many years, c running as a constant thread through various changes in his views about realism, is a doctrine that he calls “conceptual relativism,” representative formulations of which are as follows:

[A]ll situations have many diff erent correct descriptions, and . . . even descriptions that, taken holistically, convey the same information may diff er in what they take to be “objects” . . . [T]here are many usable extensions of the notion of an object . . .1 [T]he logical primitives themselves, and in particular the notions of object and existence, have a multitude of diff erent uses rather than one absolute “meaning.”2

In these passages Putnam seems to be saying that the quantifi cational apparatus in our language and thought—such expressions as “thing,” “object,” “something,” “(there) exists” —has a certain variabil- ity or plasticity. There is no necessity to use these expressions in one way rather than various other ways, for the world can be correctly described using a variety of concepts of “the existence of something.” One of his favorite examples concerns a disagreement between mereologists and anti-mereologists as to how many objects there are in

1. Hilary Putnam, “The Question of Realism” in Words and Life (Harvard University Press, Cambridge, Mass., 1994), pp. 304–305. 2. Hilary Putnam, “Truth and Convention: On Davidson’s Refutation of Conceptual Relativ- ism,” Dialectica, Vol. 41 (1987), 69–77, at p. 71.

68 quantifier variance and realism 69 some domain.3 Suppose we are evaluating the truth of the sentence, “There exists something that is composed of Clinton’s nose and the Eiff el Tower.” Mereologists will accept this sentence, whereas anti- mereologists will reject it. Putnam’s doctrine of quantifi er variance implies that the expression “there exists something” can be interpreted in a way that makes the sentence true or in a way that makes the sentence false. Since both interpretations are available to us, we have a choice between operating with a concept of “the existence of something” that satisfi es the mereologist or operating with a diff erent concept that satisfi es the anti-mereologist. The doctrine of quantifi er variance may be philosophically unset- tling. Our initial reaction may be that, if we are free to choose between diff erent ways of conceiving of “the existence of something,” then this threatens a robust realist sense that there are things in the world whose existence does not in any way depend on our language or thought. The nature and force of this threat is one of the topics of this paper. Before proceeding let me emphasize that this paper is not an exer- cise in Putnam-exegesis. I take it as obvious that the doctrine of quantifi er variance is a central part of Putnam’s overall position, but I make no attempt here to trace Putnam’s evolving views as to how this doctrine relates to diff erent versions of realism. The possible threat to realism posed by this doctrine is something that I am trying to work through on my own terms, though I have little doubt that virtually every point I am going to make can be found somewhere in Putnam’s writings.

I think we should begin by making sure to repudiate a thoroughly confused, though somehow tempting, formulation of what the threat to realism is: “Since, according to the doctrine of quantifi er variance, our linguistic decisions determine whether or not there exists something composed of Clinton’s nose and the Eiff el Tower, it evidently follows that this thing’s existence or non-existence depends on our language and thought. If this point generalizes to every application of our concept of ‘the existence of a thing’—as the doctrine seems to imply—then quantifi er variance evidently confl icts with the realist idea of things existing independently of language and thought.”

3. “Truth and Convention,” pp. 70 ff . 70 quantifier variance and realism: essays in metaontology

The fallacy in this formulation lies in the claim that the doctrine of quantifi er variance implies that our linguistic decisions determine whether or not there exists something composed of Clinton’s nose and the Eiff el Tower. What the doctrine does imply is that our linguistic decisions determine the meaning of the expression “there exists something”; hence, they determine the meaning of the sentence “There exists something composed of Clinton’s nose and the Eiff el Tower.” Hence, the truth or falsity of this sentence depends in part on our linguistic decisions. It is merely a use-mention confusion to conclude that whether or not there exists something composed of Clinton’s nose and the Eiff el Tower depends on our linguistic decisions. Consider the two sentences: “There exists something composed of Clinton’s nose and the Eiff el Tower” and “Whether or not there exists something composed of Clinton’s nose and the Eiff el Tower depends on our linguistic decisions.” Quantifi er variance implies that the expression “there exists something” can be interpreted in a way that makes the fi rst sentence true or in a way that makes it false. But there is no relevant way to interpret “there exists something” that would make the second sentence true. The second sentence expresses an absurd form of linguistic idealism that is not at all implied by quantifi er variance. There is the familiar joke about how many tails a dog would have if the word “tail” were used to refer to legs—the correct answer, everyone seems to agree, is “one,” because how many tails a dog has does not depend on our linguistic decisions. The meaning and truth-value of the sentence “A dog has one tail” depends on our linguistic decisions, but how many tails a dog has does not depend on our linguistic decisions. Why does it seem harder to grasp this point when we are talking about the meaning of quantifi er expressions rather than a general term such as “tail”? I think that part of the diffi culty stems from our wanting to say, when we are formulating the doctrine of quantifi er variance, that the relevant variations would still leave us with a kind of quantifi er expression, an expression that, as Putnam put it in the earlier quote, continues to signify a notion of the existence of something. This formulation may be misconstrued as implying that, although the meaning of the quantifi er expressions and the notion of existence remain fi xed, what we are somehow going to accomplish with our linguistic decision is to alter the truth-values of our existential claims. And that might indeed amount to a lunatic form of linguistic idealism. What needs to be stressed, however, is that the doctrine of quantifi er variance only allows for the possibility of a change in the meaning of quantifi er expressions, quantifier variance and realism 71 yielding a diff erentor extended notion of existence; only in this way does the doctrine allow our linguistic decisions to aff ect the truth-values of existential sentences. Once the meaning of the quantifi er is fi xed there is no further eff ect that our decisions can have on the truth-values of typical existential sentences. To say that even after there has been a change in the meaning of quantifi er expressions we still have a “kind of quantifi er expression” and a “notion of existence” is merely to indicate a degree of similarity between the concepts we started with and those we end up with. Nothing is being said here to imply the idealist view that what exists in the world depends on our linguistic or conceptual decisions. Suppose that I start out with the anti-mereologist’s position that the sentence “There exists something composed of Clinton’s nose and the Eiff el Tower” is false. If I accept quantifi er variance I will allow that I can make intelligible to myself a change in the meaning of the expression “there exists something” which would have the eff ect of rendering the sentence true. I might characterize this change as simply “giving up the quantifi er” and “giving up the notion of the existence of something,” but it seems more natural to characterize it rather as “acquiring a new kind of quantifi er” and “acquiring a new notion of the existence of something.” The second characterization seems more natural because it seems clear that the imagined change in the meaning of the expression “there exists something” will leave the expression’s general role in the language largely intact. In particular, the purely syntactic and formal logical properties of the expression will not be changed at all (the formal principles of quantifi cational logic will be unaltered). It therefore seems natural to follow Putnam in treating relevant variations in the meaning of such expressions as “there exists something” as yielding an altered quantifi cational apparatus and an altered concept of the existence of something. If I start out with the anti-mereologist’s stance, what exactly is involved in changing the meaning of the quantifi er with the eff ect of making the mereologist’s sentences come out true? There are important complications in the answer to this question that I will return to later, but the basic idea is quite simple. In general, we explain the meaning of a logical constant by describing the role it plays in determining the truth-conditions of sentences. Thus we explain the meaning of “and” by saying that sentences of the form “p and q” are true if and only if both the sentence “p” and the sentence “q” are true. If we were to explain some imagined change in the meaning of “and” we would do so by describing a change in the truth-conditions of sentences 72 quantifier variance and realism: essays in metaontology containing “and.” Analogously we explain the relevant change in the meaning of the quantifi er, which will render the mereologist’s sentences true, roughly as follows: In the new meaning, any sentence of the form “There exists something composed of the F-thing and the G-thing” is true if the expression “the F-thing” refers to something and the expression “the G-thing” refers to something. We are here using the quantifi er in one of its meanings to explain the other meaning. It is evidently not part of the doctrine of quantifi er variance to claim that the meaning of the quantifi er can be somehow analyzed or defi ned in terms not involving the quantifi er—any more than the meaning of “and” can be analyzed or defi ned in terms not involving conjunction. The issue of quantifi er variance should therefore not be confl ated with familiar questions about the analyzability of (the criteria for) the identity or existence of a thing. Quantifi er variance is not a matter of substituting one “defi nition” for another; it’s a matter of substituting one range of truth-conditions for another.

I suspect that the deepest source of the illusion—and I am maintaining that it is an illusion—that quantifi er variance confl icts with realism stems from the analogy we are led to draw between diff erent kinds of quantifi ers, the analogy, that is, that leads us to speak, not simply about eliminating the quantifi er and the concept of existence, but about a variation that still leaves us with a diff erent quantifi er and a diff erent concept of existence. Suppose, again, that I start out with the anti-mereologist’s position that the sentence “There exists something composed of Clinton’s nose and the Eiff el Tower” is false. In my use of “there exists something” a sentence of the form “There exists something composed of the F-thing and the G-thing” counts as true only if “the F-thing” and “the G-thing” refer to things that are connected (united) in some special ways. Let me call my use of “there exists something” the A-use. I now imagine a diff erent use of “there exists something”—I will call this the M-use—in which a sentence of that form counts as true so long as “the F-thing” and “the G-thing” refer to things, no matter how they are connected. Now, how am I to describe in my language, the A-language, what is going on in the M-language? When speakers of the M-language assert “There exists something composed of Clinton’s nose and the Eiff el Tower” they speak the truth. quantifier variance and realism 73

That is implied by the semantic rule that has been explicitly laid down in my description of the truth-conditions in the M-language of sentences of the form “There exists something composed of the F-thing and the G-thing.” I assume, however, that this explicit stipulation carries with it a natural way of fi lling in the truth-conditions for an indefi nite variety of other sentences in the M-language. Take, for example, the sentence “There exists something that is now being touched by exactly two people.” This sentence is not rendered true in my A-language by a situation in which one person is touching Clinton’s nose and one person is touching the Eiff el Tower,4 but my implicit assumption is that the sentence would be true in the M-language with respect to that situation. When I refl ect on a range of examples of this sort I fi nd myself tempted to say (in my A-language) something like the following: Although there doesn’t exist anything composed of Clinton’s nose and the Eiff el Tower, it’s somehow as if there does exist such a thing relative to the M-language. This “as if” formulation is a simile; it is a way of expressing the felt holistic analogy between the A-use of the quantifi er expressions and the M-use. As such it may be perfectly innocuous. But similes can often lead to metaphors, and metaphors, if we are not careful, can sometimes lead to lunacy. The move from simile to metaphor might take the form of simply dropping the words “as if” from the previous formulation. We then wind up saying, “Relative to the M-language, there exists something composed of Clinton’s nose and the Eiff el Tower.” As a metaphor this may still be okay; it may in fact qualify as what John Wisdom called an “illuminating paradox.”5 It can, however, turn into utter confusion and philosophical madness if one forgets that it is merely a metaphor. One may then wind up claiming that language or thought literally creates everything that exists, that nothing could exist if there were no people speaking or thinking. What we need to be clear about—and this may require a continual eff ort on our parts—is that the doctrine of quantifi er variance does not imply any such idealist formulation.6

4. Subject to a slight qualifi cation to be made later. 5. John Wisdom, Philosophy and Psychoanalysis (University of California Press, Berkeley, 1969). 6. Ernest Sosa’s “existential relativity” has it that things exist relative to conceptual schemes, but Sosa takes great pains to disassociate himself from the idealist claim that language or thought literally creates things. An important question considered in Sosa’s discussion, which I’m bypassing here, is how “far down” quantifi er variance can go. See Ernest Sosa, “Existential Relativity,” Midwest Studies in Philosophy, 22 (1999), and my “Sosa’s Existential Relativism,” in Ernest Sosa and His Critics (Blackwell, Oxford, 2004). 74 quantifier variance and realism: essays in metaontology

The push towards the simile and metaphor, and the accompanying risk of falling off the edge of idealist madness, is enhanced when we consider the following kind of sentence: “In the M-language the expression ‘thing composed of Clinton’s nose and the Eiff el Tower’ refers to a thing composed of Clinton’s nose and the Eiff el Tower.” This sentence (about the M-language) cannot be true in my A-language. Since (as I say in my A-language) there isn’t any thing composed of Clinton’s nose and the Eiff el Tower, there can’t be any such thing referred to by any expression in the M-language. But if we make the natural assumption that the truth-conditions for sentences in the M-language containing the word “refers” will respect the disquotational principle, then the previous sentence about the M-language, although false in the A-language, counts as true in the M-language. To take this one small step further, the sentence “In the M-language the expression ‘thing composed of Clinton’s nose and the Eiff el Tower’ refers to something” counts as false in the A-language but true in the M-language. As goes “existence” so goes “reference.” The lesson here is that if we are imagining that the quantifi er expressions in the M-language function diff erently from our A-quantifi ers, then we can—and naturally will—imagine a correlative diff erence in the use of the word “refers.” If we alter our concept of “a thing” then we alter our concept of “reference to a thing.” But here especially we may feel impelled to say, “It’s as if, relative to the M-language, there is something being referred to by the expression ‘thing composed of Clinton’s nose and the Eiff el Tower.’ ” Again we must struggle to keep this simile in tow and not let it lead us into idealism.

III

Let me now briefl y consider a few formulations that figure in the literature on realism and ask how they relate to quantifi er variance. I will assume that these formulations are being posed in the A-language. (In the next section I will argue that people who speak plain English are in fact using the A-language, not the M-language.) 1. The truth of a (contingent) statement in any language depends on what things exist and what properties these things have. This claim is not threatened by quantifi er variance. People who speak the M-language will have a diff erent concept of “a thing” than mine, but the truth of their statements will depend on how it is with things in the world, in my quantifier variance and realism 75

(A-language) sense of “things in the world.” For instance, if one person is touching Clinton’s nose and another person is touching the Eiff el Tower, then the existence and properties of these things will render true the M-statement, “There exists a thing that is being touched by exactly two people.” Might there, however, be truths that do not depend on how things are, truths that are somehow not about things in my sense (or perhaps in any sense that I can make intelligible to myself)? I think that this question takes us into the vicinity of Kantian noumena and mysticism. It’s not a question that I’m addressing here. This last question should, however, be distinguished from another one: Might there be things whose nature and unity I cannot understand? The doctrine of quantifi er variance does not preclude there being such things.7 (This relates to my earlier point that the doctrine is not to be confused with issues about the “analyzability” of existence and identity.) 2. The truth of any (contingent) statement in any language depends on the existence and properties of the things referred to by the non-logical expressions in the statement. This claim as it stands is immediately refuted by such sentences as “The average professor has fewer children than the average plumber,” the truth of which evidently does not depend on there being any things referred to by the expressions “the average professor” and “the average plumber.” Let’s suppose, however, that this sentence can be viewed as in some sense “merely a transformation” of some other “basic” sentence. Claim 2 might then be roughly understood as saying that, in any language, the truth of a basic sentence depends on the existence and properties of the things referred to by the non-logical expressions, and the truth of other sentences depends on their equivalence by way of transformation rules to some basic sentences. Is claim 2 threatened by quantifi er variance? If we suppose that the notion of “reference to a thing” is kept fi xed to the language in which 2 is formulated (which I am now imagining to be the A-language), then I suspect that 2 is threatened. In terms of my A-concept of “reference to a thing” I cannot explain the truth of the M-statement “There exists a thing that exactly two people are touching” by appealing

7. So the doctrine need not prevent us from agreeing with Thomas Nagel’s claim in The View from Nowhere (Oxford University Press, N.Y., 1986), p. 98: “We can speak of ‘all the things we can’t describe,’ ‘all the things we can’t imagine,’ ‘all the things humans can’t conceive of,’ and fi nally, ‘all the things humans are constitutionally incapable of ever conceiving.’” 76 quantifier variance and realism: essays in metaontology to the reference of “touching.”8 (Quantifi er variance might thus be said to induce a certain kind of systematic diff erence of meaning in the word “touching” and, by the same token, virtually any other general word.9) I might try to treat as “basic” only those sentences of the M-language that have the same truth-conditions as homophonic sentences in the A-language and derive the rest by transformations, but it is not obvious what such transformation rules would look like. Claim 2 requires that it be possible to formulate in my A-language something approximating to a Tarski-style theory of truth for the M-language, that is, very roughly put, a fi nitary account in terms of reference relations that yields for each of the indefi nite number of M-sentences what the conditions are for its truth. It seems quite possible that this requirement can’t be met. If claim 2 constitutes a certain kind of “correspondence theory of truth” (what might be called a “referential correspondence theory”) then this kind of correspondence theory may indeed fall to quantifi er variance—assuming, again, that claim 2 is being formulated in the A-language and that what is meant by “reference to a thing” is kept fi xed to this language. But I think it is therefore clear at this point that this kind of correspondence theory is not essential to a straightforward realist view of the world. The possible falsity of claim 2 does not threaten the basic realist idea (now being expressed in the A-language) that the world consists of things whose existence and properties are independent of language or consciousness. Putnam takes Donald Davidson’s animadversions against “the very idea of a conceptual scheme” to constitute a repudiation of quantifi er variance.10 I don’t fully understand what Davidson is driving at in his

8. I am perhaps making some sort of controversial assumption here related to the inscrutability of reference, or better put, the scrutability of non-reference, for it might be questioned whether some kind of ingenious reference scheme assigning a reference to “touching” might somehow do the trick. I’m quite sure that this is impossible, but perhaps I should limit myself to saying more cautiously that I don’t see what such a scheme could be. See, further, my Dividing Reality (Oxford University Press, N.Y., 1993), pp. 102–109, and my “Objectivity Without Objects” in the Proceedings of the Twentieth World Congress of Philosophy: Epistemology, Vol. 5 (Philosophy Documentation Center, Bowling Green State University, 2000). 9. There may yet be a sense in which the qualities (features, points of similarity) signifi ed by general words remain the same despite quantifi er variance. I am leaving a number of questions open here. 10. This seems to be Putnam’s assumption in “Truth and Convention.” See Donald Davidson, “On The Very Idea of a Conceptual Scheme,” in Inquiries into Truth and Interpretation (Oxford Univer- sity Press, London, 1984). quantifier variance and realism 77 talk against “conceptual schemes,” but I doubt that it has really any defi nite connection to the issue of quantifi er variance. Let us note that the M-language is “translatable” into my A-language in at least the following sense: For any M-sentence I can fi nd an A-sentence with the same truth-conditions, where two sentences have the same truth- conditions if, relative to any context of utterance, they hold true with respect to the same possible situations.11 This kind of intertranslatibility between the A-language and the M-language holds even if it’s impossible to formulate in the A-language a fi nitary theory of truth for the M-language. (“For any M-sentence I can fi nd an A-sentence with the same truth conditions” does not entail “There is a fi nitary truth-theory which yields, for any M-sentence, an A-sentence with the same truth- condition.”) It is indeed because of this intertranslatibility that the proponent of quantifi er variance maintains that the two languages are equally capable of truthfully describing the world. I don’t know, therefore, whether the A-language and the M-language qualify as “diff erent conceptual schemes” in Davidson’s sense. If they do, he has certainly said nothing to show why there couldn’t be diff erent conceptual schemes in this way. Another idea associated with Davidson is that any (learnable) language must be describable in terms of (something approximating to) a Tarski-style truth-theory.12 Interpreted in one way, this claim, too, is not threatened by quantifi er variance. If it is possible to formulate in the A-language a truth-theory for the A-language, then it is possible to formulate in the M-language a truth-theory for the M-language, each theory being formulated in terms of each language’s meaning of “reference to a thing.” If, however, Davidson is claiming that it must be possible to formulate in our (“home”) language a truth-theory for any possible language, this seems clearly untenable, for we can surely conceive of people whose sensory apparatus diff ers from ours to the extent that we cannot describe the truth-conditions of some of their sentences.13 A question that might be raised, however, is this: If we start out speaking the A-language (as I think we in fact do), and we are not able to formulate in this language a truth-theory for the M-language, how

11. In other words, they have the same “character” in the sense of David Kaplan, “Demonstra- tives” in J. Almog, J. Perry, and H. K. Wettstein, eds., Themes from Kaplan (Oxford University Press, N.Y., 1989). 12. See Inquiries into Truth and Interpretation. 13. See Thomas Nagel’s criticism of Davidson in The View from Nowhere, pp. 94–98. 78 quantifier variance and realism: essays in metaontology can we come to learn that language? How do we come to agree (as I assume we do) on the truth-conditions of an indefi nite number of M-sentences? The answer is that we are relying on our shared sense of the analogy between our A-quantifi er and the M-quantifi er. Starting with the A-language I teach someone the M-language ostensively, by giving a few representative examples of how the M-language works. That anyone can then be expected to go on in the same way is part of what Putnam means when he says that our quantifi cational concepts admit of “many usable extensions.”14 3. The truth of any (contingent) statement in any language depends on what the facts are in the world. This might also be put by saying that a statement’s truth depends on the way the world is, the statement’s being true only if the world is the way the statement says the world is. Claim 3 is another perennial version of “the correspondence theory of truth.” Is it threatened by quantifi er variance? In the philosophical literature we fi nd facts (as well as states-of- aff airs, propositions, and properties) sometimes understood in a coarse- grained unstructured manner and sometimes in a fi ne-grained structured manner.15 A structured fact is what Putnam describes—and rejects—as a “sentence-shaped thing in the world.”16 A fact in this sense is built up in a certain way out of things and properties (and perhaps logical oper- ations). If we have logically equivalent sentences, such as “This is round” and “This is either round and red or round and not red,” then each sentence expresses a diff erent structured fact, or, as another alternative, they express the same structured fact but only one sentence (the fi rst) succeeds (or succeeds better) in properly picturing the fact’s structure. In the coarse-grained sense, however, these sentences express the same unstructured fact, and it therefore makes no sense to ask which sentence does better at depicting this fact. If claim 3 is understood in terms of unstructured facts it does not confl ict at all with quantifi er variance. Quite the contrary, the basic idea of quantifi er variance can be nicely formulated by saying that the same (unstructured) facts can be expressed using diff erent concepts of “the existence of a thing,” that statements involving diff erent kinds of

14. “The Question of Realism,” p. 305. 15. See George Bealer, Quality and Concept (Oxford University Press, N.Y., 1982), pp. 181–187, and David Lewis, On the Plurality of Worlds (Basil Blackwell Ltd, N.Y., 1986), p. 56. It must be borne in mind that I have defi ned the notion of “truth-conditions”—and consistently use it here—in the coarse-grained unstructured sense. 16. “The Question of Realism,” p. 301. quantifier variance and realism 79 quantifi ers can be equally true by virtue of the same (unstructured) facts in the world. The notion of a structured fact does, however, raise certain problems for quantifi er variance—but one must be careful not to misun- derstand what these problems are. If I change what I mean by “a thing” then I must also change what I mean by “the way a fact is built up out of things and properties”; hence, I must change what I mean by “a structured fact.” But, note carefully, that it does not follow that there are no structured facts independent of language—any more than it follows that there are no things independent of language. Indeed, when one looks at this carefully I think one sees that claim 3 can be sustained even if the facts on which the truth of statements depend are taken to be structured. I can say in my A-language that the truth of an M-statement depends on the structured facts, even though what I mean by “the structure of the facts” is not what a speaker of the M-language means. A problem arises, however, if one wants to say that corresponding to any true statement in any language there is, as Putnam puts it, “a unique sentence-shaped thing in the world,”17 in other words, that each true sentence states one structured fact. How can we say which one that is? If we stick to our own language then we have the trivial disquotational formulation, “The sentence ‘p’ states the structured fact that p,” and perhaps we don’t have to worry further about which structured facts are identical with which. But if I am speaking the A-language, how am I to say which structured fact is stated by the true M-sentence “There exists a thing that exactly two people are touching,’’ when I am able to “translate” that sentence into a variety of structurally diff erent sentences in my language having the same truth- conditions? I don’t doubt that some moves might be made here (for instance, it might be suggested that it is indeterminate which structured fact, in my A-sense of “structured fact,” the true M-sentence states), but I am inclined to agree with Putnam that, once we’ve accepted quantifi er variance, there is no point in trying to hold onto language-shaped facts that are in the world independent of language. However, we can retain the notion of an unstructured fact. I think this is indeed our most basic notion of “reality,” “the world,” “the way it is,” and this notion can remain invariant through any changes in our concept of “the things that exist.”

17. “The Question of Realism,” p. 301 (my emphasis). 80 quantifier variance and realism: essays in metaontology

In the two passages quoted from Putnam at the outset of this paper he says in the fi rst one that “there are many usable extensions of the notion of an object,” and in the second that “the notions of object and existence have a multitude of diff erent uses rather than one absolute ‘meaning.’ ” These two characterizations of quantifi er variance are subtly diff erent: the fi rst seems to talk about the possibility of having diff erent meanings attached to the quantifi er, whereas the second seems to say that in actuality the quantifi er is used with many diff erent meanings. In the preceding discussion I was arguing in behalf of quantifi er variance in the fi rst sense, not the second; that is, I was arguing in behalf of possible quantifi er variance, not actual quantifi er variance. In the discussion surrounding the second passage—the passage that seems to describe actual quantifi er variance—Putnam criticizes an anti-mereologist who gives the following speech: “I know what you’re talking about if by an object you mean a car, or a bee, or a human being, or a book, or the Eiff el Tower. I even understand it if you refer to my nose or the hood of my car as ‘an object.’ But when philosophers say that there is an ‘object’ consisting of the Eiff el Tower and my nose, that’s just plain crazy. There simply is no such object. . . . and it’s crazy to suppose that every fi nite universe contains all the objects that those [mereologists] would invent, or, if you please, ‘postulate.’ You can’t create objects by ‘postulation’ any more than you can bake a cake by ‘postulation’.”18 Now I am myself an anti-mereologist who considers the above speech to be quite reasonable. At least it seems reasonable if it is taken in the following spirit: “I assume we’re all speaking plain English, and that we’re employing the quantifi er in the sense of plain English. Given that, I understand perfectly well what it means to talk (in plain English) about such things as cars, bees, human beings, books, and the Eiff el Tower, or even to talk about such marginal things as noses and car-hoods. But it’s crazy to say (in plain English) that there exists something composed of my nose and the Eiff el Tower. And you can’t create any such thing by ‘postulating’ it or by changing your language—that would be an absurd form of linguistic idealism.” As a believer in possible quantifi er variance I would add: “I can of course make intelligible to myself a possible language that diff ers from plain English in the meaning of its quantifi er. In that imagined language the

18. “Truth and Convention,” p. 72. quantifier variance and realism 81 sentence ‘ There exists something that is composed of my nose and the Eiff el Tower ’ would be trivially true. If you’re speaking that language for some reason, that’s okay. Just don’t pretend to be speaking plain English.” Putnam seems to reject this criticism of the mereologist because he thinks that, in plain English, there actually are diff erent senses of the quantifi er, and that the mereologist’s claims are true in one of those plain English senses. It seems to me, however, that the linguistic evidence indicates that fl uent speakers of English do not speak the mereologist’s language. Speakers of English use what I was calling in previous sections the A-language; they do not use the M-language. My attitude towards the mereologist might be compared to the attitude that I and most current philosophers have towards the traditional epistemologist who says, “Only sense-data can be (strictly) perceived; physical objects cannot be perceived.” Most philosophers would nowadays agree that the sense-data philosopher is saying something that is absurd. This is not to deny that we can imagine a use of “perceives” diff erent from the one in plain English which would render the sense-data philosopher’s remark true. Indeed, this philosopher’s confusion might be characterized as “language gone on holiday”;19 the philosopher has somehow confused himself into speaking a new language without realizing it. But confusion this is; the philosopher has made a mistake. What holds for the sense-data philosopher holds as well, I think, for the mereologist. Putnam, it seems, doesn’t see it that way. I’m quite sure he would agree with what I have said about sense-data philosophy, but he apparently views the issue of mereology diff erently. He says that if an anti-mereologist attacks the mereologist’s view as being obviously wrong, then this shows that the anti-mereologist is a “metaphysical realist” who wrongly rejects Carnap’s “principle of tolerance.”20 I think, however, that the attack can stem from the anti-mereologist’s being a philosopher of common sense. I think it’s clear that Putnam’s use of “metaphysical realism” has sometimes been confusing. I think he uses this term as the name of a certain attitude in metaphysics as much as the name of a defi nite philosophical position. To the extent that there is a position it is the denial of even the possibility of quantifi er variance. Metaphysical realism says that there is somehow one metaphysically privileged sense of the

19. See Ludwig Wittgenstein, Philosophical Investigations, I. 38. 20. “Truth and Convention,” p. 73. 82 quantifier variance and realism: essays in metaontology quantifi er, that any departure from this privileged sense would leave us without adequate resources to state the truth properly. There is, so to speak, the quantifi er that God would use, and to get things metaphysically right, that’s the quantifi er we have to use. I think it is clear, however, that most people whom Putnam would call metaphysical realists have never heard of quantifi er variance, and would not readily acknowledge that they have any special views about the necessity to use one kind of quantifi er rather than another. It’s their attitude when they philosophize that suggests that they tacitly deny the possibility of quantifi er variance. What is the nature of this attitude? Putnam has characterized it in various ways throughout his writings, but I would like to put it as follows: these philosophers argue too much. They descend upon us as a legion of ontological lawyers, their briefcases overfl owing with numerous arguments and counterarguments, a case for one entity, a case against another. Questions that appear to be trivial beyond the pale of conversation are somehow converted by them into occasions for deep theoretical debate. “Metaphysical realists” are affl icted with a kind of hyper-theoreticalness. I would certainly not claim that there is an inevitable connection between this affl iction and the issue of quantifi er variance, but I think that in many cases a potentially helpful diagnosis of the affl iction would be roughly as follows: If whenever you make an existential claim in metaphysics you are tacitly or unconsciously assuming that the claim has to be couched in terms of a quantifi cational apparatus that is in some sense the uniquely right one—the one that God would use—then this assumption is likely to lead you to futile and interminable pseudo- theoretical arguments. In his discussion of mereology Putnam implies that the proper remedy for the hyper-theoreticalness of “metaphysical realism” is Carnapian tolerance. I am suggesting that another remedy is “ordinary language philosophy” or an appeal to common sense. Each of these two remedies is appropriate in diff erent cases. Carnapian tolerance is appropriate where an existential sentence being disputed by philosophers is actually vague or ambiguous in plain English (or whatever natural language the philosophers are speaking), and each disputant has in eff ect become attached to one permissible interpretation of the sentence. In this kind of case once the disputants realize that the quantifi er admits of relevantly diff erent interpretations in their language they should each say— thereby exhibiting Carnapian tolerance—that both of them are right, taking the quantifi er in the relevantly diff erent meanings. But there are other cases—and I think the case of mereology is an example—in quantifier variance and realism 83 which the disputed sentence admits of only one relevant meaning in plain English, and one of the disputants is saying something that— interpreted in plain English—is trivially absurd. What may prevent this disputant from simply acquiescing to ordinary usage and common sense is the implicit assumption that any question of ontology must be highly theoretical, because any such question turns on how to describe the world in terms of the metaphysically privileged sense of the quantifi er. Once the possibility of quantifi er variance is accepted, and the notion of a metaphysically privileged quantifi er is abandoned, there is nothing to inhibit us from simply expressing the trivial common sense truth in terms of the quantifi er we actually have in our language. In these kinds of examples acknowledging quantifi er variance leads not to Carnapian tolerance but rather to the common-sense philosopher’s ridicule of needless philosophical paradox.21 What may make it diffi cult for Putnam to see that the mereologist’s position, stated in plain English, seems to be trivially absurd is the fact that there is a closely related position that is not absurd. If we have Clinton’s nose and we have the Eiff el Tower then perhaps we also have the total quantity (or mass) of matter that composes these two things. To talk about that matter may seem tolerably correct in plain English. If so, there may be, in terms of plain English, something that can be said to consist of Clinton’s nose and the Eiff el Tower. What needs to be understood, however, is that this is not the thing that mereologist’s mean by “the sum of Clinton’s nose and the Eiff el Tower.” The matter that makes up the nose and the tower at one moment may not be exactly the same matter that makes up these objects at another moment; hence the matter cannot be identifi ed with the mereologist’s “sum” of the nose and the tower. Once this point is grasped I think it seems quite clear that there is no defensible interpretation of the plain English quantifi er which would make true the sentence “There exists something that remains composed of Clinton’s nose and the Eiff el Tower while these two objects alter their matter.” Perhaps the paradoxical eff ect of the mereologist’s position is clearer if we consider non-contemporaneous objects, such as Socrates’s nose and the Eiff el Tower, which the mereologist will say has a “sum.”

21. I allude casually here to Carnapian tolerance because Putnam mentions it. I am not seriously discussing or defending Carnap’s own thinking on these matters. He seemed to imply that realists ought to “tolerate” phenomenalists, which I think is out of the question for reasons that go beyond the present discussion. See Rudolph Carnap, “Empiricism, Semantics, and Ontology,” in Meaning and Necessity, 2nd edition (University of Chicago Press, Chicago, 1956). 84 quantifier variance and realism: essays in metaontology

The paradox is still clearer when we add to the mereologist’s position another doctrine that almost always goes together with it: the doctrine of temporal parts. If we have temporal parts and also mereological sums then we wind up with the position, familiar from Quine, Lewis, and many other philosophers, in which any set or series of bits of matter constitutes an object—that is, constitutes an object that can be said to occupy a place and to have ordinary qualities of shape, color, texture, and so on. Imagine that a tree stands in a certain yard, and that nothing extraordinary happens to it. Con- sider the succession of (stages of) bits of matter that make up the whole tree in the daytime and just the trunk during the nighttime. According to the position just mentioned this succession corresponds to an object on a par with the tree and the trunk. We could then give a name to such an object, say, “a shmree.” A shmree, on this view, is a brown wooden object in the yard that loses its branches every night and regains them every morning. Surely to claim that there is such an object seems to violate common sense to the highest degree. Philosophers who are attracted to this view, I am suggesting, consciously or unconsciously reject the possibility of quantifi er variance, and therefore take themselves to be engaged in some high- level theoretical speculation about whether the metaphysically privileged concept of existence—God’s quantifi er—would encom- pass shmrees. We who accept quantifi er variance, on the other hand, consider the question about the existence of shmrees to be utterly trivial. It seems immediately clear that our ordinary concept of “a thing” renders the sentence “There is a thing in the yard that keeps gaining and losing branches” false. We can indeed make intelligible to ourselves a diff erent concept of “a thing” which would render the sentence true, but, once we reject as unintelligible the idea of a metaphysically privileged concept of “a thing,” there is no reason for us to resist acknowledging the concept that we in fact have. And, then, there is no reason to resist acknowledging the obvious and trivial truths that fl ow from that concept. We can say, in the general spirit of ordinary language philosophy: It’s obviously false to assert that there is a thing in the yard that gains and loses branches, because that’s not the way we talk. But the appeal here to “the way we talk” is not to be misconstrued as suggesting the idealist view that, had we talked diff erently, there would have been something in the yard that gains and loses branches. The suggestion rather is that, had we talked diff erently, the sentence “There is a thing in the yard that gains and loses branches” would have been true. The quantifier variance and realism 85 appeal to “the way we talk” is a reminder that we have the concepts that we have and that, especially when we do philosophy, we need to remain responsible to the concepts we claim to be using, and not confuse ourselves into slipping inadvertently into a diff erent way of talking. If the denial of the possibility of quantifi er variance leads to a certain kind of hyper-theoreticalness in ontology, the acceptance of quantifi er variance leads to an attitude that might be called “ontological shallowness,” or what Quine at one point calls “steadfast laymanship.”22 That is the attitude I am commending. What I have tried to explain is that, depending on what kind of case we are dealing with, ontological shallowness will either take the form of Carnapian tolerance or the form of insisting that the ontological judgments of common sense be respected in philosophy.

I am primarily concerned here with the ontology of physical objects—indeed, highly visible physical objects (roughly, what Aus- tin called “moderate-sized dry goods”23)—and the issue of quantifi er variance with respect to that domain. I’m not addressing questions that involve abstract things (such as the familiar question whether numbers exist in the same sense as physical objects). I have maintained thus far that with respect to the domain of physical objects we can understand possible variations in the quantifi er. Are there, however, examples of actual variations? Do we, in fact, when describing physical objects in plain English, alter the meaning of the quantifi er? Insofar as English is conceived of as an evolving language, whose major sortal-categories change somewhat from time to time, I’m inclined to think that this change carries with it some change in the meaning of “a thing,” and therefore in the meaning of “the existence (and reference to) a thing.” Moreover, insofar as a given speaker’s rep- ertoire of sortals changes over time, this may qualify as a change in the speaker’s concept of “a thing” (modulo issues about the “communal” nature of each person’s concepts). In any case, the “sortal-dependence”

22. W. V. Quine, Word and Object (MIT Press, Cambridge, Mass., 1960), p. 261. 23. J. L. Austin, Sense and Sensibilia (Oxford University Press, London, 1962), p. 8. 86 quantifier variance and realism: essays in metaontology of our concept of “a thing” is an issue I have discussed at length elsewhere, and I will not try to add to it here.24 It’s essential to distinguish between quantifi er variance and what David Lewis calls quantifi er restriction. Lewis points out that if, looking into the refrigerator, I assert, “There is no beer,” I am restricting the quantifi er to (the domain of) what is in the refrigerator. The unrestricted quantifi er, by contrast, says what exists “strictly” or “simpliciter.”25 That the quantifi er is often contextually restricted in the way that Lewis says may be granted by all. The question of quantifi er variance, however, pertains to the unrestricted quantifi er, to our concept of “existence simpliciter.” We will see in a moment that Lewis, while often appealing in his work to restrictions on the quantifi er, tacitly denies the possibility of there being variations in the unrestricted quantifi er. I have often heard it suggested—though not, to my knowledge, on the basis of anything Lewis himself says—that, once we acknowledge the phenomenon of quantifi er restriction, it is no longer clear that the ontological commitments of Lewis and other philosophers, in favor of mereological sums and temporal parts, confl ict with the views of common sense. Take, again, the sentence, “There exists a brown wooden thing in the yard that keeps losing its branches every night and gaining them back every morning.” Lewis says that this sentence is true, and ordinary people say it is false. But that’s because Lewis is using the quantifi er unrestrictedly, whereas ordinary people are restricting the quantifi er to the domain, roughly speaking, of “familiar bodies.” So there is really no confl ict, after all. This suggestion is misguided, I think. Ordinary people must have a concept of “existence simpliciter.” They must understand how to use the quantifi er unrestrictedly; otherwise no such use could be part of the English language. With respect to the cited sentence, if we explain to ordinary people that the “brown wooden thing” in question need not be any kind of familiar thing, it need not be an interesting thing or the sort of thing one would normally talk about, they still regard the sentence as insanely false, though now qualifi ed in absurdly irrelevant

24. The notion of sortal-dependence has been introduced to contemporary philosophy by Wiggins’s seminal work, but I cannot tell what his attitude is towards quantifi er variance. See David Wiggins, Sameness and Substance (Harvard University Press, Cambridge, Mass., 1980). I discuss sortal- dependence in The Concept of Identity (Oxford University Press, N.Y., 1982), especially chs. 2 and 3, and “Basic Objects: A Reply to Xu,” Mind and Language, Vol. 12, 3 (1997), 406–412. 25. Lewis, On the Plurality of Worlds , p. 3. quantifier variance and realism 87 ways. For they take it for granted that there is no brown wooden thing in the yard of any sort whatever that keeps gaining and losing branches. They would in fact claim to be able to see with their eyes that this is the case. So there surely is the starkest confl ict between the views of common sense and the ontological position of Lewis and other philosophers. I’ve tried to indicate earlier that there is a certain general connection between the tacit rejection of quantifi er variance and the tendency amongst many philosophers to reject the ontological judgments of common sense. In Lewis’s case, however, there is a more specifi c connection between these two rejections. His main reason for believing in mereological sums, contrary (as he admits) to the views of common sense, is that he holds a certain view about quantifi er vagueness that, as I will try to explain, comes out of his tacit rejection of quantifi er variance. Let me give an illustration to bring out Lewis’s reasoning about mereological sums. Suppose we are constructing a very simple table out of a top and a leg. If we accept the common sense position, and do not believe in arbitrary mereological sums, we will say that at the beginning of this process, when the top and the leg were far apart (or, perhaps, when no one even had the plan to connect them), there wasn’t any thing made up of the top and the leg. At the end of the process there is a thing, the table, made up of the top and the leg. Now, given the gradualness of this process (the gradualness, say, of driving a nail into a piece of wood or gluing two pieces of wood together), there must surely be a time at which it is indeterminate whether we have a table yet. At that time, then, it is indeterminate whether we have anything composed of the top and the leg.26 But that, says Lewis, makes no sense. It can be indeterminate whether there exists something having a certain property only if there exists a thing such that it’s indeterminate whether it has that property. In order for it to be indeterminate whether there exists something composed of the top and the leg, then, there must exist a thing such that it’s indeterminate whether it is composed of the top and the leg. What thing could that be? It seems as if we are being led by the common sense view to say incoherently that there exists a certain thing such that it’s indeterminate whether there exists

26. A variation of Lewis’s argument, with convincing arguments that there must be examples of the relevant kind of indeterminateness, is given in Theodore Sider, “Four Dimensionalism,” The Philosophical Review, Vol. 106, no. 2, 197–231. 88 quantifier variance and realism: essays in metaontology that thing. “What is this thing such that it sort of is so, and sort of isn’t, that there is any such thing?”27 The only coherent way to view the matter, Lewis concludes, is to assume, contrary to common sense, that at every point in the process there is something composed of the top and the leg. At some point in the process it may be indeterminate whether the thing composed of the top and the leg qualifi es as a table. That’s the only kind of indeterminateness that can make sense here. If we restrict the quantifi er to tables (or “familiar objects”) we can correctly say, “It’s indeterminate whether there exists anything composed of the top and the leg,” meaning that it’s indeterminate whether the sum of the top and the leg is a table (or any other “familiar object”). But the quantifi er can be restricted in that manner only against the background of arbitrary mereological sums. The key assumption in Lewis’s argument is the principle mentioned earlier: If it’s indeterminate whether there exists something having a certain property, then there exists something such that it’s indeterminate whether it has that property. If not for this principle we can simply say that it’s indeterminate whether there exists something having the property of being composed of the top and the leg, though there isn’t anything such that it’s indeterminate whether it’s composed of the top and the leg. So why should we accept this principle? If we deny the principle then we are saying that a sentence of the form “There exists something that is F ” can be vague, not by virtue of the vagueness of the application of the term “F ” to things, but by virtue of the vagueness of the rest of the sentence, that is, by virtue of the vagueness of the quantifi er expression “there exists something.” Lewis’s argument depends on his denying the possibility that the quantifi er can be vague. Lewis is explicitly aware of this. And he tries to explain why the quantifi er can’t be vague as follows: “The only intelligible account of vagueness locates it in our thought and language. . . . Vagueness is semantic indecision. But not all of language is vague. The truth- functional connectives aren’t. . . . Nor are the idioms of quantifi cation, so long as they are unrestricted. How can any of these be vague? What would be the alternatives between which we haven’t chosen?”28

27. Lewis, On the Plurality of Worlds, pp. 212–213. 28. Ibid, p. 212. quantifier variance and realism 89

I accept Lewis’s assumption that vagueness is a matter of semantic indecision.29 I won’t say anything here about truth-functional connectives, but what is the problem Lewis is raising about the vagueness of the quantifi er? Since the meaning of a logical constant is given by its role in determining the truth-conditions of sentences, the vagueness of the quantifi er would consist in our semantic indecision with respect to the truth-conditions of certain sentences, for example, the sentence, “There exists something composed of the top and the leg.” With respect to a situation in which the top and the leg are borderline attached, we are undecided about whether to count the sentence as true or false. I can think of only one reason why Lewis rejects this explanation of the vagueness of the quantifi er. The explanation requires us to be able to make sense of there being two possible meanings for the quantifi er, the “precisifying” meaning that would make the sentence true and the one that would make it false. Quantifi er vagueness requires the possibility of quantifi er variance, and Lewis rejects the former because he rejects the latter.30

In this fi nal section I want to try to explain a bit further what I mean by commending a “shallow” approach to ontology. From the standpoint of shallow ontology Quine’s portentous notion of “ontological commitment” already conveys an unfortunate aura of theoretical hype and pseudo-depth. One of the most striking characteristics of “deep ontology’’ is a certain kind of maddening modesty and caution in its formulations. Whereas the shallow ontologist will address a typical question of ontology either by shrugging it off with Carnapian tolerance for many diff erent answers, or by insisting with Austinian glee that

29. But the argument I am about to go through would be unaff ected if we accepted instead Williamson’s epistemic view of vagueness. Lewis and Williamson agree that, for the quantifi er to be vague, it must admit of diff erent “interpretations.” The argument is unaff ected by whether these interpretations are taken to be diff erent possible “precisifi cations,” as in Lewis, or diff erent guesses (which we can’t verify) about the one correct interpretation, as in Williamson. See Timothy Williamson, Vagueness (Routledge, London, 1994), especially pp. 164, 237, 257–258. 30. An argument similar to Lewis’s, but somewhat more diffi cult to unravel, stems from the vagueness of identity sentences. Such an argument is given in Sydney Shoemaker, “On What There Are,” Philosophical Topics, 16 (1988). I try to show why that argument doesn’t work, appealing in eff ect to quantifi er variance, in my “The Vagueness of Identity,” Philosophical Topics, 26 (1999). 90 quantifier variance and realism: essays in metaontology the answer is laughably trivial, the deep ontologist will tend to treat the question with the kind of somber speculative anxiety appropriate to matters of high theory. “All things considered, I am tentatively inclined to be ontologically committed to apple trees but not to apples.”31 The challenge for the shallow ontologist posed by this kind of formulation is how to keep a polite straight face while listening to it. Even from the shallow perspective not every ontological question— within the domain of highly visible physical objects—can be quickly answered. Here is a good philosophical experiment: Look at your hand while you are clenching it, and ask yourself whether some object called a fi st has come into existence. As shallow ontologists the rstfi thought that must come to mind when we ask this question is this: There can’t be anything deep or theoretical here. The facts are, so to speak, right in front of our eyes. Our task can only be to remind ourselves of relevant ways in which we describe these facts in our language. We might consider, for instance, comparisons and contrasts between how we talk about “making a table” and “making a fi st.” (For example, a table can be made and can be destroyed, but do we talk about “destroying a fi st?”) Our task is to “command a clear view of the use of our words,” as Wittgenstein put it,32 that is, a clear view of how the relevant concepts operate. We’re engaged in what Austin called “linguistic phenomenology,” which need not always be an easy task.33 Finally—and perhaps most importantly— we are open to the vagueness of “existence”: The best answer might turn out to be that a fi st sort of comes into existence and sort of doesn’t. The deep ontologist approaches this question in a very diff erent spirit. She is engaged in the highly theoretical enterprise of deciding whether her “ontological commitments” should include fi sts that come into existence when hands are clenched. She will anguish over the “theoretical price” of having such things or not having them. She might eventually venture her best theoretical conjecture about whether fi sts come into existence. Or perhaps, after furrowing her brow for an appropriate period, she will simply announce that it “darkens the understanding” to suppose that fi sts keep popping in and out of existence.

31. For the uninitiated, who may possibly think that at least in this example I am tilting at wind- mills, the example actually expresses the position defended in Peter van Inwagen’s extremely infl uential Material Beings (Cornell University Press, Ithaca, N.Y., 1990). 32. Philosophical Investigations, I, 122. 33. J. L. Austin, “A Plea for Excuses,” in Philosophical Papers (Oxford University Press, London, 1961). It goes without saying that we need not be sympathetic to everything that went under the name “ordinary language philosophy.” quantifier variance and realism 91

Do the deep ontologists care about what common sense says? Sometimes they claim to be trying to balance the demands of common sense against the demands of philosophical theory.34 The trouble is that theory seems always to win. When I consider the writings of some of the most prominent deep ontologists of recent years— Chisholm, Lewis, van Inwagen, just to mention three—I can rattle off many cases in which they veto a commonsensical judgment in behalf of a philosophical argument, but I would be hard pressed to recall an example in which the reverse happens. My impression is that, in matters of ontology, virtually any theoretical problem, however marginal or fl imsy, if it cannot be adequately answered, suffi ces, by the lights of these philosophers, to trump the most deeply entrenched beliefs of common sense. For a shallow ontologist like me, the opposite is the case. Given any well entrenched ontological judgment of common sense (about highly visible physical objects), I could not imagine giving it up for the sake of some philosophical argument. If I had nothing more defi nite to say about the argument I would simply repeat Moore’s famous point that the force of the common sense judgment shows that there must be something wrong with the argument (even if I don’t know what it is).35 I am trying to roughly characterize two ways of approaching an ontological question, what I am calling the “shallow” and “deep” ways. This diff erence is no doubt partly a matter of intellectual temperament and style, but there is more to it than that. I’m especially interested here in how it relates to the issue of quantifi er variance. Let me try to sketch now a general model that shows that common sense judgments of ontology must be taken very seriously. The model, it will be seen, presupposes the possibility of quantifi er variance. Suppose that two philosophers named Xstein and Ystein are engaged in an ontological dispute. I assume that both philosophers claim to be speaking plain English. We might attempt a certain strategy of semantic ascent. Perhaps both philosophers can be gotten to agree that there are two possible languages, call them Xglish and Yglish, such that every sentence asserted by Xstein is true in Xglish and every sentence asserted

34. Lewis often presents his methodology as involving this kind of balancing. See, for example, Philosophical Papers, vol. I (Oxford University Press, N.Y., 1983), p. x, and On the Plurality of Worlds, pp. 134–135. 35. See G. E. Moore, “A Defense of Common Sense,” p. 41, and “Four Forms of Scepticism,” p. 222, both in Philosophical Papers (Collier Books, N.Y., 1962). See also Ned Markosian, “Brutal Composition,” Philosophical Studies 92 (1998), 211–249. 92 quantifier variance and realism: essays in metaontology by Ystein is true in Yglish. Each philosopher agrees that he can understand what the truth-conditions are of the sentences in both languages, and that, in terms of those truth-conditions, he could express in either language everything (every unstructured fact) he wants to assert. In these circumstances, if it were known that Xstein is speaking Xglish and Ystein is speaking Yglish, then these philosophers could simply agree that both are speaking the truth in their respective languages. Of course, this happy result cannot be quite right, since each philosopher claims to be speaking English, which can’t be both Xglish and Yglish. Now suppose, further, that typical fl uent speakers of English assert the sentences that Xstein asserts and reject the sentences that Ystein asserts. In other words, they assert sentences that are true in Xglish but false in Yglish. This would seem to be overwhelming evidence that the language they speak is Xglish, not Yglish. As usual, I am focused on sentences having to do with highly visible physical objects. By any reasonable standards of interpretation—for example, the standards explained by Lewis36—we ought to say that these people are uttering truths in Xglish, rather than that they are uttering falsehoods in Yglish. We have concluded that English is Xglish. We have, therefore, also concluded that Xstein is right and Ystein is wrong. Since they both claim to be speaking English, it is Xstein, not Ystein, who is saying the truth in English. It appears that Ystein has allowed “language to go on holiday”; Ystein’s philosophical theorizing has apparently confused him into in eff ect slipping into a language of his own creation, the language Yglish. Of the lamentable Ystein we must say, following Wittgenstein, that his theorizing has led him to “have a new conception and interpret it as seeing a new object [or seeing the absence of an old object].”37 As an illustration of this model: Let Lewish be the language in which all of David Lewis’s ontological utterances (with respect to highly visible physical objects) are true; let Inwagish be the language in which all of Peter van Inwagen’s ontological utterances are true; and let Shmenglish be the language in which the typical ontological utterances of non-philosophers are true. (By “true” I always mean “true in the strictest and most philosophical sense,” whatever that is supposed to mean; and I am always assuming an unrestricted use of the quantifi er.) The sentence “There exists a brown wooden thing in the yard that keeps gaining and losing branches” is true in Lewish but false in both

36. See his “Radical Interpretation,” Philosophical Papers, Vol. I. 37. Philosophical Investigations, I, 401. quantifier variance and realism 93

Inwagish and Shmenglish. The sentence “There exists apples” is true in both Lewish and Shmenglish but false in Inwagish. And so on. I expect the reader to be able to fi ll in the truth-conditions of the indefi nite number of sentences in these three languages. Both Lewish and Inwagish correspond to Yglish, and Shmenglish corresponds to Xglish. Both Lewis and van Inwagen correspond to Ystein, and I—the shallow ontologist, who simply asserts the same sentences that are asserted by ordinary speakers of English—correspond to Xstein. The argument proceeds as before. Since the typical fl uent speaker of English asserts sentences that are true in Shmenglish, but false in both Lewish and Inwagish, the evidence seems to be overwhelming that English is Shmenglish, not either Lewish or Inwagish. Therefore, I am right, and both Lewis and van Inwagen are wrong. Only my ontological sentences are true in English (= Shmenglish); theirs are not. Their sentences are true in either Lewish or Inwagish, but not in English, which is the language they both claim to be speaking. It will be objected that something essential has been left out of my model. What about the theoretical arguments these philosophers give for their positions? Why haven’t those arguments been taken into account? But how can they be taken into account? That’s my main point here. Those arguments have no obvious bearing on what has now become the critical question: whether English is Shmenglish. Remember, if English is Shmenglish, then it follows directly that both Lewis and van Inwagen are mistaken; that simply follows from the way Shmenglish has been defi ned. It appears now that all of the deep ontologist’s theoretical maneuvers have no relevance to rebutting the seemingly obvious fact that English is Shmenglish. I’ve already considered Lewis’s argument about the vagueness of the quantifi er. As another and more typical ontological argument, let’s briefl y consider Lewis’s reason for believing in temporal parts. Lewis’s argument derives from a problem about what he calls “temporary intrinsics.”38 The problem, roughly put, is as follows. If an object can have a shape at one moment and lose it at another moment, as common sense says, then the shape is being treated logically and semantically as if it were a relation binding the object to a time. But shapes are intrinsic properties, not relations. The only way to hold onto the intrinsicness of shapes is to claim, contrary to common

38. Lewis, On the Plurality of Worlds, pp. 202–204. 94 quantifier variance and realism: essays in metaontology sense, that what has the shape is a temporal part of the object, and this temporal part never loses the shape. Now I don’t myself see this as a genuinely worrisome problem.39 But I don’t want to enter into the details of this issue. My point now is something else. How does Lewis’s problem of temporary intrinsics bear on the question whether English is Shmenglish? Remember, once again, that if English is Shmenglish then it follows that Lewis is wrong in asserting (in English) that there are temporal parts. It is given that the following sentence, which Lewis holds to be false in English, but to which typical speakers of English assent, is true in Shmenglish: “A bent wooden stick can sometimes be straightened without any wooden thing going out of existence.” In order for the problem of temporary intrinsics to show that English is not Shmenglish the problem would somehow have to be refl ected in the speech behavior of typical uentfl speakers of English, perhaps by their being disposed to assent to the sentence, “Shapes are intrinsic properties, and an intrinsic property cannot be treated logically and semantically as if it were a relation binding an object to a time.” In all likelihood, only typical fl uent speakers of English with philosophy degrees from Princeton would have any strong tendency to assent to this sentence. But, furthermore, whatever evidence Lewis might be able to marshal from the speech behavior of typical speakers of English vis-à-vis the problem of temporary intrinsics, this would have to be capable of defeating the overwhelming evidence that English is Shmenglish that derives from the unambiguous disposition of typical speakers to assent to countless sentences such as, “A bent wooden stick can sometimes be straightened without any wooden thing (of any sort whatever, whether familiar or not, interesting or not) going out of existence.” I cannot see how Lewis could argue successfully that English is not Shmenglish, let alone argue successfully that English is Lewish. I am aware that my argument from the Xstein-Ystein model requires various refi nements—I intend to provide these elsewhere— but I think that the argument even in its sketchiest form indicates that there is something fundamentally incoherent in the attempt of philosophers to deny basic and seemingly trivial ontological beliefs of common sense. The point at which my argument depends on quantifi er variance is in its assumption that we have these three possible languages, Lewish,

39. Because to say that a shape is intrinsic might mean intuitively that a thing’s having a certain shape at a certain time does not depend on how it is related to other things at that time. quantifier variance and realism 95

Inwagish, and Shmenglish, each with its own quantifi cational apparatus and concept of “the existence of a thing.” A philosopher might try to block the argument by holding that Shmenglish is not a possible language. The burden would fall on this philosopher to explain why the seemingly most obvious interpretation of the language we speak is not a possibility. I don’t see how this burden can be met.40

40. In Dividing Reality I sought to uncover some general constraints that would make it impossible for a natural language to diverge (too much) from the taken-at-face-value semantic structure of English. My conclusion in that book was that no interesting constraints seem to be defensible. Had the sought-after constraints emerged, although they would certainly have allowed some forms of quantifi er variance (in particular, the quantifi er variance involved in quantifi er vagueness, and that involved in sortal-dependence), they may have implied that either Lewish or Inwagish is not a possible language. An argument for the impossibility of Shmenglish would have to move in the opposite direction: to show that the taken-at-face-value semantic structure of English is not a possibility. In responding to an earlier version of my argument from Shmenglish a few years ago, Sider in fact held that Shmenglish is not a possible language (because, very roughly, its quantifi er fails to correspond to the world’s logical joints). (See, further, the Introduction to Theodore Sider, Four Dimensionalism: An Ontology of Persistence and Time [Oxford University Press, N.Y., 2001].) Will that, I wonder, be the deep ontologist’s standard response to my question? six c C

against revisionary ontology

My impression is that the period of the last twenty years or so has been an amazingly bad time for common sense ontology. It seems to me that every third philosopher I talk to has serious doubts about whether tables exist, and of the other two at least one of them thinks that if there are such things as tables then there are also such things as the object composed of the Eiff el Tower and Plato’s nose. As I overheard a philosopher saying recently, “Either everything exists or nothing exists.” The “everything exists” people are sometimes called “universalists.” They hold that for any two things there is a third thing composed of the two. Typically they also hold that if a thing persists through a period of time then there is a second thing, a “temporal part” of the fi rst, which exists only during that period. The world consists, then, of all of the sums of these temporal parts, so that, as Quine often puts it, every materially occupied space-time portion of the world, however discontinuous or gerrymandered, is an object on an equal footing with any other.1 The “nothing exists” people—perhaps Peter Unger is the most famous recent example (or would

1. Universalists who accept temporal parts are often called “four-dimensionalists.” Examples are W. V. Quine, Word and Object (Cambridge, Mass.: MIT Press, 1960), esp. 171; David Lewis, On the Plurality of Worlds (Oxford: Basil Blackwell, 1986), esp. 202–4, 211–13; Mark Heller, The Ontology of Physical Objects: Four-Dimensional Hunks of Matter (Cambridge: Cambridge University Press, 1990); ch. 1 reprinted in J. Kim and E. Sosa, eds., Metaphysics: An Anthology (Oxford: Blackwell,1999); Theodore Sider, Four-Dimensionalism: An Ontology of Persistence and Time (New York: Oxford University Press,

96 against revisionary ontology 97 be if he existed)—typically hold that none of the complex entities that people ordinarily seem to be talking about really exist, though they may possibly allow for the existence of metaphysical simples.2 Even more amazing in a way than the anti-commonsensical stances of the universalists and nihilists is the view of people like Chisholm and van Inwagen who, while agreeing with common sense that the truth lies somewhere between “everything exists” and “nothing exists,” claim to have some way of picking and choosing amongst ordinary things, admitting some of them but dismissing others, in a manner I can’t help associating with the game played by small children who demand from you an invisible ticket before they’ll admit you into their room.3 These attacks on common sense amaze me in part because I think they are so badly misguided, but also because when I entered the pro- fession of philosophy during the heyday of the so-called ordinary language movement, I think almost no one would have predicted that before the end of the millennium—even given some predictable end- of-millennium madness—the existence of tables would again be called into question. I think most of us assumed back then that philosophers like Moore, P. F. Strawson, Austin, and Wittgenstein had once and for all established the undeniability of our most basic common sense beliefs—or, if not “once and for all,” at least for more than ten years. We were, as it now appears, overly optimistic. Looking back at those heroes of common sense we can draw a rough distinction, I think, between two general approaches. The fi rst

2002). Universalists who reject temporal parts are rare, but include Judith Jarvis Thomson, “Part- hood and Identity across Time,” Journal of Philosophy 80 (1983): 201-20; reprinted in Kim and Sosa, eds., Metaphysics: An Anthology, and James van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time,” Midwest Studies in Philosophy XI: Studies in Essentialism (1986). An argument from universalism to temporal parts is given in Richard Cartwright, “Scattered Objects,” in Analysis and Metaphysics, ed. K. Lehrer (Dordrecht: Reidel, 1975); reprinted in Kim and Sosa, eds., Metaphysics: An Anthology. 2. Peter Unger, “There Are No Ordinary Things,” Synthese 41 (1980): 117–54, and “I Do Not Exist,” in Perception and Identity: Essays Presented to A. J. Ayer with His Replies to Them, ed. G. F. Macdonald (New York: Macmillan, 1979). 3. Roderick Chisholm is a “mereological essentialist” who holds that the only things that exist are things that have the same parts from moment to moment (and from world to world); see Person and Object (LaSalle, Ill.: Open Court, 1976); ch. 3 reprinted in Kim and Sosa, eds., Metaphysics: An Anthology. This position is defended further in van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time.” Peter van Inwagen believes that the only composite things that exist are living things; see Material Beings (Ithaca, N.Y.: Cornell University Press, 1990). A variation of this position, defended with diff erent (and clearer) arguments, is given in Trenton Merricks, Objects and Persons (Oxford: Oxford University Press, 2001). 98 quantifier variance and realism: essays in metaontology approach, typifi ed by Moore, comes out of basic epistemology. A simple point that Moore made in a number of his papers is that our common sense convictions have more epistemic weight than any fancy philosophical arguments. Suppose you think you have an argument against the existence of tables. Maybe there is some mistake in the argument that you haven’t seen. That’s possible, isn’t it? What is more likely, that there is a mistake in the argument or that there aren’t any tables? Moore thought that any sane person who considers that question would soon realize that it is more reasonable to abandon the argument than to abandon tables.4 The second approach, which one is likely to associate with the name “ordinary language philosophy,” comes out of linguistic rather than epistemological considerations. In arguing against revisionary ontology in the present paper I’m going to develop a version of this second approach. This is not because I think that the more simple Moorean appeal to basic epistemology is insuffi cient. It ought to suffi ce, but the trouble is that revisionists tend to be unmoved by it. The revisionary literature often gives the impression that revisionists are people who have heard rumors about the existence of tables at family aff airs, and apart from a natural reluctance to off end their non- philosophical relatives, they themselves have no initial intuitive feel- ings one way or another about whether tables might exist. The simple Moorean appeal to the convictions of sane common sense, therefore, tends to be shrugged off by these philosophers. The argument from language that I’m going to develop will, I hope, be harder for them to ignore. The basic argument can be summarily put as follows. It is widely acknowledged that a central constraint on interpreting a language is a “principle of charity.”5 Suppose we have two candidate interpretations for a set of sentences that fl uent speakers of a language would typically be prepared to assert (or assent to). If one of these interpretations im plies that the speakers are correct in asserting these sentences, and the other interpretation implies that they are incorrect, then the

4. G. E. Moore, “A Defense of Common Sense,” esp. 41, and “Four Forms of Scepticism,” esp. 222, both in Philosophical Papers (New York: Collier Books, 1962). A Moorean argument against revisionary ontology is interestingly developed in Ned Markosian, “Brutal Composition,” Philosophical Studies 92 (1988): 211–49. 5. Quine, Word and Object, 59; Donald Davidson, Inquiries into Truth and Interpretation (London: Oxford University Press, 1984), esp. “Belief and the Basis of Meaning” and “Thought and Talk”; David Lewis, “Rad- ical Interpretation,” in Philosophical Papers I (New York: Oxford University Press, 1983), and “New Work for a Theory of Universals,” Australasian Journal of Philosophy 61 (1983): 343–77, at 370–77. against revisionary ontology 99

principle of charity imposes a presumption in favor of the fi rst interpretation. Revisionists seem to fl out this presumption, overriding it for no apparently good reason. Given a set of commonsensical ontological assertions, revisionists seem to perversely insist on interpreting these assertions in a way that makes them incorrect, even though it seems perfectly possible to interpret them in a way that would make them correct. I’ll refer to the above argument as “the argument from charity.” This argument is one basis—not the only basis—for explaining why there is always a strong (if defeasible) presumption in favor of common sense. In evaluating this presumption I want to place a great deal of emphasis on a certain distinction. The beliefs expressed by typical speakers of a language may be attacked in two diff erent ways. One way is to claim that these beliefs are empirically false. For example, typical speakers of English may at one time have held the empirically false beliefs that the Earth is fl at and that whales are sh.fi The second way is to claim that these beliefs are false on a priori conceptual grounds. I take the second way to be characteristic of revisionary or anti-commonsensical ontology. A nihilist like Unger and a mereological essentialist like Chisholm think that it is an a priori mistake to claim that there are tables (i.e., composite tables that persist through mereological changes), and universalists like Lewis think it is an a priori mistake to deny that the Eiff el Tower and Plato’s nose compose an object. Quine may not fi t into this, if he completely rejects the distinction between the empirical and the a priori. But I assume that distinction in all of my work, as well as in the present paper, and the revisionary ontologists that most concern me assume it too.6 The reason why this distinction is important to the argument from charity is that it is generally understood that charity in linguistic interpretation has more to do with rationality—with good reasons—than with truth. People are expected to make empirical mistakes if their sensory data are limited—it may in fact be positively irrational for them not to make such mistakes—so that an interpretation of language that ascribes such mistakes to people may be compatible with, or even required by, the principle of interpretive charity. Revisionary ontology, however, as this is understood in the present discussion, has nothing to

6. Within the context of this discussion I don’t think it’s important to worry about distinctions— coming out of Kripke’s work—between “a priori necessary,” “conceptually necessary,” “epistemically necessary,” “metaphysically necessary.” Although it’s the fi rst two that are most crucial to my argument, I think that the typical revisionist is claiming that common sense beliefs are necessarily false in all these ways. 100 quantifier variance and realism: essays in metaontology do with sensory data. If revisionists interpret the ontological assertions of common sense as a priori necessarily false then, assuming there are other available interpretations, this does prima facie violate the principle of charity, since people are not normally thought to have good reason to assert what is a priori necessarily false. The presumption that typical fl uent speakers of a language do not make assertions that are a priori necessarily false is safer—because it demands less—than the more general Moorean presumption that the assertions are true. The central problem for revisionists is that they seem to fl out even this presumption. Consider the universalist’s claim that any two things, no matter how related (even if not contemporaneous, I’m assuming), make up a thing.7 When I have presented this position to ordinary people I have sometimes gotten the following response: “These philosophers don’t mean by ‘a (one) thing’ what we ordinarily mean.” That response seems to me essentially right. By ignoring the obvious—that is, the charitable—interpretation of what the words “a (one) thing” mean in English, these philosophers are merely “abusing the language.” Of course they are entitled to introduce a technical language if they wish, but then they have to say that, and not pretend that they are expressing in plain English a substantive and controversial philosophical discovery. To be a bit more accurate, it is not the expression “a thing” taken by itself that the universalist misuses, but that expression in such larger contexts as “a thing that is made of fl esh at one time and made of metal at a later time.” On any reasonably charitable interpretation of the English language that expression doesn’t refer to anything, or at least it doesn’t refer to anything merely by virtue of a nose existing at one time and a tower existing at a later time. This is because, on any reasonably charitable interpretation, an English expression of the form “the thing that is made up of the F and the G” does not refer to anything unless the things referred to by “the F ” and “the G” are connected or united in some relevantly special way. If we compare the universalist’s “fusions” or “sums” with sets, the essential contrast, from the point of view of the present discussion, is that typical speakers of English have no disposition to deny standard assertions about sets. To the extent that the language of set theory is technical, speakers unschooled in this language may be apt to greet an assertion about sets with incomprehension

7. Thomson has urged in conversation that ordinary people quite easily accept portions of matter, even, for example, the scattered portion of matter that makes up Clinton’s nose and the Eiff el Tower. That may be. But I take it for granted that the universalist’s “fusions” are not limited to contemporaneous portions of matter. against revisionary ontology 101

(“What do you mean by ‘a set having a member’?”), but certainly there is no tendency to deny the assertion. By contrast, typical speakers of English will regard as sheer lunacy the following assertion: “Something (some physical object) was made of fl esh thousands of years ago and it (that same thing) is now made of metal (by virtue of Plato’s nose and the Eiff el Tower).” To repeat what I said a moment ago, if universalists intend to introduce a technical language—so that the sentence just mentioned is not to be understood in plain English—then they are not revisionary ontologists in the sense that concerns me, and I have no argument with them (nor, obviously, do they have any argument with common sense).8 To spell out a bit further what I mean here by a revisionary ontologist:

Revisionary ontology. Many common sense judgments about the existence or identity of highly visible physical objects are a priori necessarily false. Perhaps there are common sense judgments about the existence or identity of things other than highly visible physical objects, such as numbers or properties. If so, they are not the topic of this discussion. I am dealing here only with an area of ontology that might be roughly called (in honor of Austin) “the ontology of moderate-sized dry goods.”9 Revisionists, in the present sense, hold that ordinary people make mistakes in their judgments about the existence or identity of the physical objects they claim to perceive in front of them, and not just mistakes, but a priori necessary mistakes. Revisionists hold this because they misinterpret the language, or so the argument from charity says. A reaction that I have often gotten to this argument is that it is a form of linguistic idealism. “We are doing ontology, not linguistics. We are considering substantive questions about what exists in the world, and if you think that such questions can be settled by an examination of the English language then you must hold the most non- commonsensical view of all, namely, the view that language determines what exists in the world.” This objection is off the mark. If we are speaking English then to know the truth-value of the sentence “There are tables” is to know whether there are tables. If we have decided that

8. Cf. my papers “Quantifi er Variance and Realism,” Philosophical Issues 12 (2002): 55–56 and “Sosa’s Existential Relativism” in Sosa and His Critics (Oxford: Blackwell, 2004). 9. J. L. Austin, Sense and Sensibilia (London: Oxford University Press, 1962), 8. 102 quantifier variance and realism: essays in metaontology on the most charitable and reasonable interpretation that sentence is true, then we have decided that: there are tables. This has nothing to do with linguistic idealism. It simply follows from the trivial disquotational constraint that (assuming “p” is a sentence in the language I am now speaking) “p” is true if and only if it is the case that p. I say that, on the most charitable and reasonable interpretation of the English sentence “There is something that was made of fl esh at one time and made of metal at a later time,” the sentence is not made true by the existence of Plato’s nose followed by the existence of the Eiff el Tower. In saying this, since I am now speaking (plain, nontechnical) English, I have concluded that there is not (by virtue of Plato’s nose and the Eiff el Tower) anything that was made of fl esh at one time and made of metal at a later time. My move from the remark about language to the remark about what exists does not commit me to the absurd view that what exists depends upon language. It rather commits me to being responsible to the language that I speak, even—or, I should say, especially—when I am doing philosophy. “But how,” it may still be puzzled, “can an appeal to anything as shallow as ordinary English usage resolve deep issues about what exists in the world?” The answer is that it can’t. The issues being debated by revisionists are not deep; they are completely trivial. That’s the key to all of this. Revisionists display to the highest degree the philosophical syndrome Wittgenstein called “language gone on holiday.”10 Because they have lost their grip on language they “have a new conception and interpret it as seeing a new object [or the absence of an old object].” 11 Revisionists suff er from the illusion that certain questions are philosophically deep, inviting complicated theoretical debates, when in fact these questions are comically trivial. “Is it possible for a table to exist?,” “Is it possible for a car to survive the change of a tire?,” “Is it possible for two things not to make up a third thing?”—the only sensible response to such questions is, “Of course, what on Earth are you talking about?” To dismiss the questions in that manner, to treat them as utterly trivial, is part of what is involved in understanding the words and concepts that enter into them. The argument from charity is a last- ditch eff ort to bring the revisionists back to their senses, that is, to bring them back to the language that they themselves claim to be using, so that they can recognize utter triviality when it stares them in the face.

10. Ludwig Wittgenstein, Philosophical Investigations (Oxford: Blackwell, 1953), I, 38. 11. Wittgenstein, Philosophical Investigations, I, 401. against revisionary ontology 103

The argument from charity can be schematically formulated as follows, where “O” is some ontological sentence affi rmed by ordinary people but denied by some revisionists:

1. Typical fl uent speakers of the language assert (or assent to) the sentence “O.” 2. Therefore, there is the charitable presumption that, on the correct interpretation of “O,” speakers have good reason to assert “O,” so that “O” is not a priori necessarily false. 3. There is nothing to defeat this presumption. 4. Therefore, “O” is not a priori necessarily false. 5. Therefore, it’s possible that O. [6. Therefore, it’s actually the case that O.]

Let me fi rst comment on the bracketed line 6. A philosopher might reject 6 on empirical grounds. Such a philosopher, who thinks that O’s truth is possible but not actual, would not be a revisionist in the present sense. Revisionists, who think that O’s truth is impossible, will typically agree, however, that if they are wrong about that, then common sense is right, and O is not merely possibly true but actually true. Van Inwagen, for instance, would probably agree that, if it’s possible for a table to exist (as a composite nonliving thing), then a table does in fact exist. Why would he want to deny that tables in fact exist once he has accepted their possibility? Again, I’m confi dent that the universalist Lewis would agree that, if it’s possible for there to be things that don’t compose anything, then there are in fact many cases of such things. The move from 5 to 6 should actually be carefully examined, but I will not undertake that here. My general assumption—for expository reasons, if nothing else—will be that, if we have refuted the revisionist’s claim that common sense ontological judgments are a priori necessarily false, then we have established the truth of those common sense judgments. In presenting the argument from charity I assume—subject to further discussion—that revisionists who deny commonsensical assertions of some sentence “O” will admit that they are able to make intelligible to themselves a prima facie charitable interpretation on which “O” is true. By an “interpretation” I simply mean the assignment of truth conditions to “O.” This need not involve formulating a sentence that is synonymous with “O,” nor even a sentence that has strictly the same truth conditions as “O.” All that is required is that one 104 quantifier variance and realism: essays in metaontology be able to indicate in some—perhaps rough and sketchy—manner what are the possible situations or worlds with respect to which “O” (as uttered in some specifi ed kind of context) holds true. Nihilists and various quasi-nihilists, for instance, deny that there are tables, but agree that there are bits of matter that are table-wise interrelated.12 It is indeed the latter condition that, according to them, somehow induces ordinary people to make the mistake of thinking that there are tables. The question posed by the argument from charity is why these philosophers don’t interpret the English sentence “There are tables” as holding true whenever that condition obtains. Again, universalists will admit that they can make intelligible to themselves a possible use of the sentence “Nothing was fi rst made of fl esh and then made of metal” which has as its truth conditions that nothing made of fl esh was connected or united in some relevantly special way to something later made of metal. The question posed by the argument from charity is why not interpret that English sentence as having such truth conditions. Starting now from the beginning of the argument: 1 reports a fact about people’s linguistic propensities. 2 comes out of the widely accepted principle of interpretive charity. If 3 is also accepted then it follows that 4 should be accepted: if we assert that there is a presumption in favor of something’s being the case, then we should assert that it is the case as soon as we agree that there is nothing to defeat the presumption.13 The move from the meta-level claim 4 to the object- level claim 5, I have already defended. If there is a hole here I think it is in 3. Revisionists might claim that the presumption is defeated in some way. I’ll presently examine, and criticize, several ways in which this claim might be formulated. Before turning to this, however, there is a complication that has to be addressed. There are some philosophers who sound like revisionists but who insist that they are not really disagreeing with ordinary people about anything. I will call them, with due respect, crypto-revisionists. I think this maneuver never works and only confuses things further. The most familiar form of the maneuver, found in Butler and more recently in Chisholm, is to claim that the assertions standardly made by

12. This is van Inwagen’s famous formulation, which has been taken over by many revisionists. 13. This entire paper hovers in the vicinity of, but does not enter into, issues about the ultimate nature of semantic content, especially the issues discussed in Saul A. Kripke, Wittgenstein on Rules and Private Languages (Cambridge, Mass.: Harvard University Press, 1982). Do 2 and 3 strictly entail 4 or is the connection in some sense probabilistic? I leave that open. It seems clear, however, that anyone who asserts 2 and 3 should assert 4. against revisionary ontology 105 ordinary folk are perfectly correct in the “loose” sense intended, though philosophers need to realize that these assertions are “strictly” false. Since ordinary people intend the statements only loosely, the philosophers have no disagreement with them.14 Now I don’t doubt that the distinction between “loose talk” and “strict talk” is potentially of interest to philosophy, and that the distinction may not always be easy to clarify, but the trouble is that Butler and Chisholm seem to make intuitive nonsense of the distinction. The distinction between loose and strict talk must surely be based on the linguistic practices of fl uent speakers of the language. Three obvious criteria to determine whether an assertion is strict or loose are: (a) Ask the speakers if they mean the thing strictly or loosely; (b) See if the speakers will make the assertion in a “strict” context, that is, a context calling for careful and thoughtful formulations—such as in a court of law or a formal document; (c) See if the speakers withdraw the assertion when reminded of relevant information. The following is a genuine example of loose talk.15 A and B have often bemoaned the fact that since C left town she calls very rarely. On this occasion A says, “She never calls.” Obviously A means this only loosely. The proof is that, fi rst, if you asked her whether she means that, strictly speaking, C never calls she would say no; second, A would not say under oath in court, “C never calls”; third, if B says to A, “Well, she does call once in a while,” A doesn’t say, “That’s right, she never calls.” Imagine now that a witness for the defense swears that the car involved in the accident was the same one that he later parked in the garage. “Are we getting the strict truth from you? You cannot be claiming that it is the same one—can you, now?—when you yourself have testifi ed that a hubcap fell off on the way to the garage!,” triumphantly cries the prosecuting attorney Chisholm. No ordinary person, even under oath, will be induced to reconsider the claim that it was the same car merely by being reminded that the hubcap fell off . (It would be question-begging to appeal to what some people might be induced to say by—what I am claiming is—bad philosophy.) The Butler-Chisholm approach seems especially implausible with respect to a statement like, “This tree (is the same one that) had more branches twenty years ago,” which mentions the very fact about mereological change that is alleged

14. Joseph Butler, “Of Personal Identity,” in The Whole Works of Joseph Butler, LL.D. (London: Thomas Tegg, 1836), 263–70; Chisholm, Person and Object. 15. The example was off ered to me by Donald Baxter to help me to understand what people like Butler and Chisholm are talking about, but I use it to the opposite eff ect. 106 quantifier variance and realism: essays in metaontology to make the statement loose. This would be analogous to A’s saying, “My friend C, who calls on holidays, never calls.” I don’t question that there may be hard cases in which it is unclear (or indeterminate) whether we are speaking strictly or loosely, but it seems obvious that, by any sensible criteria, the assertions ordinary people make about the existence and identity of physical objects are typically intended to be true in a perfectly strict sense. This does not imply, of course, that such assertions contain no elements of vagueness. Strictness in the sense of the present discussion must not be confused with precision or non-vagueness. The statement “Rockefeller was wealthy” is normally intended to be strictly true (rather than “loosely true but strictly false”), although “wealthy” is highly vague. I think many people even in the revisionary camp have come to accept that the “loose”–“strict” distinction is of little help to them.16 There is, however, another form that crypto-revisionism has recently taken. Lewis has a formulation of the universalist position which says that the ontological assertions of common sense are correct if the quantifi ers—such words as “something” and “anything”—are restricted roughly to ordinary or familiar things, but the assertions are false when the quantifi ers are interpreted as unrestricted.17 I think that Lewis himself did not mean to claim that universalists are therefore not in disagreement with common sense, but I have often heard other universalists make that claim. I think it is easy to show that the claim is false. Imagine that I hold up two pieces of wood, one white and one brown, and I ask some ordinary people whether they can see any wooden thing that is fi rst white and then brown. Of course they say no. Now the universalist—I mean here a “four-dimensionalist” universalist like Lewis, one who believes in temporal parts—believes that any early part of the white piece together with a later part of the brown piece add up to a wooden thing that is fi rst white and then brown. Might this belief be reconciled with what the ordinary folk say by supposing that they only meant—what is surely the case—that there is no ordinary or familiar thing that is fi rst white and then brown? Well, let’s ask them: “Now, you’re not supposed to restrict yourself to ordinary things, or

16. Closely related to this distinction, and equally obscure, is van Inwagen’s attempt to distinguish between what a sentence expresses “in the ordinary business of life” and what it expresses “in the philosophy room.” Van Inwagen, Material Beings, 98–107, and “Reply to Reviewers,” Philosophy and Phenomenological Research 53 (1993): 709–19, at 711. For a critique of van Inwagen’s crypto-revisionism, see Merricks, Objects and Persons, 162–70. 17. Lewis, On the Plurality of Worlds, 3 and 212–13. against revisionary ontology 107 familiar things, or anything like that. Take into account anything whatsoever no matter how peculiar a thing it is. Do you now see some wooden thing that is fi rst white and then brown?” They still answer “no.” I assume everyone agrees that that would be the outcome of the experiment. But let’s take it one step further, with a fi nal attempt at reconciliation: “Ah, you didn’t get the riddle. You see, you kept restricting your attention to just one sort of thing even though you were told not to. The correct answer is that any early part of this one together with any later part of the other one make up a wooden thing that is fi rst white and then brown.” Do they say, “Oh yes, how stupid of us! Give us another riddle.”? No, they throw you out the window. That there is a stark confl ict between common sense beliefs and the universalist’s position seems to me undeniable. I have spoken to philosophers who remain unconvinced. The trouble is, they say, that ordinary people just can’t unrestrict the quantifi er, no matter what you tell them. Ordinary people don’t really disagree with the four-dimensionalist, but in the previous example the quantifi er remained restricted to familiar things, or interesting things, or something of that sort. Does this even make sense? If ordinary people can’t unrestrict the quantifi er how could the so-called unrestricted quantifi er ever have become a part of the English language? If it is part of the English language then fl uent speakers must know how to use it. Imagine that someone says, as in one of Lewis’s examples, “There’s no beer,” meaning to restrict the quantifi er to beer in his fridge. I say to him, “But of course there is beer in the grocery store.” He replies, “No, there is no beer in the grocery store,” still restricting the quantifi er to beer in his fridge, and saying in eff ect, “There is no beer in my fridge in the grocery store.” I say, “But you’ll surely agree that there is beer outside your fridge?” “No,” he replies, “there is no beer outside my fridge,” still restricting the quantifi er to beer in his fridge, and saying in eff ect, “There is no beer in my fridge outside my fridge.” Finally he exclaims, “There is no beer of any sort whatever, anywhere in the world, that can be referred to by anyone!” and he says that because he has never before had occasion to quantify over beer outside his fridge, so he keeps restricting the quantifi er to beer in his fridge. That story, it seems to me, makes as much sense as the suggestion that ordinary people are not really disagreeing with the four-dimensionalists, but they just keep restricting the quantifi er to familiar and interesting objects even when other objects that they supposedly don’t reject are explicitly and emphatically brought to their attention. 108 quantifier variance and realism: essays in metaontology

III

Let me return to the main thread of the argument. I will henceforth assume that we are dealing, not with crypto-revisionists, but with outright revisionists, philosophers who think it is their job to correct the mistakes ordinary people make about the existence and identity of the moderate-sized dry goods in front of their eyes. I have suggested that the only hope revisionists have to rebut the argument from charity is to claim that there are some considerations that defeat the presumption in favor of an interpretation of the language that makes the ordinary assertions come out true. I will consider these potential defeaters under two headings: confl icts of charity and constraints beyond charity. A confl ict of charity occurs when typical speakers of the language are disposed to assert confl icting sets of sentences. If we adopt an interpretation that justifi es the fi rst set of assertions, this will make it impossible to justify the second set; and if we adopt an interpretation that justifi es the second set, this will make it impossible to justify the fi rst set. How do we then decide which interpretation is correct? To which set of assertions do we bestow our interpretive charity? The standard revisionist literature does not address the argument from charity, but it does often invoke various general principles that are claimed to refute commonsensical assertions about the existence and identity of objects. Insofar as some of these principles are themselves accepted by ordinary people we may have a confl ict of charity. To illustrate this idea let me consider a principle that shows up in a great deal of the revisionary literature, the principle that two things cannot wholly occupy the same place at the same time.18 It seems that ordinary people are inclined to accept this principle. The principle seems, however, to confl ict with various commonsensical assertions about existence and identity. If a sculptor sculpts a statue out of a lump of clay, ordinary people say that the statue but not the lump of clay has just come into existence. It follows (by Leibniz’s Law) that the statue and the lump of clay are two things, and they wholly occupy the same place, contrary to the “no-two-things-in-the-same-place” principle. There are numerous examples of this sort, in which the principle seems to confl ict with the ordinary person’s assertions about existence and identity. Here it seems that we have a confl ict of charity. We can

18. Revisionists from every camp have tried to capitalize on this principle, including Cartwright, van Cleve, van Inwagen, Heller, and Merricks. against revisionary ontology 109

interpret the English language in a way that makes the ordinary person’s assertion of the principle come out true and numerous ordinary assertions about the existence and identity of objects come out false, or we can interpret the language to the opposite eff ect. I think when the options are presented in that way it’s immediately clear that the revisionists have no leg to stand on. What possible reason could we have to accept the fi rst interpretation over the second? The most that one can imagine claiming is that there is some degree of indeterminateness here, so that the revisionists’ bizarre assertions are after all not determinately false (in plain English). But even that very modest claim—far more modest, obviously, than what revisionists want—cannot, I think, be sustained. The correct interpretation of English is most plausibly the second one, that is, the one that makes ordinary assertions about existence and identity come out true and the “no-two-things-in-the-same-place” principle come out false. Let me try to explain why this is so. It is a standard assumption in general discussions of the nature of language that the linchpin of language-learning and language-interpretation consists of examples, especially perceptual examples.19 Faced with two candidate interpretations, and a confl ict of charity, we must therefore choose the interpretation that does best in sustaining people’s assertions about examples, rather than the interpretation that sustains some general principles. General principles are made to be qualifi ed or refi ned in the face of counterexamples (that’s the very idea of a counterexample). This point is a commonplace in philosophy. A famously dramatic instance is the successive qualifi cations of the principles traditionally defi ning knowledge in the face of Gettier’s counterexamples. In a confl ict between accepted principles and accepted examples we normally hold onto the examples and qualify or refi ne the principles. So it should be with the “no-two-things-in-the-same-place” principle. Once we become aware of the standard counterexamples to the principle surely the natural move is to qualify the principle in some way so as to accommodate the counterexamples. There are several ways to do this. A famous qualifi cation off ered by Locke is to say that the principle applies only to things of the same sort, which takes care of at least most of the standard counterexamples.20 Another possibility is Lewis’s plausible suggestion that we sometimes “count by temporary

19. The role of examples in linguistic interpretation is especially emphasized in Tyler Burge, “Intellectual Norms and Foundations of Mind,” Journal of Philosophy 83 (1986): 697–720. 20. John Locke, An Essay concerning Human Understanding (1690), book 2, ch.27. 110 quantifier variance and realism: essays in metaontology identity.”21 The statue and lump of clay are indeed two things in the Leibnizian sense (i.e., in the sense of “identity” satisfying Leibniz’s Law), but the “no-two-things-in-the-same-place” principle remains true in the sense of saying that if what are two things in the Leibnizian sense occupy the same place then they are “temporarily one thing,” meaning roughly that they temporarily share all of their properties. In the next section I will give additional reasons—besides the primacy of examples—for accepting the interpretation that sustains ordinary ontological assertions while qualifying the “no-two-things- in-the same-place” principle. Consider, however, a revisionist who has no wish to contest my ideas about linguistic interpretation but who responds to what I have been saying as follows: “The problem is that to my mind it seems completely clear that two things cannot wholly occupy the same place at the same time. What could make them two things (or, worse, two sorts of things) if they share the same (minute) parts? That merely darkens my understanding. In this case, therefore, my concerns about ordinary usage and correct linguistic interpretation must defer to my deep metaphysical intuitions.” This “deep” response, however, misses the essential point of the resolute shallowness of my argument. Perhaps it will help if I formulate the argument in a slightly diff erent way. I am claiming that the correct interpretation of the English language assigns truth conditions to sentences in a way that makes the ontological sentences typically asserted by ordinary people come out true. My assumption at present (subject, as I said, to further discussion) is that the revisionists will admit that they understand well enough the interpretation I have in mind, but they don’t think that is the correct interpretation. Let me give the name Shmenglish to a hypothetical language for which it is stipulated that this is the correct interpretation. Shmenglish sentences are the same (phonetically and syntactically) as English sentences, but it is stipulated that Shmenglish sentences have the truth conditions that I (perhaps mistakenly) take English sentences to have. On my present assumption revisionists will admit to understanding well enough how Shmenglish operates. Quasi- nihilists like van Inwagen and Merricks, for instance, will understand that the Shmenglish sentence “A statue was created from a lump of clay” counts as true with respect to any situation in which a continuous succession of (atomic) bits of matter arranged lump-of-clay-wise

21. David Lewis, “Survival and Identity,” in The Identities of Persons, ed. A. Rorty (Berkeley: University of California Press, 1976), 26–27. against revisionary ontology 111

temporarily coincides with a succession of bits of matter arranged statue-wise—or something roughly to that eff ect. If the people presently living in North America were speaking Shmenglish then, as even revisionists would admit, their assertions of such sentences as the one just mentioned would be correct. Consider now the following little argument: 1. In Shmenglish the sentence “Two things can wholly occupy the same place at the same time” is true. 2. On the most reasonably charitable interpretation of the dominant language in North America, this language is Shmeng- lish; that is to say, English is Shmenglish. 3. Therefore (we conclude in English): Two things can wholly occupy the same place at the same time. I assume that (1) will be accepted by revisionists. For quasi-nihilists, to take them again for illustrative purposes, the Shmenglish sentence “Two things can wholly occupy the same place at the same time” is made true by such facts as the one about lumps-of-clay-wise and statue-wise arrangements of matter that I sketched a moment ago. If there is a fl aw in this argument it is evidently not brought out by a philosopher’s announcing her deep metaphysical commitment to the “no-two-things-in-the-same-place” principle. Since this philosopher will admit (I am now assuming) that Shmenglish is an intelligible language in which the principle (that is, the sentence expressing the principle) is false, the question that has to be addressed is whether English is Shmenglish. Interpretive charity indicates that it is. And some philosopher’s professedly unshakable commitment to the “no-two-things-in- the-same-place” principle seems to have no serious bearing on this question. Of the various other general principles invoked by revisionists, some seem to be highly contrived philosophical constructions, which have as such no direct relevance to issues of interpretive charity.22 None have, I think, the intuitive commonsensical appeal of the “no-two- things-in-the-same-place” principle. If any of these principles generate genuine confl icts of charity in interpreting the language they should be

22. Two examples: fi rst, the principle that a quality cannot be formally treated as a relation between a thing and a time, invoked in behalf of temporal parts in Lewis, On the Plurality of Worlds, 202–4; second, the principle that a composite thing cannot be causally redundant in a certain technical sense, invoked in behalf of quasi-nihilism in Merricks, Objects and Persons, ch. 3. 112 quantifier variance and realism: essays in metaontology dealt with as outlined above. Philosophers who are inclined to assert these principles should fi rst remind themselves that if they were speaking Shmenglish it would be wrong to assert them. Next they should remind themselves that they are speaking Shmenglish (if they are speaking English). So they should stop asserting these principles.23 That linguistic interpretation leans heavily on examples comes out in one of the most famous passages in the Philosophical Investigations.24 Wittgenstein considers the impulse to think that games must have something in common, or they would not be called “games.” “Don’t think, but look!” Wittgenstein says. Look at examples of games and you fi nd that they don’t satisfy some set of necessary and suffi cient conditions. The revisionist might be compared to someone who answers: “I have looked and indeed have found that what common sense calls ‘games’ do not have anything in common. Since this is absurd I conclude that common sense is wrong in its judgments as to which activities are games.” We might even imagine this answer giving rise to counterparts of our three main revisionist positions, universalism, nihilism, and quasi-nihilism: “Every human activity is really a game”; “No human activity is really a game” (two existentialist positions); and, as one tempting counterpart of quasi-nihilism, “The only real game is dreydl.”

The primacy of (perceptual) examples in resolving confl icts of interpretive charity is evidently only one part of a more complicated story. Certainly there are other considerations that enter into resolving such confl icts, the barest sketch of which must include at least the following. If one interpretation has the eff ect of sustaining some of the typical speaker’s assertions while abandoning others, and a second interpretation has the reverse eff ect, the correct interpretation is the one that tends as far as possible to favor assertions that are:

A. assigned good reasons (rather than truth) B. numerous

23. I suspect that for some philosophers it will be hardest to give up certain principles involving the vagueness of existence and identity which generate problems for common sense. I’ve tried to directly address these problems in “The Vagueness of Identity,” Philosophical Topics: The Philosophy of Sydney Shoemaker 26 (1999): 139–58; and “Quantifi er Variance and Realism.” 24. I, 66. against revisionary ontology 113

C. perceptual D. specifi c (as in examples) E. strongly held F. widely held G. hard to qualify H. hard to explain away by a theory of human error

Some of these criteria may be closely related. For instance, assertions that are on the order of general principles (hence, do not satisfy D) tend to be both easily qualifi ed and easily explained away (hence, do not satisfy either G or H). I think that almost all of the criteria favor an interpretation of English that makes ordinary ontological assertions come out true or reasonable and various principles invoked by revisionists come out false or unreasonable. Criterion H is worth looking at for a moment. If an interpretation abandons some assertions as being false or unreasonable, that interpretation is more credible insofar as we have some plausible way of explaining why people might be expected to make mistakes in those assertions. On my interpretation of English people tend to make mistakes about general principles such as the “no-two-things-in-the- same-place” principle because they generalize too hastily from a few paradigmatic examples that immediately come to their minds. This seems to be a very obvious and plausible explanation. By contrast, revisionists have no plausible way of explaining why people make the mistakes revisionists allege. According to (quasi-) nihilists, for example, people mistakenly judge tables to be in front of them when there are no tables in front of them. Why would people make a mistake like that? The illusion of depth that generates the whole revisionary project also generates the illusion that there is an answer to this question: People make these mistakes because it requires deep philosophical insight to avoid them. But what sense does this make? If there isn’t an object of a certain sort in front of people why would they have to be as philosophically acute as Trenton Merricks to avoid the mistake of perceptually judging that such an object is there? Or take this from the (four- dimensionalist) universalist’s angle. If when two sticks of diff erent color are held up, as in my earlier example, there are numerous highly visible wooden objects there that change color, why would one have to be as deep as Mark Heller to avoid the mistake of thinking that there are no such objects there? If acknowledging the existence of the sticks requires no great philosophical depth, why would acknowledging the existence of these other stick-sized wooden objects require depth? The fact is, 114 quantifier variance and realism: essays in metaontology

I think, that revisionists standardly delude themselves into thinking that they can plausibly explain why people make the mistakes they allege. It is closer to the truth to say that such mistakes would require some highly selective and seemingly arbitrary forms of genetically or cultur- ally transmitted idiocy. Criterion A harks back to a point made very early in this paper: the principle of charity pertains primarily to good reasons rather than to truth. This criterion is, I think, the main reason why the correct interpretation of a language can sometimes have the eff ect that typical speakers make assertions that are empirically (as opposed to a priori) false, even about (perceptual) examples. Suppose that typical speakers are prepared to assert “The Earth is fl at.” Why not interpret that as being true, as meaning something like “The Earth is locally fl at (or looks fl at)?” We have to take into account many other assertions these people will make, such as, “If the Earth is fl at, then if you keep moving in as straight a line as possible (over land and sea) on its surface you’ll reach a point where you can’t go any farther (you fall off ),” and “If the Earth is fl at it’s shaped more like a large pancake than like a large grapefruit.” The details are not easy to spell out, but I think it’s fairly clear that we are faced with essentially two interpretive choices. On one interpretation “The Earth is fl at” turns out to be true, but numerous conditionals asserted by the speakers turn out to be incomprehen- sibly unreasonable. On the second interpretation the conditionals are correct, and, although “The Earth is fl at” is false, people have (tolerably) good reasons for asserting it, given their sensory data. The second interpretation is, therefore, the credible one, by criterion A. Other examples of attributing empirical falsehoods to typical speakers may involve further complications.25 The essential point to stress in all of these examples is that they do not involve contravening the principle of charity in the basic sense of criterion A, for we are not denying that, given their sensory experience, the speakers may have good reason to assert what they do. In contrast, revisionists imply that typical speakers of the language make many a priori false ontological judgments for no good reason. It should be clear that because of criterion A we never seek mere “models” as our interpretations (in the sense of Putnam’s “model theoretic” argument against Realism). Suppose that by mapping words of English to certain sets of objects on Alpha Centauri, objects that are

25. If typical speakers assert “Whales are fi sh” and we interpret the sentence as being false, this may require an appeal to Lewis’s “naturalness-presumption,” to be discussed later. against revisionary ontology 115 causally unrelated to people on Earth, we can make the sentences typically asserted come out true. That mapping would not even qualify as a candidate interpretation, since people on Earth would evidently have no reason to assert sentences that are made true by causally remote objects on Alpha Centauri. A credible interpretation must delineate perceptual sentences that speakers typically assert with good reason when their sensory apparatus is suitably aff ected by their environment, and then must delineate reasonable connections between these assertions and others typically made. The revisionist’s interpretation of English drastically fails to meet these constraints in the most charitable way possible.26 As Davidson and others have stressed, interpreting a language is part of the larger project of understanding human behavior. Criteria A through H deal only with assertions, but other forms of linguistic and non-linguistic behavior would ultimately need to be taken into account. I want to especially stress that these criteria are inadequate for interpreting normative language, which is linked essentially to attitudes and non-linguistic behavior. Suppose that typical speakers are prepared to assert, “We’re not morally required to make large sacrifi ces to help starving strangers,” and that Peter Singer says he disagrees with this. Here we certainly must not try to appeal to an argument from ordinary language to refute Singer. Questions of value cannot be resolved by appealing to ordinary English usage. On one meta-ethical picture Singer would admit that the “descriptive meaning” he assigns to the words “morally required” is diff erent from that assigned by typical speakers, but the important point is that the “evaluative meaning” of those words is fi xed by people’s attitudes and behavior. Singer is engaged in a “disagreement of attitude” with ordinary people, and that can’t be resolved by straightening out misinterpretations of language. The meta-ethical picture I’ve just appealed to may be overly crude, but I think the essential point is clear enough. Questions in the ontology of moderate-sized dry goods, however, typically have no evaluative signifi cance whatever. For instance, my friend van Cleve is a mereological essentialist who claims to believe that when you change a tire on a car you have replaced one car with

26. Merricks suggests that, although the ontological assertions of ordinary folk are not really supported by any sensory data, and are refutable by a priori philosophical arguments, they are nevertheless a “justifi ed starting-point” because they are “a matter of conventional wisdom and local cus- tom” (Objects and Persons, 74–75). Charity, however, requires us to look for an interpretation of these assertions that assigns “good reasons” to them in a more robust sense. 116 quantifier variance and realism: essays in metaontology another one. I would still not hesitate to let him fi x my fl at: why should I care if he likes to call this “destroying the original car?” It’s because questions in the ontology of moderate-sized dry goods have no evaluative signifi cance—and also no empirical signifi cance—that typically the only thing required to answer them correctly is: Speak plain English! 27

I turn, fi nally, to “constraints beyond charity.” We can introduce this topic by considering a puzzling argument in Richard Cartwright’s paper “Scattered Objects,” a seminal early universalist formulation.28 Cartwright begins by giving examples of scattered objects, such as a pipe whose parts are spatially disconnected while it is being cleaned, or a multivolume book whose volumes might be in diff erent places. He then notes that, since some spatially disconnected objects make up an object, it would be “arbitrary” to exclude any disconnected objects from making up an object. He thereby draws the universalist conclusion that any two objects, however disconnected, make up an object. What kind of argument is this? It seems clear that from the standpoint of common sense an expression of the form “the thing made up of the F and the G” refers to something only if the things referred to by “the F ” and “the G” satisfy some suitable kind of unity relation, whether this be spatial, causal, historical, or whatever. Ordinary people would certainly not count Plato’s nose and the Eiff el Tower as making up something, since they fail to satisfy any relevant unity relation. What can Cartwright mean by rejecting this ordinary judgment because it is “arbitrary?” Consider that it is given that in Shmenglish, as I have defi ned that language, the sentence “There is something made up of Plato’s nose and the Eiff el Tower” is false. I claim that all the linguistic evidence indicates that English is Shmenglish. How can Cartwright rebut this claim by appealing to “arbitrariness?”

27. Persons, however, may be a specimen of moderate-sized dry goods that have to be treated diff erently. To the extent that the question “Will that future person be me?” is linked to the normative question “Ought I to have some special concern for that person?” I’m not sure if an appeal to ordinary English usage completely suffi ces to answer even the rst fi question (let alone, the second). For a slightly diff erent angle on essentially the same point, see Alan Sidelle, “On the Prospects for a Theory of Personal Identity,” Philosophical Topics: The Philosophy of Sydney Shoemaker 26 (1999): 351–72. 28. The argument I’m referring to is at 293 and 298 in Kim and Sosa, eds., Metaphysics: An Anthology. against revisionary ontology 117

Cartwright seems perilously close to carrying on a burlesque battle with the English language. Obviously, if we don’t like something about the language—because, perhaps, it strikes us as “arbitrary”—that doesn’t entitle us to say that people who are using the language in the conventionally correct way are making mistakes. Of course we can overtly stipulate a change in the language, but that’s exactly what revisionists like Cartwright say they are not doing. “Arbitrariness” in philosophical discussions is closely related to “too much complexity” (or “too much complexity of the wrong kind”). Perhaps what is really bothering Cartwright is that the unity relation that seems to enter into the ordinary meaning of the expression “thing made up of the F and the G” is too complex. If it is unclear whether this is the problem that leads Cartwright to universalism it is certainly one of the main problems that leads van Inwagen to his brand of quasi- nihilism.29 One of van Inwagen’s main complaints against common sense is that it seems impossible to give any kind of accurate analysis of what counts as the unity relation for common sense. If ordinary judgments about objects were correct the unity relation would have to turn out to be highly disjunctive and intractably complex. I think it’s quite likely that van Inwagen is right about this. He talks primarily about the composition relation—that is, the relation that holds between the contemporaneous parts of a unitary thing—but I also have in mind the relation that binds the successive stages of the history of a single thing. An examination of cases suggests, I think, that those successions of (sets of) bits of matter that common sense treats as answering to the history of a persisting object probably do not share some conjunction of conditions that neatly mark them off from sets that are not so treated. In all likelihood, our common sense ontology requires a unity relation that is messily disjunctive and even in some ways grue-like.30 But so what? The truth can obviously be told with grue-like expressions: the grass that grew a few years ago was grue, and the grass that grows now is bleen; surely, no one questions that. Van Inwagen, like Cartwright, seems to be battling the language. Let me put the point as before: It’s given that in Shmenglish the sentence “There exists a table” is true, and all the linguistic evidence seems to indicate that English is Shmenglish. So how could van Inwagen’s complaints about the

29. Material Beings, esp. 64–71 and 122–23. 30. Markosian, “Brutal Composition,” suggests that the unity relation is in principle not ame- nable to any fi nitary non-circular analysis. This is, I think, a very interesting possibility, but a lot will turn on what one means by a “non-circular analysis.” 118 quantifier variance and realism: essays in metaontology

complexity of the unity relation possibly lead us to his conclusion that “tables don’t exist,” which implies that English is not Shmenglish?31 Perhaps the complexity of the common sense unity relation aff ects the revisionists in the following way: they simply claim not to understand what this relation is. They don’t understand what it is that gets ordinary people to distinguish between “things that make up something” and “things that don’t make up something.” This is a peculiar impasse. From my point of view such revisionists would merely be pretending not to understand English. Of course, they won’t see it that way. They understand English well enough, they would say, but it’s my make-believe language Shmenglish that they don’t understand. What they don’t understand, they say, is the supposedly “most charitable interpretation” of the assertions made by typical North Americans, the interpretation that would make these assertions come out true and that is stipulated to be the correct one for Shmenglish. “Give us a clear analysis of the unity relation for objects that is supposed to operate in Shmenglish,” they demand. And I’m apparently trapped, because I can’t give a clear analysis; the most I can do is make some rough sketchy gestures.32 Let’s consider an analogous case involving an imaginary philosopher named Shmgettier. In this fantasy Gettier never existed. Shmgettier, however, presents a couple of novel examples. One of them concerns someone Smith who has evidence that Jones owns a Ford. . . . Well, we know the example. It’s what we call in the real world a Gettier case. But Shmgettier’s approach to this case is very diff erent from that of the real life Gettier. Shmgettier’s point is that our best theory of knowledge, roughly that knowledge is true rational belief, teaches us that this case is a case of knowledge. Of course common sense says otherwise, but, Shmgettier declares, common sense is often wrong, and here is a case where philosophical theory shows us it is wrong. “But,” we object to Shmgettier, “you’re merely misusing the English word ‘knowledge.’ As that word actually functions in the language it obviously doesn’t apply to the kind of case you mentioned.” “What ‘kind of case’ is that?” replies Shmgettier. “I simply don’t understand how you think the English word ‘knowledge’ functions, if it doesn’t

31. This question applies even more obviously to those numerous (quasi-) nihilists who repudiate tables by appealing to Ockham’s razor: How could that appeal possibly show that English is not Shmenglish? 32. Actually I made 311 pages worth of rough sketchy gestures in The Concept of Identity (New York: Oxford University Press, 1982). against revisionary ontology 119 function as I say. Give me a clear analysis of how you think it functions.” Of course, that’s exactly what we’re not able to do. So does Shmgettier win this argument? No, I think almost everyone will agree he doesn’t. The lesson we learn from this case (and numerous others like it) might be put as follows: Charity trumps problems of analysis. When we interpret a language we proceed on the presumption that ordinary speakers are making correct judgments about examples. That presumption is not defeated by the diffi culty (or impossibility) of giving a clear analysis of how the language functions.33 I think that, in fact, many revisionists ought not even to be tempted by the move I’ve just been discussing, the move of saying, “We simply don’t understand how Shmenglish is supposed to function.” The overall story told by revisionists will often provide the resources to say how Shmenglish functions. Four-dimensionalist universalists, for example, will almost always want to distinguish between the broad category of space-time worms that are physical objects and the sub-category of such objects that are talked about outside philosophy. Quine calls this sub-category “bodies.”34 One can then defi ne Shmenglish simply as a language in which the most general word “(some)thing,” as this applies within the domain (roughly) of space-occupying perceptible objects, is equivalent to “body.” There can therefore be no complaint that the truth conditions of the sentences of Shmenglish have not been made intelligible. We have, then, a kind of reductio argument: Since four-dimensionalists distinguish between “bodies” and “other (space- occupying perceptible) objects,” they can understand how Shmenglish operates by appealing to the former category. It must be clear to them, therefore, that Shmenglish is the dominant language of North America, i.e., English is Shmenglish. But, then, four-dimensionalists are hoisted with their own petards, for, speaking English, as they claim to be doing, they can no longer distinguish between “bodies” and “other (space- occupying perceptible) objects.” Considerations of complexity may play another kind of role in linguistic interpretation. Lewis has the well-known view that, in interpreting a language, there is the presumption that words express properties that are natural to a relatively high degree.35 Degrees of

33. Note that, unlike Shmgettier’s question, classical skeptical issues are typically about our “right to be sure,” and this apparently normative component may not be resolvable merely by appealing to ordinary language. 34. W. V. Quine, The Roots of Reference (LaSalle, Ill.: Open Court, 1974), 54. 35. David Lewis, “New Work for a Theory of Universals,” 370–77. 120 quantifier variance and realism: essays in metaontology

naturalness are, for Lewis, the same as degrees of complexity relative to “perfectly” natural properties. This naturalness-presumption, as I will call it, ought to be viewed as a constraint beyond charity. The point is somewhat obscured by the fact that in Lewis’s own discussion he seems to treat the naturalness-presumption as a corollary of the principle of charity. I have been talking about charity at the level of assertions: an interpretation is credible to the extent that it makes typical assertions come out reasonable (if not true). Perhaps Lewis thinks that there is another level of charity: the semantic structure of the language itself ought to be made reasonable. If this is the correct explanation of the naturalness-presumption, it would have to be the case that, other things being equal, it is unreasonable for people to have a language whose words express highly unnatural properties. I have devoted a book- length discussion to this question, and, as far as I can make out, there would in fact be nothing unreasonable about this.36 This makes me somewhat skeptical about the naturalness-presumption, though I think it would also be possible to take this presumption as irreducible and unrelated to charity at any level. The presumption is, in any case, a constraint beyond charity in my sense, i.e., charity at the level of linguistic behavior. Perhaps we should back up for a moment and ask what the basis is for the principle of charity itself. The essential answer is given by Davidson: The simplest and most plausible explanation of any human behavior, including linguistic behavior, is, other things being equal, that there is a good reason for the behavior.37 The principle of interpretive charity is simply an application of this general idea. If we accept the naturalness-presumption, it might be possible to apply it to the word “unity (of a thing),” with the result that there is a presumption in favor of construing the unity relation as being not too complex. Even so, charity (at the level of assertions) trumps the naturalness-presumption, as Lewis would certainly agree. We see this in countless cases, such as the case of Shmgettier. Shmgettier’s interpretation of “knowledge” assigns to this word a much simpler, and therefore more natural property than that assigned by the correct interpretation. Shmgettier is wrong because charity trumps naturalness. Certainly one would have to say the same thing about “unity.” The revisionist’s interpretation of that word would be wrong for just the same reason that

36. Dividing Reality (New York: Oxford University Press, 1993). 37. Truth and Interpretation, esp. 159–60. against revisionary ontology 121

Shmgettier’s interpretation of “knowledge” is wrong. Shmgettier’s essential folly is to perversely interpret the language in a way that implies that typical speakers make a priori false judgments about examples. That seems to be precisely the revisionist’s folly. The most signifi cant response in the literature (known to me) to the kind of challenge that I am here posing to revisionism is found in some recent work by Theodore Sider.38 Sider holds that there is, in addition to the principle of charity, and in addition to the naturalness-presumption, a third constraint: language ought to respect the world’s “logical joints.”39 The part of the English language that is most closely connected to “logical joints” are such expressions as “(there) exists” and “(some)thing,” the so-called quantifi ers. Let us say that expressions in a language are quantifi er-like if they function formally like the English quantifi ers (roughly speaking, they satisfy the rules of the predicate calculus). Sider’s third constraint implies that the quantifi er-like expressions in any possible language must be genuine quantifi ers, and this means that they must answer to reality’s logical joints. There cannot, therefore, be diff erent languages whose quantifi er-like expressions have diff erent (referential) meanings. Since quantifi er-like expressions play the same formal role in diff erent languages, they might be trivially said to share the same “formal meaning.” Sider’s position, however, is that such expressions must also share the same meaning in the sense of referring to (ranging over) the same objects.40 When it comes to quantifi er-like expressions, therefore, charity falls out of the picture. Given that a language contains quantifi er-like expressions it is already established that they have the only possible meanings that such expressions can have. Our task as ontologists is to try as best we can to describe what really exists, and in doing so we

38. “Criteria of Personal Identity and the Limits of Conceptual Analysis,” in Philosophical Perspec- tives 15 (2001), and the Introduction to Four Dimensionalism: An Ontology of Persistence and Time. Sider is, as far as I know, alone amongst current revisionists in recognizing that they face a serious problem with respect to interpretive charity. 39. Sider bases linguistic interpretation on what he calls “use” and “eligibility.” The fi rst corresponds roughly to what I call “charity,” and the second to what I call “constraints beyond charity.” The latter constraints divide up, for Sider, into the naturalness-presumption and the additional constraint about “logical joints.” 40. Sider does allow that there could be explicit stipulations that assign diff erent (referential) meanings to quantifi er-like expressions, but such diff erent meaning would be by necessity derivative of the primary meaning. He may also allow for the possibility of languages that have no quantifi er- like expressions at all. Let me ignore these complications. 122 quantifier variance and realism: essays in metaontology know in advance that we are at the same time describing what is referred to by the quantifi er-like expressions of any possible language. Ontology teaches us, therefore, that Shmenglish—the imagined language in which the sentences typically asserted by speakers of English come out true—is a metaphysically impossible language. As charitable as we would like to be to ordinary people’s ontological assertions, therefore, we cannot avoid the conclusion that many of these assertions are false. I think that Sider’s underlying assumption is that the truth conditions of the sentences of any possible language must be generable by (something approximating to) a Tarski-style referential semantics, with the quantifi er-like expressions ranging over everything (“everything” as defi ned by the true ontology).41 Of course, if we are operating from inside Shmenglish there would be no special problem in generating the imagined truth conditions of its sentences by what looks like a Tarski- style treatment. But that does not suffi ce, according to Sider, for ontology teaches us in advance that operating from inside Shmenglish is not even a possibility. It seems to me that Sider’s position is the only hope for revisionists, but I fi nd it to be a dim hope. Of course I am bothered (as Sider himself is) by the heavy metaphysical apparatus of “logical joints,” and also by the quite mysterious epistemology that must go with it. But what is more centrally of concern to me is Sider’s contention that, as regards quantifi er-like expressions, charity doesn’t count. Sider’s third constraint, therefore, goes far beyond Lewis’s naturalness-presumption. It is, in fact, not a presumption, but an absolute and indefeasible requirement. Why should we suppose that there is any such absolute requirement on language? How indeed can there be such a requirement? Sider seems willing to say that in an important sense “we can’t mean what we want” by the quantifi er-like expressions.42 Worse, in refusing to interpret charitably the English language we actually speak, Sider’s position implies that in an important sense we can’t possibly mean what we in fact seem to mean. Why should anyone accept this? The position seems to me primarily designed to provide revisionists a carte blanche to fl y in the face of common sense.

41. On possible connections between this formulation and the philosophical positions in Davidson, the Tractatus, and the Philosophical Investigations, see Kripke, Wittgenstein on Rules and Private Languages, 71–72, note 60. 42. A more focused discussion of the specifi c issue of assigning diff erent meanings to quantifi er expressions is given in “Quantifi er Variance and Realism” and “Sosa’s Existential Relativism.” against revisionary ontology 123

A number of philosophers have reacted to my point of view by cau- tioning me about the diffi culty in determining what ordinary people— typical fl uent speakers of the language—really believe about the existence and identity of moderate-sized dry goods. Certainly there are diffi culties; I’ve limited myself here to saying only enough to reveal the basic error of revisionism. Part of the diffi culty is to find the proper balance in both acknowledging and not exaggerating the genuine elements of vagueness and indeterminateness in the ordinary language of existence and identity. A familiar experience for a philosophy professor is that some students can easily be induced to accept almost any ontological position. What follows? That all bets are off , that there are no rules of language? No, what follows is that interpreting a language is a delicate matter. First and foremost one has to avoid kicking the language around, pushing it where one wants it to go. Austin called his method “linguistic phenomenology,” which I think suggests the idea of allowing the linguistic phenomena to come to light without willful manipulation.43 The most important linguistic data are not acquired primarily from the confl icting utterances characteristically generated under philosophical pressure, but primarily from the relatively sponta- neous and unrefl ective utterances shared by virtually all speakers, including philosophers. If our philosophical business is the ontology of moderate-sized dry goods—and that is, obviously, not the only important part of metaphysics—we need to put aside the currently fashionable revisionist’s mannerisms: the comical battles with ordinary language, the lawyer’s briefs for and against entities, the mock-theoretical debates about what objects exist in front of our eyes. If our business is the ontology of moderate-sized dry goods then our only philosophical task is to explore the currents and undercurrents of our ordinary ways of thinking and talking about the perceptible world around us.44

43. J. L. Austin, “A Plea for Excuses,” in Philosophical Papers (Oxford: Oxford University Press, 1961), 130. 44. I owe a debt of gratitude to Ned Markosian, Trenton Merricks, and Ted Sider for many important criticisms of a much earlier draft of this paper, which led to extensive revisions. seven c C

comments on theodore sider’s FOUR DIMENSIONALISM

C Theodore Sider has given us a terrifi c book, bursting at the c seams with new arguments and new takes on old arguments. Whether or not one is convinced by his conclusions, the thorough- ness, lucidity, fair-mindedness—and the sheer exuberance—of his discussions make Four Dimensionalism a major contribution to contemporary metaphysics. Sider defends a four-dimensionalist position committed to temporal parts and unrestricted mereological composition. The following may serve to illustrate both what the position amounts to and how it stands to common sense. Suppose that an ordinary tree is in the yard, and consider with respect to that situation the sentence A: “Something in the yard is a highly visible brown wooden object that contains branches during the daytime and contains no branches during the nighttime.” I will take it as obvious that non-philosophers would tend to regard A as insane. Four-dimensionalists, however, accept A by virtue of the object that is the sum of the daytime parts of the tree and the nighttime parts of the trunk. I am going to try to present a reductio ad absurdum refutation of four- dimensionalism. Sider has a position in his book, spelled out especially in the Introduction, that would answer my argument. Although I don’t agree with that answer I think it’s one of the most important ideas in the book, and I would like Sider to clarify it. If we are four-dimensionalists we will draw a rough distinction between those objects that are typically countenanced by non-philosophers and

124 comments on theodore sider’s FOUR DIMENSIONALISM 125 the additional objects that we accept as four-dimensionalists. The latter are mereological constructions out of (temporal parts) of the former. For present purposes, I will simply call the latter “Siderian objects.” In the situation imagined earlier, the tree and its trunk and branches are non-Siderian objects, and the object that is the sum of the daytime parts of the tree and the nighttime parts of the trunk is a Siderian object. Here, then, is my reductio argument. 1. Assume that four-dimensionalism is correct, so that (with respect to the imagined situation) it is the case that A. 2. We can imagine a language—let’s call it Shmenglish—in which the most (contextually) unrestricted quantifi er expression “(there exists) something” is equivalent in reference to “(there exists) something that is not a Siderian object;” so in Shmenglish it is correct to say, “It is not the case that A.” 3. The most infl uential and best understood methods of linguistic interpretation require us to say that the dominant language in North America is Shmenglish; that is to say, Shmenglish is plain English. 4. We conclude (in plain English): It is not the case that A. We conclude, in other words, that there are no Siderian objects, and four-dimensionalism is false. Let me comment briefl y on the argument. Probably it does not really qualify strictly as a reductio, because we might be able to skip premise 1 and give a straight argument starting from 2. But imagining ourselves starting as four-dimensionalists makes it especially easy to explain how Shmenglish is supposed to operate. And the reductio format vividly brings home the lesson that one’s best eff orts to do the kind of “deep ontology” that Sider wants to do necessarily degenerate into absurdity because of purely linguistic considerations. Premise 2 doesn’t imply that whenever speakers of Shmenglish utter the word “something” they contemplate the non-existence of Siderian objects. The Shmenglish speakers never give a thought to Siderian objects. The point is rather that the semantic rules implicit in their linguistic behavior imply that their quantifi er expressions do not range over Siderian objects. I don’t care if you say, “Then those expressions aren’t really quantifi ers” (as Sider seems to say on p. xx). They are, in any case, “quantifi er-like.” So the most general (contextually) unrestricted quantifi er-like expressions available in Shmenglish don’t refer to Siderian objects. (Of course, if my argument succeeds, and Shmenglish turns out to be plain English, then the quantifi er-like 126 quantifier variance and realism: essays in metaontology expressions of Shmenglish are presumably what philosophers call “quantifi ers.”) Step 3 appeals primarily to the familiar “principle of charity,” which says, roughly, that an interpretation is to be favored to the extent that it makes the prevailing judgments in the linguistic community come out true or reasonable. Typical speakers of English regard innumerable sentences like A as insanely false. Sider interprets English in a way that makes these speakers wrong. They are wrong about what they perceive in front of their faces, and not just wrong, but a priori necessarily wrong, insofar as their judgment that there is a tree with branches in the yard a priori necessarily entails, according to him, that A is true. A less charitable interpretation, it would seem, is hard to imagine. Charity impels us to interpret the language of these speakers as Shmenglish, in which A is indeed insanely false. Since Sider holds that four-dimensionalism is a priori necessarily true, he will agree, I think, that speakers of Shmenglish can express (object-level) facts a priori necessarily equivalent to (and, therefore, in a sense identical to) all the facts he can express. Shmenglish is therefore not to be compared to some bizarre language in which the quantifi ers contain a blind spot leading people to bump into objects over which they don’t quantify. If Shmenglish is defective it’s certainly not defective in that way. The immediately relevant question, in any case, is not whether Shmenglish might be in some ways an imperfect language, but whether it is the dominant language in North America. Considerations of interpretive charity seem to indicate decisively that it is. The fi nal step 4 of the argument descends from the meta-level and draws a conclusion in plain English. There is no trickery in this. We’re supposed to be speaking plain English, aren’t we? I certainly would have no disagreement with the proposal that we might change our language, perhaps to suit some philosophical purpose, so that A is converted into a true sentence. But Sider is evidently claiming that A is true in the language we already have. If Shmenglish is the language we already have, he is wrong. The answer to this argument implicit in Sider’s Introduction is, to my knowledge, completely original—and completely audacious. Sider holds that Shmenglish is a metaphysically impossible language! There can be no question of interpreting the language of our community as Shmenglish, because it’s impossible for any linguistic community to have that language. The reason for this is that “the world comes equipped with ‘logical joints’ as well as extra-logical ones” (p. xxii), and comments on theodore sider’s FOUR DIMENSIONALISM 127 the quantifi er-like expressions of any language must cut at the logical joints. Ontology is the study of the world’s logical joints, and if ontological inquiry leads us to four-dimensionalism then it follows that the quantifi er-like expressions in any language must range over Siderian objects, so that Shmenglish is not a possible language. If I have correctly expressed Sider’s view—and I’ll raise a certain doubt about this later—then it is open to two objections. The fi rst has to do with the nature of the “logical joints;” the second has to do with the compulsion to connect the quantifi er-like expressions to the “joints.” The “extra-logical joints” are Lewis’s “natural properties,” which, as Sider notes, are “classes of objects that count as genuinely similar” (p. xxii). That is obscure enough, but at least we have the intuitive notion of similarity to help us. What do we have to help us understand what the world’s “logical joints” are? This notion seems extremely dark. Might Sider be suggesting that existence itself is a natural property? Apart from raising with a vengeance all of the old questions about how existence can be a property, this suggestion would raise addition- ally baffl ing questions about how existence can make for similarities. It’s worth noting that Lewis himself has another notion that operates together with natural properties: natural things, that is, “well-demarcated things.” l Many Siderian objects are examples of what Lewis (who is of course a four-dimensionalist) would call unnatural things. Sider’s logical joints, therefore, cut across Lewis’s natural things. This may be merely an oddity, but I think it adds to one’s sense of perplexity about what these “logical joints” are. Putting these doubts aside, the more urgent question is why a language like Shmenglish should be deemed impossible just because its quantifi er-like expressions fail to cut at the logical joints. Sider draws on Lewis’s theory of linguistic interpretation, according to which charity, in the sense indicated earlier, is not the only consideration that determines the correctness of an interpretation. There is also the presumption that words express natural properties (indeed, natural properties of natural things, but let’s ignore that). Interpretation, as Sider puts it, is a function of both “use” (= “charity”) and “eligibility.” Lewis says that natural properties are the most eligible candidates for the interpretation of general words, and Sider is now adding that the class of all the objects there are, as uncovered by our best ontological theory, is the most eligible interpretation of quantifi er-like expressions. That’s

1. David Lewis, “New Work for a Theory of Universals,” Australian Journal of Philosophy 61, 343-77, at p.372. 128 quantifier variance and realism: essays in metaontology why it can’t be correct to interpret the quantifi er-like expressions of any language as failing to range over Siderian objects. I assume that the way I have summarized this view makes it immediately apparent what the diffi culty is. If interpretation is governed by a tradeoff between use and eligibility then, granted that the interpretation of “something” in Shmenglish is not maximally eligible, why can’t considerations of use trump eligibility to arrive at that interpretation? Certainly in Lewis’s theory of how to interpret general words it very often happens that use trumps eligibility. It’s not that Sider ignores use in his discussion of quantifi er-like expressions, but, at least in his Intro- duction, the only element of use that he takes into account with respect to these expressions are their “inferential role” (pp. xxii-xxiii), for instance, their satisfying the rules of predicate logic. But surely there is more to use/charity than that. The fi erce denial by members of our linguistic community of innumerable sentences like A must surely provide an overwhelming argument from use/charity to the Shmenglish- interpretation of the quantifi er-like expressions. Only if there is reason to believe that eligibility considerations must always trump such arguments could there be a basis for saying that Shmenglish is an impossible language. What could this reason be? Let me admit at this point that I am not entirely certain that Sider is claiming that Shmenglish is an impossible language. This is something that I would like Ted to clarify for us. It seems quite clear to me that in his Introduction he was indeed saying that it’s metaphysically impossible for quantifi er-like expressions in diff erent languages to have diff erent semantic functions, that inferential role together with eligibility considerations necessarily determine the same meaning for such expression in any possible language. But some of his remarks later in the book (especially pp. 185-6) may leave a diff erent impression. One view—I hope it’s not Sider’s, and don’t think it is—might be that Shmenglish is a possible language but it happens not to be our language. Let me indicate how this view might work. I taught a metaphysics course last year attended by a student of Lin- guistics with no previous background in philosophy, and when I explained to the class what four-dimensionalists say about sentence A this student had a seizure of uncontrollable laughter. She told me afterwards that she thought that four-dimensionalism was an absolutely hilarious aberration of the English language. Now let’s imagine a linguistic community in which everyone reacts to sentences like A the way this student did. (Everyone in the community says, “That’s an absolutely hilarious aberration of the language.”) In this community comments on theodore sider’s FOUR DIMENSIONALISM 129 the language would indeed be Shmenglish, according to the view under consideration. But Shmenglish is not the language of our community, because people in our community generally don’t react to four-dimensionalism that way. I have to say that it’s not easy for me to see this as being an interesting position, and I hope (and believe) it is not Sider’s. A four- dimensionalist who concedes that Shmenglish is a possible language, but denies that it is our language, seems to be scraping the very bottom of the barrel. If it is possible for considerations of use to trump eligibility in favor of Shmenglish, then surely that’s what we have in our own linguistic community. The typical reaction in our community to A, if not exactly that of my student, is unmistakably negative in the extreme. Sider remarks that in his experience “unprejudiced folks” have no problems with four-dimensionalism (p. 218). It’s not diffi cult to imagine the kind of philosophical surgery he uses to remove their preju- dices. One has to avoid a kind of Stalinist semantics in which the true meanings of words are revealed by what typical fl uent speakers say under torture. Of course “use” has to take everything into account, including people’s philosophical ruminations and kvetching, but the primary focus is always on what people say spontaneously, especially about concrete examples (and especially perceptual examples). The considerations of use in favor of the Shmenglish-interpretation of the language of our community are overwhelming. If Shmenglish is possible, it’s actual. I don’t believe that Sider’s position requires him to argue with me about that conditional. I take our real disagreement to be over its ante- cedent. That is by far the more interesting disagreement. But I want him to tell us whether I have this right. Something else on which I think Sider and I can agree is that all of his arguments for four-dimensionalism are directed towards eligibility and the “logical joints”; none are directed towards use and charity. Consider his argument from the assumption that the quantifi ers can’t be vague (pp. 120-39). If Shmenglish were our language, our quantifi ers obviously would be vague, since there is no sharp line between “Side- rian” and “non-Siderian” objects. Sider’s point must be that, though our use of the quantifi ers makes them seem vague, the quantifi ers can’t really be vague because it can’t possibly be indeterminate where the logical joints are. His argument from vagueness presupposes that, with respect to the quantifi ers, the “joints” determine eligibility, and eligibility must always trump use. The need for this presupposition can also be seen as follows: If Sider thought it was possible for use to trump 130 quantifier variance and realism: essays in metaontology eligibility in interpreting the quantifi ers, then he would have to allow that there could be cases of borderline trumping, which would constitute vagueness in the meaning of the quantifi ers. His confi dence that there is no vagueness in our quantifi ers shows that, for him, once an expression’s inferential role identifi es it as quantifi er-like, eligibility considerations completely determine what the expression means. Sider might have written a book entitled “The World’s Logical Joints,” leaving the existence of perceptible objects as common sense has it. This could have been a new topic in philosophy (perhaps the topic that traditional ontology was always confusedly aiming for). There could even have been a nice division of labor: the patently portentous topic of “the world’s logical joints” inviting just the kind of complicated theoretical arguments that Sider and his friends are so adept at, leaving for people like me the shallow task of commenting on the ordinary language of existence. He should indeed have written such a book if he thought that questions about the “joints” are separable from questions about existence. But his basic assumption, if I have this right, is that these questions are not separable. They are not separable because the “joints” completely determine the interpretation of quantifi er-like expressions. That is why Shmenglish is not a possible language, and that is why the existence of perceptible objects cannot possibly be left as common sense has it. One of my main objections to Sider’s account is that the claim that Shmenglish is impossible is not supported by Lewis’s model of use versus eligibility. The claim has more in common, I think, with the kind of intuitive sentiment expressed by Kripke when he questions whether we can really make intelligible to ourselves (“from the inside”) a language in which concepts like plus are replaced by concepts like quus.2 If the quantifi er-like expressions in diff erent languages can be governed by diff erent semantic rules, then this implies that the speakers of these languages have in some sense diff erent concepts of existence. I think Sider has the intuitive feeling that this makes no sense. There can be only one concept of existence, the one we have. I certainly don’t deny that there is an initial intuition like that. But it can’t be sustained on refl ection. If there is only one possible semantic rule that can apply to quantifi er-like expressions, then any describable rule is either necessary or impossible. Assuming that the

2. Saul A. Kripke, Wittgenstein On Rules and Private Language (Harvard University Press, Cambridge, 1982), p.98, note 78. comments on theodore sider’s FOUR DIMENSIONALISM 131

Shmenglish rule is not necessary—and I think there are a number of considerations that bear that out—then that rule is impossible. But why should that be? Are we to say that we can’t make intelligible to ourselves what it would be like to speak Shmenglish? That seems rather like saying that, as it turns out, it is plus, rather than quus, that we can’t make intelligible to ourselves. Kripke’s question was whether we can really make intelligible to ourselves a “form of life” that strikes us as radically alien, where our own form of life is a “given.” One cannot announce on the basis of a complicated array of theoretical arguments that the Shmenglish quantifi er seems alien to us, and is therefore unintelligible. If there is any language that obviously does not seem alien to us, it is Shmenglish. Since the Shmenglish quantifi er seems obviously possible, then, assuming that it is not also necessary, there is no alternative to saying that there are diff erent possible semantic rules that can apply to quantifi er- like expressions. In that sense (and only in that sense) it is possible for there to be diff erent concepts of existence. I am grateful to Sider for paying attention to a foundational issue that is almost always ignored or obfuscated by other members of his anti-commonsensical camp. He makes it clear that before philosophers enter the trenches to fi ght it out over what exists they have to explain why those fi ghts don’t really amount to assigning diff erent meanings to the words “what exists.” I obviously intend to pay Sider a high com- pliment when I say that the meta-ontological view outlined in his Introduction is, to my knowledge, the most serious defense of anti- commonsensical ontology that has ever been attempted. He has not yet convinced me, however, that the defense is credible. eight c C

sosa’s existential relativism

Ernest Sosa’s notion of “existential relativism” opens new ground in one of the hardest areas of metaphysics.1 In this chapter I want to explore two questions: fi rst, how existential relativism relates to David Lewis’s ontological position, and, second, how it relates to Hilary Putnam’s conceptual relativism.

I. Existential Relativism and Explosionism

Existential relativism is initially presented by Sosa as one response to our ability—or apparent ability—to conceive of diff erent ways of breaking the world up into objects. He asks us to consider by way of illustration the possibility of operating with the concept of a “snowdiscall.” Thinking in terms of our ordinary conceptual scheme, we say that when a (suitably sized) chunk of snow is made spherical some object called a snowball comes into existence. We can imagine an alternative scheme in which “any shape between round and disc-shaped” stands to the term “snowdiscall” in the way that in the ordinary scheme “round” stands to “snowball.” When a chunk of snow is made disc-shaped, in the alternative scheme, but not in the ordinary one, it would be correct to say, “Something—i.e., a snowdiscall—has come into existence.” Obviously we can repeat this illustration for any shape we want. It follows that, for any shape, there will be a scheme in which it is correct to say “Some object has come into existence” whenever a chunk of snow acquires that shape. One philosophical response to these refl ections is to hold that all of the objects countenanced by the various schemes really exist. When a

1. All of my references in what follows are to Sosa’s “Existential Relativity,” in Midwest Studies in Philosophy 22 (1999) (cited as ER). A related discussion is in the fi nal part of Sosa’s “Putnam’s Prag- matic Realism,” The Journal of Philosophy 90 (1993): 605–26 (cited as PPR).

132 sosa’s existential relativism 133 chunk of snow becomes discshaped one object—a snowdiscall—comes into existence, when the chunk is made round another object—a snowball—comes into existence, and so on for every shape that the chunk can acquire. Reality suff ers from a kind of “explosion” of objects, as Sosa puts it (ER, 133). I will indeed call this position “explosionism.” Sosa himself calls it “absolutism,” but for reasons that will be explained later I prefer a diff erent terminology. To get a better feel for the scope of explosionism—for how far and wide the explosion of reality will be on this view—it may be helpful to look at an illustration a bit more extreme than Sosa’s. Suppose that an ordinary tree stands in a certain yard. We can consider the succession of (stages of) objects that consists of the whole tree during each day and only the trunk during each night. Imagine a conceptual scheme in which this succession corresponds to the persistence of a “shmree.” Although nothing exceptional is happening to the tree, in this other scheme it will be correct to say, “A certain brown wooden object—a shmree—shrinks in size by losing all of its branches each night, and then grows in size by retrieving the branches each morning.” The explosionist will say that there is in reality that shrinking and growing object in the yard. It is clear that explosionism is the ontological position held by David Lewis and many other philosophers. According to them, any succession of (stages of ) bits of matter constitutes an object. To put it another way: if we start out with our ordinary objects, any way of mentally dividing these objects spatially or temporally yields additional objects; and any way of mentally summing these objects yields still additional objects. Any object that we can cook up, so to speak, was really already there waiting for us. Explosionism seems to fl y in the face of common sense. If this is not entirely clear with respect to Sosa’s rather modest example of snowdiscalls, it surely seems clear with respect to the example of shmrees. Ordi- nary people would regard it as sheer lunacy to suggest that when an ordinary tree stands in a yard there is some brown wooden object there that keeps gaining and losing branches every morning and evening. Existential relativism is presented by Sosa as an alternative to explosionism. According to this position “what . . . exists relative to one conceptual scheme may not do so relative to another” (ER, 133). Is there a brown wooden object in the yard that keeps gaining and losing branches? According to the existential relativist there is such an object relative to the shmree-scheme but there is no such object relative to the ordinary scheme. Since ordinary judgments are made relative to the ordinary scheme, the ordinary person is quite right in regarding it as 134 quantifier variance and realism: essays in metaontology lunacy to judge that there is an object that gains and loses branches— relative to the ordinary scheme. On the other hand, it would be lunacy to judge that there is no such object relative to the other scheme. Existential relativism seems to me to be an exciting position. And the fi rst question that I want to raise in this chapter is why Sosa eventually gives it up. He evidently intends his fi nal position—a “middle ground,” as he calls it (ER, 141)—to be a kind of compromise between existential relativism, as initially explained, and explosionism. He calls his fi nal position “conceptual relativism,” but it seems to me that this terminology is likely to mislead, since, as Sosa himself stresses, this position is not relativist in any serious sense—indeed, it seems to be no more relativist than Lewis’s position. Here is the way that Sosa characterizes his fi nal position:

Absolutism [i.e., what I am calling explosionism] is true. Moreover, existential claims are true or false only relative to the context of speech or thought, which restricts the sorts of objects relevant to the assess- ment. Such restrictions are governed by various pragmatic or theoretical considerations. (ER, 142)

This, I think, is exactly David Lewis’s position.2 When we do philosophy, Lewis explains, our quantifi er is not restricted in any way. In terms of the unrestricted quantifi er explosionism is true: there exist snowdiscalls and shmrees as well as snowballs and trees. In ordinary contexts, however, we restrict the quantifi er in a way that excludes such things as snowdiscalls and shmrees. This seems to be what Sosa means when he says, “Speaking loosely and popularly we may hence say that there are only snowballs there, even if strictly and philosophically one would recognize much that is not dreamt of in our ordinary talk” (ER, 143). I think Sosa means that when we speak “strictly and philosophically” we employ the unrestricted quantifi er. There appears, then, to be no real diff erence between Sosa’s fi nal position and Lewis’s position. Why did Sosa give up on existential relativism, which, we recall, was initially introduced as an alternative to explosionism? If we remain existential relativists then we regard the conceptual scheme adopted by the explosionist as merely one possible scheme amongst many. Relative to this scheme there exist all the objects that exist relative to all the other schemes. This scheme is not, however, metaphysically privileged, as compared to the others. It gives us one kind of “relative” truth, not the “absolute” or “philosophically strict” truth, as the explosionists— and now Sosa—believe.

2. David Lewis, On the Plurality of Worlds (New York: Basil Blackwell, 1986), especially at p. 213. sosa’s existential relativism 135

I am not sure that I understand the argument Sosa gives in favor of explosionism—and against existential relativism—in the fi nal section of “Existential Relativity.” One thing that he says is that his preference for explosionism is motivated by his “failure to fi nd attractive and well-motivated restrictions on allowable matter-form pairs” (ER, 143), that is, on allowable ways of conceptually dividing and summing bits of matter to form what will be considered unitary objects. Earlier Sosa had explained that any conceptual scheme will contain a “selection function” (ER, 138) that determines how unitary objects can be conceptually constructed out of bits of matter. In explosionism the selection function is completely permissive: any construction counts as a unitary object. What seems to be driving Sosa towards explosionism is the following train of thought. If our selection function allows snowballs then it seems quite arbitrary to exclude snowdiscalls. Surely the diff erence of shape between these items cannot have any ontological signifi cance. But, then, if we are allowing snowballs and snowdiscalls, why exclude shmrees? Why exclude any construction? Explosionism seems to be the best escape from metaphysical arbitrariness. This argument for explosionism seems to me unpersuasive, and for reasons that Sosa himself elegantly explains in earlier sections of his paper. People can speak about snowballs or speak about snowdiscalls, Sosa explains, and yet be grasping the same “facts” from diff erent conceptual perspectives (ER, 140–1). Another way to put this point, I think, is that any truth expressible in one scheme is (a priori) necessarily equivalent to some truth expressible in another scheme. It follows that there is no substantive disagreement between people who employ dif- ferent schemes. As Sosa says: “Of course there may be reasons why it is better to select one [scheme] rather than another, pragmatic reasons at least; rather as there may be reasons why it is better to be at one location rather than another. But this would not show that the actual judgments of ‘what is nearby’ made by those poorly positioned are inferior to the judgments made by those better positioned. . . . Similarly, to have diff erent positions in ontological space might reveal a lack of coinci- dence in [selection functions], but little else in disagreement about what there can be or what there cannot be” (ER, 139). These points, it seems to me, apply as well to the scheme chosen by the explosionists. Perhaps there are pragmatic reasons to choose that scheme, at least when we are doing philosophy. It may be the simplest and therefore in a sense the least arbitrary scheme we can adopt. (I mean that the permissive selection scheme of explosionism may be logically simple, though it evidently generates a large number and variety of objects.) 136 quantifier variance and realism: essays in metaontology

But it does not follow that the judgments of “what exists” made by those positioned in other schemes are inferior in truth—inferior in describing the facts—to the judgments made from within the scheme of explosionism. There is, I think, no persuasive argument here to give up on existential relativism and hold that explosionism is the “absolute philosophical truth.” My own preference for existential relativism—or some closely related form of relativism that I will describe later—over explosionism stems from my earlier suggestion that explosionism fl ies in the face of common sense. I need now to explain this a bit further. According to both Lewis and Sosa, when ordinary people say, “It’s not the case that there is something brown and wooden in the yard that keeps gaining and losing branches” they are right, because they are using the quantifi er expression “there is something” as restricted to “standard objects,” as we might say. They are therefore not in confl ict with the philosophical claims made by Lewis and Sosa that there is something there that gains and loses branches, since these philosophical claims are made with an unrestricted sense of “there is something.” But this reply, it seems to me, does not really meet the point. Suppose we say the following to ordinary people: “Take a look at the tree in the yard. Is there something brown and wooden there that loses its branches each night and gains them back each morning? Now make sure that you don’t restrict the range of objects you consider in answering this question. Take into account objects of any sort whatever, no matter how strange they are, and no matter how unlikely you would normally be to refer to them.” Of course, they still insist that there obviously is no such object. I think this shows that even when they use the expression “there is an object” in a completely unrestricted sense they regard the explosionist claims as absurd. I have sometimes heard it suggested that ordinary people, no matter how hard they try, cannot use the quantifi er in a completely unrestricted sense, so that, no matter what they say, they are not really disagreeing with philosophical explosionists like Lewis and Sosa. That retort seems extremely puzzling to me. How could a semantic rule become entrenched in the English language with the eff ect that fl uent speakers of the language, with the exception of a few philosophers, constantly use quantifi er expressions in a contextually restricted fashion, without intending it, or knowing it, or even recognizing it when it is pointed out to them? I think that makes little sense. It needs to be emphasized that explosionists like Lewis and Sosa claim to be speaking plain English—“strict” plain English with unrestricted sosa’s existential relativism 137 quantifi ers. I agree that the contexts of ordinary life do indeed often restrict the quantifi er in various ways, but ordinary people must be able to understand how to use in plain English the quantifi er in unrestricted ways, and when they do so they adamantly reject explosionism. Explosionism does, therefore, confl ict with common sense. It should be emphasized, further, that the claims in confl ict concern the category of highly visible macroscopic objects—if there are such things as shmrees they belong to that category. There is, I think, something especially peculiar in philosophers’ disagreeing with common sense with respect to the existence of objects within that category. The situation seems quite diff erent with respect to existential relativism. Suppose we say to ordinary people: “Look at the tree. Do you think that, relative to some conceptual scheme that can’t be expressed in English, there exists something there that gains and loses branches?” The ordinary response will surely not be to answer this question in the negative, but rather to ask what the words “there exists something relative to some conceptual scheme that can’t be expressed in English” means. If we, as existential relativists, give an affi rmative answer to the question, therefore, we are not saying anything clearly against common sense. That is, I think, how it should be: philosophers have every right to say things that ordinary people, unversed in philosophy, do not understand, but they should try to avoid saying things that ordinary people do understand and regard as absurd. Let me now say why I rejected Sosa’s use of the word “absolutism” to express the explosionist position. Sosa’s terminology seems confusing because philosophers who believe in “the absolute philosophical truth” need not be explosionists. Eliminativism, a position that Sosa himself cites, is one example of a belief in “absolute truth” being combined with the denial of explosionism. Another example is van Inwagen’s view that the only composite things are living things. I take it that to believe in the “absolute philosophical truth” in the sense relevant to the present discussion is to believe that, amongst the various conceptual schemes and selection functions that we seem able to make intelligible to ourselves, one is somehow uniquely privileged, uniquely right in some sense. Suppose, now, that we believe in the “absolute truth” in this sense. I suppose we are then left trying to come up with the best philosophical hypothesis, the best guess, as to what that truth is. The simplicity and hence non-arbitrariness of explosionism may seem quite signifi cant on this supposition. Explosionism, it may be argued, is, because of its simplicity, the best bet, the one most 138 quantifier variance and realism: essays in metaontology reasonable to accept. The appeal to simplicity, therefore, may be a good argument for explosionism, given that one is already committed to the “absolute truth.” The appeal to simplicity, however, cannot itself support this commitment. As I argued before, the existential relativist can acknowledge the special simplicity and non-arbitrariness of the explosionist’s scheme but not be moved by this to regard that scheme as expressive of the “absolute truth.” Believers in the “absolute truth” may be led to explosionism, or they may be led somehow to other positions, such as van Inwagen’s. One thing that seems quite clear is that they will not be led to the ontology of common sense. Our common-sense selection function, as far as one can make it out, seems to be an amorphous and intractably complex mess, containing in all likelihood disjunctive conditions and grue-like exceptions. How could that possibly be the uniquely correct selection function? It follows that if we want to be able to defend our common-sense ontology we need to reject the “absolute truth.” Our common-sense ontology must be seen as one scheme amongst many, far from the simplest, but still able to express in its own way the same facts expressed by the others. It begins to look, therefore, as if existential relativism—or perhaps some variation of it—may be a requirement for common-sense philosophy. The connection goes in the other direction as well: once we are committed to existential relativism there is no good motive to repudiate our common-sense ontological beliefs, complex and messy though they may be.

II. Existential Relativi sm and Quantifi er Relativism

What I next want to talk about is the relationship between existential relativism and Putnam’s “conceptual relativism.” The terminology here has evidently become diffi cult, since Sosa is now using “conceptual relativism” as the name for a position that includes explosionism, which is certainly not a position encompassed by Putnam’s relativism. I am therefore going to refer to Putnam’s relativism in this chapter as “quantifi er relativism.” The motivation for this terminology will, I hope, become apparent. Let me introduce quantifi er relativism by quoting a couple of passages from Putnam. “The logical primitives themselves, and in particular the notions of object and existence, have a multitude of diff erent sosa’s existential relativism 139

[possible] uses rather than one absolute ‘meaning’.”3 “All situations have many diff erent descriptions, and . . . even descriptions that, taken holistically, convey the same information may diff er in what they take to be ‘objects.’. . . [T]here isn’t one single privileged sense of the word ‘object’. . . , but there is only an inherently extendible notion of ‘object.’”4 Putnam is saying that diff erent languages, diff erent conceptual schemes, might operate with diff erent selection functions, and hence with diff erent concepts of what it means for there to exist an object. In his surrounding discussion Putnam makes it clear that he is not merely saying that the same unrestricted concept of existence, operating within every language, might be contextually restricted in diff erent ways. Rather he is saying that diff erent languages might possess diff erent concepts of “unrestricted existence.” In the fi rst quote from Putnam I myself inserted the word “possible.” There are really two claims here: fi rst, that the quantifi er expressions in plain English (or some other actual language), when they are in no way contextually restricted, have multiple senses; second, that there is the possibility (if not the actuality) of languages in which the contextually unrestricted quantifi er has a diff erent sense from that of plain English. It is the second claim that I am here calling quantifi er relativism. The fi rst claim might follow from the vagueness of our unrestricted quantifi ers, which I in fact believe in, but I am not going into that here.5 The diff erence between existential relativism and quantifi er relativism is that the former makes an object-level claim about the world, whereas the latter makes a meta-level claim about diff erent possible languages or conceptual schemes. Quantifi er relativists are not saying that diff erent objects exist relative to diff erent conceptual schemes; rather, they are saying that diff erent concepts of an “object” might be employed in diff erent conceptual schemes, schemes that are all adequate for describing the world. In each scheme, however, the relevant descriptions would take the non-relativized form, “There exist such- and-such objects,” not, as in existential relativism, the form, “There exist such-and-such objects, relative to scheme S.”

3. Hilary Putnam, “Truth and Convention: On Davidson’s Refutation of Conceptual Relativ- ism,” Dialectica 41 (1987): 71. 4. Hilary Putnam, “The Question of Realism,” in Words and Life (Cambridge, MA: Harvard University Press, 1994): 304–5. 5. See my “The Vagueness of Identity,” in Philosophical Topics 26 (1999). 140 quantifier variance and realism: essays in metaontology

Sometimes philosophers initially react to quantifi er relativism by regarding it as trivial. “Isn’t it trivial that we could use the words ‘the existence of an object’ in any way we please?” Of course that is trivial; we could use those words to assert, for instance, that it is raining outside. That is not Putnam’s point, however. Putnam is saying that there could be a language whose (contextually unrestricted) quantifi er expressions do not have the same sense as those of English, so that the people who employ that language (as their primary language, as the language “in which they think”) could be said to operate with a diff erent concept of “the existence of an object” than the ordinary one, and yet they can describe the world as accurately and truthfully as we do. To appreciate how untrivial this claim is, one should consider how it relates to the correspondence theory of truth. One classic understanding of that theory might be put roughly as follows:

The referential correspondence theory of truth. In any possible language the truth of a statement depends on the referential relations between its (non-logical) words and objects that exist in the world.

Another way to express this theory of truth is by saying that any possible language must admit of a standard referential semantics. If we are quantifi er relativists it seems we must reject this theory of truth. Sup- pose we (who are speaking plain English) are considering the truth within the shmree-scheme of the statement, “There exists an object in the yard that is brown and wooden and that keeps gaining and losing branches.” Without going into the details, it seems clear that we cannot (in plain English) give a standard referential semantic analysis of what makes the sentence true in the shmree-language. Such an analysis would require us to say that there is a suitable object that is referred to by the terms “brown,” “wooden,” and “gains and loses branches.” What could that object be? From the perspective of our own scheme there is no such object for these terms to refer to in a suitable truth-making way. The referential correspondence theory requires that it be possible to match up the reference of the terms and quantifi er expressions of our own language to those of any other describable language; that is what is required to give in our own language a referential analysis of the semantics of another language. Quantifi er relativism claims that sometimes this will not be possible. Since the shmree-language embod- ies a diff erent concept of “the existence of an object” we cannot in terms of our own concept of “the existence of an object” describe the sosa’s existential relativism 141 truth of sentences in that language as deriving from the referential relations between its words and objects that exist in the world. Quantifi er relativists must reject the referential correspondence theory of truth, but they still have available to them another perennial version of the correspondence theory:

The factive correspondence theory of truth. In any possible language the truth of a statement depends on the world’s being the way the statement says the world is. Another way to put this is that the truth of a statement must depend on its correspondence to the facts. We recall Sosa’s explanation that the same facts (in one intuitive sense of the word “facts”) can be grasped from diff erent conceptual perspectives. Quantifi er relativists can hold, therefore, that, whichever concept of “the existence of an object” we adopt, our concept of “a fact,” our concept of “the way the world is,” remains the same, and it is this concept that defi nes the nature of truth for any conceptual perspective. (I will not at present attempt to explain Putnam’s own attitude—apparently negative—towards the factive correspondence theory.) A question might now be posed as follows: If we are quantifi er relativists we claim that the expression “there exists an object,” as used in the shmree-scheme, does not mean that there exists an object, that is, it does not mean what it means in our language. Why, then, it might be asked, should we even call this expression, as it functions in the shmree- language, a “quantifi er?” And why, then, should we say that speakers of the other language are operating with a diff erent concept of “the existence of an object,” rather than saying that they operate with no concept of “the existence of an object?” I think that quantifi er relativists should readily concede the force of this question. Indeed, we could just as well express the basic idea of quantifi er relativism by saying that the expression “there exists an object” functions in the shmree-language in some ways like a quantifi er, and it expresses in that language a concept in some ways like our concept of “the existence of an object.” We are merely trying to indicate these similarities when we say that the expression is “their quantifi er” and that it expresses a “diff erent concept of ‘the existence of an object.’” The similarities in question are partly a matter of formal logic, since we are tacitly assuming in our construction of the shmree-language that all of formal quantifi cational logic will remain intact in that language. It seems on an intuitive level, however, that there are “holistic” similarities that go beyond purely formal matters. When we refl ect on how the 142 quantifier variance and realism: essays in metaontology word “shmree” functions in the other language we are inclined to say that the word does something like referring to an object. We know that (in terms of our concept of an object) there isn’t any suitable object for the word to refer to, but it is in some sense as if there were an object being referred to, a brown and wooden object in the yard that keeps gaining and losing branches. Of course, to say that it is “as if” there is such an object, though there is no such object, is only a way of expressing our sense that the word “shmree” functions in the other language in some ways similar to how our word “tree” functions in ours. Whether it is even in principle possible to spell out the relevant holistic similarity seems doubtful. How should we now say that quantifi er relativism is related to existential relativism? Consider, fi rst, the following possible analogy. It is correct to make the meta-level statement, “In base 5, ‘The sum of 3 and 2 is 10’ is a true sentence.” I think most people would, however, regard it as merely a use-mention confusion to try to make the object-level statement, “Relative to base 5, the sum of 3 and 2 is 10.” Is the existential relativist committing the same confusion? The quantifi er relativist makes the correct meta-level statement, “In the shmree-scheme, ‘There exists an object in the yard that keeps gaining and losing branches’ is a true sentence.” Are existential relativists committing a use-mention confusion when they make the object-level statement, “Relative to the shmree-scheme, there exists an object in the yard that keeps gaining and losing branches?” Existential relativists have a response, however. They will say that the cases are very diff erent. We have no trouble describing, from our ordinary base 10, how other bases function. In base 5, we say, the expression “10” refers to the number 5, the expression “20” refers to the number 10, and so on. There is, therefore, no reason here to move to an object- level description, which would indeed be merely a use-mention confusion. In the case of alternative concepts of “the existence of an object,” however, we cannot readily describe how these alternative concepts function. In the shmree-scheme the word “shmree” seems to have the function of referring to a certain kind of object, but, in terms of our ordinary scheme, there is no such object. We are, therefore, pushed in this case to make the object-level claim that, relative to the shmree-scheme, there is in fact the required sort of object. I can accept this explanation, but only in a certain spirit. My inclination is to think that there is at bottom no genuinely substantive dif- ference between quantifi er and existential relativism. The central insight of both forms of relativism is that there are many possible sosa’s existential relativism 143

selection functions, many possible perspectives on “the existence of objects,” which all are adequate for truthfully describing the same facts, the same “way the world is.” None of these schemes is, therefore, metaphysically privileged; none of them qualifi es as somehow presenting the uniquely “absolute truth,” the “strict and philosophical truth.” (This point holds, we recall, even for the explosionist scheme, which the relativist regards as merely one scheme amongst many—albeit, perhaps, an especially simple one—for describing the facts.) This basic relativist insight is fi rst formulated in the meta-level terms of quantifi er relativism. That formulation attempts to remain close to plain English but strains to make itself intelligible. “In the shmree-scheme,” the quantifi er relativist says, “it is in some sense as if there exists a suitable object for the word ‘shmree’ to refer to.” This awkward and stammering formulation is then converted by the existential relativist into the object-level formulation, “Relative to the shmree-scheme, there does exist a suitable object.” I am viewing existential relativism as essentially a technical formulation of something that is hard to put in plain terms. The existential relativist departs from plain English, but does so for the purpose of presenting the basic relativist insight in a more vivid and intuitive manner. As so viewed, quantifi er relativists and existential relativists need have no disagreement with each other.

I want to conclude by commending existential relativism to Ernest Sosa. This was an important idea that I wish he had not abandoned so quickly. Certainly, I must regret his reversion to the “strict and philosophical truth” of explosionism. Existential relativism and its close cousin quantifi er relativism deserve to be elaborated, if for no other reason than that they appear to provide the essential philosophical backdrop for a defense of the ontological judgments of common sense. nine c C

physical-object ontology, verbal disputes, and common sense

Two main claims are defended in this paper: fi rst, that typical disputes in the literature about the ontology of physical objects are merely verbal; second, that the proper way to resolve these disputes is by appealing to common sense or ordinary language. A verbal dispute is characterized not in terms of private idiolects, but in terms of diff erent linguistic communities representing diff erent positions. If we imagine a community that makes Chisholm’s mereological essentialist assertions, and another community that makes Lewis’s four-dimensionalist assertions, the members of each community speak the truth in their respective languages. This follows from an application of the principle of interpretive charity to the two communities.

My central claim in this paper is that many familiar questions about the ontology of physical objects are merely verbal. Nothing is substantively at stake in these questions beyond the correct use of language. A derivative claim is that, since they are verbal, the proper way to resolve these questions is by appealing to common sense or ordinary language. The fi rst claim is evidently connected to Carnap’s famous distinction between “internal” and “external” questions, but I’m not sure how close the connection is.1 Although it’s not my aim to engage here in Carnapian exegesis, let me mention one immediately important diff erence between my approach and Carnap’s. Whereas Carnap evidently intended his distinction to apply to all issues of ontology, including those involving abstract things such as sets and properties, my claim about verbalness is restricted to questions about the existence and identity of highly visible physical objects.

1. Rudolph Carnap, “Empiricism, Semantics, and Ontology,” in Meaning and Necessity, 2nd edition (University of Chicago Press, Chicago, 1956).

144 physical-object ontology, verbal disputes, and common sense 145

I’ll later say something briefl y about why the claim may not be extendible to the wider range of ontological issues, but this question must remain largely for a separate discussion. It should be understood, furthermore, that throughout this paper I make the assumption, which I think is common in disputes about physical-object ontology, that all sides to a dispute have at their disposal constructions involving sets (or properties). Questions about the ontology of physical objects have been prominent in the recent literature. Some of the doctrines that have been most frequently debated include the following: Nihilism. There are no composite objects.2 Quasi-Nihilism. Some few composite objects exist, including persons and perhaps some other living things, but there are no tables, ships, moun- tains, rivers, planets, pebbles, leaves, eyes, or almost any other of the variety of composite objects that people ordinarily seem to be talking about.3 Mereological Essentialism. An object cannot persist with any of its parts replaced.4 The doctrine of mereological sums. Any two objects compose an object. The doctrine of temporal parts. If an object persists through an interval of time, there is a temporal part of the object that exists only during that interval and that spatially coincides with the object during that interval. Four-dimensionalism. This is the conjunction of the doctrines of mereological sums and temporal parts. It implies that, if we start with the objects ordinarily talked about, any sum of temporal parts of these objects, however discontinuous or gerrymandering, constitutes an object on a par with them.5 A fi nal position that is often ignored in the current literature is:

2. Nihilism is discussed (and rejected) in Peter van Inwagen, Material Beings (Cornell University Press, Ithaca, N.Y., 1990), ch. 8. Nihilism may be suggested in Peter Unger, “There Are No Ordinary Things,” Synthese 41 (1980), 117–54. The dispute over nihilism that I view as verbal occurs when both sides agree that there are simples and disagree about whether there are composites. As Theodore Sider pointed out to me, my arguments in this paper do not seem to imply that the question whether there are simples is verbal—a question which, it may be noted, is not about the existence of highly visible objects. Cf. Sider’s “Van Inwagen and the Possibility of Gunk,” Analysis, 53 (1993), 285–89. See also note 29, below. 3. Two quasi-nihilists are van Inwagen, Material Beings, and Trenton Merricks, Objects and Persons (Oxford University Press, Oxford, 2001). 4. Roderick Chisholm, Person and Object (Open Court, LaSalle, Illinois, 1976), ch. 3; reprinted in J. Kim and E. Sosa, eds., Metaphysics: An Anthology (Blackwell Publisher Ltd., Oxford, 1999); James van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time,” Midwest Studies in Philosophy XI (1986). 5. I adopt the expression “four-dimensionalism” for the highly popular and infl uential conjunction of the two doctrines, but the reader should be aware, fi rst, that some authors use the expression for the doctrine of temporal parts, without regard to the doctrine of mereological sums, and, second, neither doctrine really has much to do with the intuitive idea of “four dimensions.” Most proponents 146 quantifier variance and realism: essays in metaontology

Common sense ontology. This denies all of the above doctrines. The composite objects we ordinarily talk about really exist; they typically persist through changes in their parts; they typically do not have sums; and they typically do not have temporal parts.6 According to my fi rst claim the dispute between these various positions is purely verbal, and this implies, according to my second claim, that the position of common sense ontology must be correct.

L et me begin by saying something about what I mean by a verbal dispute. The following is a fairly simple example. I know someone, whom I’ll call A, who claimed that a standard drinking glass is a cup. “Just as a cat is a kind of animal,” she said, “a glass is a kind of cup.” Everyone else whom I’ve asked about this agrees with me that a glass is not a cup. Clearly, this dispute is, in some sense, merely about language. It’s tempting to try to elaborate the sense in which this is so by saying that A’s assertion is true in her idiolect, and mine is true in mine. This appeal to private idiolects may, however, have the drawback of suggesting that, in asserting what she does, A does not express a false belief or thought about cups and glasses. An infl uential view of Tyler Burge may imply, on the contrary, that A’s beliefs and thoughts about cups and glasses are determined by what her asserted sentences mean in the public language, not in A’s alleged private idiolect.7 I intend to skirt this entire of either of these doctrines also accept the other and are therefore four-dimensionalists in the defi ned sense. Two exceptions are Van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time,” and Judith Jarvis Thomson, “Parthood and Identity Across Time,” Journal of Philosophy 80 (1983), 201–20; reprinted in Kim and Sosa, eds., Metaphysics: An Anthology. Four-dimensionalists include W. V. Quine, Word and Object (The MIT Press, Cambridge, MA, 1960), esp. p. 171; David Lewis, On the Plurality of Worlds (Basil Blackwell. Oxford, 1986), esp. pp. 202–4, 211–13; Mark Heller, The Ontology of Physical Objects: Four Dimensional Hunks of Matter (Cambridge University Press, Cambridge, 1990); chapter 1 reprinted in Kim and Sosa, eds., Metaphysics: An Anthology; Theodore Sider, Four-Dimensionalism: An Ontology of Persistence and Time (Oxford University Press, Oxford, 2002). An argument from the doctrine of mereological sums to temporal parts is given in Richard Cartwright, “Scattered Objects,” in K. Lehrer, ed., Analysis and Metaphysics (Reidel, Dordrecht, 1975); reprinted in Kim and Sosa, eds., Metaphysics: An Anthology. 6. A highly original formulation of common sense ontology is given in Ned Markosian, “Brutal Composition,” Philosophical Studies 92 (1988), 211–49. 7. Tyler Burge, “Individualism and the Mental,” Midwest Studies in Philosophy IV (1979). Since I’m not clear about the intended scope of examples to which Burge’s idea applies, I’m not entirely confi dent that it applies to the trivial example of the glass and cup. Certainly it would apply to other examples I’ll consider in this paper. physical-object ontology, verbal disputes, and common sense 147 issue. I think we can bring out a relevant sense in which the dispute is merely a matter of language without getting involved with questions about private idiolects. Let’s assume that all of the disputes considered in this paper are intended to take place in plain English. Each disputant claims to be speaking the truth—that is, the strict and literal truth—in plain, nontechnical English. Now let’s imagine a linguistic community which is in all other ways as close as possible to our actual English-speaking community but in which everyone agrees with A. We’ll call this the A-community. To say that members of the A-community agree with A means, roughly, that they accept all of the disputed sentences which A accepts. They will accept such general sentences as “A glass is a kind of cup.” With respect to perceptual sentences, involving demonstratives or indexicals, something a bit more complicated has to be imagined. The basic idea is that they say the same things A would say in relevantly similar perceptual circumstances. In the A-community anyone shown a standard drinking glass will assent to the sentence “This (here) is a cup.” This characterization of the A-community is not completely precise, but I think it’s good enough for my purposes. By A-English we’ll mean the language that would most plausibly be attributed to the imagined A-community. I think it’s obvious that in A-English the sentence “A glass is a cup” is true. In that language the word “cup” denotes, roughly, any vessel designed for drinking, that is, all the things we call “cup” plus drinking glasses.8 My dispute with A is verbal because the disputed sentences asserted by A are true in A-English, and, by the same token, the disputed sentences asserted by me are true in the language corresponding to my position. The only real question at issue is which language is (closest to) plain English. I take it that the answer to that question is that the language corresponding to my position is plain English. That’s the sense—the only sense— in which I’m right and A is wrong. In this kind of example it seems obvious that the proper way to resolve a verbal dispute is by appealing to common sense or ordinary language. If Burge’s view applies to this example, it implies that A is not speaking (or thinking in) A-English (or in a corresponding private idiolect); rather A has the mistaken thoughts and beliefs that are expressed by her assertions in plain English. I’m not entering into that

8. Throughout this paper I will allow a single expression to have diff erent meanings in diff erent languages. In another sense we can say that the diff erent languages contain distinct expressions that are phonetically and syntactically indistinguishable. 148 quantifier variance and realism: essays in metaontology question. What is important for my purposes is that the sentences asserted by A are true in A-English, so that the only real question is whether A-English is plain English. This is why the dispute with A is merely verbal. In eff ect I am redefi ning “A’s idiolect” to mean the (imagined) public language associated with A’s position. This redefi ned sense of “A’s idiolect” captures the relevant sense in which “A is right in A’s idiolect (and we are right in ours).” I say that on the most plausible interpretation of the A-language the sentence “A glass is a cup” is true in that language. By the “most plausible” interpretation I mean one that is permanently most plausible; I’m taking it as given that the members of the imagined A-community are not going to change their linguistic behavior in any relevant way. Might it nevertheless be the case that the most plausible interpretation is not the correct one? Is it possible that, contrary to all evidence, the members of the A-community really mean by “cup” what we mean, but they have for some reason the intractable inclination to falsely judge that glasses are cups? That makes no sense, I think, for reasons related to the “Wittgenstein paradox” discussed by Kripke.9 But that’s another issue that I’m not entering into here. I’ll simply assume in what follows that once we agree on the most plausible interpretation of a language, no further questions will be raised about whether that is the correct one. But why is it plausible to suppose that in the A-language the word “cup” doesn’t mean what it means in our language, so that the sentence “A glass is a cup” is true in that language? The basic answer to this question comes out of a widely accepted principle of linguistic interpretation that has often been called the “principle of charity.”10 This principle, put very roughly, says that, other things being equal, an interpretation is plausible to the extent that its eff ect is to make many of the community’s shared assertions come out true or at least reasonable. As Davidson says, interpreting a language is part and parcel of explaining people’s behavior and psychology, and such an explanation is, other things being equal, simpler and more plausible if it depicts people as

9. Saul A. Kripke, Wittgenstein on Rules and Private Languages (Harvard University Press, Cambridge, MA, 1982). 10. Quine, Word and Object, p. 59; Donald Davidson, Inquiries into Truth and Interpretation (Oxford University Press, Oxford, 1984), esp. “Belief and the Basis of Meaning” and “Thought and Talk”: David Lewis. “Radical Interpretation” in Philosophical Papers I (Oxford University Press, N.Y., 1983), and “New Work for a Theory of Universals,” Australasian Journal of Philosophy 61 (1983), 343–77, at pp. 370–77. physical-object ontology, verbal disputes, and common sense 149 reacting in some reasonable way to the facts they confront.11 If we tried to interpret the word “cup” as meaning in the A-language what it means in our language, we would have to depict the A-speakers as inexplicably making false and unreasonable judgments about cups. The principle of charity tells us to avoid this implausible result by interpreting the word “cup” diff erently. The general idea of interpretive charity is to make the community’s assertions come out as far as possible true or reasonable. I think it’s obvious, however, that some assertions demand more charity than others. Perceptual assertions are evidently central to linguistic acquisition and interpretation. What I’ll call “charity to perception” is the very strong presumption that any language contains sentences used to make perceptual reports, and that these reports are generally accurate (to a fair degree of approximation), especially when they are widely accepted in the community. It’s hard to imagine any procedure for interpreting a language that does not take charity to perception very seriously. If we interpreted the word “cup” as meaning in the A-community what it means for us, we would regard as false such perceptual reports as, “Here is a cup,” said with respect to a glass. Charity to perception urges us to avoid such an interpretation. One other salient kind of charity that I want to mention is what I’ll call “charity to understanding.” Certainly there must be the strong presumption that typical speakers of a language have a suffi ciently adequate grasp of their linguistic and conceptual resources so that they don’t generally make a priori (conceptually) false assertions, especially when these assertions seem to be relatively simple, not ostensibly involving any complicated calculations or computations. If “cup” were interpreted in the A-language to mean what it means in our language, the A-speakers would inexplicably make such a priori false assertions as, “A glass is a cup.” Worst of all, they would make a priori false perceptual assertions, such as, “There’s one cup here, and it’s a glass.” An interpretation that simultaneously violates both charity to perception and understanding can typically not be taken seriously. The general characterization of a verbal dispute is one in which the controversial sentences are most plausibly interpreted as having diff erent truth conditions in the diff erent languages associated with the contending positions, so that each position turns out to be correct in its associated language. In the sense intended throughout this discussion sentences have the same truth conditions if (relative to the same

11. Davidson, Inquiries into Truth and Interpretation, esp. pp. 159–60. 150 quantifier variance and realism: essays in metaontology

context of utterance) they hold true in the same possible worlds. The verbalness of a dispute may turn on the reference of a term such as “cup,” or it may turn on the meaning of logical constants—as in a verbal dispute as to whether “Either John or Mary is ill” is true when both John and Mary are ill—or any semantic or syntactic feature of language that aff ects the truth conditions of sentences. The primary focus is always on whole sentences and how to assign truth conditions to them in the most charitable way possible. When I speak throughout this paper about interpreting a language this is always to be understood in the narrow sense of assigning truth conditions. I leave it open what there is to understanding a language beyond knowing the truth conditions of its sentences, but, whatever this additional element may be, it will have a bearing on my argument only insofar as it might aff ect the plausibility of certain truth-condition assignments. These points will be important to bear in mind when we consider later the diff erent ways to assign charitable truth conditions to sentences in the ontological cases. Charity in the sense I’ve explained—that is, the presumption that sentences widely accepted by the linguistic community are true or reasonable—is not the only interpretive consideration mentioned in the literature. Considerations of semantic compositionality, complexity, and property-naturalness may play a role, and will be addressed later. A point that I want to stress at present is that most disputes, whether in ordinary life, in science, or in mathematics are substantive rather than verbal. It’s essential to realize that if we pick almost any intuitively substantive dispute, and set ourselves the task of fi nding some plausibly charitable assignment of truth conditions that will make both sides come out right in their associated languages, we generally fi nd that we have not the faintest idea of how to proceed. We are emphatically not merely looking for “models” here (in the sense of Putnam’s “model-theoretic argument”).12 We are looking for genuinely plausible truth-condition assignments that make the most charitable sense of what members of a community say, and especially the most charitable sense of how what they say rationally refl ects their perceptions (hence, their causal connections to their environments) and their understanding. I repeat that if we consider virtually any familiar question from outside philosophy—the question whether God exists, whether there is extra-terrestrial life, whether Goldbach’s conjecture

12. Hilary Putnam, Reason, Truth, and History (Cambridge University Press, Cambridge, 1981), ch. 2. physical-object ontology, verbal disputes, and common sense 151 is true—we fi nd that we have not the faintest idea of how to plausibly interpret both sides of the questions as coming out right in their associated languages.13 In the minority of cases where this is possible we have a merely verbal dispute. Leaving aside various complications for the moment, I want to mention another fundamental element of interpretive charity that in many relatively easy cases immediately settles that a dispute is substantive. This element, which I’ll call “charity to retraction,” is illustrated in the following case. My friend B claims, “When a ball is thrown into the air at a certain speed, it hits the ground at a much greater speed.” I say it hits the ground at roughly the same speed. This is surely not a verbal dispute. If I imagine the B-community, in which everyone accepts the disputed sentences that my friend accepts, I’m obviously imagining a community that is ignorant of basic physics, not a community that asserts true sentences in a diff erent language. If I tried to interpret B-English so as to make the community’s assertions come out right, I would fi nd myself quickly faced with cascading complications, but even apart from that, in the present example there is an obvious reason not to seek any such interpretation. My friend B may suff er from a perversely irrational confi dence in his untutored physical intuitions, but, like most people, he is by no means beyond the reach of reason. There are any number of experiments that, if he confronted them, would get him to retract his original assertion. We imagine the members of the B-community as behaving in the same way. When we try to interpret a language in a manner that is charitable to what people say, we need to take into account not just what they actually have said, but what they will or would say in the face of additional evidence. If we tried to interpret the sentence “It hits the ground at a much greater speed” as being true in B-English, we would imply that the B-speakers make a mistake when they are disposed to retract the sentence in the face of additional evidence. This is surely not charitable. A plausibly charitable interpretation must take account of the strong presumption that reasonable people are expected to improve the accuracy of their judgments in the face of additional evidence. This is obviously a

13. When considering the interpretation of a mathematical sentence such as the Goldbach sentence it must be borne in mind that such sentences are always related to other kinds of sentences. A plausible assignment of truth conditions that would make the Goldbach sentence come out false must make the sentence “The number of stars is even (and more than two) and not the sum of two primes” come out true in some worlds and false in others. It can never be simply a matter of formulating a rule—supposing that were possible—that makes some mathematical sentences true (or false) in every world. 152 quantifier variance and realism: essays in metaontology

complicated principle that would ultimately need to incorporate facts about probability and confl icting evidence, but a very simple application of it is “charity to retraction.” Certainly we must, other things being equal, favor an interpretation that makes the community’s retractions in the face of additional evidence come out right. This consideration suffi ces to explain why we must regard the imagined B-community’s assertion, “It hits the ground at a much greater speed,” as false, and why my dispute with my friend B cannot be viewed as merely verbal. Many disputes, both of an a priori and empirical nature, are immediately shown to be substantive by the consideration of charity to retraction. Suppose it’s known that there are eighteen rows of coins on the table, each row containing seven coins. Someone says, “So there are 146 coins,” and I disagree. This person would surely retract his claim if he checked his computations or counted the coins, so this dispute is obviously not verbal. Or take the case of someone who says, “A whale is a big fat fi sh.” Many people disagree. If this dispute is not viewed as merely verbal, one reason might be that this person would retract her claim if she were shown empirical evidence about how whales diff er from other things she regards as fi sh. This example, however, introduces complications that I’ll want to come back to later. Finally, consider my disagreement with a student who starts out by saying that there is no diff erence between doing something by accident or by mistake. Assum- ing that she would retract her claim once she is shown Austin’s famous examples, she is making a substantive a priori (conceptual) mistake, in the sense that she is wrong even in “her own idiolect.”14 If we imagine a community that agrees with her, they would all be wrong, and would realize it if Austin arrived on the scene.15 Perhaps I’ve said enough in this section to provide a preliminary framework within which to start to consider now the ontological examples.

14. J. L Austin, “A Plea for Excuses,” in Philosophical Papers (Oxford University Press, London, 1961), p. 185, note 1. 15. In the initial example, A would presumably retract her claim that glasses are cups if she found out more about how the community uses the word “cup,” but this source of retraction cannot be relevant to what I mean by “charity to retraction (in the face of new evidence),” since there evidently could not be any such retraction if the community agreed with A. Retraction in the sense relevant to interpretive charity must be generalizable to the imagined community that agrees with a speaker, since (in interpreting the speaker’s “idiolect”) it is the language of this community that we are trying to interpret. physical-object ontology, verbal disputes, and common sense 153

III

I’m claiming that the familiar disputes about physical-object ontology that fi ll the current literature are all verbal. Let’s initially focus on the dispute between mereological essentialists and four-dimensionalists. In order to fi x our ideas—since diff erent theorists may present these positions in diff erent ways—let’s take Roderick Chisholm as representing mereological essentialism and David Lewis as representing four- dimensionalism. We imagine, then, the RC-community in which everyone accepts the disputed ontological sentences accepted by Chisholm, leaving everything else as close as possible to our actual community. And the imagined DL-community accepts the sentences accepted by Lewis. RC-English and DL-English are, respectively, the languages spoken in these imagined communities. My claim is that all of Chisholm’s accepted sentences are true in RC-English, and all of Lewis’s accepted sentences are true in DL-English, so that the only real issue is which, if either, of these languages corresponds to plain English. (I’ll argue in a later section that neither does, so that both of these philosophers are making verbal mistakes.) Let’s begin by looking at this from the four-dimensionalist’s standpoint. I want to get these philosophers to agree that all of Chisholm’s mereological essentialist claims are true in RC-English. If we are four-dimensionalists trying to understand what members of the imagined RC-community are saying, we quickly realize that, on the assumption that they mean the same thing by their sentences that we mean, they are frequently making extreme mistakes, both of an a priori and perceptual sort. The following is a representative example. Suppose that in the room there is a brown wooden pencil on the table and a pink rubber ball on the fl oor. We four-dimensionalists will say that the following sentence is true: “There is something in the room that is fi rst brown (and wooden, and cylindrical), and later it— that same thing—is pink (and rubber, and round).” This sentence is true because there are any number of things in the room that are composed of an early part of the pencil and a later part of the ball. But when members of the RC-community are presented with this scenario and are queried about the sentence they adamantly insist that the sentence is false, even though the relevant things are right in front of their eyes. If they mean what we mean, not only are they making a mistaken perceptual report, but they are a priori contradicting themselves, since they admit, “There is a brown (and wooden, and cylindrical) 154 quantifier variance and realism: essays in metaontology thing in the room, and later there is a pink (and rubber, and round) thing in the room,” from which the truth of the sentence they deny follows a priori.16 It seems evident that, as four-dimensionalists, we should try to reject the assumption that the RC-speakers mean what we mean, and instead look for an interpretation of RC-English that respects charity to perception and understanding. Such an interpretation doesn’t seem hard to fi nd. Lewis has pointed out that people often use contextually restricted quantifi ers. For example, someone says, “There is no beer,” where the conversational context indicates that the quantifi er is meant to range only over beer in the fridge, excluding beer elsewhere.17 It seems perfectly intelligible to suppose that there can also be semantically restricted quantifi ers, that is, quantifi ers that, because of the semantic rules implicit in a language, are restricted in their range in certain specifi c ways. If the quantifi ers in a language are semantically restricted, they are always limited in their range, regardless of the conversational context. It seems evident (we should say, as four-dimensionalists) that the quantifi ers in RC-English are semantically restricted, excluding from their range such things as the object that is composed of the early part of the pencil and a later part of the ball. This must be why the RC-speakers reject the sentence, “There is something in the room that is fi rst brown and later pink.” How exactly to characterize the semantic restriction on the quantifi ers might have to be fi ne-tuned, but the rough idea seems to be that the range of the RC-quantifi ers excludes any physical object that is composed of matter but is not itself a mass of matter (in roughly the sense of Locke18). Excluded, therefore, are (proper) temporal parts of a mass of matter and mereological sums of temporal parts of diff erent masses. As four-dimensionalists, we believe that physical objects com- prise (1) masses of matter, (2) objects (such as temporal parts and sums of masses of matter) that are composed of matter but are not themselves masses of matter, and (3) perhaps other objects such as fi elds of energy. The RC-quantifi ers exclude from their range the second kind of physical object. In RC-English the word “(some)thing” is roughly

16. My assumption here is that four-dimensionalists regard the principles of mereological sums and temporal parts as a priori truths. 17. Lewis, On the Plurality of Worlds, p. 3 and pp. 212–13. 18. John Locke, An Essay Concerning Human Understanding (1690), book 2, chapter 27. Ignoring certain complications, let’s assume that mereological essentialism holds for masses in Locke’s sense. Furthermore, we can assume for the purposes of this discussion that the particles of physics count either as masses of matter or as not being composed of matter. physical-object ontology, verbal disputes, and common sense 155 equivalent to “(some)thing that is either a mass of matter or is not composed of matter.” Given that this is what “(some)thing” means in RC-English, it makes perfectly good sense that the RC-speakers will reject the sentence “There is something in the room that is fi rst brown and later pink.”19 Moreover (we ought to continue, as four-dimensionalists), the semantic restriction on the RC-quantifi ers does not prevent the RC- community from adequately describing the physical facts. Chisholm, we will assume, accepts set-theoretical constructions, and so, therefore, do the imagined members of the RC-community. (It would not aff ect matters if some property-theoretic constructions were used in place of sets.) Where we four-dimensionalists talk of sums of temporal parts of objects, the RC-speakers talk of sets of pairs of objects and times. For anything we can say about the world they can evidently say something (a priori necessarily) equivalent. Their description of the physical facts seems therefore to be as adequate as ours. It’s a delicate matter whether distinct facts might be equivalent, but it doesn’t seem necessary to enter into that question. What seems important is that, for any fact we can express, the RC-speakers can express an equivalent fact. The RC-speakers will, of course, make the platitudinous disquotational assertion, “If something exists it is referred to by the word ‘something.’ ” Given what they mean by “something” this sentence is trivially true. We cannot therefore ask the RC-speakers, “Is there a semantic restriction on the RC-quantifi ers?” since that question is merely another form of the question whether there exists such things as mereological sums and temporal parts, a question which has diff erent answers in the diff erent languages. Hence, four-dimensionalists ought to conclude that, on the most plausibly charitable interpretation of RC-English, all of Chisholm’s disputed assertions are true in that language. Let’s now adopt the standpoint of mereological essentialists who are trying to understand what members of the imagined DL-community are saying. We’re faced with the same kind of problem that the four-dimensionalists faced when they tried to understand our language. Members of the DL-community accept the sentence, “There is something in the room that is fi rst brown and later pink,” though

19. One reason why this account of the range of the RC-quantifi ers has to be fi ne-tuned is that, while Chisholm certainly rejects temporal parts of masses, I’m not sure what his view is about mereological sums. Van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time,” is a mereological essentialist who accepts mereological sums. 156 quantifier variance and realism: essays in metaontology there is nothing in the room that is fi rst brown and later pink. Charity to perception and understanding indicates that they must evidently mean something diff erent by this sentence than we mean. Can we make intelligible to ourselves a charitable interpretation of DL-English that makes the ontological sentences that they accept come out true? I think it’s clear that we can, though it may not be clear how exactly to spell out this interpretation. One thing we need to bear in mind is that our primary focus is on the truth conditions of sentences rather than on the reference of terms. We want to assign charitable truth conditions to such DL-sentences as, “There is something in the room that is fi rst brown and later pink.” We need not immediately concern ourselves with what to say about the reference of such DL-expressions as, “fi rst brown and later pink.” As regards the truth conditions of the sentence, “There is something in the room that is fi rst brown and later pink,” it seems that the DL-speakers accept the sentence with respect to any situation in which there is fi rst something in the room that is brown, and later there is something in the room that is pink. The charitable interpretation, then, is that in DL-English a sentence of the form “There is something that is fi rst F and later G” is true with respect to any situation in which there is fi rst something that is F and later there is something that is G. In fact, since there is often no dispute between us and the DL-speakers with regard to sentences of the form “There is fi rst something that is F and later there is something that is G” we can ask the DL-speakers how this undisputed sentence relates to the disputed sentence “There is something that is fi rst F and later G.” They will tell us that the sentences are (a priori necessarily) equivalent. We should believe them. That is, we should make the charitable assumption that in DL-English these sentences really are equivalent, so that the undisputed sentences can be taken as providing us the truth conditions for the disputed sentences in DL- English. There is another way to look at this. It seems obvious that the principles of mereological sums and temporal parts are in some sense central to the DL-community’s linguistic behavior. From our own mereological essentialist perspective we can usefully regard those principles as working in eff ect as semantic rules that generate truth conditions for the disputed sentences. It works like this. Suppose we have a disputed sentence X, and we are trying to decide whether X, as understood in DL-English, is true with respect to some situation or world w. We ask ourselves whether there is some undisputed sentence U, such that U is true with respect to w, and the truth of X physical-object ontology, verbal disputes, and common sense 157 follows (by standard logic) from the conjunction of U and the principles of mereological sums and temporal parts. Sentence X is true with respect to w if the answer to this question is yes. In other words, a disputed sentence is true in DL-English if it follows from the undisputed facts in conjunction with the two principles. An immediate consequence is that the principles themselves are (a priori) true with respect to every possible situation, which is of course what the DL- community says. “But why,” it may be asked, “should we regard the principles as working in eff ect as semantic rules, rather than as false beliefs shared by the members of this community, which lead them to all kinds of other perceptual and a priori mistakes?” I think the question answers itself. Why should we not regard the principles in that charitable manner? By so regarding the principles we make good sense out of why the DL- speakers say what they say, instead of having to assume that they have some incurably irrational tendency to make a priori mistakes about what they perceive in front of their faces. A kind of DL-sentence that may seem especially challenging to us, if we are mereological essentialists, involves the word “reference,” for example, the sentence, “The expression ‘thing that is fi rst brown and then pink’ refers in DL-English to something in the room that is fi rst brown and then pink.” Since, as mereological essentialists, we say that there is nothing in the room that is fi rst brown and then pink, hence no such thing that can be referred to, we certainly cannot accept this sentence (about DL-English) in our language. What we have to say is that this sentence, if understood in DL-English, correctly describes the mentioned expression’s function in DL-English, but, if understood in our language, incorrectly describes the expression’s function in DL-English. Viewing DL-English from our perspective as mereological essentialists it must strike us that, since the sentence “There is something in the room that is fi rst brown and then pink” is true in DL-English, the expression “fi rst brown and then pink” seems in some sense to function in that language as if it refers to something that is fi rst brown and then pink (though, since there is no such thing, the expression cannot really refer to such a thing). One should not ask, “So which is it? Does the expression really refer to something, or does it merely behave ‘as if’ it refers to something?” That question is just the same old question about whether mereological sums and temporal parts exist, in a slightly diff erent guise, and therefore has diff erent answers in the two languages. As goes quantifi cation so goes “reference.” Since the DL-speakers don’t mean what we do by “(there 158 quantifier variance and realism: essays in metaontology exists) something,” they can’t mean what we mean by “reference (to something).”20 A question might be raised, however, as to whether we have presented a “compositional semantics” for the charitably construed DL- English, a semantic analysis, that is, which explains how the truth conditions of sentences are determined by the meanings and ordering of the words in them. Obviously no attempt has been made here to present any formal or rigorous semantics, but, on an intuitive level, the account that has been sketched above does, it seems to me, satisfy the demand for a compositional semantics in the only sense in which such a demand has any clear force. The basic point is that truth conditions cannot be assigned to sentences one at a time. We must have some intelligible way of arriving at the truth conditions of a sentence by looking at its composition. That mereological essentialists do have such a way of arriving at the charitable truth conditions of the sentences of DL-English is shown by the following experiment. Take a group of mereological essentialists. Present them with any series of disputed sentences paired off with undisputed sentences, and ask them whether four-dimensionalists will say that the paired sentences are equivalent. The mereological essentialists will agree on the answers to these questions, and their answers will be right. That shows that they have a general way—whether or not they can spell it out rigorously—for charitably interpreting the truth conditions in DL-English of the disputed sentences: they need only take the charitable truth conditions of the disputed sentences to be given by the undisputed sentences that the four-dimensionalists would regard as equivalent. I think it would, in fact, be interesting to see how a philosopher might try to argue that a charitable interpretation of DL-English, which makes the disputed sentences accepted by the DL-speakers come out true, is unintelligible because no compositional semantics could be given for such an interpretation. The burden would fall on this philosopher, fi rst, to explain what the required compositional semantics is, second, to show that it’s not possible to provide such a semantics for the charitably interpreted DL-English, and, third and by far most important, to explain why we should think that an interpretation is unintelligible just because it resists a particular kind of semantic

20. I discuss “as if reference” further in Dividing Reality (Oxford University Press, Oxford, 1993), pp. 12–13, 188; “Quantifi er Variance and Realism,”Philosophical Issues, 12 (2002), pp. 55–56; and “Sosa’s Existential Relativism,” in J. Greco, ed., Ernest Sosa and His Critics (Blackwell Publishing Ltd., Oxford, 2004), pp. 231–32. physical-object ontology, verbal disputes, and common sense 159 analysis.21 It may be suffi ciently clear that, to the extent that we cannot provide a certain kind of compositional semantics for some language (or, perhaps, to the extent that we cannot provide a word-for-word translation from some language into ours), we will be unable to straightforwardly express in our terms the fi ne-grained intentional content of some of the assertions made in that language. What needs to be shown is the bearing this point might have on my argument. I am claiming that if we (as mereological essentialists) consider the linguistic behavior of people in the DL-community, we are obliged to assign charitable truth conditions to their asserted sentences, and to conclude, if need be, that some of their intentional states may not be fully expressible in our own language. I’m not insisting that there can’t be an argument against this. Let’s see someone produce the argument, and then we can try to assess it. In applying charity to the interpretations of RC-English and DL- English in the above discussion I emphasized charity to perception and understanding. What about charity to retraction? Should that also play a role here? Of course ontologists do occasionally retract their positions, but, as Lewis remarks, a stage seems eventually to be reached in ontology when “all is said and done,” when “all the tricky arguments and distinctions and counterexamples have been discovered,” so that each position has achieved a state of “equilibrium.”22 I’m assuming that in the ontological disputes under discussion the “all is said and done” stage has been reached. In imagining the RC and DL-communities, therefore, we imagine that the members of this community are not disposed to retract their ontological assertions in the face of additional evidence or arguments. Since charity to retraction plays no role the verbalness of the dispute is especially clear in this kind of case. Lewis’s view is that when we have reached the “all is said and done” stage we are left with a “matter of opinion” in which one side “is making a

21. It seems obvious that, if we are mereological essentialists, we can’t do a straight-out Tarskian referential semantics for the charitably interpreted DL-English, but no one, I think, has ever claimed that such a semantics is feasible for any natural language without the addition of various equivalence transformations. See Davidson, Inquiries into Truth and Interpretation, pp. 29–30; W.V. Quine, The Roots of Reference (Open Court, Illinois, 1973), pp. 93–95. We might attempt to assign truth conditions to the disputed sentences in DL-English by way of equivalence transformations from the undisputed sentences, treating the latter the same way in DL-English as in our language. Furthermore, to whatever extent we mereological essentialists can provide a referential semantics for our language, the DL- speakers can as easily provide for their language what they (but not we) will call a “referential” semantics. See also Kripke, Wittgenstein On Rules and Private Languages, pp. 71–72, note 60. 22. David Lewis, Philosophical Papers, Volume 1 (Oxford University Press, Oxford, 1983), p. x. 160 quantifier variance and realism: essays in metaontology mistake of fact.”23 That, I am saying, is the wrong way to view the matter. Prior to the “all is said and done” stage, when retraction is still a live option, ontological disputes may not be verbal in the intended sense. I’ll argue in section V, however, that, since these disputes are verbal after all is said and done, any retractions ought to be in the direction of the common sense position. Let me say something about how the “all is said and done” stage is reached. It’s useful to roughly delineate a special kind of disputed sentence that fi gures in ontology, which I’ll call “ontological axioms.” The way it typically works in ontological disputes is that each camp will try to defend its position by appealing to its favored axioms. The axioms themselves can’t be defended—they are, so to speak, the bottom line for each camp—but they can sometimes be eff ective in arguments, leading some people to retract their positions. The reason for this is that the axioms are often tricky. Their connection to the other disputed sentences, which I’ll call the “(plain) ontological sentences,” may not be immediately transparent. The latter sentences describe in fairly straightforward terms what objects exist in the world or some perceived part of the world. It may happen that some people who accept the plain ontological sentences of one camp are seduced into committing themselves to the axioms of another camp, only to realize too late that they now have no option but to switch camps.24 One of the favorite axioms of the mereological essentialists is: “Two things cannot spatially coincide at any time.”25 This supports the correctness of mereological essentialism in the following tricky way. Suppose an object can persist while losing a part. It used to be larger, but the mass of matter that now composes it was always the same size. So it can’t be identical with the mass of matter, though they spatially coincide. Someone who starts out blithely assuming that a car can survive the change of a tire, but then innocently accepts the “no-coinciding-things” axiom, because he doesn’t sense any danger, may soon be made to realize by the pouncing mereological essentialists that he has gotten himself into a problem.

23. Ibid., p. xi. 24. Some of the examples that I give of “axioms” may not be viewed by their proponents as the bottommost line, but they are very close to the bottom, and it’s the way they support the plain ontological sentences that most concerns me. 25. See Van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time,” p. 149. Van Cleve’s use of the principle is actually far more subtle than the argument I’m about to sketch. physical-object ontology, verbal disputes, and common sense 161

The “all is said and done” point has been reached when ontologists have gone around the dialectical block enough times to feel secure that they are prepared to reject any axioms that might undermine their ontological assertions. The DL-community is therefore to be imagined as rejecting the “no-coinciding-things” axiom, which evidently cannot be reconciled with the four-dimensionalist’s world of overlapping and criss- crossing space-time chunks. Mereological essentialists ought to agree that this rejection is perfectly reasonable in the DL-language. The charitable interpretation of the language that made the sentence “There exists something that is fi rst brown and then pink” come out true will also make the “no-coinciding-things” axiom come out false. I’ll mention a number of other axioms later, but the general lesson that I hope is reasonably clear is that the presence of the axioms does not complicate in any essential way my argument for the verbalness of the ontological disputes. If we take any two opposing camps, the members of each camp will be able to fi nd a plausibly charitable interpretation of the language associated with the other camp which makes its ontological sentences come out true and any axioms that threaten these sentences come out false. Both the question about the plain ontological sentences and the question about the axioms are merely verbal.

I want to briefl y explain in this section why my approach may diverge from Carnap’s treatment of “interna l” and “external” questions. Con- sider the dispute between a platonist who accepts sets (or properties or numbers) and a nominalist who rejects all such abstract items. Carnap holds that this dispute simply turns on the choice of a language. The dispute does not, however, appear to be verbal in the sense I’ve been discussing. I consider a dispute verbal only if it satisfi es the following condition: Each side ought to acknowledge that there is a plausibly charitable interpretation of the language associated with the other side’s position which will make that position come out true. It does not appear, however, that nominalists can acknowledge that there is any such interpretation for the platonists’ position. For example, platonists will regard the following sentence as contingent, that is, as true in some worlds and false in others: “There are two (perhaps infi nite) 162 quantifier variance and realism: essays in metaontology sets X and Y, whose members are (perhaps infi nite) sets of persons, satisfying the condition that, for any set X′ in X, there is a set Y ′ in Y such that all persons in X′ love all and only persons in Y ′, and some person in Y ′ loves some person in some set in X other than X ′.” Nom- inalists do not appear to have any plausible way of assigning truth conditions to this sentence which, from their point of view, will make the platonists come out right in viewing it as true in some worlds and false in others.26 The simplest paradigm of a verbal dispute—the simplest way it can happen that each side of a dispute can fi nd a charitable interpretation that makes the other side come out right—is where, for each disputed sentence D, there are two undisputed sentences U1 and U2, one true and one false, such that one side holds that D is (a priori necessarily) equivalent to U1 and the other side holds that D is equivalent to U2. Each side can then assign charitable truth conditions to D in the other side’s language simply by assuming that in that language the other side’s asserted equivalence holds. In the previous section I tried in eff ect to show that the simple paradigm applies to the dispute between mereological essentialists and four-dimensionalists.27 There are evidently examples of verbal disputes in which the simple paradigm fails. Sup- pose there are two versions of English such that “red” means in one what “green” means in the other, and vice versa. The speakers of one language ought obviously to assign charitable truth conditions to the disputed sentences accepted by the speakers of the other language. (“In their language the sentence ‘Grass is red’ holds true of any situation in which grass is green”), though these assignments can’t be made in terms of undisputed sentences. In the case of nominalism versus platonism, however, it’s not just that the simple paradigm evidently fails. The question is whether the nominalists can in any terms acceptable to them assign charitable truth conditions to the disputed sentences accepted by the platonists. We should take note of another kind of verbal dispute that departs from the simple paradigm. Imagine a philosopher named Shmgettier

26. For illustrative purposes I’ve constructed more-or-less randomly a moderately complex sentence, without carefully checking whether this sentence might somehow reduce to a sentence in predicate logic. If necessary, a more complex example could obviously be constructed. 27. For the disputed sentence, “There is something in the room that is fi rst brown and then pink,” four-dimensionalists take it to be equivalent to the undisputed true sentence, “There is fi rst something in the room that is brown, and later there is something in the room that is pink,” whereas mereological essentialists take it to be equivalent to the undisputed false sentence, “There is in the room a mass of matter (or something not composed of matter) that is fi rst brown and then pink.” physical-object ontology, verbal disputes, and common sense 163 who, when presented with standard Gettier examples, insists that they are cases of knowledge because they satisfy the three traditional conditions. Assuming that Shmgettier has no disposition to retract, it seems that our dispute with him is merely verbal. If we imagine a linguistic community that agrees with Shmgettier, the plausibly charitable interpretation of the languages of that community and ours implies that they assert epistemic sentences of the form “S knows that p” on the basis of the three conditions, whereas we assert them on the basis of some other conditions. A complication, however, is that the Shmgettier- speakers may reasonably claim to be unable to fi nd a formulation in their own terms of what these other conditions are. Their inability to fi nd such a formulation would follow on the supposition that the speakers of our language are unable to formulate any clear analysis of what is required for knowledge beyond the three conditions. Never- theless this dispute is verbal because the Shmgettier-speakers can at least formulate a rough approximation to these conditions. They can at least say something like this about our epistemic assertions: “In this other language a sentence of the form ‘S knows that p’ is true only if some condition is satisfi ed related to the cause of S’s belief that p, or perhaps related to the potential defeasibility of the belief, or perhaps some combination of these conditions is required, perhaps satisfi ed to some required degree along some relevant dimension.” Even this kind of rough sketch suffi ces to allow the Shmgettier-speakers to acknowledge that there is a plausibly charitable interpretation of our language which makes our epistemic assertions come out true.28 It seems questionable, however, whether nominalists can formulate in terms acceptable to them even the roughest sketch of plausible truth-conditions for the platonist’s assertions that would make these assertions come out true. I make this remark with some reservations, because I am insuffi ciently knowledgeable of various logical maneuvers found in some nominalist literature, such as, substitutional quantifi cation, plural quantifi cation, meta-linguistic quantifi cation, infi nitary sentences, schemata, fi ctionalism, and other devices. I can’t rule out the possibility that, given a suffi ciently resourceful application of such devices, the dispute between platonists and nominalists might ultimately dissolve into verbalness. This dispute is, however, not verbal in any straightforward way. If resolving the dispute is merely a matter of

28. It may be held that the deepest disagreements in epistemology concern our “right to be sure,” and that such disagreements aren’t verbal for the same reason that ethical disagreements aren’t, namely, that they involve “disagreements in attitude.” Our disagreement with Shmgettier, however, has nothing to do with the right to be sure. 164 quantifier variance and realism: essays in metaontology choosing a language, as Carnap says, that would have to be developed along lines that go substantially beyond the present discussion.29

I turn now to my second claim in this paper: Since the disputes about the ontology of physical objects are merely verbal (when “all is said and done”), the correct position must be that of common sense ontology. I want to begin, as before, by having a specifi c representative of this position, and then imagining a corresponding community and language. Let me take myself as the representative. We imagine, then, the EH- community in which everyone accepts the disputed sentences I accept. The argument up to this point implies that these sentences are true in EH-English. The only question now is whether plain English is EH- English, RC-English, DL-English, or something else. The ontological sentences I accept, and that are therefore accepted in the EH-community, include all of the ones typically accepted by the non-philosophers in our actual community.30 That’s why I call my position one of common sense. But it may not follow immediately that EH-English is plain English, for the EH-community diff ers from our actual community in two relevant ways. First, the non-philosophers in our actual community may not agree with me about some of the axioms. Second, many philosophers in our actual community don’t agree with me about any of the disputed sentences, whether they be axioms or plain ontological sentences, whereas even the philosophers in the imagined EH-community agree with me about all of these sentences. I’ll return to these diff erences shortly.

29. A more general question is which fundamental disputes in metaphysics are merely verbal. I am certainly not inclined to think that all are (even after “all is said and done,” and no retractions are forthcoming). To mention two examples very briefl y, consider the position of dualists who reject both identity- theoretic and functionalist accounts of mental states. It seems to me that the identity-theorists and functionalists cannot begin to formulate in terms acceptable to them even the roughest sketch of a plausibly charitable interpretation that will make the dualist position come out right. As another possible example, consider the dispute (referred to in note 2, above) about whether simples exist. Can either side of this dispute fi nd a charitable assignment of truth conditions to their opponents’ assertions? 30. I accept sentences involving sets and properties. If some of these arc not understood and therefore not accepted by the non-philosophers in our actual community, neither are their negations. I don’t accept any ontological sentence (in plain English, without special stipulations) whose negation is typically accepted by the non-philosophers in our community. physical-object ontology, verbal disputes, and common sense 165

Many axioms have been put forth to refute the common sense position. Let me mention a few of the most important ones. 1. Two things cannot spatially coincide at any time. (This was already discussed.) 2. A composite thing must have causal powers beyond the causal powers of its parts.31 3. The conditions for a thing’s unity through space and identity through time cannot be intractably complex or “arbitrary.”32 4. Qualities such as shape, color, and texture cannot function semantically as relations between things and times.33 5. It cannot be indeterminate whether two things compose a third thing.34 6. If it’s indeterminate whether a is identical with b, then there must exist a plurality of things such that it’s indeterminate which of them is a (or which of them is b).35

I reject all of these axioms and so do the imagined members of the EH-community.36 I’ve already said in my earlier discussion of RC-Eng- lish and DL-English that ontological axioms do not bring in any essentially new diffi culties. If considerations of charity show that the question about the truth of the plain ontological sentences is verbal, then considerations of charity show in the same way that the question about the truth of the axioms is also verbal. Axioms 5 and 6, however, may require special attention. These axioms—especially axiom 6—seem often to be treated in the literature as if they simply derive from standard logic. Since the EH-community follows me in accepting all of standard logic, it may seem that these axioms must be accepted. Let me concentrate here on axiom 6; the treatment of axiom 5 easily follows along the same lines.

31. Merricks, Objects and Persons, ch. 3. 32. Cartwright, “Scattered Objects,” in Kim and Sosa, eds., Metaphysics: An Anthology, p. 293 and p. 298; Ernest Sosa, “Existential Relativity,” Midwest Studies in Philosophy XXII (1999), p. 143; Van Cleve, “Mereological Essentialism, Mereological Conjunctivism, and Identity through Time,” p. 145; van Inwagen, Material Beings, esp., pp. 64–71 and pp. 122–23. A rare heroic instance of resistance to axiom 3 is found in Markosian, “Brutal Composition.” 33. Lewis, On the Plurality of Worlds, pp. 202–4. Axiom 4 is apparently Lewis’s main reason for accepting temporal parts. Cf. Sider, Four Dimensionalism: An Ontology of Persistence and Time, pp. 95–101. 34. Lewis, op. cit., pp. 212–13. Lewis says that axiom 5 is his reason for accepting the doctrine of mereological sums. Lewis’s argument is clarifi ed and elaborated in Sider, op. cit., pp. 120–32. 35. A modifi ed version of four-dimensionalism is defended on the basis of axiom 6 in Sydney Shoemaker, “On What There Are,” Philosophical Topics 16 (1988), p. 217. 36. In truth, I’m not sure if I fully understand axioms 2 or 3, but I reject them conditionally: If they turn out to make sense and to threaten the common sense position, they must be false. 166 quantifier variance and realism: essays in metaontology

I’ve criticized axiom 6 at length in a previous article.37 In the context of the present discussion I want to present something akin to a reductio of it—or, more accurately, something akin to a reductio of the use made of the axiom in arguing against the common sense position. I’ll imagine the four-dimensionalists using the axiom in arguing against common sense. Let’s assume (as the premise of the “reductio”) that four-dimensionalism is right. I’ll now show that, on that assumption, axiom 6 is false in EH-English. Hence the appeal to axiom 6 as an argument against the common sense position has no force. The prima facie problem for common sense from axiom 6 is brought out in the following sort of example. It may be determinate that there is exactly one ship in the harbor on Monday and one ship in the harbor on Tuesday, but indeterminate whether the ship of Monday is the ship of Tuesday (because it’s indeterminate whether a change of planks has been too extensive, too discontinuous, and so on). Axiom 6 would then require that there exist a plurality of things such that it’s indeterminate which of them is the ship of Monday (or the ship of Tuesday). But, on the common sense position, there are no such things. To satisfy axiom 6 we need to accept the four-dimensionalist’s posit of numerous ship-like space-time chunks, some of which persist from Monday to Tuesday, some of which exist only on Monday, and some of which exist only on Tuesday. It’s indeterminate whether the ship of Monday is the ship of Tuesday because it’s indeterminate which of these space-time chunks will count as a ship. Axiom 6 is thought to follow from the Evans-Salmon result: “There cannot be things x and y such that it’s indeterminate whether x is y.” 38 The Evans-Salmon result does indeed seem to be a truth of logic (coming out of Leibniz’s Law), and hence something that the EH- community must accept. But accepting axiom 6 does not really follow from accepting the Evans-Salmon result.39

37. “The Vagueness of Identity,” Philosophical Topics 26 (1999). I criticize axiom 5 in “Quantifi er Variance and Realism,” pp. 65–66. 38. Gareth Evans, “Can There Be Vague Objects?,” Analysis 38 (1978), 208; Nathan Salmon, Ref- erence and Essence (Oxford University Press, Oxford, 1982), pp. 243–46, and “Modal Paradox: Parts and Counterparts, Points and Counterpoints,” Midwest Studies in Philosophy II (1986), pp. 110 ff . 39. Contrary to what seems to be implied in: Evans, “Can There Be Vague Objects?” (as interpreted in David Lewis, “Vague Identity: Evans Misunderstood,” Analysis 48 (1988), 128–30); Salmon, “Modal Paradox: Parts and Counterparts, Points and Counterpoints,” pp. 110 ff .; Timothy Williamson, Vagueness (Routledge, London, 1994), pp. 253–54; Shoemaker, “On What There Are,” p. 217; Rich- mond Thomason. “Identity and Vagueness,” Philosophical Studies 43 (1982), pp. 329–32. A position that seems closer to mine is Robert Stalnaker, “Vague Identity,” in Philosophical Analysis: A Defense of Examples, ed. D.F. Austin (Kluwer Academic Publishers, Dordrecht, 1988), p. 354. physical-object ontology, verbal disputes, and common sense 167

If we are four-dimensionalists trying to interpret the linguistic behavior of the EH-community in the most charitable way possible, we’re faced with the same kind of problem we had with respect to the RC-community. Here again it seems clear that we are dealing with a semantically restricted quantifi er. Whereas the quantifi er in RC-English is restricted roughly to masses of matter (or things not composed of matter), the restriction in EH-English—subject to some necessary fi ne-tuning—is to (ordinary) bodies (or things not composed of matter).40 (Let’s suppose, for simplicity, that a mass of matter counts as a body.) Now what axiom 6 says in EH-English is that, if it’s indeterminate whether the ship of Monday is the ship of Tuesday, there must exist a plurality of bodies (or things not made of matter) such that it’s indeterminate which of them is the ship of Monday (or Tuesday). That is surely false. It’s clear therefore that, given the restricted way the quantifi ers function in EH-English, axiom 6 is false in EH-English. Of course if we are operating from inside EH-English we will not formulate the argument that was just put from outside the language. (We cannot coherently say in EH-English, “The quantifi ers of EH- English are semantically restricted.”) From inside EH-English we simply reject axiom 6.41 And four-dimensionalists must approve of our doing this. They can’t then turn around and try to refute us by claiming that we violate logic by rejecting axiom 6. My argument up to this point has attempted to establish that RC- English, DL-English, and EH-English are intelligible languages, in each of which the ontological sentences accepted by the associated community are true. My next step is to show that, given this, plain English is EH-English, at least to a good approximation, so that the position of common sense ontology is true (in plain English). I’m going to move quite quickly through this step, since I think it’s virtually irresistible. I can scarcely imagine a philosopher saying, “Yes, there are these three possible languages, and people in North America might have spoken any one of them. As it happens, they speak RC-English (or DL- English).”

40. Quine, “The Roots of Reference,” p. 54, distinguishes between the four-dimensionalist’s general notion of a “physical object” and the more specifi c notion of a “body.” Some form of the latter notion fi gures in virtually all four-dimensionalist literature. See Lewis’s notion of a “natural thing” in “New Work for a Theory of Universals,” p. 372, and his notion of an “ordinary thing” throughout The Plurality of Worlds (e.g., pp. 203–4). 41. Since axiom 6 is false in EH-English and the Evans-Salmon result is true, it emerges that the context “It’s indeterminate whether a is b” is referentially opaque in EH-English, even where there do not exist a plurality of referential candidates for “a” (or “b”). All of this should really be worked through at the level of precisifi cations, as I try to do in “The Vagueness of Identity.” 168 quantifier variance and realism: essays in metaontology

If there are these three possible languages that North Americans might have spoken, then it seems obvious that they are in fact speaking a language in which the ontological sentences they accept are true, rather than a language in which those sentences are false. One diff erence between the imagined EH-community and our actual community is that, whereas everyone in the former community rejects the axioms that challenge the ontological sentences accepted in both communities, the non-philosophers in the latter community are generally perplexed by the axioms. In the case of one (and I think only one) of these axioms—that is, the “no-coinciding-things” axiom 1— they may be inclined to accept it, before they realize the problems this generates. Suppose a mereological essentialist says, “What we’re faced with here is a confl ict of charity: We can interpret the language in a way that is charitable to what the community says about the ontological sentences, yielding EH-English, or in a way that is charitable to what the community says about the ‘no-coinciding-things’ axiom, yielding RC-English. The latter interpretation is more plausible.” How could the latter interpretation be more plausible? At most what this philosopher might claim is that there is an element of indeterminateness in (plain) English, refl ected in the two possible interpretations. But even that seems extremely far-fetched. I think everyone agrees that perceptual sentences are central to both the acquisition and interpretation of language. To interpret the language of our community as RC- English is to imply that typical speakers make countless mistakes—not just mistakes, but a priori mistakes—in their perceptual reports. Such an extreme simultaneous violation of charity to perception and understanding seems out of the question. The more plausible interpretation, surely, would imply that the accepted ontological sentences are true, but people are understandably prone to make mistakes about a few tricky general statements like the “no-coinciding-things” axiom.42 A second diff erence between the imagined EH-community and our actual community is that in the latter community many philosophers disagree with me, both about the ontological sentences and about the axioms. But they also disagree with each other, thereby can- celing each other out as a source of linguistic interpretation. I’m not of course suggesting that when experts have prolonged disagreements about intractably diffi cult questions, we can ignore what they say. From my point of view, however, we are dealing here with the exceptional

42. I say something more about “confl icts of charity” in “Against Revisionary Ontology,” Philo- sophical Topics 30 (2002). physical-object ontology, verbal disputes, and common sense 169 case in which, fi rst, the questions at issue concern matters of linguistic interpretation, and, second, many philosophers don’t realize this, and therefore produce endlessly confl icting arguments on behalf of their favored entities—arguments that are irrelevantly convoluted and hyper-theoretical—while ignoring the genuine task of interpreting the language properly. The scenario that we are faced with is roughly this: (1) The non- philosophers in the community are essentially agreed on which ontological sentences (of the sort at issue) to accept; (2) the non-philosophers are generally perplexed by the ontological axioms, and may occasionally accept axioms that confl ict with the ontological sentences they accept; (3) many philosophers in the community engage in endless hyper-theoretical debates about both the ontological sentences and the axioms. Philosophers—unfortunately, at present a minority—who understand that the only genuine question at issue in these ontological debates is one of linguistic interpretation must conclude, upon examining (1), (2), and (3), that the language of this community is one in which the ontological sentences accepted by the non-philosophers are true. Plain English, in short, is EH-English, at least to a good approximation. A familiar maneuver found throughout the anti-commonsensical literature is to claim that while many of the ontological utterances of non-philosophers are strictly and literally false, they are nevertheless true (or in some manner acceptable) in some “loose” sense. I don’t think it’s necessary to examine the details of these maneuvers here.43 If the general framework of my argument is accepted, then I think it’s suffi ciently clear that these maneuvers are misguided. Of course there are distinctions—of various sorts—that need to be made between “strict” and “loose” talk, but these distinctions themselves must be based on a charitable interpretation of what people say. They cannot be imposed imperiously by philosophers just to save themselves the embarrassment of fl ying too fl agrantly in the face of common sense. The following seems suffi ciently clear: If you simply set yourself the task of interpreting in the most charitable way possible the language of our community, you cannot avoid the conclusion that the ontological sentences typically accepted by the community are true in that language, in the strictest and most literal sense.

43. I do examine some details in “Against Revisionary Ontology.” For illuminating critiques of such maneuvers, see Merricks, Objects and Persons, pp. 162–70, and John O’leary-Hawthorne and Michaelis Michael, “Compatibilist Semantics in Metaphysics: A Case Study,” Australasian Journal of Philosophy 74, 1 (1996), 117–34. 170 quantifier variance and realism: essays in metaontology

Amongst the numerous anti-commonsensical ontologists who currently dominate the literature, Theodore Sider is, to my knowledge, alone in seriously addressing the issues o f linguistic interpretation that I’ve been discussing in this paper.44 Sider has a clear and defi nite response to my argument: The ontological disputes are not verbal because it is impossible for there to be the diff erent languages I describe. The RC, DL, and EH-communities must all be using the same language, and are disagreeing with each other on matters of ontology. Since I think that Sider’s position is the only possible hope for anti-commonsensical ontology, I regard it as being of singular interest. I will try to explain, however, why I judge the position to be ultimately untenable.45 A basic premise in Sider’s account might be put by saying: Charity is not enough. What I call “(interpretive) charity” Sider calls “use,” by which he evidently means the charitable interpretation of the use people make of their language. In interpreting a language, according to Sider, we cannot simply appeal to charity or use, because some interpretations are more “eligible” than others. A more eligible interpretation may be the best candidate, all things considered, even though another interpretation is more charitable. On behalf of this general picture Sider cites Lewis’s infl uential idea that there is an a priori presumption that words express (relatively) natural properties rather than unnatural ones.46 Lewis has given us one principle of eligibility, related to the general words in a language, and Sider now wants to propose a second principle, related to quantifi cational expressions. Sider’s principle is this: The quantifi cational expressions of any possible language

44. See “Criteria of Personal Identity and the Limits of Conceptual Analysis” in Philosophical Perspectives 15 (2001), and the Introduction to Four Dimensionalism: An Ontology of Persistence and Time. 45. Sider has responded to some of my criticisms of his position in a recent publication that did not appear prior to my submitting the present paper: see his “Replies to Gallois, Hirsch, and Markosian” in the book symposium on Four-Dimensionalism, in Philosophy and Phenomenological Research 68, 3 (2004), 674–87, at pp. 679–82. I discuss Sider’s position further in my contribution to that symposium, and also in “Ontological Arguments: Interpretive Charity and Quantifi er Variance,” in J. Hawthorne, T. Sider, and D. Zimmerman, eds., Contemporary Debates in Metaphysics (Blackwell, Oxford, 2008). 46. Lewis, “New Work for a Theory of Universals,” pp. 370–77. Lewis’s terminology is a bit puzzling, since he seems to treat his naturalness-presumption as a corollary of the principle of charity, although there is no obvious connection between this presumption and the charitable presumption that the sentences accepted by the community are true or reasonable. (The apparent absence of any connection between the two presumptions is the main topic of my Dividing Reality.) I will continue to use “charity” in the sense of the latter presumption (which is equivalent to Sider’s “use”) and to distinguish it from the naturalness-presumption. physical-object ontology, verbal disputes, and common sense 171 must answer to the “logical joints in reality.” To fi nd out about the logical joints in reality requires that we do substantive ontology, for a quantifi er answers to reality’s logical joints only if it refers to everything that exists. Sider’s constraint implies, therefore, that the quantifi ers of any language must have the same semantic function as our quantifi ers, namely, to refer to everything that exists. Suppose our ontological refl ections have led us to believe (as they have in fact led Sider to believe) that four-dimensionalism is correct. In trying to provide a charitable interpretation of the linguistic behavior of the RC-community we fi nd that we would have to interpret their quantifi ers as semantically restricted in certain ways. But that interpretation is ruled out by Sider’s constraint as ineligible. The RC-quantifi ers must refer to everything that exists, and the RC-community must therefore be making substantive ontological mistakes when they deny the existence of temporal parts.47 Suppose, on the other hand, that our ontological refl ections have led us to accept mereological essentialism. Then, in trying to provide a charitable interpretation of the linguistic behavior of the DL-community, we fi nd that we would have to regard the DL-quantifi ers as having the semantic function within certain sentences of not really referring, but only behaving “as if” they refer. Sider’s constraint rules out this interpretation, and forces us to regard the DL-community as simply being wrong when they affi rm the existence of temporal parts. Hence the issue between the mereological essentialists and the four- dimensionalists is not verbal at all. I noted earlier that certain problems of compositional semantics may be thought to intrude when mereological essentialists try to provide a charitable interpretation of DL-English. Certainly no such problems intrude from the other direction, when four-dimensionalists are trying to interpret RC-English—semantically restricting the quantifi er raises no such problems. Sider’s constraint, therefore, evidently does not derive from considerations of compositionality. It is rather a straight-out constraint on the semantic function of quantifi cational expressions.

47. It might be objected that the RC-language can be charitably interpreted without violating Sider’s constraint on quantifi ers, by supposing that the RC-quantifi ers refer to temporal parts and sums, but every predicate in RC-English (including such predicates as “can’t be described”) have built in as part of their meanings “thing that is either a mass of matter or not composed of matter.” I think, however, that Sider can very plausibly respond that there is really no distinction between a blanket restriction on every predicate and a restriction on the quantifi ers. 172 quantifier variance and realism: essays in metaontology

Occasionally Sider seems to slip into a trivially ineff ectual defense of his constraint. He sometimes says that if an expression doesn’t have the semantic function of referring to everything that exists then it’s simply not a quantifi er.48 That’s irrelevant. Call it something else if you want. The relevant question is why we, as four-dimensionalists, ought to rule out a charitable assignment of truth conditions to the sentences of RC-English. If there could be a charitably construed RC-English then—as Sider is well aware—it would be the speakers of that language who might worry whether we are really using “quantifi ers.” So this is neither here nor there. In order to avoid a facile appeal to the word “quantifi er,” let me say that the “quantifi er-like” expressions in a language are expressions that function formally like our quantifi ers, that is, roughly, they satisfy the formal rules of predicate logic. Sider’s constraint implies that the most plausible interpretation for the quantifi er-like expressions in any language is that they refer to everything that exists. I think it’s obvious that Sider’s talk of “logical joints” is merely giving us a kind of metaphorical representation of his quantifi cational constraint; it does nothing to explain the constraint. The image is of joints or grooves in the world into which properly functioning quantifi er-like expressions easily fi t. In the absence of the constraint the idea of there being such joints is meaningless. Another metaphor that Sider uses is far more dangerous. He says that, in formulating his constraint, he is appealing to the view that the world comes “ready-made” with a domain of objects.49 Here again Sider is merely repeating his constraint in pictorial language. I, who reject his constraint, also believe in “ready-made objects” in the following prosaic sense: There are numerous objects in the world—rocks, rivers, trees, apples, planets, electrons—whose existence do not depend in any way on the existence of language or thought. These objects typically do not have temporal parts or sums, and that too does not depend in any way on the existence of language or thought. If I had been speaking DL-English instead of plain English, I would have correctly said, “These objects have temporal parts and sums, and that does not depend in any way on the existence of language or thought.” It’s essential in this area of philosophy to avoid a gross but somehow tempting use-mention confusion. My view is that if we consider expressions in languages with the same formal role as our expression

48. “Criteria of Personal Identity,” p. 16; Introduction to Four-Dimensionalism, pp. xv ff . 49. “Criteria of Personal Identity,” p. 16; Introduction to Four-Dimensionalism. p. xvii. physical-object ontology, verbal disputes, and common sense 173

“the existence of things in the world,” the semantic functions of these expressions may vary depending on the specifi c rules of the diff erent languages. I am most emphatically not saying that the existence of things in the world depends on there being certain rules of languages. The sentence, “The existence of things in the world depends on there being certain rules of languages,” is absurdly false in EH-English, as well as in RC-English and DL-English. My rejection of Sider’s quantifi cational constraint does not imply idealism.50 At one point Sider says that those who reject his constraint must be “committed to a fairly radical conception of the nature of existence.”51 I can’t imagine what he means by that, unless he has gotten himself to think that the rejection of his constraint implies some form of idealism. Is there, then, anything to be said in favor of Sider’s constraint? Let’s fi rst consider his basic premise that charity is not enough. I agree with this, but I think it’s important not to exaggerate its import. Sider cites with approval Lewis’s claim that the answer to “Putnam’s paradox” cannot appeal to charity or use, but can appeal to the principle that eligible interpretations involve (suffi ciently) natural properties. The paradox is that if you take some false theory, such as the phlogiston theory, so long as it is consistent (and so long as there are enough objects in the world), there will be objects and properties in the world that could provide a truth-making interpretation of the theory. Why, then, isn’t the theory true? Lewis’s answer is that an eligible interpretation must involve (suffi ciently) natural properties. That doesn’t seem to be the right answer. Suppose there are perfectly natural properties of objects on Alpha Centauri—objects that have no special connection to people on Earth—that could provide a truth-making interpretation of the phlogiston theory (together with any other relevant sentences accepted by the phlogiston theorists). That still doesn’t make the theory true. Even if Lewis is right in claiming that the naturalness-presumption solves Putnam’s problem, he also states (and Sider agrees) that another solution might appeal to a requirement that there be some kind of appropriate causal connection between the elements of the

50. A potential source of confusion is that the doctrine Putnam calls “conceptual relativism” implies both the rejection of Sider’s constraint and (often) some form of idealism. The two don’t have to go together, however. For further discussion, see “Quantifi er Variance and Realism” and “Sosa’s Existential Relativism.” 51. “Criteria of Personal Identity,” p. 17. 174 quantifier variance and realism: essays in metaontology

interpretation and the linguistic community in which the theory is formulated. But the principle of charity itself requires some such causal connection. A charitable interpretation of a theory must make it come out reasonable, hence, must make it come out connected in some reasonable way to the community’s perceptions (or perceptual reports), hence, to certain (reported) causal connections between the community and its environment.52 Ideally a charitable interpretation seeks both truth and reasonableness for a theory, but if both cannot be jointly achieved, reasonableness without truth is far more charitable than truth without reasonableness. The reason why the phlogiston theory is false is because any interpretation that makes it reasonable (to some signifi - cant degree) makes it false. The theory is false because it comes out false on the most charitable interpretation of it.53 It’s therefore not obvious at all that Putnam’s paradox shows that charity is not enough. I think that many philosophers are under the impression that Lew- is’s naturalness-presumption is needed to explain why we interpret “fi sh” to exclude whales. Earlier I imagined a scenario in which people who say “Whales are fi sh” are disposed to retract this if they are given evidence of the biological diff erences between whales and other things called “fi sh.” If our interpretation is guided by charity to retraction, then we don’t need Lewis’s naturalness-presumption. But what about a case in which there is no community-wide disposition to retract? Melville’s protagonist Ishmael insisted in Moby Dick that, despite what is said by tendentious scientists who hang around in lab- oratories, people who hang around with whales understand that they are really big fat fi sh. Melville/Ishmael was well aware of the biological diff erences between whales and other things he called “fi sh,” but didn’t care.54 If we imagine a Melville-community in which everyone agrees with Melville/Ishmael about this, it seems obvious that in Melville- English “Whales are fi sh” is (strictly and literally) true. People have the

52. Charity requires that certain sentences be interpreted as conveying reports about perception. They can’t be treated merely as sentences containing the word “perception” to be interpreted in terms of some relation on Alpha Centauri. 53. My point does not depend on charity to retraction, although that may certainly play a role, albeit complicated somewhat by the fact that many phlogiston theorists refused to retract. 54. Herman Melville, Moby Dick, sec. 32. After quoting Linnaeus’s grounds for not classifying whales amongst the fi sh, and reporting that his shipmates remained unconvinced, Ishmael states: “Be it known that, waiving all argument, I take the good old fashioned ground that the whale is a fi sh . . . This fundamental thing settled, the next point is, in what internal respect does the whale diff er from other fi sh. Above, Linnaeus has given you those items.” physical-object ontology, verbal disputes, and common sense 175 right to use the word “fi sh” to mean, roughly, “creature that lives in the water and has a fi sh-like appearance,” and charity would surely determine that the Melville-community uses “fi sh” in that way. It might be suggested that in our own community people are ambivalent about whether or not to agree with Melville, and Lewis’s naturalness- presumption gives the word “fi sh” a little nudge in the direction of excluding whales. Maybe so, but it’s far from obvious. Once we bring in Kripke’s picture of a reference chain going back to an initial baptism of a natural kind, the baptism presumably being interpreted as charitably as possible, the dependence on Lewis’s naturalness- presumption becomes even more obscure. I think it’s not clear that Lewis’s naturalness-presumption fi gures critically in this kind of example at all. Nevertheless, I do agree that at some point interpretations must go beyond charity. Lewis applies his naturalness-presumption to a form of Kripke’s “Wittgenstein paradox,” and here I am inclined to agree. As I understand it, the essential problem in the form presented by Lewis is how to determine the correct interpretation of certain very complicated or diffi cult sentences (say, in mathematics) when this cannot be settled by appealing charitably to the judgments people (would) make about the sentences, because people may fi nd the sentences too hard to make any judgments about them. In this form the problem doesn’t really give us a case in which the naturalness-presumption points to one interpretation (in terms of plus) and charity points to another (in terms of quus). Rather we have a case in which there is no possible appeal to charity, so naturalness takes over. I suspect that there may be other kinds of cases—I won’t try to produce them here—in which a somewhat less charitable interpretation is favored over one that would require going down an unnaturally twisted quus-like path. I am not, however, aware of any convincing example in which an appeal to naturalness (or compositionality, or general simplicity) would have us reject an interpretation that is demanded by the clearest forms of charity (e.g., charity to perception and understanding).55

55. Williamson’s “epistemic vagueness” may provide another kind of example in which the correctness of an interpretation is not determined by charity (use), at least at a level knowable by us, but this would certainly not be an example in which charity is trumped by other factors. See Williamson, Vagueness, esp. pp. 205–12. 176 quantifier variance and realism: essays in metaontology

Even if I have, in the above sketch, overstated the role of charity, the following is undeniable: Lewis’s interpretive constraint related to naturalness is only a defeasible presumption. It is defeasible by considerations of charity, and is in fact very often defeated in this way. Almost no word of ordinary language expresses what Lewis would regard as a perfectly natural property, and numerous words (“game,” to take a famous example) express properties that he would regard as quite unnatural. The naturalness-presumption is regularly trumped or compromised by charity. The contrast with Sider’s proposed quantifi cational constraint could therefore not be greater. What Sider wants is not a presumption, but an absolute and indefeasible constraint on how quantifi er-like expressions can possibly function. If we are four-dimensionalists viewing the linguistic behavior of the RC or EH-community, the hypothesis that their quantifi er-like expressions are semantically restricted in certain describable ways is supported by the most overwhelming demands of charity to perception and understanding. Sider wants us to rule out that hypothesis because it violates his constraint. There is nothing that we know about the nature of language or interpretation that prepares us for this kind of absolute constraint or that makes it seem credible.56 If we tried to formulate a defeasible presumption related to quantifi er- like expressions on the analogue of Lewis’s naturalness-presumption, it would not support Sider’s position, for it would not exclude charitable interpretations of the diff erent ontological sentences accepted in the diff erent communities. Let me also mention a major problem we face if we try to formulate such a presumption. If we have a defeasible presumption related to quantifi er-like expressions, it must be an open possibility that, in our own language, this presumption has been defeated to some degree (as evidently happens with respect to Lewis’s presumption). The problem is that the following is trivially true:

56. Sider does allow that quantifi er-like expressions can trivially vary in meaning as a result of outright stipulation, but in the texts cited in note 44 above he seems to imply that quantifi er variance cannot occur any other way. Here especially, however, see what Sider says in “Replies,” p. 680. Without being able to enter into any details, let me simply state that, to the extent that Sider is at present prepared to hold that four-dimensionalism is true in plain English, he would defend this position by making the following claim: “It is metaphysically impossible for the quantifi er-like expressions to have diff erent meanings in the diff erent linguistic communities, on the assumption that speakers in these communities accept and reject their ontological sentences in the same manner that speakers in our actual community do.” I think that my arguments against Sider in the present discussion would not be signifi cantly altered if they are taken to address that somewhat qualifi ed claim. physical-object ontology, verbal disputes, and common sense 177

“The quantifi er-like expressions in our language (e.g., the expression ‘everything that exists’) refer to everything that exists.” This makes it seem that, in whatever language we formulate the presumption, it will be trivially correct to say that the presumption holds to the highest degree in that language. It is diffi cult to see, therefore, how a defeasible presumption related to quantifi er-like expressions can be coherently formulated. There appears to be another important contrast between Sider’s quantifi cational constraint and Lewis’s naturalness-presumption. Lewis presents his presumption as a solution to certain existent problems in the philosophy of language. Sider’s proposed constraint seems, by contrast, designed to create problems rather than solve them. A main point of the constraint is to get us to think that there is some problem in defending our common sense judgments about what we perceive in front of our faces. A philosophical proposal that would have the eff ect of making it seem reasonable to worry, given that there are apple trees, whether there are apples is probably not geared towards solving our problems.57 A good rule of thumb in philosophy, I think, is that whatever seems to be actual is at least possible. Sider’s important insight is that anti-commonsensical ontologists must hold that the language we seem to be actually using—a language in which the ontological sentences typically accepted in the community would be trivially true—is not even a possible language. There appears, however, to be little to recommend this extraordinary idea beyond the wish of these philosophers to fi nd some way to justify their attacks against common sense.58

57. Quasi-nihilists like van Inwagen and Merricks are prone to believe that, whereas apple trees exist, apples don’t. 58. My thanks to Danny Kornman, and to members of the philosophy departments at Boulder and Rutgers, for many valuable comments on this paper. ten c C

ontological arguments: interpretive charity and quantifier variance

In the fi rst section of this chapter, I introduce a certain kind of defense of common sense ontology, one that derives from ordinary language considerations. The defense presupposes that quantifi er expressions can have diff erent meanings in diff erent languages, and this idea is discussed in the second section.

I. Charity in the Philosophy Room

In philosophy we often operate at two levels. At one level we use the language of our community—English, in the present instance—to make assertions about various philosophical topics. At another level we may be thinking about the nature of language, in particular about how linguistic behavior determines meaning. The interaction between these two levels can become problematical when we fi nd ourselves at the second level disagreeing about the meanings of our own assertions at the fi rst level. Let me try to illustrate the issue I’m driving at by considering the famous debate between Locke and Butler about the identity of a tree.1 Locke held that we have to distinguish between a tree and the masses of matter that successively constitute it. A tree may lose a branch and still retain its identity as that same tree, though it is now made up of a diff erent mass of matter. Butler insisted, on the contrary, that if we have a diff erent mass of matter then, strictly speaking, we don’t have the same tree. According to Butler, no object can persist through a change of parts. Some philosophers (not all, by any means) will share my own immediate intuitive feeling that this dispute between Locke and Butler is not substantive, that it is in some sense merely verbal. Locke

1. Locke (1690: bk. 2, ch. 27); Butler (1836).

178 ontological arguments 179 and Butler agree that we are faced with a situation in which some tree-composing masses of matter are related to each other in certain qualitative, spatiotemporal, and causal ways. They don’t seem to disagree at all about what those relations are. (They are not, for example, disagreeing about “entelechies” or life-forces or anything like that, because their disagreement carries over to non-living things, such as rivers and ships.) It seems, therefore, that they are merely disagreeing about whether a situation in which masses of matter are interrelated in certain agreed-upon ways ought to be described in our language with the words “It’s the same tree,” or with the words “It’s a diff erent tree.” Let “Lockean English” be an imagined version of English in which the semantic rules prescribe the former description, and “Butlerian English” an imagined version of English in which the semantic rules prescribe the latter description. Then it seems that all that these philosophers can be disagreeing about is whether the language we speak is in fact the fi rst imagined language or the second. Let’s suppose provisionally that this really is their disagreement; I’ll consider other alternatives later. If this is their disagreement, how can we try to resolve it? How can we determine whether the language we speak is Lockean English or Butlerian English? We need to imagine, of course, that the semantic rules for Lockean and Butlerian English assign diff erent truth conditions, not just for the mentioned sentences about trees, but for numerous sentences that imply identity through mereological change (i.e., change of parts).2 Butlerian English is a “mereological essentialist” version of English, in which no sentence can count as true if it entails the sentence “Some objects persist through a change of parts.”3 Let me call this latter sentence “L” (“L” because it is accepted by Locke but not by Butler), and the “L-sentences” will be L together with the various particular identity sentences that entail it. I’m presently assuming that we can make intelligible to ourselves two versions of English: in Lockean English the L-sentences are true, whereas in Butlerian English they are false. Depending on which version of English we are speaking, we

2. I’m making the standard assumption that a language must contain some form of “truth condition” rules, i.e., rules that determine for the assertion of any sentence what conditions or states-of- aff airs are required for the assertion to count as true. 3. Mereological essentialism is the view that an object cannot persist through a change of parts. Butler’s mereological essentialism has been defended in the more recent literature by a number of philosophers, including Chisholm (1976: ch. 3); Van Cleve (1986). 180 quantifier variance and realism: essays in metaontology ought to either accept or reject the L-sentences. So which version are we speaking? A highly infl uential principle of linguistic interpretation, sometimes called the principle of charity, says, roughly, that if we are trying to decide between two interpretations of a language, there is a presumption in favor of the one that succeeds better in making people’s assertions come out true, or, if not true, at least reasonable.4 The principle appeals to “ordinary use” in the sense of requiring us to interpret a language in the way that makes most rational sense out of how the language is used. If we apply the principle to the current linguistic behavior of our English-speaking community with respect to the L-sentences, it seems fairly obvious that we get a presumption in favor of the Lockean interpretation. Speakers of Butlerian English would treat the L-sentences as only loosely true but strictly false. This is clearly not the norm in our current community, where people would not hesitate to swear in court to the strict and literal truth of an L-sentence like “The tree that he banged into is the one that used to have more branches.” (How odd that in Butler’s famous passage he says that “when a man swears to” the identity through mereological change of a tree, he means it only loosely; surely the general presumption is that when people swear that something is so they mean it strictly.) But the appeal to charity doesn’t end here. A central element of charity is what might be called charity to retraction. If the community retracts a set of sentences that were previously accepted, then considerations of charity must favor an interpretation which makes the sentences false. This is because there must surely be the presumption that people are more likely to get things right at the end of the day, after being able to examine more evidence or consider more arguments. Now the mereological essentialists are urging the community to retract its commitment to the strict and literal truth of the L-sentences. If the community agrees to retract, then it may follow by charity to retraction that, on the most plausible interpretation, the language being spoken is Butlerian English. We seem, then, to have reached the following peculiar impasse. If we decide that we ought to retract the L-sentences, then the very fact of our having made that decision would establish the charitable presumption that we were speaking Butlerian English—in which case we were right to retract. If we decide that we ought not to retract, then

4. Quine (1960: 59), Davidson (1984: esp. “Belief and the Basis of Meaning” and “Thought and Talk”); Lewis (1983a and 1983b: 370–7). ontological arguments 181 the very fact of our having made that decision would establish the charitable presumption that we were speaking Lockean English—in which case we were right not to retract. It appears that, whichever way we decide, we are right just because we have so decided. There is an evident connection here to the “Wittgenstein paradox,” which I’m not going to try to explore.5 As regards the question whether our language is Lockean or Butlerian, I will, however, suggest an answer. One route out of the impasse is to be clearer, and perhaps more modest, in our understanding of what the word “we” means when it is said, “If we decide that we ought to retract, etc.” If the linguistic community decides to retract, if retracting becomes the norm of the community, then it may indeed follow by charity to retraction that Butlerian English is the community’s language. But if “we” are merely some philosophers who have decided to retract, that has virtually no eff ect on the interpretation of the language of the community. Even, therefore, if some philosophers fi nd themselves at the fi rst level inclined to follow Butler’s urging to retract the L-sentences (even if this is their “intuition”), they must realize at the second level that it is a mistake to retract. They must acknowledge that, given the lack of authority that philosophers have in our culture, and given that, even if they were to have authority, they cannot agree amongst themselves on whether the L-sentences ought to be retracted (a fact that is not unrelated to their lack of authority), the community at large is going to stick with the L-sentences. Assuming that mereological essentialists intend to be speaking the language of the community—that is, plain, non-technical English—they must realize at the second level that this language is not Butlerian English. They must, therefore, stop being mereological essentialists. They must correct their fi rst-level intuitions because of their second-level realization that these intuitions do not—and will not— refl ect the linguistic practices of the community. Of course there is nothing to prevent philosophers from knowingly adopting Butlerian English as a technical language that they believe serves some philosophical purpose (for example, the purpose of eliminating some of the vagueness that affl icts identity sentences in Lockean English). But these philosophers are not mereological essentialists in the sense that Butler was. They would not say, as he did, that when “we”—that is, we speakers of English—swear to its being the same tree we ought to mean this only loosely.

5. I refer to the paradox discussed in Kripke (1982). The paradox begins with the observation that our past use of an expression cannot determine a unique rule for correctly using it in the future. 182 quantifier variance and realism: essays in metaontology

The focus of interpretive charity is not only on what people actually say, but on what they would say, especially what they would say “with their eyes wide open,” as Sider has put it (2004: 680). That is indeed the rationale of charity to retraction. But ontological arguments of the sort given by mereological essentialists about the existence and identity of physical objects are not eye-opening in the relevant sense. They lead to irresolvably confl icting assertions or to ambivalence and confusion. The focus of interpretive charity must therefore be on the pre-philosophical assertions of commonsense. One has missed my point entirely if one is tempted to answer that it’s necessary to be intel- lectually courageous and resolute and to continue to seek the ontological truth. Given my assumption that the confl icting ontological assertions are true or false depending on whether we speak Lockean or Butlerian English, the only truth we seek is which of these languages is plain English. To fi nd this out we have to learn to ignore the tower of Babel in the ontology room and focus our charity on what ordinary people say. Once we learn to do this, we are likely to fi nd that our fi rst-level philosophical intuitions gradually merge with the pre- philosophical convictions of commonsense. Even in ontology, one can learn to acquiesce to the language instead of battling it. Let’s imagine, however, a linguistic community in other ways like ours, but in which philosophers invariably agree that the L-sentences should be retracted. Everyone in this community, we may imagine, has the disposition to retract the L-sentences under the infl uence of philosophical refl ection. Would it then follow by charity to retraction that the language of this community is Butlerian English, even though the vast majority of its members, having no knowledge of philosophy, persist in asserting the L-sentences with complete confi dence? I am not sure how to answer this. For my immediate purposes it suffi ces to say that this imagined community is not ours. The role of philosophers in the imagined community should, however, be contrasted with the role of scientifi c experts in our actual community, for example, in getting people to retract the sentence “Whales are fi sh.” There is a critical diff erence between the bearing of interpretive charity in scientifi c and philosophical examples. Apart from charity to retraction perhaps the most essential element of charity is what might be called charity to understanding. This is the presumption that members of the linguistic community generally understand what they are talking about to the extent at least that they do not make a priori (conceptual) mistakes about seemingly uncomplicated judgments. The scientists who claimed that people were mistaken about whales being fi sh did not imply that ontological arguments 183 people were making a priori mistakes. The claim was rather that people were making empirical mistakes for relatively good reasons. It is in general a modest violation of interpretive charity—and certainly not a violation of charity to understanding—to suppose that people make reasonable empirical mistakes. My assumption throughout this chapter, however, is that issues of ontology are a priori. The mereological essentialists are claiming that ordinary people make innumerable a priori mistakes with respect to seemingly simple observational facts. This claim constitutes an overwhelming violation of charity to understanding. I doubt that this claim could stand even in the imagined community in which all philosophers agreed on it. It certainly cannot stand in our actual community. Mereological essentialists have often followed Butler in granting to ordinary people that the L-sentences they assert about mereological change, while false in the “strict and philosophical sense,” are true in some “loose and popular sense.” Is that not suffi ciently charitable to ordinary speech? The problem at this point, however, is not so much a matter of charity as the intelligibility of the distinction these philosophers want to make between “strict” and “loose” talk. A philosopher cannot simply announce from on high that some sentences generally accepted in the community are only “loosely true.” This has to come out of an analysis of the use of these sentences in the community. One might as well say that ordinary people are just joking when they describe mereological change. (The reason why no one laughs is because the joke is old; small children, on the other hand, do have inexplicable laughing fi ts.) Butler seems to want to say that the statement “It’s the same tree” is something like an exaggeration. But there are ordinary criteria for calling an utterance an exaggeration and these criteria are patently absent in assertions about mereological change. People, for instance, will not typically swear to an exaggeration, as already noted, and they will typically withdraw an exaggeration when challenged. Nothing like this happens in our linguistic community with respect to assertions about mereological change. It must continually be borne in mind that our present assumption is that Butlerian and Lockean English are two possible languages. Mereo- logical essentialists must therefore be able to explain by what criteria they take our community to diff er from the possible community that speaks Lockean English, in which the L-sentences are (not exaggera- tions but) strictly true. It is not enough for these philosophers to rely on the strength of their private intuitions in the philosophy room. They 184 quantifier variance and realism: essays in metaontology have to explain why, all things taken into account, plain English is not Lockean English, as it superfi cially appears to be. I should hope it is clear that what I have just been saying about the fallacy of mereological essentialism applies as well to every ontological position that wants the community to retract its ordinary assertions about the existence and identity of physical objects. I have in mind the plethora of currently popular positions that seem to fl y in the face of common sense. These include the doctrine of unrestricted mereological composition, according to which any two objects compose a third object;6 the doctrine of temporal parts, according to which ordinary objects like trees have earlier and later parts;7 ands sundry “nihilist” doctrines that deny the existence of all or most composite objects.8 From my perspective, the philosophers who advance these various anti-commonsensical doctrines seem to be driven by an obsessive fascination with highly rarifi ed general principles and philosophical riddles, which invariably override any pre-philosophical intuitions of common sense. The intuitive sensibilities of these philosophers baffl e me, but that is not my main point here. If these philosophers intend to use a technical language diff erent from plain English, then they are required to stipulate what this technical language is (and to explain what interesting questions can arise once such a stipulation is made). If, however, they intend to speak plain English (which is what generally appears to be the case), then they seem to have failed drastically to interpret this language in a plausibly charitable manner. Their various doctrines seem to be trivially absurd in plain English, as charitably interpreted. I don’t mean to suggest that a sharp line can be drawn between the pre-philosophical intuitions of commonsense and the intuitions that philosophers profess to have in the philosophy room. What we have rather is a gradation of cases, going from the most unrefl ective perceptual judgments of ordinary people to the most esoteric judgments of professional philosophers. Consider the general principle, “Two things cannot occupy exactly the same place at the same time.” Philosophers disagree about this principle. Locke had to deny it in order to distinguish between the tree and the mass of matter that makes it up. Mereological

6. See chapter 8 of Sider, Hawthorne, and Zimmerman (2008); also see Van Cleve (1986), Thom- son (1983), Cartwright (1975), Lewis (1986: esp. 212–13), Sider (2001b: 120–32). 7. See chapter 6 of Sider, Hawthorne, and Zimmerman (2008); also see Lewis (1986: 202–4), Heller (1990; ch. 1), Sider (2001b). 8. See chapter 8 of Sider, Hawthorne, and Zimmerman (2008); also see Unger (1980), van Inwa- gen (1990), Merricks (2001). ontological arguments 185 essentialists accept it, and often use it as an argument for their position. What should we say is the pre-philosophical intuition about this principle? Certainly people are initially inclined to accept it, but they are also often confused by it (“What about mixing wine and water?”), and they are taken by complete surprise when the principle is applied to an object and the matter that makes it up; that is clearly not the kind of case they had in the back of their minds when they wanted to accept the principle. Considerations of interpretive charity have to take all of this into account. To the extent that the principle seems attractive to the members of our linguistic community, charity would point to an interpretation of our language as being Butlerian English, in which the principle is true. On the other hand, to the extent that members of our community accept innumerable L-sentences such as, “That tree used to have more branches,” charity points to an interpretation in terms of Lockean English. We seem to be faced with what might be called a “confl ict of charity”: we can choose an interpretation that is charitable to the principle or one that is charitable to the L-sentences. I think it’s clear how this confl ict needs to be resolved. The demands of charity in behalf of the L-sentences far outweigh those in behalf of the principle. The dominant tendency of our language is Lockean, not But- lerian. There are several reasons for this. Most obviously, there is one principle contending with numerous L-sentences. Furthermore, people are almost always confused by the principle, whereas they frequently accept L-sentences with complete confi dence and clarity. The last point is related to the fact that the principle is abstract, whereas the L-sentences are often rooted directly in perceptual experience. I’ve highlighted the importance of charity to understanding and retraction, but another central consideration is charity to perception, by which I mean the presumption that any language contains sentences used to make perceptual reports, and that these reports are generally accurate (to a fair degree of approximation). Charity to perception is central because of the undeni- ably close connection between perceptual experience and language learning. Although it need not be implausible to attribute some specifi c perceptual mistake to the community, such as the mistake of perceiving whales as fi sh, there must be a strong presumption against attributing to the community massive perceptual errors about the existence and identity of the objects typically encountered, especially errors that are alleged to be of an a priori conceptual nature. It is, all things considered, far more plausible to suppose that the members of our linguistic community are mistakenly attracted to a certain abstract principle than that 186 quantifier variance and realism: essays in metaontology they make innumerable mistakes about what they perceive in front of them. Philosophers ought to admit this, and ought therefore to admit that plain English is Lockean English, no matter what the strength might be of their own attraction to the principle.9 The generalities and riddles that often lead to anti-commonsensical ontology are not alien to ordinary thinking. If they were, they could not catch on even in the philosophy room. They refl ect tendencies inherent in our language that are, however, defeated by other more predominant tendencies. Lockean English is certainly the dominant language of our community, but I think it need not be wrong to say that Butlerian English is in some sense contained germinally within plain English, as a kind of recessive undercurrent. Our ordinary concepts prepare us for Butlerian English, and may even push us in that direction, at least when we are engaged in thinking about certain kinds of questions.10 In investigating the ontology of physical objects, part of our philosophical task is to try to understand these “dialectical tendencies” that I think can be viewed as in some sense implicit in what we ordinarily mean. What we must not allow to happen is that we are carried away into a language of our own creation without realizing it, identifying the predominant language with what is only an undercurrent in it. (Defi nition of “ordinary language philosophers”: philosophers who have been trained to monitor their fi rst-level intuitions by paying attention to their second-level judgments about the demands of linguistic interpretation.)

II. Quantifi er Variance

My provisional assumption has been that we can make intelligible to ourselves the two imagined versions of English, Lockean English and Butlerian English, but that assumption is by no means trivial. These languages would have to be viewed as diff ering not just in the meaning of such terms as “tree” and “ship,” but also in the meaning of such quantifi er expressions as “(some)thing” and “(there) exists.” Consider, for example, the sentence, “In addition to masses of matter, there exist other material things.” This sentence would qualify as true in Lockean

9. For further discussion of “confl icts of charity,” see Hirsch (2002a). 10. Compare with J. N. Findlay’s notion of “logical dynamics” (1962: 76–7). ontological arguments 187

English but, I assume, false in Butlerian English. (Butler implied, I think, that a tree or a ship is simply identical with a particular mass of matter.) The diff erent semantic rules that would have the eff ect of rendering the sentence true in one language and false in the other must in some sense provide diff erent rules for “counting what things there are in the world.” If there could be these two languages they would have to embody in some sense diff erent concepts of what it is for “there to exist something.” Not all philosophers would admit that it makes sense to suppose that there could be diff erent semantic rules with that eff ect. Hilary Putnam has been a vigorous defender of the possibility of quantifi er variance, that is, the possibility of quantifi er expressions varying their meaning from one language to another. In two typical passages he says:

[T]he logical primitives themselves, and in particular the notions of object and existence, have a multitude of diff erent uses rather than one absolute “meaning.” (1987: 71)

Putnam often calls his position “conceptual relativism,” and he has often seen the overall import of this position as being in some signifi - cant sense anti-realist. Criticisms of Putnam have frequently centered on his anti-realism, and the specifi c doctrine of quantifi er variance has therefore not often been properly isolated. In the present discussion I want to make sure that quantifi er variance is clearly disassociated from anti-realism. It is only the former that interests me and that is presupposed in my discussion of Lockean and Butlerian English. The doctrine of quantifi er variance, properly isolated, does not entail anti-realism. One should not say, “If the sentence ‘In addition to masses of matter, there exist other material things’ is true in Lockean English but false in Butlerian English, then it is a matter of linguistic choice what exists in the world.” It is rather a matter of linguistic choice what is meant by the expression “what exists in the world.” Whether we speak Lockean or Butlerian English, the anti-realist statement “It is a matter of linguistic choice what exists in the world” can be—and, I think, ought to be—viewed as absurd. Anti-realism, therefore, is not at all what we are talking about here. 188 quantifier variance and realism: essays in metaontology

Once this point is understood, the temptation might be to make a mistake from the other side. “If all you’re saying is that expressions like ‘there exists (something)’ can have diff erent meanings in diff erent languages, that’s completely trivial. The expression might be used to mean ‘Hello,’ for instance.” The non-trivial question, however, is whether the expression can have diff erent meanings in diff erent languages while still retaining the general formal role of a quantifi er expression, that is, roughly put, the formal role described in quantifi - cational logic.11 In Lockean and Butlerian English, the expression “there exists (something)” might be said to have the same “formal meaning,” but, for these languages to function in the intended ways, the expression must have diff erent meanings in the languages in the sense of contributing diff erently to the truth conditions of sentences in the languages. If there could be the two languages, therefore, there must be an important sense in which the speakers “operate with diff erent concepts of existence.” It would be pointless to resist this formulation and to insist that, even if there could be the two languages, only one of them (at most) contains a “real quantifi er” that expresses “the real concept of existence.” This is pointless because the important question, for the argument in the last section and for related arguments, is whether it is possible for there to be languages that stand to each other in the manner of Lockean and Butlerian English. In what follows, therefore, I will make the quite natural stipulation that if there could be such languages then they could be said to contain quantifi er expressions with diff erent meanings, and these expressions would signify diff erent concepts of existence. The important question is whether quantifi er variance in this sense is an intelligible possibility. Suppose it is not a possibility. Then the issue between Locke and Butler cannot be as I represented it in the last section. If quantifi er variance is impossible, then we don’t have two possible interpretations of English, the Lockean and the Butlerian, and the disagreement between Locke and Butler is not in any sense merely verbal. The impossibility of quantifi er variance would make ontology seem far more substantive and profound. That may be a reason for philosophers to wish for it, but not a reason to accept it. One thing we must not do is to allow the issue to degenerate into a battle of metaphors: “the world as containing ready-made objects”

11. The formal role is illustrated by the principle that “This is F ” entails “There exists something that is F,” a principle that holds both in Lockean and Butlerian English. ontological arguments 189

versus “the world as an amorphous lump divided up by language.” These metaphors have had the eff ect in the literature of reinforcing the confusion that the issue is realism versus anti-realism. Once this confusion is put aside, the opponents of quantifi er variance must squarely face the burden of showing why there could not possibly be languages related to each other in the manner of Lockean and Butlerian English. I myself don’t see how this burden can be met. Certainly I don’t deny that there is something perplexing in the idea that the world might be described with diff erent concepts of “there exists something.” If we are realists, we are committed to the following formulation: “There exist things in the world independent of language, and language comes along and enables people to state truths about the things in the world.” The diffi culty is in seeing that this formulation does not entail “There exist things in the world independent of language, and any language that comes along must contain the same concept of ‘the existence of things in the world’ that I have just expressed.” The doctrine of quantifi er variance, in the realist sense that I’m assuming, accepts the fi rst formulation but rejects the second. This is initially perplexing, but seems, on refl ection, unavoidable.12 There are, I would agree, describable languages that strike us intuitively as being so bizarre that we can’t take them seriously as being in some sense real possibilities. The imagined language in which “grue” and “bleen” replace “green” and “blue” certainly strike many people in this way.13 As regards such “radically alien” languages, Kripke expresses the doubt whether we can in some sense make them really intelligible to ourselves (“from the inside”) (1982: 98n78). That’s a doubt I share. Some examples of quantifi er variance may also generate this kind of doubt. Sider imagines a language in which “There exists something that is F” means that Quine says that there exists something that is F (2001b: xx). This language may indeed strike us as “too strange to be really intelligible.” It seems in fact easier to cite a reason for the intuitive craziness of this language than it is for the grue-bleen language. It is well known that for any fact stateable in our language,

12. One source of perplexity that I cannot go into here concerns the problematical eff ects of quantifi er variance on referential semantics. For further discussion, see Hirsch (1999; 2002b; 2004; 2005). A novel referential argument against quantifi er variance, which I will not be able to address here, is given by Matti Eklund in chapter 9.2 of Sider, Hawthorne, and Zimmerman (2008). 13. See Goodman (1973: 74), where the predicate “grue” is defi ned as applying to all things examined before a given time just in case they are green but to other things just in case they are blue. “Bleen” is defi ned in the corresponding manner. 190 quantifier variance and realism: essays in metaontology there is an a priori necessarily equivalent fact stateable in the grue- bleen language. In this sense, the two languages have the same fact- stating power. The language with the intuitively crazy Quine-linked quantifi er, however, would be drastically defi cient in fact-stating power. That is one reason (I doubt it is the only one) why this example of quantifi er variance may strike us as strange to the point of not being fully intelligible. I assume that Butlerian and Lockean English have approximately the same fact-stating power. It’s immediately clear that, for any sentence in Butlerian English, there is an a priori necessarily equivalent sentence in Lockean English. (Henceforth, I’ll use “equivalent” as short for “a priori necessarily equivalent.”) The point in the reverse direction is somewhat less clear. In order to fi nd in Butlerian English a sentence equivalent to the true sentence of Lockean English, “A tree persisted while losing a branch,” one would need to provide in terms of the former language an analysis of what counts in the latter as “the identity of a tree.”14 Such an analysis is, of course, hardly a trivial matter, but I think we can assume that there is in principle a sentence in Butlerian English that is at least more or less equivalent to the Lockean sentence. It will help to bring into the discussion another possible language. I will take four-dimensionalism to be the highly popular position that combines the doctrine of temporal parts and the doctrine of unrestricted mereological compostion. Suppose that an ordinary tree, containing a trunk and some branches, stands in the yard. The doctrine of temporal parts implies that the tree is made up of daytime parts and nighttime parts; the same for the trunk. The doctrine of unrestricted mereological composition implies that there is an object composed of any combination of these parts, for example, the daytime parts of the tree and the nighttime parts of the trunk. According to four-dimensionalists, therefore, there is in the yard a highly visible brown wooden object that has branches during the day and has no branches during the night. I would want to deal with four-dimensionalism in the same way I have dealt with mereological essentialism. We can imagine a language— let’s call it 4D-English—in which all of the sentences asserted by the four-dimensionalists are true. The semantic rules for 4D-English per- mit one to describe any object in terms of “temporal parts,” and any pair of objects in terms of their “sum.” These rules allow one, therefore,

14. Perhaps Butler himself gives us a too easy solution: we might translate the Lockean sentence into the Butlerian sentence, “A tree loosely persisted while losing a branch.” ontological arguments 191 to describe the ordinary situation of a tree standing in a yard with the words: “There is in the yard a highly visible brown wooden object that has branches during the day and has no branches during the night.” From my point of view, we have three possible languages: Butlerian English, Lockean English, and 4D-English. These languages have essentially the same fact-stating power. In saying this, I’m assuming that all of these languages have the resources of set theory. So the 4D-English sentence just cited is equivalent in both Lockean and Butlerian English to the sentence, “There is a set containing pairs of objects and times, some of the times being during the day and some during the night, such that each object in a pair is brown, highly visible, and in the yard, and if the time in a pair is during the day the object in that pair has branches, and if the time in a pair is during the night the object in that pair does not have branches.” I can imagine a commonsense ontologist who holds that, while But- lerian and 4D-English have essentially the same fact-stating power as plain English, these languages are intuitively “too strange to be really intelligible.” I’m not inclined to adopt that extreme position, though I do have a bit of sympathy for it. Since I take it that these are three possible languages, it seems that the only relevant question is which of them is plain English. Considerations of interpretive charity seem to indicate decisively that Lockean English is plain English. This is why the pre- philosophical assertions of commonsense are correct in plain English.15 The possibility of quantifi er variance brings with it another possibility: quantifi er vagueness. If the quantifi ers in diff erent languages can have diff erent semantic functions, then it may be indeterminate (or unknowable because of vagueness) precisely how the quantifi er functions in a given language.16 Lewis and others have assumed that quantifi er vagueness is not possible, and have argued against commonsense ontology on the grounds that, in this ontology, it must sometimes be indeterminate (or unknowable because of vagueness) how many objects exist in a given situation.17 This argument has no force once quantifi er variance is accepted.

15. In associating plain English with Lockean English (that is, the version of English in which John Locke’s assertions come out true), I have in mind only what Locke said about the identity and existence of physical objects. (I ignore his views about personal identity, modes, and various other matters.) 16. I intend this formulation to be neutral as between supervaluationist and epistemic views of vagueness. See Williamson (1994). 17. See Lewis (1986: 212–13); Sider (2001b: 120–32). 192 quantifier variance and realism: essays in metaontology

The disagreement between mereological essentialists, four- dimensionalists, and defenders of commonsense is, on my view, merely verbal—merely a matter of language—in the following sense: Each side can charitably interpret each other side’s position in terms of a language in which all the other side’s assertions come out true.18 I am not suggesting, however, that every ontological dispute is verbal in this sense. Suppose that a Platonist conjectures, “The set of bits of matter is in a one-to-one correspondence with the set of integers.” A nominalist who rejects sets (and other abstract things) seems to have no way of making any charitable sense out of this conjecture. The reason for this is that nominalists apparently cannot make intelligible in terms acceptable to them a “Platonist language” which has essentially the same fact-stating power as their own “nominalist language.” This point may not be clear if one focuses exclusively on sentences of pure set theory, that is, sentences that are either necessarily true or necessarily false. The Platonist conjecture just mentioned, however, which is intended to be true in some worlds and false in others, is, on the face of it, not equivalent to any sentence that nominalists can accept.19 Carnap’s (1956) doctrine of “external questions” implies that the issue between Platonists and nominalists is merely a matter of language. There may be some intuitive sense in which this is correct, but nothing in the present chapter supports this position. As far as my argument goes, the relevance of quantifi er variance to ontology is limited, roughly, to issues about the existence and identity of perceivable objects. The position I’m defending here has been signifi cantly challenged by Sider. His basic idea is that there is a constraint on linguistic interpretation requiring quantifi ers to correspond to what he calls the world’s “logical joints.” Ontological arguments reveal where these joints are located. Since these arguments favor four-dimensionalism, as Sider believes, it follows that only the quantifi ers of 4D-English correspond to the world’s logical joints. The correct interpretation of plain English, therefore, must be 4-D English. Although charity to ordinary use counts

18. The relevant notion of a verbal dispute is clarifi ed further in my “Physical-Object Ontology, Verbal Disputes, and Commonsense.” (There may, of course, be other signifi cant ways of defi ning what is meant by a verbal dispute.) 19. I make this remark with some reservations because I have no clear grasp of various linguistic resources nominalists sometimes allow themselves, such as, plural quantifi cation, substitutional quantifi cation, schemata, infi nitary sentences, and fictions. One would have to examine whether such resources might allow nominalists to make charitable sense of the Platonists’ assertions. ontological arguments 193 against this interpretation, considerations of charity are trumped by the constraint binding the quantifi er to the logical joints.20 Sider compares his quantifi er constraint to Lewis’s idea that there is a presumption in favor of interpreting general words as expressing (highly) natural properties (Lewis 1983: 370–7). The comparison seems, however, to work against him. Lewis’s presumption is frequently defeated by considerations of charity to ordinary use. The word “game,” for instance, would probably express a far more natural property if one- person games like solitaire were excluded, but ordinary use prevails. There are numerous examples of the same sort. To my knowledge, neither Lewis nor any other philosopher has presented an example of interpreting a general word in which clear-cut considerations of charity to ordinary use do not prevail. Why should it be diff erent with words like “something” and “exists?” Even if there is a special interpretive presumption binding these words to the “logical joints,” and even if this presumption is satisfi ed only by the 4D-quantifi er, the use of these words in our community decisively trumps any such presumption and shows that 4D-English is not our language.21 There is a further move that Sider has recently made to which I am less sure how to respond (Sider 2004: 680–2).22 He claims that even if it is agreed that plain English is not 4D-English, that need not bother four-dimensionalists. Granted, their assertions would then be false in plain English, but ontologists don’t care about plain English. The tacit understanding amongst ontologists is that in the philosophy room they use an ontologically ideal language, whether or not it is plain English, in which the quantifi ers correspond to the logical joints. Disagree- ments of ontology are disagreements about the logical joints, and therefore cannot be resolved by appealing to ordinary use. One immediate point is that if philosophers who appear to be writ- ing in plain English really aren’t, that needs to be stipulated at the outset. It seems unclear, in any case, why it is necessary to switch to a new language, Ontologese (as Sider calls it), in order to argue about the logical joints. Why can’t we simply add the technical expression “logical

20. See Sider (2001a; 2001b: Introduction; 2004: 679–82). 21. Sider suggests that, in order for ordinary use to trump the presumption in favor of the 4D quantifi er, roughly the following condition has to be satisfi ed: typical speakers, after being exposed to four-dimensionalist arguments, reject the four-dimensionalist claims, not just as false, but as “linguistically deviant” (2004: 680). No such extraordinary condition has, to my knowledge, ever been imposed on the interpretation of general words, and if it were, the eff ect, I think, would be to force us to reinterpret many words (including, probably, the word “game”). 22. See also Dorr (2005). 194 quantifier variance and realism: essays in metaontology joints” to our language and then proceed in plain English? But this question leads to another one: Do we really have any good idea what is meant by the world’s “logical joints?” Let’s be quite clear that, in the present argument, Sider agrees that plain English is not 4D-English. So the general picture is that ontologists like Sider start out in plain English, and they are able to discover, from within plain English, that the quantifi ers of plain English fail to correspond to the world’s logical joints. But how can we understand the nature of this failure? It is a trivial disquotational truth (in plain English) that the expression “everything that exists” refers to everything that exists. What could be better for a quantifi er than to refer to everything that exists? Sider suggests that the ideal quantifi er must express existence (in italics), and the plain English quantifi er fails to do this. But the plain English word “existence” trivially expresses existence. It’s clear that, for Sider, there is a distinction between existence (in the sense of plain English) and existence. The latter is evidently a technical concept that the ontologist must employ in order to explain what is wrong with plain English. There appears, however, to be no prospect within plain English of explaining what this technical concept means. The “logical joints” and “existence” must apparently be grasped in the philosophy room as undefi ned primitives. Once again, the contrast with Lewis’s naturalness presumption is important. When Lewis suggests that general words are linked to natural properties, he makes a serious eff ort to indicate in plain English what is meant by a “natural property.” His most fundamental explanation, and the one that is widespread in the literature, is that a property is natural if the sharing of it makes for similarities. This explanation faces well- known diffi culties of detail, but there is at least a defi nite intuitive connection between the technical notion of a natural property and the familiar notion of similarity. This intuitive connection is far-reaching, allowing naturalness to apply not just to monadic properties but also to relations, and even to the properties and relations of abstract things.23 Lewis’s notion of a natural property is not off ered as an undefi ned primitive, as is Sider’s notion of the world’s logical joints. The latter notion

23. Sider argues that, since Lewis’s naturalness presumption must apply at abstract levels (e.g., in favoring “plus” to “quus”), this shows that property-naturalness cannot be straightforwardly explained in terms of similarity, any more than the logical joints can (2004: 682). I think, however, that it seems suffi ciently straightforward to say that the triplets of numbers that are related by “plus” are all similar to each other—they have something in common—in the intuitive sense in which the triplets related by “quus” are not all similar to each other. ontological arguments 195 seems, for this reason, far more obscure than the former. Moreover, given the notion of a natural property, Lewis’s interpretive presumption makes immediate intuitive sense. The basic function of general words is to classify things, and there is a strong intuitive link between classifying things and bringing similar things together under a single word. This explains, at least in a rough intuitive way, the link between general words and natural properties. Sider has no comparable way of explaining the alleged link between quantifi er expressions and the “logical joints.” I think that Sider’s idea is that ontological arguments themselves reveal what the arguments are about. Those who have an aptitude for ontology become engaged with these arguments in a meaningful way. They thereby display their understanding of Ontologese and what is meant by the “logical joints,” though they have no way to explain these matters in other terms. This is an intriguing idea, and, though I person- ally fi nd it to be excessively obscure, I certainly cannot refute it.24 I can only urge the practitioners of this form of ontology to make it abun- dantly clear that one thing they are not engaged in is debating in plain English about what exists.25

References Butle r, Joseph. 1836. “Of Personal Identity,” in The Whole Works of Joseph Butler, LL.D. (London: Thomas Tegg), pp. 263–70. Carnap, Rudolph. 1956. “Empiricism, Semantics, and Ontology,” in Meaning and Necessity, 2nd edn. (Chicago: University of Chicago Press). Cartwright, Richard. 1975. “Scattered Objects,” in K. Lehrer, ed., Analysis and Metaphysics (Dordrecht: Reidel); repr. in J. Kim and E. Sosa, eds., Meta- physics: An Anthology (Oxford: Blackwell Publishing, 1999). Chisholm, Roderick. 1976 Person and Object (LaSalle, ILL: Open Court, 1976); repr. in J. Kim and E. Sosa, eds., Metaphysics: An Anthology (Oxford: Black- well Publishing, 1999). Davidson, Donald. 1984. Inquiries into Truth and Interpretation (Oxford: Oxford University Press). Dorr, Cian. 2005. “What We Disagree About When We Disagree About Ontol- ogy,” in Mark Kalderon, ed., Fictionalism in Metaphysics (Oxford: Oxford University Press), pp. 234–86.

24. It would be instructive, I think, to compare and contrast Sider’s position to that of an epistemologist who tries to fend off ordinary language critiques of the traditional claim that we do not perceive external objects, by insisting that the word “perceives,” as used in the philosophy room, is not to be understood in plain English, but in a superior language of Epistemologese. 25. My thanks to Matti Eklund for comments on this paper. 196 quantifier variance and realism: essays in metaontology

Findlay, J. N. 1962. Hegel: A Re-examination (New York: Collier Books). Goodman, Nelson. 1973. Fact, Fiction, and Forecast, 3rd edn. (Indianapolis and New York: Bobbs Merrill). Heller, Mark. 1990. The Ontology of Physical Objects: Four Dimensional Hunks of Matter (Cambridge: Cambridge University Press). Hirsch, Eli. 1999. “The Vagueness of Identity,” Philosophical Topics 26. ———. 2002a. “Against Revisionary Ontology,” Philosophical Topics 30. ———. 2002b. “Quantifi er Variance and Realism,” Philosophical Issues 12. ———. 2004. “Sosa’s Existential Relativism,” in J. Greco, ed., Ernest Sosa and His Critics (Oxford: Blackwell Publishing). ———. 2005. “Physical-Object Ontology, Verbal Disputes, and Common- sense,” Philosophy and Phenomenological Research 70(1): 67–97. Kripke, Saul A. 1982. Wittgenstein on Rules and Private Languages (Cambridge, MA: Harvard University Press). Lewis, David. 1983a. “New Work for a Theory of Universals,” Australasian Jour- nal of Philosophy 61: 343–77. ———. 1983b. “Radical Interpretation,” in Philosophical Papers I (New York: Oxford University Press). ———. 1986. On the Plurality of Worlds (Oxford: Blackwell Publishing). Locke, John. 1690. An Essay Concerning Human Understanding. Merricks, Trenton. 2001. Objects and Persons (Oxford: Oxford University Press). Putnam, Hilary. 1987. “Truth and Convention: On Davidson’s Refutation of Conceptual Relativism,” Dialectica 41: 69–77. ———. 1994. “The Question of Realism,” in Words and Life (Cambridge, MA: Harvard University Press). Quine, W. V. 1960. Word and Object (Cambridge, MA: MIT Press). Sider, Theodore. 2001a. “Criteria of Personal Identity and the Limits of Con- ceptual Analysis,” Philosophical Perspectives 15: 188–209. ———. 2001b. Four-Dimensionalism: An Ontology of Persistence and Time (New York: Oxford University Press). ———. 2004. “Replies to Gallois, Hirsch, and Markosian,” Philosophy and Phenomenological Research 68(3): 674–87. Sider, T., J. Hawthorne, and D. W. Zimmerman, eds. 2008. Contemporary Debates in Metaphysics (New York: Oxford University Press). Thomson, Judith Jarvis. 1983. “Parthood and Identity Across Time,” Journal of Philosophy 80: 201–20; repr. in J. Kim and E. Sosa, eds., Metaphysics: An Anthology (Oxford: Blackwell Publishing, 1999). Unger, Peter. 1980. “There Are No Ordinary Things,” Synthese 41: 117–54. Van Cleve, James. 1986. “Mereological Essentialism, Mereological Conjunctiv- ism, and Identity Through Time,” Midwest Studies in Philosophy 11: 141–56. Van Inwagen, Peter. 1990. Material Beings (Ithaca, NY: Cornell University Press). Williamson, Timothy. 1994. Vagueness (London: Routledge). eleven c C

language, ontology, and structure

Organicists and defenders of common sense disagree over a range of ontological sentences, such as, “There exists a table,” which organicists regard as false and defenders of common sense regard as true. We can imagine two linguistic communities: in the O-community typical speakers accept the sentences accepted by organicists, whereas in the C-community typical speakers accept the sentences accepted by the defenders of common sense. I have maintained that this dispute (given a certain proviso) is merely verbal.1 Each community ought to agree that, on the most plausibly charitable interpretation, the other community speaks the truth in its own favored language. Mathew McGrath objects to this conciliatory outcome.2 He does not challenge my claim that each community ought to apply considerations of interpretive charity to the other community’s language, but he claims that an essential element of charity has been left out of the picture, an element that he calls “charity to expressibility.” McGrath follows my standard terminology in calling the language of the O-community O-English and the language of the C-community C-English, but I think that for the purposes of this discussion it will be

1. The proviso is that the two sides are not disagreeing over whether composites must be composed of simples. I do not regard that issue as verbal. See, below, note 7. 2. McGrath (2008)

197 198 quantifier variance and realism: essays in metaontology helpful to introduce a slightly diff erent terminology. Let O∗-English be an imagined language in which the sentences accepted by the O-community (in the relevant contexts of utterance) are true, whereas C∗-English is an imagined language in which the sentences accepted by the C-community are true. It is a substantive question, then, whether the language of the O-community (i.e., O-English) is O∗-English, and whether the language of the C-community (i.e., C-English) is C∗- English. According to my argument, the C-community ought to agree that, on the most plausibly charitable interpretation, the language of the O-community is O∗-English, and therefore members of that community speak the truth in their language. By the same token, the O-community ought to agree that members of the C-community speak the truth in C∗-English. So the apparent dispute has been resolved. Let me formulate McGrath’s objection from the standpoint of the C-community. If we are members of the C-community trying to interpret the language of the O-community, then if we say that their language is O∗-English we achieve the charitable result that they are right in asserting such sentences as “It’s not the case that there exists a table,” since that sentence in O∗-English is (a priori necessarily) equivalent to the true sentence, “It’s not the case that some table-wise interrelated things compose a living thing.” The trouble is that this charitable result is at the expense of the implausibly uncharitable attribution to the O-community of an expressively defi cient language. From our C-standpoint O∗-English is expressively defi cient because it is impossible in that language to express such facts as that there exists a table. What we have then is a confl ict of charity: there are reasons of interpretive charity both for and against attributing O∗-English to the O-community. McGrath goes on to argue that there are additional considerations that may tilt against this attribution. I’ll come back to those considerations. It’s clear that the “facts” that, from our C-standpoint, cannot be expressed in O∗-English, according to McGrath, are “structured” facts (s-facts) rather than “unstructured” facts (u-facts). Two true sentences (as uttered in a certain context) might be said to express the same u-fact if they hold true in the same possible worlds. A u-fact can be associated with a true Lewisian proposition (i.e, a set of possible worlds containing the actual world). For two true sentences to express the same s-fact something more is required. The rough idea is that the two sentences must be built up (semantically) in the same way. Certainly we Cs will allow that in O∗-English one can express the u-fact that there language, ontology, and structure 199 exists a table, for that u-fact is identical with the u-fact that there exist some table-wise interrelated things. But O∗-English is expressively defi cient because in that language one cannot express the s-fact that there exists a table. Let me make a few preliminary observations before trying to respond to McGrath’s argument. It is not the case that all ontologists who take themselves to be engaged in substantive issues believe in structured facts. David Lewis is a prominent example of one who doesn’t.3 Even if we are dealing with Os and Cs who do believe in structured facts, it is not inevitable that they will have any disagreement about structure. For they may both hold that only simple things and their (perhaps simple) properties and interrelations enter into the structure of the facts. (That is, a common sense ontologist who believes in s-facts may believe in the existence of tables while disbe- lieving that composites like tables can be constituents of s-facts.) McGrath’s argument would seem to be relevant only with respect to instances of Os and Cs who believe in s-facts and disagree about their structure. Let us then confi ne ourselves in what follows to such instances.4 Let’s also be clear about what McGrath is claiming and what he is not claiming. As I understand him, he is not claiming that even if the dispute about what exists is verbal there remains a substantive dispute about the structure of the facts. That claim would seem immediately implausible. One’s views about the structure of the facts are so inti- mately tied up with one’s views about what exists that it couldn’t be plausible to claim that disputes about the former are substantive while disputes about the latter are not. If the dispute about what exists is verbal this must be because the Os speak O∗-English and the Cs speak C∗-English. But, that being the case, each side speaks truly when it describes what it calls “the structure of the facts.” We can suppose that the Os and the Cs agree on the necessary truth of sentences coming under some such rough schematism as the following:

3. Lewis (1986), pp. 56–57. It would be a mistake to think that Lewis’s distinction between natural and unnatural properties commits him to structured facts. His view is that one can employ various kinds of set-theoretical constructions to add structure to unstructured facts. Such constructions might, for some purposes, be in terms of natural properties. Cf, section V, below. 4. In his note 27 McGrath seems to imply that from our C-standpoint O∗-English is expressively defi cient even if we do not believe in s-facts, simply because in O∗-English it is impossible to refer to or describe such objects as tables. But it seems to me that this impossibility could not qualify as an expressive defi ciency unless it implies that some facts cannot be expressed. 200 quantifier variance and realism: essays in metaontology

(S) There exists something composed of things that are R-interrelated if, and only if, there obtains a structured fact whose ordered constituents are: existence; composition; and some additional constituents amounting to things being R-interrelated.

If the Os speak O∗-English then they correctly deny the sentence, “There exists something composed of things that are tablewise interrelated.” Applying the schematism (S) they also correctly deny in O∗-English the sentence, “There obtains a structured fact whose ordered constituents are: existence; composition; and some additional constituents amounting to things being tablewise interrelated.” Like- wise, if the Cs speak C∗-English they correctly assert both of these sentences in their language. The dispute about structure would be as verbal as the dispute about existence. But McGrath (as I understand him) is making a diff erent point. He seems to agree that there is no substantive dispute of any sort on the assumption that the Cs speak C∗-English and the Os speak O∗- English. He claims, however, that neither side should acquiesce to this conciliatory assumption. For each side should view the assumption as requiring an implausibly uncharitable attribution to the other side of an expressively defi cient language.

I I

I will eventually try to argue that neither side ought to view the conciliatory assumption as attributing any defi ciency to the other side’s language. But my most immediate response to McGrath is that it does not seem to me that “charity to expressibility” can bear the weight he wants to place on it. The basic idea of interpretive charity is that there is a presumption that people use their language in a way that makes good sense. Suppose again that we are on the C-side, and suppose for now that McGrath is right that we ought to regard O∗-English as expressively defi cient. That may be, he says, a reason not to interpret the O-community as speaking O∗-English. But what is the alternative? The better alternative, he thinks, might be to interpret them as speaking C∗-English, the language that we take to be our own. How can that make better sense of their use of language? If they speak C∗- English they have the capacity (we’re assuming) to express s-facts that they could not express in O∗-English. But they use this capacity to language, ontology, and structure 201 deny the additional s-facts rather than to assert them!5 That doesn’t seem to picture them as making a more reasonable use of their language. If we interpret the O-community as speaking C∗-English rather than O∗-English this interpretation makes many of their assertions come out false rather than true without making any of their assertions come out true rather than false. It’s hard for me to see that this interpretation is charitable to any signifi cant degree. Charity to expressibility has not fi gured in my work. What does fi gure is something that might be called “non-charity to excessive expressibility.” In order for a dispute to count as verbal in my sense it must allow for a conciliatory outcome, given only that the disputants are prepared to interpret each other charitably. My assumption is that a conciliatory outcome could not require one side to make the following sort of speech: “Although my opponents seem to have essentially the same intellectual and sensory capacities as I have (this is not like a case of someone wandering into the country of the blind), I will assume—charitably—that they speak the truth even though I have no idea what that truth is. I’ll charitably assume that they are asserting true sentences, though I cannot even roughly describe the truth conditions of these sentences.” That is not what I want to count as a conciliatory outcome. I therefore require of a verbal dispute that each side be able to formulate, at least in rough terms, the truth conditions (in any context of utterance) of the other side’s asserted sentences.6 Each side must be able to roughly express the u-facts that it interprets the other side to be expressing (in any context of utterance).7 Now I can imagine someone holding that more is required for a conciliatory outcome. Each side must be able to express not just the

5. McGrtath points out in correspondence that certain compound s-facts will be asserted, for example, the s-fact that, if a table exists, it is composed of simples. This is certainly a relevant point, but I doubt that an appeal to such compound (and trivially derivative) s-facts can sustain McGrath’s argument. 6. I take non-actual contexts into account. One doesn’t have the required kind of understanding of the truth conditions of the other side’s sentence “This is water” if one doesn’t understand what u-fact would be expressed by the sentence on Twin Earth. Only a rough description of the truth conditions of the other side’s asserted sentences is required. There can be verbal disputes in which people use a word (e.g, the word “game,” as in McGrath’s example in his note 32) in somewhat diff erent ways and neither side can spell out with precision the truth conditions of the other side’s assertions involving that word. 7. This is why I am not prepared to regard as verbal either a dispute about the existence of simples or a dispute about the existence of sets. I cannot readily see how in each of these examples there can be a conciliatory outcome which will allow the anti-simples (or anti-sets) side to express even roughly the u-facts that it charitably interprets the pro-simples (pro-sets) side to be expressing. 202 quantifier variance and realism: essays in metaontology u-facts, but the s-facts that it interprets the other side to be expressing.8 It is doubtful that this requirement can be satisfi ed if the two sides are speaking the diff erent ∗-languages. The requirement, however, seems to me implausible. Each side has a suffi ciently adequate grasp of what the other side’s assertions amount to as long as it understands what u-facts are being asserted by the other side. In any case, McGrath’s main argument from charity to expressibility does not imply any such requirement. I will try to show in section IV that neither side ought to regard the other side’s language as expressively defi cient, in which case the argument from charity to expressibility does not get off the ground. But the argument seems to me to fail even if we do regard, from each side, the other side’s language as defi cient. It seems more plausible to view people as making correct assertions in an inferior language than as making incorrect assertions in a superior language.9

III

In this section I want to address some other considerations, apart from charity to expressibility, that are said by McGrath to undermine my appeal to charity in support of the conciliatory interpretation. McGrath argues, to begin with, that there is a distortion in the picture I present of the linguistic communities that are supposed to correspond to the diff erent ontological positions. These are supposed to be communities in which the corresponding ontological positions will not appear deviant. But, for that to be the case, the communities cannot be pictured as consisting of ordinary people making perceptual judgments in the ordinary non-inferential manner. That is surely not how ontologists behave when they present their positions in the philosophy room. McGrath’s criticism applies to the part of my position that says that the ontological disputes between professional philosophers are merely verbal. It does not apply to the part of my position (discussed in McGrath’s section 3) that attempts to defend non-philosophers from

8. Some such requirement may be implied in McGrath’s “side note” on p. 432. 9. Imagine someone who says: “There is this game on T.V. called ‘Jeopardy’ in which the winner is supposed to be the one who displays the most encyclopedic knowledge of arcane information. For some reason everyone involved with the game violates the rules by discounting correct answers that are not put in the idiotic form of a question.” We do not say that, because it is more plausible to suppose that people are following the rules of a less rational game than that they are breaking the rules of a more rational game. language, ontology, and structure 203 the attacks of philosophers, that attempts, that is, to defend the commonsensical assertions of typical members of our linguistic community from the attacks of revisionary ontologists (such as, organicists, mereological essentialists, and four dimensionalists). My defense of common sense consists in an “argument from charity”: the most plausibly charitable interpretation of typical assertions in our community about the existence of physical objects makes these assertions come out true. For the purposes of this argument, why do I need to bring in other imagined linguistic communities at all? Perhaps this is needed to make it clear that I am not playing favorites, that I am prepared to say for any imagined linguistic community, consisting of non-philosophers who make assertions corresponding to the diff erent revisionary positions, what I say about the assertions in our own community. But the deeper reason for bringing in the imagined communities is to illustrate the possibility of “quantifi er variance.”10 It often seems that an implicit assumption of revisionary ontology is that there is only one possible use of quantifi er-like expressions in any language. If that were so, a charitable interpretation of the quantifi ers in our language might become moot. By considering the diff erent imagined communities, consisting of non-philosophers who make assertions corresponding to the diff erent ontological positions, we see that the assertions in all these communities are true. The ostensible disputes between these non-philosophical communities are merely verbal. That suffi ces to defend the commonsensical assertions made by the non-philosophers in our community. McGrath’s criticism does, however, apply to my claim that the disputes between professional philosophers are verbal. The criticism is that in trying to imagine linguistic communities in which the diff erent philosophical assertions would appear non-deviant, it is a distortion to imagine communities consisting of non-philosophers. I will accept this criticism.11 My picture of the communities does indeed appear to involve the distortion that McGrath points out. The question now is whether removing this distortion will seriously undermine my conciliatory argument. I think it will not.

10. Hirsch (2002). 11. The fact is, however, that in a traditional ontological dispute neither side would be likely to grant that, if there were a linguistic community in which ordinary folk spontaneously assert the sentences asserted by the other side, those assertions would be (strictly and literally) true. The traditional ontologists’ assumption that their disputes are substantive do not generally rely on McGrath’s point. 204 quantifier variance and realism: essays in metaontology

Let us follow McGrath’s advice, then, and henceforth take the imagined linguistic communities corresponding to each position to be a “community of persistent philosophers.” (Let’s also follow him in not worrying too much about whether this grisly scenario is really possible.) Consider, again, the O-community, pictured now as consisting of persistent philosophers. The members of this community do not casually assert, “There is no table here,” when confronted with tablewise interrelated things. On the contrary, they are in such circumstances tempted to assert, “There is a table here,” and may even assert this sentence in some “loose” or “fi ctional” manner. But they appeal to some philosophical principles to arrive at the conclusion that (“strictly,” “really”) there are no such things as tables. Indeed, if we are in the C-community, although at the end of the day we reject the philosophical principles that the O-community appeals to, we ourselves feel a defi nite attraction to these principles. What we seem to have here is the makings of a principled dispute, not a verbal dispute. It is therefore no violation of interpretive charity for us to take the O-community to be speaking the same language we speak and to be making a substantive and reasonable philosophical mistake. It seems to me that an essential point is being missed here. The “philosophical principles” that the O-community appeals to are certain general sentences that need to be interpreted, and that ought therefore to be interpreted in the most plausibly charitable manner. The way it generally works in ontology is that each side eventually settles on various general sentences that cohere with its ontological claims.12 For example, organicists typically assert the sentence, “Two things cannot occupy the same place at the same time” whereas common sense ontologists typically deny this sentence. The philosophical C-community ought therefore to acknowledge the charitable presumption that the philosophical O-community employs O∗-English, in which the general sentence is true, and the latter ought to acknowledge the charitable presumption that the former employs C∗-English, in which the sentence is false. The fact that each side appeals to “philosophical principles” invites, rather than defeats, the presumption that each side employs a language in which its principles are correct. I do agree with McGrath’s claim that charity to perception is weaker in the philosophical communities than in ordinary communities in which perceptual judgments about the existence of highly visible macroscopic objects do not appear to be based on conscious inferences

12. Cf. Hirsch (2005), pp. 81-82. language, ontology, and structure 205 from general principles. But this point must not be overstated. The distinction between conscious and unconscious inference is notori- ously nebulous. Probably all perception is theory-laden in some ways or to some degree. The fact is that the philosophical communities do make perceptual judgments, such as, “There is no composite object here,” said by members of the O-community in the presence of a table. Nothing that McGrath has said about the distinctive nature of the philosophical communities seems to undermine the claim that there is an overwhelming presumption in favor of charitably interpreting each side as employing a language in which its assertions, both general and perceptual, come out true. I have the feeling that something that impresses McGrath about the philosophical communities is the uncertainty and tentativeness with which many ontologists hold their positions. Even after “all is said and done” (when “all the tricky arguments and distinctions and counterexamples have been discovered”13), each side may show great respect and intuitive sympathy for the other side’s point of view. Isn’t that a sign that something substantive is at stake? I think not. Let’s bear in mind that closely related to the idea of a “verbal dispute” is the idea of a “verbal question.” Imagine, as a variation of William James’s famous example, that, having explored all of the relevant considerations for and against, someone remains ambivalent about whether or not to say that the guy who runs around the squirrel on the tree has gone around the squirrel. Here we have outright uncertainty, but this evidently does not show that there is some substantive question at stake. A verbal question is roughly one in which, if diff erent answers were given, we would have a verbal dispute. If ontologists remain perpetually uncertain and tentative about their confl icting positions, we might say that we have a verbal dispute mixed with a bit of a verbal question. The important point is that considerations of interpretive charity reveal that there is nothing substantive at issue. There is one further consideration of charity that often attracts a great deal of attention from philosophers, though I think McGrath may agree with me that it has little weight.14 I have encountered critics who say something like this: “You claim that you are being charitable to the disputing ontologists, but it’s obvious that you’re being uncharitable to them. In fact you are insulting them! These philosophers would much prefer to be substantively wrong than to be trivially right in a

13. Lewis (1983b), p. x. 14. See his note 35. 206 quantifier variance and realism: essays in metaontology verbal dispute.” This critic seems to think that interpretive charity is a matter of generosity, as if I am in competition with Peter Singer. The presumption of charity is supposed to be an a priori principle that is partially constitutive of linguistic meaning. Insofar as the disputing ontologists assert the sentence, “We are not engaged in a verbal dispute,” this sentence will fi gure, together with all the other asserted sentences, in arriving at the most charitable interpretation. I would suspect that meta-level, quasi-technical (self-aggrandizing) assertions probably have low priority as supplicants for charity. In any case, it can’t be seriously suggested that the charitable presumption in favor of the correctness of this one assertion threatens to trump the presumption in favor of all of the other assertions made by the ontologists. All things considered, there appears to be an overwhelming argument from charity for the conclusion that the diff erent communities speak diff erent languages. They are engaged in a verbal dispute, and if they say otherwise, they are mistaken about this.15 The verbal dispute between the O and C-communities is centered on diff erent meanings of quantifi er expressions, such as, “(there) exists” and “something.” It may be useful to glance at a dispute that centers on another kind of logical expression. Imagine that two people named Ik and Ek disagree over the sentence, “Either Mary was present or John was present,” when they agree that both were present. Ik holds that the sentence is true, and Ek holds it is false. Ik and Ek, it turns out, always disagree about the truth-values of sentences of the form “p or q” when both p and q are true. Evidently they are engaged in a verbal dispute. Ik uses disjunction only inclusively, whereas Ek uses it only exclusively. If that is so, can each of them express the s-facts that the other can express? The answer to this is unclear.16 What is clear is that, whatever the

15. Some critics have apparently taken me to be predicting that the contending philosophers, once they hear my spiel, will admit that they are engaged in a merely verbal dispute and go home. Actually, I am not optimistic about this. When I say that they ought to admit that it is a verbal dispute I only mean that if they accept what I, but perhaps not they, take to be the correct approach to linguistic interpretation, they must make this admission. 16. I think it can be argued that, unless all truth-functionally equivalent sentences express the same s-fact, the answer must be no. (The argument that follows is an adaptation of an argument shown to me by Matthew McGrath.) Let us fi rst make the plausible assumption A: If C is a truth-functionally compound sentence containing S as a constituent, and S’ expresses the same S-fact as S, then the result of substituting S’ for S in C is a compound sentence C’ such that C and C’ express the same s-fact. Suppose fi rst that Ik’s “p or q” expresses the same s-fact as Ek’s “not both not-p and not-q,” and Ek’s “p or q” expresses the same s-fact as Ik’s “Not both p and q, and not both not-p and not-q.” Since language, ontology, and structure 207 answer, the dispute remains verbal. It suffi ces that, on the conciliatory outcome, each can express the u-facts that the other can express. But suppose each of them defends his position by appealing to general principles. For example, Ik argues from the principle, “It is not the case that either p or q iff it is the case that neither p nor q, and it is the case that neither p nor q iff both not-p and not-q.” Ek argues from the principle, “It is the case that either p or q iff it is either the case that p and not-q or it is the case that q and not-p.” We may imagine that each side has a number of principles of this general sort to appeal to. Does that make their dispute substantive? Surely, it need not. If we imagine a stage has been reached when “all is said and done,” so that they are both aware of all the arguments and principles that they expect to be relevant to the issue, then their dispute about the principles is itself merely verbal. Ik’s favored principles are true for inclusive disjunction and Ek’s for exclusive disjunction. Each side asserts general principles that are true in that side’s version of the language. But suppose they both say, “Of course there are inclusive and exclusive uses of disjunction in ordinary language, but that has nothing to do with the issue we are working on. We are trying to determine whether it is really (strictly and philosophically) the case that either Mary was present or John was present.” Apart perhaps from raising the threshold of obscurantism, this speech does not seem to change anything. So they are now engaged in a verbal dispute about what is really (strictly and philosophically) the case disjunctively. Suppose Ik says to Ek, “I certainly appreciate the force of your arguments, but, on balance, I continue to believe tentatively that either Mary was present or John was present.” Ek is equally reverential and tremulous in declaring his disjunctive commitments. This, surely, won’t get us to say that they are engaged in a substantive dispute. What we can

there is presumably no diff erence in their use of conjunction and negation, this “translation” implies that, in each person’s language, a disjunctive sentence expresses the same s-fact as a truth-functional construction out of negation and conjunction. Why would one say that unless one held that all truth- functionally equivalent sentences express the same s-fact? Suppose, instead, that Ik’s “p or q” expresses the same s-fact as Ek’s “p or q, or p and q,” and Ek’s “p or q” expresses the same s-fact as Ik’s “p or q, and not both p and q.” It then follows from A (together with the assumption that conjunction and negation operate the same way in both languages) that Ik’s “p or q” expresses the same s-fact as Ik’s “Either it is the case that both p or q and not both p and q, or it is the case that both p and q,” and Ek’s “p or q” express the same s-fact as Ek’s “It is both the case that p or q or both p and q, and also the case that not both p and q.” One could not say that unless one held that all truth-functionally equivalent sentences express the same s-fact. 208 quantifier variance and realism: essays in metaontology say perhaps is that their verbal dispute overlays an incompletely resolved verbal question in both their minds. And suppose, fi nally, that they both agree on the sentence, “We are not engaged in a verbal dispute.” But they are.

There remains a signifi cant question raised by McGrath. Even if we view the philosophical communities as speaking diff erent languages and engaged in a verbal dispute, can we agree with McGrath’s suggestion that one of the languages might be expressively defi cient relative to the other? There is an immediate problem with this suggestion. The suggestion is that, if we are on the C-side we should say, “The O∗-language is expressively defi cient because it does not allow one to express the s-fact that there exists something composed of tablewise interrelated things,” and, if we are on the O-side, we should say, “The C∗-language is expressively defi cient because it does not allow one to express the s-fact that there does not exist something composed of tablewise interrelated things.”17 But these complaints seem to be equally valid and therefore cancel each other out. The complaints seem equally valid because (on the assumptions we have made about how the two languages operate) in C∗-English it is true to say, “There obtains the s-fact that there exists something composed of tablewise interrelated things,” and in O∗-English it is true to say, “There obtains the s-fact that there does not exist anything composed of tablewise interrelated things.” What we seem to arrive at is that the languages are expressively equal, each allowing for some s-facts to be expressed that can’t be expressed in the other. I can think of one answer McGrath might give, but I doubt that it is suffi cient. Consider the following two sentences: 1. It is not the case that there exists something composed of tablewise interrelated things. 2. There obtains an s-fact whose ordered constituents are: negation; existence; composition; and some additional constituents amounting to things being tablewise-interrelated.

17. The O-side’s complaint appeals to “negative facts,” which McGrath mentions on p. 431. Certainly many believers in s-facts will want to acknowledge such facts. language, ontology, and structure 209

We are assuming that in both O∗-English and C∗-English (1) is necessarily equivalent to (2). Suppose we are on the C-side, speaking C∗-English. Since (1) is true in O∗-English, so is (2). But, since both (1) and (2) are false in our language, (2) cannot in our language describe the s-fact that (1) expresses in O∗-English. We need not conclude, however, that in O∗-English (1) expresses an s-fact that cannot be expressed in our language. Another alternative is to hold that in O∗-English (1) does not express any s-fact. The reason why (2) is true in O∗-English but false in our language is that the expression “s-fact” does not mean the same in O∗-English as in our language. Since the meaning of “existence” and related expressions varies from one language to the other, so does the meaning of “the structure of the facts.” Let me call this the “restricted” view of s-facts. A u-fact can be viewed as divided up into various s-facts. The restricted view will allow that there may be u-facts inexpressible in our language, and such a u-fact will be divisible into inexpressible-in-our-language s-facts, but u-facts that are expressible in our language cannot be divided up into s-facts that are not expressible in our language. That is why (1) in O∗-English, which expresses a u-fact expressible in our language (i.e., the u-fact that tablewise interrelated things do not compose a living thing), cannot be said to express in O∗-English an s-fact inexpressible in our language. If we hold the “unrestricted” view of s-facts, on the other hand, we will say that (1) in O∗-English does express an s-fact that cannot be expressed in our language, an s-fact that can be correctly described by (2) in O∗-English but not in our language.18 On the restricted view at least part of McGrath’s position seems coherently statable. Each side can correctly assert, “There are s-facts that can be expressed in our language but not in the other side’s language, whereas there are no s-facts that can be expressed in the other side’s language but not in our language.” There still seems, however, to be a major problem here. Granted that each side can correctly make the assertion just mentioned, how can each side correctly draw the conclusion, “The other side’s language is defi cient”? McGrath

18. There are other possible views, but they cannot help McGrath’s argument. It might be held that both languages express the very same s-facts but express and describe them in radically diff erent ways. If we are speakers of C∗-English, what we should say is that the two O∗-English sentences, “It is not the case that there exists something composed of tablewise interrelated things” and “It is not the case that there exists some living thing composed of tablewise interrelated things” express the same s-fact, i.e., the s-fact that it is not the case that there exists some living thing composed of tablewise interrelated things. The reason why it is true to say in O∗-English, “Those two sentences express diff erent s-facts” is that they don’t mean by “(diff erent) s-fact” what we mean. 210 quantifier variance and realism: essays in metaontology certainly would not want to say that what we have here is a verbal dispute about which language is defi cient, each side asserting the truth on its own use of the word “defi cient.” Nor would I want to say that. I assume that value disputes are generally not verbal, since they amount to (or at least bring with them) confl icts between pro-attitudes and con-attitudes that make a conciliatory outcome impossible. McGrath’s idea, clearly, is that each side can criticize the other side’s language for its expressive failures. But how can that be? It cannot be that each side, just because it is speaking its language, can correctly criticize the other side’s language. This begins to look like little more than each side’s exhibiting an absurd form of linguistic chauvinism. One may certainly have the initial feeling that McGrath is right, that if we speakers of C∗-English view O∗-English as capable of expressing fewer s-facts than we can express, then we can reasonably criticize O∗-English as being expressively defective. But this cannot possibly be right, for, if it were, it would lead to the incoherent result that they have as much reason to criticize our language. McGrath’s position runs afoul of the seemingly self-evident principle that a substantive dispute must not be based on a verbal dispute. Our verbal dispute with speakers of O∗-English about whether tables exist, or whether there obtains the s-fact that tables exist, cannot generate a substantive dispute about which language is better. But why isn’t it right for us to criticize their language as being expressively defective, if in their language they are able to express fewer s-facts than we can in ours? I think this puzzling question may motivate us to reject the restrictive view of s-facts, or even to reject s-facts altogether. If we hold on to the restrictive view we must say that there is no value as such in expressing s-facts. We must say that the speech act that in O∗-English is called “expressing an s-fact,” even if this act does not constitute (what we call) expressing an s-fact, has no less value than (what we call) expressing an s-fact. This is why we cannot say that their language is defective, and they cannot say that ours is. If this formulation seems too paradoxical, then we must give up the restrictive view. If we do not believe in s-facts we will say that a u-fact is in a sense divided up in diff erent ways by the structures of the sentences that can express it. We may perhaps consider there to be something of value in being able to express u-facts in structurally diff erent ways (in being able to see the facts under the aspect of diff erent sentential structures). That value, such as it might be, is all the value that the restricted view can assign to expressing s-facts. This is why, on that view, the speech act that language, ontology, and structure 211 in O∗-English is called “expressing an s-fact” has no less value than (what we call) expressing an s-fact.19 However these details are worked out, the essential point is that we cannot allow ourselves to be led to the consequence that McGrath’s position seems to imply. We cannot say that, from the standpoint of each language, the other language is defective. And if we cannot say that, then it seems that the whole issue of structured facts gives us no reason to think that any of the ontological languages are better than any others.20

Throughout the main part of his paper McGrath contends with my claim that many ontological disputes are verbal. Towards the end of his paper he focuses on a quite diff erent question, a question about whether some of the ontological languages are better than others, better, that is, in some sense related to philosophy. Diff erences in s- fact- expressibility, which I have just been discussing, is the main criterion he mentions for evaluating the philosophical virtue of a language. Traditional ontologists did not concern themselves much with questions about the philosophical virtues of alternative languages. In fact, the idea of alternative “ontological languages”—i.e., languages such as O∗-English and C∗-English, in which diff erent ontological assertions come out true—seemed not to fi gure at all in traditional

19. It will come as no surprise that I am inclined to view the dispute between believers and disbelievers in s-facts as merely verbal. Very roughly, on the most plausibly charitable interpretation, one side’s assertions about “the existence of diff erent s-facts” has the same truth conditions as the other side’s assertions about “the existence of diff erent possible sentential structures for expressing u-facts.” But this position would have to be developed, and I have not relied on it in anything that I have said here. 20. Perhaps a capsule summary of the main argument in the text can be helpful. Consider the following two sentences (which make reference to the two languages C∗-English and O∗- English). 1. Our language expresses more s-facts than the other language. 2. Therefore, (we have reason to believe that) our language is better than the other language. McGrath’s argument implies that (1) is true in both C∗-English and O∗-English. It seems clear that (2) cannot be true in both languages. Might (2) be true in only one of them? But this seems impossible to understand. What could determine in which language (2) is true? The answer can’t be that this is determined by ontology, since we are assuming that speakers of the two languages do not have any substantive disagreements of ontology. It must follow that (2) is not true in either language. Neither side ought to claim that its language is better because of s-fact-expressibility. 212 quantifier variance and realism: essays in metaontology ontology. In recent years this has changed dramatically. The “neo- ontologists,” as I would like to call them, have as a central part of their program the attempt to defi ne the philosophically or ontologically best language, the language that ought to be used in the philosophy room, and to determine which sentences are true in that language.21 The clearest announcement of this program has, I think, come from Theodore Sider. Sider wears two hats these days, the traditional hat and the neo-hat. As a traditional ontologist he tries to defend four- dimensionalism as holding true in plain English. But he has said that even if this defense fails there is a more important question that concerns him. This is the question whether four-dimensionalism holds true in the philosophically best language, a language that he sometimes calls Ontologese.22 Let me off er one friendly piece of advice to the neo-ontologists. Sider and his followers often express the intention to actually carry on their conversations in Ontologese when they are in the philosophy room. This would allow them to simulate traditional ontology by engaging in object-level disputes about what exists, the disputes taking place now in Ontologese. My advice is that they should stick to the meta-level and engage in disputes about which sentences are true in the philosophically best language, rather than attempting to speak that best language. Imagine someone who moves to Israel and announces the fi erce intention to speak the world’s oldest language. We do not say that, though it sounds exactly as if he is speaking Hebrew, he is really speaking the world’s oldest language (Sumerian?) and botching it. Even if there were a sign on the door of the philosophy room saying, “All who enter here must intend to speak the philosophically best language,” there is no reason to think that the philosophers who enter the room with that intention will wind up speaking the same language, let alone the philosophically best language. That would depend on what they wind up saying, on what object-level sentences they utter. If one philosopher

21. As Matti Eklud (2009) shows, in my various writings on “quantifi er variance” I have tended to run together two ideas. The fi rst idea is that many typical ontological disputes between Xs and Ys are merely verbal, because the Xs speak the truth in X∗-English and the Ys speak the truth in Y∗-English. The second idea is that X∗-English and Y∗-English are equally good languages. I would now say that that the fi rst idea is most relevant to traditional ontology, and the second to neo- ontology. 22. Sider (2004) and Sider (2009). Sider is open to treating Ontologese as (plain) English governed by the context of the ontology room rather than as simply a technical language. I think that will not aff ect the main question that I am going to raise about how disputing ontologists can come to understand Ontologese in the same way. See, also, the treatment of “the language of ontology” in Dorr (2005). language, ontology, and structure 213 talks like a typical organicist and another talks like a typical common sense ontologist then, despite their protestations that they are both speaking the philosophically best language, it’s probably more plausible to hold that the fi rst is speaking O∗-English and the second is speaking C∗-English. The best hope of avoiding the traditional verbal disputes in the neo-ontology room is by sticking to the meta-level, and arguing about what is true in the philosophically best language.23 But will this be enough? In Sider’s infl uential view Ontologese (the philosophically best language, the language that ought to be used in the philosophy room) is the language whose quantifi er-like expressions correspond to the world’s “natural joints,” to the world’s “distinguished structure.” There seems to be no defi nite connection between what Sider calls “structure” and the structured facts that fi gure in McGrath’s discussion. For one thing, Lewis is one of Sider’s main exemplars of a believer in “structure,” though, as mentioned earlier, Lewis explicitly repudiates structured facts. Moreover, it does not seem that Sider is saying that what makes some ontological languages inferior is that they can’t express certain facts. His idea seems rather to be that “an ideal inquirer . . . must carve the world at its joints in her thinking and language,” and this cannot happen if one is speaking one of the inferior languages, even if one can express all the facts in those languages.24 If Sider is not talking about structured facts, it’s not really clear to me how the image of “structure” is working within his view, or what it can mean to say that a language is aligned to (“cleaves to,” as he puts it) structure in this sense.25 It is clear, nonetheless, that the kind of

23. Sider (2009) suggests that we can insure that everyone in the philosophy room is speaking the same language of Ontologese by issuing several stipulations. One of these has to do with “naturalness,” a notion that I am about to criticize. A second is that “no philosophically contentious sentences count towards your use of [‘exists’].” I think that is not a coherent stipulation (unless it means that one should only be kidding around when one makes contentious assertions). Charity to use is an external constraint constitutive of interpretation. It cannot be controlled by stipulation. Imagine a religious community that stipulates that more interpretative charity should be bestowed on their sincere religious utterances than on their sincere secular utterances. 24. Sider (2009). 25. Might Sider be talking about a structure of dependence between u-facts, wherein one u-fact obtains by virtue of another u-fact obtaining? I fi nd it diffi cult to fi t that into his formulations. I’m not sure, to begin with, what it could mean to say that a language is aligned to structure in this sense. To the extent that I am able to make out what such an alignment might amount to, it looks to me that the ontological language corresponding to nihilism (rather than Sider’s favored language corresponding to four-dimensionalism) is likely to be the one most aligned to a structure of dependence, for ontologists of many stripes may well agree that all facts obtain by virtue of facts about simples. (The possibility of gunk will complicate this, but Sider is surely not staking everything on that possibility.) 214 quantifier variance and realism: essays in metaontology

“ structure” that Sider considers to be most relevant to ontology—what he calls “quantifi cational structure”—has something very much to do with existence. He says indeed that “existence is a logical joint in reality.”26 It seems to follow immediately that the verbal disputes about what exists that occupied traditional ontologists are simply going to be transformed by the neo-ontologists into verbal disputes about reality’s logical joints and quantifi cational structure. Sider wants to ask, “Which ontological language is aligned to the world’s quantifi cational structure?” but it seems that in asking this he is trying to stand “both inside and outside language” at the same time.27 I would certainly expect him to accept the following principle P: “If the sentence ‘There exists something composed of Socrates’ nose and the Eiff el Tower’ is true in a certain language, then that language might be the one aligned to the world’s quantifi cational structure only if part of the world’s quantifi cational structure consists in the fact that there exists something composed of Socrates’ nose and the Eiff el Tower.” Let FD∗- English be an imagined language in which all of the four-dimensionalists’ ontological assertions come out true. Now, the sentence, “There exists something composed of Socrates’ nose and the Eiff el Tower,” is true in FD∗-English, but not in either C∗-English or O∗-English. It follows in a few seemingly obvious steps that the sentence, “FD∗- English is aligned to the world’s quantifi cational structure” might be true in FD∗-English, but cannot be true in either C∗-English or O∗- English.28 This illustrates the point that the neo-ontological disputes about which sentences are true in Ontologese, the language that is aligned to the world’s quantifi cational structure, merely recapitulates the traditional ontologists’ verbal disputes about what exists. It will be seen that the above argument assumes the following principle Q: “There exists something composed of Socrates’ nose and the

26. Sider (2004), p. 646. 27. Kripke (1982), p. 12. 28. Let me spell this out for the case in which we speak C∗-English. 1. The principle P 2. The sentence, “There exists something composed of Socrates’ nose and the Eiff el Tower,” is true in FD∗-English. 3. It is not the case that there exists something composed of Socrates’ nose and the Eiff el Tower. 4. There exists something composed of Socrates’ nose and the Eiff el Tower iff part of the world’s quantifi cational structure consists in the fact that there exists something composed of Socrates’ nose and the Eiff el Tower. 5. (from 3 and 4) It is not the case that part of the world’s quantifi cational structure consists in the fact that there exists something composed of Socrates’ nose and the Eiff el Tower. 6. (from 1, 2, and 5) FD∗-English is not aligned to the world’s quantifi cational structure. language, ontology, and structure 215

Eiff el Tower iff part of the world’s quantifi cational structure consists in the fact that there exists something composed of Socrates’ nose and the Eiff el Tower.” Might Sider deny Q? I would fi nd that barely intelligible, especially since Sider has told us that existence is a logical joint in reality. There is a twist in the above argument that may be easy to miss. The argument assumes that the principles P and Q hold true in all of the ontological languages. But how can Sider deny that? Let’s review the dialectical situation. Sider’s reason for shifting to Ontologese is that this is supposed to be a remedy for the possibility that traditional ontological disputes are verbal. That is, the shift to Ontologese is a remedy for the problem that diff erent ontologists may speak diff erent languages. The remedy consists in explaining to these philosophers how Onto- logese works and getting them to carry on their discussions in (or, on my advice, about) Ontologese. A central part of the explanation of Ontologese is the idea of quantifi cational structure, an idea that seems clearly to imply the principles P and Q. Therefore, the ontologists, who we are now imagining to speak the diff erent languages, are in eff ect instructed by Sider to accept P and Q. It follows from Sider’s account, therefore, that P and Q are true in all the languages. The challenge for Sider is to address a group of philosophers who mean diff erent things by (their most strict and literal use of) the quantifi er, and get them all to mean the same thing by “quantifi cational structure.” If it is not immediately clear that this is an impossible task, the above argument is an attempt to make it clear. What Sider evidently wants is that such principles as P and Q should hold in a privileged sense of “existence” (in the sense of Existence, with capitals and italics), the sense belonging to the best ontological language. The trouble is that precisely what the neophyte neo-ontologists need from him is an explanation of what it means for one ontological language to be better than another. As things stand each ontological camp will have as much reason as any other to regard as privileged the sense of “existence” in its ontological language. This is why one gets the feeling that Sider is trying to issue his explanations from a position outside any of the ontological languages. This, however, he cannot do. The philosophers who enter the neo-room have no alternative but to understand his explanations in terms of the diff erent ontological languages they initially bring with them. Let me, then, restate the earlier argument in more general terms. Sider’s explanations of quantifi cational structure, addressed to the philosophers congregating in the neo-ontology room, who we imagine start out speaking the diff erent ontological languages, 216 quantifier variance and realism: essays in metaontology imply that the following two sentence-schema are correct in all the languages. 1. A necessary and suffi cient condition for an ontological language L to be aligned to the world’s quantifi cational structure is that any sentence in L of the form “There exists such-and-such” is true iff part of the world’s quantifi cational structure consists in the fact that there exists such-and-such. 2. There exists such-and-such iff part of the world’s quantifi cational structure consists in the fact that there exists such-and- such. It follows that in C∗-English it is true to say, “C∗-English, but not O∗-English or FD∗-English, is aligned to the world’s quantifi cational structure”; in O∗-English it is true to say, “O∗-English, but not C∗-English or FD∗-English, is aligned to the world’s quantifi cational structure;” and so on for all the other ontological languages. General- izing, it is true to say in each language: 3. This ontological language (the one currently being used), but no other ontological language, is aligned to the world’s quantifi cational structure. The answer to Sider’s question, “Which ontological language is aligned to the world’s quantifi cational structure?” depends on which language one is speaking. The question is purely verbal; it has no substantive content.29 We should note a distinction between similarity-based and ontology- based “natural joints.” The notion of a natural property or kind is closely related to similarity. Sharing the property of being green strikes us as making things similar in a sense in which sharing the property of being grue does not make things similar. A natural property might be defi ned intuitively as a property whose (actual and possible) instances are more similar to each other than to other things.30 Similarity-making

29. It would be a misunderstanding to suggest that Sider can answer this objection by appealing to “reference magnetism” (i.e., the presumption in favor of a more “eligible” interpretation). His reason for shifting to Ontologese is precisely not to have to rely on the (dubious, as I think) claim that reference magnetism trumps charity. It should also not be suggested that, even if reference magnetism does not trump charity, there can still be a substantive dispute about the direction of the reference magnet (about which language is more eligible), for there cannot be a substantive dispute based on the verbal dispute about quantifi cational structure. (Sider [2009] indeed states that the explanation of Ontologese need make no appeal to reference magnetism.) 30. See Quine (1969) and Lewis (1983a). Several ways of refi ning the notion of a similarity- making property are formulated in Hirsch (1993), appendix 2. language, ontology, and structure 217 considerations might also be applied directly to propositions and facts. Suppose we associate a proposition (as in Lewis) with a set of possible worlds, and assume (as in Lewis’s theory of counterfactuals) that a similarity relation is defi ned over worlds. Then a “natural proposition” might be defi ned as a set of worlds more similar to each other than to other worlds. A “natural fact” is a true natural proposition.31 My skepticism here about natural joints does not pertain to natural properties or propositions defi ned in terms of similarity, but only to Sider’s notion of quantifi cational joints or structure, a notion that is grounded in ontology rather than similarity.32 Because it is grounded in ontology, disputes about its application can be no more substantive than traditional ontological disputes.33 There is a further diffi culty. Ontologese is supposed to be the ontologically best language, the language that ought to be used in the ontology room. According to Sider that language is the one that cuts at the joints, that is aligned to the world’s quantifi cational structure. Now if I am right in claiming that disputes about which ontological language cuts at the joints are merely verbal, these disputes cannot ground substantive disputes about which language is best. That repeats a point that I made in connection with McGrath’s evaluation of the languages. But there is an additional diffi culty in Sider’s case. McGrath based his evaluation of the languages on the plausible principle that a language is better if it can express more facts. I tried to show that the principle can

31. A “basic proposition” might be defi ned as a natural proposition that does not entail a conjunction of natural propositions (neither of which entails the other). Other developments along these lines might be contemplated. 32. Sider (2009), however, very briefl y suggests that quantifi cational structure can actually be explained in terms of similarity. If I understand him, the suggestion is that the distinguished quantifi er expression is the one that would allow us to say that two true sentences of the form “There exists something that is F” and “There exists something that is G” express facts that are similar with respect to (being about) existence. Apart from the dubiousness of the notion of “similarity with respect to existence” I think there is another reason why this suggestion cannot work. Let “shmexistence” be defi ned as existence restricted to things sharing some natural property (e.g., to living things or to simples or to neutrons). Then I think it is clear that if it is correct to say that true sentences of the form “There exists something that is F” and “There exists something that is G” express facts that are similar with respect to existence, it must also be correct to say that true sentences of the form “There shmcxists something that is F” and “There shmexists something that is G” express facts that are similar with respect to shmexistence. Sider’s test seems therefore incapable of picking out the distinguished quantifi er. 33. At the end of Hirsch (2008) I criticize Sider’s notion of ontological joints as being excessively obscure, leaving it open that disputes about the joints may yet be obscurely substantive. In the present paper I have arrived at a more defi nite negative judgment. 218 quantifier variance and realism: essays in metaontology lead to trouble when we are dealing with s-facts, but the principle is clearly plausible. On the other hand, what is the plausibility of Sider’s principle that we ought to speak a language that is aligned to the world’s quantifi cational structure? Are we like lazy surgeons who like to cut at the joints? Perhaps the reference magnet exerts its attraction, but is it more virtuous to succumb or resist? Sider says that the ideal inquirer ought to discern the world’s structure. Suppose I am wrong in claiming that the question about the world’s quantifi cational structure is merely verbal, and suppose that the answer to this substantive question is that FD∗-English is the language that is aligned to the world’s structure. Then it would be important for the ideal inquirer to acknowledge that fact. Therefore, if she speaks C∗-English she ought to assert, “FD∗-English is the language that is aligned to the world’s structure.” Why would it be an additional virtue, beyond knowing what the world’s structure is, for her to switch to FD∗-English? Why ought she to try to speak a language that in some sense replicates that structure? Sider seems to suggest that, apart from such standard cognitive virtues as knowledge, true belief, and rational belief, there is the additional irreducible virtue of: speaking a language that is aligned to the world’s structure. This suggestion seems to me interesting but ultimately mystifying.34 My main criticism of Sider’s position, however, is that there is no substantive question about the world’s quantifi cational structure. The diff erent answers to that question in the diff erent ontological languages cannot be the bases for judging that (or explaining why) one ontological language is better than another. The impression I’m left with after looking at both McGrath’s and Sider’s attempts to evaluate the diff erent ontological languages is that the neo-ontologist’s search for the best language is chimerical.35

References Dorr, C. 2005. “What we disagree about when we disagree about ontology.” In M. Kalderon (ed.) Fictionalist Approaches to Metaphysics. (Oxford: Oxford University Press). Eklund, M. 2009. “Carnap and Ontological Pluralism.” In D. Chalmers, D. Manley, and R. Wasserman, (eds.), Metametaphysics. (Oxford: Oxford Uni- versity Press).

34. Related issues are discussed throughout Hirsch (1993), especially chapters 6 and 7. 35. My thanks to Mathew McGrath and Theodore Sider for helpful comments on this paper. language, ontology, and structure 219

Hirsch, E. 1993. Dividing Reality. NY: Oxford University Press. ———. 2002. “Quantifi er Variance and Realism.” Philosophical Issues 12, 51–73. ———. 2005. “Physical-Object Ontology, Verbal Disputes, and Common Sense.” Philosophy and Phenomenological Research 70, 67–97. ———. 2008. “Ontological Arguments: Interpretive Charity and Quantifi er Variance.” In J.Hawthorne, T. Sider, and D.W. Zimmerman (eds.) Contem- porary Debates in Metaphysics. Oxford: Blackwell. Kripe, S. 1982. Wittgenstein on Rules and Private Language. Cambridge, Mass: Harvard University Press. Lewis, D. 1983a. “New Work for a Theory of Universals.” Australasian Journal of Philosophy 61,343–77. ———. 1983b. Philosophical Papers, Volume 1. (NY: Oxford University Press). ———. 1986. On the Plularity of Worlds. (Oxford: Basil Blackwell). McGrath, M. 2008. “Conciliatory Metaontology and the Vindication of Common Sense,” Nous 42, v. 3, 482–508. Quine, W.V. 1969. “Natural Kinds.” In Ontological Relativity and Other Essays. New York: Columbia University Press. Sider, T. 2004. “Precis of Four-Dimensionalism, and Replies to Critics,” Philoso- phy and Phenomenological Research 68, 642–47, 674–87. ———. 2009. “Ontological Realism.” In D. Chalmers, D. Manley, and R. Wasserman (eds.), Metametaphysics. (Oxford: Oxford University Press). twelve c C

ontology and alternative languages

I. Introduction

C Rudolph Carnap said that issues of ontology amount to nothing c more than choosing one language or another.1 My position is roughly Carnapian, but with three qualifi cations. First, whereas Car- nap’s formulation sometimes seems to suggest an anti-realist or verifi - cationist perspective, my position is robustly realist. I take it for granted that the world and the things in it exist for the most part in complete independence of our knowledge or language. Our linguistic choices do not determine what exists, but determine what we are to mean by the words “what exists” and related words. These words, corresponding to the quantifi ers of our language, can vary in meaning from one language to another.2 Second, I do not understand Carnap’s explanation of why ontological issues reduce to linguistic choices. He says that ontological questions are “external,” but this seems merely to give a name to the problem. It’s evident that he considers some a priori (or non-empircial) issues to be external and some not, but it seems unclear how he explains which are and which aren’t. If the explanation is verifi cationist, appealing

1. Carnap 1956. 2. The possibility of quantifi er variance is also central to many of Hilary Putnam’s ontological writings, but again with an anti-realist slant foreign to my own thinking. See, e.g., Putnam 1987, Putnam 1994.

220 ontology and alternative languages 221 to the idea that ontological issues are hard or impossible to resolve, I reject that explanation. In my view, an issue in ontology (or elsewhere) is “merely verbal” in the sense of reducing to a linguistic choice only if the following condition is satisfi ed: Each side can plausibly interpret the other side as speaking a language in which the latter’s asserted sentences are true. Much of this paper is an attempt to clarify this condition. It follows that I cannot agree with Carnap’s apparent blanket assumption that all issues of ontology are verbal. It may well be that some ontological issues satisfy the condition just mentioned and some do not. I will suggest towards the end of this paper that the dispute between platonists and nominalists does not, on the face of it, satisfy the condition, and this dispute may therefore not be verbal. One central kind of ontological dispute that I think is verbal concerns which sets of (successions of ) bits of matter constitute a unitary physical object. Disputes of this kind have been pervasive in the recent literature. Per- durantists, endurantists, mereological essentialists, four dimensionalists, and sundry nihilists have engaged each other in lengthy and often highly theoretical disputes, though many of these disputes are, on my position, merely verbal. In this paper I will focus on the issue between perdurantists and endurantists to illustrate the sense in which it seems to me that disputes in the ontology of physical objects are verbal. My arguments here in regard to the verbalness of this issue are to be understood as generalizable to many of the other issues in physical-object ontology. The third diff erence from Carnap is that a central part of my project is to defend common sense, whereas this doesn’t seem to be a concern for Carnap. Revisionist ontological views, such as perdurantism, have dominated the recent literature. I argue that the revisionists are in eff ect merely choosing to use a language diff erent from ordinary language, and, insofar as they are not aware of this (and take themselves to be using ordinary language), they are making a certain kind of verbal mistake. The structure of the paper is as follows. In section II I allude to the religious dispute between Jews and Christians in order to bring out the point that merely having alternative languages that make both sides right doesn’t make a dispute verbal; it is essential that the two sides be plausibly interpreted in terms of these languages. The nature of linguistic interpretation, and its role in understanding the verbalness of certain disputes, is broached in section III. Section IV deals with a number of semantic problems for my view. The fi nal section V addresses the issue 222 quantifier variance and realism: essays in metaontology of platonism versus nominalism. There is in addition an appendix containing a small parable about verbal disputes in ontology. Let me make explicit a few assumptions that operate throughout this paper. Endurantists may well allow that such items as processes and histories must have temporal parts. I want to stipulate that such items are to be ignored throughout this discussion. The primary dispute between endurantists and perdurantists concerns the temporal parts of ordinary physical bodies (trees, rocks, tables) that can be straightforwardly said to occupy a volume of space, to be composed of matter, and to have shape, size, and color. If it can be shown that this primary dispute is verbal, it seems clear that any residual issues about such items as processes and histories can be dealt with along the same lines. I take endurantists and perdurantists to be engaged in an ontological dispute about the existence of temporal parts. That is all they (need to) disagree about. They do not, in particular, (need to) disagree about the conditions for the diachronic unity of an ordinary body (both sides may accept an analysis of diachronic unity in terms of spatiotemporal and causal continuity, or both sides may reject such an analysis). My general assumption is that ontological disputes concern matters of a priori necessity. Perdurantists assert, and endurantists deny, that a tree (rock, table) is made up of a succession of temporal parts. Both sides are to be understood as defending their claims on grounds of a priori necessity. Even if perdurantists intended to make an empirical claim about temporal parts, and endurantists intended to deny this claim on empirical grounds, the dispute would arguably be merely verbal, but that is not a question that I’m going to try to settle here.3

3. John Hawthorne (Chalmers, Manley, and Wasserman 2009) states that many recent ontologists—he mentions Lewis, in particular—do not appeal primarily to a priori arguments. He and I must be using “a priori” in diff erent senses. In Lewis (1986) the main (or only) argument for perdurantism derives from an analysis of the nature of intrinsic change. His main (or only) argument there for unrestricted mereology derives from an analysis of vagueness. These arguments are paradigmati- cally a priori in the sense of appealing not to sensory experience or empirical facts, but to reasoning, understanding, and intuitive insight. The same holds for Sider’s arguments for perdurantism and unrestricted mereology, which appeal primarily to his analyses of vagueness and various puzzle cases (Sider 2002). I think it is clear that many of the most prominent arguments in recent ontology (in Chisholm, Cartwright, Thomson, Lewis, Shoemaker, Unger, Sosa, Van Inwagen, Van Cleve, Sider, and many others) are a priori in the relevant sense. Insofar as some (not all) of these philosophers will now and then gesture towards the facts of empirical science, I ignore that element of their arguments and focus on what is far and away the more dominant a priori element. I think what Hawthorne must mean is that revisionary ontologists often adopt the speculative tone of high-level theorists rather than the tone of philosophers engaged in straightforward conceptual or linguistic analysis. That may well be, but their main arguments, whatever their speculative or theoretical tone, are a priori rather than empirical. ontology and alternative languages 223

II. Alternative Languages

We can imagine a proponent of endurantism (call her Edna) who, for some reason, decides to try to pass herself off as a perdurantist. As she doesn’t want to lie outright she hits on the idea of secretly adopting a language of her own that would allow her to assert the same sentences that perdurantists assert while secretly remaining true to her endurantist convictions. In her diary she writes: “Henceforth I will use the expression ‘temporal part of an object’ when I want to talk about how an object is at a certain time. I’ll say, ‘Lincoln had in 1860 a temporal part that was bearded’ to describe the situation in which Lincoln was bearded in 1860. In general, I’ll use a sentence of the form ‘a has at time t a temporal part that is F ’ to be true of any situation in which a is F at t (where ‘F ’ is a term—unlike ‘adolescent’—that applies to an object at a time by virtue of how it is at that time).” We can also imagine a perdurantist (call him Pedro) who adopts a secret language that will enable him to sound like an endurantist. Pedro writes in his diary: “Henceforth I will in every context restrict my quantifi er to objects accepted by endurantists—roughly, objects other than (proper) temporal parts of ordinary bodies.”4 In Edna’s secret language the sentences asserted by perdurantists are true. That, of course, was the point of the language. Let us call this language P-English. On the other hand, in Pedro’s secret language, in which the quantifi ers were restricted to objects acknowledged by endurantists, the sentences asserted by endurantists are true. Let us call this language E-English. When we ask whether objects have temporal parts it seems that our answer ought to be “yes” if our language is P-English, and “no” if our language is E-English. In this sense the question seems to be merely verbal. It just depends on which language we are speaking. Couldn’t one do this trick for any question? Couldn’t one make any dispute appear verbal by somehow fi nding alternative languages that will make each position come out right? The answer is that, for any typical dispute outside philosophy, we cannot fi nd relevant alternative languages. Before pursuing this point, let me fi x some terminology. I’ll follow Lewis in taking a “proposition” to be a set of possible worlds.5 And I’ll follow Kaplan in taking a sentence’s “character” to be a function that assigns to the sentence, relative to a context of utterance,

4. A larger tale about Edna and Pedro is presented in the appendix. 5. Lewis 1986. 224 quantifier variance and realism: essays in metaontology a proposition (the proposition being the set of worlds in which the sentence holds true).6 The character can also be said to give the sentence’s “truth conditions” (relative to a context of utterance). By the “interpretation” of a language I’ll mean a function that assigns to each sentence of the language a character. Note that interpretation in this sense is defi ned in terms of the characters of sentences, not in terms of the reference of expressions. I assume that, at least for our present purposes, a language is individuated by an interpretation; that is, distinct languages do not have the same interpretation. In a dispute between two positions I’ll say that an “alternative language” for a given position is a language in which proponents of that position could express all of the (object-level) propositions they believe while asserting only sentences that proponents of the other position would assert.7 (Note that sentences are here being individuated phonetically, so that the same sentence can have diff erent characters in diff erent languages.) P-English is in this sense an alternative language for endurantists, and E-English is an alternative language for perdurantists. No restrictions on the semantic structure of an alternative language are assumed beyond the following: the set of characters of the sentences in the alternative language is the same as the set of characters of the sentences in the original language, the characters being merely redistributed over the sentences in shifting from one language to another. (To the extent that one can associate a sentence’s character with its “meaning” the stipulation is that in shifting from one language to another nothing is gained or lost in what can be “meant” by one’s asserted sentences.) I am claiming that E-English and P-English are alternative languages that render the dispute between endurantists and perdurantists verbal. I now want to explain, however, that in most disputes, even if there are alternative languages, they cannot function in the manner required to render the dispute verbal.

6. Kaplan 1989. I use the patently technical expression “character” throughout this paper, rather than the more common “truth conditions,” because I have found that some people apparently fi nd it too diffi cult to hear the latter in a coarse-grained sense, even when so stipulated. It is of course a major substantive question whether the coarse-grained notion can actually accomplish what I want in this discussion. Although it may have little bearing on the kinds of examples that concern me here, I should note that I take non-actual contexts to fi gure in character. “This is water” and “This is 2H O” do not have the same character, because they have diff erent truth conditions as uttered on Twin Earth. 7. Special problems about meta-level semantic propositions will be addressed in section IV. Until then let us restrict ourselves to object-level propositions. ontology and alternative languages 225

In the middle ages many Jews pretended to be Christians to avoid persecution. They simply lied about this. Was there an alternative language for them in their dispute with the Christians, which they could have adopted as a secret language that would have enabled them to express all of their believed propositions while sounding like Chris- tians? (Of course, this is merely a philosophical illustration; I have no reason to believe that Jewish practices would encourage any such shenanigans.) In the secret language the sentence “Jesus walked on water” might express the proposition that Moses descended from a mountain, and “Moses descended from a mountain” might express the proposition that Jesus did not walk on water. Then the Jews could assert both of these sentences as the Christians do. Could such a program be carried out successfully to cover all of the sentences in the language? In the previous example, the obvious idea would be that in the secret language “Jesus” refers to Moses, “Moses” refers to Jesus, “walked on water” refers to things that descended from a mountain, and “descended from a mountain” refers to things that did not walk on water. But suppose, now, that a certain man is observed to descend from a mountain. The Jews would soon perish if they had to assent to the sentence, “That man walked on water.” Might it be that in the secret language “that man” does not refer to the man being pointed at? Evi- dently complications are going to ramify here very quickly. Let’s bear in mind, however, that since we make no prior assumptions about the semantic structure of the secret language, it need not turn out that the character of a sentence in this language is determined in the standard manner by referential relations between words and objects. My question must not be confused with Putnam’s famous point in his model-theoretic argument against realism.8 Putnam’s point would be that, so long as the sentences asserted by Christians are logically consistent, there will be a model that makes these sentences true (assuming enough things in the world). Putnam would conclude that on a realist notion of truth the Jews could not have had any reason to deny the Christian’s assertions. It was only their pragmatism that caused many of these Jews to submit to death rather than agree with the Christians. My presuppositions, however, are completely realist, and I am not asking a question about models. I am asking whether the Jews could possibly express their believed propositions using the same sentences Christians use to express their believed propositions (where the

8. Putnam 1981. 226 quantifier variance and realism: essays in metaontology set of characters to be distributed over these sentences are kept fi xed). An affi rmative answer to this question does not follow from there being models for both the Jews and the Christians that would make the sentences they assert come out true. There are really two questions to be asked here. The fi rst is whether there is in logical space a possible secret language for the Jews. The second is whether the Jews could possibly learn such a language. The answer to the fi rst question would be trivial if we ignored the element of context in character. We could then imagine all of the sentences acceptable to Christians arranged in a sequence C, and all of those acceptable to Jews arranged in a sequence J. In the secret language, the character of the nth sentence in C can simply be the character of the nth sentence in J. A trivial version of this scheme, which ought to be learnable (given certain obvious assumptions), would make corresponding sentences in the sequences contradictories if the sentences are disputed, and make them identical if the sentences are not disputed. (So any disputed sentence will secretly express the negation of the proposition it normally expresses.) Once we take context into account, however, this simple kind of scheme can’t work. Consider the sentence A: “That man has been damned to hell.” This sentence will be acceptable in some contexts but not in others, both to Jews and Christians, though the two groups will disagree in some contexts. In the secret language A’s character must be such that Jews will accept A in the secret language in all and only contexts in which Christians accept A in the standard language. In order to satisfy the condition that the two languages contain sentences with the same characters, there must then be a sentence B such that A’s character in the secret language is the same as B’s character in the standard language, and Jews will accept B in the standard language in all and only contexts in which Christians accept A in the standard language. What could sentence B be? Let us take this a step further. There must now be a sentence C such that B’s character in the secret language is the same as C’s character in the standard language, and Jews will accept C in the standard language in all and only contexts in which Christians accept B in the standard language. And then there must be a sentence D that stands to C in the way that C stands to B and B stands to A, and so on. What kind of a sequence of sentences are we looking at here? Is there any such sequence in the language? If not, there is no possible secret language for the Jews. A variation of a “fi ctionalist” strategy may solve the problem, as long as we can assume that both sides are fully apprised of the implications ontology and alternative languages 227 of the other side’s position.9 (Perhaps we will also have to resort to additional stipulations to deal with a few special examples involving meta-linguistic assertions.) Let a sentence of the form “On the assumptions of Judaism, A” have the same character in the secret language that A has in the standard language.10 And, if B is a sentence that does not begin with the words “on the assumptions of Judaism,” let B have the same character in the secret language that “On the assumptions of Christianity, B” has in the standard language. If in some context the Jews disagree with the Christians’ belief that a certain man was damned to hell, they can pretend to agree by asserting the sentence “That man was damned to hell,” which in the secret language means that this is what the Christians think. They could also express their secret belief that the man was not damned to hell by asserting the sentence “On the assumptions of Judaism, that man was not damned to hell,” which the Christians will have no trouble accepting. So it appears that there is indeed this peculiar alternative language for the Jews. And, by the same reasoning, there will be one for the Christians. It does not, however, follow absurdly that this dispute is merely verbal. Even if there are alternative languages for this dispute, they are not the right kind. In a verbal dispute it must be possible to make the following speech to each side: “You should agree that, on the most plausible interpretation of the other side’s language, the other side asserts the truth by your own lights.” But it could not be suggested to the Jews that their secret language is a plausible interpretation of the Christian’s assertions. The Jews cannot hold that when the Christians assert “That man is damned to hell” and “On the assumptions of Juda- ism, that man is not damned to hell” they are asserting propositions that hold true in worlds in which the man is not damned to hell but, on the assumptions of Christianity, he is. Though the Christians’ assertions, so interpreted, would be true by Jewish lights, this interpretation is obviously absurd.11

9. For the purposes of this discussion, a fi ctionalist strategy is relevant only if fi ctionalist sentences are viewed as strictly and literally true. The fi ctionalist strategy employed here corresponds to “meta-fi ctionalism” in Yablo 2002. 10. I’ll assume that the context will make clear when quotation marks are to be construed as Quinean square quotes. “On the assumptions of Judaism, A” refers to the sentence consisting of the words “on the assumptions of Judaism” followed by the sentence A. 11. Why exactly is it absurd? The most obvious answer, perhaps, is that Jews and Christians don’t just diff er over what sentences they assert, but also over their non-linguistic behavior and attitudes associated with these sentences. They diff er, for example, over their treatment and attitudes toward someone about whom one side asserts and the other side denies, “He is damned to hell.” See, further, note 24, below. 228 quantifier variance and realism: essays in metaontology

One lesson to be derived from this example is that an issue can be perfectly substantive even though there is, for each side, a possible language in which the sentences asserted (in any context) by that side come out true. More is needed for an issue to degenerate into “merely a matter of choosing a language.” It is required that each side (in a sense to be explained) ought to fi nd it plausible to interpret the other side as speaking the truth in the other side’s language.

III. Verbal Disputes and Interpretive Charity

Since languages are here individuated by interpretations (i.e., by assignments of characters to sentences) P-English and E-English are considered diff erent languages, though they can also be called diff erent versions of English. Even when disputants use an English expression in diff erent conventionally correct senses they can be said to (momentarily) use “diff erent languages” in the relevant sense. One may be initially tempted, then, to defi ne a verbal dispute as one in which each side speaks the truth in its own language. That cannot be right, however, since a dispute can be verbal even when the disputants are mistaken about the non- linguistic facts. There may be a verbal dispute about whether Harry is running around the squirrel or merely running around the tree containing the squirrel, when in fact Harry is home in bed and the runner is someone who looks like Harry. Moreover, if we have a case in which the dispute at the fi rst level will inevitably lead to a dispute at the second level as to whether each side in the fi rst level speaks the truth in its own language, then, whatever might be our own judgment about the second level question, I don’t think we want to treat the dispute as merely verbal. At least we will not want to treat it as verbal if we are thinking of a verbal dispute as one that can be dissolved or defl ated by getting the parties to agree that “it’s just a matter of choosing a language.” I’ll try to explain later that in the dispute between nominalists and platonists, whereas the platonists will be able to interpret the nominalists as speaking the truth in their own language, the nominalists may be unable to make that concession to the platonists. One might suggest that in this case the platonists and nominalists have a second level dispute about whether their fi rst level dispute is verbal, but I would prefer to say that the fi rst level dispute in this case is defi nitely not verbal, since it can’t be dissolved or defl ated in the characteristic manner of verbal disputes. I would therefore defi ne a verbal dispute as follows: It is a dispute in which, given the correct view of linguistic interpretation, each party ontology and alternative languages 229 will agree that the other party speaks the truth in its own language. This can be put more briefl y by saying that in a verbal dispute each party ought to agree that the other party speaks the truth in its own language. There can still be a philosophical question about whether a given dispute is verbal, but this question would have to do with issues of linguistic interpretation. There is a snag in this defi nition that I simply want to circumvent, since I think it doesn’t have a critical bearing on what I am driving at. Burge and others have claimed that, if a person belongs to a certain linguistic community, then the language of that community is the person’s language, and no matter what idiosyncrasies the person’s linguistic behavior exhibits, the conventionally correct character must be assigned to the person’s asserted sentences.12 Burge’s point would make it diffi - cult to say in general that in a verbal dispute each disputant has her own language.13 To get around this, I’m going to stipulate that for the purposes of this discussion, the language of side X in any dispute is the language that would belong to an imagined linguistic community typical members of which exhibit linguistic behavior that is relevantly similar to X’s. We can, if we wish, think of X as forming its own linguistic community. If side X is perdurantism then X’s language is the language that would belong to an imagined linguistic community typical members of which “talk like perdurantists,” i.e., they assert the sentences that perdurantists assert and endurantists reject. I claim that the dispute between endurantists and perdurantists is verbal. On the correct view of linguistic interpretation, endurantists will agree that perdurantists speak the truth in P-English, and perdurantists will agree that endurantists speak the truth in E-English. Central to what I take to be the correct view of linguistic interpretation is an appeal to “use,” but it must be understood that the only way to understand that appeal is in terms of what has been called the “principle of charity.”14 Imagine a linguistic community in which typical speakers assent to a certain sentence S only when an apple is present. Why wouldn’t it be an appeal to use to say that by S they mean that an elephant is present and

12. Burge 1979. 13. Burge’s point might straightforwardly accommodate verbal disputes in which the two sides use an ambiguous expression in diff erent but conventionally correct ways. I am assuming, however, that there can be verbal disputes in which a “verbal mistake” is made, i.e., (at least) one side is not using language correctly. 14. Quine 1960, p. 59; Davidson 1984; Lewis 1983a (“Radical Interpretation”); Lewis 1983b, at pp. 370–77. 230 quantifier variance and realism: essays in metaontology they systematically mistake apples for elephants? The answer is that an appeal to use is an attempt to make the most sense out of people’s use of language. Central to linguistic interpretation is the presumption that the correct interpretation is the one that makes people’s use of language as reasonable as possible. In interpreting a language there is therefore an overwhelming, if in principle defeasible, presumption that typical speakers make perceptual assertions that are reasonably accurate, and that they do not assert relatively simple sentences that are a priori false. The principle of charity to use does not depend on human generosity. It is, as I conceive of it, constitutive of the phenomena of language and meaning. Now, let’s imagine a community of people who “talk like perdurantists.” Shown an ordinary wooden stick they will assent to the sentence, “In front of us there are a succession of highly visible wooden objects that persist for a moment and then go out of existence.” If they are speaking E-English, in which this sentence is false, they are making a mistake about what visible objects exist before their eyes. Moreover this perceptual mistake is linked to their general a priori mistake of asserting the false E-English sentence, “Any persisting object a priori necessarily consists of a succession of temporal parts.” (Note the relevance here of my general assumption that ontological disputes concern matters of a priori necessity.) Barring some extraordinarily unlikely explanation for why they would be expected to make mistakes of this sort, charity to use requires us to interpret them as speaking P-English, in which these sentences are true. An analogous point evidently holds for those who “talk like endurantists.” Their language, as indicated by charity to use, is E-English. Of course, the people in our own community talk like endurantists. Plain English, therefore, is E-English. Insofar as both endurantists and perdurantists claim to be speaking plain English, the endurantists are right. Perdurantists are making a verbal mistake. They have somehow managed to philosophize their way out of the communal language. Lewis points out that a stage seems eventually to be reached in ontology when “all is said and done,” when “all the tricky arguments and distinctions and counterexamples have been discovered,” so that each position has achieved a state of “equilibrium.”15 I am thinking primarily of this stage when I say that the dispute between endurantists and perdurantists is verbal. Prior to this stage, if an endurantist, say, is disposed to change her mind in response to some perdurantist arguments, then charity to use may favor interpreting her language as P-English, so that

15. Lewis 1983a, p. x. ontology and alternative languages 231 the change of mind is deemed reasonable and her earlier judgment deemed mistaken. But after the “all is said and done” stage has been reached, there is nothing to be said but that each side speaks the truth in their own language. In saying this I am rejecting Lewis’s claim that when we have reached the “all is said and done” stage we are left with a “matter of opinion” in which one side “is making a mistake of fact.” My diff erence with Lewis about the status of the dispute at the “all is said and done” stage aff ects earlier stages. Suppose that prior to the “all is said and done” stage some perdurantists are trying to fi nd arguments in behalf of their position. If they listen to me their attitude will be, “If we can persevere and stick to our guns we’ll be right at least in our language.” But they also realize that what they had initially wanted was to be right in plain English. That’s the language they initially took themselves to be speaking. They realize, further, that no matter what fancy arguments they may come up with, there will be no general disposition in the wider linguistic community to convert to perdurantism. Even those few members of the general community who are disposed to take philosophy seriously are as likely to turn out to be nihilists, or idealists, or something else inimical to perdurantism. Since typical speakers talk like endurantists and have no disposition to change their minds in the direction of perdurantism, it is established that plain English is E-English. Why, then, should these perdurantists knock themselves out trying to come up with fancy arguments that might allow them to perservere in speaking their specialized language of P-English? Better simply to jump quickly to the “all is said and done” stage by reverting to endurantism and speaking plain English (= E-English). Lewis’s main argument for perdurantism is the “problem of temporary intrinsics.”16 I won’t review that argument here. Put in very rough terms the argument proceeds from the assumption, fi rst, that intuitive qualities such as shapes must not be treated as relations to times, and, second, that endurantism would require this treatment. I suspect that even if I were a real ontologist I would fi nd this argument to be weak, but the point I want to make at present is that the argument can be dismissed virtually sight unseen. For it is hardly likely to have the power to convert typical English speakers. Surely the typical speaker’s disposition to talk like an endurantist far outweighs any disposition to worry about temporary intrinsics. So plain English remains E-English. Insofar as the philosophers involved in the dispute claim to be speaking plain English, that’s all they need to know.

16. Lewis 1986, pp. 202–4. 232 quantifier variance and realism: essays in metaontology

Let me put this in a slighdy diff erent way. Lewis’s argument from temporary intrinsics appeals to the truth of some such philosophical sentence as, “Shapes and other intuitive qualities cannot be treated as relations to times.” Insofar as (the relatively few) speakers of English who are able to understand and evaluate this sentence are inclined to accept it (let’s assume that they are, though this is questionable), charity to use favors an interpretation that makes the sentence true. But if the truth of the philosophical sentence confl icts with endurantism, as Lewis claims, then this sentence is false in E-English. The overwhelming considerations of charity to use that indicate that the language of our community is E-English—the charity to use that appeals to the typical speakers’ confi dent assertion of innumerable endurantist sentences—must be weighed against charity to the philosophical sentence. This seems to be no contest. Since the preponderance of evidence clearly indicates that our language is E-English, the philosophical sentence must be deemed false (assuming that it actually confl icts with endurantism).17 This point generalizes to other philosophical arguments against endurantism. These arguments must always appeal to philosophical sentences that are false in E-English. Charity to use then dictates that the sentences are false in plain English. It turns out, then, that at any stage of this dispute the primary concern is with “what one ordinarily says.” The fancy arguments found in Lewis and others become irrelevant. (I mean they become irrelevant in the dispute; the confl icts and tensions they reveal in both ordinary and philosophical thought certainly need to be addressed.) I hope that I am not encouraging a verbal dispute about what counts as a verbal dispute. In saying that the issue between perdurantism and endurantism is “merely verbal” I am making essentially two claims: First, if there were a linguistic community typical members of which assert the sentences that, after all is said and done, perdurantists (endurantists) assert, then charity to use dictates that the language of this community is P-English (E-English). It follows, second, that speakers of either language ought to allow that speakers of the other language assert sentences that have the same characters and hence the same truth-values as the sentences that they themselves assert. It seems to me that these two claims imply the Carnapian idea that there is at bottom nothing to the issue of perdurantism versus endurantism but the choice of either P-English or E-English. I do not deny, certainly, that perdurantist philosophers and endurantist philosophers are related

17. Cf. my discussion of “confl icts of charity” in Hirsch 2002b, Hirsch 2005, and Hirsch 2007. ontology and alternative languages 233 to each other in some ways that are signifi cantly diff erent from the ways that the members of the two imagined linguistic communities are related to each other. For one thing (this is Burge’s point again) the philosophers grew up in the same linguistic community and may therefore be committed to the language of that community. Further- more, even after all is said and done, philosophers cannot shed their histories and return to the pre-philosophical innocence of typical members of the imagined communities; the linguistic behavior (and phenomenology) of philosophers is never quite like that of typical speakers. If someone holds that these points imply that there is a sense in which the philosophers are not engaged in a verbal dispute, I need not disagree. I only insist that there is a sense in which they are engaged in a verbal dispute, and it is this sense that supports the Carnapian conclusion. Charity to use may not be the only relevant interpretive principle, but I think it is by far the dominant one. In Lewis we fi nd the idea that there is a presumption in favor of interpreting words as expressing natural properties.18 Sider has made the novel suggestion that Lewis’s idea be extended to the logical apparatus of a language, so that there will be in eff ect an interpretive principle that has a special bearing on ontology.19 I’ve discussed Sider’s position at length elsewhere and will only briefl y recapitulate a few points here.20 He claims that a language may fi t better or worse to the world’s “logical joints,” and that there is a presumption in favor of an interpretation that yields a better fi t. It is this presumption that will determine whether a linguistic community is speaking E-English or P-English. Since ontological arguments reveal, as he thinks, that P-English is a better fi t to the logical joints than E-English, there is a presumption in favor of an interpretation yielding the former language, and this presumption determines that P-English is the language even in a linguistic community like ours, in which people seem to talk like endurantists. I have two basic problems with this position. First, the notion of “logical joints” seems to me obscure. Sider often says that the quantifi er-like expressions in a language answer to the world’s logical joints only if they express Existence

18. Lewis 1983b, pp. 370–7. 19. Sider (2001); Sider (2002), Introduction; Sider (2004), pp. 679–82. The novelty of Sider’s suggestion should not be understated. Lewis, in fact, distinguishes between natural and unnatural things, and assumes that our quantifi ers range over both kinds of things. There is nothing in Lewis to suggest Sider’s idea that the quantifi cational apparatus itself is constrained by considerations of naturalness. 20. Hirsch 2004; Hirsch 2005; Hirsch 2007. 234 quantifier variance and realism: essays in metaontology

(capitalized and italicized). This is evidently less an explanation than an invitation to accept “logical joints” and “Existence” as primitive notions. Can they be explained ostensively, by citing examples? Apparently not, since there is no agreement on what the examples are. I think one must feel some skepticism about the intelligibility of an allegedly primitive notion when there is no agreement as to what examples come under the notion.21 My second problem is that, even if we grant what Sider says about the presumption imposed by the logical joints, this presumption must be defeasible by considerations of charity to use. In regard to natural properties everyone will agree, I assume, that considerations of charity determine that most of the words in our language express properties that are not highly natural. By the same token, whether the logical joints favor E-English or P-English, it is the linguistic behavior of members of a community that will primarily determine which language they are speaking. I think it’s clear, therefore, that ontology cannot reveal that the members of our community really speak P-English and make innumerable a priori mistakes about what visible objects exist in front of their eyes.

IV. The Demand for a Semantics

The picture that I’m presenting of the relationship between the two languages E-English and P-English is that there is a certain set of sentences and a certain set of characters, and the characters are distributed diff erently over the sentences of the two languages in such a manner that the sentences asserted by endurantists in one language have the same characters as the diff erent sentences asserted by perdurantists in the other language. An objection that has often been raised in conversation is that this picture ignores a fundamental asymmetry in the relationship between the languages, an asymmetry that in fact makes it impossible for endurantists (speakers of E-English) to make good sense of P-English. By a semantics for a language I mean a fi nite description of the language which entails, for each sentence, an adequate specifi cation of its character. I’ll not try to pin down the relevant notion of adequacy; its intuitive meaning is probably suffi ciently clear for present purposes.

21. In Hirsch (2008, section 5) I suggest that the question which language best fi ts the logical joints is itself verbal. Here I limit myself to saying that this question seems unduly obscure. But the question whether there exist temporal parts is verbal, in the sense that any answer given by a linguistic community is correct in the language of that community. ontology and alternative languages 235

Now, if we start out speaking P-English then, assuming that we have a semantics for P-English, there seems to be no major obstacle in providing a semantics for E-English. We need only stipulate that the domain of things over which the quantifi ers of E-English range is limited in certain ways, so that temporal parts rejected by endurantists are excluded. The stipulated limitation is semantic, holding in any context, with the eff ect therefore that the most contextually unrestricted quantifi er available in E-English—that is, the widest sense of “everything” in E-English—does not range over such temporal parts. On the other hand, if we start out in E-English and are trying to provide a semantics for P-English, we obviously cannot say that the domain of objects in P-English contains more objects than the ones that are referred to by our word “everything.” How, then, can we make any sense in E-English of what it means in P-English to talk about temporal parts of things, things that (as we say) don’t exist? The problems presented by this case arise in various other issues of ontology (for instance, in the issue of unrestricted mereological composition). The general problem is how one side in a dispute can interpret another side that seems to posit “more objects.” If we look back at what Edna said in her diary I think we can see one perfectly reasonable way of formulating in E-English a “semantics” for P-English:

(1) In P-English a sentence of the form “a has at time t a temporal part that is F” has the same character as the E-English sentence “a is F at time t” (where F is a term that applies to an object at a time by virtue of how the object is at that time). Other forms of sentences in the language operate in the obvious ways.

We imagine that (1) is appended to a semantics in E-English for E-English, so that the characters of the E-English sentences can be assumed as given. It will of course be objected that (1) is a poor joke of a semantics, since it implies nothing about the character of any P-English sentence that does not have the specifi c form mentioned in (1). Doesn’t it? One thing that I think is clear is that any reasonably intelligent speaker of E-English who is guided by (1), perhaps supplemented with a few well chosen examples, will immediately be able to provide, for virtually any sentence of P-English, a sentence of E-English with the same character (as stipulated by (1)). Even if that fact doesn’t qualify (1) as a semantics, it should, I think, disarm the complaint that speakers of E-English would have no way of making charitable sense of P-English. One might view (1), supplemented with some examples, as 236 quantifier variance and realism: essays in metaontology a kind of “ostensive defi nition” enabling speakers of E-English to go on the same way in assigning characters to the sentences of P-English. It may be objected that (1) tells us little about the characters of intentional ascriptions in P-English. I think, in fact, that if we know the syntactic structures and characters of the sentences of a language other than intentional ascriptions, then we know a great deal (perhaps everything) about the fi ne-grained belief states expressed in the asserting of those sentences. I’m not going to try to defend that here (or related views about other intentional states). I’m going to put the objection aside for the present on the grounds that the general obscurity in the semantics of intentionality makes its relevance to the specifi c issue of understanding P-English doubtful.22 A somewhat more signifi cant worry, I think, is that (1) gives us no clue as to how to formulate in E-English the characters of P-English sentences involving “reference.” Take the sentence: “In P-English the word ‘bearded’ refers to (many things including) any bearded temporal part of Lincoln.” We must assume that this sentence is trivially true in P-English, but it cannot be true in E-English. Since, as we say in E-English, there are no such things as temporal parts of Lincoln, there cannot be any such things for a word of P-English to refer to. One possibility for explaining in E-English the characters of P-English sentences involving “reference” is along the following lines:

(2) P-English functions as if persisting objects are necessarily made up of temporal parts, and a P-English sentence of the form, “In P- English the expression e refers to things that are F” has the same character as the E-English sentence, “In P-English it’s as if the expression e refers to things that are F.”

This “as if” formulation expresses a holistic analogy between the way expressions function in P-English and the way they function in E-English, despite the former expressions not having the kind of referential function they have in E-English (in the E-English sense of “reference”). The “as if” formulation has something in common with a fi ctionalist approach. The following might be suggested as a substitute for both (1) and (2):

22. I’ll also ignore questions about what the truth conditions are in P-English for sentences involving the transworld identity of temporal parts. This can be stipulated in diff erent ways, depending on what perdurantists say. See, e.g., Sider 2002, pp. 218-24. ontology and alternative languages 237

(3) Any P-English sentence S has the same character as the E-English sentence “On the assumptions of perdurantism, S.”23 Assuming that the E-English semantics for E-English can deal with sentences of the form, “On the assumption of A, B,” the conjunction of that semantics and (3) may qualify as a respectable semantics in E-English for P-English.24 I think, however, that there is an important sense in which (3) depends upon (1). Since the E-English sentence “Persisting objects are necessarily made up of temporal parts” expresses an impossibility, and this sentence summarizes the assumptions of perdurantism, the form of E-English sentence cited in (3) is in eff ect a counterpossible. (“On the assumptions of perdurantism, S” is roughly equivalent to “If perdurantism were true, then S would be true.”) With- out entering into the serious diffi culties attending the analysis of counterpossibles, I think the following is reasonably clear: The truth of a “fi ctionalized” sentence of the form “On the assumption A, B” (where A is impossible) depends on the truth of some “straight” sentence C such that, from the conjunction of A and C, B follows, in some suitable sense of “follows.” (It obviously cannot be a sense in which everything follows from an impossibility.) Thus the truth of the E-English sentence “On the assumptions of perdurantism, Lincoln had in 186o a part that was bearded” depends on the truth of some straight E-English sentence, such as, “Lincoln was bearded in 1860.”25 The latter sentence

23. Cf. Dorr 2005. 24. I dismissed as obviously absurd an interpretation by Jews of Christians’ assertions along the lines of (3). One diff erence between the two cases was implied in note 11: Jews and Christians diff er over their non-linguistic behavior and attitudes, but there are no such diff erences between perdurantists and endurantists. A further point can now be made. My premise throughout this paper is that perdurantists and endurantists regard their respective positions as a priori necessary, and as having no bearing on their judgments about what experiences people have had and will have. Jews and Chris- tians do not regard their positions as a priori necessary, and do regard them as having a bearing on what experiences people have. (A related point is that Jews and Christians, but not perdurantists and endurantists, will give confl icting pictorial representations of certain aspects of reality.) All of these considerations enter into charity to use. Whereas charity to use encourages the interpretation given by (3), it discourages in the extreme a parallel interpretation by Jews of Christians’ assertions. 25. I am here supposing (vaguely and ignorantly) that counterpossibles are to be treated in some manner of “relevance logics” (as in Anderson and Belnap [1975]), but I would think that the main point in the text holds for any intelligible treatment. For example, Nolan (1997) suggests that a counterpossible “If A then B” is true iff B is true in the impossible worlds in which A is true that are closest to the actual world. Suppose that W1 is an impossible world in which the assumptions of perdurantism hold, that the sentence “Lincoln had in 1860 a part that was bearded” is true in W1, and that W1 is as close to the actual world as any world can be in which the assumptions hold and this sentence is true. And suppose that W2 is another impossible world in which the assumptions of perdurantism hold, that the sentence “Lincoln had in 1860 a part that was bearded” is false in W2, and 238 quantifier variance and realism: essays in metaontology is of course the sentence delivered by (1) as the E-English counterpart of the P-English sentence “Lincoln had in 1860 a part that was bearded.” The truth, therefore, of every fi ctionalized sentence of the form cited in (3) depends upon the truth of a straight sentence of the form delivered by (1) for the P-English sentence under consideration. In this sense it seems to be (1) that really gives the E-English speaker a sense of how P-English works. But this is not to deny that (3) may have the virtue, evidently lacking in (1), of providing a passably respectable semantics in E-English for P-English. Many philosophers, however, will consider (1)–(3) to be completely off the track. These formulations do not explain how the characters of the sentences of P-English are generated by the underlying referential relations between the words of that language and objects in the world. In fact, (2) makes it clear that the word “reference” in P-English could not mean what it does in E-English. Although the speakers of P-English will assert, “The characters of the sentences of P-English depend upon the reference of the words in P-English,” E-English speakers cannot make that assertion. They therefore can make no sense of P-English. The puzzle expressed in the last paragraph ought to lead, not to the vanquishing of P-English (by speakers of E-English), but to the rein- forcement of a basic insight into the nature of language. The insight is that “only within the context of a sentence does a word have meaning.” What must be given up is a picture of language in which the characters at the level of sentences are generated by some underlying referential mechanisms at the level of words. This “bottom-up” picture is misguided because the references of words depend upon the characters of sentences. One might perhaps have thought that there is somehow a two-way dependence, so that one can picture language both bottom- up and top-down. Apart from the inherent instability of this picture, it is shown to be untenable precisely by considering the relationship between such languages as E-English and P-English. As a speaker of E-English (= plain English) I must view the expression “temporal part” as functioning syncategorematically in P-English, in the way that logical words are generally viewed. The same will hold for many other words that are attached to “temporal part” (e.g., “bearded” in “bearded

that W2 is as close to the actual world as any world can be in which the assumptions hold and this sentence is false. If we say that W1 is closer to the actual world than W2 is (so as to make true “On the assumptions of perdurantism, Lincoln had during 1860 a part that was bearded”), this can only be because we understand that “Lincoln was bearded in 1860” is true in W1 but not in W2. ontology and alternative languages 239 temporal part of Lincoln”). There is nothing wrong with that. (I am not, as a philosopher once suggested to me, insulting the speakers of P-English by saying that their words do not refer to anything, in my E-English sense of “refers to something.”) The key to understanding the relationship between E-English and P-English is what I have called elsewhere “quantifi er variance”: this refers to the possibility that quantifi er-like expressions in diff erent languages may have diff erent semantic functions; they may contribute diff erently to the characters of sentences.26 Our E-English concept of “(the existence of ) something” is not the same as the corresponding concept expressed in P-English. All of the other diff erences between the languages depend on that one. On the picture I’m presenting the essence of language is the distribution of a set of characters over a set of syntactically structured sentences. There appear to be no general considerations about reference that impose a priori necessary constraints on how that distribution can work. It may indeed be that, for virtually any language that we can make intelligible to ourselves, it will seem most natural from within the language to explain the language in terms of what will be called a “referential semantics.” As a speaker of E-English I can well understand the correctness of the P-English speakers’ assertion of the sentence, “The characters of the sentences of P-English depend on the reference of the words of P-English.” I cannot, however, assert that sentence in terms of my use of the words “what exists” and “what can be referred to.” All of this ought to be seen as being quite in order. Our concept of “reference” varies with our concept of “what exists,” but it should be emphasized that our concept of “truth” does not thereby vary. Whether we speak E-English or P-English we mean the same by “The sentence S is true in the language L.” Our concept of saying the truth, of “saying it as it is,” is the stable Archimedian point around which everything else revolves. Matti Eklund has in several places illuminatingly discussed under the heading “neo-Fregeanism” the idea that the truth values of sentences are explanatorily prior to the reference of terms.27 I am evidently a neo-Fregean in this respect (though perhaps not in other respects that fi gure in Eklund’s discussion). Eklund, however, wishes

26. Hirsch 2002; Hirsch 2005; Hirsch 2007. 27. Eklund 2006a; Eklund 2006b; Eklund 2009. I focus here on only one aspect of Eklund’s complex discussion of neo-Fregeanism, and on only one of his objections to positions like mine. 240 quantifier variance and realism: essays in metaontology to combine the neo-Fregean idea with the following “Tarskian principle”: (T) For a sentence of the [syntactic] form “F(a),” of any [possible] language, to be true, the [syntactically] singular term “a” must refer.28 The conjunction of neo-Fregeanism and (T) immediately entails that endurantism is false. If we are neo-Fregeans we must allow that P-English is a possible language, assuming, as I now am, that the truth conditions (characters) of the P-English sentences have been made intelligible. We cannot, as perhaps we could if we were both anti-neo-Fregeans and anti- perdurantists, reject the possibility of P-English on the grounds that it lacks the required kind of referential and ontological structure. (Let us note in passing that, even more obviously, if we are neo-Fregeans we must allow that E-English is a possible language.) Given that P-English is a possible language, we observe that it contains true syntactically singular sentences (e.g., “The daytime part of that tree was never wet”) which, according to (T), requires the singular terms to refer to temporal parts. Hence endurantism is refuted. The conjunction of neo-Fregeanism and (T) leads Eklund in the direction of a maximally expansive ontology that countenances, not just temporal parts (and their mereological sums), but any object that we can conceive of as required as a referent of a true syntactically singular sentence of any imaginable language.29 Let me say that the sentence (T) as written above is a “strong” T-sentence, and that a “weak” T-sentence is the result of replacing “of any [possible] language” by “of this language.” (“For a sentence of the [syntactic] form ‘F(a),’ of this language, to be true, the[syntactically] singular term ‘a’ must refer.”) And let me say that a “T-language” is a language within which at least the weak T-sentence is true, whereas a “strong T-language” is one in which the strong T-sentence is true. Given that P-English is a possible language, it follows that E-English, even if it is a T-language, is not a strong T-language. Eklund’s argument (in plain English, as I take it) for an expansive ontology assumes that plain English, the language we ordinarily speak, is a strong T-language, and is therefore not E-English. What reason does he have for that assumption?

28. (T) is from Eklund (Chalmers, Manley, and Wasserman 2009) with my added bracketed expressions, which are clearly intended by Eklund. It should be noted that Eklund acknowledges that there may be counterexamples to (T) that have no bearing on the ontological issues. 29. An argument similar to Eklund’s is found in Hawthorne 2006. Eklund in fact raises various questions about whether the notion of a maximally expansive ontology can be made coherent, but I am concerned here with the more modest application of his argument from (T) to the denial of endurantism. ontology and alternative languages 241

If the reason is simply that the strong T-sentence strikes Eklund as intuitively right, I have already answered this in my comments in the previous section about Lewis’s argument from temporary intrinsics. The strong T-sentence is a philosophical sentence that is false in E-English, and therefore false in the language we ordinarily speak, on the most plausibly charitable interpretation. One might try to bolster the intuition in behalf of the strong T-sentence by arguing that, if our language were not a strong T-language, it would be drastically (and implausibly) defi cient in its capacity to explain the semantics of various possible languages. In E-English, for example, we could not explain the truth conditions (characters) of the sentences of P-English by appealing to the reference of singular and general terms. But this argument (apart from other problems with it) is inconsistent with Eklund’s neo-Fregeanism. If we are neo-Fregeans we do not explain truth conditions by appealing to reference. Therefore, if we are neo-Fregean speakers of E-English, we do not fi nd ourselves faced with the explanatory defi ciency just mentioned.30 Nevertheless, I would not deny that it may be helpful to shift to a strong T-language (or something approximating to it) for certain philosophical and semantic purposes. That is not the issue. If Eklund is merely claiming that strong T-languages are in some ways more useful than other languages, this is squarely a Carnapian remark, and I need not disagree with him. Two points that I am making, however, are: First, endurantism is true (and Eklund’s expansive ontology is false) in the sense that E-English is plain English; the language we ordinarily speak is not a strong T-language. Second, the dispute between endurantism and a more expansive ontology is merely verbal, turning on whether or not we choose to use a language in which the strong T-sentence is true. Let me add a few qualifying remarks. Certainly, our understanding (as speakers of E-English) of P-English is “compositional,” in the fundamental sense that the words and their ordering in a P-English sentence is what allow us to determine the sentence’s character. The manner of compositionality of the P-English sentences is what is described in (1)–(3). Furthermore, it may be permissible in some cases to attribute a referential function to expressions in P-English. For

30. If we are neo-Fregeans we may try to explain the truth conditions of complex sentences on the basis of the truth conditions of simple sentences, including one-word sentences. We may also exploit various devices for providing a fi nitary description of the truth conditions of the sentences of a language. One such device might be a Tarskian semantics (supplemented by various equivalence transformations). Another kind of device might be (3), above. But we do not as neo-Fregeans view facts about the truth conditions of sentences to be explanatorily based on facts about the reference of terms. 242 quantifier variance and realism: essays in metaontology instance, it seems to do no harm to take “apple” to refer to apples in the P-English sentence, “This apple is now red.” (However, thinking of “apple” as referring to apples will not help us to give a straightforward referential analysis of a P-English sentence like, “The current temporal part of this apple is more red than a certain temporal part of that tomato.”) Certainly, we are not required to view the words and sentences of P-English as fl oating free of the objects that exist in the world (in our E-English sense of “objects that exist in the world”). On the contrary, the truth-values of sentences in P-English depend on the objects that exist and their properties and interrelations. For instance, the truth-value of the P-English sentence, “The current temporal part of this apple is more red than a certain temporal part of that tomato,” depends on the existence of a certain apple and a certain tomato, and on their properties and interrelations. We can even say, in a rough intuitive sense, that the sentence is “about” nothing but a certain apple and a certain tomato. The words of P-English are semantically related to objects (in our E-English sense of “objects”) in various ways that are indirectly and holistically described in (1)–(3). Finally, when I say that the essence of language is nothing more than the distribution of sentential characters over syntactic structures, I do not deny that some such distributions will intuitively strike us as crazy or, put somewhat more temperately, as too strange to be taken seriously as possible natural languages. What to make of these intuitions I don’t know, but their force should not be denied.31 Examples of languages that seem “too strange” might include Quine’s language without variables;32 his language with a single dyatic predicate;33 perhaps a language with (too many) words like “grue” and “quus.” In none of these examples is there any kind of special semantic problem. It is obvious that many of the perdurantist’s assertions initially strike typical members of our linguistic community as bordering on madness. (This point is even more obvious when unrestricted composition is added to perdurantism, as it almost always is.) I think it is indeed prob- able that, as an empirical fact, humans are innately disposed to favor languages like E-English over languages like P-English.34 Even if this is so, however, I expect that it is quite easy for most people to adjust to P-English. The obvious analogies between spatial and temporal inter- vals, and between ordinary objects and processes, can make P-English

31. See throughout Hirsch 1993. 32. Quine 1966. 33. Quine 1954. 34. Spelke 1984; Xu 1997; Hirsch 1982; Hirsch 1997. ontology and alternative languages 243 seem quite benign after a bit of practice. Therefore, while some common sense philosophers might consider P-English to be “too strange,” I don’t myself see it that way. In any case these “transcendental” intuitions about which languages may be “too strange to be taken seriously” are probably themselves too fragile to be taken very seriously. These intuitions, for whatever they are worth, must be attempts to measure the distance (along some dimensions) between some putative languages and our paradigms of intelligibility, where some distances may strike us as too great. The paradigms themselves are of course ordinary languages, such as plain (E-) English. These languages require one to talk like an endurantist, and they are no more nor less strange than ordinary life.

5. Platonism versus Nominalism

As I stated at the outset, my position is roughly Carnapian, for it was Carnap who said that ontological disputes amount to nothing more that a choice of language. I now want to bring out the point that Car- nap evidently intended his doctrine to apply to various disputes that I may not count as verbal. An example is the dispute between platonists and nominalists about the existence of abstract items, such as sets, properties, and numbers. In considering this dispute it is important not to focus exclusively on sentences of pure set (property, number) theory, which are either necessarily true or necessarily false. Consider the sentence:

(4) There are two [nondenumerably] infi nite sets X and Y, whose members are [nondenumerably] infi nite sets of angels, satisfying the condition that, for any set X’ in X, there is a set Y’ in Y such that all angels in X’ love all and only angels in Y ’, and some angel in Y’ loves some angel in some set in X other than X’.

Platonists consider (4) to be true in some worlds and false in others, whereas nominalists (of the sort that concern me here) regard (4) as unintelligible or necessarily false. Let me assume that nominalists can accept some relevant version of a semantic treatment in terms of characters.35 In order for the dispute about (4) to be verbal in my sense it must be possible to produce a sentence that nominalists can plausibly

35. For example, contexts and worlds might be treated as concrete things, and “A has the same character as B” might be defi ned to mean “For any context x and world y, A, relative to x, is true with respect to y iff B, relative to x, is true with respect to y.” 244 quantifier variance and realism: essays in metaontology regard as having the same character that (4) has in the platonists’ language and that makes the platonists’ attitude towards (4) come out right. An attempt to produce such a sentence (temporarily ignoring the nondenumerable case) might yield something like this:

(5) There are angels x-11, x-12, x-13,..., and angels x-21, x-22, x-23,..., and angels x-31, x-32, x-33, and angels..., and angels y-11, y-12, y-13,..., and angels y-21, y-22, y-23,..., and angels y-31, y-32, y-33,..., and angels..., such that either it is the case that {x-11, x-12, x-13,... all love only y-11, y-12, y-13,..., and either (y-11 loves either x-21, or x-22, or x-23,..., or x-31, or x-32, or x-33, or..., or...) or (y-12 loves either x-21, or x-22, or x-23,..., or x-31, or x-32, or x-33, or..., or...) or (y-13 loves either x-21, or x-22, or x-23,..., or x-31, or x-32, or x-33, or..., or...)} or it is the case that {x-21, x-22, x-23,... all love only y-11, y-12, y-13,..., and either (...

I think it is safe to assume that typical nominalists will not consider there to be any possible sentence that fi lls in the infi nite number of infi nitely long gaps in (5). If we bring in the nondenumerable case the prospect for fi nding the required sentence is even more obviously hopeless. There appears therefore to be no straight (nonfi ctionalized) sentence that would allow the nominalists to charitably interpret what platonists mean by (4). Nor, I think, does it do any good to bring in the fi ctionalized sentence:

(6) On the assumptions of platonism, (4).

Since I am assuming that nominalists regard the assumptions of platonism (e.g., various axioms of set theory) to be unintelligible or necessarily false, (6) would have to be understood as a counterpossible. The truth of (6) in some possible world would require, along the lines indicated earlier, that there is a straight sentence S that is true in that world such that, from the conjunction of S and the assumptions of platonism, (6) follows (in the relevant sense of “follows”). But nominalists cannot acknowledge there being any such sentence S. If they regard (4) as unintelligible or necessarily false they will have the same attitude towards (6).36

36. I have been informed, however, that many nominalists who reject (4) and do not acknowledge any such sentence as (5) will nevertheless accept (6) (as possibly true). If this is so, these nominalists, I would say, are only engaged in a verbal dispute with the platonists. But their position seems to me hard to understand. Will they perhaps try to treat (6) in terms of closest impossible worlds (as in Nolan [1997]; see note 25, above)? Suppose that W1 is an impossible world in which the assumptions of ontology and alternative languages 245

It seems to follow that the dispute between platonists and nominalists is not verbal in my sense. From their perspective nominalists have no way of charitably interpreting platonists as speaking the truth in their own language. What, then, can Carnap mean by saying that, for nominalists to convert to platonism, nothing more is needed than the choice to speak a diff erent language? Platonists claim that (4) has a character that makes it true in some worlds. The character of (4) is treated by platonists as akin to a “limit” approached by the succession of characters of increasingly fi lled in versions of (5). Nominalists deny that any sentence can have such a character. It is this denial that Carnap must be describing as “choosing not to speak a certain language.” Does that description merely beg the questions against nominalism? My aim here is not to rebut Carnap’s claim that the issue between platonists and nominalists is in some sense verbal; in fact, I would be inclined to look for some kind of defense of this position. What I am trying to bring out, however, is the diff erence between the relatively simple conception of a verbal dispute that I have employed throughout this paper and the more diffi cult conception that would be required to explain Carnap’s position. Superfi cially it may seem that platonists stand to nominalists the way that perdurantists stand to endurantists, the former position in each pair believing in things that are denied by the latter. A critical diff erence, however, is that perdurantists can off er endurantists something that platonists apparently cannot off er nominalists: a plausible interpretation wherein the opposing camps are seen as merely asserting diff erent sentences with the same characters.

platonism hold, that (4) is true in W1, and that W1 is as close to the actual world as any world can be in which the assumptions hold and this sentence is true. And suppose that W2 is another impossible world in which the assumptions of platonism hold, that (4) is not true in W2, and that W2 is as close to the actual world as any world can be in which the assumptions hold and this sentence is not true. If nominalists hold that that there is no such sentence as (5) that can be said to be true in both the actual world and in W1, what sense can they make of saying that W1 is closer to the actual world than W2 is, which is what would be required to make (6) true? Since W1 is the only one of the three worlds in which (4) is true, it seems that W2 should turn out to be the world closer to the actual world. (Note that this problem applies even if the nominalists somehow regard the non-existence of sets as contingent, so that (6) can be construed as simply a counterfactual rather than a counterpossible). Might the answer be that closeness of worlds does not depend on the expressible truths in these worlds? But that would not explain why the truth of (4) in W1 makes W1 closer to the actual world in which (4) is not true. The explanation, it seems, would have to be that there is a certain inexpressible truth about the actual world which, in conjunction with the (impossible) assumptions of platonism, entails the truth (but not the falsehood) of (4). That seems to me hard to understand. 246 quantifier variance and realism: essays in metaontology

Might a nominalistically acceptable interpretation of the platonist’s (4) be statable in terms of such devices as substitutional or plural or meta-linguistic quantifi cation? That remains to be shown; I myself don’t see how to accomplish that.37 Might it be suggested more simply that (5) as it stands, dots and all (or perhaps with the dots replaced by “etc.”s), is a passable English sentence of a sort, which ought to be acceptable to nominalists, and which explains the character of the platonist’s (4)?38 This may be a move in the direction of a Carnapian treatment of the issue (at least for the denumerable case). More would evidently need to be said about the status of such formulations as (5). I will, however, simply conclude this discussion by emphasizing the importance of not assuming that the Carnapian approach is either correct for all ontological issues or correct for none. One has to examine each case and see whether a plausible interpetation is available that will make both sides come out right.39,40

Appe ndix: Duran’s Dilemma (a parable)

Edna was a young philosophy professor who believed in endurantism. “It’s absurd to suppose that objects like trees and pebbles have temporal

37. A diff erent kind of maneuver suggested as a possibility by Steve Yablo in correspondence is to render (4) nominalistically in terms of something like, “Our world is angelically indiscernible from a world in which (4) is true,” or “The angels don’t falsify (4); it may be false for mathematical-ontology reasons but not for angelic reasons.” Let’s look at a somewhat simpler example. Suppose you make the statement, “There are ten positively charged electrons each of which loves a diff erent angel.” I respond, “Although your statement is false (there are no positively charged electrons, and electrons don’t love things), our world is angelically indiscernible from a world in which your statement is true. The angels don’t falsify your statement; it may be false for physics reasons but not for angelic reasons.” Absent some specifi c explanation (e.g., “I agree that there exist ten things of a certain kind each of which loves a diff erent angel,” or “I agree that there exist ten particles of a certain kind each of which does a certain kind of action to a diff erent angel,” or “I agree that there exist ten angels,” etc.), my response is incomprehensible. There seems to be no nominalistically acceptable way of explaining Yablo’s formulation. The basic question here is whether nominalists ought to be able to make sense of “what (4) says about angels (putting aside the stuff about sets).” I’m at present skeptical about this. Similar diffi culties pertain to the formulation in Rosen (2001), pp. 75-6. Of course, if anything goes we can simply introduce an irreducible “nominalist operator” N that transforms the nominalistically objectionable (4) into the acceptable N((4)). 38. I am indebted here to Mark Moyers. 39. Two other examples in which I think it is doubtful that the Carnapian approach applies are the issue of the existence of souls and the issue of the existence of physical simples. See Hirsch 2005, section IV. 40. For help with this paper my thanks to Cian Dorr, Matti Eklund, John Hawthorne, Ted Sider, to the participants in the 2007 Metametaphysics conference at Boise State University, Idaho, and especially to Steve Yablo. ontology and alternative languages 247 parts,” she thought. Another thing she thought was that her commitment to endurantism was blocking her from moving to one of the better philosophy departments, in which, she felt, the dominant perdurantists were loathe to hire an endurantist. Eventually this situation seemed intolerable to her. Although she was unwilling simply to lie about being an endurantist, she hit on the idea of secretly adopting a language of her own that would allow her to sound like a perdurantist while remaining an endurantist. In her diary she wrote: “Henceforth I will use the expression ‘temporal part of an object’ when I want to talk about how an object is at a certain time. I’ll say, ‘Lincoln had in 1860 a temporal part that was bearded’ to describe the situation in which Lincoln was bearded in 1860. In general, I’ll use a sentence of the form ‘a has at time t a temporal part that is F’ to be true of any situation in which a is F at t (where ‘F’ is a term—unlike ‘adolescent’—that applies to an object at a time by virtue of how it is at that time).” In this manner, Edna began to pass herself off as a perdurantist. It’s of course questionable whether this trick really enabled her to avoid “lying,” but we’ll not dwell on that. At the time that Edna adopted her secret language she was preg- nant and had been married for several years to a philosopher named Pedro. Pedro now found himself in a very odd situation, for, it turns out, he was secretly a perdurantist who had been pretending to be an endurantist out of his love for Edna. He had fi rst become aware of her when he was standing in the back of a crowded elevator at a philosophy conference and heard her repeatedly referring to “those regrettable perdurantists.” That night he wrote in his diary: “Hence- forth I will in every context restrict my quantifi er to objects accepted by endurantists—roughly, objects other than (proper) temporal parts of ordinary bodies.” Having, so to speak, banished temporal parts from his world, he was able to draw Edna into it. But now he found himself in the grotesque situation of being a secret perdurantist married to a secret endurantist. He decided to tell Edna that he was going to join her in her fake perdurantism. She accepted this decision without comment, but soon thereafter their marriage unraveled. Who can say why? Did Edna perhaps detect an element of deceit in Pedro’s off er? He found a job in a department in the Midwest. Edna gave birth to a boy who was named Duran. Three years later Pedro was found dead in front of his T.V. set from causes that could not be determined. An autopsy revealed that recently he had tattooed on his chest large portions of Kripke’s sermon from “Substitutional Quantifi cation.” At his funeral Edna 248 quantifier variance and realism: essays in metaontology was inconsolable. “He didn’t even do technical philosophy,” she kept saying. As the years passed Edna did not move from Harvard, but her fake perdurantism had by now become second nature and she never gave it up. She was very close to her son Duran, who for some reason stayed completely clear of philosophy, with no objections from his mother. It’s perhaps not surprising that he grew up talking like his mother. He talked like a perdurantist. Often this made no diff erence, since his mother, like many real perdurantists, didn’t mind saying such things as, “Lincoln was once bearded.” Sometimes, however, he would say such things as, “Lincoln had a bearded temporal part.” His friends typically treated this as merely an interesting variation of hiphop language, but a few of his teachers recorded that he seemed to suff er from a hitherto undiscovered form of madness. When he was twenty he announced that he was moving to the Midwest with his fi ancée to join her father who owned a tattoo parlor. While trying to prepare herself emotionally for his departure a bewildering and chilling question entered Edna’s mind, and once it was lodged there, she could scarcely believe that she hadn’t considered it earlier. Was Duran a real perdurantist? It seemed to her that she had been tacitly assuming all along that he was like her, at bottom an endurantist. But could it be that she had raised her child to belong to the other camp and to be asserting falsehoods? Her conversation about this with Duran did not go smoothly. “Duran, honey,” she began, “when you say, for instance, that a certain temporal part of Lincoln existed only in 1860, do you really mean this?” “Why, Mom? Don’t you really mean it?” he replied. Edna immediately saw that her question was misguided. Of course she really meant it in her secret language. “The question I’m trying to get at, honey, is this. When I say it I just mean that a certain part of Lincoln’s life stretched throughout 1860. Is that all you mean?” “I don’t get it, Mom,” he replied, “don’t those two things amount to the same thing?” At this point Edna felt that she had to give Duran at least an inkling of the philosophical division between perdurantists and endurantists. After completing her explanation she said, “What I’m trying to determine, honey, is whether you’re a perdurantist or an endurantist.” “How can I tell, Mom?” Duran replied. Duran and his fi ancée had made the obscure decision to travel by Greyhound bus, so Edna found herself early one morning at a crowded bus depot waiving goodbye to them. As the bus pulled out Duran shouted something out the window that Edna couldn’t hear. Someone behind her said, “He was yelling, ‘How can I tell, Mom?’” ontology and alternative languages 249

When she turned around she could not locate the person who said this to her.

Referenc es Anderson, A. R. and N. D. Belnap (1975). Entailment: The Logic of Relevance and Necessity, Vol. I (Princeton, NJ: Princeton University Press). Burge, T. (1979). “Individualism and the Mental,” in P. A. French, T. E. Uehling, and H. K. Wettstein (eds), Midwest Studies in Philosophy IV (Minneapolis, MN: University of Minnesota Press), pp. 73–121. Carnap, R. (1956). “Empiricism, Semantics, and Ontology,” in Meaning and Necessity, 2nd edn (Chicago, IL: University of Chicago Press), pp. 205–28. Chalmers, D., D. Manley, and R. Wasserman, eds. (2009). Metametaphysics (New York: Oxford University Press). Davidson, D. (1984). Inquiries into Truth and Interpretation (London: Oxford University Press). Dorr, C. (2005). “What we disagree about when we disagree about ontology,” in M. Kalderon (ed.), Fictionalist Approaches to Metaphysics (Oxford: Oxford University Press), pp. 234–86. Eklund, M. (2006a). “Metaontology,” Philosophy Compass 1:3, pp. 317–34. ———. (2006b). “Neo-Fregean Ontology,” Philosophical Perspectives 20, pp. 95–121. ———. (2009). “Carnap and Ontological Pluralism,” in Chalmers, D., D. Manley, and R. Wasserman (eds.), Metametaphysics (New York: Oxford University Press). Hawthorne, J. (2006). “Plenitude, Convention, and Ontology,” Metaphysical Essays (Oxford: Oxford University Press). ––——. (2009). “Superfi cialism in Ontology.” in Chalmers, D., D. Manley, and R. Wasserman (eds.), Metametaphysics (New York: Oxford University Press). Hirsch, E. (1982). The Concept of Identity (New York: Oxford University Press). ———. (1993). Dividing Reality (New York: Oxford University Press). ———. (1997). “Basic Objects: A Reply to Xu,” Mind and Language 12, pp. 406–12. ———. (2002a). “Quantifi er Variance and Realism,” Philosophical Issues 12, pp. 51–73. ———. (2002b). “Against Revisionary Ontology,” Philosophical Topics 30, pp. 103–27. ———. (2004). “Comments on Theodore Sider’s Four Dimensionalism,” Philosophy and Phenomenological Research 68:3, pp. 658–64. ———. (2005). “Physical-Object Ontology, Verbal Disputes, and Common Sense,” Philosophy and Penomenological Research 70, pp. 67–97. ———. (2007). “Ontological Arguments: Interpretive Charity and Quantifi er Variance,” in T. Sider, J. Hawthorne, and D. Zimmerman (eds), Contem- porary Debates in Metaphysics (Maiden, MA: Basil Blackwell), pp. 367–81. ———. (2008). “Language, Ontology, and Structure,” Noûs, 42, pp. 509-28. 250 quantifier variance and realism: essays in metaontology

Kaplan, D. (1989). “Demonstratives,” in J. Almog, J. Perry, and H. K. Wettstein (eds), Themes from Kaplan (New York: Oxford University Press), pp. 481–563. Lewis, D., (1983a). Philosophical Papers I (New York: Oxford University Press). ———. (1983b). “New Work for a Theory of Universals,” Australasian Journal of Philosophy 61, pp. 343–77. ———. (1986). On the Plurality of Worlds (New York: Basil Blackwell). Nolan, D. (1997). “Impossible Worlds: A Modest Approach,” Notre Dame Journal of Formal Logic 38, pp. 535–71. Putnam, H. (1981). Reason, Truth, and History (Cambridge: Cambridge University Press). ———. (1987). “Truth and Convention: On Davidson’s Refutation of Conceptual Relativism,” Dialectica 41, pp. 69–77. ———. (1994). “The Question of Realism,” in Words and Life (Cambrdige, MA: Harvard University Press), pp. 295–312. Quine, W. V. (1954). “Reduction to a Dyadic Predicate,” Journal of Symbolic Logic 19, pp. 180–2. ———. (1960). Word and Object (Cambridge: MA: MIT Press). ———. (1966). “Variables Explained Away,” in Selected Logic Papers (New York: Random House), pp. 227–35. Rosen, G. (2001). “Nominalism, Naturalism, Epistemic Relativism,” Philosophical Perspectives 15, pp. 69–91. Sider, T. (2001). “Criteria of Personal Identity and the Limits of Conceptual Analysis,” in Philosophical Perspectives 15, pp. 189–209. ———. (2002). Four-Dimensionalism: An Ontology of Persistence and Time (New York: Oxford University Press). ———. (2004). “Replies to Gallois, Hirsch, and Markosian,” Philosophy and Phenomenological Research 68:3. pp. 674–87. Spelke, E. S. (1984). “Perception of Unity, Persistence, and Identity: Thoughts on Infant’s Conceptions of Objects,” in J. Mehler and R. Fox (eds), Neonate Cognition: Beyond the Buzzing Blooming Confusion (Hillside, NJ: Lawrence Erlbaum Associates). Xu, F. (1997). “From Lot’s Wife to a Pillar of Salt: Evidence that Physical Object is a Sortal Concept,” Mind and Language 12, pp. 365–92. Yablo, S. (2001). “Go Figure: A Path Through Fictionalism,” in Midwest Studies in Philosophy 25:1, pp. 72–102. index

absolute truth, 44, 137–38 on nominalism vs. platonism, absolutism, 133, 134, 137 245–46 admissible precisifi cation, 50, 54n.16 ontological issues, 220–21, 243 adult ontology, 33–34 as quantifi er variantist, xvi ambiguity, 45–51, 54n.16, 62 Carnapian tolerance, 81, 82, 83n.21, anti-realism, xiv, xvi, 187, 189 85, 89 a priori argument, 115n.26, 222n.3 Cartwright, Richard, 116–17 arbitrariness, 116–17, 135 change(s) argument from charity, 99, 101–4, intrinsic, 222n.3 108, 202–3, 206 in meaning of quantifi er “as if” formulation, 73–74, 142–43, expression, xvi, 70, 71–72 157–58, 236 mereological, 179, 180, 183 assertions, 106, 112–15, 118, 120, 203, minimizing, 28, 29, 34 206 of single object, 24 Austin, J.L., 85, 89, 90, 97, 101, 123 charity, 119, 120, 122, 175, 176, 193, 205–6 basic proposition, 217n.31 See also argument from charity; behavior, 115, 120, 148 confl icts of charity; being, 60n.23 constraints beyond charity; bodily unity, 6–9, 12, 14, 17–18, 20, interpretive charity; principle 25–26 of charity body-mindedness, 6, 8–10, 17–18, charity to expressibility, 197, 200, 23, 25 201, 202 Bower, T.G.R., 22 charity to perception, 149, 154, Burge, Tyler, 146, 147, 229, 233 156, 159, 168, 175–76, 185, Butler, Joseph, 32, 104–5, 178–79, 204 183 charity to retraction, 151, 152, 159, Butlerian English, 179–91 174, 180–82 charity to understanding, 149, 154, Carey, S., 29–31, 33, 34 156, 159, 168, 175–76, 182–83, Carnap, Rudolph, 164, 232, 241 185 internal/external questions, 144, charity to use. See interpretive 161, 192, 220 charity

251 252 index children conciliatory outcome, 197, 201–2, concept/experience of unity, 22, 203, 207, 210 26 confl ict of charity, 108–9, 111, 112, development of, 21–22 185, 198 learning of language, 23 connectedness, 4, 6 quality space, 23 conscious inference, 204, 205 See also infants constraints beyond charity, 108, 116, Chisholm, Roderick, 91, 97, 99, 120, 121n.39 104–5, 153, 155 context, 105, 226, 227, 238 Chomsky, Noam, 6, 8, 20n.14 contextually restricted quantifi er, 154 Christianity, 221, 225–27, 237n.24 contingent statement, 74–78 classifi cation, 7–8, 10, 19, 195 conventionalism, 10–15, 20–21 cohesiveness, 4, 20, 22 correspondence theory of truth, 76, common motion, 22, 28 78, 140–41 common sense, 82–83 counterpossibles, 237, 244, 245n.36 and body-mindedness, 9, 10, 17 crypto-revisionism, 104, 106, 108 and criteria of unity, 4–6, 17, 21 defense of, 221 Davidson, Donald, 41, 42, 76–77, 115, and explosionism, 133, 136, 137 120, 148 and four-dimensionalism, 166, deep ontology, 89–91, 93, 95n.40, 125 167, 192 defeasible presumption, 99, 176, 177, Moore on, 98 230, 234 ontology, 43, 85, 87, 91–92, 94, 96, degrees of articulation, 28 103, 106, 117, 138, 143, 146, 191 dependence, 213n.25 and physical reality, 7, 177 determinateness, 50, 54n.16 presumption in favor of, 99 disagreement of attitude, 115 regarding objects, 93–94 disquotational constraint, 102 restrictivism, 50, 53, 58, 60, 61n.26, disquotational principle, 74 63, 64n.33 distinguished quantifi er expression, sanity, 37, 42 217n.32 unity relation, 117–18 distinguished structure, 213 community of persistent doctrine of mereological sums. See philosophers, 204 mereological sums complexity, 117, 118, 119–20, 150 doctrine of temporal parts, See composites, 3–5, 145, 146, 165, 184, temporal parts 197n.1, 199 doctrine of unrestricted mereological compositional semantics, xv n.8, 150, composition, 184, 190 158–59, 171, 241 dualism, 164n.29 composition relation, 117 Duran’s dilemma, 246–49 conceptual relativism, 38–39, 42, dynamic cohesiveness, 4, 6, 16, 22 43–44, 68, 134, 138–39, 173n.50, 187 Eklund, Matti, xvn.8, 239–40, 241 conceptual schemes, 13, 29, 39, 73n.6, eliminativism, 137 76, 77, 139 empirical falsehoods, 99, 114 index 253 empiricism, 15–20, 26, 222n.3 existence simpliciter, 86 endurantism, 221–24, 229–32, 233, existential relativism, 132–43 234, 240–41, 243, 245–48 experience, 11–12, 185 English language explosionism, 132–38, 143 disputed sentence in, 83 expressively defi cient language, as evolving, 85 198–99, 200, 202, 208, 210 and imagined language, 36–37, external questions, 144, 161, 192, 220 80–81, 92–94, 100, 111, 116–18, extra-logical joints, 126, 127 147–48, 153–59, 179 interpretations of, 102, 109, 113, factive correspondence theory of 115 truth, 141 no-two-things-in-same-place fi ctionalist strategy, 226–27, 236 principle, 109, 113 four-dimensionalism precisifi cation, 55, 59, 63 and common sense, 166, 192 truth conditions, 110 defi nition of, 96n.1, 145 unrestricted quantifi er, 86, 107, and disputed sentences, 153–55, 136–37, 139, 140 158 See also quantifi ers doctrines of unrestricted epistemology, 98, 122, 163n.28 temporal parts and unrestricted equivalence, xi mereological composition, 190 Evans-Salmon position, 45–46, 47, 52, and interpretive charity, 171 60, 62, 63n.32, 66, 166, 167n.41 of Lewis, 106, 153–55 examples, 109, 112, 114, 121 and mereological essentialism, existence 153–55, 158, 162, 192 concepts of, 71, 72, 84, 130, 131, and revisionary ontology, 106, 107, 139, 187–88 119 everything/nothing exists camps, Sider’s position on, 124–31, 96–97 176n.56, 212, 213n.25 vs. existence, 194 and truth conditions to sentences, as natural property, 127 172 of objects, 106, 108–9, 123, 130, Four Dimensionalism (Sider), 124 140–43, 144–45, 182, 184, 185, “From Lot’s Wife” (Xu), 27 191n.15, 192, 203, 204 and quantifi cational structure, games, 112 214–15 general principles, 108, 109, 111, 113, and quantifi er expressions, 70–71, 184, 205, 207 74 general sentences, 204 and similarity, 217n.32 genuinely plausible truth-condition of temporal parts, 171 assignments, 150 unrestricted, 139 genuine quantifi ers, 121 and vagueness, 60, 67, 90 Gestalt Psychology (Köhler), 11 verbal disputes on, 200, 214 Gettier counterexamples, 109 Existence/existence, 131, 194, 215, Gettier examples, 163 233–34 Goldbach sentence, 151n.13 254 index good reasons, 99, 120 Sider’s version of, 170 groups, 11–12 of strict and loose talk, 179 of theory, 174 Hawthorne, John, 222n.3 and truth conditions of sentences, Heller, Mark, 113 156, 158–59, 162, 172 histories, 222 and verbal disputes, 161–63, Hume, David., 32 228–34 intrinsic change, 222n.3 idealist madness, 73–74 intrinsic properties, 94 identity Butler-Locke debate on, 178–79 James, William, 18, 19, 205 criteria, 8, 13 Judaism, 221, 225–27, 237n.24 Lewis on, 109–10 of objects, 101, 106, 108, 109, 123, Kant, Immanuel, 6, 17 182, 191n.15 Kantian noumena, 75 through time, 4, 28 Kaplan, David, 41, 223 vagueness of, 45–67 knowledge, 109, 118, 120–21 impressionism, 14 Köhler, Wolfgang, 6, 11–12, 14–16, indeterminateness, 87n.26, 88, 24 109, 123 Kripke, Saul A., 99n.6, 130–31, 148, infants 175, 189, 247 ontology, 33–34 relation to objects, 21–24, 26, 27, language(s) 29, 32–34 aligned to world’s structure, 218 surprise reactions, 30–34 alternative/imagined, 13–14, innate sense of similarity, 7, 8 36–42, 80–81, 110, 116–19, 122, innate sense of unity, 7, 8, 20, 23, 26 179–91, 198, 211, 223–28 internal questions, 144, 161 character of, 234 interpretive charity, 204, 206, concepts of unrestricted 211n.19, 213n.23 existence, 139 basic idea/element of, 151, 182, contingent statements, 74–78 200 without criteria of unity, 13 of disputants, 201 defi cient, 198, 200–201, 209–11 for each community, 197, 198 existence of world independent and four-dimensionalism, 171 of, 73, 102, 172–73, 189, 220 ontological arguments, 178–86, infl uence on experience, 11 191, 192 interpretation of, 108–10, 112, of ontological sentences, 176 114–15, 119, 121, 123, 127–28, of physical objects, 203 148–49, 151, 169–70, 180–81, presumption in favor of, 205 198, 221, 224, 227–30 purpose of, 149 learning, 8, 10, 23, 78, 109, 185 and revisionary ontology, 100, metaphysically impossible, 122, 102–4, 111, 112 126, 128 and semantics, 237n.24, 241 ontological, xii, 211, 213–18 index 255

philosophically best, 212–13, problem of temporary intrinsics, 217–18 93–94, 231–32, 241 primary, 11 semantics, 41, 42 quantifi er-like expressions, set of possible worlds, 223 121–22, 127–28, 130–31, 172, theory of counterfactuals, 217 177, 203 theory of linguistic interpretation, quantifi ers of, 171, 220 127–28 referential apparatus, 41, 42 universalism, 99, 103, 106 semantics, 41, 42, 122, 140, 158–59, view on vagueness, 43, 50n.12, 171, 173, 234–43 60n.25, 88–89, 191, 222n.3 sense-data, 17 linguistic decision, 70, 71 space-time, 13, 17 linguistic idealism, 41, 101 “too strange,” 189, 191, 242–43 linguistic phenomenology, 123 truth-conditionally equivalent, Locke, John, 109, 154n.18, 178–79, xi-xii, xiv 184, 191n.15 truth theory for, 77 Lockean English, 179–91 unity in, 10, 15 logical constant, 71, 89 See also English language; logical joints, 95n.40, 121–22, 126–27, sentences 129–30, 171–72, 192–95, 214–15, language-shaped fact, 79 233–34 learning loose talk, 105–6, 169, 183 and empiricist position, 19, 26 judgments of unity based on, 15 mathematical sentence, 151n.13 of language, 8, 10, 23, 78, 109, 185 matter Leibniz’s Law, 45–46, 63, 108, 110, articulated, 29, 34 166 bits of, 104, 133, 135, 221 Lewis, David constituting object, 84 “count by temporary identity,” portions of, 100n.7 109–10 underlying substance, 6 example of restricted quantifi er, maximal object, 27–28 86, 107 McGrath, Mathew, 197–200, explosionism, 133, 134, 136 201nn.5–6, 202–5, 208–11, 213, four-dimensionalism, 106, 153–55 217, 218 model of use vs. eligibility, 130 Melville, Herman, 174–75 naturalness-presumption, 119–20, mereological essentialism, 145, 173–77, 194 153, 154n.18, 155–58, 159n.21, on natural properties, 127, 170, 173, 160–62, 171, 179–85, 192 176, 193–95, 233 mereological sum, 84, 86, 87, 88, 145, on natural things, 127 154–57 as nonbeliever in structured facts, mereologist, 68–69, 81–84 199, 213 Merricks, Trenton, 110, 113, 115n.26 ontological position, 87, 91–93, metaphor, 73–74, 188–89 133, 159, 230–31 metaphysical realism, 81–82 perdurantism, 231 metaphysical simples, 97 256 index minimizing change, 28, 29, 34 focusing on, 20–26 mistakes, 99, 108, 113, 114, 183, four-dimensionalist, 48, 125 185–86, 229n.13, 230, 234 identity of, 101, 106, 108, 109, 123, Moby Dick (Melville), 174 182, 191n.15 Moore, G.E., 91, 97, 98 made up of matter, 84, 221 motion, 12–13 maximal, 27–28 mysticism, 75 objectivity without, 36–44 ontology of physical, 85, 144–77, Nagel, Thomas, 75n.7 192, 203, 221–22 natural fact, 217 Putnam’s concept of, 40, 80 natural joint, 213, 216, 217 Quine’s views on, 5, 9, 13, 96 naturalness, 213n.23, 233n.19 ready-made, 172, 188 naturalness-presumption, 119–21, scattered, 116 170n.46, 173–77, 194 sentences dealing with, 92 natural property, 127, 170, 173, 176, Siderian, 125, 127, 128 193–95, 216, 217n.32, 233 sortal-dependent, 30, 34 natural proposition, 217 as space-time portions, 5, 42, 43, natural thing, 127, 233n.19 96 necessity invariantism, xv stationary, 27 negative fact, 208n.17 temporal parts of, 93–94, 172, 223, neo-Fregeanism, 239–40 247 neo-ontology, 212–15, 218 unitary, 44, 221 nihilism, 97, 104, 145, 184, 213n.25 Ockham’s razor, 118n.31 nominalism, 161–63, 192, 221, 228, ontological axioms, 160, 165 243–46 ontological commitment, 86, 89, 90 non-charity to excessive ontological dispute, xii, 160, 161, 170, expressibility, 201 211, 221, 222, 230, 243 normative invariantism, xv ontological language, xii, 211, no-two-things-in-same-place 213–18 principle, 108–11, 113, 160–61, ontological sentence, 110, 160, 165, 168, 184–85, 204 169, 176 numbers, 243 ontology adult, 33–34 objectivity, 36–44 and alternative languages, 220–46 objects common sense, 43, 85, 87, 91–92, basic, 27–35 94, 96, 103, 106, 117, 138, 143, changes in, 24 146, 164–77, 191 classifi cation of, 7, 19 deep, 89–91, 93, 125 common sense regarding, 93–94 issues as merely verbal, See verbal composite, 3–5, 145, 146, 184 dispute(s) concepts of, 139–40, 142 of Lewis, 87, 91–93, 133, 159, existence of, 106, 108–9, 123, 130, 230–31 140–41, 143, 182, 203 of objects, 85, 144–77, 203 and existential relativism, 132–38 of Quine, 32, 33, 89 index 257

revisionary, 96–123, 203, 221 metaphysical realism, 81–82 shallow, 85, 89, 90, 91 paradox, 173, 174 as study of logical joints, 127 on quantifi cational concepts, 78 ontology-based natural joints, 216 on quantifi er expression, 70, 71, “On What There Are” (Shoemaker), 187 48 on quantifi er variance, xvi, 69, 76, ordinary language philosophy, 82, 82, 187 97, 98 “Question of Realism,” 36, 38 ostensive defi nition, 236 Reason, Truth, and History, 39 on structured facts, 78, 79 perception, 174, 205 See also charity to perception qualitative continuity, 5, 6 perceptual example, 109, 112, 114, quality space, 7, 8, 23, 26 129 quantifi cational apparatus, 68, 71, 82, perceptual experience, 185 95, 233n.19 perdurantism, 221–24, 229–32, 234, quantifi cational constraint, 171–73, 237, 242, 245, 247–48 176, 177, 192–93 permissive views, 48–49 quantifi cational joint, 217 persistent philosophers, 204 quantifi cational structure, 214–18 Philosophical Investigations quantifi er expression, 70, 187–88, (Wittgenstein), 112 195 philosophical principles, 204 quantifi er-like expression(s) phlogiston theory, 173, 174 and defeasible presumption, 177 Piaget, Jean, 21–22 formal function of, 172 platonism, 161–63, 192, 221–22, 228, in philosophically best language, 243–46 213 precisifi cation, 50–66, 89n.29 and revisionary ontology, 121–22, primary language, 11 203 principle of charity, 98–100, 104, 114, semantic functions, 239 120–21, 126, 148–49, 170n.46, semantic restriction hypothesis, 174, 180, 229–30 176 principled dispute, 204 and Sider’s four-dimensionalism, problem of temporary intrinsics, 127–28, 130–31 93–94, 231–32, 241 and Sider’s logical joints, 233 processes, 222 quantifi er relativism, 138–43 properties, 164n.30, 243 quantifi er restriction, 42, 86, 106–7, Putnam, Hilary 134, 137, 154, 167 argument against realism, 225 quantifi ers concept of object, 39, 80, 140 genuine, 121 conceptual relativism, 38, 43, 68, of language, 171, 220 138–39, 173n.50, 187 of plain English, 126, 194 criticism of, 40–41 Sider on, 125, 172 on Davidson, 76–77 vagueness of, 43, 87–89, 129–30, and mereologist’s position, 83 191 258 index quantifi er variance, xi, 186–95 and space-time, 9, 13, 14, 18, 19, 32 and Carnap, xvi and anti-realism, xiv, xvi,187, 189 as corollary of Urmson’s dictum, Reason, Truth, and History (Putnam), xii, xiv 39 defi nition of, xiii reference, 74, 75–76, 157–58, 236, doctrine of, xii 238–39 and imagined communities, 203 reference magnetism, 216n.29, 218 Lewis’s rejection of, 87 referential ambiguity, 46–51, 62 misunderstandings of, xiii-xiv referential apparatus, 41, 42, 44, and ontology of physical objects, 59–62, 65, 67 85, 192 referential correspondence theory , possible, 80, 85 76, 140, 141 problems raised by structured referential semantics, 36, 38, 42, 122, facts, 79 140, 159n.21, 189n.12, 239 and Putnam, xvi, 69, 187 restricted quantifi er, 42, 86, 106, 107, vs. quantifi er restriction, 86 134, 137, 154, 167 and realism, xiii, xvi, 68–95 restricted view of structured fact, and semantics, 239 209, 210 quasi-nihilism, 104, 111, 113, 117, revisionary ontology, 98–123, 203, 118n.31, 145 221, 222n.3 “Question of Realism, The” Roots of Reference, The (Quine), 8 (Putnam), 36, 38 Russell, Bertrand, 49, 59 Quine, W.V., 19, 99 “bodies,” 119 “Scattered Objects” (Cartwright), and body-mindedness, 8, 9, 10–11 116 language theories, 189–90, 242 semantically restricted quantifi er, notion of object, 5, 9, 13, 96 154, 167 ontology, 32, 33, 89 semantic indecision, 49, 60, 61, 62, sense of similarity, 7, 10 88–89 steadfast laymanship, 85 semantics, 41, 42, 122, 140, 158–59, views on quality space, 23 171, 173, 234–43 views on translation of foreign sensation, 16 language, 14 sense-data philosopher, 81 Quinton, Anthony, 7 sensory organization, 12 sentences rationality, 99 characters of, 223–26, 228–29, reality/realism 235–36, 238–41, 245 existence as logical joint in, confl icting sets of, 108 214–15 context of, 238 physical, 7 dealing with physical objects, 92 Putnam’s argument against realism, disputed, 82–83 225 equivalent, xi, 36, 78 and quantifi er variance, xiii, xvi, general, 204 72, 68–95 identity, 51 index 259

interpretation of, 98–99 Sosa, Ernest, 73n.6, 132–38, 141, 143 ontological, 110, 160, 169, 176 space precisifi cations of, 53–61, 64, 65 no-two-things-in-same-place vague identity, 48–49, 89n.29 principle, 108–11, 113, 160–61, See also truth-conditions/values of 165, 168, 184, 204 sentences quality, 7, 8, 23, 26 sets, 100, 164n.30, 192, 201n.7, 243 unity through, 4 shallow ontology, 85, 89, 90, 91 space-time shape, 94n.39 language, 13, 17 Shoemaker, Sydney, 12–13, 14, 48–49, portions as objects, 5, 23, 24, 25, 42, 53–54, 61n.26, 89n.30 43, 96, 119 Sider, Theodore reality, 9, 13, 14, 18, 19, 32 arguments for perdurantism and spatial connectedness, 4 unrestricted mereology, 222n.3 spatiotemporal continuity, 5, 6, 12–13, on charity, 170, 182 16, 22, 222 constraints on linguistic Spelke, E.S., 27, 31, 33, 34 interpretation, 192–93 steadfast laymanship, 85 four-dimensionalism, 124–31, 212 Strawson, P.F., 32, 97 invariantism, xiv strict talk, 105–6, 169, 183 logical joints theory, 95n.40, structure, 213–14, 217 121–22, 171–72, 192–95, 215, structured fact, 78, 79, 198–200, 202, 233–34 206–7n.16, 208–11, 213, 218 ontological arguments, 189, 213–17 substance, 6 philosophically best language, substantive dispute, 199, 207, 210, 212–14 216n.29 on physical objects, 145n.2 supervaluationist approach, 46, 50, quantifi cational constraint, 171–73, 54n.16, 55, 56n.19, 191n.16 176, 177 Siderian objects, 125, 127, 128 Tarski-style semantics, 122, 159n.21, similarity, 6–10, 23, 26, 127, 141, 241n.30 194–95, 216–17 Tarski-style theory of truth, 76, 77 similarity-based natural joints, 216 temporal composites, 4–5 similes, 73–74 temporal parts, 86, 96, 111n.22, 154 simples, 97, 145n.2, 164n.29, 201n.5, doctrine of, 84, 145, 184, 190 201n.7, 213n.25 existence of, 171 simpliciter, 86 and four-dimensionalism, 96n.1, simplicity, 138 124, 125, 145, 154n.16 Singer, Peter, 115 and linguistic behavior, 156, 157 SKSW experiment, 31–34 of objects, 93–94, 145, 146, 154, something, xii, 43, 68–72, 88, 136, 155, 172, 223, 247 155, 189 rejected by endurantists, 235 sortal-dependence, 31, 85, 86n.24, and semantics, 235, 238, 240 95n.40 and universalism, 106 sortals, 28–30, 34 See also doctrine of temporal parts 260 index temporary intrinsics, 93–94, 231–32, through space, 4 241 through time, 5, 12, 16 theory of counterfactuals, 217 unity relation, 116–18, 120 thing, 84, 85–86, 100 universalism, 96, 97, 100, 104, 106, Thomson, Judith, 100n.7 113, 116, 117 thought, 73 unnatural things, 127, 233n.19 time, 5, 12, 16 unrestricted existence, 139 See also space-time unrestricted mereology, 184, 190, truth, 75–78, 99, 137–38, 147, 201, 239 222n.3 truth-conditionally equivalent unrestricted quantifi er, 107, 134, languages, xi-xii, xiv 136–37, 139–40 truth-conditionally equivalent unrestricted view of structured fact, sentences, xi 209 truth-conditions, 78n.15 unstructured fact, 78–79, 92, 198–99, truth-conditions/values of sentences 201–2, 207, 209–10, 213n.25 in alternative languages, 59, 73, 74, Urmson, J.O., xi, xii, xiv 76–79 charitable, 156, 158–59, 162, 172 vague identity sentence, 48–49 via composition, 158 vagueness, 43–44 and linguistic decisions, 70, 71 epistemic, 47n.4, 89n.29, 175n.55 and precisifi cation, 56 of identity, 45–67 and reference of word, 37–38 of identity sentences, 48–49, relating to physical objects, 92 89n.30 and revisionism, 110, 119 in language of existence and and semantics, 60, 89, 242 identity, 123 Sider’s view of, 122 Lewis on, 43, 50n.12, 60n.25, and verbal disputes, 150, 201, 206 88–89, 191, 222n.3 truth-functionally equivalent of quantifi er, 43, 87–89, 129–30, sentences, 206–7n.16 191 of referential apparatus, 61–62, 65 unconscious inference, 205 van Inwagen, Peter, 90n.31, 91–93, 97, Unger, Peter, 96, 99 103, 104n.12, 106n.16, 110, 117, unity 137, 138, 177n.57 bodily, 6–9, 14, 17–18, 20, 25–26 verbal dispute(s), 146–64 criteria of, 3–6, 8, 10, 14–17, 19–21, between endurantists and 23, 26 perdurantists, 222, 224 experience of, 11 example of, 146–47 group, 12 on existence, 199, 214 innate sense of, 7, 8, 19–20, 23, 26 and interpretive charity, 161–63, judgments of based on previous 197, 228–34 learning, 15 ontological, 192, 202, 203, 211, 221 in language, 10, 15 on physical objects, 153, 222 sense of, 3–26 simple conception of, 245 and similarity, 6–10 and substantive dispute, 210 index 261

and truth conditions of sentences, Wisdom, John, 44, 61, 73 150, 201, 206 Wittgenstein, Ludwig, 13, 36, 44, 90, and unstructured facts, 207 92, 97, 102, 112 and verbal questions, 205, 208 Wittgenstein paradox, 148, 175, verbal mistake, 229n.13, 230 181 verbal question, 205, 208, 223 words, 36–37, 38, 60, 119, 127–28, 195, 238 whales, 174–75, 185 Wiggins, David, 29, 86n.24 Xu, Fei, 27–34 Williamson, Timothy, 63n.32, 89n.29, 175n.55 Yablo, Steve, 246n.37