Referring to the World: An Opinionated Introduction to the Theory of Reference 0197537359, 9780197537350


291 13 3MB

English Pages [308] Year 2021

Report DMCA / Copyright

DOWNLOAD PDF FILE

Table of contents :
2883a782_cover(full permission)
Title_Pages (1)
Contents
Dedication
Foreword
The_Mystery_of_Reference_and_Objective_Representational_Content
Inner_Fitness_and_Outer_CauseThe_Two_Factors_of_Content
Against_Jazz_Combo_Theories_of_Meaning_and_Reference
Puzzles_of_CoreferenceTheme_and_Variations
Concepts_Conceptions_in_the_Psychology_of_the_Referring_Mind
Representing_RepresentationsThe_Priority_of_the_De_Re
The_Things_We_Do_with_Empty_Names
Works_Cited
Index
Recommend Papers

Referring to the World: An Opinionated Introduction to the Theory of Reference
 0197537359, 9780197537350

  • 0 0 0
  • Like this paper and download? You can publish your own PDF file online for free in a few minutes! Sign Up
File loading please wait...
Citation preview

Referring to the World

Referring to the World An Opinionated Introduction to the Theory of Reference K E N N E T H A .   TAY L O R

1

3 Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford is a registered trade mark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016, United States of America. © Oxford University Press 2021 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. Library of Congress Cataloging-​in-​Publication Data Names: Taylor, Kenneth Allen, 1954–2019, author. Title: Referring to the world : an opinionated introduction to the theory of reference / Kenneth A. Taylor. Description: New York, NY, United States of America : Oxford University Press, [2021] | Includes bibliographical references and index. Identifiers: LCCN 2020049975 (print) | LCCN 2020049976 (ebook) | ISBN 9780195144741 (hb) | ISBN 9780197537343 (epub) | ISBN 9780197537350 Subjects: LCSH: Reference (Philosophy) Classification: LCC B105 .R25 T395 2021 (print) | LCC B105. R25 (ebook) | DDC 121/.68–dc23 LC record available at https://lccn.loc.gov/2020049975 LC ebook record available at https://lccn.loc.gov/2020049976 DOI: 10.1093/oso/9780195144741.001.0001 1 3 5 7 9 8 6 4 2 Printed by Integrated Books International, United States of America

Contents Foreword 

ix

1. The Mystery of Reference and Objective Representational Content 

1

2. Inner Fitness and Outer Cause: The Two Factors of Content 

25



31 44 48



1. Preliminaries  2. The World Rushes In  3. Semantic Presentationalism and the Epistemic One-​Sidedness of Reference  4. Objectuality and Objectivity Again  5. Kantian and Fregean Roots 

25 26

3. Against Jazz Combo Theories of Meaning and Reference 

57



Preliminaries  Jazz Combo Theory and the Priority of the Sentence  The Cause-​Norm Gap  Jazz Combo Theories and the Social-​Dialectical Nature of Objectivity  5. Some Conceptual Tools and Distinctions 

67 72



6. Closing Argument: Against the Semantic Priority of the Sentence 



1. 2. 3. 4.

5.1. 5.2. 5.3. 5.4. 5.5.

Dynamic Priority versus Semantic Priority  Pragmatics and Partiality  Objectual versus Objective Representations Again  Syntactic Correlativity of the Sentence and its Constituents  Semantic Bootstrapping 

4. Puzzles of Coreference: Theme and Variations 

1. 2. 3. 4. 5. 6. 7. 8. 9. 10.

The Many Coreference Puzzles  Combinatorial Syntax and Singular Reference  Names as Devices of Explicit Coreference  Names, Their Spellings, and the Drainage Thesis  Drainage as Ambiguity  Drainage as Indexicality  Drainage as Predicativity  Coreference Revisited  Comparisons  Conclusion 

57 59 62

73 76 77 79 82

83

87

87 94 99 102 105 113 118 123 136 140

viii Contents

5. Concepts, Conceptions in the Psychology of the Referring Mind 

142

6. Representing Representations: The Priority of the De Re 

177

7. The Things We Do with Empty Names 

235

Works Cited  Index 

287 293



1. Concepts and Conceptions  2. Rationality and the Psychology of the Referring Mind  1. 2. 3. 4.

1. 2. 3. 4. 5. 6. 7. 8. 9. 10. 11.

From World to Mind and Back Again  Commitments Ascribed versus Commitments Undertaken  Substitution Puzzles Reconsidered  From De Re Ascriptions to De Re Attitudes  Preliminaries  Empty Names and Fictional Discourse  Empty Names as Merely Objectual Singular Representations  Empty Names in Veridical and Non-​Veridical Language Games  Truth versus Truth-​Similitude  Truths about Fictions  Notional Objects and Illusions of Subject Matter  What Objects Are Not  Taking Stock  Believing with Singular Purport in the Non-​Existent  Ascribing Belief in the Non-​Existent Again 

143 167 177 183 208 218 235 237 242 244 258 261 262 264 271 273 278

 

The slow miracles of thought take shape through patience into grace. —​Elinor Wylie, “O Virtuous Light”

Foreword On December 2, 2019, our beloved colleague and friend, Ken Taylor, announced to all of his Facebook friends that the book he had been working on for years, Referring to the World, “finally existed in an almost complete draft.” Completing references, eliminating redundancies, and writing a conclusion was all that remained. But that same day, while at home in the evening of December 2, Ken died unexpectedly. Claire Yoshida and Kiyoshi Taylor located the computer files for Ken’s book for us. The book you have before you consists of those files. We have corrected typos, added some extra commas in some of Ken’s long sentences and some extra quotation marks for the many terms he introduces, made numbering of examples and such things consistent, and filled in the references as best we could. Some of the footnotes were notes from Ken to himself about references to add, or points to elaborate on, without enough detail for us to confidently fill in the missing information. We left these as they were written and they are set in small caps to indicate this. The index was compiled by Ken Perry of CSLI Publications. Krista Lawlor and Debra Satz helped in a number of ways. The book abounds, as one would expect, with insights, examples, trenchant analyses of criticisms of scores of philosophers, and subtle theorizing that weaves this all together in a new and exciting framework for thinking about reference and many related issues. Sadly, it lacks a concluding chapter. Anna-​Sara Malmgren, John Perry, Mark Crimmins, Robin Jeshion

Ken Taylor’s Facebook Post, December 2, 2019. Now this calls for a minor celebration. My long delayed book, Referring to the World, finally exists in an almost complete draft. It just needs a preface and a conclusion and a beefed up scholarly apparatus of notes and citations. This creature started out its life years and years and years ago as a completely different book. It was first commissioned and conceived of as an opinionated introduction to the theory of reference. it was supposed to be a SHORT

x Foreword introduction too . . . of no more than 60,000 words. I thought I could write such a thing pretty quickly. Maybe I could have. But the sprawling book that I actually wrote became more and more opinionated and much more focused on arguments designed to back up those often philosophically controversial opinions. Somewhere along the line—​I don’t know when exactly—​it gradually became less and less of an ­introduction to anything except my own views. Some sort of Sorites Paradox lies therein perhaps. To be fair to me, it still does contain substantial traces of its original introductory design. Because throughout most of it, it tries hard to be scrupulously dialectally fair to competing points of view—​all of which, though plausible and appealing, are ultimately wrong. The opinions and the supporting arguments have been developed over way too many years, in a series of articles, some of which were written with the book in mind others of which were not. But now they are all gathered together in a single if somewhat sprawling argumentative thread. Maybe some will find them more convincing that way. But maybe not. We shall see . . . I hope. It will still take some time to spit and polish it all up into a form suitable for shipping off to the publisher. But god it feels good to get this way too long delayed book project in close to final form. I think I’ll pour a glass of wine to mark the occasion, before plunging back into the work that is still to be done. One of my favorite poems includes the line “the slow miracles of thought take shape through patience into grace.” Amen to that!

1 The Mystery of Reference and Objective Representational Content This book is an investigation of the mystery of reference in both private thought and public talk. Just what is it for some bits of either our shared public language or our inner thoughts to refer to or stand for bits of the world? In virtue of what does the relation of reference obtain between some bit of the world and some bit of either outer language or inner thought? What about apparent reference to putatively non-​existent objects, like Santa Claus or Sherlock Holmes? We appear to think and talk about objects that do not exist. But there are no such objects. So just how do we manage to think and talk about them? Or consider abstract objects, like numbers, that are thought by some to exist outside the spatial-​temporal order. We appear to think and talk about such objects as well. But it is a mystery how, if at all, the reach of our thought can extend beyond even the bounds of space and time. Such questions are tied up with some of the deepest, most challenging, most enduring problems of philosophy. They concern how it is possible for the mind and language to make contact with an apparently mind-​independent reality. In one form or another, this problem has bedeviled the imaginations of philosophers as diverse as Aristotle, Plato, Descartes, Hume, Berkeley, Kant, Russell, Frege, Heidegger, Husserl, and Wittgenstein. To be sure, the commonality of concern is sometime masked by stark differences of philosophical idiom. The philosophers of the early modern period, for example, worried incessantly about our “ideas” and about their “resemblance,” or a lack thereof, to a possibly mind-​independent reality. Partly because an idea is more likely to resemble some other idea than it is to resemble any mind-​ independent reality, the idiom of ideas proved not to be an altogether stable and fruitful means of explaining the relationship between mind and world. Moreover, partly because of the temptation to see ideas as essentially private, it was hard for the moderns to offer any convincing account of how ideas could serve as the basis of shared contact with a mind-​independent world.

Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0001

2  Referring to the World Early in the twentieth century, with the beginning of the so-​called linguistic turn in philosophy, the idiom of ideas was supplanted. The idiom of language, of words and sentences, sometimes with a stress on their logical forms, sometimes with a stress on their everyday use and abuse, came to occupy center stage. This mere shift of idiom was widely advertised as a great and clarifying revolution in philosophy. Though this revolutionary shift did not on its own suffice to instantly render the mind’s power to think and talk about a mind-​independent world more conceptually transparent, it did help to dispel any lingering temptation to explain the mind’s ability to think and talk about the world in terms of the resemblance between its representations and its representeds. Words and sentences are clearly the vehicles or instruments of the mind’s thought and talk. They are the mind’s representations. They are not the mind’s objects, not its representeds. Once this insight has been gained, it is a short step to seeing that there need not be any deep resemblance between the vehicles of thought and talk and the objects of thought and talk. Taking that short step does not, however, settle the question of how possibly words and thoughts manage to be about objects in a mind-​independent  world. Though this book addresses a number of discrete questions about reference, the answers I proffer to them stem from a single underlying view about the source and nature of what I call “objective representational content.” By that I mean the property that our words and our thoughts have of being “semantically answerable” to how things are by objects in the world. Consider the following example. There is a person Smith with whom I am acquainted. I have many beliefs about her. One such belief is well-​expressed by my utterance of the form of words, "Smith has a generous spirit and a quick and incisive intellect." Now the truth or falsity of my words depends entirely on how things are by Smith and her properties. My words are true just in case Smith, and no one else, has just those qualities, and no other, of spirit and intellect. How things are by Smith with respect to her qualities of spirit and intellect determines how things are in the way of truth and falsity by my thoughts and my words. I mean to point only to unremarkable facts of this kind when I say that our thoughts and our words are semantically answerable to how things are by objects and their properties. A major goal of this book is to explain how possibly both sentences in our public language and the representational vehicles of private thought, whatever they are, achieve this remarkable property of semantic answerability to the world.

The Mystery of Reference  3 My explanation of how our thoughts and words manage to be semantically answerable to the world will turn, in part, on the fact that the vehicles of thought and talk are structured wholes built out of parts that are themselves semantically valued. For example, the sentence “Smith has a generous spirit and a quick and decisive intellect” contains as constituent parts the name “Smith” and the predicate “. . . has a generous spirit and a quick and decisive intellect.” Intuitively, the name has the semantic function of standing for or referring to a certain object, viz., Smith, while the predicate has the semantic function of expressing a certain property. Moreover, the grammar of our language entails that a sentence containing a name and a predicate arrayed in just the way that this name and this predicate are arrayed may be used to make a statement and express a belief about the referred to object to the effect that it has the expressed property. Such an approach proceeds in three steps. First, it appeals to a theory of what we might call “semantic value” for the basic constituents of our language. The theory of semantic values should include at least a theory of reference for the referring expressions of our language and a theory of “expression” for the predicates of the language. Second, it should include principles of what we might call syntactic combination. Such a theory would explain how less complex expressions can be combined to yield more complex expressions. Finally, the theory would include principles of semantic composition. Principles of semantic composition would determine the semantic values of whole sentences as functions of the semantic values of the constituent parts of those sentences and the way those parts are organized to yield wholes. Though the approach just gestured at will strike many as highly intuitively compelling, there are those who take it to be misguided from the very start. One goal of this book is to assess both the considerations that weigh in favor of my own preferred approach and also those considerations that weigh against it. As such, though the book sets out to defend a peculiar and novel approach to the problem of objective representational content, it should also be serviceable as something of an introduction to the foundational issues facing any theory of reference. My aim is not, however, to provide a survey of the ins and outs of various alternative approaches to the problem of objective representational content. My aim is, rather, to think through, in a systematic way, the range of problems with which any theory of objective representational content must deal and to give a variety of reasons for preferring one particular approach—​what I call two-​factor referentialism—​to its main

4  Referring to the World rivals. While I do attempt to give rival approaches a fair hearing, I make no claim to neutrality. Prominent among those who reject the kind of explanatory approach defended in this book is Donald Davidson. Davidson readily admits that our beliefs and our utterances are true or false of a world that is mostly independent of mind. And to that extent he agrees that our thought and talk have what I call objective representational content. But he claims that concepts such as reference play no essential role in explaining the relationship between mind and/​or language, on the one hand, and the world, on the other. As he puts it, “We don’t need the concept of reference; neither do we need reference itself, whatever that may be.” (Davidson, 1977, 256) Similarly, Robert Brandom claims that the supposed reference relation is “a philosopher’s fiction, generated by grammatical misunderstandings.” (Brandom, 1994, 324) Though Brandom affords what he calls an “expressive role” to what he calls “reference talk,” he, like Davidson, denies that there is a genuine reference relation. Moreover, he denies as a consequence that any such relation has any role in explaining the relationship between mind and/​or language and the world. Views that seek to explain where objective representational contents come from, but without appeal to anything like a reference relation, pose a far-​reaching challenge to the explanatory strategy favored in this book. Our strategy takes reference to be both a real relation in nature—​though admittedly one that is itself in need of explanation—​and a relation that can be used, once it has itself been explained, to shed further explanatory light on the nature of objective representational content. This approach presupposes that although reference is itself neither an explanatorily nor metaphysically primitive notion, it is more metaphysically and explanatorily basic than the objective representational contents of whole sentences or entire thoughts. Call a theory with this character a “bottom-​up” explanatory theory of reference. In Chapter 3, we shall dig more deeply into Brandom and Davidson style arguments for thinking there can be no bottom-​up explanatory theory of reference. For now, it is enough to note that one major part of what is at issue between friend and foe of a bottom-​up explanatory theory of reference is a dispute about the significance of what might be called the priority of the sentence. It is hard to deny that names and predicates perform their semantic work primarily in the context of the production of sentences. Except in very special circumstances, the utterance of an isolated name or an isolated predicate would not yet constitute the staking out of a determinate claim about an

The Mystery of Reference  5 object and its properties.1 Strikingly, this seemingly trivial observation leads some to the consequential conclusion that a referring expression, taken on its own, outside the context of a sentence in which it is deployed, is entirely semantically idle. Wittgenstein (1953) famously put the point as follows: Naming is so far not a move in the language-​game—​any more than putting a piece on a chessboard is a move in chess. We may say: nothing has so far been done, when a thing has been named. It has not even got a name except in a language game.2

Wittgenstein apparently thinks that an expression functions as a name or referring expression only if it plays a (yet to be specified) role in an already constituted “language game.” If we take a language game to be something like a totality of significant linguistic moves or speech acts, as they are often called—​where these include acts of asserting, promising, requesting, to name just a few—​then Wittgenstein’s claim would seem to be twofold. First, acts of referring happen only in the performance of already significant speech acts—​paradigmatically, perhaps, acts of assertion. Wittgenstein seems to want to say, that is, that referring is not some separable act, somehow antecedent to any speech act. It is not that one first refers to Socrates and then makes an assertion about him. Rather, one refers to Socrates only by and in the course of asserting something or other about him.3 Now many are tempted to conclude, on the basis of such considerations, that there can therefore be no separable reference relation, specifiable independently of the specifying of a space of actual and possible speech acts.4 On this sort of view, talk of the reference of a name must be understood entirely derivatively, entirely in terms of the role played by the name in a totality of sentences, the potential utterances of which constitute a space of already linguistically significant speech acts. Reference is not, in other words, explanatorily prior to asserting and other speech acts, but is, at best, 1 Editorial note: Some of the footnotes were notes from Ken to himself about references to add, or points to elaborate on, without enough detail for us to confidently fill in the missing information. We left these as they were written and they are set in SMALL CAPS to indicate this. (See Foreword.) Reference John Perry’s “Davidson’s Sentences and Wittgenstein’s Builders” (1994) and the primitive language games described therein. Note that when there is an articulate background of prior “issues” (Who just entered the room is a question in the air), the utterance of a name does count as a significant linguistic performance. 2 Wittgenstein, 1953, §4. 3 Reference to Searle on referring vs. predicative speech acts. 4 Give examples of those who endorse the view. Quine, Davidson, Brandom, others?

6  Referring to the World explanatorily dependent on them and, at worst, an altogether explanatorily idle notion. In Chapter 3, I shall say what can and can’t be said for the principle of the priority of the sentence and for the correlative claim that reference is at best a derivative and at worst an entirely idle notion. I shall grant that there is a sense in which the principle is true, perhaps even trivially so. It seems exactly right to say, for example, that neither we theorists nor the child acquiring her first language can gather direct evidence about the reference of a name, say, or the properties expressed by a predicate, say, except by witnessing the use of the predicate or name in the context of sentences of various sorts. This fact points to what I call the dynamic priority of the sentence. The dynamic priority of the sentence is a straightforward consequence of the hardly unexpected fact that significant linguistic performances typically involve the production of whole sentences. But I shall show that the dynamic priority of the sentence does not support skepticism about the bottom-​up explanatory utility of reference in an account of objective representational content. Arguments against bottom-​up explanatory theories of reference from the supposed priority of the sentence purport to show that not just reference but any other semantic property attaching in the first instance to mere constituents of sentences rather than to whole sentences themselves must be devoid of both independent metaphysical standing and foundational explanatory utility. My main argument against the priority of the sentence is a blocking argument. That is, I attempt to show that to the extent that the priority of the sentence is true, it does not, in fact, entail that reference has no foundational explanatory utility. To be sure, even if my blocking argument succeeds in keeping open the question of whether reference has some foundational explanatory work to do in the theory of objective representational content, that will not suffice to solve the more fundamental puzzle of just what in the order of things the reference relation could possibly be. Solving that puzzle requires us to explain just how names and other referring expressions could possibly manage to have the property of standing for objects (and the correlative puzzle of how predicates manage to express properties) to begin with. Pointing the way toward the unraveling of that puzzle is one of the main burdens of this book, though I do not flatter myself that I have penetrated to the bottom of the mystery of reference. Before launching head-​on into the main arguments of this book, I spend much of the remainder of the current chapter dwelling on just how puzzling the fact of objective representational content, and with it the reference relation, really is. Viewed in one way, our words are just marks on the paper

The Mystery of Reference  7 or perturbations in the air or patterns of nerve firings in our brains. Why should such marks and perturbations be semantically answerable to any independent existents at all? What endows different kinds of expressions with their distinctive roles in bringing about this semantic answerability? What exactly is a predicate, for example, such that it has the semantic role of expressing a property? What exactly is a name such that it has the semantic role of standing for an object? And what exactly is a sentence such that its truth or falsity determinately depends on how things are by this or that object and its properties. Standing in the philosophical wake of Kripke’s Naming and Necessity, one’s first thought may be that the key to answering such questions lies with matters causal and/​or informational.5 Our linguistic output is enmeshed in myriad causal and informational chains. At one extreme of such chains stand thinking and speaking agents—​stocked with their beliefs, desires, and intentions. At the other extreme stand events in the world. Worldly events initiate an inward rush of energy upon the portals of sensation. This inward rush gives rise to perception, which eventually occasions vast torrents of language, thought, and action in response. And it is not an entirely unpromising thought—​one that lies at the foundation of all forms of empiricism—​that it is in the causal interaction of mind and world via the portals of sensation that objective representational thought is somehow first constituted. But this suggestive initial thought carries us only so far. That is because the inward rush is more occasion than direct cause of either outer linguistic production or inner content bearing thought episodes. Outer linguistic production is not directly caused by events in the outer world but by our own inner intentions, desires, and beliefs. Our episode beliefs and judgments do not, as such, directly “encode” information about the world. A speaker may say that it is raining outside because she believes that it is raining outside and intends to inform her hearer of that fact. And she may want to inform her hearer that it is raining so that the hearer will act on that information by, for example, grabbing her umbrella or putting on a raincoat. But the crucial point is that she may say, believe, and intend all these things even in the absence of rain. Now many philosophers hold that inner episodes of believing, intending, or desiring themselves involve a play of inner representations, connected to the outer world via intricate causal and informational chains, ultimately anchored in perception of and action in and upon an external world. Because of the causal role of beliefs, intentions, and desires in bringing about 5 Kripke (1980).

8  Referring to the World utterances, many a philosopher has held that the semantic answerability of sentences in a public language depends on their being instruments for expressing certain already contentful inner thoughts. This approach makes thought prior to language in the semantic order. There is, I think, something deeply right about this approach—​though as always in philosophy, there are those who intensely disagree.6 But let us bracket such disagreements for the moment. If for the nonce we take it to be given that thought is prior to language in the semantic order, it follows immediately that if we could but tell a systematic story about how the association between sentences, on the one hand, and already semantically answerable thoughts, on the other, is achieved, we would have gone part way toward explaining how sentences, at least, manage to achieve objective representational content. But telling this much of the story would still leave us far from our ultimate goal of explaining the ultimate source and nature of semantic answerability, for reasons that will become clear shortly. Yet telling even this incomplete fragment of the ultimate story would itself be no trivial matter. For one thing, the number of thoughts potentially thinkable by minds like ours is infinite. So too is the number of sentences whose objective representational contents we potentially understand. Yet our minds are finite. The mapping between thought and language could not, therefore, be given by a mere list. Such a list would have to be infinite. No merely finite mind could grasp such a list via the exercise of merely finite cognitive powers. If the infinite mapping between thought and language is to be graspable by a finite mind, there must be a finite principle, graspable by us, that somehow generates and explains that infinite mapping. A promising strategy for dealing with the infinity problem was articulated by Frege early on and given fuller expression in more recent times in the work of Jerry Fodor.7 That strategy posits a language of thought. The idea is that to think a thought is to deploy an inner sentence of so-​called mentalese, the inner idiom of thought. Like the sentences of our shared language, the “sentences” of mentalese are supposed to be structured wholes built out of discrete and semantically significant parts. If there is at least a rough correspondence between the constituents of the sentences of mentalese and the sentences of the public language, and if rules of composition at the level of 6 Examples. Connection to those who accept the priority of the sentence. 7 Reference to Frege. Reference to Fodor. Quote passage where Frege seems to argue for something like LOT as a contingent psychological hypothesis about cognizers like us.

The Mystery of Reference  9 thought correspond to the rules of composition at the level of language, then, perhaps, we can give a systematic account of why a given sentence expresses just the thought it expresses and no other.8 Though it is not without its detractors, the language of thought hypothesis has, on my view, a great deal to recommend it and for much of the course of the argument of this book I will help myself to the assumption that there must be something like a language of thought. To be sure, this assumption does not solve our hardest problems. For even if we could uncover some finite principle that generates and explains the potentially infinite mapping between outer public languages and an inner mentalese, we would have succeeded only in pushing the deeper problem back one step. We would still be left with the unanswered question of what renders our inner thoughts themselves, rather than just the outer sentences that express those thoughts, semantically answerable to a mind-​independent reality. This book proposes some features of the language of thought that arguably make it possible for our thoughts and our words to be semantically answerable to a mind-​independent world of objects and their properties. I don’t pretend to have a knock down argument in favor of the story I tell in this book. But I shall argue that my story is superior to the competing alternatives.9 Now the problem of objective representational content has so thoroughly bedeviled the philosophical imagination that some philosophers have rejected the very idea that thought and language are semantically answerable to a mind-​independent reality. The most explicit and prominent advocates of such views are idealists of various stripes.10 Idealists typically find the idea that our thoughts are answerable to something lying outside of the mind nearly incoherent. The most immediate object of thought, the idealist maintains, must be something like our own ideas or our own sense data. To be sure, on some versions of this tale, we may have an indirect hold, via the ideas on which we have a direct hold, on that which lies outside the mind. But no matter the precise form it takes, idealism is an extreme and desperate doctrine. Though many a philosopher has claimed to have decisively refuted it, I myself suspect that it is very likely with us to stay, in some form or other, at least as a permanently tempting philosophical option, available 8 More on the language of thought hypothesis. 9 That, in a way, is small praise. Explaining where objective representational content comes from and how it fits in to the order of things, turns out to be a very hard problem indeed. As a consequence, even moderately convincing hypotheses are thin on the ground. 10 Berkeley and descendants.

10  Referring to the World when the going gets rough. I do not pretend to offer a decisive refutation of all forms of idealism. Because I doubt that idealism can be decisively taken off the table of dialectically available options, I will be content if I manage to diminish whatever dark allure idealism continues to enjoy. Even if I cannot demonstrate the outright falsity of idealism, it will be enough for my dialectical purposes if we can show that nothing in the philosophical common ground between idealists and non-​idealists is threatening to the claim that our words and thoughts are semantically answerable to a world largely independent of the mind. There is, I shall argue, a coherent story to tell about how possibly mind and language achieve semantic answerability to a mind-​ independent world. That story is not threatened by any philosophical or empirical premises plausibly available to both idealists and non-​idealists alike. Moreover, I shall argue that that coherent story is plausibly true. Showing that nothing in the common ground between idealists and non-​idealists threatens the idea that our thought and talk are semantically answerable to a mind-​independent world does not constitute a direct refutation of idealism. But it does show that nothing in the philosophical common ground makes idealism inevitable. And since idealism is a corner into which no one should willingly retreat, merely stripping it of any claim to inevitability should suffice to diminish its dark allure. One need not go all the way over to idealism, however, to experience a sense of puzzlement about the mind’s ability to reach beyond itself and hold itself answerable to an external reality. Viewed in one way, episodes of thinking amount to nothing but the energized sloshing of a vat of chemicals locked up in our skulls. How possibly could such energized goings-​on confined to our skulls be semantically “answerable” to anything at all, let alone to realities beyond themselves? Of course, viewed in one way, the mind is surely more than just a random vat of energized chemicals sloshing about. The mind is, I  think, rightly and informatively characterized as a field of inner representations. Its inner states are symbolic in that they have both (syntactic) form and (semantic) content. Understood in this way, standing for objects, representing or being about mind-​independent facts, expressing mind-​independent properties are of the very essence of both the mental and the linguistic. Though this is all arguably correct as far as it goes, it only goes so far. Without some further explanation of how configurations of neuronal states and processes constitute themselves as symbolic and representational, the mere insistence that they do it somehow or other has all the philosophical benefits of theft over honest toil.

The Mystery of Reference  11 To acknowledge the need for honest toil in our philosophizing about the representational powers of mind and language is not to deny those very powers, however. The mind’s representational powers are at the very core of our cognitive and social lives. Without the ability to cognize the world via inner representations, all knowledge would be impossible. Without our shared ability to represent a common external world, and to meta-​represent our own representations of that common world, human social life as we know it would not be possible at all. We could never communicate with one another about the world, never engage in rational collective action on a common world. Conceiving of a brain as a mind, rather than as a mere vat of energized chemicals is, in part, conceiving of it as the seat of representations and representational capacities.11 As a thoroughly modern naturalist, I endorse the view that the mind’s representations and representational capacities must ultimately arise out of merely material processes and states of our brains. But qua philosopher, rather than neuroscientist, it’s not my job to tell you the details of how actually the brain manages to constitute the mind. Philosophy is typically more concerned with the how possibly than with the how actually. Moreover, in the particular case at hand, there are good reasons to postpone a frontal assault on the how actually questions for another day. That’s because the mystery of objective representational content isn’t fully ripe for a frontal assault by those concerned with how the actual processes and states of the brain constitute the representational capacities of the mind. This is not to deny that there is, in our times, much revealing scientific, as opposed to philosophical, work currently being done on the mind and its representational powers. But for better or for worse, at our stage of the history of inquiry, problems about the source and nature of objective representational content remain pretty much a matter of determining “how possibly at all.” As such, addressing them is very much the business of philosophy. The problem of objective representational content remains mostly a how possibly problem despite the fact that the cognitive revolution of the last century has opened up an entirely new conceptual space. That revolution was in part a rediscovery of the lost wisdom of an earlier day. It amounted to a reclaiming of modern philosophy’s foundational insight that the mind is a field of inner representations, with real causes and effects in the world. As strange as it now seems, for a large chunk of the last century, especially

11 Consciousness too.

12  Referring to the World during the period when both philosophical and scientific behaviorism reigned supreme, this simple but crucial insight was widely derided and dismissed. It was thought both to be a source of deep philosophical confusion and to be a bar to scientific progress. But the cognitive revolution has achieved much more than simply reclaiming the lost wisdom of earlier centuries. It has conceptually augmented and refined that ancient wisdom. The best fruits of this revolution have been not just the actual detailed theories and models of the cognitive workings of the mind-​brain, but also the generation of whole new families of heretofore unimagined—​perhaps even unimaginable—​ hypotheses. Still, even with the ever-​ increasing consolidation of the explanatory gains of the cognitive revolution, there remains considerable conceptual distance between the concepts with which we cognize the biochemistry, functional anatomy, and even computational architecture of the working brain and the concepts with which we cognize the representational and semantic capacities of mind and language. Partly as a consequence, we still don’t know how to re-​identify the brain states and processes that we pick out via our neurochemical, neuroanatomical and even neurocomputational concepts as the contentful mental states that we pick out via our ordinary mentalistic concepts. As a consequence, we are more or less stuck with philosophizing, with asking how possibly at all, at least when it comes to explaining how both mental and linguistic content arises out of facts about the intact working brain. The key to making philosophical progress toward the day when a more frontal assault on the details of the “how actually” question is in order is to reduce conceptual distances. We can do so by self-​consciously and reflectively evolving a set of intermediate or “bridging” concepts. With proper deployment of an appropriate set of bridging concepts, we may hope to gain fuller imaginative acquaintance with the possibility that what we have heretofore conceived of as two metaphysically independent domains is really a single domain, antecedently conceived in two distinct ways, with two distinct and heretofore uncoordinated sets of concepts. To that end, this book articulates a set of concepts specifically designed to more fully enable us to imagine what in the natural order reference and related semantic relations might be. We do not initially conceive semantic notions like “standing for,” “being about,” “expressing,” and the like in terms that readily allow us to see how the semantic arises out of or reduces to the natural. Because we lack antecedently available, ready-​to-​hand concepts that bridge the gap between the semantic and the natural, we simply don’t initially know how to re-​identify reference

The Mystery of Reference  13 and other semantic properties as aspects of the natural order. If we are to bridge that gap, we need some new concepts. I hope to provide a few such concepts in this book. By providing such concepts and showing that they plausibly have application to possible worlds not entirely unlike our own, I hope to convince you at least that objective representational content really could arise of out of the natural order, with no supernatural or non-​natural intervention or additions. Even if I succeed in convincing you that the semantic could arise out of the natural order, I won’t have told you the details of how it actually does so. But it is not unreasonable to hope that the stock of concepts I develop in this book may ultimately prove to have application not just to nearby possible worlds but to our very own as well. And if that is so, they may also prove to be of use in answering the “how actually” question about objective representational content. But that is an argument for another day. For the nonce, I will be satisfied if the reader gains fuller imaginative acquaintance with the possibility that reference and related semantic properties and relations are a real part of the natural order. She should thereby be left with no lingering temptation to appeal to certain misbegotten doctrines to explain how objective representational content fits into the natural order of things. This last point bears stressing. Precisely because of the distance and lack of systematic coordination between our concepts of material nature, including the material nature of the brain, on the one hand, and our concepts of mind, language, and their contents, on the other, philosophers have sometimes felt less than fully confident about the places of mind and language in the natural order. Such qualms are most loudly and explicitly voiced by self-​professed dualists, like Descartes, who famously thought that “mental substances” and their states were metaphysically distinct from material substances. Nothing with the properties characteristic of thought and thinking, he argued, could possibly be resident in a merely material substance. These days, Cartesian dualism is a distinctly minority position. But even some self-​professed materialists, who have no truck with Cartesian substance dualism, sometimes deny the possibility of illuminating, non-​question-​begging explanations of the objective representational contents of our thoughts. Consider, for example, the family of philosophers we might call the interpretationists, including Quine, Dennett, and Davidson. Each of these philosophers is deeply committed to the metaphysical doctrine of materialism in one way or another. Nothing non-​natural has any place in any of their fundamental ontologies. But each has argued that there is really no

14  Referring to the World independent fact of the matter about the contents of our thoughts and/​or words. Each does allow that there is an interpretive practice of ascribing contents to thoughts and words. They also insist that our interpretive practices are well-​nigh indispensable. Dennett and Davidson, for example, see our interpretive practices as inextricably tied up with our cognition of ourselves as rational cognizers and actors. Rational cognizers are pushed and pulled not merely by the blind forces of material nature but by the force of the better reason. Absent our interpretative practices, we could appeal to only the blind operations of nature in explaining human behavior. If we would cognize human cognition and action as rational, as responsive to the force of the better reason, we must, the interpretationist will say, interpret. That is, we must assign (objective) representational contents to our thoughts and our words. But for all their insistence on the indispensability of interpretations, interpretationists typically maintain that interpretations are infected with a deep and ineliminable observer relativity. Different interpreters may be fully justified in ascribing incompatible mental contents to one and the same subject. And this supposed observer relativity and indeterminacy of all interpretations is supposed to render interpretations if not quite fictional or illusory, nonetheless not quite part of the “objective” furniture of the universe either. There are other self-​professed materialists who, though conceding both the determinacy and objective reality of objective representational content, seem, nonetheless, to regard the objective representational contents of our thoughts as all but inexplicable. John Searle is the perhaps unintended paradigm here. Searle has argued that what is distinctive about thought is that it enjoys what he calls intrinsic, as opposed to derived intentionality. As such, thoughts are supposed to stand in sharp contrast to sentences. Unlike sentences, thoughts do not have their representational content derivatively, in virtue of being associated with some other already contentful things. This claim amounts to an assertion of the priority of thought over language. About this much, Searle seems to me to be entirely correct. But he also holds something stronger and more questionable. He apparently believes that just because thoughts enjoy intrinsic rather than derived intentionality, there can be no illuminating explanation of how objective representational content arises out of that which is not yet intentional. Contra the dualist, however, Searle does not claim that intentionality sits outside natural order. Indeed, on Searle’s view, the brain literally secretes intentionality and consciousness. Moreover, unlike interpretationists like Quine and Davidson, Searle insists

The Mystery of Reference  15 that intentionality is a genuine and determinate causal power, somehow resident in or at least caused by the brain. But it seems to me that Searle’s confidence that he has located intentionality in nature is misplaced. He is insufficiently puzzled about just how the vast vat of neurotransmitters, electrochemical impulses, and synaptic junctions that make up our brains could have the startling property of secreting states with objective representational content. Since Searle is famous for arguing that all other supposedly naturalistic theories of mind—​ save his own—​fail to respect the intrinsic intentionality of the mental, that charge may seem somewhat surprising. But despite Searle’s self-​declared naturalism, his insistence that the mind is intrinsically intentional misleads him into treating the intentionality of thought as if it were a metaphysical surd, one neither subject to nor in need of deeper explanation and explication in other terms.12 Searle’s notion of “intrinsic intentionality” in fact brings to mind an important and philosophically challenging view first articulated by Brentano. Brentano was no materialist. He was, rather, a kind of idealist and also a kind of dualist. (Brentano, 1995/​1874) Brentano’s dualism is shown by the fact that he took the mental to be fundamentally and categorically distinct from the physical. As a consequence, he also took the mental to be irreducible to the physical. In particular, he claimed that the objects of our thoughts are defined by the property of what he called “intentional inexistence”: Every mental phenomenon is characterized by what the Scholastics of the Middle Ages called the intentional (and also mental) inexistence of an object, and what we would call, although not in entirely unambiguous terms, the reference to a content, a direction upon an object (by which we are not to understand a reality . . .) (emphasis added) or an immanent objectivity. Each one includes something as an object within itself, although not always in the same way. In presentation something is presented, in judgment something is affirmed or denied, in love [something is] loved, in hate [something] is hated, in desire something is desired, etc. This intentional inexistence is exclusively characteristic of mental phenomena. (1995/​1874,  68)

12 This is in the main because of his views about the connection between intentionality and consciousness. It is really consciousness, in particular the first-​person point of view, that is the primary metaphysical surd for Searle. Insert quote from Searle and explicate quote

16  Referring to the World The first point to appreciate here is that on Brentano’s view a thought may involve “a direction upon an object,” as he calls it, even when there is no independently real object, no “reality,” to use his term, upon which the thought is directed. Think, for example, of the thought that Santa Claus lives at the North Pole or the thought that Sherlock Holmes is an incredibly good detective. There is no Sherlock Holmes in reality. But that does not prevent us from thinking thoughts as of Sherlock Holmes. In Brentano’s vocabulary, Sherlock Holmes is an “immanent” object—​an object for thought—​but not a reality.13 Now the view that it is possible to have thoughts putatively about “non-​ existent” objects like Sherlock Holmes is pretty widely endorsed—​though there is a great deal of disagreement about just what the contents of any such thought come to. What is striking about Brentano, however, is that he endorses an even stronger view—​and it is this stronger view that amounts to a kind of idealism. For him, it is not just that some of the objects of thought are merely “immanent” and, therefore, “intentionally existent.” He makes the startling claim that every object of thought has a “merely” immanent existence as an object for thought alone and no existence whatsoever “outside” of thought. Indeed, he takes relatedness to a merely immanent—​and thus intentionally inexistent—​object to be of the very essence of thought. Thought and only thought exhibits relatedness to intentionally inexistent objects. No physical relation can have this character, since it is impossible to stand in any physical relation to a non-​existent object. If, for example, I hit the car, then the car surely exists and exists, moreover, independently of my hitting it. Moreover, the fact that the objects of thought are one and all intentionally inexistent, while no object of any physical relation can have this property leads Brentano to conclude that thought is categorically distinct from and irreducible to anything merely physical.14 Now I  don’t profess that it’s completely unproblematic to say just what Brentano’s thesis comes to. Nor do I pretend to offer any deep explication and analysis of that thesis here. However exactly the thesis is spelled out, it seems pretty clearly mistaken. Intuitively, it seems just wrong to say that the objects of thought are one and all intentionally inexistent, one and all merely immanent objects, existing for thought alone and not as self-​standing mind-​ independent existents in their own right. Anybody who elects to defend 13 It is a delicate matter of Brentano scholarship, into which I will not plunge here, to decide exactly what Brentano means by “immanent” objects internal to thought and objects that exist “outside” of at least human thought. Insert something that give some flavor of debates. 14 It is fair to wonder what Brentano thinks entitles him to posit physical objects at all. Elaborate

The Mystery of Reference  17 such a claim must discharge a very heavy burden of proof. Certainly, neither Brentano, nor anybody else, comes close to discharging any such burden. On the other hand, the weaker and less radical thesis that at least some of the objects of thought are “intentionally inexistent” in something like Brentano’s intended sense is not so obviously mistaken. Versions of that thesis have been accepted by philosophers with no obvious affinity for either mind-​body dualism or content idealism. But despite its widespread acceptance, many have often found it deeply challenging to say exactly what thoughts apparently about non-​existent objects come to. Frege—​to take just one prominent example—​grudgingly grants that we can think thoughts apparently about Santa Claus or Pegasus. But he dismissed such thoughts as merely “mock thoughts.” Such thoughts may have a place in fiction—​where we are concerned, he claims, only with the emotions that such thoughts may arouse—​but they have no place at all in science or in logic—​where our concern is with truth and falsity. In a related vein, Russell argued that the very idea of a non-​existent object lands us in contradiction.15 But he held that in fact sentences that superficially appear to require reference to such objects in reality do no such thing at all. Putatively referring expressions that at first blush appear to have the semantic job of standing for some non-​existent object—​expressions like “Santa Claus” or “Pegasus”—​not only don’t have any such function in reality but are really a kind of grammatical illusion. When we subject sentences containing such expressions to rigorous logical analysis, we find that those expressions simply disappear. They do not show up in what we might call the true logical grammar of either external language or inner thought. I suggest that there is something half right and half wrong both about Frege’s and Russell’s squeamishness about thoughts about apparently non-​ existing objects, on the one hand, and about Brentano’s insistence that the objects of thought are one and all intentionally inexistent objects for thought alone, on the other. And I  claim that any adequate account of the objective representational contents of our thought and talk must pay due respect both to the squeamishness of Frege and Russell and to Brentano’s insistence. Contrary to Brentano, however, I shall argue that we have little reason to deny that thought is primarily and paradigmatically directed toward real existents that are mostly independent of the mind rather than at merely “immanent” objects that are objects for thought alone. Indeed, I suspect that the

15

Russell reference and example

18  Referring to the World squeamishness of Frege and Russell about thoughts about putatively non-​ existent objects is partly explained by their tacit grasp of the fact that the primary and paradigmatic objects of thought are real existents. What Brentano correctly grasped, however, albeit through a glass dimly, is that a certain part of business of determining the objective representational content of our thoughts and our words lies entirely on the side of the thinking subject, with no contribution at all from the external world. If we are to pay due respect to both Brentano’s insistence and the squeamishness of Frege and Russell, we need some new distinctions. Most of the needed distinctions will come in subsequent chapters, but the single most crucial one is worth getting on board right away. I distinguish what I call merely objectual representations and what I call fully objective representations. To a first approximation, a fully objective representation is one that actually stands for a real existent or expresses a real property. A merely objectual representation, on the other hand, is one that is “fit” or “ready” for the job of standing for a real existent or expressing a real property, even though that representation may not yet do so. I will sometimes say that objectual representations are referentially fit and that fully objective representations are referentially successful. I will eventually say a great deal more by of way of clarifying and explaining this distinction. I will argue, for example, that the factors that go into making a representation fit for the job of referring to an object lie entirely on the side of the cognizing, thinking subject. The fitness-​making factors turn out to be internal, role-​oriented and syntactic. These claims need substantial further elaboration and defense. Such elaboration and defense will come in subsequent chapters. For the nonce, however, I offer a foretaste of the use to which this distinction, once flushed out, will eventually be put. I use the notion of a merely objectual representation to re-​introduce the question of why our thought and talk should take the form of thought and talk about objects at all. I say “re-​introduce” because although this question once occupied center stage in the theory of objective representational content for philosophers as diverse as Kant, Frege, and Quine, in more recent times, it has largely faded from view.16 One important aim of this book is to shine the klieg lights once again on the form-​content distinction as applied to our thought and talk about objects. That is because it is only by doing so that we can make conceptual progress on the question of how our thoughts



16 Robert Brandom is the clear exception. Brandom quote.

The Mystery of Reference  19 manage to achieve full-​blown semantic contact with mind-​independent objects and their properties. I shall argue that it is in virtue of our having representations that are already referentially fit that our thought and talk take on the form of thought and talk as of objects. The claim is not that the mere playing of inner structural roles is sufficient to render inner representations referentially fit ipso facto suffices to render such expressions referentially successful. There is a gap between fitness and success, and a good theory of reference will need to explain how possibly that gap is bridged. I shall argue that bridging the gap between mere referential fitness and full-​blown referential success is made possible by the influence on mind of certain extra-​representational factors lying largely outside the cognizing subject, on the side of the world itself. In particular, I shall argue that referential success requires that already referentially fit representations bear appropriate causal/​informational connection to outer objects and events. Because it posits two distinct but cooperating factors that work jointly to constitute reference, I  call the view defended in this book two-​factor referentialism.17 According to two-​factor referentialism, both an internal fitness-​making factor and an extra-​representational causal/​informational factor are necessary for successful reference, but neither suffices, on its own, to constitute reference. In the absence of causal connections to objects and events in the world, the fitness-​making factors yield the form of thought as of objects, but they leave our thoughts devoid of semantic contact with any real existents. Absent the internal fitness-​making factors, causal/​informational connections between objects and events in the world, on the one hand, and internal states of our bodies, on the other hand, would yield nothing but semantically inert to’ings and fro’ings. This last point bears brief elaboration. In arguing that reference and semantic answerability more generally are partly constituted out of the relentless causal and informational to’ing and fro’ing in which mind and language are immersed, we immediately face the fact that a plethora of relations obtain between our words or thoughts, on the one hand, and objects, events, and properties in the world, on the other. Our thoughts and words temporally precede or follow upon events in the world in a breathtaking array of patterns.

17 What makes the view a “two-​factor” view is the claim that it takes two factors to constitute reference. What makes the view a version of referentialism, however, has more to do with matters semantic than with the metaphysical nature of reference determination.

20  Referring to the World Our thoughts are sometimes caused by or in turn cause events in the world. This all happens via myriad causal pathways. If reference ultimately rests at least partly on the obtaining of some special causal/​informational relation or other between certain sorts of words and/​or thought constituents and things, we must be able to say which of the many different relations that hold between words or thoughts and objects or events in the world play a role in constituting or determining reference. But this latter is no easy task. It is, in fact, the rock on which every variety of causal theory of reference and objective content has so far run aground. Neither bare relations of causation, nor bare relations of temporal precedence and succession are as such straightforwardly sufficient on their own to explain how our thoughts come to be semantically answerable to the world. A blow to my brain may cause any number of thoughts in my head. But those thoughts would not for that reason stand in any semantic relation to that blow. The caused thoughts would not, that is, be for that reason alone, semantically answerable to how things are by that very blow and its properties. Suppose, for example, that a blow to my head causes me to think that the hamster’s cage needs cleaning. Though the blow plays a decisive causal role in bringing about my thought, it clearly doesn’t follow that my thought is in any way about that blow to the head. How things are by the blow has no bearing whatsoever on whether my thought is true or false. Similarly consider my utterance of “Though Socrates was a great philosopher, he would never have gotten tenure at Stanford, since he never wrote a thing.” In making this utterance I have evidently referred to Socrates and have said something plausibly true about him. Yet, I have no direct causal commerce with Socrates. Indeed, my tokening of the name “Socrates” is causally connected to past Socrates-​involving events only by lengthy and indirect informational and/​or causal chains, stretching far back in history to a time when Socrates still lived. Though causal theorists are wont to think that somehow or other the presence of such connecting causal chains secures a referent for my use of the name “Socrates,” a plethora of events must of necessity intervene between my tokening of “Socrates” here and now and any long since expired Socrates-​involving events.18 There are, for example, 18 Explain that the (epistemic) distance between Socrates and my use of “Socrates” is the sort of thing that led Russell to deny that we can refer to Socrates and that “Socrates” is a name. [Also explain the distinction between knowledge by acquaintance/​knowledge by description and foreshadow my later discussion of the role of “knowledge-​wh” in making reference possible.]

The Mystery of Reference  21 all the events involved in the writing and production of the books in which the deeds of Socrates are recorded. There is a rich oral history, passed down through the ages from philosophy professor to philosophy student. Socrates himself is involved in none of those events, though many other objects are. Despite the fact that my tokening of “Socrates” is much more intimately causally connected to events that are not Socrates-​involving than it is to any Socrates-​involving event, we want to say that it is Socrates, and no else, to whom I refer with my use of the name “Socrates.” If reference to Socrates is partly constituted out of causal/​informational relations to Socrates then there must be some distinguished causal/​informational relation that my use of “Socrates” bears to Socrates and to nothing else. I shall argue that causation does make a contribution to the constitution of reference, but only a contribution of what might be called an incremental kind. Not just anything can be made to refer to an object by standing in a causal relation to it. Set up whatever causal connection you like between the states of a window and the states of a brick. I  claim that unless you have thereby made the brick function as a representation of the right sort, implementing that causal connection will not suffice to turn the window into the referent of the brick. Reference is not merely a purely causal affair, in other words; it is also a distinctively mental cum linguistic affair. Only representations in a mind or in a language—​and representations of a distinctive sort—​can be made to refer to an object by bearing appropriate causal connections to it. This fact points the way toward a strategy for explaining how causation or any other extra-​representational factor can enter into the constitution of reference and objective representation content. The question we should put to ourselves is this: What is a representation such that being causally related to objects renders it semantically answerable to how things are by those very objects? I shall argue that it is only when representations are already objectual that the addition of extra-​representational causal factors incrementally suffices for the full objectivity of our representations. The distinction between objectual and objective representations will also play a crucial role in allowing us to give due deference both to Brentano’s insistence that the objects of thought are one and all intentionally inexistent and to the squeamishness of Frege and Russell about countenancing thoughts about such objects. The truth that Brentano dimly grasps is that something lying entirely on the side of thought itself renders our representations objectual and thereby prepares the way for those representations to make semantic contact with real existents. But he failed to appreciate that no object

22  Referring to the World at all—​not even an inexistent one—​is yet given through the mere objectuality of our representations. The squeamishness of Frege and Russell, on the other hand, reflects their grasp of the fact that it is only when our representations achieve full-​blown objectivity that they are rendered semantically answerable to anything at all. If the having of objective representational content is of the very essence of thought, then Frege and Russell can be seen as doubting that whatever is devoid of full-​blown objective representational content is not yet thought in the fullest sense. That, I  think, is precisely what drove Frege to dismiss thoughts of non-​existent objects as merely “mock” thoughts. Though there is something right about this dismissal, it is also true that what Frege dismissed as merely mock thoughts occupy a much grander place in our cognitive lives than he imagined. Not all of our thoughts are about an external and mind-​independent world. Much of our thought and talk is about the contents of the mind itself (and about that which is dependent on the mind). If a theory of the objective representational contents of our thoughts and words is to be complete, it will also have to show how our thought and talk can be objects for themselves. Now language has an amazing power to open up the contents of our minds to another in a detailed and articulated fashion. Peruse one’s favorite work of philosophy, science, or mathematics, and you will be able to infer from the words on the page a great deal about the contents of the author’s mind: what she believes, what she intends to prove, what she takes to be an argument for what. Just imagine trying to infer those very same thought contents without the aid of language. It almost certainly cannot be done. I do not mean to suggest that absent the aid of the spoken or written word our minds are entirely impenetrable to one another. A wealth of data on the development of the theory of mind in early childhood demonstrate that the growing child has a rapidly unfolding, perhaps even innate, understanding of at least the general structure of a human mind.19 The developing child comes to know early on that minds contain beliefs, desires, intentions, perceptions, and emotions—​to name just a few of the denizens of our mental lives. Moreover, she comes to know fairly early on how the denizens of mental life relate both to one another and to the world that she shares with those who surround her. This rapidly unfolding knowledge of mind seems to predate language and even to aid in its acquisition. It also plays an indispensable role



19 References

The Mystery of Reference  23 in the developing child’s mastery of the instruments of social life and shared cognition.20 Granting that the minds of others are never impenetrable black boxes to us does not entirely dissolve our puzzle, however. After all, the developing child is surrounded by others who already share and deploy the instruments of social life—​including a shared language. This makes the child’s task of acquiring mastery of those instruments considerably easier than it otherwise might be. But our problem is not how we might acquire the instruments of social life from others who already possess them. Our problem is to explain how possibly there could be such instruments in the first place. To appreciate the depth of our problem, it will help to imagine agents who must freshly constitute a language together in the absence of any already constituted instruments of social life. Such agents may indeed already know, like the developing child, the general structure of a human mind and may already know something about how the denizens of such minds relate to one another and to the world. They may even possess an innate universal grammar of the sort posited by some linguists. What they cannot know are the articulated contents of one another’s episodic mental states—​at least not in anything like the degree of detail that language alone possesses the power to reveal. Still, it must somehow be possible for agents thusly related to constitute a language together and for that language, once constituted, to function as a shared window on the articulated contents of the mind. Saying how such a thing is possible is no easy matter. Nor do I profess to offer here anything like a complete story. What I shall argue is that part of the story about how we constitute language together as a shared window on the articulated contents of our private mental states turns on the antecedent semantically answerability of our thoughts to a shared, external world. That is, I shall argue for the perhaps surprising conclusion that part of what enables us to talk and think in such an articulated manner about the contents of our own and others’ minds is the fact that our inner representations are semantically bound down to outer objects. We cognize inner mental contents by cognizing those contents in relation to the external world to which they are paradigmatically semantically answerable. The mind opens up its inner contents to view by opening itself up to a largely shared, largely mind-​ independent external world. Indeed, I shall show that we describe and refer to our inner thoughts largely by redeploying devices we already deploy in

20

References

24  Referring to the World describing and referring to the external world. Once we have the power to say or to think, for example, that the cat is on the mat, we are already a long way toward having the power to talk or think about the contents of a mind, to describe a mind as believing that or fearing that the cat is on the mat. To say this is not to deny the distinction between the inner world of thought and the outer world to which thought and language are answerable. It is only to insist that our ability to cognize thought and its content is, in a way, parasitic upon our ability to cognize the world and to cognize ourselves and our inner representations in relation to that outer world. This priority of our cognitive hold on the outer world over our cognitive hold on mental contents is reflected, I shall argue, in the relative priority of what philosophers typically call de re ascriptions of mental contents over so-​called de dicto ascriptions of mental content. But I shall leave the further explications of these notions to subsequent chapters.

2 Inner Fitness and Outer Cause The Two Factors of Content

1. Preliminaries In this chapter I begin to develop and defend one of the central foundational claims of this book. I argue that two distinct and independent factors contribute to the determination of the objective representational contents of our thought and talk. One factor is extra-​representational, causal, and informational. It is rooted largely outside of the subject, on the side of the world. The other factor is internal and structural. It lies entirely on the side of the subject. Both of these factors are necessary for objective representational content, but neither suffices on its own to determine content. Absent the external causal factor, we could have the inner form of thought as of objects, but no actually existent objects would thereby be made available to thought. Absent the internal factor, there could still be a relentless inward rush of energy upon the portals of sensation, but the inward rush would leave no outwardly oriented semantic footprints upon the shores of consciousness and thought. The plan of the current chapter is as follows. Section 2 gives motivational force to a certain initial question. Just how does the mind manage, on the basis of the inward rush of mere energy upon the portals of sensation, to represent to itself a world populated with a dizzying variety of putatively mind-​independent objects? In section 3, I distinguish what I call “semantic referentialism” from what I  call “semantic presentationalism.” Semantic presentationalism is the view that the semantic contents of our thoughts and/​ or the meanings of various expressions are determined by presentational properties. Semantic referentialism is the view that the objects themselves may somehow directly constitute thought contents, without the mediation of presentational properties. Though I ultimately defend a modified version of referentialism—​what I call “two-​factor referentialism”—​the main aim of section 3 is less to refute presentational theories of content than to undermine one of the main motivations for such approaches. Semantic presentationalists Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0002

26  Referring to the World typically motivate their view by appeal to the argument from the epistemic one-​sidedness of all reference. Beginning here and extending over the course of the next several chapters, I  argue that the epistemic one-​sidedness of reference is best understood as a phenomenon having to do with the form and structure of thought rather than with the contents of our thoughts. If that is right, epistemic one-​sidedness is entirely consistent with semantic referentialism—​at least of the two-​factor variety defended in this book. In section 4, I return to the notion of a merely objectual representation, which was introduced briefly in Chapter  1. Here my aim is to begin to lay the groundwork for the argument of Chapter 4, where I shall argue that names, which I take to be paradigmatic referring expressions, are what I call devices of explicit co-​reference and that this is a matter of inner syntactic role and structure. I also outline the Fregean and Kantian roots of my distinction between objectual and fully objective representations. Though my explanation of the source and nature of this distinction differs considerably from either of theirs, it is important to see both what they each got right and what they each got wrong.

2.  The World Rushes In The external world influences the inner contents of our thoughts primarily through its relentless assault upon the portals of sensation. We open our eyes, unstop our ears, breathe in through our nostrils, reach out with our hands, and a torrent of energy comes rushing in. In response to this inward rush of energy, episodes of perception and thought are stirred up in the mind, as if unbidden. Strikingly, the perceptual and thought episodes stirred up in the mind by the inward rush are present to the mind whose perceptions and thoughts are purporting to semantically answer to mind-​ independent objects and events somehow revealed to the mind via the inward rush itself. In Chapter 1, we began to contemplate just how initially puzzling it is that this relentless inward rush of energy gives way to outwardly oriented representational contents. Every molecule and cell in our bodies is subjected to a vast array of external influences. Yet while most of these assaults from without may leave causal traces, they apparently leave no semantic footprints. The puzzle is only deepened by the fact that our thought and talk are sometimes apparently semantically answerable to objects with which we ourselves have had no direct causal or cognitive commerce.

Inner Fitness and Outer Cause  27 Neither through the present testimony of the senses nor through memory has any one currently alive ever had perceptual or cognitive contact with energy emanating directly from any Socrates-​involving event. Nonetheless, by the modest effort it takes merely to competently deploy the name “Socrates” in inner thought or outer talk, we manage to direct our thought and talk toward that long dead person. This can seem like a nearly magical semantic accomplishment.1 A natural thought is that perhaps we can manage to dispel the mystery to a degree by taking a closer look at exactly what rides inward on the energy wave. At first blush, the inward rush may seem rather modest, at least in comparison to that in which it ultimately eventuates. But perhaps it is best to work backward from a consideration of what ultimately eventuates as a result of the inward rush toward the receptive surfaces of the mind. Eventually, the inward rush somehow unleashes a torrent of inner representations that come in a dizzying variety—​representations of color, size, shape, number, location, temperature, change, persistence, cause, effect, and on and on. So perhaps what is present in this torrent of outer representations is there from the beginning, riding inward toward the waiting shoreline of consciousness and thought. Something like this thought—​that whatever is ultimately present in the mind, must have somehow originated in the senses—​is, I think, the foundational thought behind various forms of empiricism. But time and time again, empiricism has proven that it cannot be the entire story about the mind and its objective representational contents. The failures of empiricism suggest that the unleashing of the torrent of outwardly oriented representations must also depend on certain distinctive inner powers, capacities, and even contents of the mind-​brain. As I have already suggested, the flow of energy across spacetime is ubiquitous, but it eventuates in outwardly oriented representations only in those peculiar regions of spacetime inhabited by minds. Whatever turns out to be the correct story about what is already there, surfing inward on the energy wave, prior to its encounter with mind, there must also be a story about the inner psychological powers, capacities, or contents by which our minds manage to either produce or extract outwardly oriented representations in response to the inward rush. One may reasonably hope that eventually we will have a complete step-​by-​step account of the intricate details of how actually the inward rush calls forth

1 Reference to Dummett and his views about reference to the past.

28  Referring to the World and engages the representational powers and capacities of the mind-​brain. Our aim, here, of course is not to answer “how actually” questions, but “how possibly” ones. We want to know how possibly inner representations might be arrayed, both in their inner structural relations to one another and in their outward causal/​informational relations to the inward rush, such that through the mind’s contact with the inward rush, inner representations achieve outward semantic orientation in the form of states that enjoy objective representational contents. There are philosophers of various stripes who reject talk of inner representations as little more than mystery mongering. I  am not one of them. But I  admit to having no knock-​down argument to offer against steadfast anti-​representationalism. Yet based on the state of our current best albeit incomplete scientific understanding of the mind, it does seem to me overwhelmingly likely that the mind actually and literally contains a field of outwardly oriented inner representations and that many of these representations are semantically answerable to how things are by objects entirely external to the mind. In the end, the question of whether the mind is well construed as a field of outwardly oriented inner representations is a broadly empirical question rather than a purely philosophical one. As such, I do not seek to decisively settle the question here. What I seek to achieve is fuller imaginative acquaintance with a certain possibility—​the possibility that the mind contains a field of inner representations, such that partly in virtue of their inner structural relations to one another and partly in virtue of their causal relations to the inward rush, they are outwardly semantically oriented toward an external world of objects and their properties independent of the mind. My central claim is Kantian in spirit, though it is not Kantian in detail. The guiding idea is that in order that its inner representations achieve outward semantic orientation, the mind must come already internally and antecedently poised or ready to refer. The mind must, that is, be antecedently ready to transform the inward rush into objective representational contents for its inner representations. Or to put the point in more Kantian terms, an antecedent inner readiness to refer is, I claim, a condition of the very possibility of our thoughts achieving objective representational contents. If the mind lacked antecedent inner readiness to transform the inward rush into outwardly oriented semantic contents, the world’s assault upon the portals of sensation would be as semantically inert as are the exertions of gravity on our physical mass or the invasion of our cells by bacteria and viruses.

Inner Fitness and Outer Cause  29 I argue not just that the mind must have a certain inner readiness to refer and that this readiness must take some form or other. I shall also offer a conjecture about what such readiness plausibly consists in. I will try to make a case that my conjecture is plausibly true of nearby possible worlds not entirely unlike our own. Admittedly, that will not suffice to prove that the world we actually inhabit is one in which the requisite inner readiness is secured in just the way that I conjecture. But if I can establish that the conjecture plausibly holds good in possible worlds not entirely unlike our own, I will have done enough to discharge the burden of showing how possibly the mind and the world might conspire to jointly constitute objective representational content. Now even if we manage to understand with what powers and capacities the mind could possibly be furnished that would allow it to transform the inward rush of energy into objective representational contents, we will not ipso facto have explained how possibly our thoughts and words can be semantically answerable to objects with which an individual cognizer has no direct perceptual, cognitive, or causal contact. There will still remain the question of how possibly we manage the feat of thinking and speaking about, say, Socrates, when no one currently alive has had or will have any direct perceptual, cognitive, or causal contact with Socrates. Once we have told the story of how possibly the mind transforms the inward rush of energy upon the portals of sensation into mental states with objective representational content, we will still have need of a story about the cognitive alchemy, whatever it is, by which a mind is able to represent to itself objects with which it has no direct cognitive, perceptual, or causal contact. One deeply influential answer appeals to reference borrowing (Kripke) and the division of linguistic/​cognitive labor (Putnam). I, here and now, situated at great cognitive and causal distance from Socrates, nonetheless have the power to refer in thought and talk to Socrates because I “borrow” the reference of the word “Socrates” from others. I use the word “Socrates” with the intention of referring to the very same person, whoever he was, that others have referred to before with their use of that very name. Of course, those from whom I borrow the reference of “Socrates” are likely to have borrowed it from others, who will have borrowed it from still others, and on and on in a long chain of reference borrowing. Chains of borrowed reference must eventually come to an end in original reference-​determining encounters, presumably themselves causal, with Socrates-​involving events. But it is, nonetheless, an important insight that at least some of our referential abilities depend on our relations to others. Taken

30  Referring to the World in its totality, that is to say, referring to the world is not solely the work of individual minds, taken one-​by-​one.2 Though this is a lesson well-​learned from Kripke and others, the phenomenon of borrowed reference has not, on my view, been either adequately problematized or adequately theorized, especially not by those of a referentialist cast of mind.3 The capacity for borrowed reference is but one manifestation of an even more fundamental capacity—​ the capacity for what I call discursive or normative community.4 Cognizing agents stand in normative or discursive community with one another when they are connected by a set of mutually endorsed institutions, practices, and norms that reciprocally shape and govern their thought, talk, and action. Standing in discursive community is a way of engaging in a certain kind of mutually reinforcing rational self-​management of our shared cognition and connation. Discursive community has certain semantic benefits. By standing in discursive community with other referring minds, the semantic reach of our thought and talk may extend to objects and properties that we ourselves have never directly perceived or encountered in our own experience. There is a great deal more to say about exactly how discursive communities are formed and sustained. I do not pretend to offer a full-​blown account of such matters in this book. 5 But I will have occasion to return to this notion at several points in this book. It will play a major role, for example, in Chapter 3 in connection with my treatment of what I call jazz combo theories of meaning and reference. Strikingly, those who tend to give explanatory priority to discursive community over the role of the individual referring mind in the constitution of objective representational content tend to downplay if not entirely dismiss the importance of reference. This, I think, is a mistake, as I will try to show in the next chapter. There I will argue that while we must acknowledge the role of discursive community in the achievement of objective representational content, there is still a central role for reference, and also for the representational innards of the referring mind in constituting objective

2 Reference to Frank Jackson. Reference to Chris Peacocke. 3 By his own admission, Kripke himself offers only a rough “picture” rather than a theory of how reference in general and reference-​borrowing more particularly works. For example, at a crucial point, he appeals to the role of referential intentions of a certain sort—​intentions, roughly, to co-​refer by one’s use of a name, say, to what others refer to by that name—​to explain how reference to an object can be secured for an expression in the absence of any direct cognitive contact on the part of the cognizer with the relevant object. But he never pretends to explain just what such intentions consists in or why and how they suffice to determine reference. 4 Reference to Brandom on discursive community (does he actually use that phrase?) 5 See Taylor (2015) for a broader discussion of normative community.

Inner Fitness and Outer Cause  31 representational content out of the inward rush of energy upon the portals of sensation.

3.  Semantic Presentationalism and the Epistemic One-​Sidedness of Reference We begin our inquiry into the nature of the mind’s antecedent inner readiness to transform the inward rush of mere energy into outwardly oriented semantic contents by discussing one phenomenon and two competing families of views about the significance of that phenomenon. The phenomenon is what I  call the epistemic one-​sidedness of reference. The two competing families of views are semantic presentationalism and semantic referentialism. Semantic presentationalism is the view that the semantic contents of various thought constituents and linguistic expressions are determined by what I call presentational properties. Semantic referentialism is the view that both objects and worldly or non-​presentational properties themselves may, in a sense, “directly” constitute semantic content. To a first approximation, think of presentational properties as properties manifested in “presentations” as of objects or properties to either thought or perception. The thought is that presentational properties are what make both worldly objects and worldly properties first available as thinkable or perceivable at all. Presentational properties are more immediately and directly grasped by the mind in its acts of thinking or perceiving than are the worldly objects and properties that are ultimately presented to the mind via those presentations. Consider a chair and a perception thereof. There is the chair itself, with its (worldly) properties of being a certain determinate size and shape. But there is also the way the chair is presented to or in perception. The presentation as of a chair in or to perception is a thing distinct from the chair itself. But we must also distinguish among different presentations of the very same chair. When a perceiver views a chair from underneath, in bad lighting, with its upward facing surfaces occluded from view, the chair is presented in one way, that is, via one presentation. When the very same chair is viewed under good lighting, from above, with its downward facing surfaces occluded from view, it is presented in a different way, that is, via a different presentation. In the two different perceptual episodes, we have the very same chair, with the very same worldly properties, presented again, but in different ways, that is, via two different clusters of presentational properties and thus

32  Referring to the World via two distinct presentations. The crucial further thought is that the representational content of a perception is determined by the way in which the chair is presented, that is, by presentation via which the chair is presented, and not directly by the presented chair or its worldly properties alone. Two presentations that are presentations of the very same chair, with the very same worldly properties but in which the chair is presented via different presentational properties, have different perceptual contents. The presentationalist need not be an idealist. She need not deny that we perceive both the chair and its worldly properties. It is just that she insists on distinguishing what we perceive from the way we perceive what we perceive. Her point is that the representational content of our perception depends not so much on the what but on the how. Perceiving the chair and its worldly properties requires that we be perceptually presented with the chair in some way or other, that is, via some presentational properties or other. The presentational properties via which we perceive the chair on this or that occasion may vary from presentation to presentation, even while the chair itself remains the self-​same object again, with the self-​same worldly properties again. It is crucial that there can be no perception of the chair that does not involve a perceptual presentation of it in some way or other, via some presentational properties or other. Although one may perceive the chair itself, one can only do so indirectly, via the mediation of such presentational properties as are manifest in a given perceptual presentation of it. At the same time, it is important to stress that this approach allows the presentationalist to countenance two different but related levels of content—​ what we might call “narrow” or presentational content and what we might call “wide” or worldly content. On this “two dimensional” view of content, two perceptual states can be said to have the same worldly content just in case they are related in the right way to the same objects with the same worldly properties. Two states with the same worldly content will, however, differ in presentational content even if the same object, with the same properties, is presented again, but in different ways, via different presentational properties. Presentationalists tend to insist that narrow presentational content, as we might call it, is in some sense primary, while wide or worldly content is secondary or derivative. Giving primacy to narrow presentational contents over worldly wide contents can be understood as a way of capturing Brentano’s insistence that the mental is characterized by the property of intentional inexistence.

Inner Fitness and Outer Cause  33 What goes for presentational views of perception goes also for presentational views of cognition. The semantic presentationalist will hold that just as it is via presentational properties manifest in perceptual presentations that worldly objects and their properties are made available to perception, so it is via presentational properties manifest in cognition—​via what we might call cognitive presentations—​that worldly objects and their properties are made available to thought. Consider the much discussed morning-​star evening puzzle, first brought to philosophical prominence by Frege, perhaps the founding father of modern presentationalism. It was a discovery of some moment, Frege claims, that the evening star and the morning star are one and the same object. Discoveries of this are, he claims, often among the most important in science: There is no doubt that the first and most important discoveries in science are often a matter of recognizing something as the same again. However self-​ evident it may seem to us that it is the same sun which went down yesterday and rose today, and however insignificant therefore this discovery may seem to us, it is certainly one of the most important in astronomy and perhaps the one that really laid the foundations of the science. It was also important to recognize that the morning star was the same as the evening star, that three times five is the same as five times three. (Frege, 1979, 142)

But exactly what does one learn when one learns, for example, that the evening star is identical with the morning star? At first blush, a true statement of identity may seem merely to assert the identity of an object with itself. But to be told of an object that it is identical with itself would seem to add little if anything to our knowledge. Imagine our amusement at someone who professed to inform us that he had discovered that Donald J. Trump is, after all, the very same person as Donald J. Trump. Nonetheless, as Frege points out in the passage above, many identity statements seem both non-​ trivial and highly informative. How do we explain the difference between trivial and informative statements of identity? Indeed, how are informative statements even possible? Frege himself argued powerfully that the key to answering such questions lies in distinguishing between sense and reference. On his view, no worldly object is ever immediately and directly available to thought. Our cognitive hold on worldly objects is always mediated by our grasp of sense—​where a sense is or contains a one-​sided “mode of presentation,” as Frege called it,

34  Referring to the World of an object. Among other things, a mode of presentation is supposed to provide a criterion of identification of a reference. We might suppose, for example, that the sense of the expression “the morning star” contains something like the following criterion of identification:6 o is the morning star just in case o is the last celestial object visible in the morning sky just before sunrise.

Call this criterion M. Frege holds that senses are the sorts of things that can be grasped or understood, presumably directly, by human cognizers. He further maintains that the grasp of a sense supports “one-​sided” recognitional abilities. One who grasps M will thereby gain the ability to recognize the morning star as the same object again when it reappears in the early morning sky, just after sunrise. It is sometimes easy to determine whether an object satisfying a criterion like M is present. Anyone who surveys the early morning sky, at the right time of day, probably can, if the sky is clear, determine whether an object satisfying M is present. There are also circumstances in which, although the morning star is present, one could not easily determine whether an object satisfying criterion M is present. Most of us probably cannot tell by mere inspection of the evening sky, which, if any, of the objects present at the time of inspection will be the last to be visible just before sunrise. But that implies that in such circumstances we cannot readily re-​identify the morning star as the morning star again simply by directly applying criterion M. It is limitations of this sort on our recognitional abilities that show the one-​sidedness of our recognitional abilities. Grasping a sense gives us cognitive grasp on an object by giving us a cognitive hold only on one facet or face of the object without thereby also giving us hold on alternative faces of the very same object. Just as the sense of “the morning star” involves a criterion of identification of a reference, so also does the sense of “the evening star.” We might suppose that the sense of “the evening star” is given by something like the following criterion of identification:

6 I  say to a first approximation because M is arguably constructed out of worldly rather than presentational constituent properties. Just as worldly objects are grasped via presentational properties so to, on Frege’s view, worldly properties must be grasped via modes of presentations of those properties.

Inner Fitness and Outer Cause  35 o is the evening star just in case o is the first celestial object visible in the evening sky just after sunset.

Call this criterion E. One who grasps criterion E will also have a one-​sided recognitional ability. If one is able to re-​identity an object as the object that satisfies criterion E, then one is ipso facto able to re-​identify that object as the evening star again. Once again, there will be many situations in which the evening star is present in which a cognizer is unable to determine whether the satisfier of criterion E is present. Of course, the recognitional ability supported by a grasp of criterion E and the recognitional ability supported by the grasp of criterion M are abilities to recognize the very same worldly object, with the very same worldly properties. Yet it can be a discovery that the object recognized via the application of E is the same again as the object recognized via the application of M. But this means that it can also be a discovery that the evening star is one and the same object as the morning star. That is because though both abilities are abilities to recognize the very same object, they give us cognitive access to different faces or aspects of that object, via different cognitive presentations of that object. Now Frege famously took thoughts to be abstract particulars in which senses occur as constituent parts. Though thoughts are abstract rather than concrete, on his view, they are also in some sense “composed” of senses. The abstractness of thoughts has to do with the fact that on Frege’s view thoughts are not located in the heads of individual thinkers. They inhabit a third realm, a realm distinct from the concrete physical world to which our senses give us access and distinct from the inner psychological realm of mere “ideas” or inner representations. For that reason, a thought cannot be identified with an act of thinking, on Frege’s view. Unlike an act of thinking, a thought as such has no “owner.” An act of thinking occurs in the head of a particular individual at a particular time. When the act has entirely elapsed that very act can never occur again. But many thinkers can “grasp” the same thought simultaneously. Moreover, a given thinker may think the same thought today that she thought before. This is not to deny that a thinker does the grasping of a thought via an inner act of thinking. But the thought grasped is one thing, while the inner act of grasping the thought is something entirely different, at least according to Frege. Now I do not intend to explore the alluring but enigmatic depths of Frege’s views about the metaphysical nature of thought here. My concern is mainly with Frege’s claim that senses contain “modes of presentation” and his further

36  Referring to the World claim that senses, rather than the referents they possibly determine, are the basic constituents out of which thoughts are in some sense composed. What interests me about Frege’s approach is the fact that in carefully distinguishing the cognitive presentation that is, on his view, a constituent of thought from the worldly objects and properties presented via those thought constituents and thereby giving center stage at the level of thought content to the presentation rather than to the presented, Frege inaugurates the urform of semantic presentationalism about thought and its contents.7 Like Frege, Russell also held that ordinary middle-​sized objects, like Socrates, are thinkable only via one-​sided presentations. Unlike Frege, however, Russell believed that some objects are directly available to thought, without the mediation of a one-​sided presentation. In effect, Russell hewed simultaneously to both a limited form of presentationalism about certain objects of our thought and talk and a limited form of referentialism about other objects of our thought and talk. What made it possible for Russell to simultaneously embrace both presentationalism and referentialism, though with respect to different classes of objects, was his distinction between knowledge by acquaintance and knowledge by description. We are directly acquainted with an object, according to Russell, if we cognize it in a direct, inferentially unmediated way. If a cognizer is directly acquainted with an object and encounters that same object again, she is ipso facto cognitively equipped to recognize that she has encountered the same object again. And she is equipped to do so without the aid of inference. A thinker presented twice over with an object with which she is directly acquainted cannot intelligibly wonder, as one does in the morning star/​evening star case, whether the “first” object is the same again as the “second” object. When we are directly acquainted with an object, we have made direct, unmediated contact with the object itself and our cognition of it is not mediated by a one-​sided presentation of the object. Or so Russell imagined. Russell took there to be precious few objects with which we are directly acquainted, however. He held that only absolute simples could be the objects 7 In the context of arguing that the truth value of a complete thought should be assimilated to reference rather than to sense, Frege says: The relation of the thought to the True may not be compared with that of subject and predicate. Subject and predicate understood in the logical sense are indeed elements of thought; they stand on the same level of knowledge. By combining subject to predicate, one reaches only a thought, never passes from sense to reference, never from a thought to its truth value. One moves at the same level, but never advances to the next. A truth value cannot be part of a thought, any more than, say, the sun can, for it is not a sense but an object.

Inner Fitness and Outer Cause  37 of direct acquaintance. And among the absolute simples, he counted only sense data, universals, and one’s own inner states—​though he took it to be an open question whether one could be directly acquainted with an enduring self. Decidedly not among the absolute simples, however, were complex middle-​ sized objects like Socrates. About our thoughts about such objects, Russell was a thoroughgoing presentationalist. This may suggest a certain deep affinity to Frege’s views. And it was this affinity that lead Kripke, for example, to talk of what he sometimes called the “Frege-​Russell” theory of reference. Still, it must be stressed that Russell’s partial presentationalism is bound up with certain non-​presentationalist doctrines. Frege’s presentationalism was a thesis not just about the nature of some thought contents. It was a sweeping general thesis about how semantic answerability is first achieved. But for Russell, one-​sided presentational contents would seem to play no foundational role in first securing semantic answerability. Rather, for Russell, semantic answerability is ultimately grounded in our capacity to refer directly to objects. And direct reference is not grounded in one-​sided presentations. Here the distinction between knowledge by description and knowledge by acquaintance turns out to be absolutely crucial. Russell holds that we can (directly) refer only to those objects with which we are directly acquainted. Among other things, this approach implies that there can be no direct reference to middle-​sized objects like Socrates with which we are never directly acquainted. But in denying that we can directly refer to Socrates, Russell did not intend to banish Socrates from the realm of the thinkable. Though we cannot think or talk directly about Socrates, since we are not acquainted with him, we can still think and talk about him in a sense indirectly. Again, this is part of the payoff from distinguishing knowledge by acquaintance from knowledge by description. Socrates may have been immediately present, and therefore immediately thinkable, only to himself. To others, he is present only as a logical complex or construction. Now logical complexes, on Russell’s view, must be built out of logically simple constituents. Though we can be acquainted with the logically simple constituents of a logical complex, we know complexes only by description. As such, though we cannot (directly) “refer” to such complexes, we can (indirectly) “denote” them. Talk of denoting rather than referring may seem at first a mere quibble. But in Russell’s hands it is directly tied to the difference between merely presentational contents and purely referential contents. Russell’s full spectrum of views about these matters is both subtle and complex. We need not explore it fully for our present purposes. It suffices to notice that on Russell’s view a

38  Referring to the World statement or thought to the effect that Socrates died at the hands of the Greeks is not a statement in simple subject-​predicate form, directly predicating some property of some free-​standing, logically simple object. A statement to the effect that Socrates died at the hands of the Greeks is more perspicuously put by replacing the name “Socrates” with a “definite description” like “the teacher of Plato and husband of Xanthippe” or “the Athenian philosopher accused of corrupting the young” or, perhaps, “the philosopher who figured most prominently in the dialogues of Plato.” On Russell’s view, the sentence “Socrates was put to death at the hands of the Athenians” is really a shorthand way of putting one or other of the following propositions: (1) The teacher of Plato and husband of Xanthippe was put to death at the hands of the Athenians. (2) The philosopher accused of corrupting the young was put to death at the hands of the Athenians (3) The philosopher who figured most prominently in the dialogues of Plato was put to death at the hands of the Athenians. Statement (1) makes a claim about a structure of properties rather than a claim about a particular really existent individual. It says, in effect, that the property of being a teacher of Plato and husband of Xanthippe, on the one hand, and the property of being put to death at the hands of the Athenians, on the other, have exactly one common instance. Though Socrates may in fact be the common instance of these two properties, (1) could still be true even if either Socrates never existed or though he did exist, he never taught Plato or married Xanthippe. One way to appreciate that Socrates is not directly semantically implicated in the content of (1) is to note that (1) can be rephrased as (4) below: (4) The teacher of Plato and husband of Xanthippe, whoever he was, was put to death at the hands of the Athenians. As phrased, (4) is explicitly non-​committal about the identity of the teacher of Plato and husband of Xanthippe. Notice, moreover, that although Socrates was both accused of corrupting the young and figured prominently in the dialogues of Plato, (2)  and (3)  make different assertions about a different structure of properties from (1). Statement (2) says, in effect, that the property of being accused of corrupting the young and the property of being put

Inner Fitness and Outer Cause  39 to death at the hands of the Athenians have exactly one instance in common. Statement (3)  says that the property of being a philosopher who figures prominently in the dialogues of Plato and the property of being put to death at the hands of the Athenians have exactly one instance in common. For all the contents of (2) and (3) require in order to be true, the person that makes (2) true could be distinct from the person that makes (3) true. And this helps us to see another sense in which none of (1) –​(4) can be said to be directly about Socrates. Socrates, the individual, is not, as such, the direct subject matter of any of (1) –​ (4). Now on Russell’s view genuinely referring expressions, or logically proper names, as he also thought of them, work differently from definite descriptions and other so-​called denoting complexes.8 Denoting complexes never have the semantic job of merely standing for some definite object, not even when, as in the case of some singular definite descriptions, there is one and only one object that happens to satisfy the relevant description. By contrast, according to Russell, the sole semantic role of a genuinely referring expression is precisely to stand for an object and to do so Punkt, without introducing any descriptive content into the proposition expressed by the sentence in which the relevant expression occurs. The supposed semantic distinction between genuinely referring expressions and denoting complexes is sometimes put by saying that sentences containing genuinely referring expressions express so-​called singular or object-​dependent propositions, while sentences containing definite descriptions or other denoting complexes, but no genuinely referring expressions, express non-​singular, general, or object-​independent propositions. A singular or object-​dependent proposition is a proposition the very existence of which depends, and depends essentially, on the existence of a certain object. If p is an object-​dependent proposition, then there is a certain object o such that p would not exist at all, and thus would not be thinkable or expressible if o failed to exist. An example may help. Suppose, for example, that there is not and has never been any such person as Donald J.  Trump. Consider an utterance of the sentence “Donald J. Trump is currently the president of the United States” in our imagined scenario. Exactly what proposition would such an utterance express? Clearly, there is no particular person that such an 8 Among the class of denoting complexes, Russell counts not only definite descriptions like “the teacher of Plato who was put to death by the Athenians,” but also quantifier expressions such as “some man,” “every donkey,” and “a stranger on a train.”

40  Referring to the World utterance could be said to be about. Nor does there seem to be any particular cluster of properties that such a sentence expresses a claim about. Indeed, it seems tempting to say that in the imagined circumstances the sentence “Donald J.  Trump is currently the president of the United States” would express no fully determinate proposition at all. If you share this intuition, then you are in effect having the intuition that as things currently stand, with Donald J. Trump being an actually existent object, the proposition that Donald J. Trump is currently the president of the United States is an object-​ dependent proposition. Contrast so-​called object-​dependent propositions with what are variously called non-​singular, general or object-​independent propositions. The identity of a general proposition is not tied to the existence of any particular object. What I express when I say that Smith met a man, for example, does not depend for its truth on Smith standing in the relation of having met x to any particular man. Its truth or falsity does not depend on whether Smith met Black or Brown or Jones in particular but only on whether Smith met some man or other. Even if there were no men at all in existence for Smith to meet, we could still both think and assert, albeit falsely, that Smith met a man. The point is that that very content—​the content that Smith met a man—​would still exist and could still be the content of a possible assertion or thought even if there were no men at all in the world. We’ve barely scratched the surface of Russell’s theory of descriptions. We’ve said nothing at all, for example, about his view that definite descriptions are a kind of grammatical illusion, or his view that such expressions have no “meaning in isolation,” or his view that such expressions disappear upon proper logical analysis. But this is not the place for a thorough review and critique of Russell’s deservedly influential and deeply important theory.9 That’s because my present concern is mainly with the extraordinarily high epistemic standards to which Russell subjected our thoughts in order that they count as making what we might call direct semantic contact with objects. Unlike Frege, while Russell did suppose that such contact was at least possible, he claimed that our thoughts can make direct semantic contact with an object only if we have direct epistemic access to that object. He held, that is, that there can be no direct reference without direct acquaintance. Of course, we have already seen that we have direct epistemic access to precious



9 See Taylor 1998, c ­ hapter 2.

Inner Fitness and Outer Cause  41 few objects, according to Russell. It follows that there must be precious few objects with which we make direct semantic contact. It is by now pretty widely agreed that no objects may be immediately present to mind, in Russell’s intended sense.10 But that does not necessarily mean that we must give up on referentialism and adopt a thoroughgoing presentationalism. Instead, we should, I think, give up instead Russell’s assumption that direct reference requires direct acquaintance. And indeed, many subsequent thinkers have held that Russell’s epistemic standards for reference were too high and that his logically simple, epistemically immediate objects of thought simply do not exist. At the same time, many have thought that Russell was on to something in imposing at least some reasonably high epistemic constraint on the direct reference and thinkability of an object as such. The underlying thought among those who insist that there must be some such epistemic constraint, even if not one as demanding as Russell’s own, remains the essentially Russellian belief that if one is to have thoughts directly “about” an object, then one must “know which” object one is thinking or talking about. And in the spirit of Russell, knowing which object one is directly thinking or talking about is a matter of being able to recognize that object as the same again when one is presented with it again over at least some range of circumstances. For better or for worse, there has been little post-​Russellian consensus on just how tight one’s epistemic hold on an object must be if one is to be able to think and talk about it. Though few would now endorse a requirement of direct acquaintance on direct reference, many have endorsed descendants of Russell’s epistemic standard on reference. David Kaplan once argued that an object as such is de re or directly thinkable only if the thinker is en rapport with the object.11 One is en rapport with an object, roughly, if one has the sort of cognitive commerce with the object that renders one’s use of a name of that object what he called vivid, where vividness has to do, roughly, with the fulsomeness and accuracy of the descriptive contents one associates with the relevant name. A name was supposed to be vivid for a speaker if a speaker has lots of accurate descriptive information about the bearer of that name. If one has a vivid name for an object, one will, presumably, be able to recognize that object as the same again in a variety of different circumstances and as it appears under

10 But not everyone agrees that no properties are directly before the mind in Russell’s intended sense. See Frank Jackson. 11 Reference to Kaplan

42  Referring to the World a variety of different guises. Kaplan’s other criterion of of-​ness had to with the kind of role the object itself played in the genesis of a speaker’s use of the name. Kaplan’s idea was that it does not suffice for the thinkability of an object that the thinker knows a lot of truths about the relevant object. The object itself has to have played a decisive and central role in generating the thinker’s knowledge of those truths. Presumably the “decisive and central role” had something to do with one’s causal and or perceptual encounters with the relevant object. But on this point Kaplan was not entirely clear. Kaplan’s rapport is an epistemically demanding criterion on the de re thinkability of an object, but it isn’t quite as strong as Russell’s direct acquaintance. Even if one possesses a vivid name of an object, there can still be scenarios in which the object is present in a surprising or unrevealing guise. On such occasions, one might fail to recognize that one is presented with the same object again. Less epistemically demanding than either Kaplanian rapport or Russellian acquaintance is what we might call ordinary “knowledge-​ wh”—​that is, ordinary knowledge—​who, what, when, or where. In a quite ordinary sense of speaking, I know who my wife Claire is, know when I am writing this sentence, know which computer I am writing it on, and know where I am now sitting. But I am not, in Russell’s sense, directly acquainted with any of these objects. Similarly, with the exception of my wife, it is also unlikely that I possess anything so strong as Kaplanian rapport with any of these objects. I don’t intend to offer a precise account of ordinary knowledge-​wh here. It is worth saying that it is clearly necessary, though clearly not sufficient, for ordinary knowledge of this sort that one has at least some causal commerce with the relevant object. And it is clearly sufficient, but perhaps not necessary, that one has frequent enough perceptual contact with the object or that one is able reliably enough to recognize the object as same again under various circumstances. How much perceptual contact is enough? How reliably is reliably enough? Clearly, our workaday conception of knowledge-​who is contextual and interest-​relative. In some contexts, I count as knowing who Smith is only if I can visually discriminate between her and others. In other contexts, even if I cannot recognize Smith as Smith when I see her, I may nonetheless count as knowing who she is merely on the basis of having read the collected works of Smith. Suppose I know little more about Smith than that she is the author of such and such important works of philosophy. At a philosophical conference, someone asks “Do you know who Smith is?” And I respond, “Yes I do. She is the noted author of The World as Representation

Inner Fitness and Outer Cause  43 and Represented.” Even if that is the sum total of my knowledge of Smith, in this context it seems that I speak truly. Suppose, for example, that I am attending a crowded reception at the conference. I hear that Smith is in attendance. Dying to make the acquaintance of the renowned Smith, I ask my friend to point Smith out to me. “Which one is Smith?” I ask. My friend does point Smith out. I now come to know what Smith looks like. But despite my newly acquired capacity to recognize Smith when I see her, there seems to be a perfectly good intuitive sense in which I knew who Smith was all along. As a consequence, though it seems correct to say that I’ve gained some additional recognitional abilities with respect to Smith, it doesn’t seem correct to say that I have for the first time come to have the ability to think de re or singular thoughts about Smith. It would be an interesting, worthwhile, and challenging philosophical task to work out the factors that determine just how tight our cognitive grip on an object has to be in various contexts in order that one count as possessing ordinary knowledge-​wh of the relevant object, and thus, presumably, as being in sufficient cognitive contact with it to render the object thinkable at all. I will not undertake that task here. On my view, acquaintance has been oversold as a constraint on the possibility of our referring to and thinking about objects.12 In fact, I shall eventually argue that an object can be directly or de re thinkable even if we are not in any strong or weak sense epistemically acquainted with it. On my own view, in other words, it will turn out that semantic directness requires neither that we have direct acquaintance nor that we stand in some more attenuated “knowledge-​which” relation to the objects of our thought. But more on this point in due course. Here, I  want to focus more on what I  see as the culprit that first led philosophers down the primrose path of thinking that direct reference must require some either strong or weak version of direct acquaintance to begin with. That culprit is the epistemic one-​sidedness of all reference. Epistemic one-​sidedness is sometimes widely, but mistakenly, taken to motivate and justify presentationalism and to spell the decisive blow against referentialism. Seen in this light, the appeal to direct acquaintance is a way of singling out direct reference as an epistemically special form of reference. But this assessment is, I think, mistaken on all counts. Direct reference is not epistemically special at all. And the thought that it must be is based on a misdiagnosis of the source and nature of the epistemic one-​sidedness of

12 Jeshion makes a similar claim.

44  Referring to the World reference. When properly understood, epistemic one-​sidedness neither justifies presentationalism nor undermines referentialism. And that means that we can give a satisfying explanation of epistemic one-​sidedness while still adhering steadfastly to semantic referentialism. Or so I shall argue in the remainder of this chapter.

4.  Objectuality and Objectivity Again I have already made it clear that on my view, reference is a distinctively mental and linguistic affair. Set up whatever causal connection you like between the states of, say, an isolated window and the states of a brick. Unless you have thereby made the brick function as either a linguistic or mental representation of the right sort within an overall system of such representations, implementing that causal connection may suffice to turn the state of the brick into a signal of the state of the window, but it will not suffice to turn the window into the referent of the brick. To say that the states of a brick may, in the right setting, be made to function as a signal of the states of the window is to allow that in the right sort of background, the brick and its states can enjoy a certain sort of derivative or perhaps even natural semantic relation to the window and its states. Two agents could simply agree, for example, to use the brick to signal whether the window is opened or closed. Perhaps when a given face of the brick is turned up, that is to be taken as a signal that the window is open and when the opposite face is turned up, that is to be taken as a signal that the window is closed. But the brick is not thereby made into an autonomously functioning or syntactically articulated representation, with distinct and assignable parts that perform distinct and assignable semantic tasks like referring to the window or predicating a property of the window. The most that we could say about the brick in relation to the window and its state is that the brick as a whole somehow manages to represent the complex consisting of the window and its state as a whole, and then only relative to a certain background of intentions. We have been imagining a wholly conventional arrangement, one that sets up a correlation between the overall state of one thing and the overall state of another. Such correlations might also be a consequence of a combination of local facts and in situ laws. Such a combination might, for example, arise as a consequence of certain naturally selected arrangements.13

13 See Millikan.

Inner Fitness and Outer Cause  45 Especially  when one of the correlated entities plays a role in the overall economy of some living, breathing organism, it is tempting to see in such correlations the ground for at least the first stirrings and rudiments of a kind of intentionality and so as a kind of natural signaling system.14 There is no doubt something to be said for this thought, at least when it is taken as a first pass at what it might take to reduce our conceptual puzzlement about the very possibility that something with at least some of the hallmarks of intentionality might subsist in the natural order.15 Still, such natural signaling systems, as we might call them, do not provide very deep insight into the nature of peculiarly linguistic or even peculiarly mental representations. Such natural signals enjoy at most what I have elsewhere called “type 1 intentionality” and not what I call “type 2 intentionality” of the sort that is characteristic of syntactically and semantically articulated representational systems. In particular, they tell us little about reference as such and its potential bottom-​up role in determining the objective representational contents of our thoughts and our words.16 Suppose we presume an articulated mind, with syntactically and semantically articulated innards, one side of the equation. And suppose we presume the sort of nomically grounded relations of covariation that suffice for what I have called type 1 or signal-​like intentionality on the other. Would that be sufficient to give us full-​blown type 2 intentionality? Many versions of externalism about mental content might answer yes to that question. But that is because externalists about content tend to be insufficiently attentive to the role of the structured innards of the organism in making objective representational content possible in the first place. One of the aims of my two-​factor referentialism is to correct that lacuna while doing justice to certain important insights of the externalists. I hold, in particular, that externalist theories of objective representational content need augmenting by an explanation of why our thought and talk should take the form of thought and talk about objects and their properties in the first place. Without being augmented by such an explanation, externalist approaches lack the resources to explain how the mind manages to transform what would otherwise be the semantically inert flow of energy inward into outwardly oriented representational 14 Fodor, Dretske, Millikan, Taylor. 15 To the extent that what we might call nomic covariationist approaches to objective representational content, in the spirit of Fodor (1987, 1990) and Dretske (1981) or selectionist theories of content in the spirit of Millikan and Dretske. 16 Reference to Millikan (2005, 2012) on natural signs.

46  Referring to the World contents. Only a mind antecedently poised to refer in virtue of being equipped with a distinctive sort of system of representations, organized in a distinctive way, will contain representations that refer to an object in virtue of being causally or historically connected to that object in the right way. Unless we attend to the distinctive inner nature of referring representations, and the representational systems and structures within which they are at home, we will be at a loss to explain how reference to the world is at all possible. That is, unless we understand how our inner representations manage to be objectual, and thus how our minds are antecedently poised or ready to refer, we will be powerless to explain how the mind’s external causal encounters with the inward rush of energy manages to render those representations fully objective and thus referentially successful. Recall that I  suggested in Chapter  1 that a representation is objectual if it is syntactically fit for the job of standing for an object. And I said that fitness for the job of standing for an object is a matter of enjoying certain broadly syntactic or structural features. Objectual representations are those that can well-​formedly flank the identity sign, occupy the argument places of verbs, serve as links of various sorts in anaphoric chains of various sorts. In Chapter 4, I will say a great deal more about the syntactic character and role of names. I take the whole class of singular terms, including demonstratives, indexicals, variables, and pronouns, to be the paradigmatic examples of referentially fit expressions. I argue that the objectuality or referential fitness of such representations is determined by their systematic roles in the entire syntactic apparatus of singular reference—​an apparatus built of terms, quantifiers, argument-​taking verbs, and the identity sign. It is important to stress what I am and am not claiming on behalf of referential fitness. First, mere referential fitness must be sharply distinguished from full-​blown referential success. Referential fitness is determined by certain sorts of facts about representation-​representation relations. But representation-​ representation relations are insufficient on their own to supply any representation with a represented object. That is why a representation can be referentially fit without actually standing for an object, without, that is, being referentially successful. I will sometimes also say that representations that are referentially fit, but not referentially successful, are objectual without being objective. Merely objectual representations are fully poised to refer, but they do not succeed in actually referring merely in virtue of being poised to refer. Now the role-​oriented property of fitness, so understood, is crucial to the bottom-​up explanatory strategy pursued throughout

Inner Fitness and Outer Cause  47 this book. That strategy, again, presupposes that reference is a real, though non-​fundamental relation in nature and it grants that because reference is not a fundamental relation, it stands in need of explanation and vindication. It insists that once reference has been vouchsafed as real, then facts about reference can in turn be deployed to cast further explanatory light on the nature of objective representational content. Although the strategy grants that reference itself is neither metaphysically nor explanatorily basic, it takes reference to be more basic than the objective representational contents of whole sentences or entire thoughts. The distinction between merely referentially fit and fully referentially successful representations figures in our bottom-​up explanatory strategy in two different ways. First, it is argued that the fitness-​making factors are essential pre-​conditions of referential success. The claim is that it is only already referentially fit expressions, expressions that by their very natures are antecedently poised or ready to refer, that are rendered referentially successful as consequence of the mind’s encounter with the mere inward rush of energy across the portals of sensation. Second, it is argued that the very structural and syntactic features in virtue of which certain expressions are antecedently poised to refer also helps prepare the way for successful reference, once achieved, to have an effect on objective representational contents of whole sentences. What is absolutely crucial here is the role-​oriented character of referential fitness. According to two-​factor referentialism, no isolated representation, all on its own and independently of its connection to other representations, can be fit or poised for the job of standing for an object. No expression has standing as a name, for example, except in virtue of playing the right kind of structural role in a system of systematically interlocking linguistic representations. But if it is right that referential fitness is a pre-​ condition of referential success and right that such fitness is determined by certain role-​oriented factors, then it follows that no object can be successfully designated except by an expression which already occupies a role in a system of systematically interlocking representations. But the crucial further claim is just because of the role-​oriented character of referential fitness, that successful reference, once achieved, can also have an upward effect on the objective representational contents of any sentence in which a referring expression may occur. That is why I say that referential fitness both prepares the way for the initial achievement of reference in the first place and renders reference, once achieved, to be more than an idle wheel spinning in the theory of objective representational content.

48  Referring to the World

5.  Kantian and Fregean Roots I will have a great deal more to say about the internal fitness making factors over the course of the next several chapters. The focus in these chapters will be to show that the notion of referential fitness, especially given a syntactic account of the source and nature of referential fitness, is the key to solving certain long-​standing puzzles in the theory of reference. In Chapter 3, I will defend the coherence and plausibility of my bottom-​up explanatory approach to reference and the problem of objective representational content. In Chapters 4 and 5, I will argue that attending to the structural and syntactic features that render singular representations poised or ready to refer is key to explaining the source and nature of the epistemic one-​sidedness of all reference and to solving what I  call the “puzzles of coreference.” In Chapter 6, I will show that the notion of a merely objectual representation, together with certain further distinctions, is central to explaining our apparent ability to refer to non-​existent objects. Before plunging into those arguments, however, I close this chapter by briefly tracing the Fregean and Kantian routes of my distinction between merely objectual and fully objective representations. That may be surprising especially in light of my rejection of Frege’s presentationalism and also in light of the fact that there is meant to be no hint of Kantian transcendental idealism in this book. The moral is just that one can find part of the truth in many places, even buried deep within the opponent’s camp. My central claim about objectuality is that it is constituted by structural or syntactic features that lie entirely on the side of the cognizing subject. Through such features no objects are yet given or constituted. Objectuality has to do not with the content or objects of thought but with the form of our thought. To explain what objectuality consists in is, in part, to explain how and why our thought and talk take the form of thought and talk as of objects. The question of why our thought and talk should take the form of thought and talk as of objects has largely faded from center stage in contemporary philosophy of language and mind. But it was not always so. Indeed, the question of why our thought and talk should take the form of thought and talk about objects at all once occupied center stage for philosophers as diverse as Kant, Frege, and even Quine.17 If we are to see both what is right and what is wrong about certain prominent views about the nature of objective

17 Insert Reference to Classic Texts of Kant, Frege, and Quine on singularity.

Inner Fitness and Outer Cause  49 representational content, it is time to shine the klieg lights once again on the form-​content distinction.18 It was Kant who was perhaps the first to grasp, albeit through a glass darkly, something like the distinction between merely objectual and purely objective representations. He even seemed to grasp the significance of that distinction for the very possibility of objective representational content. Consider Kant’s attempt to distinguish bona fide judgments from merely subjectively valid associations. Kant took it that judgments purport to be about objects and how things are by such objects. By contrast, he took mere subjectively valid associations to concern only our own psychology. In judging that bodies are heavy, we purport to represent how things are by the objects themselves, Kant held. Representing how things are by the objects is different from reporting on the merely subjective regularity that when we lift a body, we feel a pressure of weight. But Kant believed there to be certain purely formal conditions, arising solely from the side of the understanding, on the very possibility of bona fide judgments with objective representational purport. It is in his account of those conditions, that Kant gets gropingly at something like the notion of an objectual representation. Without attempting a full elucidation, it bears stressing that Kant’s account of the merely formal conditions on the “objective validity” of our judgments, as he called it, is not intended as an account of how judgments actually succeed in reaching out to the objects. He held that in order that our judgments achieve full blooded objective validity, a “given manifold” of sensory intuition must be “brought under” what he called the necessary synthetic unity of apperception. But because the understanding contains no manifold of its own, but only the formal grounds of the synthetic unity of an alien manifold, he insisted that the formal conditions on objective validity which arise on the side of the understanding can suffice only for the objectuality of judgments, and not for their full-​blooded objectivity. Witness too in this connection, Kant’s distinction between merely thinking an object and cognizing an object. In full-​blown cognition of an object, he held that there must be both a given intuitive element and a formal conceptual element. In bare thought, devoid of intuitive content, we have, Kant claims, merely “empty concepts of objects, through which we cannot

18 Reference Jeshion on semantic instrumentalism. (where did this term originate? Kaplan?) Reference Brandom as recent philosopher concerned with form-​content distinction. Brandom quote.

50  Referring to the World even judge whether the latter are possible or not—​mere forms of thought without objective validity.”(Kant 1781/​1998. 255 (B 148)) Now Kant lays heavy stress on the emptiness of representations that are merely objectual—​to use my own vocabulary rather than his own—​but not objective, especially in the “Transcendental Analytic” of the Critique of Pure Reason. There he claims that such representations are cognitively useless. Nonetheless, we should not conclude too much from Kant’s claims about what may appear to be the cognitive disutility of merely objectual representations. What Kant seems really to want to say is that merely objectual representations can play no cognitive role in the theoretical realm. But in an autonomous practical realm, the realm in which, on his view, religion and morality are at home, Kant would seem to afford merely objectual representations a central, even indispensable role. He holds, for example, that we have no fully objective representation of ourselves as free, autonomous moral agents. As a consequence, we can never cognize ourselves as free, autonomous agents. But at the same time he also would seem to hold that merely objectual representation of ourselves as free, autonomous beings—​representations through which no concrete real existent is given to thought z—​somehow lies at the very foundations of morality. In fact, Kant evidently takes it to be one of his signal philosophical achievements to show how morality can be rationally mandatory despite the fact that morality is apparently by his lights a game ultimately played with what would seem to be merely objectual representations that never achieve full-​blooded contact with concrete real existents. Frege (1960) too tacitly acknowledges something rather like the distinction between fully objective and merely objectual representations. But I  should say that his grasp of this distinction emerged gradually and was more begrudging. We can begin to trace the evolution of his understanding by considering his account in the Foundations of Arithmetic of what it takes for a singular term to stand for an object. He claims, If we are to use the sign a to signify an object, we must have a criterion for deciding in all cases whether b is the same as a, even if it is not always in our power to apply this criterion. (Frege, 1884/​1980, 73)

Now the central thought here is that an identity statement expresses what is contained in a recognition judgment—​a judgment to the effect that an object “given” in one way is the same again as an object “given” in another way. We have already seen that Frege held that “the first and most important

Inner Fitness and Outer Cause  51 discoveries in science are often a matter of recognizing something as the same again.” A natural and appealing first reading of Frege here is that he is claiming that we have succeeded in using a sign to successfully designate a determinate object just in case we have fixed a “significance” for each identity statement in which that sign may occur. It is important to stress that Frege’s talk of ways of being “given” objects cannot be understood entirely on analogy with the sensory givenness of a perceived object. He observes that we have neither “ideas” nor (sensory) “intuitions” of numbers. Yet he insists that they are still in some sense “given” to us. At times Frege seems to hold that not just in the case of numbers but in the general case as well it is not through ideas and intuitions that objects are given. Rather, objects are somehow given through the use of singular terms, paradigmatically in identity statements and recognition judgments. Indeed, Frege seems to hold that the concept of an object in general, as Kant might have put it, is simply the concept of that which is given through the use of a singular term. It is as if Frege believes at this stage of his thinking that objects are nothing but the “shadows” cast by the uses of singular terms. At least a view of this sort would make sense of the idea that even in the general case the givenness of an object consists in nothing more than the intelligibility and meaningfulness of certain identity statements. For example, the givenness of numbers is supposed by Frege to consist entirely in the intelligibility and meaningfulness of numerical statements of identity ultimately of this form: the number which belongs to the concept F is the same as that which belongs to the concept G. (Frege, 1884/​1980, 73)

When we have defined senses for identity statements of this sort, Frege claims, we have succeeded in specifying a means of “arriving at a determinate number and of recognizing it again as the same.” (Frege, 1884/​1980, 73) There is not yet a basis in Frege’s account of the epistemic givenness of an object for anything like our distinction between merely objectual singular representations and fully objective singular representations. Nor is it clear that there could be such a basis in Frege’s early work. Frege does admit that there are concepts under which no object falls—​e.g., the concept of not being self-​identical or the concept of a fraction smaller than 1 such that no fraction smaller than 1 exceeds it. He even admits that there are perfectly meaningful general or concept terms and phrases that express such empty concepts. But he claims that when we attempt to form a (complex) singular

52  Referring to the World term from any such “empty” phrase by adjoining the definite article—​as in, “the largest proper fraction”—​the resulting expression is “without content” and “senseless.” The difference between general terms and such complex singular terms has apparently to do with a difference in associated existential commitments. He says of empty general terms, for example, that no harm can come from the use of them as long as “we do not assume that there is anything which falls under them—​and that we are not committed merely by using them.”(Frege, 1884/​1980, 87) By contrast, he claims that the use of the definite article to form a singular term from a general or concept expression cannot be “justified” unless two propositions are proven antecedently, viz.: (1) At least one object falls under the relevant concept (2) At most one object falls under the relevant concept. Where such propositions cannot be proven, adjoining the definite article to a concept word to form a would be singular term yields an expression apparently entirely devoid of sense or content. At this early stage of his thinking, Frege did not have firmly in mind anything like my distinction between merely objectual singular representations and fully objective singular representations. It is not clear that he had conceptual resources that would enable him to distinguish intrinsic referential purport from full-​blown referential success. But with the subsequent emergence of his vital distinction between sense and reference, Frege clearly positions himself to appreciate something like the distinction I am after. Armed with the distinction between sense and reference, Frege is able to maintain that there can be fully “contentful” singular terms which, nonetheless, stand for no objects. Recall from our earlier discussion of Frege’s presentationalism that the sense of a term contains a mode of presentation and criterion of identification of an object. An object is the referent of the relevant term if it is presented under the relevant mode and satisfies the relevant criterion. The crucial further claim is that a term can have a fully determinate sense, even if no object satisfies the relevant criterion of identification. Singular terms with sense, but no reference, are associated with fully determinate modes of presentation and criteria of identification, despite their lack of reference. As such, contrary to his earlier views, Frege can maintain that even non-​ denoting singular terms are fully “contentful” and enjoy fully determinate referential purport. To put it in our own vocabulary, such expressions would

Inner Fitness and Outer Cause  53 be referentially fit, without being referentially successful, objectual without being objective. Notice that once the distinction between sense and reference has been introduced, and once he has allowed that some terms have sense but no reference, Frege is no longer in a position to maintain, as we put it above, that objects are nothing but the shadows of singular terms. An expression that has a sense but lacks a reference is precisely a term with full-​blown referential purport through the use of which no object is so far given. Frege himself insists, for example, that any such term still does contain a way of being given an object. But he now acknowledges that even when a term is associated with a determinate way of being given an object, there is no guarantee that an object is actually given in the relevant way. Consequently, he can no longer maintain that the givenness of an object consists entirely in the intelligibility and meaningfulness of relevant identity statements. Identity statements can be intelligible and meaningful even when the relevant terms altogether lack reference. At this stage of his thinking, Frege explicitly allows that sense-​having but reference-​lacking singular terms may have distinctive uses. To be sure, he restricts their legitimate uses to the realm of fiction and myth. But this still implies that such expressions are not just idle wheels spinning, devoid of any use whatsoever. But it was precisely this that his earlier views would seem to imply. Now we have already said that on Frege’s view in the realm of fiction and myth, we never descend to the level of truth and reference. Rather we remain at the level of sense—​that is, at the level of thought. But in the realm of fiction and myth, he claims, we are concerned not with the truth or falsity of such thoughts, but only with the feelings and emotions to which, in the bare entertaining of them, our thoughts give rise. The deeper point here is that the very possibility of fully determinate senses through which no object is actually given would seem to be at odds with Frege’s earlier thought that the givenness of an object consist in nothing but the meaningfulness and intelligibility of certain identity statements. I said earlier that although Frege does anticipate the distinction between objectual and objective representations, his acknowledgment of that distinction is at best begrudging and tentative. Indeed, he remained deeply ambivalent about what I am calling merely objectual representations. In places, he dismisses so-​called non-​denoting proper names as “mock” proper names and insists that sentences in which such names occur express merely “mock” thoughts:

54  Referring to the World Names that fail to fulfill the usual role of a proper name, which is to name something, may be called mock proper names. Although the tale of William Tell is a legend and not history, and the name “ William Tell” is a mock proper name, we cannot deny it a sense. But the sense of the sentence “ William Tell shot an apple off his son’s head” is no more true than is that of the sentence “William Tell did not shoot an apple off his son’s head.” I do not say that this sense is false, but I characterize it as fictitious. . . . Instead of speaking of fiction, we could speak of “ mock thoughts.” Thus, if the sense of an assertoric sentence is not true, it is either false or fictitious, and it will generally be the latter if it contains a mock proper name. . . . Assertions in fiction are not be taken seriously, they are only mock assertions. Even the thoughts are not to be taken seriously in the sciences: they are only mock thoughts. If Schiller’s Don Carlos were to be regarded as a piece of history, then to a large extent the drama would be false. But a work of fiction is not meant to be taken seriously in this way at all: it’s a play. . . . The logician does not have to bother with mock thoughts, just as a physicist, who sets out to investigate thunder, will not pay any attention to stage thunder. When we speak of thoughts in what follows, we mean thoughts proper, thoughts that are either true or false. (Frege, 1897, 229)

Frege’s striking ambivalence toward the merely objectual tempts him to conclude that “mock” singular terms ought, in fact, to be banished from the purified language of science and mathematics. No language that admits merely objectual singular representations, he seems to believe, could be adequately serviceable as an instrument for the pursuit of truth and knowledge. In a “logically perfect” language, merely objectual singular linguistic representations would be inadmissible as ill-​formed. It should be clear that I  do not share Frege’s ambivalence toward the merely objectual. I take his ambivalence to be a sign that he did not fully grasp the significance of the distinction that he so grudgingly and ambivalently acknowledged. On my view, Frege’s failure to fully grasp the distinction, even though he had conceptual tools at his disposals that would enable him to do so, was the source of a number of profound and consequential mistakes on his part. I  shall argue in Chapter  6, for example, that merely objectual representations, and the distinctive sorts of language games we play with them, occupy a much grander role in our shared mental lives than Frege imagined. It is not just that I afford greater cognitive significance to

Inner Fitness and Outer Cause  55 explicit fiction and myth than Frege did. Merely objectual representations occupy center stage even in a domain that Frege himself took to have the most impeccable epistemic credentials. Frege supposed that numbers are objects and that number words, including the numerals, are fully objective singular representations. He was half right and half wrong in his estimation of the language of mathematics. Number words are objectual representations that function within mathematical discourse as what I call devices of same-​ purporting thought and talk. Though they do not thereby make full-​blooded contact with a realm of independently existing objects, they do play a distinctive sort of role in a distinctive and cognitively significant sort of language game—​what I call a “non-​veridical” language game. Non-​veridical language games are games paradigmatically played with representations that are merely objectual rather than fully objective. The governing concern within such games is typically what I call truth-​similitude, rather than flat-​out truth. In Chapter 6, I shall argue that not only explicit fiction but also pure, as opposed to applied mathematics, should be regarded as non-​veridical language games, played with merely objectual representations, governed by a concern for truth-​similitude rather than flat out truth. More generally, non-​veridical language games are the stuff of which shared imaginings are made. We engage in a shared imagining whenever we imagine together places, people, societies, or entire world-​orders alternative to the actual. We undertake a shared imagining when we represent to ourselves moral and aesthetic ideals nowhere realized in the history of the world. We imagine together in deliberating together about alternative futures. Even much of our thinking about a shared past is, I think, really a kind of shared imagining. When one identifies oneself as an American or an Israeli or a Serb, for example, one typically participates in the practice of imagining. One imagines a past, a past that is supposed to be partly constitutive of one’s identity as American, or Serb, or Israeli. These identity-​constituting but imagined pasts are often mythical, bearing no particularly deep relationship to the facts about the past. Partly because of their mythical character, our imagined pasts are often contested. But it would be a mistake to think that those who contest an imagined past are ipso facto merely attempting to shine the light of truth in the dark closet of myth. For the past often figures in such contests of self-​imagining, I think, only as a prop in the struggle to constitute a shared identity. The deeper point is that the capacity for shared imaginings is a distinctively human capacity. It lies at the very foundation of the human capacity to

56  Referring to the World produce distinctively human cultural and social life. This book does not purport to explore either the rich extent and variety of our shared imaginings or the multiplicity of sources from which the contents of our shared imaginings derive. What I seek to show instead is that the referential apparatus of shared languages—​the whole apparatus of names, deictic expressions, quantifiers, variables, and anaphora—​is fundamental to our capacity for shared imaginings. The point is that although the referential apparatus of our language is, in one sense, made for talking about actual existents, we can and do deploy that apparatus even in the absence of any actual existent. We do so when we play non-​veridical language games with merely objectual representations. Our ability to deploy the referential apparatus of our language even in the absence of actual existents is part of the explanation of the very possibility of our capacity for shared imaginings. To be sure, when we deploy the referential apparatus of our language in the absence of real existents, our linguistic play may still enjoy the feel of objectivity. But we should not mistake the illusion of objectivity for real thing. That way, I shall argue, lies the philosophical darkness of Platonism and Idealism.

3 Against Jazz Combo Theories of Meaning and Reference 1. Preliminaries In Chapter  1, I  acknowledged that there are philosophers who reject the bottom-​up explanatory approach to reference and the problem of objective representational content that is pursued throughout this book as misguided from the start. Among the naysayers are the likes of Donald Davidson (1977), Robert Brandom (1994), and many others. Davidson, recall, grants that our thought and talk have objective representational contents, but he insists that reference plays no role in explaining the very possibility of such contents. “We don’t need the concept of reference,” says Davidson; “neither do we need reference itself, whatever that may be.” (Davidson 1977, 256) And Brandom dismissively claims that the reference relation is “a philosopher’s fiction, generated by grammatical misunderstandings.” (Brandom 1994, 324)  It goes without saying that I reject this dismissive approach to reference and its potential role in the theory of objective representational content. In this, chapter I say more about why I do. To aid in my argument I introduce what I call the “jazz combo theory” of meaning and reference. Based on a certain extended metaphor, the jazz combo theory represents my attempt to crystallize the spirit of a diverse family of theories which have in common only that they reject reference, and with it the bottom-​up explanatory approach to the problem of objective representational content. By bundling together an entire and diverse family of theories in this way, I hope to avoid point-​by-​point engagement with details of the plethora of individual approaches and focus instead on their shared general spirit. To get an initial feel for what I have in mind, consider the following rather extended metaphor. Imagine a collection of “musical” players. Each player has an instrument capable of making various sounds—​some pleasant, some not so pleasant. They have never played together before. There are no antecedent musical “norms” by which they are mutually guided. They share no Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0003

58  Referring to the World antecedent musical “language” within which they may express and make mutually manifest their musical intentions. There are no extant musical genres of which they have joint mastery. No player even has anything like a fixed musical style. They come together possessing only the bare capacity to make various sounds on their instruments. But now suppose that somehow or other each player forms the intention to make sounds that will be coordinated with the sounds made by every other player. Each player intends, and intends to make it manifest that she intends, to make sounds in such a way that other players will be willing and able to “play along” in ways that all players find mutually satisfying. In the absence of any antecedent musical common ground, musical norms, or musical language, the initial attempts at playing together are likely to produce a cacophony of mere noise rather than anything that deserves to be called music. Their initial cacophony would not be assessable in accordance with any musical norms at all. Nothing in the initial sound stream will yet have status as an attempt, successful or unsuccessful, to play this or that melody or to play in this or that key or this or that meter. No sound or sequence of sounds will yet count as in tune or out of tune, as appropriately or inappropriately stressed, as well or badly phrased. Yet, as each player continually adjusts to the play of every other player, there may eventually emerge a set of shared musical “norms.” Once such norms have emerged and have come to guide and govern players in their shared attempts to play together, what would otherwise have been a mere cacophony is thereby promoted into genuine music. In fact, it seems intuitively plausible that the eventual mutual (tacit) endorsement of such musical norms is what would first make it the case that some sound sequences count as at least attempts at playing this or that melody, as attempts at playing with this or that meter, as in tune or out of tune, as appropriately or inappropriately stressed or phrased. Once such norms have emerged and are mutually endorsed, players will now be entitled to hold one another to those norms. It may be that such norms remain merely implicit in their shared musical practices. In that case, “holding” one another to norms would be not a matter of explicit criticism but of practical responses, of refusing to play along with another who, for example, plays out of tune. But there may also emerge a set of devices for making matters like the intended meter and key signature explicit and with it there might emerge an explicit critical vocabulary for evaluating the play of others. The crucial and deep point is that the emergence of determinate musical content, on this picture, is simultaneous with the emergence

Against Jazz Combo Theories  59 of determinate musical norms, whether those norms are left merely implicit in a practice or are elevated into an explicit critical vocabulary. Our extended metaphor is intended to motivate the thought, prevalent among jazz combo theorists of all stripes, that the objective representational content for the sentences of our language, and possibly even of our thoughts, are, in a sense, coterminous with a set of mutually owned linguistic or discursive norms. In our metaphor, it was only against a backdrop of mutually owned musical norms that functioned to guide and govern various musical performances that what would otherwise have been mere noise was constituted as music. So too, the jazz combo theorist holds, it is only against a backdrop of mutually owned linguistic or discursive norms that somehow come to guide and govern our practices that what would otherwise count as productions of meaningless strings devoid of any linguistic function, come to be constituted as determinate linguistic acts, with determinate propositional contents. On this sort of view, a sentence enjoys a propositional content, then, solely in virtue of being liable to determinate patterns of normative assessments. Propositional contents and the norms governing a set of discursive practices are two faces of the same discursive coin.1

2.  Jazz Combo Theory and the Priority of the Sentence Now jazz combo theorists reject bottom-​up approaches to objective representational content in favor of more top-​down approaches largely because they take bottom-​up approaches to be incompatible with the principle of the semantic priority of the sentence and take top-​down approaches to be compatible with that principle. To a first approximation, the principle of the semantic priority of the sentence says that sentences and their meanings are explanatorily prior to the meanings of expressions shorter than a sentence. Frege (1980) was, perhaps, the first to endorse the semantic priority of the sentence in the form the context principle. According to Frege, it is only in the context of a sentence that a word has meaning and/​or reference at all. The 1 To be sure, jazz combo theorists have offered a variety of different accounts, some greatly detailed, some mere sketches, of just what sorts of practices and norms figure decisively in the initial constitution of propositional content. Some, like Davidson, appeal to perfectly general norms of rationality and also to certain constraints on the very possibility of what he calls “radical interpretation.” Others, like Brandom, appeal to more specifically linguistic practices and norms. But jazz combo theorists generally agree that the constitution of content is possible only where cognizing agents enjoy what I call rational solidarity or normative community one with another.

60  Referring to the World context principle is widely believed to entail that sentences are the primary and fundamental units of linguistic significance. Indeed, it is sometimes taken to be a consequence of this fact that any expression shorter than a sentence enjoys linguistic significance in only a derivative or secondary sense. 2 Jazz combo theorists frequently note that only the utterance of a free-​ standing sentence constitutes a fully determinate and contentful linguistic act. Paradigmatically, we assert, promise, command, and request by uttering free-​standing sentences. Except in special circumstances—​or relative to a background of issues, as John Perry (1994) puts it—​the bare utterance of an 2 Consider, for example, Brandom’s (1994) claim that the meanings of sub-​sentential expressions, and even those expressions themselves, have what might be called a problematic status. He puts it this way: It is not obvious why there should be subsentential expressions at all, for they cannot have the same sort of fundamental pragmatic role to play that sentences do. So we ought to start by asking . . . “What are subsentential expressions, and why are there any?” (363–​64) Though Brandom does offer detailed answers to both the “what” and the “why” question, it is enough for our current purposes to note that in pursuing answer to these questions, he insists that subsentential expressions and their meanings have a dependent or secondary status. Subsentential expressions, he claims, cannot have semantic contents in the same sense in which sentences can. They can be taken to be semantically contentful only in a derivative sense, insofar as their occurrence as components of sentences contributes to the contents . . . of . . . sentences. (364) Like Wittgenstein before him, Brandom seems to take the fact that it is paradigmatically tokened by claiming that significant illocutionary acts are performed to establish that sentences enjoy semantic content in a way not metaphysically or explanatorily dependent on the meanings of any antecedently meaningful subsentential constituents. As he puts it: the category of sentences has a certain explanatory priority over subsentential categories of expression, such as singular terms and predicates. For sentences are the kind of expression whose freestanding utterance . . . has the pragmatic significance of a performing a speech act. . . . Accordingly, there is available a sort of answer to the question—​What are sentences and why are there any?—​that is not available for any subsentential expression, namely, sentences are expressions whose unembedded utterance performs a speech act, such as making a claim, asking a question, or giving a command. Without expressions of this category there can be no speech acts of any kind, and hence no specifically linguistic practice. (363) Though my main aim in this chapter is less to refute the view that the sentences are semantically prior to their constituents than to show that no such view is forced on us by any non-​question-​begging principles that are plausibly part of the philosophical common ground between bottom-​up theorist and jazz combo theorists, I should say that the jazz combo approach to the order of metaphysical and explanatory priority between sentences and their subsentential constituents strikes me as pretty nearly incoherent. It is hard to fathom how possibly, on this view, utterances of particular sentences are supposed to constitute performances of peculiar and determinate speech acts. Suppose I we ask why it is that by uttering the sentence “Claire is puzzled” I manage to assert of Claire and no one else that she is puzzled. Try as they might, defenders of the semantic priority of the sentence will, I predict, be unable to provide a compelling answer to this question. To be sure, Brandom offers a quite elaborate defense of his views about the secondary and derivative status of subsentential constituents and their meanings. But my aim is not to launch a full-​scale attack on that defense here. I will, however, argue later in the chapter that Brandom and other jazz combo theorists seem to mistakenly infer the semantic priority of the sentence from what I call the “dynamic priority” of the sentence.

Against Jazz Combo Theories  61 expression shorter than a sentence—​a name, say—​does not yet constitute a determinate linguistic act. This starting point has the ring of a truism. But if it is a truism, it is one of which jazz combo theorists make much. Many take this truism to directly entail, more or less on its own, that no free-​standing expression shorter than a sentence is semantically significant in its own right. Expressions shorter than complete sentences may indeed have semantic significance, but only insofar as they are understood as potentially occurring in sentences. Commenting on the context principle, Michael Dummett (1981), for example, remarks that a sentence is . . . the smallest unit of language with which a linguistic act can be accomplished, with which a “move can be made in a language game”: so you cannot do anything with a word—​cannot effect any conventional (linguistic) act—​save by uttering some sentence containing that word (save for the cases in which, as in the answer to some questions, the remainder of the sentence is understood from the context). And we have seen that, for Frege, the sense of a word or expression always consists in the contribution it makes to determining the thought expressed by a sentence in which it occurs. That is, the sense of the word consists in a rule which, taken together with the rules constitutive of the sense of the other words, determines the condition for the truth of a sentence in which the word occurs. The sense of a word thus consists—​wholly consists—​in something which has a relation to the truth-​value of sentences containing the word. (194)

Precisely because the semantic priority of the sentence has been taken by jazz combo theorists to entail that sentences have explanatory priority over constituents, it has also been thought to more or less directly entail the falsity of the sort of bottom-​up approach to objective representational content that is on offer in this book. The problem is that bottom-​up approaches would seem to require that semantic facts about expressions shorter than sentences be prior to and independent of semantic facts about sentences. Take causal theories of reference as a case in point. Such theories would seem to require that the causal factors that putatively play a role in constituting reference ultimately serve as part of the explanation of why a sentence containing a certain referring expression has the objective representational contents that it does. But such causal factors could enjoy explanatory priority over the contents of sentences only if they, and the referential relations that are thereby at least partly constituted, were metaphysically antecedent

62  Referring to the World to and independent of facts about sentences and their objective representational contents. But if expressions smaller than the sentence have no independent semantic significance, there could not be a metaphysically separable reference relation, specifiable prior to any facts about the totality of potential speech acts.

3.  The Cause-​Norm Gap Bottom-​up theories are also alleged to run afoul of the priority of the ­sentence in another way. Not only do they allegedly get the explanatory priorities wrong, but they also allegedly fail to bridge the supposed gap ­between cause and norm. The jazz combo theorist takes it to be a defining characteristic of sentences and their utterances that their contents are ­propositional. Partly in virtue of its propositional content, an assertion is subject to a­ ssessment as true or false, as warranted or unwarranted. An assertion is the sort of thing for which reasons can be demanded or given and which can itself stand, if true, as a reason for still other assertions. It would seem to be part of the very nature of that which is propositionally contentful to be subject to such assessments. But such assessments, the jazz combo theorist now adds, are intrinsically normative. To characterize an assertion along such dimensions as truth or falsity, being warranted or unwarranted, or being reasonable or unreasonable, the thought goes, is to assign a normative status or significance to the relevant assertion. Such assessments have to do not with mere causal histories but with the propriety or impropriety of such performances. As Brandom (1994), for example, puts it: Our ordinary understanding of states and acts of meaning, understanding, intending, or believing something is an understanding of them as states and acts that commit or oblige us to act and think in various ways. To perform its traditional role, the meaning of a linguistic expression must determine how it would be correct to use it in various contexts. To understand or grasp such a meaning is to be able to distinguish correct from incorrect uses. . . . To say this is in no way to deny that occurrences of intentional states of meaning, understanding, intending, and believing have causal significances. It is simply to point out that understanding them as contentful involves also understanding them as having normative significances. (13–​14)

Against Jazz Combo Theories  63 It is not just the thought that the ascription of propositional content is intrinsically and directly tied up with proprieties of use, rather than with causal histories, that leads to the charge that bottom-​up theories face a gap between cause and norm. Nor should we think, the jazz combo theorist will insist, that normative proprieties can simply be reduced to anything like causal histories. Causal histories, she will say, can play no role in endowing sentences with the pattern of proprieties marked by the ascription of propositional content. Causation has to do not with normative proprieties, not with what ought to happen, but merely with what does happen. As such, no appeal to mere causal connections between words and things could bridge the gap between cause and norm. Such connections do nothing to either explain or constitute the proprieties of use that that are marked by the ascription of propositional contents to sentences. To quote Brandom (1994) again: The “ought” involved in saying that a stone subject to no other forces ought to accelerate toward the center of the earth at a rate of 3 feet per second shows itself to have the force of an attribution of causal necessity by entailing that the stone will so act. The claim that it in this sense ought to behave a certain way is incompatible with the claim that does not do so. In contrast, no such entailment or incompatibility is involved in claims about how we intentional agents ought to behave, for instance what else one of us is committed to believe or to do by having beliefs and desires with particular contents. (31)

It is, of course, a very large issue whether the normative can be reduced to anything non-​normative. Though I argue elsewhere that the normative ultimately just is part of the natural order—​as long as humankind and the doings of humankind are part of that order too—​nothing I say here in this chapter depends directly on the reducibility of the normative to the natural.3 I concede, if only for the sake of argument, that in initial concept, at least, there is a distinction between a causal “must” and a normative “must.” I further concede that to the extent that propositional contents do directly and intrinsically encode normative proprieties of use, it follows that any bottom-​up theory shoulders the burden of somehow bridging the cause-​norm gap. After all, the ultimate explanatory target of any bottom-​up theory of reference will not be reference itself but the propositional contents of the sentences of our 3 But see Taylor (2015), Taylor (2003) and Taylor (in progress).

64  Referring to the World language. A theory of reference is of ultimate explanatory interest to the extent, and only to the extent, that reference plays a role in determining and explaining the propositional contents of sentences. It may therefore seem fair to require that a bottom-​up approach must show itself to have explanatory payoffs at the level of the sentence. And if that is right, then so as long as we accept that propositional contents directly and intrinsically mark or encode normative proprieties of use, then unless the causal theorist can show that and how bare facts about causal connections have explanatory payoffs for such proprieties, her theory will be an idle wheel spinning. A first response on behalf of a bottom-​up approach might be to point out that it is not altogether clear that assessments of linguistic performances as true or false, for example, really are per se normative. Indeed, it is easy to see how the normative significance of an assertion might come apart from its status as true or false. Suppose we distinguish, at least in principle, between the normative status of being correctly asserted in a given context, from the simple descriptive property of being true. Though there could be a normative principle which states that all and only true sentences enjoy the normative status of being correctly asserted, it is equally possible for there to be normative principles which count some true sentences as incorrectly asserted and some false sentences as correctly asserted. If that were so, then truth as such would be neither necessary nor sufficient for the normative status of being correctly asserted. Nor does this possibility seem far removed from our actual practices. The norms governing cooperative conversation seem, for example, not to count tautologies as correctly asserted, at least not in any given context, simply because they are true. Nor do they count sincerely and reasonably believed falsehoods as incorrectly asserted, when they are asserted, simply because they are false. The point is just that the truth or falsity of a sentence may be to some degree independent of its normative status as correctly or incorrectly asserted in a given context or on a given occasion. A sentence’s normative standing as correctly or incorrectly asserted might be thought to depend on a variety of factors—​including a set background of practices, interests, and aims. But suppose that we make an even further concession. Suppose that we go so far as to concede that norms of truth-​telling or truth-​seeking do govern certain cooperative enterprises and concede that to the extent that one cares about the aims of such enterprises, one ought indeed to speak and seek the truth. Still, if one is indifferent to those aims, it may be a matter of no

Against Jazz Combo Theories  65 normative significance whether one’s utterances are true. If the normative significance for us of truth and falsity are dependent in this way on our own possibly contingent cares and concerns, then it still does not yet follow that propositional contents as such directly encode or mark proprieties of use. The point of the preceding arguments is to show that it is not inevitable that propositional contents be viewed as directly encoding or marking normative proprieties of use. If not, then we need not necessarily suppose that a bottom-​up explanatory theory must serve somehow to directly bridge the cause-​norm gap. A bottom-​up theory need not purport to provide the normative foundations of a total theory of either the dynamics of rational inquiry or of the nature of cooperative communication more generally. Hence, even if we grant that the objective representational contents that a bottom-​up theory purports to partially explain do, in the context of various games of rational inquiry or cooperative communication, achieve a certain liability to normative assessment in virtue of the character and purposes of rational inquiry and/​or cooperative communication, it does not follow that a theory of objective representational contents must deliver contents that are per se normative. Consider the following vivid illustration of the way in which propositional contents might fail to directly encode or mark proprieties of use. Imagine a language game called “exaggeration.” In the game of exaggeration, players attempt to outdo one another at defending more and more wildly exaggerated claims. Moves in exaggeration are propositionally contentful and they do enjoy certain normative statuses. Some moves are permissible; other moves are not. Some moves outdo or exceed previous moves; other moves fail to exceed previous moves. Some moves are winning moves; others are losing moves. Now a theory that seeks to explain where the propositional contents of moves within the game of exaggeration first come from clearly need not directly entail anything at all about the normative status of this or that move within the game of exaggeration. The propositional contents of moves within the game are antecedent to and independent of the normative statuses of those moves. Mutatis mutandis, one might think, for moves within games of rational inquiry or cooperative communication. Though moves within such games may indeed enjoy both normative significance and propositional content, a theory of where those objective representational contents ultimately come from need not directly explain normative statuses within such games. The jazz combo theorist is likely to complain that these last remarks miss a crucial and deep point. The game of exaggeration is a wholly optional

66  Referring to the World game, played with sentences and utterances that have somehow already been endowed with propositional contents. These propositional contents are not coterminous with normative statuses within the game of exaggeration. Sentences and utterances do not first achieve their propositional contents through the play of the game of exaggeration. And it does not follow from the fact that some language games are in this way pragmatically and semantically posterior to the determination of objective representational content that all language games are. Brandom, for example, has insisted that there are certain bedrock or “downtown” language games within which sentences first achieve propositional content. He claims that to opt out of our bedrock content-​ conferring practices would be to opt out of thought and language altogether. Within these bedrock linguistic or discursive practices, he seems to believe, normative status and propositional content are so thoroughly intertwined as to be inseparable. Thus Brandom (2000) says: Language (discursive practice) has a center; it is not a motley. Inferential practices of producing and consuming reasons are downtown in the region of linguistic practices. Suburban linguistic practices utilize [sic] and depend [sic] on conceptual contents forged in the game of giving and asking for reasons, are parasitic on it. Claiming, being able to justify one’s claims, and using one’s claims to justify other claims and actions are not just one among other sets of things one can do with language. They are not on a par with other “games” one can play. They are what in the first place makes possible talking and therefore thinking: sapience in general. Of course, we do many other things as concept users besides applying concepts in judgment and action and justifying those applications. But  .  .  .  those sophisticated late coming linguistic and more generally discursive activities are intelligible only against the background of the core practices of inference-​and-​assertion. (14–​15)

Presumably within such bedrock practices, sentences enjoy a sort of foundational semantic priority. Jazz combo theorists take bottom-​up approaches, including causal approaches to reference, to be obvious non-​starters precisely because they take it that there are certain bedrock content-​conferring linguistic practices such that (a) these practices involve games played with whole sentences; (b) within these practices contents are conferred de novo; and (c) within these practice, content is coterminous with normative status or significance.

Against Jazz Combo Theories  67

4.  Jazz Combo Theories and the Social-​Dialectical Nature of Objectivity This bundle of views is given a certain unity and coherence because of the fact that jazz combo theorists take the constitution of propositional content to be a largely social-​dialectical matter, in very much the manner of our initial metaphor. As such they take the conferring of content to be intrinsically tied up with the social-​dialectical game of “giving and asking for reasons,” as Brandom puts it. Jazz combo theorists tend to see the supposed normativity of content in terms of our answerability to rational pressures emanating from fellow co-​inhabitants of a normative or discursive community, rather than in terms of direct answerability to a mind-​independent world. To be sure, jazz combo theorists may acknowledge that beliefs are caused by the inward rush of an external world upon the portals of sensation, as I put it in Chapter 2.4 What jazz combo theorists deny is that our bare causal commerce with the world suffices for the de novo constitution of content. Bare causation can do nothing to render thought and talk semantically answerable to the world with which we have causal commerce. Witness, in this regard, Davidson’s enigmatic claim that nothing justifies a belief but another belief. Davidson makes this claim in the course of defending what he calls a “coherence theory of truth and knowledge.” I will not try to unpack the full content and force of that theory here. Part of what Davidson seems to have meant was to deny that our beliefs, at least taken

4 Davidson (1981), for example, famously grants to the deliverances of sensation a role in causing beliefs, while denying them any role in justifying our beliefs. As he puts it: Sensory stimulations are indeed part of the causal chain that leads to belief, but cannot without confusion, be considered to be evidence, or a source of justification, for the stimulated belief. (151) To be sure, Davidson does allow that our beliefs are not typically “about” either sensations or experiences but about free-​standing distal objects. And he apparently takes free-​standing distal objects to be the ultimate causes of about belief. For in the following quoted passage, he goes on to say: We must, in the plainest and methodologically most basic cases, take the objects of a belief to be the causes of that belief. And what we, as interpreters, must take to be is what they in fact are. Communication begins where causes converge: your utterance means what mine does if belief in its truth is systematically caused by the same events and objects. (151) Nonetheless, Davidson seems to want to deny that causal relations between beliefs and their objects in any way serve to explain what aboutness or answerability to the world consists in. Davidson seems to think, in fact, that though the very idea that our beliefs are objectively true is tied up with our interpretive practices there is nothing substantive to say in other terms about just what objectivity comes to. Davidson altogether rejects, in other words, the very explanatory aim that motivates bottom-​up explanatory accounts of objective representational content.

68  Referring to the World one-​by-​one, are directly semantically answerable to a mind-​independent world. He says: Nothing, however, no thing, makes sentences and theories true:  not experience, not surface irritations, not the world, can make a sentence true. That experience takes a certain course, that our skin is warmed or punctured, that the universe is finite, these facts, if we like to talk that way, make sentences and theories true. But the point is better put without mention of facts. The sentence “My skin is warm” is true if and only if my skin is warm. Here there is not reference to a fact, a world, an experience, or a piece of evidence. ( Davidson, 1974, 194)

Davidson does not, I  take it, intend to be denying that there is a mind-​ independent world. He is even willing to admit there is a certain sense in which our beliefs can be said to be true or false of the world. But he takes great pains to deny that there is any substantive metaphysical relationship between mind and world or between our words and the world in virtue of which our words and our thoughts can be said to “answer” to the world when they are true. We have already seen that he denies the need to posit a substantive reference relation between singular terms and objects in the world to explain how our thoughts and words manage to be true or false of a mind-​ independent world. Likewise, he insists that we do not need to think of whole sentences or whole thoughts as standing in any substantive semantic relation to various local chunks of the world—​like facts or states of affairs. One upshot of this way of thinking is to deny the very possibility that we might explain the semantic in terms of merely causal relations between words and things. It takes something more than the mere causal rush of energy upon the portal of sensations to enable thought and talk to reach out to the world. Indeed, jazz combo theorists tend to take relatedness to others to be explanatorily prior to relatedness to the world. Our words and thoughts achieve the status of being true of the world, because we ourselves are somehow answerable, in the first instance, to one another.5 Brandom, for example, insists that objectivity is a “structural aspect” of what he calls the “social-​perspectival form of conceptual contents.” As he puts it:

5 It is something like this belief which is behind what Brandom calls his “deontic scorekeeping” model of discursive practice, according to which the bedrock or downtown content constituting practice is what he calls the Socratic game of giving and asking for reasons.

Against Jazz Combo Theories  69 The permanent possibility of a distinction between how things are and how things are taken to be by some interlocutor is built into the social-​inferential articulation of concepts. The distinction is in the first instance available to each scorekeeper regarding the commitments of others, of those to whom the scorekeeper attributes commitments. . . . Although grounded in essentially social, other-​regarding scorekeeping, however, the possibility of a distinction between how things actually are and how they are merely taken to be by some interlocutor remains a structural feature, even . . . in the case of attributions to oneself. (Brandom, 1994, 597)

In a similar vein, Davidson says: The ultimate source (not ground) of objectivity is, in my opinion, intersubjectivity. If we were not in communication with others, there would be nothing on which to base the idea of being wrong, or therefore, of being right, either in what we say or in what we think. The possibility of thought as well as of communication depends  .  .  .  on the fact that two or more creatures are responding, more or less simultaneously, to the input from a shared world and from each other. (Davidson 1997, 83)

Now Davidson does concede to causation a role in determining the contents of at least some of our thought and talk. But he stresses again that it is only in the context of what he calls triangulation, which he takes to be a kind of social interaction and coordination, that causation can make a difference to content at all. Thus Davidson (1999) says: Social interaction, triangulation, also gives us the only account of how experience gives a specific content to our thoughts. Without other people with whom to share responses to a mutual environment, there is no answer to the question what it is in the world to which we are responding. The reason has to do with the ambiguity of the concept of cause. It is essential to resolve these ambiguities, since it is, in the simplest cases, what causes a belief that gives its content. In the present case, the cause is doubly indeterminate: with respect to width, and with respect to distance. The first ambiguity concerns how much of the total cause of a belief is relevant to content. The brief answer is that it is the part or aspect of the total cause that typically causes relevantly similar responses. What makes responses relevantly similar in turn is the fact that others find those responses similar; once more it is the

70  Referring to the World social sharing of reactions that makes the objectivity of content available. The second problem has to do with the ambiguity of the relevant stimulus, whether it is proximal (at the skin, say) or distal. What makes the distal stimulus the relevant determiner of content is again its social character; it is the cause that is shared. The stimulus is thus triangulated; it is where causes converge in the world. (129)

It is, I think, plausible enough that the very idea of objectivity is tied up with the idea that different thinkers might simultaneously have different perspectives on one and the same world. To acknowledge this much is to acknowledge that the very distinction between how things are taken to be by this or that thinker and how things actually are in themselves, independently of our takings, itself depends on the idea that there are multiple, simultaneous perspectives that remain, nonetheless, commensurable, despite their diversity. Viewed in this context, talk of how the world is “in itself ” would seem to amount, at least in part, to a shorthand way of talking about the coordination among the totality of distinct, simultaneous, but still commensurable perspectives.6 I have no need to deny that the very idea of objectivity has a certain social-​ dialectical significance for the practice of adjudicating among the competing claims of those who stand in discursive community one with another. As I indicated in the previous chapter, I hold that discursive community has an important role to play in making objective representational content possible. But conceding the social-​dialectical significance of the very idea of objectivity within discursive communities does not suffice, on its face, to either answer or silence a certain distinctively metaphysical question about the ultimate source and nature of objective representational content. We want to know how possibly something lying entirely on the side of the thinking subject—​or even communities of such subjects—​could be semantically answerable to that which is largely independent of mind and subjectivity. To be told that certain social-​dialectical practices play distinctive roles in structuring and sustaining our thought and talk as of a mind-​independent world, leaves the question of how possibly our thought and talk manage to 6 On the other hand, some jazz combo theorists, like Richard Rorty (1979), at least seem at times to want to give up on the very idea of objectivity, even construed in social-​dialectical terms. He does have room for a notion of “solidarity” that involves a kind of intersubjectivity. It is fair to wonder whether the Davidsonian and Brandomian social-​dialectical notion of objectivity amounts to anything more than Rortyesque solidarity. See also Price (2003).

Against Jazz Combo Theories  71 be actually semantically answerable to a mind-​independent world entirely unaddressed. It is one thing to be told that in the act of interpreting one another we take ourselves to be jointly answerable to a mind-​independent world. It is another thing entirely to be told what exactly such answerability could possibly consist in. And it is still another thing to be told whether we are entitled to take our thought and talk to be semantically answerable to a mind-​independent  world. It is tempting to read jazz combo theories as, in effect, eschewing the very coherence of these last two questions. To read jazz combo theorists this way is to read them as expressivists of sorts, for whom the content of the very idea of objectivity is wholly exhausted by its social-​dialectical role.7 If the social-​ dialectical role of the idea of objectivity is all that confers significance upon that idea, it may seem to follow that there can be no stepping outside of a certain social-​dialectical practice in hopes of “vindicating” that idea. And if that is right, then the project of externally vindicating our idea of objectivity would be a non-​starter. It would be a non-​starter just because there is no neutral vantage point from which we might hope to simultaneously regard thoughts, our words, and the mind-​independent world in order to assess what sort of relations they stand in, or fail to stand in, to one another. I own up to hankering after a substantive metaphysical account of what in the order of things the references relation is. I want to know what in nature constitutes and determines the semantic relation between our words and/​ or thoughts and the mind-​independent world. But it is just that sort of thing that our imagined expressivist says is neither needed nor attainable. I concede to the expressivist that there is no neutral standpoint, sitting entirely outside our ongoing discursive practices, from which we might hope to decisively vindicate, once and for all, the very idea of objectivity. But that concession does not directly imply that our hankering must ultimately come to naught. What is true is that our ultimate account of what objective representational content consists in must be couched in a semantic meta-​vocabulary that is available to us from a stance of critical self-​reflection upon the very practices that give the very idea of objectivity its point. Perhaps we cannot stand entirely apart from those practices, but that gives us no reason to deny that through a process of critical self-​reflection upon those practices we might gain richer imaginative acquaintance with the very possibility that our 7 Cassierer (1923) seems to have thought something like this about objectivity. See especially ­chapters 5–​6 of Substance and Function.

72  Referring to the World thought and talk might make semantic contact with a world largely independent of mind.

5.  Some Conceptual Tools and Distinctions Over the course of the current section, I introduce, or reintroduce, a bundle of conceptual tools and distinctions, each of which plays a significant role in the closing argument of this chapter. My argumentative aim is to show that it is at least conceivable that there could be something substantive in the order of things that constitutes reference. My argument will consist partly of blocking maneuvers and partly of enabling maneuvers. Both the blocking maneuvers and the enabling maneuvers depend on distinguishing among various senses of priority and arguing that although sentences enjoy a certain form of priority over their constituents, they do not enjoy other forms of priority over their constituents. In particular, I will argue that although sentences enjoy dynamic or pragmatic priority in the theory of illocutionary force, sentences are syntactically correlative with their constituents, and semantically posterior to their constituents. The blocking maneuvers are intended to show that dynamic or pragmatic priority—​the sole form of priority enjoyed by sentences over their constituents—​does not entail the falsity of bottom-​up approaches to objective representational content, because it does not entail that sentences enjoy semantic priority over their constituents. It is precisely the consistency of the triad consisting of dynamic priority, syntactic correlativity, and semantic posteriority that enables us to see how possibly a bottom-​up explanatory theory might be true. It should be said that my argument does not attempt to establish that this possibility is actual. That would take an argument of a different kind, one that is less speculative and foundational and more constructive. I am in no position to offer a positive constructive account of the bottom-​up constitution of content at the moment. Instead I offer a speculative argument designed mainly to increase our imaginative acquaintance with the possibility that reference is a substantive, real relation that has some place or other in the vast and layered labyrinth of existence in its totality. The goal is not yet to precisely locate reference within that labyrinth. That is a thing that a priori philosophical arguments are hardly ever fit to do, at least not on their own.8 8 See Taylor (2019)

Against Jazz Combo Theories  73

5.1.  Dynamic Priority versus Semantic Priority Linguistic performances involve at least two distinguishable factors—​ propositional content and illocutionary force. The illocutionary force of an utterance can be partly explained by appeal to the notion of what is sometimes called the direction of fit between content and world.9 The content of a promise typically neither fits, nor purports to fit, the way the world stands antecedent to the act of promising. If I promise to come to your recital tomorrow, then the representational content of my promise involves my being at your recital, to be sure, but I don’t thereby represent the world as currently “fitting” that representational content. I do, however, thereby represent myself as intending to bring it about that the world will eventually come to fit that content at the appointed time and that it will come to do so as a consequence of something I undertake to do.10 A sincere assertion, by contrast, does purport to represent the way the world independently stands. If I assert that the recital is scheduled to happen tomorrow, when it isn’t so scheduled, I have misrepresented how the world stands. If I promise to come to the recital tomorrow, but intend not to show up, I have misrepresented myself. Any adequate bottom-​up theory of reference will have more or less direct ramifications for the content side of the force-​content complex. But such a theory need have no direct implications for the theory of force. Theories of reference play a role in explaining content, but no role whatsoever in explaining force. Precisely because the theory of reference plays no role in explaining force, bottom-​up theories of reference are entirely consistent with the dynamic priority of the sentence over the word. To acknowledge the dynamic priority of the sentence over the word is to grant that the performance of a determinate illocutionary act, with a determinate illocutionary force, typically involves the production of some sentence or other. It is to acknowledge that linguistically significant tokenings of subsentential expressions—​ tokenings that carry significance for the propositional contents of particular 9 See Anscombe (1963) and Searle (1984) for two classic expositions of the notion of direction of fit. 10 There is a use of “I promise” in which what one is doing is offering one’s assurance that a certain state of affairs (currently) obtains as in the following discourse: A: I’ve looked in the cellar, but I didn’t see any more beer there. B: But there is more beer there. I promise you. Did you look on the shelves in the far corner? In this little dialog, B is offering A not a commitment to a future action, but a kind of epistemic assurance that the proposition that there is more beer in the cellar is true.

74  Referring to the World illocutionary acts—​typically occur only in the performance of some determinate and propositionally contentful illocutionary act. A speaker typically refers to Socrates by tokening the name “Socrates” in the course of performing some illocutionary act or other via the production of some “Socrates”-​ involving sentence or other.11 Illocutionary acts—​acts of asserting, promising, requesting, inquiring—​performed with sentences are not built up out of “shorter” illocutionary acts first performed with individual words. Indeed, though words do make direct contributions to the propositional contents of illocutionary acts performed with sentences containing them, individual words typically make no independent contributions to illocutionary force. As far as illocutionary force is concerned, then, any bottom-​up theory of force would be a complete and obvious non-​starter. We have just distinguished dynamic priority from semantic priority. But it may be conjectured by the jazz combo theorist that the dynamic priority of the sentence with respect to force somehow entails or presupposes the semantic priority of the sentence with respect to content. In fact, most extant arguments in favor of the semantic priority of the sentence seem to begin with the observation that sentences enjoy what I am calling dynamic priority and conclude on that basis that sentences must ipso facto enjoy semantic priority. I will not claim that nothing can be said for this sort of argument. Indeed, I concede that there is a degree of at least initial tension between the claim that the sentence is dynamically prior while the constituent is semantically prior. It is not altogether implausible to maintain that a name, say, first achieves determinate reference by being tokened in the course of utterances of some sentence(s) or other. Plausibly, one has to do something—​something language involving—​in order to fix a reference. It is hardly sufficient to merely mouth a name in the physical presence of the object. One would seem to have to have and express, at a minimum, certain reference-​fixing intentions. And perhaps these are plausibly expressed in uttering a sentence rather than just a bare word.

11 To be sure, there are circumstances in which the production of a name constitutes the performance of a determinate illocutionary act. Upon seeing Socrates enter the room, one can greet him merely by uttering his name. Similarly, one may call out to Socrates for help, request that Socrates come, express one’s pleasure or surprise at Socrates’ arrival, offer a fact about Socrates in answer to a “who” question simply by uttering his name with an appropriate intonation. Such non-​sentential linguistic performances would seem to require an appropriate and special background. Consequently, the possibility of such non-​sentential performances may well be less fundamental than the possibility of sentential linguistic performances. But this fact poses no special challenge for the causal theorist. See Perry (1994).

Against Jazz Combo Theories  75 This way of thinking may seem to lead directly to Brandom’s (1994) dictum that semantics must answer to pragmatics and thereby to undercut the semantic priority of the sentence. But before drawing any such inference, we need to distinguish claims about the dynamics of reference determination from claims about the order of semantic dependency. For it is equally intuitively compelling that unless and until the constituents of a sentence are rendered semantically significant, the sentence itself can enjoy no fully determinate and complete semantic significance of its own. If the semantic values of sentences in no way depend on the (independently determined) semantic values of their constituents, it is hard to see where sentential semantic values could possibly come from. And it is also hard to see why substitution of one constituent for another, with the same or different semantic value could possibly affect the semantic value of the whole.12 I take these two plausible feeling claims—​the claim that sentences have priority even in the dynamics of reference determination, and the claim that constituents are prior in the semantic order—​to be claims about different, though related matters. Though it is not altogether obvious that they need be viewed as incompatible, they do seem to pull our initial intuitions in somewhat opposed directions. One gives pride of place to the sentence even in reference determination. The other gives pride of place to the constituent. When one is faced with two equally plausible principles that at least appear to be in tension, even if not outright conflict, one really would like to make it more explicit how one can comfortably have it both ways. And that is precisely what I shall try to pull off in the remainder of this chapter. I shall be arguing that the key to reconciling any apparent tension or conflict, between the dynamic priority of the sentence and the semantic priority of the constituent, is the notion of what I call “semantic bootstrapping.” Semantic bootstrapping happens when one or more partially contentful sentences are tokened in such a way as to yield semantic values for their not yet semantically valued constituents and thereby for the very utterances containing those expressions as well. But doing this will take a considerable bit of stage setting. 12 Though I will not argue the point here, I must say that the jazz combo view of the order of metaphysical and explanatory priority between sentences and their subsentential constituents strikes me as frankly incoherent at bottom. Just how and why, on this view, is it possible for utterances of particular sentences to constitute performances of peculiar and determinate speech acts? Suppose I ask why it is that by uttering the sentence “Claire is puzzled”; I manage to assert of Claire and no one else that she is puzzled. Try as they might, defenders of the semantic priority of the sentence will seem unable to provide a compelling explanatory answer to this question.

76  Referring to the World

5.2.  Pragmatics and Partiality The idea that some meaningful sentences may be only partially semantically valued should be familiar enough. Many contemporary referentialists hold, for example, that sentences containing empty names express so-​called gappy propositions or, as I myself prefer to call them, “propositions-​in-​waiting.”13 A sentence that expresses a proposition in waiting is merely partially semantically valued in a pretty straightforward sense. Because it contains one or more “gaps” where propositional constituents would otherwise go, a proposition-​ in-​waiting is not yet a (full) proposition. Sentences expressing such are, in an obvious sense, semantically incomplete. Just as, say, a would-​be automobile, with too many missing parts isn’t really and completely an automobile just yet—​but also isn’t, for that reason alone, a non-​entity—​so, a would-​be proposition with gaps where constituents might eventually go isn’t really a full-​ blown proposition. That’s precisely why I call them propositions-​in-​waiting. Because they contain gaps where actual propositional constituents are called for, propositions-​in-​waiting don’t have fully determinate truth conditions. A  sentence that expresses a mere proposition-​in-​waiting can’t be used, at least not strictly and literally, to make a fully determinate claim about how the world is. Now the crucial point for our current purposes is that despite the fact that they are devoid of both truth conditions and truth-​value, semantically incomplete sentences that express propositions-​in-​waiting may, nonetheless, serve important communicative functions. I return to argue at greater length for this claim in Chapter 6, where I take up issues about non-​referring singular terms and the many games we play with empty singular terms. There I will focus both on what I call non-​veridical language games and on what I call one-​and-​a-​half stage pragmatics and show examples of the communicative significance of semantically incomplete, merely partially valued sentences. Another form of communicative significance for merely partially valued sentences—​what I call semantic bootstrapping—​is introduced later in this chapter. Because I claim that merely partially valued sentences can be used with communicative significance in a fair number of ways, I trust that my appeal to semantic bootstrapping will not seem arbitrary or ad hoc. 13 The literature on so-​called gappy propositions is by now extensive. For a representative sample, see several of the essays collected in Taylor (2003), Taylor (2010), Taylor (2014), Braun (1993, 2000), Adams and Detrich (2004), Adams, Stecker, and Fuller (1997) For a dissenting voice, see Everett (2003). I spell all of this out more fully in Chapter 6.

Against Jazz Combo Theories  77 By way of fleshing out the potential communicative significance of merely partially semantically valued sentences just a bit more, let me say a little more about one-​and-​a-​half stage pragmatics—​though I  give a fuller treatment in Chapter  7. One-​and-​a-​half stage pragmatics externalities are typically generated on the journey up from not fully saturated sentence meanings, in the sense of Recanati (2004), to contextually determined propositional contents. This may happen in two different ways: either when so-​called primary pragmatic processes like saturation fail to come off or as contextually generated by-​products of successful saturation. The contents generated by one-​and-​a-​half stage pragmatics do not serve to “saturate” the gaps in a mere proposition-​in-​waiting—​and that is why I call them “externalities.”14 Such externalities are generated on the side. Because they are externalities that are not themselves the direct consequence of saturating unsaturated slots, such externalities are less intimately associated with the sentence and its constituents than situationally saturated propositional contents are. For example, they typically are neither entailed nor presupposed by fully saturated contents, even when such contents are situationally generated. On the other hand, they are more intimately associated with the sentence and its meaning than Gricean implicatures are. Gricean implicatures are generated post-​ propositionally—​ that is, post-​ saturation—​ via so-​ called secondary prag15 matic processes. Since such processes come into play only after full-​blown propositional contents have been fixed via saturation, Gricean implicatures are less intimately associated with the sentence and its meaning than my one-​ and-​a-​half stage externalities are.

5.3.  Objectual versus Objective Representations Again I first introduced the distinction between merely objectual and fully objective representations in Chapter 1, developed it further in Chapter 2, and will have occasion to turn to it again in subsequent chapters. Since I have said a fair amount about the distinction already and will say even more soon, I can afford to be relatively selective here. Up to this point in the overall argument, I have leaned on this distinction in the service of two related points. I have

14 See Taylor (2003), especially essays 6 and 9, Taylor (2007), Taylor (2014) and Chapter 6 in this volume or further exploration of the notion of a one-​and-​a-​half-​stage pragmatic externality. 15 For the distinction between primary and secondary pragmatic processes, see Recanati (2004).

78  Referring to the World claimed that objectuality or referential fitness is a precondition for referential success, and I have suggested that objectuality helps to explain the epistemic one-​sidedness of reference. Here, I need the distinction as a plank in my argument that the dynamic priority of the sentence does not entail the semantic priority of the sentence. On the assumption that I can successfully carry off that argument, the distinction between objectuality and objectivity will have helped us to see both what is true and what is false in jazz combo theories of meaning and reference, at least to the extent that such theories turn on arguments having to do with the priority of the sentence. What the jazz combo theorist correctly perceives—​though through a glass darkly—​is the fact that the referential fitness is a holistic property. Referential fitness accrues not to representations taken one by one, in isolation from other representations, but as a result of their roles in interlocking systems of representations. No isolated representation, just taken on its own, independently of any connections to any other representations, can be “fit” for the job of referring to or standing for an object. No expression even has standing as a name except in virtue of playing the right kind of role in an entire system of interlocking linguistic representations. This thought in and of itself will no doubt seem quite amenable to the jazz combo theorist, even though it is only a thought about the significance of referential fitness and not yet a thought about referential success. But once we move beyond mere referential fitness to referential success, we can still acknowledge that the jazz combo theorist has grasped an important truth. That is because it follows from the claim that referential fitness is a precondition of referential success, that no object can be successfully designated or referred to except by an expression that occupies a role in a system of interlocking representations. It is worth restating in brief compass the broad outlines of the overall picture that emerges if these claims are accepted. It is a picture according to which the constitution of objective representational content is the joint work of mind and the world. Objective representational contents are neither simply thrown out at the world from the side of mind nor merely impressed upon the mind from the side of the world. Objective representational content happens only where the antecedent representational innards of referring minds, typically arrayed together in discursive community, meet the relentless rush of energy from without upon the portals of sensation. Absent this inward rush, we might still have form of thought as of objects, but our thoughts would be devoid of semantic contact with

Against Jazz Combo Theories  79 any real existents and thus devoid of objective representational content. On the other hand, absent the internal, fitness-​making factors, the inward rush would amount to little more than energized, but semantically inert to’ing and fro’ing. Though the world is awash in information, flowing every which way, only in very special corners of the universe does the flow of information give rise to reference. Reference is the work of a distinctive kind of thing—​representations, linguistic and mental, that enjoy antecedent referential purport. Reference happens only when information flowing inward encounters and is encoded by representations that have been rendered antecedently ready to refer. This rough picture, which no doubt needs much greater refinement, allows us to grant that there are important grains of truth contained both within the broader doctrine of semantic holism, and in Wittgenstein’s pithy but opaque remark that nothing has so far been done when a thing has merely been named. What this approach gets right is the fact that reference is not the business of referring expressions taken one-​by-​one. Reference happens only where referring expressions play quite particular roles—​about which we shall say more in Chapter 4—​in certain sorts of systems of representations. I part company with jazz combo theorists, however, when it comes to the lessons that are to be drawn from these grains of truth. The jazz combo theorist wrongly concludes on the basis of such starting points that the sentence must be semantically prior to its constituents. But this, I shall argue, is a mistake. Though sentences are pragmatically or dynamically prior to the constituents which they contain, the constituents are semantically and explanatorily prior to the sentences which contain them. And nothing we have conceded to the jazz combo theorist gives us any reason whatsoever to deny this. Or so I shall try to argue in the remainder of this chapter.

5.4.  Syntactic Correlativity of the Sentence and Its Constituents The syntactic or structural correlativity of sentences and their constituents is closely connected to the notion of an objectual representation. Indeed, objectuality and syntactic correlativity might be said to be two sides of the very same coin. Even so, syntactic correlativity is worth separate mention, since it plays a distinctive role in making semantic bootstrapping possible. Semantic bootstrapping, which we shall turn to next, is the key to seeing why

80  Referring to the World the dynamic priority of the sentence does not entail the semantic priority of the sentence. We begin by considering the linguistic category NAME. Viewed from a syntactic or structural perspective, of the sort that I shall elaborate and defend more fully in Chapter 4, names are nothing but invariants over a totality of potential sentences. The class of names is the class of expressions that interact with other expressions in characteristic ways in both the context of individual sentences and multi-​sentence discourses. For example, names combine with predicates to form sentences; they anchor anaphoric chains both within and across sentence boundaries; they make explicit the preservation of subject matter from tokening to tokening of the same name again in either the same or different discourses. What is true for the category NAME is true for other categories of expressions as well. The category SENTENCE is itself inextricably bound to categories like NAME, VERB, QUANTIFIER. From the syntactic or structural point of view, these are correlative notions, functionally interdefineable only as an entire system. If names, and singular referring expressions more generally, are nothing but recurrent sub-​sentential expressions that combine with predicates to yield sentences, then sentences are nothing but linguistic complexes, of a distinguished sort, built out of names and other sub-​sentential expressions. We have, that is, neither an independent hold on the syntactic category NAME nor an independent hold on the syntactic category SENTENCE. The crucial further point is that precisely because of the syntactic correlativity of sentences and their constituents, sentences are not first epistemically given to us as unstructured blocks. Rather, they are epistemically given as structured wholes, built out of discernible and repeatable parts. When I say that sentences are given as structured wholes, I do not mean to say that in the bare audition of a spoken sentence, say, we automatically recognize that a sentence, as opposed to a mere word or string of words, has been uttered. Nor do we automatically recognize what sentence has been uttered when we recognize that some sentence or other has been uttered. The claim is rather that the recognition that a sentence has been uttered is tantamount to the recognition that a complex whole built out of semantically significant parts has been produced. Though it may take further cognitive processing to discern exactly what those parts are and how they are structured, nonetheless, to perceive a sound stream as a sentence is to perceive it as a complex structured whole, built of semantically significant parts.

Against Jazz Combo Theories  81 Now there are strings that may appear to function as one-​word sentences, devoid of internal complexity. A language containing only such sentences would not exhibit syntactic correlativity of sentences and their constituents because such sentences would be devoid of constituent structure. Languages in which all sentences are unstructured blocks, devoid of constituent structure, may be barely logically possible. But it is surely non-​accidental that in all humanly possible languages, internally unstructured sentences are the exception rather than the rule. It is no accident that the grammar of any natural language generates a potential infinity of grammatical and meaningful sentences. Competent speakers potentially grasp, at least tacitly, such potentially infinite totalities. Given the finiteness of minds, the ability to grasp a potential infinity of sentences and sentence meanings must involve only the exercise of finite cognitive capacities. The only remotely plausible available hypothesis for explaining the very possibility of our grasping a potentially infinite totality of sentences and their meanings through the exercise of merely finite cognitive capacities is the hypothesis that the infinite totality of sentences and sentence meanings is compositionally generated from a finite store of primitive vocabulary items and a fixed set of recursive rules of syntactic combination and semantic composition. This widely endorsed answer is the only hypothesis that renders the capacity of a finite mind to grasp a potential infinity of sentences and their meanings non-​mysterious. The claim that the totality of sentence contents is compositionally generated is best understood as a substantive metaphysical conjecture. The conjecture is that sentences and their constituents do their semantic work partly in virtue of the way whole sentences are constructed, brick-​by-​brick, out of antecedently available materials in accordance with compositional semantic rules and combinatorial syntactic principles laid down by the grammar of the language to which they belong. The combinatorial nature of the syntax of the language is a direct expression of the syntactic correlativity of sentences and their constituents. It is certainly fair to wonder just where both the rules of semantic composition and the rules of syntactic combination come from and how they relate to one another. Are the principles of syntactic combination and semantic composition two entirely separate and independent domains or is one somehow derivable from the other? What about the lexicon? What role, if any, does it play in determining either the rules of syntactic combination or the rules of semantic composition? These are deep, important, and difficult questions that would need to be squarely addressed in any final account

82  Referring to the World of the metaphysics of meaning. Though I myself favor the UG16 hypothesis, I will not stop to address such questions in detail here. For the nonce, I will only say that I take grammars, including the lexicon, to be psychologically real in the sense that the existence and nature of the compositional principles that give rise to sentence meanings out of the meanings of their constituents is ultimately grounded in facts about human psychology. If it is right that the facts about grammar, including facts about semantic composition and syntactic structure, are grounded in facts about human psychology, the human mind itself turns out to be the ultimate source and site of semantic composition and syntactic combination. If so, then to describe the grammar of language is to give an abstract description of the mechanism by which the mind builds complex representations and their meanings out of antecedently given and meaningful simpler constituents.

5.5. Semantic Bootstrapping Earlier on, I briefly detailed a few ways in which merely partially semantically valued sentences may still be pragmatically significant. There is great deal more that can be said about these matters. And I shall take them up in more detail in Chapter 6. Here though I focus on another sort of pragmatically significant use of merely partially semantically valued sentences, what I call semantic bootstrapping uses. What is pragmatically distinctive about bootstrapping uses is that, in the right sort of speech situations and relative to appropriate backgrounds of issues, they may function to fix values for sentential constituents that do not yet enjoy semantic value. As such, semantic bootstrapping uses have the distinctive pragmatic function of enabling us to bootstrap our way out of semantic partiality. Imagine that there is a yet to be named dog present. Let what the dog is to be named constitute the background of issues relative to which bootstrapping utterances are to be pragmatically interpreted. Smith and Jones are debating whether to call the dog “Fido” or “Rover.” The following dialog ensues: Smith: Come here, Fido. You want to play fetch the stick, Fido? Jones: Don’t call him “Fido.” “Fido” is such a common name. You aren’t Fido, are you, boy? You’re Rover. Aren’t you Rover, boy?

16 From this point on, Taylor uses “UG” as short for “Universal Grammar” (Chomsky 1965, 2007).

Against Jazz Combo Theories  83 Smith: Watch him come, when I call him “Fido.” Hey Fido. Come here Fido. [The dog comes] Jones: Oh that proves, nothing. He’ll respond to “Rover” too. Come, Rover. Come to me. [The dog doesn’t come.] Smith: See, he won’t come. That’s because he’s not Rover. He’s Fido. Jones: Okay, then. Fido it is. You are now Fido, doggie. At the start of our little dialog, “Fido” does not yet refer, at least not semantically, to the dog that is soon to become Fido. Strictly speaking, then, the dialog’s initial sentences containing the name “Fido” are semantically incomplete and do not semantically express any fully determinate propositions. We have already seen, however, that to say that an utterance fails to semantically express any determinate proposition is not to deny that it pragmatically conveys some proposition or other. The crucial point, though, is that by the end of our little dialog, not just “Fido” itself but also sentences containing that name have come to have more nearly complete semantic contents.

6.  Closing Argument: Against the Semantic Priority of the Sentence Throughout the previous sections of this chapter, we have clearly conceded a great deal to the jazz combo theorist. To take just one example, in merely endorsing the syntactic correlativity of sentences and their constituents and maintaining that objectuality is not a property that expressions taken one-​by-​one enjoy on their own, independently of their roles in a system of interlocking representations, we have endorsed a kind of holism, at least at the level of syntax, if not yet at the level of semantics. But in taking objectuality to be a pre-​condition of referential success, we come closer still. For we thereby endowed holism at the level of syntax with the power to percolate up to what might be called the pre-​semantic interface between syntax and semantics. We have also conceded to the jazz combo theorist, that there is a kind of priority that sentences do enjoy of their constituents—​dynamic or pragmatic priority. To grant that sentences enjoy dynamic or pragmatic priority over their constituents is to grant that it is primarily via the production of sentences that significant linguistic acts are typically formed.

84  Referring to the World What still divides our view from that of the jazz combo theorists is the question of whether coherent sense can be made of the idea that sub-​sentential constituents are independently meaningful or have meaning only in a sort of derivative and secondary sense. We say that they are independently meaningful; the jazz combo theorist says that they are not. My closing argument against the jazz combo approach is that despite all of our concessions, we have no reason to assert and every reason to deny that constituents are meaningful in only a derivative and secondary sense. That argument turns on the set of distinctions introduced earlier on. These distinctions enable us to see where both concession and resistance are warranted. The basic thought is that one should no more suppose that the sentence is semantically and explanatorily prior on the basis of the dynamic priority of the sentence than one should suppose that events are ontologically prior to objects solely on the basis of the fact that objects participate in causal relations only as constituents of causally related events. The inference from dynamic priority of the sentence to the semantic priority of the sentence is invalid because the following triad is consistent: Dynamic Priority of the sentence Syntactic/​Structural Correlativity of constituent and sentence Semantic Priority of the constituent In fact, not only is the triad consistent, but it is precisely the combination of the dynamic priority of the sentence and the syntactic correlativity of the constituent and the sentence that make the semantic priority of the constituent possible in the first place. On the one hand, it is dynamic priority of the sentence that makes it possible to determine reference by means of boot-​ strapping uses of merely partially semantically valued sentences. On the other hand, the syntactic correlativity of sentence and constituent makes it possible for the reference of a constituent, once determined, to function as more than an idle wheel spinning. Because of syntactic correlativity, reference can be “percolated upward” into the propositions expressed by sentences in which the relevant expression occurs. Let us conclude by reviewing the bidding of our overall argument. Begin by recalling the earlier claim that the category name is, on our view, a role-​ oriented category. We said that no expression has standing as a name except in virtue of playing the right kind of syntactic role in a system of syntactically interlocking linguistic representations. We insisted that it is only through

Against Jazz Combo Theories  85 the playing of this role that representations are rendered “fit” for the job of standing for an object. We noted from this premise it follows that no isolated representation, all on its own and independently of its connection to other representations, can be “fit” for the job of standing for an object. It was a short step to the conclusion that no object can be successfully designated except by an expression that already occupies a role in a system of interlocking representations. Except that an expression is already referentially fit, neither causation nor any other extra-​representational fact would suffice to render it referentially successful. Reference is not something achieved by representations that stand in splendid isolation one from another Let w be an isolated string that plays no role in any interlocking system of linguistic representations and let occurrences of w be caused as regularly and as systematically as one likes by occurrences of o-​involving events. The obtaining of that causal relation would not suffice to constitute o as the referent of w. Precisely this kind of consideration typically leads jazz combo theorists to reject bottom-​up approaches to objective representational content. But the bare consistency of our triad helps us to see that bottom-​up approaches ought not to be rejected on such grounds alone. What is true is that the determination of reference involves an interaction of two independent factors. On the side of the representations themselves, there are the internal, role-​ oriented factors that make for referential fitness. There is also a contribution to be made to the constitution of reference and objective representational content more broadly, by extra-​representational factors of a broadly causal/​ informational nature. And it is this fact that renders a bottom-​up approach overwhelmingly plausible. Jazz combo theorists, with their commitment to the semantic priority of the sentence, have simply failed to appreciate the sort of division of labor that the constitution of reference and objective representational content involves. They have focused too heavily on that which falls on the side of language and on the structure of discursive community as exemplified in our controlling initial metaphor. They fail to appreciate how the structure of language itself opens up room for extra-​representational connections between word-​ involving sentence tokenings and object-​ involving events to be directly implicated in the constitution of reference and, eventually, of objective representational contents. Only a theory reference that simultaneously gives due weight to the dynamic priority of the sentence, the syntactic correlativity of sentences and their constituents, and the semantic priority of the constituent has the potential to explain how and why the two factors of reference determination only jointly suffice to constitute

86  Referring to the World reference. Though they have not been my target here, it also bears stressing that standard issue “one-​factor” causal theorists about reference make a complementary error. They focus too heavily on that which falls on the side of the world, and not enough on the interaction between worldly factors and structural cum syntactic factors lying on the side of the subject and/​or language. Given the dynamic priority of the sentence and the fact that names and other sentential constituents are themselves paradigmatically tokened in the context of the tokening of sentences, it follows that there could obtain reference constituting causal connections between a name n and an object o only if there obtain causal connections of some privileged sort between the tokening of some or all n–​involving sentences and some or all o-​involving events. Consequently, if one is to offer a bottom-​up, explanatory theory that assigns a central role in determining reference to causal/​informational relations between words and things, one thereby takes on the burden of explaining, in non-​question-​begging terms, and without presupposing that sentences have fully determinate propositional contents antecedently to the determination of reference for the constituents those sentences contain, just why the obtaining of causal/​informational connections of the privileged sort suffice to constitute o as the referent of n. Such was really the burden shouldered by bottom-​up theorists from the very beginning. It was never plausible that bare names, independently of their roles as sentential constituents, and bare objects could stand in reference constituting extra-​representational connections to one another. Such connections could make a reference-​constituting difference only relative to some background structure that prepared the way. But this should hardly be surprising. Just as objects participate in causal relations only indirectly by serving as invariants across causally connected object-​involving events, so too names participate in extra-​representational reference constituting relations and events by being tokened in the course of utterances of sentences. And I hope to have made a strong case for the claim that the fact that this is true is a more or less direct consequence of the consistent triad consisting of the dynamic priority of the sentence, the syntactic correlativity of sentences and their constituents, and the semantic priority of the constituent.

4 Puzzles of Coreference Theme and Variations

1.  The Many Coreference Puzzles In Chapter 2, I laid out some initial motivations for distinguishing referential fitness from full-​blown referential success. I suggested that there are certain inner factors that render our representations antecedently fit or poised to refer. I argued that these fitness-​making factors lie entirely on the side of the referring mind. I also claimed that they are broadly syntactic and role oriented. Qua syntactic, the internal fitness-​making factors are not in and of themselves content determining.1 As not yet semantic, though the internal factors do suffice to render our representations objectual and thus fit or poised to refer, they do not thereby render our representations referentially successful and thus do not yet serve to endow our representations with objective representational content. It simply takes more than objectuality to render our representations fully objective and thus referentially successful or to endow our thought and talk with objective representational content.2 Chapter 3 carried the argument for the centrality of the inner fitness-​making factors over to the metaphysics of reference and reference determination. It did so by assigning such factors a central role in our defense of the semantic priority of the constituents against the jazz combo theorist, who mistakenly insists that constituents have meaning in only some derivative and secondary 1 There are other “two-​factor” approaches according to which both factors are “semantic.” Two-​ factor approaches of this sort typically truck in some version of the distinction between so-​called wide content and so-​called narrow content. I have not yet explicitly argued against semantic two-​ factorism, as it might be called, since I have been more concerned with outlining my own version of two-​factor referentialism. 2 I need to say more about why the dispute between my own syntactic approach and semantic two-​factorism is not a merely verbal dispute. That discussion begins in earnest toward the end of this chapter and is carried further over the course of a number of subsequent chapters. But the first hints of my reasons for rejecting semantic two-​factorism should already be evident from what I had to say in Chapter 2 about the role of the inner fitness-​making factor in explaining the epistemic one-​ sidedness of reference.

Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0004

88  Referring to the World sense. The current chapter builds on the momentum of the previous two chapters by showing that two-​factor referentialism can also make significant headway on the many coreference puzzles. Now coreference puzzles have long been regarded as bedrock data against which the adequacy of any theory of singular reference must ultimately be tested. Frege was, of course, the founding father of the tradition of thought that saw explaining such puzzles as a primary condition of adequacy on a theory of reference. It was largely because he believed that no referentialist semantics for singular referring expressions could explain the coreference puzzles that introduced the distinction between sense and reference. As the founder of a tradition, it is perhaps not altogether surprising that Frege’s understanding of these matters still enjoys a certain pride of place. At least the spirit, if not the details, of Frege’s approach still enjoys status, the status of something like an entrenched default. I say the spirit rather than the details because one need not be a strict Fregean to allow that solutions to the coreference puzzles are likely to have far-​reaching significance for the proper understanding of not just the epistemic one-​sidedness of reference, but of reference. Even many referentialists, who have no truck with Fregean senses, or even semantic presentationalism more generally, have followed Frege in taking there to be a deep connection between the coreference puzzles, the epistemic one-​sidedness of reference, and the ultimate metaphysical nature of reference.3 To be sure, strict Fregeans, neo-​Fregeans, and flat-​out anti-​Fregeans are likely to have rather different understandings of the source and nature of both epistemic one-​sidedness and reference in general. Indeed, they are even likely to describe the phenomenon of epistemic one-​sidedness in rather different terms. Frege himself characterized the phenomenon in terms of a notion of “cognitive significance” which he himself never explicitly defined. Others talk in terms of differences in truth conditional contributions or differences in modal profiles or in even in terms of pragmatic or conversational dynamics. Hardly anyone has appealed to syntax in the way I shall 3 To be sure, because of what they see as the pitfalls of Frege’s own approach, they have often introduced semantic innovations of a rather far-​reaching and non-​Fregean character. I  have in mind such innovations as the unarticulated constituents of Crimmins (1992) and Crimmins and Perry (1989), the hidden indexicals of Schiffer (1977), the contextually variable restrictions on correlation functions between the Russellian Annotated Matrices of Richard (1990), the quasi-​singularizations of Recanati (1993), and even the relational semantics of Fine (2007). There is of course much to be said both for and against this raft of semantic innovations. But I shall not attempt any such detailed survey and evaluation here. My aim is not to directly refute all competing theories but to motivate, develop, and defend an approach rather different in spirit from most of them.

Puzzles of Coreference  89 do in this chapter. Bracketing for the nonce the question of precisely how coreferring terms may differ from one another which would justify our calling reference epistemically one-​sided, just the fact that they may differ at all is important enough to merit explanation. So perhaps all that needs to be said by way of the initial characterization of the phenomenon is that coreferring terms may differ in some yet to be fully specified way. That way, we don’t have to buy into any particular theory of the nature and source of epistemic one-​sidedness from the very start. The final right to conclusively describe what that phenomenon truly consists in will go to the winning theory of reference. That is, the winning theory of reference will have earned the right to describe the phenomenon of epistemic one-​sidedness in its privileged theoretical vocabulary. Of course, I will be arguing that two-​factor referentialism is the winning theory and that it thus wins the right to describe epistemic one-​sidedness in its own terms. That description will depart in significant ways from entrenched Fregean orthodoxy. And to the extent that Fregean orthodoxy has a hold on philosophical common sense, that may prove at least initially jarring. But sometimes philosophy must disturb rather than accommodate philosophical common sense. And that is what I shall do in this chapter. In particular, I will argue that Fregean orthodoxy misunderstands both the nature and source of epistemic one-​sidedness. And out of this misunderstanding, a long, varied, and estimable tradition of taking the coreference puzzles to be of signal importance primarily for understanding the semantics of singular reference was launched. I break from that tradition and reject semantic approaches to the coreference puzzles. The true and ultimate source of the coreference puzzles and with them the epistemic one-​sidedness of reference lies, I shall argue, not in semantics but the logical syntax of our thought and talk.4 Let us turn now to a brief tour of the many coreference puzzles. While I do not propose to solve each and every one of them in this book, since that would take us down too many rabbit holes, I will address a fair number of them over the course of this and subsequent chapters. Despite the fact that 4 I do not claim to be the first or only philosopher of language to urge a reorientation of the philosophical study of the coreference puzzles away from semantics. By now, there is a long and distinguished countertradition of work arguing that the cognitive significance phenomena may have more to do with pragmatics than with semantics proper. See, for example, Soames (1985), Salmon (1983), Wettstein (1986), Taschek (1992), Perry (2001), and a plethora of others. While I think there is much insight in this countertradition, I think it too has paid insufficient attention to matters of syntax and form.

90  Referring to the World I do not aim for comprehensiveness, it will help to have a wide variety of coreference puzzles firmly in mind as we go about our theory construction. The granddaddy of all coreference puzzles is, perhaps, Frege’s puzzle about the possibility of informative identity statements. One may think of it as a theme on which there are many variations. Frege’s original coreference puzzle concerns the question of how possibly a statement of the form ⌈a = a⌉ could differ in “cognitive value,” as he called it, from a true statement of the form ⌈a = b⌉. Statements of the first sort are trivial, while statements of the second sort may, when true, convey new information. Yet, if a is identical with b, then a statement asserting the identity of a with b merely purports to assert the identity of an object with itself. We are now supposed to wonder whether that is not precisely what the trivial statement ⌈a = a⌉ also purports to assert. If it is, we are then supposed to wonder how the one statement can be trivial and the other informative, given that the two statements assert the very same thing about the very same object. Frege’s puzzle about informative identity statements is a first cousin of a broader class of puzzles—​what I will call “substitution puzzles.” Frege was also the first to take serious note of this broader class and to elevate their resolution into a sine qua non for the adequacy of a theory of reference. Substitution puzzles arise because of the apparent fact that substituting one name for another with the same reference in the context of, for example, a propositional attitude statement or an indirect discourse report may turn an apparent truth into an apparent falsehood. Even though Mark Twain is none other than Samuel Clemens, it seems possible for (1) below to be true while (2) below is not: (1)  Mike believes that Mark Twain was one of the founders of the Bohemian Club. (2) Mike believes that Samuel Clemens was one of the founders of the Bohemian Club.

Inspired by the spirit though not necessarily the letter of Frege, many philosophers have since concluded that contexts of this sort must be semantically special in some way or other. To some, such contexts have appeared to disrupt the “normal” semantic functioning of singular referring terms therein embedded. Frege himself originally argued that terms embedded in such contexts undergo a reference shift so that they denote therein what they customarily express when they occur outside of any such context. Quine

Puzzles of Coreference  91 went in a rather different and darker direction. He famously argued that expressions apparently embedded in such contexts do not, in fact, function as referring expressions at all. Such contexts, he claimed, are “referentially opaque.” In dismissing such contexts as opaque and as “creatures of darkness,” he meant to cast doubt on the semantic and metaphysical coherence of the idiom of so-​called de re belief. Views that see a disruption in the normal semantic function of names within certain contexts have proven highly controversial, however. And many have sought to solve the substitution puzzles without abandoning what Donald Davidson has called our pre-​Fregean semantic innocence, according to which terms embedded in “opaque” contexts suffer no change whatsoever in their semantic functioning or character. We shall eventually come down on the side of semantic innocence ourselves, but in a way that rejects not just the letter, but even the spirit of Frege. Substitution puzzles were once thought to arise solely in the context of so-​ called opaque or oblique contexts. But it eventually became clear that something rather like substitution failures seem also to occur even in so-​called referentially transparent contexts. And this may be taken to suggest, even if it does not quite decisively show, that it is not actually the peculiar opacity of the supposedly opaque context that causes substitution to go awry even there. Consider the following variant on the theme first made famous by Jennifer Saul (Saul 1997): (3) Clark Kent entered the phone booth and Superman came out. (4) Clark Kent entered the phone booth and Clark Kent came out.

Whether (3) and (4) differ in truth-​value or in some other way is unclear. Intuitions are bound to differ and probably not in a directly diagnostic way. One thing for certain, though. There are clearly circumstances in which one would be prepared to assert (3) but not prepared to assert (4), despite the coreference” of “Clark Kent” and “Superman.” It is initially puzzling why this should be so, especially given that no sort of embedding is at issue here that might be thought to induce opacity and non-​innocence with it. We have seen so far that some coreference puzzles arise for terms that occur within certain apparently semantically special contexts, while other coreference puzzles seem to have little to do with the workings of such contexts. There is yet a third variation on the theme that appears to involve an intricate interplay between so-​called transparent and so-​called opaque contexts. Consider the following somewhat updated version of an intriguing variation

92  Referring to the World on our coreference theme first made famous by Mark Richard (1983). A man and a woman are swiping through profiles on the dating app, Tinder. After swiping right on each other’s profiles, they end up exchanging text messages on their cellphones in an attempt to arrange a hookup for the evening. They are in separate cafés. Each is seated by a window. They are in close enough proximity to see each other out of their respective windows. But they do not (yet) realize that they can see each other. As they are texting back and forth, the man notices, just out of the corner of his eye, a woman sitting near the window in a café across the street. He sees that she is in danger. Trying to catch her eye, he frantically waves to her. Out of the corner of her eye, the woman notices that a man in a café across the street is frantically waving at her. The woman to whom the man is frantically waving is one and the same person as the woman with whom the man is eagerly texting. It is crucial to our little puzzle that neither realizes this fact. The woman texts the following message to her Tinder match: (5) The man watching me believes that I am in danger.

He echoes her text back to her, but with one change. He exchanges her first person “I” for his second person “you” as in (6) The man watching you believes that you are in danger.

Both (5) and (6) would seem to be straightforwardly true. Though we do not have a puzzle quite yet, a slight substitution—​this time not of one embedded referring expression for another—​as in our Mark Twain/​Samuel Clemens case above—​but of one unembedded expression for another—​quickly lands us in pickle. Witness (7) below, as might be texted by our male protagonist: (7) I believe that you [addressing the woman he is texting] are in danger.

Again, judgments are likely to divide here over what exactly is going on—​over whether (7) is flat out false or just pragmatically infelicitous in some other way. And it seems doubtful that such judgments wear their true natures on their sleeves. One thing is certain, though. (7) is not squarely on all fours with (6), and whether the difference is a matter of truth/​falsity or assertability/​ non-​assertability, we will not attempt to say at the moment. Clearly, though, (6) is something our guy is in a position to knowingly utter, while (7) is not.

Puzzles of Coreference  93 However exactly we choose to characterize the difference, it is puzzling that there should be this difference between (6) and (7) at all. While many coreference puzzles involve two distinct referring expressions for the same object, Kripke (1979) has given a variation on the theme in which a puzzle arises even for distinct occurrences of what appear, at least to superficial appearances, to be the same name again. Suppose that Smith lives among people who are quite divided about the aesthetic merits of the city of London. Some of them—​call this group the “A community” —​think that London is a city of outstanding beauty, while others—​call this group the “B community” —​find London horrendously ugly. Smith herself has no prior acquaintance with London. Nor has she made any prior use of the name “London.” Suppose the name “London” is first introduced to her via interaction with members of the A community. The members of the A community appear knowledgeable to Smith. Perhaps naively but not irrationally, she is inclined to take apparently knowledgeable people at their word. She uses the name “London” with the intention of referring to whatever the members of the A community refer to when they use that name. And she inherits certain beliefs about London from that community. In particular, she comes to believe what is expressed by (8): (8) London is very beautiful.

Subsequently, Smith comes in contact with other apparently knowledgeable people—​the members of the B community. These apparently knowledgeable people believe that London is among the ugliest cities in the world. Now Smith believes—​mistakenly and perhaps somewhat naively, but certainly not irrationally—​that knowledgeable people are unlikely to hold such divergent opinions about one and the same city. She therefore confidently concludes that the apparently knowledgeable people encountered later, and the apparently knowledgeable people encountered earlier, are not talking about one and the same city. She mistakenly concludes that there are two different cities, each of which is called “London.” We might say that she uses “London” as if it were ambiguous, as if that sound-​shape pattern were associated with two referentially independent though homophonous names such that when fully disambiguated they refer to two distinct objects. Smith is, however, wrong about all this. In fact, she has encountered the same name twice. And let us take it that the members of the A community and the B community do not share her confusion; they are mutually aware that they are using the

94  Referring to the World same name for the one city, London, and are mutually aware that they disagree about the aesthetic qualities of that city. Indeed, unbeknownst to Smith, representatives of the two subcommunities get together from time to time to debate the aesthetic merits of their shared city over pints of stout. The puzzle arises as soon as we, and also Smith’s fellow non-​confused citizens, speak about Smith’s beliefs about the one and only London. On the one hand, it may seem correct to say that Smith believes that London is one of the most beautiful cities in the world. After all, in those moments when she defers to the usage and beliefs of the A-​community, she assents to what is expressed by (8) above. But we can with equal justice conclude that Smith believes that London is very ugly, can we not? After all, when she is deferring to the usage and beliefs of the B community, she readily assents to such sentences as: (9) London is very ugly.

We know that Smith herself would probably deny that she believes that (the single city) London is both ugly and not ugly. Whether she is deferring to the A-​community or the B-​community in her use of “London” she would deny (10): (10) London is both beautiful and ugly.

Yet deferring to first the A-​community and then the B-​community, she might well assert: (11) Though London is very beautiful, London is very ugly.

Again, it is puzzling why this should be so, and what we should conclude about her state of mind given that this is so. It is also puzzling what we should say about the semantics of “believes” in light of all these puzzling data.

2.  Combinatorial Syntax and Singular Reference We will revisit the coreference puzzles—​at least the Fregean theme if not the many variations—​toward the end of this chapter. But rest assured, we shall have much to say about at least some of the variations over the next several chapters.

Puzzles of Coreference  95 I have already advertised my preference for syntactic approaches—​though they will need to be mixed, it shall turn out, with a decent helping of pragmatics here and there—​over strictly semantic approaches to the many coreference puzzles. But in order to go beyond mere advertising, I need to say more about precisely what I  mean by syntax. Claims about the syntax of singular reference have mostly not taken center stage in either the philosophy of language or the philosophy of thought. But careful attention to syntax of singular reference yields, I shall argue, significant theoretical payoffs for both our understanding of public natural languages and our understanding of thought itself. Begin with the very idea of syntax. Some philosophers tend to use the word “syntax” in a much narrower way than I do throughout this book. Many philosophers tend to conceive of syntax in terms of shape-​like “intrinsic” properties of a symbol or representation. The most prominent exemplar of that tendency is Jerry Fodor, who says: Syntactic properties . . . [are] among the “local” properties of representations, which is to say that they are constituted entirely by what parts a representation has and how those parts are arranged. You don’t, as it were, have to look, “outside” a sentence to see what its syntactic structure is, any more that you have to look outside a word to see how it is spelled.5 (Fodor 2000, 20)

Syntactic properties thus understood would seem to be intrinsic or essential. As Fodor puts it, “the identity of a sentence never survives alteration of its syntax or its logical form.” (2000, 22) I will not take direct issue with Fodor’s usage. He is free to stipulate what he means by “syntax.” Moreover, there is a role to play for such shape-​like intrinsic properties, even in my own more inclusive account of syntax. But such properties constitute only a small part of syntax as I conceive it. For the most part, when I speak of the syntax of a class of symbols and representations in this book, I intend to include certain properties that accrue to representations at least partly in virtue of their (non-​semantic) relations one to another. The sort of properties I have in mind might also be called “structural” or “combinatorial” properties. Though I will typically construe the realm of the syntactic broadly enough to include both intrinsic shape-​like properties of the sort that are Fodor’s focus, and 5 In fairness to Fodor, though, he also says that syntax has an “external” face since “the syntax of a representation determines certain of its relations to other representations.” But I take it that syntax properly so-​called is constituted by the sort of representation-​representation relations toward which Fodor seems here to gesture.

96  Referring to the World structural or combinatorial extrinsic properties, there will be times when I need to distinguish between these two different sorts of properties. I mostly rely on context to signal my intentions, but when I need to be precise and explicit, I use phrases like “syntactic in structural or combinatorial sense” to signal that I am getting at something more than mere shape. Correlatively, I will use phrases like “syntax in the narrow Fodorian sense” when I mean to explicitly focus on shape-​like intrinsic properties. When philosophers of mind and language have thought it important to pay attention to matters of syntax, as they sometimes have, they have mostly tended to focus on syntax in the narrow Fodorian sense. That peculiar professional proclivity makes it unsurprising that philosophers as a class have tended to underestimate the importance of syntax for understanding the distinctive character and function of names and other singular referring expressions. It is only given a structural/​combinatorial understanding of syntax that the question of what, from a syntactic point of view, being a name consists in first becomes substantive and interesting. Given just a narrow Fodorian understanding of syntax, there is almost nothing substantive to say about the syntax of names as such. Names don't, for example, come with characteristic intrinsic shapes or even spellings that might serve to distinguish them from other sorts of expressions. Once we have adopted a narrow Fodorian understanding of syntax, it is all but inevitable that we will conclude that whatever is linguistically distinctive about names and other singular referring expressions will be explained by semantics rather than syntax. Certainly, names and other singular referring expressions such as demonstratives or indexicals do have distinctive semantic characters. And no doubt we can appeal to semantics to explain much about their distinctive roles in our thought and talk. Still, it is, I claim, a mistake to hurry past syntax. Such haste has, on my view, caused both referentialists and presentationalists to miss much that is crucial. Or so I shall argue. I start by attempting to motivate my focus on combinatorial syntax, as opposed to narrow Fodorian syntax, a bit more fully and systematically. Think, first, about the contribution of knowledge of syntax to a speaker’s overall linguistic knowledge. Suppose we ask what it is that a speaker knows about an expression e when she knows that e is a name but doesn’t know the meaning or reference of e. We may allow that she knows certain minimal facts about the semantic character of names considered as a linguistic type. She may know, for example, that names are for referring with rather than for predicating with. That won't affect the argument. What we need to suppose that she

Puzzles of Coreference  97 does not yet know is any substantive (metasemantic) truths about just how names do their referring, or precisely what kind of contribution they make to the semantic contents of the sentences in which they occur. What is there left for her to know about names in virtue of which they count as names? Armed only with a narrow Fodorian conception of syntax, there wouldn’t be much else to say. Perhaps the most that we could say is that the speaker knows that e is a member of an open-​ended list of expressions and that any expression on that list counts as a name in L. But then we are left with the question what if anything, from a syntactic point of view, determines which expressions belong on that list. If we restrict ourselves to intrinsic shape-​like properties, all that names would appear to have in common that makes them one and all count as names, is the property of being on that very list. This constricted understanding of syntax may seem unproblematic on its face. Again I’m not trying to show that the narrow Fodorian conception is incoherent or that it has no systematic theoretical use. My point is just that if we restrict ourselves to that understanding of syntax, there is the potential for serious downstream theoretical costs. Such a restriction would blind us to certain essential things about the nature and role of singular representations in both language and thought, things that semantics too is powerless to explain. And this peculiar form of blindness would undermine our ability to solve the problem of objective representational content. Fortunately, we need not opt for such blindness. We are free to acknowledge that a speaker who knows only that e is a name knows much more about e than that it belongs on some open-​ended and arbitrary list. She also knows that e is the sort of expression that can occupy the argument places of verbs, that can well-​ formedly flank the identity sign, and that can anchor anaphoric chains both within and across sentence boundaries. The crucial point is that such properties are not in any obvious sense semantic properties of e. They have nothing to do with how the reference of e is determined. They say nothing about the exact nature of e’s contribution to the propositional content of any sentence in which it occurs. These are facts about how e interacts with other linguistic expressions both within and across sentence boundaries. As such, they have to do not with representation-​world relations, which is the domain of semantics, but with representation-​representation relations, which is the domain of combinatorial syntax.6 6 There are, of course, semantic relations among representations—​relations of synonymy, for example. But such relations presuppose that expressions already stand in various representation-​world relations.

98  Referring to the World Once you are willing to characterize a representation as a name by reference to its combinatorial properties and structural relations to other representations, the property of being a name is no longer well construed as merely a matter of intrinsic shape-​like properties of the sort that an expression may enjoy all on its own, independently of its relation to other expressions.7 An expression will count as a name only to the extent that it plays the right kind of combinatorial role in an entire system of interlocking expressions. To be sure, if a name is not identifiable as a name by something like its narrow shape-​like properties, then it will be impossible for a system to treat it as a name. And that is why we cannot simply dispense with narrow Fodorian syntax in favor of combinatorial syntax. Combinatorial syntax gets to ride on the back of narrow Fodorian syntax, at least in certain kinds of systems. But it simply does not follow from the fact that names must somehow be marked as names if they are to figure in systems that operate on combinatorial syntax that the property of being a name is simply a matter of shape. To make that mistake would be to mistake an abstract role for the particular concrete occupant of that role. In the course of discussing the syntactic correlativity of various syntactic categories in the previous chapter, we already noted that what is true of names and other singular referring expressions is true for other classes of expressions as well. But it is worth repeating the point for emphasis. From a combinatorial point of view, verbs, quantifiers, names, determiners are one and all correlative notions, inter-​definable only as parts of an entire system. We no more have independent hold on the syntactic category NAME than we do on any other syntactic categories such as SENTENCE or VERB PHRASE or NOUN PHRASE. From our combinatorial perspective, names are nothing but recurrent sub-​sentential expressions of a distinguished sort that function as minimal noun phrases and thereby combine with verb phrases to yield sentences. Verb phrases are expressions that combine with noun phrases to yield sentences. Sentences, in turn, are nothing but linguistic complexes built, in accordance with the rules of a grammar, out of various sub-​sentential constituents. Only when we are armed with this broader, more holistic, and

7 This is not to say that narrowly Fodorian syntax is entirely irrelevant to syntax in the broader sense. If a competent speaker is to be able to recognize and competently deploy an expression in syntactically appropriate ways, then there must be something about the expression that enables such recognition and deployment. And intrinsic shape-​like properties may serve as good cognitive handles on syntax in the broader sense.

Puzzles of Coreference  99 combinatorial notion of syntax, can we thereby fully appreciate that the distinctive syntactic character of names cannot be captured by appeal merely to properties of local shape or intrinsic form.8 Certain further conclusions more or less directly follow. I will soon argue, for example, that if we are to distinguish one name from another in a way that allows us to do significant theoretical/​explanatory work, we must also distinguish names from their spellings. To say that a name must be distinguished from its spelling is not to downplay the importance of spelling. But I will argue that for a wide variety of theoretical and explanatory purposes, names are best individuated one from another not by their spellings—​which is just another shape-​like property of the sort that is central to the narrow Fodorian conception of syntax—​but by their roles in interlocking systems of expressions. Again, to say this is not to deny that important features of various expression types can indeed be marked by differences in spelling. With verbs, for example, we mark differences of tense, number, or aspect of the very same word by differences in spelling. But it would clearly be a mistake to conclude that a mere spelling pattern is what constitutes a verb as a verb. Just so, the property of being a name is no more a matter of mere spelling than the property of being a verb is. To be a verb is to be an expression that plays a distinctive role in an overall system of expressions. Just so for names.

3.  Names as Devices of Explicit Coreference Much of the remainder of this chapter lays the groundwork for addressing coreference puzzles from the perspective of two-​ factor referentialism. Recall that according to two-​factor referentialism, reference depends on two interlocking factors or sets of factors. One is internal and syntactic; the other is external, causal, and informational. Objective representational content arises only from the interaction of the internal and the external. Sticking, though, with the internal for the nonce, a central claim of two-​ factor referentialism is that in the internal combinatorial syntax of both thought and language, names function as devices of explicit coreference. To say that names function as devices of explicit coreference is to make a claim 8 Brandom (1994) comes closest to realizing this fact, but he does not separate semantics from syntax in the same way that I propose to do in this book.

100  Referring to the World about their characters as representational vehicles and not yet a claim about their representational content—​neither about what those contents are nor about how those contents are determined. The claim is just that numerically distinct tokens of the same name type enjoy, in virtue of their roles in the combinatorial syntax of both language and thought, an intrinsic purport of coreference with one another. According to this view, names are, in effect, a distinguished sort of anaphoric or “reference transmitting and preserving” device within both talk and thought. Though I take garden variety anaphoric expressions and names to be diverse species of a common genus, there is an important difference between reference transmission and preservation within garden variety anaphoric chains and reference transmission via names. In garden variety anaphoric chains, there are typically multiple distinct expressions standing in some relation of anaphoric dependence such that interpretation of the anaphorically dependent expression is “controlled” by the interpretation of an antecedent expression. We might also say that the anaphorically dependent expression “inherits” its reference from an antecedent that somehow has this reference independently and antecedently—​though here we must set to one side donkey anaphora. But is not quite correct to say that the reference of one name token is anaphorically inherited, in this same way, from another name token on which it is somehow anaphorically dependent. I do not mean to suggest that reference is passed from name to name in the way that a dollar bill might be passed from person to person. The more precise claim is rather that name types are such that the tokens of a given name enjoy an intrinsic purport of coreference one with another, whenever and wherever those tokens occur. This means that by their very (linguistic) nature the multiple actual and possible tokens of a given name type stand or fall together with respect to referential purport, success, and failure.9 And I shall argue that the fact that the tokens of name necessarily stand or fall together with respect to referential purport, referential success, and referential failure provides an important clue to the source and nature of referential 9 Some philosophers have been tempted to assimilate names to variables under assignment. But I do not think that this is quite right. The reference of every token of a given name is determined in one fell swoop by facts about the name type of which the token is a token. What we might call the type-​reflexivity of names is also reflected in what I take to be the fundamental semantic rule governing all names—​if e ∈ NAME, and t is a token of e, then t rigidly designates the object o such that o BEARS e, where e is an expression type, rigid designation is a semantic/​modal relation, and BEARS is in this worldly causal relation distinct from the semantic relation of rigid designation. It is the business of the metaphysics of reference to say what the name-​bearer relation consists in. It is the business of semantics to characterize the nature of designation.

Puzzles of Coreference  101 fitness or objectuality, as well as the source and nature of the epistemic one-​ sidedness of all reference. Indeed, this fact also turns out to be central to our explanation of the variety of things we do with empty names. But more on this point in due course. It is also important to keep in mind that intrinsic purport of coreference must be sharply distinguished from coincidental coreference. Two token names that are not co-​typical may refer to the same object, and thus be coreferential, without being explicitly or intrinsically coreferential. Tokens of “Hesperus” and tokens of “Phosphorus” coincidentally corefer, for example, but they are not explicitly coreferential. The fact that “Hesperus” refers to Venus is both linguistically and metaphysically independent of the fact that “Phosphorus” refers to Venus. And here is an important truth, one that deserves to be announced with bells and whistles, or at least italics. Whenever m and n are distinct names, they are referentially independent. When m and n are referentially independent, nothing in their intrinsic linguistic characters guarantees that if m refers to o then n refers to o as well. To say that distinct names are, as a matter of their linguistic character, referentially independent, is not to say that they must ipso facto fail to corefer. It is just that when two referentially independent names do corefer, their coreference will be a coincidence of usage. Taken together, the correlative relations of referential independence and explicit coreference partially characterize the broadly syntactic or structural roles of members of the linguistic category NAME. Part of what it is to be a name is to be an expression type such that tokens of that type are explicitly coreferential with one another and referentially independent of the tokens of any distinct name (type). Returning to our earlier query, we can now see that part of what one knows, when one knows of e just that it belongs to the category NAME, is that whatever e refers to, if it refers to anything at all, tokens of e are guaranteed to corefer one with another and to be referentially independent of any distinct name e', whatever e' refers to. A name (type) is, in effect, a collection of actual and possible tokens such that all tokens in that collection are “guaranteed” to corefer. We may call collections of this sort chains of explicit coreference. I take it to be a linguistically universal fact about the linguistic category NAME that numerically distinct tokens of the same name type share membership in a chain of explicit coreference and numerically distinct tokens of two distinct name types will be members of disjoint chains of explicit coreference—​even if the two tokens are coincidentally coreferential.

102  Referring to the World We shall soon see that the distinction between the intrinsic purport of coreference and merely coincidental coreference between referentially independent designators is crucial to understanding the coreference puzzles and to our ultimate explanation of the epistemic one-​sidedness of reference. But first we need a bit more stage-​setting. We need to say more about what names are and are not. We begin by distinguishing names from their spellings.

4.  Names, Their Spellings, and the Drainage Thesis On ordinary ways of thinking, which are richly suffused with common sense and untainted by the concerns of deep theory construction, names are more or less wholly individuated by some combination of spelling and pronunciation alone. Ordinary intuition would seem to count “Claire” and “Clare” as variant spellings of one and the same name—​presumably because they are pronounced the same—​rather than as two wholly distinct names, while counting “Claire” and “Clara” as distinct names, despite the fact that they overlap in spelling as much as the former do and are pronounced nearly, but not quite the same. I won’t bother much with determining whether it is spelling or pronunciation that plays the major part in distinguishing names for one another. For our purposes, talk of “spelling” is just a shorthand for whatever combination of spelling and pronunciation does the distinguishing. Now in most contexts in which the question whether x and y are the same or different names is at issue, this commonsensical understanding is mostly unproblematic. But when we turn to the task of theory construction, things are different. Though much philosophical orthodoxy about how to individuate names hews rather closely to common sense, there are important gains to be gotten by departing from common sense and distinguishing names from their spellings. Even so, there will still be an important place for spellings. On our final theory, every name will still have a spelling—​possibly, even a range of alternative spellings. It is just that no name will be construed as strictly identical with that which is individuated by a spelling—​or even a range of spellings—​ alone. Spellings should rather be thought of as being associated with names in a one-​many fashion. That is, we allow for the possibility that what function as numerically distinct names within language and/​or thought may nonetheless be associated with the very same spelling (at least in unreformed ordinary language). This approach makes names out to be more “abstract”

Puzzles of Coreference  103 or “logical” entities than mere spellings. Again, treating names in this way may do mild violence to common sense, but it enables us to achieve a much deeper understanding of the many coreference puzzles while pointing the way to deeply satisfying solutions to them.10 To begin to appreciate the theoretical costs and benefits of different ways of individuating names, suppose that we try our best to hew to both common sense and philosophical orthodoxy. We will tend to see one name where I am committed to saying there are many distinct names that happen to share a spelling. One is immediately confronted with another question. What role do names so understood actually play in a language? And this, I suggest, is where the trouble begins. One standard answer to this question is captured by the following remarks by Perry and Korta (2011): Names are nambiguous. . . . [I]‌f you name your new child “Larraitz,” you seek to establish a practice that makes it possible to refer to the child with that name; you don’t make it impossible to use the name to refer to other people named “Larraitz.”

If names are indeed individuated by spelling alone, it is perhaps natural to conclude that names are essentially “nambiguous” to use Perry and Korta’s colorful term. And once you go down this path, it’s hard to see in what sense names as such can be said to play any role whatsoever in the syntax and semantics of a language. It is especially hard to see why we should think of names as such as referring expressions at all. Consider the string “Claire” on the conception of names according to which names are individuated by spelling alone. To whom does the name so individuated refer? There would seem to be no determinate answer to that question. Suppose we ask, for 10 A distinction due to Fiengo and May (2006) may be of some initial help in motivating my approach. They helpfully distinguish between names and expressions. On their view, names are indeed individuated by spelling and pronunciation. Expressions are not. On their view, expressions may contain names, but expressions can’t be identified with names. Names are what they call “lexical items” rather than what they call “syntactic expressions.” On their view, it is expressions rather than names that play various structural roles in sentences and discourses. Similarly, on their view, expressions, and not names, bear reference. Though there are important disagreements between my view and theirs, on this score, at least, I take my view and their view to be more or less notational variants. What they call a name, I call a spelling of a name. And what they call an expression, I call a name. On this way of thinking about a name, although John Etchemendy and John Perry have first names that are spelled the same, they don’t have the same first name at all. Why endorse such a view, especially when I have admitted that it does violence to both common sense and received philosophical wisdom? The answer is that the Fiengo-​May-​Taylor view does less total violence than the alternatives.

104  Referring to the World example, whether Claire is the same person as Claire. Again, if names are individuated by spelling alone it would seem that we have not yet asked a question that could have a determinate answer. The thing wholly individuated by mere spelling does not carry with it either intrinsic purport to refer to one definite object or even an intrinsic purport of coreference upon repetition. It is not until all “nambiguity” has been bleached out that such questions come to have a determinate form and content. But here’s the thing. While commonsense does count names that are spelled the same as instances of the same name, it is also part of common sense that the function of names within the language is to refer (and also to corefer). These two elements of common sense are in tension with each other. If we insist on the innocent-​seeming claim that names are wholly individuated by spelling alone, we will end up, in one way another, draining names of all intrinsic singular referential purport. With that thing, whatever it is, that is wholly identified by the spelling “J’^’o’^’h’^’n,” there are associated many distinct individuals. Taken on its own, independent of any particular use or context, the thing wholly identified by the spelling “J’^’o’^’h’^’n” would seem to apply either to all of those individuals or to none of them at all. None of the individuals associated with the thing wholly identified by the spelling “J’^’o’^’h’^’n” stands in a special or privileged relation to it that would suffice to distinguish it as the unique referent of the thing wholly identified with the spelling “J’^’o’^’h’^’n.” If “J’^’o’^’h’^’n” refers to one of those individuals, it must equally refer to each of them simultaneously. And if it doesn’t refer to each of them, it must refer to none of them. Once we individuate names by spelling alone, we must deny that names as such even purport to either refer or corefer. Something clearly must go. There are ways to solve this apparent dilemma, even while sticking to the problematic criterion of individuation. But on any way of doing so, it will follow that taken independently of any particular use or context, the thing wholly constituted by the spelling “J’^’o’^’h’^’n” does not function as a singular referring expression at all. In other words, to rigidly identify names with mere spellings is to drain names of their singular referential purport. There are many versions of what I will call the drainage thesis. I have already mentioned the “nambiguity” thesis of Perry (2011). Names have also been said to be (or to be usable as) peculiar sorts of demonstratives (Burge 1973) or indexicals (Recanati 1993), and to be (or to be usable as) predicates (Burge, 1973, Fara, 2015, and Matushansky, 2009). On none of these views does singular referential purport attach to what each identifies as a name directly

Puzzles of Coreference  105 and as such. Predicates “apply” to many objects simultaneously without referring to any of them. Indexicals and demonstratives do refer, though only in particular contexts or on particular occasions of use and not in a context-​ independent way. On the other hand, if names as such are intrinsically ambiguous (or “nambiguous”) then names must either be said to refer to many objects simultaneously or to refer to no particular object at all unless and until they are fully disambiguated. All versions of the drainage thesis go wrong from the start, precisely because they accept the assumption that names are wholly individuated by spellings alone. Once you have accepted that assumption, the game is essentially over. You will fail to see that names are dedicated devices for referring and coreferring. You will deny that names enjoy an intrinsic purport to refer and to corefer. You will fail to appreciate that names have a distinctive coreference profile that distinguishes them from other sorts of singular referring expressions. And you will misdiagnose the many coreference puzzles. You will hurry pass syntax and turn prematurely to semantics where syntax alone would have told you all you really needed to know. Your premature turn to semantics will blind you to the fact that names and other singular referring devices can be objectual and thus antecedently poised or ready to refer without being fully referentially successful or objective. And because you have failed to fully appreciate the distinction between objectuality and full-​blown objectivity—​that is, between referential fitness and referential success—​you will not be fully alive to the crucial question of how and why space is opened for extra-​representational factors to step in and bridge the gap between mere fitness and actual success. Such downside theoretical costs are far too much to bear to make fealty to what is after all only one half of common sense worth the bother. So over the next several sections, I take up the various versions of the drainage thesis and show in greater detail how each goes wrong.

5.  Drainage as Ambiguity The most initially plausible form of the drainage thesis is the claim that names are ambiguous. This approach may appear to let us have our cake and eat it too. While it has to say that names as such may be devoid of intrinsic singular purport, it can allow that once they are fully “disambiguated” they do enjoy singular referential purport. It is not clear that it follows directly from the

106  Referring to the World ambiguity thesis that fully disambiguated names are devices of explicit coreference that are referentially independent of every distinct but also fully disambiguated name. So, there may still be deep truths about names that it fails to capture. But there is no principled reason that the ambiguity theorist cannot accept these claims as additional but independent claims. Perhaps the ambiguity thesis is best considered a mere notational variant of my own approach. What it calls a “name,” I call a mere “spelling of a name.” What I take to be a name properly so-​called, it acknowledges as a disambiguation of what it understands to be a name properly so-​called. Both approaches would also seem to agree that no matter what it is called, that entity that is individuated by spelling alone does not enjoy intrinsic singular referential purport and does not yet function as a device of explicit coreference, while a related, but more “abstract” entity—​the name fully disambiguated—​that is not individuated by spelling alone, may well do so. On either of these ways of individuating names, it is a short step to the conclusion that what mere spelling picks out is not yet a linguistically significant unit, with a distinctive linguistic function. To see this, imagine pressing the ambiguity theorist on the question whether it is the disambiguated name or the non-​disambiguated name that occupies the argument place of a verb or that may well-​formedly flank the identity sign or that may serve to anchor anaphoric chains. If the ambiguity theorist answers that it is only the fully disambiguated name that plays any significant linguistic role, then her view collapses into ours. Perhaps a bewitchment with mere spelling might lead her to insist that it is the non-​disambiguated name that plays significant linguistic roles. But such a view would be implausible in the extreme. If she admits that surface spelling is not where the real linguistic action is, while allowing that at the level of true logical-​syntactic form, names are tractable linguistic units with distinctive linguistic functions, the all but inevitable conclusion is that we ought really to reserve the title “name” not for that which is individuated by spelling alone but for the more abstract occupant of a certain distinctive linguistic role. Only an excessive deference to the divided “wisdom” of common sense could keep us from taking the theoretical plunge. One might still worry that our dispute with the ambiguity theorist is a merely verbal one over who gets the keep the word “name.” I don't deny that this worry enjoys a certain rough justice, but there may still be a substantive issue at stake between the ambiguity theorist and myself. I appeal to a broadly conceived notion of syntax to explain certain aspects of the distinctive character of what the ambiguity theorist dubs a fully disambiguated

Puzzles of Coreference  107 name and what I dub a name simpliciter. That names have intrinsic singular referential purport, that they function as devices of explicit coreference, that they may anchor anaphoric chains but never be anaphorically dependent are consequences, on my view, of the broadly structural-​syntactic role of names. It is an open question what, if anything, plays the sort of explanatory role that I assign to syntax broadly construed for the ambiguity theorist. Perry, for example, appeals to so-​called permissive conventions to explain how a not yet disambiguated name ever manages to refer to one particular object.11 Perhaps Perry would insist that various other aspects of not yet disambiguated names are to be explained by appeal to such things as “conventions” governing their use. Now I concede that there is a derivative and secondary sense in which names are governed by various conventions of use. It is a wholly conventional matter that I am called by the name “Kenneth A. Taylor,” for example. But the features and properties of names on which I focus, and which I take to be definitive of names as linguistic types, are not merely conventional in any useful sense. On my view, they are more or less directly specified by the Universal Grammar (UG) that plausibly determines the space of all humanly possible languages. In slightly more detail, my conjecture is that UG makes the category NAME available to humanly possible languages for exploitation. The claim is not that necessarily every human language contains a set of names. For all I conjecture, NAME may be an optional rather than a mandatory linguistic category. That is an empirical issue on which I do not presume to pronounce here—​though names being such an awesome and powerful linguistic device, it seems highly likely and non-​accidental that (many, most, or all?) languages will contain a stock of them. More important, I suspect that it is non-​accidental, and not a matter of conventional at all, that names behave in certain ways and not others within any language that contains some. I suspect, for example, that the fact that the category NAME is partially defined by the twin properties of explicit coreferentiality of co-​typical name tokens and referential independence of type-​distinct name tokens is due to constraints rooted in UG rather than in mere linguistic convention. If so, then the question of whether my (potential) dispute with the ambiguity theorist is substantive or merely verbal will depend on what stance the ambiguity theorist takes on such issues as these. I will hazard no guess here as to where the ambiguity theorist might ultimately stand.

11 See also Recanati (1993).

108  Referring to the World There is a further reason for thinking that the behavior of names is not a matter of mere convention. The role of names in shared public languages is more or less a direct reflection of the role of name-​like representations in the de facto private language of thought. (And that may be just why UG bothers to specify their natures.) Just as names function in shared public languages as devices of explicit coreference, so our ability to deploy name-​like expressions in thought grounds our capacity for thinking with intrinsic purport as of the same again. I can think of my son Kiyoshi today and think of him again tomorrow with a kind of inner assurance that I purport to think of the same person twice. I do so merely by deploying an inner name-​like representation *Kiyoshi* across distinct thought episodes. If one deployed no such devices in thought, it would always be an open question whether, in purporting to think now of a particular o and now of a particular o', one has thought of two distinct objects or has thought of the same object twice. I do not deny that it may sometimes be an open question for a cognizer whether two of her thought episodes share a (putative) subject matter. But it is surely not always so. And we need to ask ourselves by what means such inner assurance that we are thinking again of the same subject matter is achieved. The answer, I conjecture, lies in the very nature of the inner vehicles though the deployment of which we think with inner purport of sameness. I conjecture, that is, there exists a distinguished class of representations that function in thought as devices of same purport. For such devices, to think with or via them again is ipso facto to purport to think of the same thing again. My claim is that name-​like representations are precisely such devices. It may be asked why should either public languages or the de facto private language of thought contain such representational devices? 12 I answer that the capacity to deploy such devices is central to the objectuality and referential fitness of both our talk and our thought. For example, if there were no such devices, no two thoughts or utterances would ever intrinsically purport to be about the same object again. But if no two (token) representations ever intrinsically purported to refer to the same again, then for any new thought or talk episode, even when the thinker was, in fact, thinking or speaking of the same object again, it would always be inwardly as if she was thinking or speaking of an object possibly never previously encountered. But this would mean that the individual cognizer would have at best a fleeting cognitive 12 See Brandom (1994), who asks a similar question, viz., what are singular terms and why are there any, but from an inferentialist perspective that presupposes what I call the priority of the sentence.

Puzzles of Coreference  109 hold on the object as such. She could not remember today what she believed yesterday.13 She could not anticipate in thought future encounters with a currently perceptually salient object, as least not as encounters with that very object again. A mind in which no two thoughts ever intrinsically purported to be of the same thing again would lack the capacity for thought as of objects at all. Thoughts as of objects are thoughts as of enduring particulars that may be encountered and cognized again from different perspectives while being the same again and while being at least on occasion cognized as the same again. And this is just one thing that would be lost if our words and thoughts never purported to be thought and/​or talked as of the same again.14 On my view, names in shared public language are externalizations of name-​like representations in the language of thought. Inner name-​like expressions function as inner devices of referential coordination among the thoughts within an individual mind. Once externalized by means of a shared natural language, they come to subserve referential coordination among discursive communities of thinkers in their thought and talk with one another. They enable linguistic agents to make the preservation of subject matter manifest in their talk exchanges with another. Indeed, I suspect that names are well-​nigh indispensable for these roles. And if this is so, that is just one reason to believe that the fundamental role and character of names in both public natural languages and the de facto private language of thought is in no sense merely a matter of convention. A mind that lacked the capacity to deploy name-​like representations in thought would hardly be recognizable as a human mind. A language that failed to contain names might not be recognizable as human language either. 13 For a suggestive and helpful discussion of mental anaphora and its role in identity thinking and content-​preservation, see Lawlor (2002). 14 As suggested in Chapter 2, Kant was perhaps the first to see clearly that an inner capacity for thinking with same-​purport was essential to enable our thought to achieve what I have elsewhere called full objectivity (as opposed to mere objectuality). Consider, for example, the following passage from the A-​Deduction Without consciousness that that which we think is the very same as what we thought a moment before, all reproduction in the series of representations would be in vain. For it would be a new representation in our current state, which would not belong at all to the act through which it had been gradually generated, and its manifold would never constitute a whole, since it would lack the unity that only consciousness can obtain for it. If, in counting, I forget that the units that now hover before my senses were successively added to each other by me, then I would not cognize the generation of the multitude through this successive addition of one to the other, and consequently, I would not cognize the number (1781, 103). Kant, of course, takes the points made above about how numbers are cognized to be perfectly general and to apply to the cognition of any object as such.

110  Referring to the World Convention does play a secondary role in populating public languages with names. Though our minds likely come antecedently equipped to produce and entertain name-​like representation in thought, in order to populate a shared natural language with a determinate stock of names and to bring it about that those names stand for particular objects, a community has to do something. This is where convention comes it. Though UG and logical syntax of thought reserve a role for names in language and in thought respectively, neither suffices on its own to guarantee that a shared language will contain names at all, let alone that it will have names for particular objects. What UG specifies with respect to a natural language is what you’ve gotten when you have succeeded in introducing names into a language. It specifies the sort of syntactic and semantic roles the introduced expression will play. It tells you, for example, that names will function syntactically as devices of explicit co-​reference and that they will function semantically as devices of direct reference. Moreover, though a language that contained no devices of explicit co-​ reference would be pragmatically impoverished in ways that I outline briefly below, there is no a priori philosophical reason to rule out the very possibility of such pragmatically impoverished languages. And this is precisely where convention must step in and fill the breach. Return briefly to the distinction between names and their spellings. I have claimed that unless we make such a distinction, we will be at a loss to explain various linguistic facts and regularities. And in fact, a closer look at certain well-​known grammatical principles governing reference and coreference directly bears that prediction out. Consider the following: (12) Johni kicked Johnj.

On the default reading of (12), the two occurrences of “John” must be read as at most coincidentally coreferential. In fact, Principle C of the binding theory, along with parallel principles from other grammatical frameworks, predicts that with the two occurrences of “John” “co-​indexed,” as linguists like to say, (12) is syntactically ill-​formed and therefore, presumably, not directly interpretable at all. (Chomsky 1981, 1995). One might initially be tempted to read Principle C as somehow prohibiting recurrences of the same name within a single argument structure. That would explain why (12) is syntactically ill-​formed. But it is important to see that as more or less standardly stated, Principle C isn’t quite correct. The standard interpretation of Principle C conflates what linguists sometimes call “disjoint” reference with

Puzzles of Coreference  111 what is, in fact, merely a requirement of referential independence. Though related, these two notions are not equivalent. Wherever there is disjoint reference, there is also referential independence. But referentially independent expressions need not be referentially disjoint in the intended sense of that phrase. Indeed, there are contexts in which an utterance of (12) could convey—​though admittedly in a sort of accidental way—​a proposition about John’s relation to himself. Suppose, for example, the speaker has encountered John twice, on two different occasions, but wrongly supposes that the first John and the second John are two distinct people. A requirement of referential independence doesn’t rule out this sort of coincidental coreference, but a requirement of disjoint reference would. On the correct reading of Principle C, it entails not disjoint reference, but referential independence. To be sure, it turns out that whenever we have two referentially independent designators, there is a strong, but overridable default reading on which the two designators fail to corefer. That is a communicatively significant fact that shall play an important role in our explanations of the many coreference puzzles. But to say that there is default reading on which referentially independent designators are read as referentially disjoint is one thing. To say that they are necessarily referentially disjoint is a different claim entirely. Notice that when we have what are obviously two distinct names, as in (13) below, the parallel construction is perfectly permissible—​even in contexts in which it is manifest that “Clark Kent” and “Superman” coincidentally corefer: (13) Superman kicked Clark Kent.15

Why should there be this difference? The answer has in part to do with the fact that the two occurrences of “John” in (12) cannot permissibly be construed as explicitly coreferential, but must be construed as referentially independent, despite the fact that they share a spelling. The correct generalization 15 Distinct names are ipso facto referentially independent. Moreover, unless it is made explicit either by context or by the addition of an identity statement explicitly linking the two, the use of referentially independent names generates a defeasible imputation of distinctness. Because of this fact, substitution of names that are merely coincidentally, as opposed to explicitly coreferential, is not, in general, guaranteed to preserve what I have elsewhere called dialectical significance. It is, unfortunately, quite easy to confuse failure to preserve dialectical significance with failure to preserve truth-​value. A clear failure to appreciate the significance of the difference between failure to preserve truth-​value and failure to preserve dialectical significance is exemplified in Saul (1997). Though Saul was one of the earliest philosophers to take explicit notice of the failure of co-​referring names to be freely substitutable even in simple sentences containing no embedding constructions, a lack of a full understanding of the distinctive coreference profiles of names led her to draw a number of erroneous conclusions. See also Saul (2007).

112  Referring to the World is that whenever two names can permissibly occur in a construction like “x kicked y,” the names must be construed as referentially independent rather than explicitly coreferential. But any given name is explicitly coreferential with itself and referentially independent of every other name. Hence to interpret the two occurrences of “John” in (12) as referentially independent just is, from a purely grammatical point of view, to count the two occurrences as occurrences of two distinct and therefore referentially independent names which happen to be spelled the same, rather than as the occurrence of the same grammatical name twice. The fact that sentences like (12) strongly default to a reading on which two occurrences of names with the same superficial spelling are, nonetheless, linguistically marked as referentially independent and therefore grammatically distinct names is, I think, due to certain deeper grammatical constraints on the way a name may claim simultaneous occupancy of the multiple argument places of a single verb. The strongly preferred way of saying that John is simultaneously the agent and patient of a single kicking is to deploy the reflexive pronoun as in (14): (14) Johni kicked himselfi.

Indeed, though a non-​reflexive pronoun can often be explicitly coreferential with an antecedent name, explicit coreference is not possible here. If we substitute such a pronoun for “himself ” in (14) we get: (15) Johni kicked himj.

As with (11), on the default reading of (15) “John” and “him” are referentially independent.16 Again, though it is possible for “John” and “him” as they occur in (15) to corefer, they can do so only if the coreference is coincidental rather than explicit. These data strongly suggest that, to a first approximation, a single name can simultaneously “control” multiple argument places of a single argument structure only through anaphoric dominance of a reflexive pronoun. We may think of the reflexive pronoun as the name’s referential doppelganger. Somewhat surprisingly, perhaps, a name is forbidden from serving as its 16 Indeed, Principle B of the binding theory predicts that (11) is syntactically ill-​formed when “John” and “him” are co-​indexed and thus explicitly coreferential. (Chomsky 1981, 1995).

Puzzles of Coreference  113 own referential doppelganger. That is, a name cannot be anaphorically dominated even by itself. And that is why when we have what appears due to mere spelling to be the same name again, deeper grammatical constraints suggest that we should regard that appearance as a mere appearance.

6.  Drainage as Indexicality Turn next to a version of the drainage thesis that assimilates names to demonstratives and/​or indexicals. Although it too denies that names as such have singular referential purport, at least when they are taken independently of context, it does have the advantage of treating names as such as significant syntactic and semantic units. Yet this approach runs roughshod over another deep and crucial fact about names—​the fact that the linguistic category NAME and the linguistic category DEICTIC are a minimal pair within the class of singular terms. These categories have complementary coreference profiles. It is a defining fact about the category DEICTIC—​which includes both explicit indexicals and true demonstratives—​that tokens of the same deictic type are referentially independent. When tokens of the same deictic type do corefer, the coreference will always be a coincidence of usage. Because token deictic expressions of the same type are referentially independent of each other, it is always “as if ” an episode of deictic reference involves reference de novo to the relevant object—​at least relative to any numerically distinct deictic. Only expression types that carry no intrinsic singular referential purport of their own, at least not taken as types, could function as dedicated devices of de novo reference. In de novo acts of referring, referential purport is determined on a token-​by-​token basis and is not inherited from any structurally “controlling” antecedent. In contrast to names, each token deictic is referentially independent of co-​typical tokens and stands on its own with respect to referential success and failure. And this is so even when the same device of de novo reference is used again to refer again to the same object. Suppose, for example, that I point to Troy and say, “That man is a fine young philosopher” and then point to Troy again and say, “That man is a fine young poet.” Though I have pointed and referred to the same man twice and have done so by using tokens of the same linguistic device again, my two acts are two interpretationally independent acts of reference. It is crucial that the claim here is not that coreferring and co-​typical deictic tokens can never be or be interpretable as coreferential. There are utterances

114  Referring to the World for which it seems all but mandatory that co-​typical deictic tokens be interpreted as coreferential. Consider the following: (16) Ted saw that man and Bill saw that man too. (17) Ted saw (that man)i and Bill saw himi too. (18) John hates that man because that man is a cad. (19) John hates (that man)i because hei is a cad.

On the default reading, an utterance of (16) would seem to be roughly equivalent to an utterance of (17). Similarly, an utterance of (18) is roughly equivalent to an utterance of (19). This may seem to suggest, contrary to our hypothesis, that a subsequent deictic can sometimes “inherit” its reference from an antecedent deictic. But that conclusion would be hasty. I suspect that what we really have here is not any form of anaphoric dependence of a more or less standard linguistic type, but what I call “coreference through demonstration sharing.” Coreference through demonstration is not, I think, a structural phenomenon, but a purely pragmatic one. It occurs when a speaker uses the reference-​fixing demonstration associated with one deictic to also fix the reference for a second and structurally independent deictic. When two token deictics share a demonstration in this way, they will corefer, but their coreference is a species of coincidental coreference. It is not the sort of structurally controlled coreference that is implicated in the use and reuse of a single proper name. In an utterance like (18) one indicator that we have an instance of coreference through demonstration will be stress patterns. When two distinct demonstrations are at issue there will be a different stress pattern than when a single demonstration is shared by two token deictics. In an utterance like (16) it is the “too” that signals that the two demonstratives are meant to share a demonstration. But context also plays a role in signaling that a demonstration is being shared. It is odd, for example, to hate one man because another man is a cad. And this fact surely plays a role in our interpretation of (18). Dedicated devices of de novo reference are extraordinarily handy things for both public natural languages and even the language of thought to include. The availability of such referential devices in both language and thought makes possible acts of reference that systematically exploit the vagaries of local context and situation. But as useful as such devices may be, it seems unlikely that there would be languages the referential devices of which were one and all devices of de novo reference. A language that contained only

Puzzles of Coreference  115 such referential devices would be expressively diminished, if not quite semantically then certainly pragmatically and syntactically. No doubt, speakers might still be able to refer twice over to the same object again in their thought and their talk. But it would be much more cumbersome to make it manifest and explicit that the same object was being referred to again. For in that case no referential device would function in language as a device of explicit coreference or in thought as a dedicated device for thinking with same-​purport. If a language is to have the representational resources to make the preservation and retransmission of reference not only possible, but also explicit and manifest, it clearly needs more than dedicated devices of de novo reference. It also needs devices whose function it is to make the preservation and transformation of reference explicit and manifest.17 And that is why it is hardly surprising that languages generally contain both proper names and indexicals and demonstratives. It is also why it would be a serious mistake to collapse the two categories of referential devices by simply assimilating names to demonstratives or indexicals. As dedicated devices of explicit coreference, proper names stand in sharp contrast with devices of de novo reference. It is true enough that when one tokens a given name again, there is a sense in which one tokens something brand-​new under the sun. That very token is a datable, locatable particular. It has never existed before and will never exist again. But to re-​token a name is to manifestly purport to repeat a reference. In re-​tokening “Cicero” again I do not (purport to) refer de novo to Cicero. Rather I purport to refer either to what others or I myself have purported to refer to before. And I intend that it be manifest to either myself or to others who also use that very name, either that they and I corefer or that I corefer with myself again. That is one reason why, if I were to be asked, “Was Cicero a great Roman orator,” I would 17 Brandom (1994) makes a similar point. Where I distinguish between devices of explicit coreference and devices of de novo references, he distinguishes between repeatable and unrepeatable tokenings. Names and anaphoric pronouns are repeatable, while demonstratives generally are not. Names, he says, must be “understood as anaphoric dependents—​elements in an anaphoric chain that is anchored in some name-​introducing tokening.” Moreover, Brandom argues—​quite correctly, in my view—​that anaphoric reference of the sort that requires repeatability is more fundamental than what I have called de novo reference. As he puts it: Deictic uses presuppose anaphoric ones. One cannot coherently describe a language in which expressions have demonstrative uses but no pronominal uses (although the converse is entirely possible). For indexical uses generally, like deictic ones, are essentially unrepeatable according to types. Different tokenings of “this” or “here” or “now” are not in general recurrences of each other, or even co-​identifiable. (464–​65) Though the precise letter of Brandom’s views are rather different from my own, there is significant overlap between the spirit of his view and the spirit of my own, at least on this score.

116  Referring to the World not answer by responding, “Yes, Tully was a great Roman Orator.” I would not do so, at any rate, unless the identity of Cicero and Tully was already part of the shared common ground between my interlocutor and me. The mere use of the same name again, already on its own, does the work of making the preservation of subject matter explicit, without the need to deploy identity statements. This is a convenient place for a brief digression about the role of intentions to corefer and so-​ called reference borrowing. Especially among anti-​ presentationalists who reject the view that anything like Fregean senses or mode of presentation somehow autonomously determine reference, it is widely held that intentions to corefer play a crucial role in at least the transmission of reference, if not in the initial securing of reference. I myself have never been in direct causal or cognitive contact with Socrates. Yet I use the name “Socrates” to refer to Socrates. How do I manage to make semantic contact with Socrates across the great causal and cognitive gap that separates me from any actual Socrates-​involving event? Presentationalists have an easy, if ultimately unsatisfactory, answer to this question. Fregean modes of presentations and their ilk are the very sorts of things that are supposed be able to reach out and determine reference across vast divides of time and space. Thanks to the awesome power of modes of presentation, one need not have any causal or informational contact at all with an object in order to make it an object of one’s thought and talk. For any anti-​representationalists who eschew senses, modes of presentation, and their ilk, no such easy answer is available. But it is here that borrowed reference is supposed to come to the rescue. I can refer to Socrates by my use of the name “Socrates” because I can “borrow” the reference of “Socrates.” I do so by using the name “Socrates” with the intention of referring again to what others have referred to before by their use of that name. To be sure, those from whom I have borrowed the reference of “Socrates” are likely themselves to have borrowed it from others who in turn borrowed it from still others. Ultimately, though, such chains of borrowed reference must culminate in what Kripke has called an original baptism. It is in the original baptism that reference is first secured. But it is via intentions to corefer that reference is transmitted from the original baptizers to others. This oft-​told tale is, I  think, getting at something true and important. Still, I suspect that it misunderstands the true role of so-​called intentions to corefer. It is as if the intention to use a certain name as a name is one thing and the intention to corefer via the use of a name is a separate thing entirely. It is

Puzzles of Coreference  117 as if the inner monologue of the user of a name goes somewhat as follows, “I have here the name ‘Socrates.’ I shall use it to refer to someone. But to whom shall I use it to refer I know, I will refer via my use of ‘Socrates’ to what my teacher refers in her use of that name.” But there is one intention too many at play here. What are typically called “intentions to corefer” should really be understood as part of what might be called the “name-​making machinery” for a language. By the name-​making machinery, I mean the set of institutions and practices by which what would otherwise be a mere sound/​shape pattern is promoted into an expression in the language with a tractable and distinctive syntactic and semantic role. Because a mere sound/​shape pattern such as “John A. Smith” is not directly given as a linguistically significant unit, something must be done to it to make it function as such. What are typically—​but on my view not entirely correctly—​called intentions to corefer enter the scene at this stage. These intentions serve to make it the case that various tokenings of a given sound/​shape pattern count as further episodes in a connected history of such tokenings. And they do this—​it is important to add for reasons that we will only be in a position to fully appreciate once we turn to the consideration of so-​called empty names—​independently of whether the name is referentially successful. As such, what I would rather call name-​making intentions serve to constitute an otherwise discrete collection of actual and possible tokenings of the sound/​shape pattern “John A. Smith” as what have earlier called a chain of explicit coreference. They thereby turn that collection of tokens into an actual name within the language. Thus it is that when an institution such as the Division of Motor Vehicles, the Social Security office, the registrar of a university, or one or more individuals issue tokens of the sound/​shape pattern “John R. Smith,” they do so with the intention of thereby adding yet another member to a certain chain of explicit coreference. They thereby make it the case, and this by means of those very intentions, that that very token counts as a token of the same name again—​where, of course, names are understood in the more abstract sense that I have been at pains to articulate and defend in this chapter. Thus, they each may play their part in the name-​making machinery of their shared language. But precisely this is why we cannot cleanly separate the intention to use an expression as a name, at least as this or that particular name, from the intention to corefer via the use of this or that name with other uses of this or that name. Those “two” intentions are the very same intention, only differently described. The deeper point is that to use an expression as a name, especially as this very name, is already to use the expression with the intention

118  Referring to the World to corefer. For that is what a name in its essential nature already is—​a device of explicit coreference. To use a name as this very name is thereby to issue a kind of social license—​a license to explicitly and manifestly corefer with the issuer of that license. It is through the transmission and retransmission of such licenses that we who have never had the slightest cognitive or causal commerce with the man himself manage to refer to Socrates via the name “Socrates.”

7.  Drainage as Predicativity Turn now to the view that names are really predicates rather than singular referring terms.18 Not entirely without reason, a small but persistent minority of philosophers and linguists have endorsed some version of this view. In fact, languages often do contain what I will call “name-​like predicates.” But I shall argue that name-​like predicates are not genuine names and that genuine names are not predicates. Moreover, I shall argue that the fact that languages often contain name-​like predicates in addition to containing genuine names is a trivial and unsurprising fact.19 The view that names are predicates is motivated by the fact that certain name-​like expressions—​to use a theoretically neutral phrase—​frequently occur fronted by determiners as in the following sentences: (17) Every John that I know is a philosopher. (18) There are so many Davids in this class. (19) I met yet another Sally just yesterday. (20) My son is not a typical Taylor.

It would be pretty hard to deny that the name-​like expressions in such constructions are functioning as bona fide predicates. And I won’t. Indeed, I  admit that the behaviors of such expressions raise many interesting,

18 An urge to which philosophers from many different philosophical milieus have been tempted to succumb. See Quine (1960), Burge (1973), Fara (unpublished manuscript), and Fara (forthcoming), among others. 19 See Jeshion (2014, 2015a, 2015b) for a far more developed argument against the names as predicates view. In rejecting the view that we must give a unified treatment of genuine names and name-​like predicates, Jeshion argues that the predicativist fails even to give a unified treatment of predicative uses of name-​like expressions. For an earlier criticism along similar lines to Jeshion’s, see Boër (1975).

Puzzles of Coreference  119 challenging, and delicate issues. A complete and exhaustive account of the linguistic behavior of name-​like predicates in all their uses would need to take on many such issues. Though I won't try to give an exhaustive account of the workings of name-​like predicates, I think it’s pretty easy to show that the existence of name-​like predicates has little, if any, direct bearing on singular referential uses of bona fide names. In both their linguistic role as devices of explicit coreference, and in their cognitive role as dedicated devices of same-​ purport, names behave like terms rather than like predicates. For example, in (21) and (22) below, the names function as complete minimal noun phrases and apparently directly occupy the argument places of the relevant verbs. And just as we ought to take the behavior of the name-​like predicates in (17)–​ (20) above at face value, so we ought to take the behavior of the names in (21)–​(22) at face value: (21) John is a philosopher. (22) Jack married Jill.

How can I have it both ways? How can I take all of (17)–​(22) at face value simultaneously? The answer is that I am to accept a kind of ecumenicalism about names and name-​like predicates. I see no reason to deny that names and name-​like predicates can happily coexist within a single language. By contrast, someone determined to provide a unified treatment of all name-​ like expressions, in both their apparently singular referential use and in their apparent use as predicates may try to assimilate singular referential uses of names to predicative uses of names. She may point to the following kind of exchange to bolster that impulse: (23) Jones: John is a philosopher, isn’t he? Smith: No he isn’t; he’s a plumber. Jones: Oh, you were thinking of that John. I thought you were talking about the other John, the one we met yesterday.

The defender of the names as predicate view surely must acknowledge that name-​like expressions at least appear to sometimes function as complete noun phrases and to directly occur in argument position. But in the name of unification, she will have to deny that this appearance is a reality. In reality, she will have to insist, when a name-​like expression appears to be a complete noun phrase and to occur in argument position, it is actually no such thing.

120  Referring to the World It is neither a complete noun phrase nor does it directly occupy the argument place of a verb. Rather, it occurs in the predicate position of some complete noun phrase, perhaps with a suppressed or unvoiced determiner. She might hold, for example, that the bare name “John” in (23) above is prefixed by an unvoiced determiner like “that” (Burge, 1973)  or “the” (Fara, 2015). This makes the apparent name equivalent to either a complex demonstrative like “that John” or an incomplete definite description like “the John.” I don’t intend to spend a lot of time arguing against this view. Except for a misplaced fetish for unity, there is little, if anything, to directly recommend it. There is no independent positive reason to deny that name-​like expressions occupying argument places are just what they appear to be—​complete, but minimal noun phrases that function as singular referring terms. But there is also no real reason to deny that name-​like expressions that apparently play the role of predicates are just what they appear to be—​predicates rather than terms. No matter of either principle or detail stands in the way of this simple and direct ecumenical approach. In fact, once we distinguish names from their spellings and allow that distinct names can, nonetheless, share a spelling, it is completely straightforward to introduce predicates that apply to all and only those objects whose names are spelled in a certain way. Indeed, it seems to be something close to an analytic truth that where S(N) is a spelling shared by many different names N1 . . . Nn . . . there will be a predicate P(N), such that P(N) applies to x just in case x has a name N that is spelled S(N). Hence if a language L already contains genuine names used as singular referring expressions and if spelling alone underdetermines a name’s type identity in L, then our theory straightforwardly allows that L may also contain name-​like predicates that apply to objects whose names share a spelling. Of course, any such predicate will be derived, rather than basic. Moreover, if the names in L have the property of being exhaustively spelled out—​so that no two names share a spelling—​any such predicate will be redundant. Note that although most natural languages do not, in fact, spell out their names exhaustively, there is no principled barrier to languages that do so. And if all names were, in fact, exhaustively spelled, we would feel no pressure whatsoever to assimilate genuine names to name-​like predicates. Not only are there no positive reasons for treating genuine names and name-​like predicates in a uniform fashion, there are positive reasons for not treating them in such a fashion. We have already argued that names in their singular, referential use function in language as devices of explicit coreference and in thought as devices of sameness-​purport. Neither complex

Puzzles of Coreference  121 demonstratives nor incomplete definite descriptions have the kind of coreference profile that is characteristic of names. Demonstratives are the simpler case. We have already seen that explicit demonstratives are devices of de novo reference. And there is no reason to deny that the same is true for the supposedly “hidden” demonstratives that would be needed to turn apparent names into name-​like predicates. The same is true of explicit complex demonstratives. That is, each token of an explicit complex demonstrative phrase is intrinsically referentially independent of every other token of that phrase. Thus (24): (24) That John came in and then that John sat down.

strongly defaults to a reading in which two different Johns are at issue. In contrast, (25) strongly defaults to a reading in which only one John is at issue. (25) John came in and then John sat down.

If, contrary to fact, there were antecedent reasons to take apparent names in argument position to be constituents of complex noun phrases with some sort of suppressed determiners, rather than free-​standing minimal noun phrases in their own right, the foregoing considerations might be taken to suggest that the supposed suppressed determiner would be more likely to be the definite article. That would mean that in (25) above we really have something equivalent to the incomplete definite description “the John.” Several brief observations are in order here. First, I distinguish between what I call “non-​contrastive” uses of incomplete definite descriptions from what I call “contrastive” uses of such descriptions. In non-​contrastive uses, incomplete definite descriptions are used to single out a contextually salient particular of a certain kind, without contrasting it with any other particular of the same general sort. In contrastive uses, incomplete definite descriptions are used to single out a contextually salient object of a certain kind at least partly by way of either an implied or explicit contrast with another object of that very kind. Consider (26) below, for example: (26) The cat is ferocious.

Even if no other cat is around with which to make a contrast, a sentence like (26) can be used to talk about some contextually salient cat—​and different

122  Referring to the World cats on different occasions of use.20 But now suppose that there are two cats about and suppose that one wants to contrast them in some way. One might utter a sentence like (27): (27) Though the black cat is quite ferocious, the brown cat is quite gentle.

When an incomplete definite description is used in a non-​contrastive way, it can be minimally specific and may consist of only the definite article plus a common noun. It need not contain any further modifiers. Indeed, whenever one adds additional modifiers to an incomplete definite description, some further contrast seems always to be implied, even in the absence of an explicit comparison. Compare (28)–​(30) below: (28) The cat is on the mat. (29) The black cat is on the mat. (30) The black cat with the damaged tail is on the mat.

(28) neither states nor implies a that there is a contrasting cat somewhere about. By contrast, (29) seems to at least weakly imply a contrast with a possible non-​black cat. And finally, (30) quite strongly implies a contrast with another cat, possibly black, but certainly without a damaged tail. But notice that while ordinary common nouns can occur as heads of both contrastive and non-​contrastive incomplete definite descriptions, name-​like common nouns can occur as heads only in incomplete definite descriptions used contrastively. Thus (31)–​(33) are all unacceptable, while (34)–​(36) are acceptable: (31) ?The John just came into the room. (32) ?The Mary loves the John. (33) ?The John came in and then the John sat down. (34) The John I met last night just came into the room. (35) The Mary I introduced you to loves the John you met yesterday. (36) The John I met last night came in, then the John you met left.

20 I have argued elsewhere that through the mechanism of what I call “one and half stage pragmatics,” utterances of sentences containing incomplete definite descriptions typically convey singular propositions. See Taylor (2003) and Taylor (2004).

Puzzles of Coreference  123 Since explicit name-​like common nouns seem never to be permissible as heads of non-​contrastive incomplete definite descriptions, we have, I think, strong reason to doubt that bare names as they occur in argument position are equivalent to such descriptions with unpronounced or suppressed definite articles. And this gives us reason to conclude that names in argument position are just what they appear to be—​singular referring terms. I do not claim that the considerations presented in this section conclusively settle the argument against a unified treatment of apparently predicative occurrences of name-​like expressions and apparently referential uses of such expressions. At the very least, though, they shift the burden of proof. Absent further argument, we are within our rights to acknowledge that while some name-​like expressions as occur in certain very special constructions do behave as predicates, other name-​like expressions behave quite differently. It is hard to see any reason to deny that such expressions are precisely what they appear to be—​singular, referring terms. Moreover, it is straightforwardly predictable that in languages that do not require each name within the language to be uniquely and exhaustively spelled, it is a fairly straightforward matter to introduce name-​like predicates that apply to objects solely in virtue of the spellings of any names those objects bear. In such a language, to be a John is to be a person with a name that is spelled J^o^h^n. But the fact that such name-​like predicates may easily be introduced into any such languages is no reason whatsoever to deny the existence of genuine names within such languages.

8.  Coreference Revisited In this section, I  begin the process of rethinking the coreference puzzles more or less from the ground floor up. I do not pretend to offer an exhaustive treatment of the entire inventory of coreference puzzles here. Though my ultimate aim is to show that all or at least many coreference puzzles yield to an at least somewhat unified approach, my current focus is mainly on Frege’s original puzzle about the very possibility of informative identity statements. At this stage of the overall argument, my main aim is to offer a fresh diagnosis of the source and significance of that granddaddy of the many coreference puzzles. That diagnosis is meant to set the stage for two central further developments. First, it will enable me to put on a fuller display what I regard as Frege’s foundational mistake in thinking about the nature of reference and

124  Referring to the World coreference. Second, it will represent a further step in my overall attempt to reorient thinking about the coreference puzzles away from semantics and toward a certain mixture of syntax and pragmatics. We begin with a deeper dive into Frege’s own understanding of the ultimate source of his puzzle about the cognitive significance of informative statements of identity. The phrase “cognitive significance” or, alternatively, “information value” was clearly something of a term of art for Frege. It connoted an intuitive, pre-​theoretical notion that he nowhere explicitly defined with precision. He does offer a measure of sameness and difference of cognitive significance. We might reasonably suppose that this intuitive measure of sameness and difference is meant to implicitly precisify the notion of cognitive significance to some greater or lesser degree. According to Frege’s intuitive measure, any two sentences Φ and Ψ differ in cognitive significance, if it is possible for a competent, rational cognizer who fully grasps both Φ and Ψ to simultaneously accept what is expressed by Φ while rejecting or being agnostic about what is expressed by Ψ. Correlatively, two terms t and t' differ in cognitive significance only if given two sentences Φ(t) and Ψ(t') which differ only in that where t occurs in Φ(t) , t' occurs in Ψ(t'), then Φ(t) and Ψ(t') differ in cognitive significance. Call this Frege’s Cognitive Criterion of Difference (CCD).21 To say precisely what we are measuring when we measure sameness and difference of cognitive significance against CCD, we would need to say more about each term in this specification. We would need to specify, for example, what “competent rational cognition” or “fully grasping” a sentence consists in. Although Frege does not undertake that task directly and explicitly, a certain understanding of the nature of competent rational cognition is implicit in his writings. And that implicit understanding plays a significant if seldom foregrounded role in many of his arguments. It would be philosophically interesting and challenging to make explicit what is left merely implicit in Frege’s many arguments. I will not attempt to do so here, though I will I have occasion to highlight at least some of the central elements of Frege’s implicit understanding at various points.

21 Notice that as stated, this implicit measure assumes but does not define what “competent, rational, cognizer” is like. Nor does it articulate any explicit understanding of what it is for such a cognizer to “fully grasp” what a sentence expresses. Certainly, Frege himself had certain substantive views about all these matters. And if you buy into the complete package of his views, you get Frege-​ style solutions to the various puzzles almost for free. It should not surprise that I will be articulating a rather different total package of views, and these are bound up with a rather different approach to the many coreference puzzles favored by Frege.

Puzzles of Coreference  125 As measured by CCD, a trivial identity a = a differs in cognitive significance from a non-​trivial but true identity a = b precisely because a rational, competent cognizer who fully understands both statements can know that a is identical to a while intelligibly wondering whether a is identical to b. Correlatively, the coreferring terms “a” and “b” differ in cognitive significance because they, in effect, make different contributions to the cognitive significance of sentences in which they occur. Now Frege famously held that in order to solve the cognitive significance puzzle, names must be construed as having two distinguishable, though related, semantic roles. Besides the role of denoting its reference, a name must also have the role of expressing a sense, where a sense, recall, is supposed to contain a “mode of presentation” of a potential or actual reference. Modes of presentation somehow “determine” reference—​though Frege never explicitly spells out what the determination relation between sense and reference consists in. But the rough intuitive idea behind much of his thinking seems to be that the relationship between a sense and the reference it determines is a relation of “fit” or “satisfaction.” The reference of a sense having expression will be whatever best “satisfies,” “fits,” or “answers to” the relevant mode of presentation. Frege took senses to be the stuff of which thoughts are made. The sense of a name is a constituent of any thought expressed by any sentence in which the name occurs. Such senses combine with other senses—​for example, the sense of verbal expression—​to yield complete thoughts. Senses thus have something of a Janus-​faced character for Frege. They are the building blocks of thought and thought content. They also point beyond thought and thereby serve to render thought semantically answerable to the world. The Janus-​ faced character of sense is crucial at many different turns in Frege’s thinking. It helps, for example, to solve what I have called the mystery of objective representational content. It is crucial to his understanding of the very nature of competent rational cognition. It helps to explain the apparently peculiar semantic behavior of expressions embedded within the context of propositional attitude statements. For our current purposes, though, we focus on only one of the many consequences of the Janus-​faced character of sense—​its role in explaining the cognitive significance phenomena and in solving the coreference puzzles. What is crucial for that purpose is the fact that senses are more finely individuated than reference. This allows for the possibility that distinct senses, containing distinct modes of presentation, may nonetheless determine the very same reference. Or starting with reference rather than sense and working up rather than down, we may say that for any given

126  Referring to the World object that may be determined by some sense or other as reference, there will be many distinct senses that may determine that very object as reference. Just given an object, in other words, there is no way “back up” from the object itself to the sense that determines it as reference. This combination of views, together with the view that senses are the very building blocks of thought, enables Frege to maintain that coreferring names need not make identical contributions to the thoughts expressed by the sentences in which they occur. And it is precisely differences at the level of sense that are thus the ultimate source of the very possibility of true but informative statements of identity. This enables the thought content expressed by a statement of the form ⌈a = a⌉ to be distinct from the thought content expressed by a statement of the from ⌈a = b⌉ even when a just is b. Frege’s solution to the cognitive significance puzzle is elegant, systematic, and comprehensive. Given all that has come before in this book, however, it will come as no surprise by now that I reject it more or less root and branch. In my estimation, Frege goes wrong from the very start in that he mislocates the source of the cognitive significance phenomena in semantics and content rather than in syntax and form. I have laid much of the groundwork for my own syntactic account of the cognitive significance phenomena in this and previous chapters. Once we have on board such notions as coreference profiles, relations of same-​purport, and the distinction between explicit and merely coincidental coreference, it is relatively straightforward to show that there will be a felt cognitive difference between identity statements in which two explicitly coreferential designators flank the identity sign and identity statements in which two merely coincidentally coreferential designators flank the identity sign. When an identity statement is flanked by two explicitly coreferential designators the statement is guaranteed to express a truth, at least so long as the designators are not empty. And it is guaranteed to do so solely on the basis of the broadly syntactic character of the relevant expression. And because of this, it will be “trivially” true in the sense that no substantive investigation will be necessary to vouchsafe its truth. By contrast, where two merely coincidentally coreferential designators flank an identity sign, the relevant identity statement is bound to have what I will call an informative feel. But form alone will not settle whether the statement is true or false. To close the question of truth or falsity, we will have to appeal to factors beyond mere form. It should be said that none of what I have just said is incompatible with the strict letter of Frege’s own views. Where Frege and I  part company

Puzzles of Coreference  127 is in our respective understandings of what I have called the felt informativeness or triviality of an identity statement ultimately consists in. For Frege, it is propositions or propositional contents themselves that are, in the first instance, informative or trivial. Representations, like the sentences “Hesperus = Hesperus” or “Hesperus = Phosphorous” or the correlates of such representations at the level of thought, are informative or trivial only in a derivative sense. Representations—​and here I mean to specify the vehicles of our thought and talk rather than the contents of our thought and talk—​ can be said to be informative solely in virtue of the informativeness or triviality of the propositions they express. Or to put it differently we might say that for Frege, the informativeness of that thought or claim that Hesperus is Phosphorous rests on, as it were, the content of that thought or claim. Similarly, for the triviality of the thought or claim that Hesperus is Hesperus. It would be hard to deny that there is a certain deeply intuitive pull to Frege’s picture. Not without reason have subsequent generations of philosophers been inspired by Frege to distinguish the content of the proposition that Hesperus is Hesperus from the content of the proposition that Hesperus is Phosphorous. Of course, there are lots of different ways of understanding just what propositions are, but however propositions are understood, it is this picture that I want to deny. On my view, informativeness and triviality—​or better the properties that explain felt informativeness and triviality—​are not, in the first instance, properties of propositions at all. They are rather properties of mental and linguistic representations. Informativeness and triviality are rooted, in other words, not in facts about propositions and their contents as such, but in facts about relations among the mental and linguistic vehicles we use to express various propositions. In denying that informativeness and triviality are properties of propositions and their contents, I do not mean to deny that informative statements of identity may also convey, and perhaps typically do convey, information packaged as propositional. My claim is that the ur-​phenomenon of cognitive significance does not have to do with propositional contents at all. This claim is entirely consistent with allowing that when we learn anew that Hesperus is Phosphorous—​especially when we learn this anew as a consequence of first hearing, understanding, and accepting a sentence such as “Hesperus  =  Phosphorous”—​we may acquire information packaged in propositional form—​information that we would not expect to acquire from hearing, understanding, and accepting the sentence “Hesperus = Hesperus.” For example, from hearing, understanding, and accepting anew the sentence

128  Referring to the World “Hesperus  =  Phosphorous” we may come to learn anew that the name “Hesperus” and the name “Phosphorous” are coincidentally coreferential. But it simply does not follow from the fact that we may learn anew that two referentially independent designators are coincidentally coreferential that we must therefore identify the proposition that Hesperus is Phosphorous with the proposition that “Hesperus” and “Phosphorous” denote the same object. Frege himself rightly pointed out the latter is a proposition about two signs and their relations to each other, while the former is a proposition about an object and its relation to itself. One ought not to conflate these two propositions. And the approach on offer here does not do so. Strikingly, Frege himself seems to have been guilty of such a conflation in his own first attempt to solve the cognitive significance puzzle. He argued that an identity statement expresses not a relation between an object and itself, but a relation between two signs. And he seemed tempted by the conclusion that what we strictly literally say when we say that Hesperus is identical with Phosphorous is that the two signs “Hesperus” and “Phosphorous” stand for the same object. Frege eventually saw the error of his ways and came to reject this solution. He rightly concluded that this analysis attributes the wrong subject matter to the statements in question. While I fully concur with Frege’s view that the proposition that Hesperus is Phosphorus is not a proposition about the signs “Hesperus” and “Phosphorus,” he goes too far when he concludes that that which explains the cognitive significance of an identity statement must therefore be an ingredient of the content of the proposition expressed by such a statement. On my view, Frege commits what John Perry has aptly called a “subject matter fallacy.” One commits a subject matter fallacy, according to Perry, when one supposes that the content of a statement or belief is wholly constituted by the conditions its truth puts on the subject matter of the statement or belief; that is, the conditions it puts on the objects the words designate or the ideas are of. (Perry 2001, 50)

In stark contrast to Frege, Perry attempts to explain cognitive significance phenomena by appeal to a distinction between what he calls referential content and what he calls reflexive content. The reflexive contents of an utterance—​there are many of these—​are, roughly, propositions about that very utterance, variously characterized, and the conditions, also variously characterized, under which that utterance is true. The referential content of

Puzzles of Coreference  129 an utterance, on the other hand, is typically a proposition about some (extra-​ linguistic) object—​the object that is the subject matter of the relevant utterance. Armed with the distinction between reflexive and referential content, Perry is able to maintain that even where there is no difference in referential or subject matter content between two sentences that differ only by the presence of coreferring names, there can still be differences in the reflexive contents of such utterances. His crucial further claim is that although what is strictly and literally said by an utterance is a matter of the referential or subject matter content of the utterance, the cognitive significance of an utterance is, in the general case, a matter of its reflexive contents. In distinguishing the cognitive significance of an utterance from what is strictly literally said or expressed by that utterance, Perry’s approach to the cognitive significance phenomena represent in my view a significant advance over Frege. Perry is entirely right to deny that the ultimate source of the cognitive significance phenomena is differences at the level of referential content. He is right also to focus on the roles of representational vehicles by means of which referential contents are expressed rather than on those contents themselves. Still, I do not think Perry has succeeded in isolating the true and ultimate source of the cognitive significance phenomena. I do not mean to reject the very idea of reflexive contents. There are such contents. And they are just as numerous and varied as Perry says. And such contents are sometimes either tacitly or explicitly conveyed as part of what might be called the total communicative content of a speech act. Nor do I mean to deny that reflexive contents play a significant role in what we might call the epistemology of linguistic understanding. I grant, for example, that it is only because we can recognize and process reflexive contents that we are able to incrementally grasp, as it were, the referential contents of our utterances. Nonetheless, Perry’s reflexive contents still seem to me to be located somewhat downstream from the ultimate source of the cognitive significance phenomena. The ultimate source of those phenomena is not any sort of proposition or propositional content—​not the general propositions of the Fregean, not the singular propositions of garden variety direct reference theorists, not Perry’s reflexive contents, not even structured propositions. To see why I say this, recall some points made in passing earlier in my overall argument. Suppose Jones says, “My, Hesperus looks lovely this evening!” And suppose that Smith wishes to express either agreement or disagreement, as the case may be, with Jones. Suppose, further, that she wishes it to be mutually manifest to both herself and to Jones that she is doing so. She

130  Referring to the World can achieve this result simply by using again the name that Jones originally used. She can utter a sentence like “Yes, Hesperus does look lovely this evening!” Or if she disagrees, she can say “No, Hesperus does not look particularly lovely this evening! It looks rather dim and anemic!” The initial point is that by using the same name as Jones originally used, Smith succeeds in doing something of great significance. She makes it explicit and mutually manifest that she and Jones are talking and thinking about the very same object. Suppose that Smith had spoken differently. Suppose that she had continued the conversation by using the name “Phosphorus” instead of “Hesperus.” Perhaps she says to Jones “My, but Phosphorus does look lovely this evening!” Smith would still thereby preserve the subject matter of the conversation. She would still be predicating the very same property of the very same object. She would, in Perry’s sense, be communicating the same referential content that she would have expressed had she used “Hesperus” instead. There is therefore a robust sense in which she would still be expressing agreement with Jones by using “Phosphorus” rather than “Hesperus.” We might call this sort of agreement “agreement in referential content.” Clearly, though, the word choice here would not serve to make the agreement in referential content explicit and mutually manifest. Indeed, if it is not already mutually manifest to both Smith and Jones that Hesperus just is Phosphorus, then, depending on the exact context, it may feel to Jones as if Smith has shifted the subject matter—​despite the fact that she has actually preserved it. Correlatively, if it is somehow manifest to Jones that Smith doesn’t intend to be shifting the subject matter and thus manifest that she still intends to be talking about the very same object as Jones, despite her choice of names, then Jones may take Smith to be implicating that “Hesperus” and “Phosphorus” corefer. At the very minimum, by shifting to a referentially independent name, the coreference of which with “Hesperus” is not already mutually manifest, Smith has left open the question of whether she has, in fact, preserved the subject matter and with it referential content. Several lessons can be drawn. First is that the distinction between explicit and coincidental coreference has deep pragmatic significance. The pragmatic significance of this distinction arises from the fact that in cooperative conversations, discourse participants are constrained not just to preserve subject matter but also to preserve subject matter in an either linguistically or contextually marked fashion. Devices of explicit coreference are dedicated tools for doing just that. Merely coincidentally coreferential expressions may

Puzzles of Coreference  131 sometime serve this function too, at least relative to an appropriate common ground. But it is crucial that they never do so merely in virtue of their dedicated linguistic character and function alone. That is why, absent very special circumstances, the use of referentially independent, merely coincidentally coreferential designators tends to generate (defeasible) imputations of distinctness. And that is why despite the coincidental coreference of “Hesperus” and “Phosphorous,” substitution of the one for the other is not guaranteed to preserve what might be called “dialectical significance.”22 There is a still deeper lesson to be drawn, one that is more directly pertinent to understanding the ultimate source and nature of the cognitive significance phenomena. Consider what happens when an identity statement that links two referentially independent but coincidentally coreferential designators is added to the common ground of a developing conversation. Focus not on the proposition expressed by any such sentence. Focus rather on the way the occurrence of such a sentence in the context of either a conversation (or an inner thought sequence) changes what I will call the dynamic landscape of the conversation (or thought sequence). Identity statements and identity thoughts change the dynamic landscape in far-​reaching ways. In a conversation, instead of transitions from statements involving the one designator to 22 But I stress that to say that substitution of merely coincidental coreferents is not guaranteed to preserve dialectical significance is not to say that such substitution fails to preserve truth value. Preservation of truth value is one thing, preservation of dialectical significance is something entirely different. The former does not require manifestness. The latter does. And in this I  take myself to be in disagreement with Kit Fine’s (2007) relational semantics. Moreover, my approach helps explains why we have different reactions to sentences like (a)  and (b) below: (a) Clark Kent went into the phone booth and Clark Kent came out. (b) Clark Kent went into the phone booth and Superman came out. Contra Saul (1997) and Saul (2007) (a) and (b) do not differ in truth-​value. (Here I bracket the fact that since there is no Clark Kent and no Superman either, each of (a) and (b) are both, on my view, truth valueless. See Taylor (2013) for a more realistic treatment of such fictional names.) Nonetheless (a) and (b) can, in the right setting, differ in dialectical significance, since (a) does and (b)  does not generate any imputation of distinctness. The author of the Superman comic books cleverly exploits the referential independence and coincidental coreference of “Clark Kent” and “Superman” to very good pragmatic effect. In particular, it is part of the background story, within which these names are at home, that although Clark Kent and Superman are one and the same person within the story, this fact is unknown to most characters in the story—​though it is known to the readers of the story. It would be an interesting exercise to trace all the pragmatic effects of this setup in greater detail. I lack the space to do so here. I will just say that it seems to me that whenever the two names are used together in the same sentence or context, the reader is supposed to attribute a sort of in the story d ­ istinctness to Clark and Superman at least for the other characters, while recognize that the characters are, in fact, confused and that in the story Clark Kent just is Superman. Moreover, the use of either name in isolation from the other evokes in the story continuity of reference and identity. But the deeper point is that this subtle exploitation of the facts about the coreference profiles of names depends entirely on the broadly syntactic character of names and does not require us to any way bring to bear any sort of semantic analysis of names and naming.

132  Referring to the World statements involving the other having the character of transitions involving de novo acts of reference, such transitions are marked as manifestly and explicitly preserving subject matter. Consequently, transitions that might otherwise have had the character of disjointed, subject shifting discourse are transformed into discourse with a more unified because explicit subject matter-​preserving character. A similar effect occurs at the level of individual thought, as a consequence of identity thinking within a single mind. Consider two different sequences of thought episodes about the one heavenly body Venus. Suppose that the A  sequence looks something like this:  {t:  >, t':  } , while the B sequence looks something like this: {t: >, t':  }. In sequence A, the two thought episodes share a subject matter, though their vehicles bear no intrinsic relation of same-​purport to each other. As such, they are dynamically independent and are not knitted together in the thinker’s mental economy so as to constitute an interconnected chain of reasoning. Things are otherwise with sequence B. Here the thought vehicles do stand in relations of intrinsic purport of sameness to each other. Thinking of Venus via the deployment and subsequent redeployment of an inner *Hesperus* representation is not thinking of the same thing twice via two cognitively independent representations. It is thinking of the same thing again as the same thing again. It is thinking of the same thing again with an intrinsic purport of sameness. It is this intrinsic purport of sameness that distinguishes streams of cognitively independent thoughts—​thoughts that may or may not be thoughts about the same thing again—​from what I am calling dynamically interconnected chains of reasoning. In dynamically connected chains of reasoning, subject matter is not just preserved but manifestly and explicitly preserved and this in virtue of the structural relations of same-​purport among the thought vehicles within that chain. Now suppose we add an identity thought to the A  sequence and form new thought sequence A*:  {t:  >, t': <  =  >, t'': }. We get a thought sequence radically different from the original. In A* the *Hesperus* thought and the *Phosphorus* thought still enjoy no intrinsic purport of coreference. Nonetheless, the two thought episodes are no longer cognitively independent and dynamically unlinked. They are dynamically linked in virtue of the identity thought. Identity thinking thus

Puzzles of Coreference  133 achieves at the level of thought what identity statements achieve at the level of talk. Via the joint deployment of referentially independent referential devices in episodes of identity thinking, representations that do not enjoy intrinsic purport of coreference with each other come to enjoy what might be called an extrinsic purport of coreference, at least in the context of the mind that so deploys them. We are now in a position to better appreciate where Frege went wrong. His fundamental mistake was to assume that sameness and difference of cognitive significance must be rooted in sameness and difference of thought content. As a consequence, he took the thought content that Hesperus is Hesperus to be distinct from the thought content that Hesperus is Phosphorus. If one fails to grasp the cognitive import of the distinction between intrinsic purport of coreference and referential independence among coincidentally coreferential designators, this may seem like a natural way of counting thought contents. But once one is armed with that distinction, it should seem unnecessarily promiscuous and to multiply thought contents beyond necessity.23 I will defend this last claim over the course of the next several chapters. The crucial point for our current purposes, though, is that the distinction between intrinsic purport of coreference, and referential independence among coincidentally coreferential designators enables us to appreciate that cognitive and communicative function of identity thought and talk is to recompile referentially independent representations in their relations to one another. Where there is originally no intrinsic purport of coreference between our representations, an identity thought or statement linking referentially independent but coincidental representations serves to reconfigure those representations in relation to each other so that they come to enjoy not an intrinsic but an extrinsic purport of coreference. Though my approach to cognitive significance departs in many ways from both the strict letter and the fundamental spirit of Frege’s approach, it is nonetheless fair to say that there is still a way in which at least the spirit of Frege still suffuses my approach. Admittedly, the way in which this is true is not altogether straightforward to explain, thanks mainly to various knots into which Frege tied himself. But untangling some of those knots will help us approach the true source of the cognitive significance phenomena from a different direction. The first thing we need to notice is that although he did not 23 See Taylor (1995) for an argument to the effect that individuating thought contents by appeal to Frege’s Criterion of Cognitive Difference will lead to an explosion in thought contents.

134  Referring to the World himself see it that way, Frege’s notion of sense is really something of a mongrel concept, one that serves many distinct and independent masters simultaneously. Sense was pressed into service by Frege for many different explanatory ends: to explain the cognitive significance phenomena; to determine reference; to explain something like intrinsic purport of coreference; to constitute the contents of our thoughts; and to be the referents of that-​clauses. Many have doubted that there is any one thing that plays the multiplicity of roles that Frege promiscuously attributed to sense.24 By disaggregating these roles and taking them one by one rather than all at once, we throw the mongrel character of sense into sharper relief. But my aim here is not to investigate each of the roles assigned by Frege to sense and to independently determine what sort of thing is best suited to play that role. Here I focus like a laser on sense understood as a potential basis for distinguishing between two sorts of representation-​representation relations: representations that enjoy intrinsic purport of coreference one with another and representations that, though possibly coincidentally coreferential, are nonetheless intrinsically referentially independent. Frege never explicitly foregrounded any such distinction himself. But it is fairly easy to see that his views imply something like this distinction. For example, despite Frege’s failure to explicitly foreground any such distinction explicitly, it is clear that on Frege’s view expressions that share a sense will necessarily share a reference and that this fact is somehow a priori knowable. They will do so even if they both entirely lack a reference—​though this is a little more complicated, since Frege’s views about expressions with sense but no reference underwent considerable evolution. Still, it seems right to say that if a sense can be fully determinate even if it fails to determine an actually existent object as reference, the sense may have what I might call fully determinate referential purport even in the absence of an actually existing reference. And that this is so seems to have been something Frege took to be epistemically manifest to the thinker herself. The next step is to appreciate that for Frege if two expressions differ in sense they will necessarily differ in referential purport. This is so even if they happen to share an actual reference. Indeed, without this claim, his explanations of the possibility of informative identity statements would simply collapse. Things get a little trickier, to be sure, when we turn to senses that fail to determine a reference. Should 24 I am not the first to deny that one thing is fit to play the multiplicity of roles that Frege attributed to sense. See Burge, (1979), for example.

Puzzles of Coreference  135 we think of “Sherlock Holmes” and “Santa Claus” as determining the same reference—​the null object, perhaps—​while differing in intrinsic referential purport? Or should we simply not speak either of reference or even intrinsic referential purport at all in such cases? Frege himself seems to be of different minds at different stages of his thinking about such matters. For our current purposes it is enough to see that whatever he ultimately wants to say about the reference of such expressions, he has the resources to say that “Santa Claus” and “Sherlock Holmes” differ in their “intrinsic” referential purport and are, therefore, in that sense, referentially independent, even if we want to say either that neither determines a reference at all or that both determine the same null reference. Bracketing all other explanatory roles that Frege assigned to sense and focusing solely on sense in its role as that which grounds a distinction between intrinsic purport of coreference and referential independence, a picture more compatible with my own syntactic approach to such matters can begin to be discerned in Frege. The starting point is to observe that sharing a sense is sharing a referential purport, but sharing a sense is not yet sharing an actually existent reference, because of the possibility of sense-​having, non-​ referring names. Similarly, because of the possibility of coreferring names that differ in sense, sharing reference is not yet sharing referential purport. If we then restrict ourselves simply to asking what a notion of sense would have to be if it is to explain both representation-​representation relations of intrinsic purport to corefer and representation-​representation relations of referential independence, we might well get a notion of sense that is purely syntactic in precisely the sense that has been at play throughout this chapter. Indeed, as long as we did not demand that sense so understood should also determine reference and constitute the contents of thought and be that which is denoted by embedded terms, there would be no real barrier, I think, to understanding “sense” in such structural-​syntactic terms. Or to put it differently, the pressure to see sense as an essentially semantic notion comes more from its supposed role in explaining representation-​world relations than from its assumed role in explaining representation-​representation relation. A stripped down, structural-​syntactic understanding of sense would, I think, fully suffice for explaining Frege’s original puzzle about the possibility of informative identity statements, at least if the puzzle were properly characterized. This is, of course, not at all how Frege himself understood matters. But I claim that he was prevented from doing so not because of the nature of the

136  Referring to the World case but solely because he heaped such a hodgepodge of explanatory burdens onto the notion of sense that he was in no position even to entertain the possibility that the determination of reference might be one thing while the determination of relations of same purport and/​or referential independence is an entirely different thing. I have been at pains to put precisely that possibility firmly on board. And given that my goal is to disaggregate various elements of Frege’s mongrelized notion of sense, my approach might well be seen less as a flat-​out rejection of the Fregean tradition and more as a corrective to that tradition. There is no deep harm in taking matters this way. From my perspective Frege did succeed in unearthing a deeply significant phenomenon that undergirds and explains much about identity thought and talk. I claim that that phenomenon is, in the end, a broadly syntactic phenomenon that hinges on facts about structure and form. Unfortunately, Frege himself more or less completely mischaracterized the phenomenon that he had unearthed. That mistake has reverberated throughout the philosophy of language with far-​reaching consequences for our theorizing about the nature and role of singular reference. Where Frege sees two names sharing a reference but differing in sense, there are really just two names that are referentially independent, but coincidentally coreferential. Where Frege sees the same sense again, we see the characteristic coreference profile of a device of explicit coreference. What Frege failed to see is that from a syntactic perspective names are quite distinctive devices of reference and coreference. To repeat a name is ipso facto to repeat a reference. To refer again to the same object, but using a different name is, in effect, to refer relatively de novo to the relevant object, that is, in a way not “anaphorically” linked with the previous act of reference. In such facts, and such facts alone, lies the ultimate source of the cognitive significance phenomena and the epistemic one-​sidedness of reference.

9. Comparisons It is worth briefly contrasting my own view with a view that is similar in spirit but differs in detail from my own.25 I have in mind the views of Fiengo and May (2006). Though they seem to agree with much of the general spirit of what I say in this chapter, they introduce a notion of de dicto content apparently to preserve certain Fregean intuitions. But given the arguments of

25 [This section was a footnote to the last sentence in Taylor’s original text.]

Puzzles of Coreference  137 this chapter it is hard to see why any view roughly in the same spirit as mine should be moved by Fregean considerations. To make clear the points of contention between their view and my own, I need to introduce three bits of their own technical apparatus: (a) the notion of a translation statement; (b) the notion of an assignment; and (c) the distinction between a de dicto and a non-​de dicto logical form. A translation statement is a statement of the form: (25) ⌈“X” translates “Y”⌉

where X and Y are expressions that may be either co-​indexed or not co-​ indexed. This notion of “translation” seems intended, at least in part, to capture speakers’ beliefs about coreference relations—​both coreference relations of the linguistically marked variety and coreference relations of the linguistically unmarked variety. Co-​indexed expressions are guaranteed to be translations of one another. Speakers have purely linguistic or grammatical grounds for their knowledge of such translation statements. But even expressions that are not co-​indexed can stand in the translation relations to each other. There are, however, no purely linguistic grounds, according to Fiengo and May, for knowledge of such translation statements. As they put it, “Translation statements may also hold between noncoindexed expressions, and these are learned case by case.” Consider next the notion of an assignment. An assignment, they say, is a “semantic belief ” about the reference of an expression. Such beliefs may be characterized by sentences of the form: (26) ⌈“[NPi X]” has the value NPi⌉

where X is a schematic letter covering the syntactic contents of NPi. Hence: (27) “[NPi Cicero]” has the value Ciceroi

and: (28) “[NPi Tully]” has the value Tullyi

represent beliefs about the semantic values of “Cicero” and “Tully,” respectively. Now Fiengo and May use a system of indices to mark what they call

138  Referring to the World expression identity. In their system, if NPs are occurrences of the “same expression” then they are co-​indexed. And if NPs are not occurrences of the same expression, they are not co-​indexed. Co-​indexed expressions are coreferential. Expressions that are not co-​indexed may also corefer, but their coreference will not be linguistically marked by the system of co-​indexing. Finally, consider the distinction between what they call de dicto and what they call non-​de dicto logical forms. A dicto logical form, roughly, is a logical form that has an assignment as an explicit constituent, while a non de dicto logical form is a logical form that does not have an assignment as a constituent. According to Fiengo and May (2006, 63): When speakers make assertions the Assignments believed by the speakers need not be part of the logical form of the sentences that speakers utter. They may be, rather, elements of the context that the speaker assumes. Assignments, however, are not always just background to assertions; they may stand as part of the truth conditions of the assertion itself, in which case they give rise to apparent exceptions to the Assignment principle, . . .

In particular, Fiengo and May claim that both in the context of identity statements and in the context of (de dicto) ascriptions of propositional attitude ascriptions, assignments make their way into asserted truth conditional contents. To be sure, they acknowledge that although we may sometimes directly ascribe belief in an assignment to an agent, typically such ascriptions are merely tacit. But they claim that assignments need not be explicitly represented in order to make their ways into truth conditional contents. Speakers sometimes . . . commit themselves to a claim about the terms under which the agent holds a belief, and provide information as to what the agent believes over and above that given by the overt primary attribution. (Fiengo and May, 64–​65.)

I reject this sort of approach. To appreciate why, begin by considering the following two statements: (29) John believes that Cicero was a Roman. (30) John believes that Tully was a Roman.

Puzzles of Coreference  139 Fiengo and May claim that (29) and (30) each have two logical forms—​a de dicto logical form and a non-​de dicto one. The two logical forms of (28) are represented by (31) and (32), while the two logical forms of (30) are represented by (33) and (34) below: (31) John believes that [Cicero1 was a Roman] (32) John believes that [[Cicero1 was a Roman] and [“Cicero1” has the value Cicero1]]] (33) John believes that [Tully2 was a Roman] (34) John believes that [[Tully2 was a Roman] and [“Tully2 “ has the value Tully2]]

The idea is that a speaker has the option of ascribing an attitude in two different ways, using two different “logical forms.” She will use a non-​de dicto logical form when her ascription is not intended to reflect the names that her ascribee herself would use to refer to the objects that the ascribed belief is about. She will use a de dicto logical form when her ascription is intended to reflect the ascribee’s own use. But I suspect that Fiengo and May have, like Frege, committed a subject matter fallacy. It may well be correct to say that in certain contexts it is possible to “convey” information about “assignments” by an utterance of an attitude ascription. But even if one can show that such information can be conveyed in a given conversational setting, by a given utterance, it wouldn’t follow from that fact alone that claims about assignments enter into the strict literal truth conditions of the relevant utterance. In fact, it is not obvious how and why such information could manage to become part of the strict literal truth conditions of attitude ascriptions. To be sure, in claiming that de dicto logical forms are just that, logical forms, Fiengo and May may mean to suggest that at some level of representation a sentence like (29) above has a structure like that of (32) above. But Fiengo and May adduce no evidence at all that (29) is structurally ambiguous. Alternatively, they might think that (29) has some hidden or suppressed constituent that possibly calls for a contextually supplied assignment as its value. But I know of no reason to think this. One could also think that assignments get into truth conditions as either Perryesque unarticulated constituents or as the consequence of something like Recanati’s free enrichment. Now I  have argued against all such approaches elsewhere and won’t repeat these arguments here. My point here is not that those approaches have been thoroughly refuted. It’s just that Fiengo

140  Referring to the World and May don’t tell us anything about how a sentence like (29)—​which makes no explicit mention of any metalinguistic information about assignments, which does not obviously contain some hidden slot or parameter that might take on an assignments as its value, which seems in no way structurally ambiguous or semantically incomplete on its face—​would manage to have a logical form and truth conditions like that represented in (32). Fiengo and May owe us some story, but their book contains no such story. Another view that shares much of the spirit of the views presented here is due to Fine’s (2007). But Fine wishes to complicate the semantics in ways that I  find unnecessary. In particular, he thinks that we have to build representations of what we might call referential coordination directly into the propositional contents of our sentences. Thus on his view, “Cicero = Cicero” and “Cicero = Tully” will, in one sense, express the same singular proposition. But the proposition expressed by the first will contain a representation to the effect that “Cicero” contributes Cicero not once but twice, while the proposition expressed by the second will contain two uncoordinated references to Cicero. According to Fine, we must distinguish representations that represent the same object from representations that represent objects as being the same. He argues that “Cicero = Cicero” represents Cicero “as the same,” while “Cicero = Tully” represents the object as being the same without representing the object as the same. Though I think Fine is on to an important distinction, what he sees as part of the semantics, I see as part of syntax. Part of the difference between Fine and me, however, may be that he sees syntax as a matter of mere typography. On my view, of course, syntax concerns considerably more. It has to do with expression classes, their formation rules, transformations that can be made on them, governance relations, anaphoric dependency relations, coreference profiles, and the like. Syntax so understood not only helps to explain various distributional facts but also to “constrain” without fully determining semantic interpretation and relations. Perhaps if Fine were to adopt this broader notion of syntax, his view and mine would be considerably closer to converging.

10. Conclusion I began this chapter by outlining themes and variations among the many coreference puzzles. I developed some machinery designed to provide a fresh take on both the nature and the origin of the many coreference puzzles.

Puzzles of Coreference  141 I applied that machinery to Frege’s original puzzle about the possibility of true but informative statement of identities, partly with the aim of showing that Frege’s mongrel notion of sense was not forced on us by his original coreference puzzle. So far, I have not attempted to give an exhaustive treatment of the many variations on the original Fregean theme. About some of these variations, I will have a great deal more to say in subsequent chapters, especially about the coreference puzzles having to do with substitution failures within so-​called opaque or oblique contexts. Over the course of Chapters 5 and 6 I will argue against the Frege-​inspired tradition that takes propositional attitude contexts to be semantically opaque and takes the semantics of embedded terms to be non-​innocent. In contrast with the current chapter, which found an explanatory role for a rather stripped down de-​mongrelized notion of sense that still bears a certain marginal resemblance to Frege’s original notion of sense, I shall argue that nothing recognizably sense-​like plays much of role in either the psychology of the referring mind or the semantics of propositional attitude statements.

5 Concepts, Conceptions in the Psychology of the Referring Mind In Chapter 4, I laid out theme and variation on the many coreference puzzles. Though I did not offer solutions to all the variations, I did treat the original Fregean theme in some detail. My principal aim was to lay bare the mongrelized character of Frege’s original notion of sense and to demonstrate that we need not allow coreference puzzles to serve as a breeding ground for such mongrels. I will have more to say about further variations on the coreference puzzles in due course. I will argue in the next chapter, for example, that the Frege inspired tradition that takes propositional attitude statements to be so-​called oblique or opaque contexts and that takes the semantics of embedded terms to be non-​innocent is founded on an illusion—​the illusion of opacity. Far too many philosophers have been bewitched by the illusion of opacity, and this bewitchment has led them to misunderstand what we are doing when we ascribe propositional attitudes. But before turning to take up the substitution puzzles in an attempt to dispel the illusion of opacity, we have to take a deeper look into the psychological organization of the referring mind. Doing so will enable us to achieve a deeper understanding of precisely what we are doing when we represent our representations in the course of the ascription of propositional attitudes. And that understanding will put us in a better position to both dispel the illusion of opacity and to appreciate the true source and nature of the substitution puzzles. We have already taken some initial but important steps toward understanding the psychology of the referring mind. We have seen, for example, that the referring mind comes antecedently equipped with the psychic capacity for thinking with inner purport of sameness. We have argued that mental names function in the logical syntax of thought as devices for thinking with such inner purport of sameness. And we have conjectured that it is precisely because names in a shared public language are the externalization of private mental names that they function as devices of referential coordination among the speakers of such languages. And we have seen that it is the Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0005

Concepts, Conceptions  143 fact that names function as devices of referential coordination, both within individual referring minds and across distinct speakers of a shared language, that explains both the cognitive import of individual identity thinking and the pragmatic import of shared identity talk. Here we seek to augment our developing picture of the psychic workings of the referring mind by going beyond facts about the inner psychic workings of the mind’s representational vehicles to consider the psychic organization of its representational contents. Our focus, therefore, will be on the referring mind’s deployment of concepts and conceptions in thought episodes.

1.  Concepts and Conceptions Both the word “concept” and the word “conception” are at home in a variety of different domains. Common sense, for example, recognizes both such things as concepts and such things as conceptions, though it is not entirely clear what common sense takes each to be or how it takes them to be related. We do no violence to common sense, for example, when we refer to the Greek, the biblical, or the liberal conceptions of justice. We seem thereby to imply that although Greeks, Christians, and political liberals all conceive of justice somehow or other, they each conceive of it differently. What is not altogether obvious is how diverse conceptions of justice are supposed to relate to concept(s) of justice. Is the concept of justice one and invariant—​as talk of “the” concept of justice would seem to imply—​while conceptions of justices are clearly many? Or are there, perhaps, as many concepts of justice as there are conceptions thereof? If that were the case, however, what precisely would be the point of distinguishing concepts from conceptions? If it is not the case, what exactly is the difference between the one fixed concept of justice, presuming there is such a thing, and the possibly varying conceptions of justice? It is not altogether clear that commonsense offers any systematic guidance on such questions. Another domain in which the notion of a concept, if not the notion of a conception, is at home is in scientific psychology. Indeed, psychologists have offered up a great number of competing theories about the nature of concepts—​though the further distinction between concepts and conceptions is seldom explicitly thematized within scientific psychology. That lacuna may have something to do with the fact that psychologists take concepts to be inner mental entities that serve to move the gears of the mind in certain

144  Referring to the World characteristic ways, paradigmatically in the course of various mental processes, such as in the making of category judgments or the performance of various discrimination tasks. On this process-​oriented way of looking at matters, the concept of a bird, say, might plausibly be identified with that mental entity, whatever it turns out to be, that plays a distinctive causal role in enabling a thinker to judge of any arbitrary object whether it is a bird or in enabling her to discriminate birds from non-​birds. If this is how one understands concepts, it is perhaps not entirely surprising that one might feel little pressure to explicitly distinguish concepts from conceptions. Once one takes concepts to be inner mental entities that help steer the wheels of the mind in judgment and discrimination task, there would have to be good reasons for supposing that concepts and conceptions steer the mind in separate and tractable ways. And it is perhaps not immediately obvious that there are two separate and distinct things in the mind that contribute in clearly separate and distinct ways to steering the mind. Philosophers, too, have theorized a great deal about the nature of concepts. Indeed, some have even had occasion to draw upon a distinction between concepts and conceptions. Rawls, to take just one example, treats the distinction between concepts and conceptions as something like the distinction between genus and species. The concept of justice articulates a general and shareable idea of justice, while the many distinct conceptions of justice articulate something more specific and contestable. Indeed, approaches that distinguish concepts from conceptions in this way tend to see more “generic” concepts such as the shareable concept of justice as essentially contestable. And they distinguish concepts from conceptions so as to explain how a contestable concept admits of various more specific determinations, as we might call them. This is a perfectly coherent way of distinguishing concepts from conceptions. Indeed, it makes perfect sense if one’s aim is to explain what is fixed and what is variable between, say, the biblical conception of justice, the liberal conception of justice, and the ancient Greek conception of justice. But my explanatory ambitions are somewhat different. They are a mix of the philosophical and the psychological. To be sure, reconciling psychological ambitions and philosophical ones is not an altogether easy task. Ever since the two disciplines parted ways around the beginning of the previous century, psychology and philosophy have often marched to rather different explanatory drumbeats. I will not rehash the history of that parting here. I will just say that one unfortunate consequence of the parting was that philosophy

Concepts, Conceptions  145 became conceptually rich but empirically impoverished, while psychology became empirically rich but conceptually impoverished. I hasten to add that there are those who celebrate rather than lament the parting and see this parting of the ways as a great leap forward for philosophy from the darkness of psychologism. But I do not propose to settle that debate here. I will just say that on my own view, philosophy has much to learn from the psychological conception of a concept and psychology has equally much to learn from philosophical thinking about concepts. This is not to deny that there are certain persistent barriers to fruitful mutual engagement. Psychology’s reticence to explicitly thematize a distinction between concepts and conceptions is one among them. To see what I mean, it will help to consider what philosophers call the publicity of concepts and thought. Publicity has to do with the fact that it is apparently possible for different thinkers to grasp and deploy the very same concepts in thinking the very same thoughts. When you think the thought that dogs are canines and I think the thought that dogs are canines, there is an intuitive sense in which we may be said to think the same thought and to deploy the same concepts in the course of doing that thinking. Philosophers have rightfully tended to place quite heavy emphasis on publicity, though not always, in my view, to good effect. Starting at least with Frege, the emphasis on publicity has sometimes led to just the sort misbegotten anti-​psychologism that I was lamenting earlier. If concepts and thoughts are essentially sharable, a certain prominent strand of philosophical reasoning goes, then they cannot possibly be strictly identical to anything inside the heads of individual thinkers. The contents of my head are uniquely mine; the contents of yours are uniquely yours. We can no more share what is in our respective heads than we can share, say, body parts. Something like this reasoning led Frege to sharply distinguish the psychological act of thinking from the thought itself. While the act of thinking does indeed take place within in the minds of individual thinkers, neither the thought itself nor the concepts or senses, as he would have called them, out of which a thought is presumably composed has any owner. Neither a thought nor its conceptual building blocks are resident, either in whole or in part, in any of the minds of the many individual thinkers who may think the thought. Nor, it must be added, is a thought itself part of the physical world that we presumably share for Frege. A thought, he argued, is not a happening that begins and ends. It is not something with which we could physically interact in any way. A thought is, rather, the denizen of a shared world, a world to which we do have some sort of shared access through the rational powers

146  Referring to the World of our minds. A thought is an enduring and unchanging resident of a sort of Platonic third realm. Though we cannot take hold of such thoughts physically, we can “grasp” them with our minds. But even when we do mentally grasp them, they remain forever unchanged by our grasp. And that is part of the reasons they remain available for other rational minds to grasp as well. Though I reject the anti-​psychologism to which an emphasis on publicity has sometimes led, I concede that publicity does exert a certain degree of pressure in that direction. It is no simple matter to locate anything fixed and stable inside the heads of a diverse array of individual thinkers with which we could straightforwardly and without remainder simply identify the concept of a dog, say. We can’t expect to look inside the head, survey its contents, and find things neatly packaged and labeled as this or that concept—​as if here in this register of the brain sits the fixed and stable concept and there in that register of the brain lie, perhaps, various conceptions of dogs. Any particular thinker who is capable of thinking thoughts about dogs is liable to have lots of different beliefs about dogs, some of them shared with other thinkers, some of them quite idiosyncratic. Some thinkers may be able to make quite fine discriminations among the different breeds of dogs, while others are hardly able to tell dogs from wolves or foxes, at least not just by looking at them. Some may take the stout African Husky to be the prototypical dog, while others take the prototypical dog to be the perky Yorkshire Terrier. What you find when you start rummaging about in the head for something stable and recurrent with which to identify various concepts, in the way that publicity would require, is liable to be a buzzing-​blooming confusion of mental states that vary substantially from thinker to thinker. If the concept of dog is to be something inner, shareable, and stably identifiable there must be something invariant from thinker to thinker that suffices to make it the case that they one and all possess the concept of dog and the very same concept of dog, at that. Because it is hard, to say the least, to locate any such thing in the head, it is not entirely without reason that generations of anti-​psychologistic philosophers despaired of this possibility and insisted that concepts, whatever they are, are not, after all, things in the head. But it is worth saying, for the sake of full disclosure, that the pressure on the psychological conception of concepts does not stop at metaphysical considerations about the difficulty of nailing down a purely internal, purely psychological criterion of concept identity. Normative considerations have also played a substantial role among many philosophers in fueling a flight away from the psychological conception of a concept. Such normatively fueled

Concepts, Conceptions  147 flights have taken many forms. I cannot hope to do justice to them here.1 I have already cited the rationalist tendency to believe that thought and its conceptual building blocks must subsist in a sort of Platonic heaven as one source of the anti-​psychologistic tendency. To add a little normative fuel to this metaphysical kindling, one has to see this third realm as inferentially articulated and normatively governed. The idea, to put it all too briefly, is that thoughts will stand in various relations of “following from” to one another, in part as a consequence of the theory that they are built out of constituent concepts. A crucial further claim is that such relations of “following from” could not be mere temporal relation of precedence and succession nor even any sort of merely causal relation. Rather, following from has to be a logical and intrinsically normative relation and thus not a merely psychological one. To be sure, those within the rationalist tendency are willing to allow that our minds, at least when functioning rightly, do enjoy the power to be causally moved in normatively appropriate ways by our somewhat mysterious grasp of the denizens of this intrinsically normative third realm. But they are liable to insist that our minds and their psychic powers to lay metaphorical hold of the denizens of this third realm could not possibly be the original source of this normativity. Again, I do not mean to dismiss the normative and metaphysical considerations just bruited as entirely without force. Indeed, in a big book in progress, A Natural History of Normative Consciousness, I try my best to do justice to such considerations and many others besides. Here, I will just insist, though admittedly without the backing of a comprehensive argument, that it would be a mistake to let them drive us all the way to anti-​psychologism. Start with the fact that even the most diehard anti-​psychologistic philosopher will concede that, whatever else they are, concepts are things that can somehow or other be grasped by minds like ours. Once we shift our focus from the nature of concepts to the nature of our grasp of them, we are immediately thrown back into the domain of the psychological, or so it would seem. Whether a thinker can be said to grasp this or that concept must surely depend, to some significant degree, on the inner psychological state of his or her individual mind/​brain. In saying “to some significant degree,” I mean to allow for the possibility that a thinker’s grasp of a concept may not depend solely on what is going on in his or her own head. That, though, is a lesson we ought already to have learned from Putnam and others. They taught us that one’s grasp of 1 I cannot do full justice to this issue here, but see Taylor (in progress).

148  Referring to the World a concept may also depend, at least in part, on what is going on in the heads of those with whom one participates in what I call a “discursive community”. Putnam himself, to take one prominent example, claimed not to be able to tell the difference, at least not by inspecting them directly, between elm trees and beech trees. Nonetheless, thanks to what he called the “linguistic division of labor,” he was able to at least somewhat competently if not entirely masterfully use the terms “elm” and “beech” to refer to the right sorts of trees in his public talk. Presumably, he could deploy both the concept and the concept in his private thought as well. At a minimum, this shows that somehow or other, despite the mildly “disordered” state of his mind—​that is, despite his confusion, ignorance, and inability to perceptually discriminate elms from beeches—​both elms and beeches were available to Putnam for thinking and talking about. But this, I think, provides an important clue about how we should think about what it could be to grasp or possess a concept. For it suggests that we should consider concepts to be not just cognitive but also social tools that somehow function to make available for our thought and talk the things to which they apply. It suggests that cognizers who share the concept or the concept —​that is, who jointly participate in the use and abuse of this cognitive and social tool—​may be alike in some ways, but radically unalike in other ways. Perhaps they are alike only in that elms and beeches are available to them all for thinking and talking about. Perhaps in many other respects they are hardly alike at all. Putnam, who knows just little about elms and beeches, and the expert arborists, who may know all there is to know about such trees, are perhaps a case in point. There is also the matter of water versus twin water to consider when we think about what it is to share a concept. According to Putnam, my doppelganger on Twin Earth has pretty much the same sorts of things going on in his head as I have going on in mine, at least modulo the absence of H2O in his head and the absence of XYX in mine. He is presumably a member of a discursive community populated with persons who are quite similar to the persons with whom I participate in a discursive community. If Putnam is right, water is available to me for thinking and talking about, but it is not available to him for thinking and talking about, at least not via the same sorts of epistemic/​causal/​informational/​discursive pathways via which it is available to me. Similarly, twin water is available to him for thinking and talking about but not available to me for thinking and talking about, again, at least not via the same sorts of pathways via which it is available to him. One might

Concepts, Conceptions  149 want to say that this is because I have the concept of water and he does not, while he has the concept of twin water while I do not. But if this is right, then not only would it be the case that thinkers whose psychic innards differ substantially in certain ways might share a concept, but it could also be the case that thinkers whose psychic innards are quite similar might fail to share a concept. The upshot of all this is that we seem to face two competing sets of pressure simultaneously. On the one hand, the fact that concepts are capable of being grasped and that our grasp of them plays a role in driving such psychological processes as the making of category judgments, among other things, pulls us toward inner psychology. But the publicity of concepts tends to pull us away from the psychological conception of a concept. It is partly by way of balancing these competing pressures that I introduce the distinction between concepts and conceptions. By stipulative fiat, I reserve the notion of a “concept” for that which unites all those who are capable of thinking and talking about a given thing or kind of thing. I reserve the notion of a “conception” for that wherein thinkers who share the capacity to think about a given thing or kind of thing, in virtue of possessing a concept, may nonetheless differ. On my way of slicing matters, concepts, whatever they are, function primarily to make objects available for thinking or talking about. Making objects available for thought and talk is what concepts do. That does not yet fully determine what concepts are. But focusing, in the first instance, on what concepts do may still have consequences for what concepts could possibly be. One thing that concepts are not, I shall argue, are things that come before the mind in the manner of a guiding light, reliably and infallibly moving the gears of the mind in just the right way. But more on this point in due course. Just as with concepts, our first approach to conceptions is by way of what they do. If concepts are cognitive and social tools that make objects available to the mind for thinking and talking about, conceptions play a role in mediating our reasoning about the things our concepts make available. The rough idea here is that once concepts have done the job of making those things available for our thought and talk, our conceptions take over to help guide our reasoning about those objects. Once the concept of justice, to return to our earlier example, makes justice available for thinking and talking about, a particular conception of justice will then play a role in mediating our reasoning about justice. So, for example, if one has a conception of justice as fairness, one is liable to make different judgments about the nature and consequences of justice and injustice than if one has a conception of justice

150  Referring to the World as something tied up with fealty to the will of God or the will of the stronger. That it is justice that one’s reasoning is reasoning about rests on the fact that one deploys the concept of justice. But one’s various beliefs and judgments about the significance of justice will depend to some degree on one’s peculiar conception of justice. Though concepts and conceptions are functionally distinct on this way of slicing matters, they are also intimately related. Indeed, they should be thought of as performing interdependent functions within the psychology of the referring mind. In fact, with apologies to Kant, I will argue below that conceptions without concepts are empty and that concepts without conceptions are idle. I will say more about the relation between concepts and conceptions later in the chapter. But it is worth saying up front that my way of dividing the cognitive labor between concepts and conceptions is driven entirely by the explanatory demands of my own two-​factor referentialism. I don’t claim to be hewing all that closely to the wisdom of common sense. Nor can I claim to be backed by the authority of philosophical or even psychological orthodoxy. I make no apologies for that fact. Whether I am ultimately entitled to either the notion of a concept as I understand it or to the notion of a conception as I understand it will ultimately depend on whether I am entitled to two-​factor referentialism. There are, of course, many alternative psychological and philosophical theories of the referring mind that divide things up differently than I do here. And some of these alternative ways of dividing things up will seem to some to be closer to either what they take to be antecedent orthodoxy or common sense wisdom than anything I say here. But sometimes philosophy should disturb rather than respect orthodoxy and common sense. And this, I think, may be one of those moments. It is also worth saying that as philosophy progresses, what once seemed an impregnable orthodoxy can begin to lose its vise-​like grip on the imagination. Once upon a time, for example, there was a thriving philosophical enterprise that went by the name of conceptual analysis. The foundational orthodoxy of that enterprise was that concepts possess definitions and that much philosophical knowledge could be gained by interrogating concepts in search of their definitions. Now conceptual analysis may, at first blush, seem like a rather trivial enterprise. What could a definition contain except what we put into in the first place? And why on earth believe that the pursuit of definitions could lead to substantive as opposed to merely trivial knowledge? It must be said, however, that it was no part of the enterprise of conceptual analysis to insist that definitions of our concepts were always and already

Concepts, Conceptions  151 fully manifest and explicit to the referring mind. Indeed, a self-​described vital task for philosophy self-​consciously done in this key was precisely to make the tacit definitions of our concepts fully manifest and explicit. This we could do, the thought went, by means of clever thought experiments, designed to elicit probative intuitions about hypothetical cases. By means of this method, it was thought that the conceptual analysis could at least make the definitions of the concepts we somehow implicitly grasp manifest and explicit to the grasping mind. It would not be fair to say that the enterprise of conceptual analysis has been consigned to the dustbin of philosophical history. There are still precincts of philosophy fully devoted to that enterprise. But those precincts no longer represent a hegemonic orthodoxy. They constitute only one locality among others in a currently fragmented philosophical landscape. My point is rather that there is by now a plethora of both psychological and philosophical grounds for at least doubting that concepts come neatly associated with fixed and invariant definitions of the conceptual analyst’s imagining, tacitly known by anyone who fully grasps the relevant concepts. Largely as a consequence of the shattering of the old orthodoxy and the absence of any single successor theory to take its place, many competing theories of concepts and conceptual competence are now afoot. Our grasp of concepts has been variously appealed to in order to explain their roles in guiding our categorizations, by reference to judged distances from prototypical instances, by stereotypes characterized by a list of defining properties—​though without the presumption that the defining properties are necessary and sufficient for category membership—​or by tacitly known theories that implicitly define the relevant concepts. Though I  propose to take a closer look at the psychology of the referring mind and to deploy the distinction between concepts and conceptions as my entering wedge, my aim is not to decisively settle the raging disputes among different theories of concepts and concept possession. Many of the alternative approaches are still live competitors and nothing that I say here is likely to end the competition. Within my own framework, much of the explanatory work that philosophers and psychologists have traditionally asked the singular notion of a concept to do is distributed between concepts and conceptions. Consequently, there is no exact mapping between theories with a singular focus on concepts and my own. Indeed, some of the long-​ standing disputes that have traditionally been framed as disputes about the nature of concepts and/​or conceptual competence are recast here as disputes

152  Referring to the World about conceptions rather than concepts. Perhaps I have made one too many distinctions. Or perhaps traditional theorists have made one too few. I will not try to settle that question here. The choice between my approach and more traditional approaches likely cannot be made on the basis of point-​by-​ point comparisons, in any case. It is more likely to come down to a choice between overall frameworks. Partly as a consequence, I won’t be wading very deeply, if at all, into certain long simmering philosophical or psychological disputes about the metaphysical nature of concepts or the psychological nature of conceptual competence. I will just try to lay out my own approach in enough detail to give the reader a sense of its explanatory virtues, which I hope are many, and its explanatory vices, which I hope are few. I said earlier that it is by having concepts that things in the world become available for talking and thinking about. That is part of what is behind the thought that concepts are the sorts of things that have extensions, a set of things that “fall under” them.2 All and only red things fall under the concept ; all and only horses fall under the concept . When one thinks about red things it is not because red things come directly before one’s mind in something like a Russellian episode of epistemically transparent “immediate acquaintance.” One thinks about red things by deploying the concept in a thought episode. One thinks the thought that John is a philosopher, not by having either John himself or the set of all philosophers come directly before the mind, but by deploying the concept and the (individual) concept in construction with one another in a thought episode. We may ask two related questions. First, there is the determination question. How is it that that a concept comes to have a determinate extension? There is also the availability question. In virtue of what is the extension of a concept made available to thought? One might argue that determination and availability are to some degree independent matters. Conceivably, there 2 Assuming that reality has a modal structure, concepts will apply not merely to actual things but also to possible things, that is, to ways the world, or at least parts of the world, might have been. To capture the modal significance of concepts, it is useful to think of a concept as associated with functions from possible worlds to extensions. I say “associated with” and not “identified with” partly out of an abundance of caution. There is perhaps much about concepts that the idiom of possible worlds fails to capture. But functions from possible worlds to extensions are not in the same way structured entities. It should also be said that perhaps not all concepts are extension-​determining concepts. There are also what might be called “intensional concepts”—​like, for example, . Such concepts don’t directly determine any extension on their own. There is not a set of things that are merely alleged. Concepts like “alleged” rather combine with other concepts to yield new concepts. An alleged murdererer, for example, may or may not be an actual murderer, but in being alleged to be a murderer, a person ipso facto falls under the concept .

Concepts, Conceptions  153 could be a story about determination that failed to directly illuminate availability. Suppose, for example, that facts about determination were entirely independent of the grasping of concepts. And suppose it were only in the grasping of a concept that the independently determined extensions of our concepts were made available to thought. Something rather like this picture is plausibly implicit in the sort of anti-​psychologistic Platonism about concepts considered earlier on. In seeing concepts as intrinsically alien to the realm of mere psychology, the Platonist would seem to construe determination and grasp as two entirely separate and independent matters. The mere determination cannot depend in any way on our capacity to grasp them and the bare facts about determination cannot explain availability. There must be a separate theory of grasping. For all I have to say here, some such picture of the dual life of concepts may well be coherent. But I will be developing and defending an altogether different picture. That picture represents availability and determination as two sides of a single coin. Modulo the anti-​psychologistic pull of publicity, which already forces on us a distinction between concepts and conceptions, concepts (and conceptions) are best viewed, not as things initially alien to the human mind that are brought into the mental realm only by being grasped by us; they are best viewed as cognitive and social instruments, always and already at home in the psychology of the referring mind. Once concepts (and conceptions) are viewed in this way, it makes no real sense to think of the extensions of our concept as somehow determined independently of the deployment of those very concepts in thought episodes. It is rather by being deployed as cognitive instruments in the context of both our individual and shared cognitive lives that concepts both come to have the extensions that they have and come to make their extensions available to thought. This way of looking at concepts was already partially surfaced in my broadly Kantian sketch, gestured at briefly in Chapter 2, of how objective representational content is determined. It is perhaps worth reiterating aspects of that rough sketch here. On my view, the mystery of objective representational content is rooted in the fact that our outwardly oriented cognition of a world that we come to see as richly populated with all manner of things—​ from quarks and gluons to peoples and nations—​initially rests on a relentless inward rush of mere energy upon the portals of sensation. It is important to stress that the inward rush is not in the first instance a rush of reasons, evidence, or even semantic information. And because it is not, objective representational content is not a thing merely impressed upon the mind and

154  Referring to the World its representations from without. Yet somehow or other this energetic inward rush does eventually give rise to outwardly oriented representational states. It is these representational states that somehow manage to represent the world, from which the inward rush apparently emanates, as populated with much more than mere energy. And we are disposed to think of much of this vast outcropping of outwardly oriented cognitive representations as more than mere projections, fictions, or inventions. I do not mean to suggest, however, that none of our apparently outwardly oriented representations are fictions, projections, or inventions. Indeed, I shall argue in subsequent chapters that many of our apparently outward representations are in fact precisely that. And though that may sound, at first blush, like a bad thing, it is not meant to be; many of these fictive, invented, or projected representations play quite important roles in our lives. Many of these representations have the form and function of merely objectual representations, but they are not, in the end fully objective representations by which we make reference to objects in a mind-​independent world. Other of these representations serve to usher into being what I call “secondary realities.” Secondary realities are not part of what is already there, independent of the inflationary ontological powers of the human mind. But they are nonetheless real in the sense that a complete inventory of what there is that left such entities off its ledger would have missed something significant. It is just that the standing of secondary entities as real depends entirely on us. The point is that our apparently outwardly facing representations come in many different varieties and play many different reference-​like roles. Some refer to real and mind-​independent entities. Some refer to secondary entities largely of our own constituting. Some may purport to refer but in fact refer to nothing at all. Representations of all these varieties have significant roles in both our cognitive and social lives. But the vast variety of representations thrown up by the mind at least in part in response to the inward rush is not our topic at present. The question before us here is how it is that at least some of our representations—​in particular, the ones by means of which we cognize and refer to the denizens of a mind-​independent world—​manage to become semantically answerable to a world whose initial assault on the mind takes the form of an inward rush of mere energy. Though I have certain hunches, I don’t profess to actually know the complete answer to this question. Answering that question fully would require an endeavor of an entirely different character than the endeavor undertaken in this book, one that could not be

Concepts, Conceptions  155 carried out to completion from the relative comfort of the philosophical armchair. At a minimum, it would require us to chart the various pathways by which mere energy first begets sensation and then percepts. We would have to understand how percepts manage to be brought under concepts and how concepts are deployed within judgments, and how judgments are integrated into comprehensive theories of the vast and layered labyrinth of existence in its totality. And we would need to understand what principles govern the formation, reformation, and deformation of such theories over time, in response ultimately to nothing but the inward rush from which all such cognition ultimately takes its beginning. The point is not that there is nothing whatsoever that is known or settled about the various steps in that journey. But little enough is decisively settled that there is still no room for the sort of speculation about how possibly things might go that is the stock and trade of philosophy. I seek only to make a modest philosophical contribution to that ultimate project by laying out a view about what concepts (and conceptions) might possibly be such that they are fit to play a role in enabling that journey. My story has internalist, externalist, and what might be called “communitarian” elements. It sees conceptual content as the joint product of mind and world. There would be no reference, and perhaps no mental contents at all, at least no singular mental content, in the absence of the inward rush. There would, to be sure, still be the form of thought, including the form of singular thought. But the bare rush of energy inward is surely not enough to make any sort of objective representational content possible. Only in a mind antecedently prepared to take the inward rush and make objective representational content out of it does the inward rush manage to give rise to states with objective representational content. I have tried to say a bit about what the antecedence readiness to make objective representational content, especially singular referential content, in response to the inward rush might plausibly consist in. But it is not just the innards of the individual mind, in the context of its own individual encounter with the inward rush upon its individual sensorium, that plays a role in turning the inward rush into objective representational content. Our psychic capacity for discursive community also plays a crucial role. The capacity for discursive community enables objective representational contents to spread from mind to mind, relatively independently of each mind’s own direct exposure to the inward rush of energy. The resulting picture is partly externalist, precisely because of the indispensable role of the inward rush in making objective representational content possible. It is partly internalist,

156  Referring to the World because of the indispensable role of the innards of the referring mind in making such content possible. And it is partly communitarian, because of the distinctive role of discursive community in spreading such contents. Turn back to publicity again, but this time with our partly internalist, partly externalist, partly communitarian picture more firmly in mind. Suppose we ask what concepts must be if they are to be the common possessions of the psychologically diverse members of a single discursive community, who participate in the determination and propagation of content in many different ways. My central claim is that there is unlikely to be any single thing that constitutes the possession of a concept within such a community. Rather, those who are alike in possessing the same concept will differ from each other in various and important ways. No doubt, there will also be ways in which they resemble one another. Hence a central challenge for any good theory of concepts is to say how those who share a concept must resemble each other and how they may differ from one another. It is largely in service of precisely this task that I introduce the distinction between concepts and conceptions. On my view, concepts function as both building blocks of thought and things that serve to make objects thinkable at all. I take this aspect of my approach to be consonant with many more traditional approaches to concepts. Conceptions, by contrast, occupy two complementary roles in the psychic organization of the referring mind that are sometimes mistakenly attributed to concepts. Rather than being building blocks of thought, conceptions are, in a sense, built out of thoughts. And while concepts serve to make their extensions available to thought, conceptions largely inherit their extensions from the concepts with which they are associated. Think of conceptions as databases of information (and misinformation) about the extensions of associated concepts. Suppose that a thinker has a variety of thoughts involving the concept . For the moment, and just for the sake of argument, assume that all such thought episodes are mediated by type-​identical mental symbols. That is, each time our thinker tokens a -​involving thought, she tokens a mental symbol shaped *CAT*. We may define a new entity—​the thinker’s cat conception—​determined in part by a (distinguished) subset of her -​involving thoughts. We symbolize conceptions as follows: “Cat” [ . . . ]. Think of a conception as a mental particular, a labeled, perhaps highly structured database of information about the extension of an associated concept. A “Cat”[ . . . ] conception may contain a variety of different kinds of information (or pointers thereto) about cats. It may contain a list of properties that some, many, most, all or typical cats are

Concepts, Conceptions  157 taken to satisfy. It may contain information (or pointers thereto) about the categorial basis of the concept —​whether is a natural kind concept, a functional concept, an artifactual concept. It may contain images of exemplary cats, a list of atypical cats, or pointers to sources where more can be found out about cats. A comprehensive theory of our actual cognitive functioning will have to say a great deal about the structure, content, organization, and evolution of our conceptions. I will not try to offer such a comprehensive theory here. The crucial point for our current argumentative purposes is the claim that conceptions play a distinctive role in mediating the deployment of concepts, at least as I understand concepts, in thought episodes. When I come to recognize that this furry animal here is none other than Nikola the dog, two distinct concepts are deployed in my thought episode and two distinct conceptions mediate those deployments. The truth-​conditional content of my thought episode is determined by the determinate construction in which the concept and the concept are deployed in the course of that thought episode. But the deployment of these concepts in a thought episode is itself mediated by both a standing “Nikola”[ . . . ] conception and by a temporary “furry animal here now”[ . . . ] conception, generated on the fly as a consequence of my current encounter with a certain furry animal. I have a standing “Nikola”[ . . . ] conception that lists the property of being a furry animal as one among Nikola’s properties. Partly because of that fact, as I become increasingly aware that the furry animal currently in front of me bears a striking resemblance to Nikola, I conclude that this furry animal here is none other than Nikola. Thereafter, I may close the temporary “furry animal here now”[ . . . ] file and feed all the information that it contains into my standing “Nikola”[ . . . ] file—​thereby adding one more entry to my stored database of information about the life history of Nikola. Now some thinkers, including some fellow referentialists, appear to want to identify the sort of mental particulars that I brand conceptions as concepts instead, at least so long as they contain information of just the right sort.3 3 Crimmins (1992), for example, claims that to possess a concept is to possess a normal “notion.” To be sure, it is not entirely clear how the notion of a notion, as deployed by Crimmins and by Crimmins and Perry (1989) maps onto my distinction between conceptions and concepts. Both Crimmins and Perry seem to think that (a) notions are constituents of beliefs and (b) notions are individuated at least partly in virtue of the beliefs in which they serve as constituents. It is the combination of (a) and (b) that leads to trouble, I think. For it leads to the thought that to have a concept is to have a notion that figures in the right set of beliefs. However, this gets matters the wrong way around. Concepts are not, I think, rightly individuated by adverting to the beliefs in which they figure. Rather, beliefs are individuated partly by adverting to the concepts that figure in them. On the other hand, conceptions

158  Referring to the World Other thinkers don’t go quite that far, but they do claim that, as a matter of metaphysical necessity, one can't have the concept unless one has a conception-​like structure of the right sort in one’s head.4 Such thinkers hold, in effect, that concepts supervene, give or take a bit, on beliefs.5 Now I readily grant that it is a psychological fact that in cognizers like us, deployments of concepts in thought episodes are typically mediated by conceptions. Indeed, I insist on that fact. For all I know, it may perhaps be a law of universal cognitive science, if there are such things, that concepts are deployable in thought only to the extent that conceptions mediate such deployments. If that were so, then the widespread tendency to conflate concepts with conceptions would not be entirely surprising. After all, if it is only by being associated with some conception or other that a concept comes to be deployed in some determinate way in thought, and if conditions of concept possession really are logically prior to conditions of concept identity, then it is natural to think that a concept is just what one has when one has a conception of a certain character. But I think there are at least two principled reasons to resist the conflation of concepts with conceptions: (1) conceptions don’t relate to their extensions in concept-​like ways (2) conceptions don’t “compose” in concept-​like ways Like a concept, a conception may have an extension. For concepts, it is trivially true that nothing belongs to the extension of a concept except instances of that concept. All and only cats belong to the extension of the concept . That cats are instances of would seem, in fact, to follow from the very nature of concepts. Indeed, I take it to be a central axiom of the general

are indeed individuated partly by adverting to collections of beliefs. Just what sorts of collections is part of the burden of this chapter to spell out. Recanati (1993) is another referentialist who uses the “file” metaphor without clearly distinguishing concepts from conceptions. 4 A paradigmatic example of such an approach are holistic theories according to which having a concept is a matter of having something that occupies an appropriate location in a system of beliefs. Such a holist need not be understood as identifying concepts with systems of beliefs. 5 The “give or take a bit” has to do with the role of knowledge-​wh in determining conceptual competence. Those who tie concept possession to knowledge and belief need not claim that one who possesses a concept has propositional knowledge of the necessary and sufficient conditions for the application of the relevant concept. It is typically enough for their purposes to maintain that one who possesses a concept will thereby have certain practical inferential dispositions and/​or recognitional capacities.

Concepts, Conceptions  159 theory of concepts that to a first approximation every instance of the following scheme expresses a trivial and transparent truth:6

applies to all and only ps.

But with conceptions, matters are different. A “cat” [ . . . ] conception may fail to be true of any actual cat and may be true of many non-​cats. This follows from the fact that conceptions are organized, localized databases of information and misinformation that are, or at least supervene upon, structures of belief. Though the concept is true of all and only cats, nonetheless, in the context of a false belief, the concept may be misapplied by a cognizer to non-​cats. When one falsely believes Nikola the dog to be a cat, this false belief infects both one’s “cat” [ . . . ] conception and one’s “Nikola” [ . . . ] conception. One comes to believe of the class of cats that Nikola is a member of it, and of Nikola that he is a cat. As a consequence, one comes to have both a “cat”[ . . . ] conception and a “Nikola” [ . . . ] conception such that each applies or, better, is applied to, something of which it is not true and fails to apply to something of which it is true. Now consider an arbitrary speaker’s “gray cat” [ . . . ] conception. What constraints are there on the relations among her “gray cat” [ . . . ] conception, her “cat” [ . . . ] conception, and her “gray”[ . . . ] conception? Evidently, there are no ways these conceptions must be related. If each of these conceptions includes a list of exceptions, there is no way in particular that the list of exceptions in “gray cats” [ . . . ] must be related to the list of exceptions in “gray” [ . . . ] and the list of exceptions in “cat” [ . . . ]. An exceptional gray cat may fail to be either an exceptional cat or an exceptional gray thing. Much the same is true for any aspect of the relation between a “complex” conception and its simpler “constituents” one cares to name. Prototypical gray cats need be neither prototypically gray nor prototypically catlike. Cat x may be an exemplar of a gray cat while failing to be either an exemplary cat or an exemplary gray thing. One may have a “false” (true) conception of gray cats, while having a “true” (false) conception of gray things and a “true” (false) conception of cats. Facts about “complex” conceptions, if there are such things, would seem to be almost entirely unconstrained by facts about the simpler conceptions out 6 Because of intensional concepts, this scheme doesn’t tell the entire story about the relationship between concepts and their extensions. Consider, for example, the concept . In general, toy guns are not a kind of gun, since something can be a toy gun without being a gun. Indeed, if x is a gun, then x precisely isn’t a toy gun.

160  Referring to the World of which they are arguably “composed.” Again, this is a reflection of the fact that conceptions are determined by structures of belief together with the fact that what one believes about gray cats is no straightforward function of what one believes about cats and what one believes about gray things. Since beliefs about gray cats are not built out of beliefs about gray things and beliefs about cats, the failure of conceptions to “compose” in any straightforward way is precisely to be expected on the assumption that conceptions are or at least are somehow grounded in structures of beliefs and are not themselves the constituents out of which beliefs are composed. With concepts, matters are more systematic. Consider the inference from x is a gray cat to x is gray and x is a cat. Why is this inference compelling? A powerful answer was made famous by Jerry Fodor (1975, 1987, 1992, 1998). He argued that the language of thought has a combinatorial syntax and a compositional semantics. On his view, the mental representation that expresses the concept is literally built out of the mental representation that expresses the concept and the mental representation that expresses the concept . Moreover, the possible worlds theoretic intension by which we might represent the concept is the set-​theoretic intersection of the possible worlds theoretic intension by which we might represent the concept and the possible worlds theoretic intension by which we represent the concept . That is because there is a general and systematic story about the way “simpler” concepts compose to yield more “complex” concepts and a systematic story about the way the intensions of more complex concepts are determined by the intensions of their constituents. There is no such general and systematic story for conceptions. The relations among concepts, conceptions, and their extensions are complex. I have already suggested that it is possible for agents who share the same concept to have quite different conceptions of the extensions determined by their shared concepts. We might analogize concepts and conceptions to the planets of a single solar system, one and all orbiting the same star, but at different mean distances from that star and at varying angles to it. Diverse conceptions associated with a single concept are like diverse planets that orbit the same star. Consider, for example, coffee. A somewhat chemically informed coffee drinker may have a coffee conception that includes the fact that coffee is part aqueous solution and part emulsion. But such facts are unlikely to form any part of the coffee conception of the chemically uninformed lover of coffee. To be sure, the chemically uninformed coffee drinker may have a vague sense that coffee is not rightly conceived of as a kind of water,

Concepts, Conceptions  161 in the way that, say, seawater is. But she may be hard-​pressed to explain why seawater is but coffee is not a kind of water, given that both are composed predominantly of water molecules. The crucial further claim is that although they have such differences in their coffee conceptions, it seems incorrect to conclude that they must have different concepts of coffee. Indeed, they might even agree in all judgments about whether various samples count as coffee. And that would seem to show that they share the concept of coffee, at least if we think that concepts are best understood as things that somehow determine a range of extensions across a range of possible worlds. But if this is right, sharing a concept while having different conceptions is something like orbiting the same star, but at different mean distances from that star. Following Fodor (2007), it may help to distinguish between having a concept and having a concept in a certain way. Consider, for example, the difference between your average at least somewhat conceptually literate person of the 21st century and, say, Homer. No doubt Homer had the concept . But if he did have that concept, he clearly had it in a rather different way from the way in which people living today have that concept. For example, Homer was in no position to even entertain the thought that the essence of water is constituted by its internal chemical structure. Confronted with the precise letter of Putnam’s twin earth thought experiment, he would have been completely baffled. He would have failed to endorse what we take to be true counterfactuals as true counterfactuals. He would have been baffled by the thought that just those counterfactuals reveal the true metaphysical nature of water. What exactly would the fact of Homer’s imagined bafflement show? Some will no doubt be tempted to conclude that Homer has an altogether different concept of water from us or perhaps that he lacks the concept of water altogether. But this is a mistake. Or so I claim. On my view, it is not that Homer lacks a concept of water. It is not even the case that he lacks our concept of water. He clearly does have a water concept—​plausibly the very same concept that we have. Suppose there is some robust sense in which we and Homer do share a concept. It is still striking that although that concept is, even for him, a concept of what is in fact a natural kind, he doesn’t really have the concept of water as a natural kind concept, at least not as we currently understand natural kinds. But this suggests that perhaps the difference between Homer and us with respect the concept is not a matter of what concept we have versus what concepts he has, at least not entirely. What concept one has doesn’t, on its own, determine which counterfactuals one either explicitly or tacitly take

162  Referring to the World to be both true and revealing of metaphysical nature and thus of which possibilities one takes to be genuine. How one has a given concept seems also to matter. What we should say about Homer is that he didn’t have the concept water in the right way. He didn’t have the concept water in a way that would have allowed him to explicitly recognize that water is a natural kind as we understand natural kinds. The thought here is that having the concept in one way can give one cognitive access to the true counterfactuals and genuine possibilities about water. Having the concept in some other way may fail to do so. Putting matters this way immediately raises the question of just what it takes not merely to have a concept of a natural kind like water but to have the concept as a natural kind concept. This is one sort of question that a good theory of both concepts and conceptions should be tasked to answer. Fodor (2007) suggests that having a concept as a natural kind concept is a matter of having the right sort of theory playing the right sort of role in “sustaining” that concept. Fodor’s thought seems to be that having a concept as a natural kind concept is a matter of having a theory of a certain kind play a decisive causal role in mediating the deployment of that concept in thought episodes. The trick here is not to think of the concept as one that is implicitly defined by a background theory, in, say, a Ramsey-​like way. Any background theory—​whether ours or Homer’s—​is, rather, to be regarded as something like a set of additional links in the causal-​informational chain linking inner tokenings of that concept to outer occurrences of water. Viewed in this way, a theory does not so much implicitly define my concept as explicitly guide my deployment of the concept in thought episodes. So perhaps what the progress of science from Homer’s time down to our own did to our concept was not necessarily to change the concept that we all had—​ though it might have done that too—​but change the way we all had that concept by changing the background theory that guided our deployment of it in thought episodes. This approach makes theories somewhat akin to scientific instruments like microscopes or spectrometers. Such instruments function to alter and enhance our perceptual and cognitive links to the objects. Theories also alter and enhance our cognitive links to the objects by hewing new causal-​informational pathways between the cognitive centers of the mind and the objects of our cognition. There is nothing unusual in this. What the progress of inquiry did for our concept of water, it likely did for many other concepts as well. Take the concept of solidity. One might be tempted to say that the advance of science

Concepts, Conceptions  163 altered the very concept of solidity so that we no longer have the solidity concept once had by our pre-​chemical forebears. Alternatively, one might insist that the concept of solidity has been more or less preserved but that it has come to be embedded in a new causal-​informational network. On this way of looking at the evolution of the concept, it went from being a concept linked to solid objects by causal-​informational pathways constructed around a network of perceptually grounded notions and capacities, to a concept linked to solid objects via enriched causal-​informational pathways that now involve ideas of chemical states and phases of matter. Only if you think that concepts embedded in different surrounding frameworks are ipso facto different concepts should you be tempted to assert the former and deny the latter. But you shouldn’t think that. What you should think instead is that deployments of the same concept again in different thought episodes can be mediated by different conceptions. And you should say that differences of mediating conception can be highly consequential for the, as it were, “cognitive significance” of such deployments. I grant that intuitions, for what they are worth, about how exactly to slice the concept-​conception pie, may differ from philosopher to philosopher. Still at least certain aspects of my approach ought not to be terribly controversial. I suspect, for example, that many are likely to concede that diverse agents may share a concept while, nonetheless, having diverse conceptions of the extension of that concept. And it should not be controversial that thinkers with different conceptions of the same extension are likely to reason in different ways. If you take water to be a natural kind, for example, you will reason about water differently from the way you would reason about it if you took it to be a functional kind or a simple essence. If you take water to be a natural kind, for example, you are liable to think that water has an intrinsic inner essence, perhaps yet to be discovered, and that nothing counts as water unless it shares in that essence, no matter its superficial appearance. You are likely to appeal to this intrinsic inner essence, even if it is not fully known, to help explain the behavior and appearances of water. On the other hand, if you think of water as a functional kind, you will take water to be constituted as water not by its intrinsic inner essence but by some role or roles that it occupies. And you are likely to reason about water quite differently from one who takes water to be constituted by an intrinsic inner essence. You may be disposed, for example, to conclude that substances with quite different intrinsic inner structures may both count as water, as long as their intrinsic differences do not prevent them from playing the same extrinsic functional

164  Referring to the World role. Finally, if like Aristotle, you think of water as what I am calling a simple essence—​that is, as something that is not itself built of anything else but is the sort of thing that in combination with other simple essences can be one building block among others of other things—​you are liable to reason about water differently still. More controversial is the question whether there is a robust and principled account of “having the same concept” according to which cognizers who think of water in the diversity of ways just alluded to can be said to have the same concept. No doubt, some will want to deny that there is. Some will say that the bare fact that if I take water to be a natural kind, while you take it to be a functional kind, and someone else takes it to be a simple essence, it follows more or less directly that we have different concepts of water and not just different conceptions. I don’t pretend to have a knockdown argument against this way of thinking about concepts. Given that anyone is free to stipulate what they mean by the term “concept,” perhaps there could not be one. Stipulation aside, however, from the perspective of my two-​ factor referentialism, there is a deep point to understanding concepts and conceptions in the way that I understand them here. So, let me say a bit more about what concepts and conceptions must be if stability of concepts is to be possible across a diversity of conceptions, where conceptions are understood as ways of having concepts. The first point is that this approach requires us to individuate concepts coarsely. Think of expressing a concept as something that a mental representations like, say, *water* does. Then think of expressing a concept as a matter of that mental representation being somehow associated with a function from worlds to extensions. To a first approximation, as long as two mental representations are associated with the same function from worlds to extensions, they will count as expressing the same concept. The crucial further claim is that representations can be associated with a given function from worlds to extensions in a variety of different ways, via a variety of different causal and informational pathways. Moreover, these differences may show up partly as differences in how we reason with the relevant concept but without necessarily affecting the content of the concept. I was gesturing at just such picture when I appealed to the role of discursive community in “propagating” conceptual contents from one cognizer to another in a given community. As a result of the power of discursive community to propagate content, one cognizer may have a given concept directly on the basis of her own encounter with the inward rush, while another may have the very same

Concepts, Conceptions  165 concept indirectly by being appropriately deferential to others within the community. Now it may well be right to say that those who have a concept indirectly are in some ways less “masterful” or “autonomous” when it comes to reasoning with the concept. But not being fully masterful or autonomous with the concept is not, on my view, a way of failing to possess the concept at all. More generally, we might think of the property of having a concept as something like a second-​order property with many first-​order witnesses. Suppose, that is, that there are a variety of first-​order witness properties P1 . . . Pn . . . the instantiation of any of which by a representation R would suffice to constitute R’s expressing concept C. Identify the property of expressing the concept C with the second-​order property of having some first-​order witness or other in a certain set of witness properties. If expressing a concept is a second-​order property with many first-​order witnesses, we can already see that it is at least logically possible that there can be stability of concepts across a diversity of conceptions or ways of having that concept. So far, I have focused on the ways in which different cognizers who share the same concept may nonetheless differ from one another. But it is important to see that something rather like what can happen in two distinct minds can also happen in a single mind. Even within the mind of a single cognizer it is possible for deployment of the same concept to be independently mediated by distinct conceptions. This will happen whenever a single mind manages to have two syntactically distinct, referentially independent inner representations H and H* such that both tokenings of H and tokenings of H* constitute deployments of C, but in virtue of different first-​order witnesses. As a consequence of the different first-​order witnesses, it may happen that the conception that mediates H-​tokening-​constituted deployments of C differs from the conception that mediates H*-​tokening-​constituted deployments of C. But even if the witness to H’s expressing of C does not differ radically from the witness to H*’s expressing of C, nothing guarantees that the conception that mediates H-​tokening-​constituted deployments of C will be dynamically linked to the conception that mediates H*-​tokening-​constituted deployments of C.  Conceptions evolve continuously on the basis of our encounters with the world. Some conceptions may systematically co-​evolve with one another. However, two conceptions that independently mediate the deployment of the same concept will often fail to co-​evolve. Consider a cognizer with two independent mediating conceptions—​one associated with the designator “Hesperus,” the other associated with the

166  Referring to the World referentially independent but coincidentally coreferential designator “Phosphorus.” Her “Hesperus”[ . . . ] conception may often be updated as a consequence of the outcome of “Hesperus”[ . . . ] mediated deployments of the individual concept without her “Phosphorus”[ . . . ] conception being updated as well. Consequently, her “Hesperus”[ . . . ] conception and her “Phosphorus”[ . . . ] conception may come to contain quite different information. That fact that a single mind can contain two conceptions that independently mediate deployment of the same concept has far-​reaching significance both for our understanding of the many coreference puzzles and for our understanding of what we are and are not doing when we ascribe propositional attitudes to one another. I already began an argument to this effect back in Chapter 4. There I argued that largely because of syntactic factors, both identity thought and identity talk have far-​reaching cognitive and pragmatic significance. I argued that within a discourse, identity talk can function to turn what might otherwise have the character of disjointed, subject shifting discourse into discourse with a more unified, because explicitly subject matter preserving, character. And identity thinking, I claimed, achieves at the level of thought something similar what identity talk achieves at the level of talk—​ and this mostly through mere syntactic reorganization. In particular, via the joint deployment of referentially independent mental representations in episodes of identity thinking, representations that do not enjoy intrinsic purport of coreference with each other come to enjoy an extrinsic purport of coreference, at least in the context of the mind that so deploys them. We are now in a position to appreciate the role of the psychic organization of our concepts and conceptions in further amplifying the significance of identity thinking. Suppose that a cognizer learns that Hesperus and Phosphorous are one and the same object. Her concepts and conceptions are likely to be reorganized as a consequence. At a very minimum, her previously dynamically independent conceptions would come to be so linked that, ceteris paribus, they now tend to co-​evolve. It might even happen that they are merged into a single new conception, a conception that preserves some elements of the two previously unlinked conceptions and purges others. On my view, such merging and purging of conceptions should be seen as a largely syntactic operation in the sense that it does not make new thought contents happen. Rather, it makes old concepts to be deployed in new ways. I argued for a similar point in Chapter 4. There I claimed that the ultimate source of Frege’s puzzle lies in facts about the logical syntax of thought rather

Concepts, Conceptions  167 than in facts about thought content. Once we have augmented our account of the psychology of the referring mind to include concepts and conceptions as well as logical syntax, we have a further basis for doubting the Fregean understanding of the significance and nature of the coreference puzzles. Frege was led by the coreference puzzles to slice thought contents rather thinly, to insist that the thought content that Hesperus is rising must be different from the thought content that Phosphorus is rising. But once we see that identity thinking can lead a cognizer to reorganize her mediating conceptions and thereby to alter the way in which she deploys various concepts, we have no need to slice thought contents nearly as thinly as Frege does. Even so, we can grant that such a reorganization of the thinker’s concepts and conceptions may well give rise to new patterns of verbal and non-​verbal speech, including new abilities to identify and re-​identify the objects and states of affairs about which she thinks. So even on our own view, even though coming to know, as one is wont to say, that Hesperus is Phosphorus is not yet coming to know a new proposition, previously unknown, it is still a significant and consequential psychological change. Indeed, it is the kind of change that any rational cognizers really ought to strive to bring about. But it is not what Fregeans and others have long imagined it to be—​a change in the contents of her thoughts.

2.  Rationality and the Psychology of the Referring Mind This picture of the psychology of the referring mind has certain far-​reaching consequences for our understanding of precisely what the rationality of such a mind might be said to consist in. Though I cannot afford to spell out a complete alternative approach to reason and rationality here, it is worth taking brief notice of how our understanding of what rational cognition might plausibly consist in must change if my picture of the psychology of the referring mind is on the right track. A central point to take note of is the fact that when a cognizer has two unlinked conceptions that independently mediate deployments of the same concept in thought episodes, her mental life may often be metaphysically or externally incoherent. One's mental life is externally coherent if and only if there is a metaphysically possible world in which one's beliefs are jointly true and one's desires jointly satisfied. But consider the mental life of Jocasta as she is about to marry Oedipus. She has two distinct conceptions of the self-​same person—​her soon to be husband and her long ago abandoned son. One conception is associated with the designator

168  Referring to the World “Oedipus,” while the other is associated, let us suppose, with the referentially independent but coincidentally coreferential designator “Tad.” “Oedipus” [ . . . ] mediates such deployments of the individual concept as are constituted by tokenings of some inner symbol O. “Tad” [ . . . ] mediates such deployments of as are constituted by tokenings of some referentially independent inner symbol S. Jocasta believes, via tokenings of S, the proposition . She also believes, via tokenings of O, the proposition . Moreover, she desires true also via tokenings of O . She simultaneously desires true but via tokenings of S . Too bad for Jocasta. There are no metaphysically possible worlds in which Jocasta’s beliefs are simultaneously true and no metaphysically possible worlds in which her desires are simultaneously satisfied. It is important to acknowledge that for all Jocasta’s external incoherence, as I am calling it, she may seem to have what we might call an internally coherent mental life. Her beliefs internally cohere with her evidence. She plans, in an internally coherent way, a course of action designed to bring about the satisfaction of her two most ardent desires. Relations of internal coherence, it might be argued, must be defined over contents more finely grained than those to which we have so far adverted. Again, it was something rather like this thought that drove Frege, and no doubt still drives many others, to think we must individuate thought contents finely enough that the thought content that Hesperus is rising is distinguishable from the thought content that Phosphorous is rising. Again, this raises deep and substantive issues about the nature of both our rational powers and the principle on the basis of which we should decide the proper grain of the rational contents of our mental states, to which I cannot hope to do justice here.7 I will just say that the intended notion of internal coherence can really be made good in a way that makes such coherence plausibly normative for us. Nor is it obvious to me that relations of internal coherence must be defined over fine-​grained representational contents rather than just over fine-​grained representational vehicles. Indeed, I have been more or less directly, if sometimes only tacitly, adverting to just such an approach to internal coherence through much of this book. But even if there is also a notion of internal coherence, defined over fine-​ grained representational contents and not just over fine-​grained representational vehicles, and even if internal coherence so understood is normative 7 But see Taylor 2003, Taylor (forthcoming).

Concepts, Conceptions  169 for us, the case of Jocasta shows, at a minimum, just how anemic a cognitive virtue internal coherence so understood would actually be. Despite her presumed internal coherence, whatever exactly that consists in, Jocasta has a highly frustrated mental life. It is not just that Jocasta is unlikely to succeed in getting what she wants. Nor is it that if she succeeds, her success will be at most accidental. The external incoherence of Jocasta's mental life guarantees her worldly failure. There is no internally coherent plan of action that can lead, in any metaphysically possible world, to the joint satisfaction of her desires. Any metaphysically possible world in which she brings about what she most wants—​marriage to Oedipus—​is ipso facto a world in which she brings about what she most wants to avoid—​marriage to Oedipus. Consider also the case of Lex Luthor. Like Jocasta, Luthor has two conceptions of the self-​same extension. The conception labeled “Clark Kent” is associated in thought with the mental representation *Clark Kent*. The other conception is labeled “Superman,” and is associated in thought with the referentially independent but coincidentally coreferential mental representation *Superman*. Now on our way of slicing thought contents, we will insist that both these representations express the individual concept   =  and we will say that the two distinct conceptions independently mediate Luthor’s deployment of this individual concept in thought episodes. Now Luthor's most ardent desire is to put an end to Superman. To that end, he engages in internally coherent planning by which he devises an elaborate scheme for bringing about Superman's demise. At this very moment, Superman is standing right there, presenting Luthor with a golden opportunity. Yet Luthor lets the opportunity slip by. Unlike Jocasta, Luthor is not guaranteed to fail. He has it within his power, at this very moment, to flip the switch that releases the trap that does in his nemesis once and for all. But Superman is present to Luthor in the wrong way. He is present in a way that “activates” the file in Luthor's head labeled “Clark Kent” and associated in thought with *Clark Kent*. This file mediates such deployments in thought of the individual concept   =  as are constituted by tokenings *Clark Kent*. Luthor comes to believe, via a tokening in thought, of *Clark Kent* the proposition . Of course, Luthor also has a file labeled “Superman.” This file mediates such deployments of the individual concept as are constituted by tokenings *Superman*. Luthor believes, via a tokening of *Superman*, the proposition (= the proposition ). In addition, Luthor desires

170  Referring to the World true, via tokenings *Superman*, the proposition . Unfortunately for Luthor, because of the combination of the referential independence of his two representational vehicles and the dynamic independence of his two mediating conceptions, his *Clark Kent*-​tokening-​ constituted belief and his *Superman*-​tokening-​constituted desire co-​exist in splendid isolation from each other, despite the coincidental coreference of his two representational vehicles. And it is precisely this unfortunate collection of facts about the psychic organization of Luthor’s mind that costs him the opportunity to satisfy his most ardent desire. The cognitive predicaments of Luthor and Jocasta are the stuff of which tragedy and missed opportunities are made. Unfortunately for us, it is a fact of cognitive life for creatures like ourselves that even within an internally coherent mind, whatever exactly such coherence might be thought to consist in, there may often occur dynamically unlinked conceptions of the selfsame extension which independently mediate deployment of the selfsame concept, which is, in turn, expressed twice over by referentially independent internal representations. And precisely because of such architectural limitations, which may be present even in a fully rational mind, we run the constant risk of external incoherence. When we are externally incoherent then not even fully internally coherent narrow right reasoning, as we might call it, will suffice to guarantee that our beliefs, desires, intentions, and actions will relate to one another in ways which enhance our worldly success. For nothing will guarantee that in such a mind our beliefs will be true, our reasoning will be truth preserving, or that our actions will be desire-​satisfying. But this way of looking at Luthor and Jocasta suggest that their predicament is not a predicament internal to reason as such. Reason may from time to time find itself in such a predicament through no fault of its own. But precisely because such a predicament is, as it were, thrust on reason rather than brought forth through reason’s own failure, neither Jocasta nor Luthor count as irrational. I do not want to deny that there is something to this thought. And yet one might be also inclined to think that being rational is more than about having clean epistemic hands when failure and misfortune strike. One might be inclined to adopt what I have elsewhere called a “consequentialist” view of rationality, according to which rationality tends to guarantee by its very nature the worldly success of our cognition and conation. By that measure the very fact that Jocasta and Luthor are in a way doomed to failure might well be taken as a sign if not of flat out irrationality, then at least of less than full, complete, or ideal rationality. To do so would be to elevate external coherence, a

Concepts, Conceptions  171 kind of coherence which seems to require cooperation from the world beyond the exercise of narrowly coherent right reasoning, into a necessary condition on being a rational mind. To adopt a consequentialist understanding of rationality would be to reject a wholly internalist understanding of rationality as a power to guarantee only that our minds are not self-​negating. Nonetheless, I do not think that we should wholly abandon internalism about rationality. If external coherence were constitutive of rationality, rational minds would have to know all the true identities upon which the success of their behavior is contingent.8 That seems to demand too much of 8 Millikan (1993) recognizes the relevance of something like external coherence to rationality, but she goes too far. Consider the following: Because rationality pivots on Kontent [roughly, broad content] and does not reside in some inner, safer realm, it is also true that no manipulation of modes without regard to whether or not these in fact have kontent, and without regard to whether they have, perhaps multiply ambiguous kontents, could possibly be a manifestation of rationality. Having an automated formal system unfolding inside one’s head is not being a rational creature if the system has no interpretation. Nor is it being a rational creature if each symbol ambiguously means, or is undifferentiated among meaning several different things. Imagine a head full of descriptions whose component terms are empty. Imagine a head full of a thousand descriptions and (indexical) names all of the same object but without the head’s knowing this. This would be rationality? (349.) Millikan is right to suggest that a cognizer who cognizes the same object twice, but without recognizing that she has done so, is no paragon of cognitive virtue. Such a mind will be externally incoherent, its internal coherence notwithstanding. But to elevate external coherence, as Millikan does, into a requirement on rationality is really to insist that only minds that are “nicely” embedded in their environments count as rational. It is hard to see how this approach can distinguish rationality from mere cognitive good fortune. Fodor (1994, 41) too apparently thinks that something like external coherence is partly constitutive of rationality. For note the following: Rational behavior is, generally, pretty successful as a matter of fact; a lot more successful, anyhow, than behavior that is simply crazy. No remotely acceptable intentional psychology could count this fact as accidental. But it would be accidental—​it looks like it would be unintelligible—​unless generally speaking, people know and respect the facts that the outcomes of their actions depend upon; including, in particular, the facts about what is identical to what. The only means that a belief/​desire psychology has to insure that, in typical situations where a = b, Smith will assign appreciably similar utilities to Fa and Fb, is to insure that Smith believes that a = b and that he makes the relevant inferences. (Fodor, 1994, 41) Again, I agree with Fodor that our plans and actions will often go astray and that we will often miss opportunities if we do not know relevant true identities. Moreover, we can be brought by reflection to know this about ourselves. Because we can know this about ourselves, we endorse external coherence as a regulative ideal on our cognitive strivings. But it demands too much to require that a rational mind know all true identities. Nothing merely internal to our own heads, taken either severally or collectively, can guarantee that when we are presented with the same object again, we will recognize that we are. No cognitive policy or practice and no bit of “social engineering” can guarantee such a happy outcome. Indeed, assume that Jones and Smith adopt exactly similar cognitive policies and practices and are subject to a similar social division of cognitive labor. Still, it is possible that, due to the luck of the environmental draw, Jones is confronted with a plethora of Frege cases and Smith is confronted with few. By Fodor’s lights apparently (and apparently also by Millikan’s) Jones may count as irrational, while Smith counts as rational. So again, rationality turns out to be partly a matter of the luck of the environmental draw. But again, though a commitment to try, ceteris paribus, to reduce the

172  Referring to the World reason. At the same time, internal coherence asks too little. Internal coherence is a merely negative, merely syntactic virtue, displayed by a will and/​ or intellect that are merely not self-​negating. It is not a virtue that makes an intellect sufficient for cognizing the epistemically good, nor a will sufficient for achieving the conatively good. Nothing in a merely internally coherent mind can guarantee that the objective representational contents of inner representations will be so ordered as to guarantee external coherence. It is fair to wonder, therefore, whether there is a middle path, a path that is still internalist but accommodates from within a still internalist perspective the felt normative pull of external coherence. I devote the remainder of this section to sketching one such path. The middle path takes external coherence to be reason’s own self-​given regulative ideal on its cognitive strivings. In giving itself external coherence as a regulative ideal, reason commits itself to striving, ceteris paribus, to so arrange its representations that representations that share an outwardly oriented representational content also come to share a common syntactic life in episodes of narrow right reasoning. In endorsing external coherence as a regulative ideal, reason takes the plenum of its inner representations to be more than a closed Cartesian circle, subject only to constraints of syntactic well-​orderedness. In aiming to be externally coherent, reason seeks to bring it about that the inner syntactic lives of inner representations mirror the outwardly oriented representational contents. It strives to make the case that, ceteris paribus, there are not two dynamically unlinked ways of thinking the same thought again, of deploying the same concept again. It is only because external coherence is one among the defining ideals of a rational mental life that such a mental life presents itself to us as a mental life worth pursuing. After all, it is only when the inner syntactic lives of our inner representations track nicely with their outwardly oriented representational contents that our cognition and conation can possibly exhibit the hallmarks of what I earlier called consequential rationality. In the absence of external coherence, the merely negative virtue of internal coherence leaves us still dangerously liable to tragedy and missed opportunities in the manner of Jocasta and Luthor. That is because it leaves us with the possibility that dependency of our cognitive and practical success on the luck of the draw does seem constitutive of rationality, it seems wrong to elevate actual achieved external coherence into a constituent of rationality. Rational minds ought to strive to be externally coherent. They ought, ceteris paribus, to prefer being externally coherent to being externally incoherent. But no merely finite mind can guarantee its own external coherence.

Concepts, Conceptions  173 representations that share an outwardly oriented representational content will persist in syntactic isolation one from another. In cognizing its own worldly limits while, nonetheless, elevating external coherence to a regulative ideal on its cognitive strivings, reason commits itself to the unending endeavor to free itself from the whims of mere epistemic fortune. To be sure, it is worth noting that there are possible worlds in which, though Jocasta’s internal mental life is, in one sense, just the same as it actually is, she is so embedded in the world that her inner symbol *Tad* and her referentially independent inner symbol *Oedipus* refer to two different individuals rather than to the same individual as in the actual world. And this possibility will suggest to some that the so-​called narrow or intrinsic contents of Jocasta’s head might remain just as they are even in a world in which Jocasta’s *Tad* and Jocasta’s *Oedipus* fail to be even coincidentally coreferential. In any such world, Jocasta’s *Tad* involving thoughts and her *Oedipus* involving thoughts will be thoughts about different objects and also, presumably, thoughts involving different (wide) contents. In some such worlds, Jocasta marries the referent of *Oedipus* without marrying the referent of the referentially independent *Tad*. In such worlds, Jocasta’s mental life is externally coherent and she enjoys, therefore, the possibility of worldly success. Unfortunately for Jocasta, as for us all, her intrinsic powers cannot guarantee that she is embedded in the world she inhabits in such a way as to make her externally coherent. And just such considerations will suggest to some that rationality should be defined not over broad contents, which depend partly on worldly relations between representations and representeds, but only over narrow contents, which supervene only on what’s in the individual cognizer’s heads. Viewed in this light, we are back to the idea that Jocasta’s “predicament” is not a predicament of reason at all, at least not a predicament of reason’s own making. Her problem lies not in her representations or in her rational manipulation of those representations but in her “extra-​ rational” connections with the world.9 It is precisely the lack of any inner guarantee of external coherence that makes it implausible, from an internalist perspective, that external coherence is partly constitutive of rationality. And that is why I grant that rationality cannot be what I have called consequential rationality. Still, I think it would be a mistake to entirely deny the normative pull of the considerations of metaphysical possibility and necessity on which the notion of external coherence 9 See Essays VII, XI–​XII in Taylor, 2002.

174  Referring to the World depends. By treating external coherence as a self-​given regulative ideal, we give due deference to such considerations. We do so without abandoning internalism and without reducing rationality to a merely negative and, as it were, syntactic virtue. External coherence is a self-​given project of reason. And it is not that the rational mind merely happens to give itself the project of achieving external coherence, as if on a whim. Reason gives itself the regulative ideal of external coherence from the very depths of its self-​conception. To be sure, reason strives to carry out this self-​given project without any internal guarantee of success, but the absence of a guarantee of success is not the absence of the possibility of success. The directive that reason gives to itself in endorsing external coherence does not command the impossible. Rather, it commands pursuit of the possible and an attitude of non-​complacence in mere internal coherence. That we who have the inner power to guarantee only our own internal coherence, nonetheless, constitutively endorse external coherence as a regulative ideal on our cognitive strivings says a great deal both about the nature of our minds and about our self-​conception of ourselves as beings in the world. First, the standing of external coherence as a self-​given regulative ideal reflects what I  call the “syntactic plasticity” of our minds. A  syntactically plastic mind contains a plenum of inferentially articulated representations, together with the power to reconfigure its representations by laying down new and extinguishing old inferential links among its representations on the basis of its experience in ways that enhance both their epistemic and non-​ epistemic worldly goodness. In such a mind, reason’s giving itself the regulative ideal of external coherence amounts to reason’s giving itself the project of constraining the configuration and reconfiguration of its syntactic landscape with an eye toward an external world not of its own constituting. The mind’s inner syntactic landscape is present to reason as simultaneously subject to reason’s own configuration but also as not directly or automatically constrained by the external world to which reason seeks to render its representation answerable. In giving itself the regulative ideal of external coherence, reason introduces, on its own authority, a constraint from the world, while retaining full awareness that it cannot be guaranteed to satisfy the constraint merely through the absence of syntactic self-​negation. Syntactic plasticity has many expressions. It is expressed as a capacity to engage in the sort of theory-​mediated, non-​demonstrative inferencing via which low level irradiations at our sensory surfaces can come to be inferentially linked to representations of things remote from immediate sensation.

Concepts, Conceptions  175 It is also expressed as capacity to engage in the sort of non-​monotonic reasoning that enables us to withdraw a conclusion based on increased information, even when we continue to endorse the premises from which the now dropped conclusion earlier followed. Yet another expression of syntactic plasticity is our ability to alter our action plans on the basis of our ever-​ changing beliefs, desires, and opportunities via the practical syllogism. Such powers of reconfiguration enable the syntactically plastic mind to construct ever better theories of the world, to learn new, deeper, more explanatory regularities, to come to see that which is merely locally and contingently reliable as merely locally and contingently reliable, and to acquire ever more powerful practical abilities to manipulate the world for its own ends. Some intentional systems inhabit environments in which the presence of the good can be reliably enough discerned in some uniform, low-​cost way. Such intentional systems can afford to be prisoners of fixed ways of locating and tracking the good. But for systems that inhabit less anticipatable environments—​that are variable in ways relevant to the system's pursuit of the good, in which the presence of the good cannot be reliably discerned or achieved in any uniform way—​syntactic plasticity will be of great value. Only cognizing agents with some degree of syntactic plasticity can face an unanticipatable world with promising prospects of success, relatively independently of the luck of the environmental draw. To be sure, no cognizer with a finite brain is likely to be indefinitely syntactically plastic.10 But a modicum of syntactic plasticity is a sine qua non of rationality. Nothing is a rational mind that entirely lacks the power to reconfigure its syntactic landscape on the basis of its experiences in ways which enhance epistemic and conative goodness. To build a rational mind, one endows a mind with a plenum of representations with inner syntactic lives and with outwardly oriented representational contents. One links those representations in ways that 10 In holding that relatively implastic cognizers often succeed, if they succeed, merely because of the luck of the environmental draw, I need not deny that, for example, naturally selected cognizers typically are embedded via natural selection only in environments in which their cognitive strategies are at least reasonably adaptive. Nor can we necessarily say that the cognitive strategies of relatively implastic cognizers succeed merely because of the luck of the environmental draw, at least not if we view the organism-​environment relation from the perspective of evolutionary time. From the perspective of evolutionary time, it’s in one sense non-​accidental that cognizers tend to be embedded in environments in which their cognitive strategies represent winning strategies in the competitive game the organism plays with its environment. But this outcome results from the fact that over the course of evolutionary time the cognizers that are left standing are the ones epistemically lucky enough to find themselves embedded in environments in which their cognitive strategy is a winning one.

176  Referring to the World subserve the cognizing agent’s pursuit of various worldly goods. One endows that mind with the power to reorder its syntactic landscape, in goodness enhancing ways, on the basis of its experience in the world and its actions on the world. And one endows that mind with a power of reflection by which it takes ownership of norms of conduct and inquiry and thereby constitutes a space of reasons as its own. There is no deep or principled reason to deny that such a mind may be built out of a kit of entirely naturalistic materials. The kit will include such relatively familiar items as inner syntactically type-​ identifiable representations in something like a language of thought, an evolutionarily instilled functional role psychology, second-​order mental states within that functional role psychology, and some version of the causal/​informational theory of conceptual content.

6 Representing Representations The Priority of the De Re

Besides referring to objects in the external world, we also somehow manage to refer to our inner mental states, and not just to our own but to the inner mental states of others. In this chapter, I shall argue that part of what enables us to talk and think in an articulated manner, about the states of both our own minds and the minds of others, is the very fact that our inner representations are outwardly semantically oriented. We cognize our inner mental states and their contents by cognizing those states and contents in relation to the external world to which they are paradigmatically semantically answerable. The mind opens up its inner contents to view by opening itself up to a largely shared, largely mind-​independent external world. Indeed, I shall argue that we describe and refer to our inner thoughts largely by redeploying devices we already deploy in describing and referring to the external world. Once we have the power to say or to think, for example, that the cat is on the mat, we are already a long way toward having the power to talk or think about the contents of a mind, to describe a mind as believing that or fearing that the cat is on the mat. To say this is not to deny the distinction between the inner world of thought and the outer world to which thought is answerable. It is only to insist that our ability to refer to our thoughts and their contents is, in a way, parasitic upon our ability to refer to the world. The priority of reference to the world over reference to inner mental representations and their contents is reflected, I shall argue, in the relative priority of what philosophers typically call “de re ascriptions” of mental contents over so-​called de dicto ascriptions of mental content.

1.  From World to Mind and Back Again Begin with the following case. Jones is wondering where the cat has gotten to this time. Her roommate Smith utters: Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0006

178  Referring to the World (1) The cat is on the mat

Taking Smith at her word, Jones quickly comes to believe two things. She comes to believe something about the cat—​that it is on the mat. She also comes to believe something about Smith—​that Smith believes that the cat is on the mat. Now suppose that Black is curious not about the whereabouts of the cat, but about Smith’s beliefs about the whereabouts of the cat. Thinking that Jones can tell, Black asks Jones about Smith’s beliefs. In response, Jones utters (2): (2) Smith believes that the cat is on the mat.

(1) as uttered by Smith and (2) as uttered by Jones clearly differ in subject matter. (1) is about the cat and its whereabouts. It is true just in case the cat is on the mat. The truth or falsity of (2), by contrast, in no way depends on the whereabouts of the cat. Its truth or falsity depends entirely on Smith’s state of mind—​on whether she takes the cat to be on the mat. She may do so wrongly or rightly. But whether she does so rightly or wrongly is entirely irrelevant to the truth value of (2). Despite this difference in subject matter, there is an intimate connection between (1) as uttered by Smith and (2) as uttered by Jones. (2), as uttered by Jones, is a way of reporting the belief expressed by Smith in uttering (1). We execute such transitions, from talk about worldly objects and their properties to talk about mental states and their contents, all the time. We glide so easily from the one to the other that the transition mostly escapes our notice. And it works both ways. We learn much about the world from reports of what others say and believe. Upon being told by Jones that Smith believes that the cat is on the mat, Black may herself come to believe, and perhaps even to know, something not just about the states of mind of Smith or Jones but also something about the world. Though transitions between thought and talk about worldly objects to thought and talk about states of mind are so familiar to us as to seem second nature to us, there is a long-​standing philosophical tradition, endorsed by philosophers with widely varying philosophical outlooks, which makes the very possibility of such transitions if not exactly mysterious, then at least a bit puzzling. As with much of our thinking about reference, that tradition originates with Frege (1918/​1977) and his treatment of the various substitution puzzles, as we have detailed at various points throughout this book. According to that tradition, in making at least certain sorts of

Representing Representations  179 attitude ascriptions—​what are often called de dicto or “notionally sensitive” ascriptions—​speakers somehow manage to refer to, describe, quantify over, or somehow pragmatically implicate the notions, representations, or modes of presentations that plausibly figure as constituents of our mental contents—​ either to the exclusion of worldly objects themselves or in addition to worldly objects.1 That is, it is facts about modes of presentation, broadly construed, and agents’ relations to them that are relevant to the truth or falsity of belief ascriptions. Tradition would have it that in ascribing the belief that the cat is on the mat to Smith, Jones relates Smith not to the cat and the mat, at least not directly, but first and most directly to certain modes of presentations, notions, or representations of the cat and the mat. That is what is supposed to explain why the subject matter of (2) differs from the subject matter of (1). And it is also supposed to explain the supposed opacity of attitude contexts, since two designators of the self-​same cat and the self-​same mat might fail to be associated with the same mode of presentation or representation of said cat and said mat. Nor does the tradition take notionally sensitive ascriptions to be secondary or derivative cases. Such ascriptions are held to be the primary, central, or unmarked case of attitude ascriptions. It has seldom been explicitly remarked upon that this traditional wisdom about attitude ascriptions leads to something of a puzzle. Begin by noting that in our ordinary thought and talk about the world, we typically make no reference to the concepts, ideas, or representations out of which mental contents are presumably constituted. I do not mean to deny that we deploy various representational entities in thinking and talking about the world. The point is that in our ordinary discourse about worldly entities, like cats and mats, we refer to and predicate properties of those worldly entities themselves. We do not refer to or predicate properties of whatever representational items we deploy in thinking and talking about those worldly entities. But it is precisely such ordinary reference to ordinary objects that typically supports our further claims about what speakers believe. But on any broadly Fregean theory, worldly objects will relate to representational items that supposedly serve as ingredients of thought content in a one-​many fashion. For every worldly object that may serve as a referent, there will be many distinct notions, ideas, modes of presentation, or their ilk that may function in our thought and/​

1 Frege did not himself distinguish between the de re and the de dicto—​neither at the level of ascriptions nor at the level of beliefs themselves. Nonetheless, it is easy to find the roots of such a distinction in Frege. It was Quine (1956) who first brought that distinction to philosophical prominence.

180  Referring to the World or talk to pick out that worldly item. And on broadly Fregean views, there is typically no (automatic) path “back up” from worldly objects to modes of presentations and their ilk. But this is precisely why the Fregean approach to attitude ascriptions generates a puzzle. To see this, we revisit Smith’s utterance of (1). In uttering (1), Smith refers to some contextually salient cat and some contextually salient mat. But in keeping with Fregean orthodoxy, we might suppose, at least for the sake of argument, that Smith can do so only via the mediation of some mode of presentation-​like entity or other. Her belief is about the cat only because the cat satisfies or answers to the mode of presentation via which Smith cognizes the cat. The crucial further point is that it is supposedly this mode of presentation that constitutes or determines what we might call the “de dicto” or rational content of her belief. And it is paradigmatically this de dicto or rational content that we somehow seek to specify or characterize, the Fregean says, when we ascribe a propositional attitude to another. But it is important to realize that propositional attitude ascriptions are, in the main, communicative acts, by means of which conversational interlocutors seek to achieve mutual understanding. Let us consider then precisely what is necessary for the purposes of achieving mutual understanding of Smith’s utterance on the part of our two conversational partners, Smith and Jones. Clearly, Smith and Jones need to achieve mutual recognition of the intended reference. If it were not already mutually manifest exactly which cat or which mat was at issue, Jones might seek clarification. Smith might offer clarification with the response, “why, the black cat,” or “the mat in the corner of the living room.” But now ask yourself whether Jones also needs to recognize which of the many possible modes of presentation of the cat Smith cognizes and refers to the relevant cat under. The answer would seem to be a resounding no. If it is already mutually manifest which cat is at issue, it is hard to imagine a scenario in which any further question about which of the many possible modes of presentation of the cat Smith actually cognizes and refers to would be conversationally relevant. Once Jones recognizes which cat and which mat are at issue in Smith’s assertions, then whatever she knows or doesn’t know about how Smith is thinking of the cat, she already knows everything she needs to know to be warranted in uttering (2) in conversation with Black as a way of ascribing to Smith the belief that Smith expresses in uttering (1). Moreover, in uttering (2) to ascribe a belief to Smith in conversation with Black, in the absence of Smith, there is no reason at all to presume that Jones must thereby be intending to inform Black of the

Representing Representations  181 mode of presentation via which the relevant cat or the relevant mat was originally presented to Smith. Indeed, given the absence of a path “back up” from worldly objects to modes, it is not at all clear how she could possibly carry out such a communicative intention, even if she happened, for some odd reason, to have one. More generally, if the truth of our thought and talk about mental content really was by default semantically sensitive to facts about modes of presentation and their ilk, as those who take de dicto ascriptions to be the unmarked case appear to believe, it is fair to wonder how speakers would possibly manage to so easily and effortlessly transition from thought and talk about worldly objects to such presumably notionally loaded thought and talk about mental contents. It is not at all obvious how we could ever be sure that we had managed to refer to the right mode of presentation, or even the right kind of mode of presentation, in ascribing a belief to another. Precisely this is the underappreciated puzzle to which the Fregean tradition gives rise. Now it is important to distinguish what I will call the “Fregean Thesis” about attitude ascriptions from what will I  call the “Fregean Mechanism” for making good on the Fregean Thesis. The Fregean Thesis is the thesis that in specifying mental contents via propositional attitude statements, we somehow manage to either refer to, wholly or partially describe, quantify over, or somehow pragmatically implicate putative facts about the representations or modes of presentation that presumably figure as constituents of our attitude contents. The Fregean mechanism, by contrast, is a claim about precisely how the Fregean Thesis is implemented. Frege himself sought to implement the Fregean Thesis in a quite specific way. He famously held that in the context of attitude ascriptions, embedded terms and predicates undergo a shift in both sense and reference, thereby coming to denote what they customarily express—​that is, a sense or a mode of presentation. But many contemporary philosophers accept the Fregean Thesis while rejecting the Fregean mechanism. Indeed, philosophers of language have offered a rather dizzying array of pragmatic and semantic alternatives to the Fregean mechanism for implementing the Fregean Thesis.2 My target in the first part of this chapter 2 Philosophers who reject the Fregean mechanism, while accepting the Fregean Thesis tend to endorse Davidson’s (1968) view that embedded terms have an innocent semantics, while also taking at face value Frege’s observation that substitution of coreferring terms within attitude contexts fails to preserve truth value. Since the Fregean Thesis and the Fregean Mechanism may seem to be a match made in heaven, much philosophical creativity has been expended on decoupling the Fregean Thesis and the Fregean Mechanism. The list of those who have sought to decouple the two is long. Some especially prominent examples are Recanati (1993, 1995, 2010), Crimmins (1992, 1995, 1998), Crimmins and Perry (1989), Richard (1990), and Schiffer (1977, 1995, 2003). Others, like Soames

182  Referring to the World is the Fregean Thesis itself and not just this or that means of implementing it. I do not deny that the subject matter of an attitude ascription like (2) differs from the subject matter of a worldly statement like (1). What I deny is the claim that the shift from thought and talk about worldly objects to thought and talk about states of mind requires the intervention of some peculiar semantic or pragmatic mechanism of whatever type by means of which we are somehow enabled to refer to, describe, quantify over, treat as unarticulated constituents, or somehow pragmatically enrich to such representational entities as Fregean modes of presentation, ideas, notions, individual concepts, or anything else of the sort. The point is that the Frege-​inspired tradition is not just false in detail. It is not just a matter of settling on the right mechanism for implementing the Fregean Thesis. The Fregean tradition, and with it the Fregean Thesis, is entirely misguided in spirit. It embodies not just a mistaken conception of the nature of thought content, as I have argued at various stages of this book already, but also a mistaken conception of the nature and point of our talk about the contents of our thought and talk. We talk about the contents of our states of mind, not by referring to peculiarly mental or representational entities such as notions or modes of presentation or even concepts as understood in Chapter  5, but primarily by referring to worldly entities themselves. To attribute to another the belief that the cat is on the mat, one refers not to representations or modes of presentations or conceptions or even concepts of said cat or said mat, but simply to the relevant cat and the relevant mat. This is not an entirely novel thought. Davidson (1968) long ago insisted that if we could but regain our pre-​Fregean semantic innocence, it would strike us as frankly incredible that embedded expressions refer to anything other than, or additional to, what they refer to when not embedded. Indeed, many subsequent philosophers have staked a claim to having recovered our pre-​Fregean semantic innocence. For the most part, however, they have done so while still stubbornly hewing to the Fregean Thesis. Typically, they reject only the Fregean mechanism for implementing the Fregean Thesis. But I will argue that an even sharper break from the Fregean tradition is needed. We should reject not just the Fregean mechanism, but the Fregean Thesis itself. I do not pretend, however, that the (1985, 1989) or Salmon (1986, 1989, 1995), refuse to take Frege’s observation at face value and instead take substitution failures as something of an illusion. On my view, both approaches are mistaken, though in different ways, and approaches of the former sort—​which take failures of substitution at face value—​are more wrong than approaches of the latter sort—​which take such apparent failures to be illusory.

Representing Representations  183 arguments that follow constitute a complete and decisive refutation of either the Fregean Thesis or of the many alternative non-​Fregean semantic and pragmatic mechanisms for implementing the Fregean Thesis. At a minimum, I hope they suffice to show that the Fregean Thesis is less well grounded in the actual behavior of embedded expressions than many have supposed. Once we see that much, the way is opened to simply and fully reclaim, without hesitation or regret, our pre-​Fregean semantic innocence.

2.  Commitments Ascribed versus Commitments Undertaken My claim is not that attitude ascriptions are never notionally sensitive or that we never manage to somehow pragmatically implicate or semantically refer to the representational entities or their contents. The claim is rather that our ability to make notionally sensitive ascriptions is parasitic on our ability to talk about worldly objects. Indeed, notionally sensitive ascriptions typically require special stage setting and/​or the deployment of special purpose linguistic constructions. That is because notionally sensitive ascriptions are the marked rather than the unmarked case of attitude ascriptions. Contrary to the widely endorsed conventional philosophical wisdom about attitude ascriptions, I shall argue that garden variety attitude ascriptions, of the sort that tend to occur in everyday discourse, have more in common with so-​ called de re ascriptions than they do with so-​called de dicto ascriptions.3 Begin by considering a bit more closely Jones’s utterance of (2), while in conversation with Black about Smith’s beliefs.4 Notice first, that in reporting Smith’s belief about the whereabouts of the relevant cat, Jones need not herself

3 I will not argue the point here as I have argued elsewhere that the de re/​de dicto distinction is itself deeply problematic and unprincipled. There is, I think, no single and coherent way to neatly divide ascriptions into those that are de re and those that are de dicto, at least not in a fully principled way. So that distinction is perhaps best consigned to the dustbin of philosophical history. See Taylor (2002). See also Crimmins (1992, 1995), Richard (1993), Bach (1997a, 1997b), and Recanati (2000, 2010) for discussion of some of the difficulties of making out a single, coherent, and principled version of the de re/​de dicto distinction. 4 Throughout I will be concerned with beliefs about particular objects—​whether those beliefs are expressed in sentences using names, descriptions, or demonstratives. I am not discussing here the ascription of fully general beliefs, which raise interesting issues of their own. Moreover, at least in the first part of this essay, I take no stand on whether beliefs about particular objects involve relations to so-​called singular propositions. My argument in section 2 is meant only to contest the claim that in ascribing beliefs about particulars we put the ascribee’s notions/​modes of presentations/​ways of cognizing those particulars at semantic issue via the mechanism of embedding.

184  Referring to the World express any view of her own about the cat’s whereabouts. In attributing a belief to Smith, Jones is attributing what I will call a predicative commitment to Smith. In attributing a predicative commitment to Smith with respect to the cat, Jones represents Smith as predicating a certain property of the cat. She does not thereby state or imply that she herself shares the predicative commitment she ascribes to Smith. That is, Jones herself need not take the cat to be located on the mat in order to represent Smith as taking the cat to be so located. Jones ascribes such a commitment to Smith but does not undertake such a commitment herself. To appreciate the difference between predicative commitments undertaken and predicative commitments ascribed, consider (3) below.5 In (3), Jones ascribes a certain predicative commitment to Smith but also expresses or undertakes a predicative commitment of her own that is distinct from the one she ascribes to Smith: (3) Smith believes that the cat is on the mat, but it is really under the table.

The fact that Jones can ascribe a predicative commitment to Smith with respect to the cat without thereby undertaking a predicative commitment of her own is a result of the dialectical function, as I call it, of an embedded predication. The dialectical function of an embedded predication is precisely to make explicit and manifest the predicative commitments that are being ascribed to the ascribee of the belief ascription. Just as there is a difference between predicative commitments ascribed and predicative commitments undertaken in the making of an attitude ascription, so there is a difference between referential and existential 5 Philosophical orthodoxy tends to construe propositional attitude ascriptions as relational. They either relate a believer to a proposition, as in so-​called de dicto ascriptions, or, they relate a believer to an object and something further—​such as a property or, perhaps, a propositional function. As such, philosophical orthodoxy tends to focus primarily on worries about the logical form and compositional semantics of attitude ascriptions. My focus in this essay is not primarily on questions of logical form or semantic content—​though my views do have consequences for such matters. I am more concerned with what might be called the pragmatics of attitude ascriptions. I want to know what we are doing when we are making an attitude ascription. In particular, I want to know what sort of communicative act we are typically performing in making an attitude ascription. Because of the excessive focus of the philosophical tradition on matters of logical form and semantic content, we have largely lost sight of the communicative dynamics in which the ascription of propositional attitudes tends to be caught up. An important outlier here is Brandom (1994). Though I do not endorse the sort of inferentialist semantics Brandom defends, I think he is right to give pride of place to pragmatics. And this, I think, leads him to a view of the priority of the de re very similar in spirit to the view defended in this chapter. It is worth noting that some linguists have taken notice of what I am called the default ascriber-​centeredness of certain sorts of expressions. See, for example, Potts (2005) and Harris and Potts (2009) on the ascribee-​centeredness on embedded appositives and embedded expressives.

Representing Representations  185 commitments ascribed or undertaken in the making of such ascriptions. Undertaking a referential commitment is a matter of undertaking to refer to a certain object and of making it mutually manifest that the relevant object is being referred to. Undertaking an existential commitment is a matter of manifestly committing oneself to the existence of various objects. Just as we can distinguish, with respect to predication, between commitments undertaken and commitments ascribed, so we can distinguish between referential and existential commitments undertaken and referential and existential commitments ascribed. It may initially be supposed that in ascribing a belief to another, the ascriber may ascribe existential and referential commitments without herself undertaking those ascribed commitments. Indeed, something like this thought is, I think, behind the belief that de dicto attitude ascriptions are the unmarked case. Indeed, de dicto ascriptions might be thought to be ascriptions in which all commitments—​be they predicative, existential, or referential—​are ascribed but not necessarily undertaken. In a de dicto ascription, one might, for example, ascribe an existential or referential commitment to the existence of some object—​Santa Claus, say—​to the existence of which one is not committed oneself. But it turns out that referential and existential commitments are rather the converse of predicative commitments. Although it is possible for an ascriber to attribute a referential or existential commitment via the use of embedded clauses without undertaking the ascribed commitments, this is much easier said than done. Indeed, in the default or unmarked case, the ascriber typically doesn’t distance herself from ascribed existential or referential commitments. That is, she doesn’t simply ascribe such commitments, she also undertakes commitments of her own. In fact, the referential and existential commitments undertaken by the ascriber need not necessarily be or be represented as fully shared by the ascribee. In many cases, they will be shared, but they need not be shared as a matter of linguistic necessity. What I mean by this will be made clearer in due course. Part of the reason that an ascriber typically needs to undertake existential or referential commitments of her own, possibly distinct from the commitments she ascribes to her ascribee, is so that she can single out relevant objects in a way that is mutually manifest to herself and her interlocutors. There are many different cats in the universe. In conversation with Black about Smith’s beliefs, Jones may intend to ascribe to Smith a belief about the whereabouts of just one of those cats. She thereby takes on the communicative burden in her conversation with Black of making it mutually manifest to

186  Referring to the World Black which cat is at issue in her ascription of a belief to Smith. In the context of a such a conversation, it matters less how Smith thinks or thought of the cat, that is, via which mode of presentation she does or did so. It matters more whether Black can be brought to recognize, in the here and now, which cat is at issue. And I shall argue below that is because of such communicative demands that garden variety attitude ascriptions, in typical conversational settings, generally have more of a de re than a de dicto feel. To help make my ultimate argument, I  begin with an intuition pump. Think of this intuition pump as something of a softening agent. I focus, in the first instance, not on referential and existential commitments but on evaluative commitments, as expressed in the use of slurring referring terms and other forms of derogatory language. The evaluative commitments expressed in the use of embedded slurring referring terms are by now widely acknowledged to be non-​displaceable.6 Because of their non-​displaceability, slurring referring terms are widely taken to be special cases, with special semantic and/​or pragmatic features. I shall eventually argue that non-​displaceability is a more widely prevalent phenomenon than philosophers typically acknowledge. Indeed, I shall argue that for expressions occupying argument positions within embedded clauses, non-​displaceability is the rule rather than the exception.7 This means that what goes for evaluative commitments associated with the use of slurring and derogatory referring expressions occupying embedded argument position goes as well for the referential and existential commitments of non-​derogatory expressions occupying argument position. And I shall argue that what has led the prevalence of non-​displaceability of commitments expressed by embedded expressions in argument place to be vastly underappreciated, is the fact that the non-​displaceability of such referential and existential commitments is often masked by the fact that such commitments may sometimes be defeasibly taken to be part of the shared

6 The philosophical and linguistic literature on slurs is relatively new but growing rapidly. For some early discussions of non-​displaceability, see Kaplan (1999), Potts, (2007), Hom (2008), Hornsby, (2001), Richard (2008), and Taylor (2002). Another class of expressions that have widely been seen to be ascriber-​rather than ascribee-​centered within attitude ascriptions are appositives. See, for example, Potts (2005), Bach (1999), and Asher (2000). 7 It may be tempting to think of non-​displaceability as a matter of scope, especially when we come to the non-​displaceability of existential commitments as expressed by embedded definite descriptions. But I doubt this is correct, either for evaluative expressions or for definite descriptions. What is at issue is whether the relevant constructions are what I call ascriber-​centered or ascribee-​centered. Ascriber-​centeredness is not obviously a result of wide syntactic scope. And ascribee-​centeredness is not obviously a matter of narrow syntactic scope. Nor would it be right to say that always when definite descriptions are used referentially they are ascriber-​centered in an attitude ascription.

Representing Representations  187 common ground in a way that evaluative commitments typically cannot be. But in due course I will have a great deal more to say about the analogies and disanalogies between evaluative commitments, on the one hand, and referential and existential commitments, on the other. We begin by exploring the widely acknowledged non-​displaceability of evaluative commitments. Suppose Smith is both a virulent racist and something of a baseball fan. Her racists tendencies lead her both to seriously underestimate the abilities of people of African descent and to use a certain infamous derogatory term that begins with the letter “n” when referring to such people. In a conversation with Jones about the dearth of baseball players of African descent currently playing in major league baseball, Smith utters the following: (4) Niggers make poor baseball players.

Presume that Jones does not share Smith’s derogatory attitude toward people of African descent, and that she assiduously avoids using derogatory terms for such people in her own thought and talk. Despite that fact, Jones may sometimes have occasion in conversation with others to report on Smith’s beliefs. Because of the non-​displaceability of derogation and her own anti-​ racist proclivities, Jones likely would not report the belief expressed by Smith in uttering (4) by an utterance of (5) below: (5) Smith believes that niggers make poor baseball players.

Despite the syntactic embedding of the slurring referring expression, the non-​displaceability of derogation means that the derogatory force of the slur would fully attach to Jones rather than to Smith. In fact, even though we know from background context that Smith herself is a racist, (5) as uttered by Jones in the relevant context would not even seem to purport to ascribe a derogatory attitude toward people of African descent to Smith—​though it still does attribute a problematic and false belief to Smith. If, for example, Jones uttered (5) and her interlocutors were taken aback or offended by her utterance, she could not escape culpability by saying, “But, I’m just reporting what Smith believes.” The point is that the use of a syntactically embedded slurring referring expression expresses only the ascriber’s derogatory attitude and does not even so much as entail, suggest, or implicate that the ascribee so much as shares that attitude.

188  Referring to the World Consider a slightly different scenario. Jones disagrees with Smith about the baseball abilities of people of African descent. She makes her disagreement known by uttering (6) below: (6) People of African descent don’t necessarily make poor baseball players.

In uttering (6), Jones clearly takes issue with Smith’s belief about the baseball abilities of people of African descent. But notice that she does not directly challenge Smith’s own derogatory attitude in uttering (6). She does, however, refuse Smith’s term for people of African descent by openly and manifestly using a neutral counterpart rather than a slur. Jones thereby distances herself from Smith’s derogation. In so doing, she thereby undercuts any purely linguistic basis that her interlocutors might have for thinking that she herself might just share Smith’s derogatory attitude. But despite the fact that Jones has fully distanced herself from Smith’s derogation by refusing her terms, (7) below, as uttered by Smith, would still seem to correctly and felicitously report the belief expressed by Jones in her utterance of (6): (7) Jones believes that niggers don’t necessarily make poor baseball players.

In reporting Jones’s neutral belief in such expressively loaded terms, Smith clearly commits further derogation. From Jones’s point of view, such further derogation is an entirely gratuitous addition to Smith’s report of her beliefs. It is as if Smith opts to spontaneously increase, on her own accord, what might be called her expressive register with respect to black people.8 But one needs to exercise caution here. Thanks to non-​displaceability, the increase in expressive register in (7) is all Smith’s own and cannot be attributed to her interlocutors. Though her heightened expressive register may be part of what I call the “overall conversational score,” it is not part of the shared common ground of the conversation. Because it is not part of the shared common ground, Smith’s derogation is in no way ascribable to Jones or to any other party to the conversation. The essential point to notice, though, is that despite its gratuitous and even moral offensiveness, (7) is not a linguistically or conversationally problematic way of reporting the belief expressed by Jones in uttering (6). From a linguistic or conversational perspective, 8 For the notion of an expressive register, see Potts (2007). See also Kaplan (1999), especially his discussion of truth plus preserving inferences.

Representing Representations  189 (7) does two distinct things. It successfully reports what Jones believes and successfully communicates Smith’s derogatory attitude. Moreover, it does so in a way that allows us to distinguish what commitment is being ascribed to Jones and what commitment is being undertaken by Smith. We shall eventually see that distinguishing commitments undertaken from commitments ascribed, no matter their variety, is work that any felicitous attitude ascription must do. Let me say a bit more about the distinction presaged above between the developing common ground of a conversation and the developing conversational score of a conversation. We may think of the common ground of a conversation as a stock of mutually accepted propositions. Qua mutually accepted, the common ground may be thought of as in a sense the joint product of those who are participating in the conversation. It is important to stress that the common ground of a conversation is not static but grows throughout the course of the conversation. One way to add a proposition to the common ground is to make a flat-​out assertion. Indeed, we can think of an assertion as a proposal, addressed to one’s interlocutors, to add a proposition to the common ground. But propositions can also be added to the common ground in a number of other ways—​for example, via accommodation and entailment in accordance with the following two principles. If P presupposes Q, P is asserted, and P is not rejected, then Q is added to the common ground. Similarly, if P is asserted and P (obviously) entails Q, then Q is added to the common ground. I have argued at greater length elsewhere that we should think of the total conversational score as including the common ground but also as not being exhausted by the state of the common ground. One element of the total score of a conversation that is not, on my view, an element of the common ground will be a set of what I call expressive registers, one for each participant in the conversation. Expressive registers track expressive commitments undertaken by participants in the course of the conversation. Though the state of the expressive register for each conversational participant is public in the sense of being known to all participants, the expressive commitments tracked in the conversational score need not be mutually accepted or endorsed by all participants. In contrast to the elements of the common ground, expressive commitments are the work of participants, taken severally rather than jointly. But to say this is not to say that expressive commitments are not undertaken openly and manifestly. They are in that sense public. But they are not necessarily undertaken jointly.

190  Referring to the World Because expressive commitments alter the conversational score without necessarily altering the common ground, conversational participants are free to modulate their own expressive registers without regard to the expressive registers of other conversational participants. It is precisely this fact that frees interlocutors to augment or diminish their expressive registers at will, without regard to whether their interlocutors share their expressive commitments. That is why it is a conversationally legitimate move for Black to report what Smith says in (8) about Jones to Brown by uttering (9). (8) Smith: That damned Jones just got tenure (9) Black to Brown: Did you hear? Smith says that Jones just got tenure.

Similarly, Black may express to Smith her approval of what Smith said without also derogating Jones. (10) Black: I’m glad you told me that Jones got tenure.

It would not do for Smith to respond to Black with: (11) That’s not what I told you. I told you that damned Jones got tenure.

It also works the other way around. Suppose Black tells Smith that Jones just got tenure. Smith looks really dejected. Black asks Smith why she is so dejected. Smith answers: (12) You just told me that that damned Jones got tenure.

Black cannot relieve Smith’s disappointment by saying back to her: (13) No, I didn’t. I told you that that that awesome Jones got tenure.

The more general point is that it is not possible for a speaker to implicate her interlocutors in her own expressive commitments just in virtue of undertaking those expressive commitments in conversation with them, at least not as a matter of conversational dynamics alone. This is not to deny that there may be certain ethical or moral pressures to explicitly challenge or direct the derogatory speech of one’s conversational partners, on pain of some sort of moral or ethical complicity in the derogation. But as a matter of pure

Representing Representations  191 conversational dynamics, expressive commitments undertaken in the course of derogating speech acts should not necessarily be thought of as being proposed as candidates for mutual acceptance in the manner of either asserted or presupposed contents. Consequently, speakers are not under the same sort of conversational pressure to accommodate expressive commitments of their interlocutors, as they are to accommodate presuppositions of their interlocutors. Though the expressive registers of each party to a conversation are mutually known matters of public record, they need not be mutually accepted in order for the conversation to proceed. It is clear that if one wants to report what is said or believed by one who has derogated another, one must not use derogatory expressions oneself, not even in embedded contexts, unless one is willing to be implicated in the derogation oneself. But suppose, returning to our earlier example, that Jones wants not only to convey the content of Smith’s belief about the baseball-​playing abilities of African Americans but also to convey that Smith has the sort of derogatory attitude toward people of African descent typically expressed by using a racial slur. And suppose she wants to do so without herself derogating people of African descent. What is she to do? One thing she might try out is what I have elsewhere called a “truncated de re report”—​as in, (14) below: (14) Smith believes that people of African descent, to whom she often refers via the infamous N-​word, make poor baseball players.

But if the goal is to ascribe an explicitly derogatory commitment to Smith, (14) does not do the intended trick. (14) does indeed ascribe to Smith a belief that only a racist would be likely to hold. But it is silent about Smith’s derogation. To appreciate this silence, consider a slightly different scenario. Let Smith and Jones reverse roles. Suppose it is Jones, the ascriber, rather than Smith, the ascribee, who has the derogatory attitude. But suppose that despite her derogatory attitudes, Jones does not believe that people of African descent make poor baseball players—​though Smith does, despite not sharing Jones’s derogatory attitude. In that case, an utterance of (14) would clearly not express Jones’s derogatory attitude toward people of African descent—​an attitude not shared by Smith. But it would correctly ascribe to Smith a certain predicative commitment—​a commitment not shared by Jones. So even in ascribing to another a belief that we suspect that only racists might hold, we are not thereby either directly attributing or expressing the sort of derogating commitment that is typically directly and manifestly expressed in the

192  Referring to the World use of an explicit slur. We can clearly express racist beliefs without the use of a slur. But just as clearly, a slur adds an additional expressive commitment even to racist beliefs. How, if at all, can we report on the totality of commitments undertaken by a derogating racist without ourselves having to participate in the derogation? What we need is a form of ascription that allows us to have both ways. That is, we need a form of ascription that allows us to attribute expressive commitments without undertaking expressive commitments of our own. Fortunately, there is a form of ascription readymade for just that task, as I  have argued at length elsewhere.9 Jones can pull off that hat trick by deploying what I have called a fulsomely de re ascription as in (15) or (16):10 (15) Smith believes that people of African descent, to whom she often refers via the infamous N-​word, make poor baseball players. (16) Smith believes that people of African descent, of whom he thinks under the title ‘Nigger,’ make poor baseball players.

In (15) and (16) Jones expands the truncated de re belief report (14) into a fulsomely de re belief report. She does so by adding some additional modifying clauses. These optional clauses are adjuncts rather than arguments. Because the modifying clauses are adjuncts rather than arguments, the truncated de re ascriptions which they modify are perfectly semantically and syntactically in order as they stand even if unmodified. But as unmodified they only partially and abstractly characterize a reality that can be more fully characterized by a more fulsome ascription. But notice too that even when a truncated de re report is expanded into a fulsomely de re report, it still only indirectly characterizes Smith’s way of thinking about people of African descent. In the expanded report, the offending word is not itself used. The resort to this circumlocution does, however, give us a place in the sentence structure to hang either a description or mention of the problematic word in a way that enables 9 See Taylor (2002). There are quasi-​quotational uses of embedded slurs that do seem displaceable. Consider the following example from Potts (2007). (a) My father screamed that he would never allow me to marry that bastard Webster. But it is striking that when the main verb is less quotational in character, the effect either wholly disappears or is considerably weakened. (b) My father will never permit me to marry that bastard Webster. (c) My father insists that I am not to marry that bastard Webster For stronger apparent counterexamples to non-​displaceability, see Harris and Potts (2009). But note that by their own admission such examples require a great deal of pragmatic stage setting. 10 See Taylor (2002, 2003, 2007).

Representing Representations  193 us to more fully characterize Smith’s state of mind. And it does so without having to undertake the relevant expressive commitments ourselves. In the current case this enables Jones herself to avoid derogating people of African descent, while still allowing her to put Smith’s derogation on explicit display. That is why I say that the fulsomely de re ascriptions are a mechanism for ascribing an expressive commitment to another without oneself undertaking any such commitment. I want now to argue that what is true for expressive commitments is also true for existential and referential commitments as well. With the behavior of embedded slurring expressions firmly in mind, let us switch gears and consider existential commitments either expressed or undertaken in the use of embedded definite descriptions. We begin by noting up front that there is an important difference between the expressive commitments undertaken in the use of an embedded derogatory term and the existential commitments untaken in the use of an embedded definite description. In contrast to embedded slurring terms, embedded descriptions seem at least prima facie capable of playing a sort of double role—​both the role of expressing the ascriber’s existential commitment and the role of attributing an existential commitment to the ascribee. Indeed, I suspect that the appearance that embedded definite descriptions play a double role is one of the main sources of the supposed distinction between de re and de dicto ascriptions. If I say, for example, that so and so believes that the current president of the United States is in over his head, such an utterance may be taken to indicate either that I myself am committed to the existence of one and only one person who is currently president of the United States or that it is the ascribee who is so committed or perhaps even that we both are so committed. The supposed double role of embedded descriptions is standardly traced to one of two different ambiguities from which descriptions are widely thought to be subject. The first is the notion of a scope ambiguity. When a description takes narrow scope in an embedded construction, it is widely claimed that we get what I am calling “ascribee-​centered” existential commitments. It is only when a description takes wide scope that we get ascriber-​centered existential commitments. A second ambiguity is the referential/​attributive ambiguity. Kripke (1971, 1977) has convincingly argued that the de re/​de dicto distinction cannot be reduced to the referential/​attributive distinction. But there is perhaps something to the thought that when an embedded description is used referentially rather than attributively by an ascriber, it might plausibly be thought to express the ascriber’s rather

194  Referring to the World than the ascribee’s commitments. It might even be said that descriptions used referentially default to a wide-​scope reading in the context of attitude ascriptions. In that case the two ambiguities referred to above might be thought to be two sides of the same coin. But I  shall show that the very idea that embedded descriptions serve double duty—​either because of scope ambiguities or because of the referential/​attributive ambiguity—​is more problematic than is widely assumed. I do not need to take a stand on whether the interpretation of embedded description is or is not dependent upon either scope ambiguities or the referential/​ attributive ambiguity. On any understanding of how embedded descriptions function, a close look at their actual behavior in attitude contexts suggests that embedded descriptions are heavily centered on the ascriber and less on the ascribee than is widely assumed. How exactly to explain the heavy ascriber-​ centeredness of embedded ascriptions is an entirely different matter. Though I shall have a bit to say about the underlying syntax, semantics, and pragmatics of attitude ascriptions along the way that might explain why this is so, that is not my primary focus here. One way to see and appreciate ascriber-​centered attitude ascriptions really is to replicate the situation that we often find in the case of embedded derogatory expressions. We have already noted that there is often a mismatch between the expressive or derogatory attitudes of the ascriber and the ascribee. And we have noticed that the mechanism for ascribing derogatory commitments that one does not undertake is via the mechanism of fulsomely de re ascriptions. So let us take a look at what happens in cases where the existential commitments of the ascriber (and her interlocutors) and those of the ascribee do not match. Will the embedded description express a commitment undertaken by the ascriber or can they be used to ascribe a commitment to the ascribee that the ascriber does not share? If definite descriptions behave like embedded slurring referring expressions, we should expect the former. If not, we should expect the latter. Consider the following scenario which is meant to parallel the situation we often find ourselves in with respect to slurs. Smith, Jones, and Black are working a party as bartenders. They are instructed by the hosts not to serve anyone who has had too much to drink. Jones spots a very inebriated man standing in the corner downing one martini after another. But she mistakenly takes the drinker to be a woman drinking gimlets rather than a man drinking martinis. With the evident intent of alerting Smith to the drinker’s state, Jones utters:

Representing Representations  195 (17) The woman in the corner drinking gimlets has had too much to drink.

Smith recognizes which partygoer Jones has in mind. But until she is about to share Jones’s intelligence report with her co-​worker, Black, it does not dawn on Smith that the inebriated partygoer is in fact a man rather than a woman and is drinking martinis rather than gimlets. Smith is in something of a communicative pickle. Suppose it is common ground between Smith and Black that there is no inebriated woman drinking gimlets and common ground that there is an inebriated man drinking martinis. Perhaps they know that not a single gimlet has been ordered during the entire evening. Perhaps they have seen the man in the corner drinking martinis but are unsure of exactly how many. Suppose, further, that Smith knows, but Black does not, the nature of Jones’s mistake. Smith is aware, but Black is not, that Jones has misrepresented the inebriated partygoer via the false description “the woman in the corner drinking gimlets.” Though false, this mistaken description may be reasonably thought to partially characterize Jones’s existential commitments. Now it is true that Jones may be said to “refer,” in some sense, to the martini-​drinking man rather than to any gimlet-​drinking woman. This is the so-​called referential use of a definite description. But the fact that Jones may be thought to use the description “the gimlet-​drinking woman” referentially to refer to a martini-​drinking man does not obviate the fact that in using the definite description “the gimlet-​ drinking woman,” she undertakes a commitment to a woman drinking gimlets. And this despite the fact that she refers to no such woman. Moreover, even if she refers to a martini-​drinking man, she undertakes no commitment to the existence of such a man. So how is Smith to report the rather problematic belief expressed by Jones in uttering (17)? One thing is clear; she cannot felicitously use the embedded description “the woman in the corner drinking gimlets” in conversation with Black to ascribe the existential commitment undertaken by Jones in uttering (17). For consider (18): (18) Jones believes that the woman in the corner drinking gimlets has had too much to drink.

as uttered by Smith to Black. Absent further stage setting or clarification, Smith would naturally be taken by Black not merely to be ascribing to Jones a commitment to the existence of a gimlet-​drinking woman in the corner, but

196  Referring to the World also thereby, it seems, to be expressing her own commitment to the existence of such a woman. In the current context, an utterance of (18) would likely send poor Black off on a futile search for a gimlet drinking woman with intention of cutting her off. An utterance of (18) by Smith would clearly convey not only that Jones believes there to be such a woman, but that Smith herself believes it and is attempting to get Black to believe it as well. (18) simply will not do as a way of reporting what Jones believes in uttering (17). The source of the infelicity of (18) as a report of Jones’s belief is, I think, obvious. The existential commitment that would be expressed by an utterance of (18) in the imagined context conflicts with what is common ground between Smith and Black—​that there is no gimlet-​drinking woman in the corner at all. If the linguistic or conversational function of an embedded definite description were simply to ascribe existential commitments to the ascribee, whether or not those shared commitments were endorsed by the ascriber and her interlocutors, such facts about the common ground between Smith and Black should not matter at all. But the ascriber’s own commitments clearly do matter, especially when they diverge from those of the ascribee. Hence, we are, I think, entitled to conclude that the linguistic and conversational function of an embedded description is not, or is at least not solely, to ascribe existential commitments to the ascribee in a way that is entirely independent of the commitments of the ascriber or her interlocutors. This does not yet show that embedded ascriptions do not primarily function in something of a dual role to both ascribe commitments to the ascribee while also expressing commitments undertaken by the ascribee. But by varying the case ever so slightly, we can throw the ascriber-​centered commitments of embedded descriptions into sharper relief. Suppose that the inebriated partygoer that Jones has in mind and to whom she intends to refer via the description “the woman in the corner drinking gimlets” is, in fact, a woman drinking gimlets. And suppose that although it is mutually manifest to Smith and Black who Jones has in mind, they, nonetheless, both mistakenly take Jones to be mistaken. Though Smith and Black mutually recognize that Jones takes the inebriated partygoer in the corner to be a gimlet-​ drinking woman, they take the inebriated partygoer to be a martini-​drinking man. Jones is right and they are wrong. Not only are they wrong, but they are blissfully unaware of their error. Now suppose again that Jones utters (17)—​ this time truly—​intending to alert Smith to the drunken reveler. How should Smith report Jones’s belief to Black? Certainly, from our more informed perspective, it seems evident that Smith would speak truly if she were to report

Representing Representations  197 Jones’s belief to Black via an utterance of (18). That certainly is how we, who are in the know, would report Jones’s belief. But (18) would again be infelicitous. The problem once again is that an utterance of (18) by Smith would express an existential commitment that she manifestly does not have and that, moreover, Black takes her not to have. Indeed, the more felicitous way for Smith to report to Black what Jones believes in the imagined setting would seem to be the by our lights false (19), rather than the by our lights true (18): (19) Jones believes that the man in the corner drinking martinis has had too much to drink.

Again, it appears that by Smith’s use of the embedded description “the man in the corner drinking martinis” in the utterance of (19), she does not ascribe an existential commitment to Jones but undertakes or expresses her own commitment to the existence of a martini-​drinking man. Jones, by contrast, is not committed to the existence of such a man. And this lack of commitment is manifest to both Smith and Black. That is, it is common ground between Smith and Black that Jones mistakenly takes the relevant person not to be a martini-​drinking man but a gimlet-​drinking woman. We can even stipulate that it is part of the common ground between them that Jones takes there to be no martini-​drinking man in the room at all. Of course, an ascription is never all or primarily about the commitments of the ascriber only. In an utterance of (19) Smith would be ascribing a predicative commitment to Jones to the effect that a certain person—​the person whom Smith and Black mistakenly take to be a martini-​drinking man—​has had too much to drink. But this just goes to show that there is an important difference in function between the elements of an ascription that do the work of singling out an object or range of objects from the point of view of the ascriber and her interlocutors and the elements that do the work of specifying the ascribee’s predicative commitments with respect to those objects once they have been singled out. While the former is typically centered on the ascriber and her dialectical partners, the latter is always centered on the ascribee. But it is important to stress that when ascriber and ascribee share existential commitments, the fact that existential commitments are centered on the ascriber, rather than the ascribee, will typically be masked. Focusing on cases in which the existential commitments of the ascriber and ascribee diverge helps us to remove the mask. But by parity of reasoning, even where there is agreement rather than disagreement between ascriber and ascribee,

198  Referring to the World it is typically not the ascribee’s existential commitments that are expressed by the use of an embedded definite description but the ascriber’s. Let us add one final wrinkle to our original scenario. Suppose that Smith, Black, and Jones one and all mistakenly take the martini-​drinking man to be a gimlet-​drinking woman. And suppose that Smith utters (18) as a way of reporting Jones’s belief to Black in that context. Even here, it seems clear that Smith would thereby be expressing her own commitment to the existence of a gimlet-​drinking woman and would not thereby succeed in ascribing such a commitment to Jones. Rather, an utterance of (18) would leave it open whether Jones has the relevant existential commitment. That is, if the fact of Jones’s commitment to the existence of a gimlet drinking woman were not already part of the common ground in the imagined context, the mere utterance of (18) by Smith in that setting would not ipso facto increment the common ground to include such a commitment on Jones’s part. What Smith would definitely ascribe to Jones by an utterance of (18) in this context is a predicative commitment to the effect that a certain person—​a person present to Smith and Black in one way, but possibly present to Jones in quite a different way—​has had too much to drink. The crucial point is that she would not thereby purport to specify how Jones thinks of the relevant person. By using the embedded description, Smith represents only herself to Black as cognizing the relevant object under the description “the woman in the corner drinking gimlets.” She thereby offers up that description to Black as a perhaps negotiable vehicle for Black and Smith to achieve mutual recognition of the object that Jones’s belief is putatively about. But she does not thereby use the embedded description to either represent, indirectly specify, or refer to Jones’s notion of the relevant person. To help drive home the importance of the difference between the ascribee-​ centeredness of predicative commitments and the ascriber-​centeredness of existential and referential commitments, let us revisit the derogation briefly. Consider the following as potentially uttered by Smith to Black: (20) Jones believes that Wanda is a bitch. (21) Jones believes that Wanda is no bitch. (22) Jones believes that that bitch Wanda is her friend. (23) Jones believes that the boss should fire that bitch Wanda.

In (20), Smith ascribes to Jones a certain predicative commitment with respect to Wanda. In (21), she denies that same predicative commitment. In

Representing Representations  199 neither (20) nor (21) does Smith herself derogate Wanda. Indeed, in contrast to evaluative expressions in embedded argument places, embedded evaluative predicates seem to be all about evaluative commitments ascribed rather than evaluative commitments undertaken. I grant that a minority of informants do report that even when a derogatory expression like “bitch” occurs in embedded predicative position rather than in embedded argument position, it tends to convey at least a weak but generalized sense of derogation on the part of the ascriber. But whatever weak and generalized sense of derogation the use of “bitch” in embedded predicate position may convey, its use by Smith in (20) and (21) clearly represents no direct derogation of Wanda herself on Smith’s part. Contrast (20) and (21), with (22) and (23), however. In (22) and (23), the use of “bitch” as part of the complex demonstrative “that bitch Wanda” does express an attitude of derogation on Smith’s part. And notice that it does not matter whether the complex demonstrative occurs in subject or object position. What matters is that it occurs in argument position. That is, embedded arguments seem to be ascriber-​centered, while embedded predicates seem to be ascribee-​centered. Let us return briefly to our bartenders and the martini-​drinking man. Suppose, as above, that Smith intends to report Jones’s belief about the martini-​drinking man in the corner to Black. Suppose that Smith intends via her report to arm Black for interaction with Jones by making it explicit just how Jones thinks of the martini-​drinking man. It is commonly thought that it is via so-​called de dicto ascriptions that we arm each other for interaction with the ascribee. De dicto ascriptions are taken to be sensitive to the inner mental life of the ascriber rather than simply to her outer-​worldly commitments. After all, such ascriptions are supposed to be in the business of somehow characterizing the modes of presentation, notions, or ideas via which the ascriber cognizes the world. But here we are supposing that Jones mistakenly takes a martini-​drinking man to be a gimlet-​drinking woman. Smith is aware that Jones is confused. But Black is not aware of Jones’s confusion. If Smith were to report Jones’s belief by an utterance of (19), she would correctly and successfully ascribe to Jones a commitment to the effect that a certain person has had too much to drink. But since her utterance would convey no information about Jones’s confused notions of the relevant person, it might reasonably be concluded that she would thereby fail to fully arm Black for interacting with Jones, precisely because we have done nothing to specify her own inner perspective by our ascription.

200  Referring to the World There is something to this thought. To fully arm Black for interaction with Jones, Smith needs a way both to ascribe the commitment just mentioned and to convey information about Jones’s confused notions, and she needs to do so without thereby committing herself to Jones’s confusions. Just as in the parallel case of expressive commitments, she can do no better than to go fulsomely de re. She might, for example, utter something like the following: (24) Jones believes of the martini-​drinking man in the corner, whom she mistakes for a gimlet-​drinking woman, that he has had too much to drink.

Smith does several things in uttering (24). She undertakes, and manifestly so, a commitment of her own to the existence of a martini drinking man. She also ascribes to Jones a commitment to the existence of a gimlet-​drinking woman. And she does so without herself thereby undertaking any such commitment. Finally, she ascribes to Jones, also without herself undertaking, a predicative commitment to the effect that a certain person has had too much to drink. In so doing, Smith not only informs Black of Jones’s commitments, but she does so in a manner that arms Black for interaction with Jones. For she explicitly conveys information about Jones’s representations and misrepresentations of the relevant objects. Turn briefly to proper names and what I  call referential commitments. Both Frege’s original case for the Fregean Mechanism, and also his implicit case for the priority of the de dicto over the de re in the ascription of attitudes, turned heavily on the apparent failure of coreferring names to be intersubstitutable in the context of attitude ascriptions. Failures of substitution raise a number of delicate issues, which I address somewhat more fully in the next section.11 But here I prepare the way by showing that the behavior of names within attitude ascriptions is rather more nuanced than Fregeans acknowledge. Even in the case of names, we observe something rather like what we have already observed with other expressions that occupy embedded argument position—​that the commitments foregrounded by the use of embedded names are typically ascriber-​centered rather than ascribee-​ centered. To see this, we examine what I call “reverse Frege cases.”12 In reverse 11 But see Taylor (2014a), Taylor (2003), and Taylor (2002). For something like the ur-​argument that the Fregean diagnosis of substitution failures goes wrong from the very start, see the landmark Kripke (1979). 12 These examples were first considered in Taylor (2002) and expanded upon in Taylor (2007). I now call them “reverse Frege cases.” In “straight” Frege cases, a believer starts out believing that what is in fact the same thing again is two different things. She may later come to correctly believe

Representing Representations  201 Frege cases, because of the referential confusion of the ascribee, the referential commitments of the ascribee and those of the ascriber diverge. We shall see that in these cases the behavior of embedded names precisely mirrors the behavior of embedded definite descriptions and embedded evaluative referring expressions. Imagine Jones, a quite hapless astronomer. Jones proudly fancies herself the first to realize that Mars and Venus are actually one and the same planet. Before her spurious “discovery”, Jones is as linguistically competent as the rest of us. Like the rest of us, pre-​discovery she used “Venus” to refer to Venus and “Mars” to refer to Mars. Her spurious “discovery” commits her to some serious reconfiguration of her conceptions of Mars and Venus. Like anyone who discovers an identity, she reorganizes her concepts and conceptions in various ways. Thanks largely to the fact that she otherwise remains a member in good standing, it is not obvious that we should say that such reconfigurations would ipso facto cause her to no longer to be numbered among the linguistically and conceptionally competent. In fact, in one sense Jones is no worse off either linguistically or conceptually or doxastically after her spurious discovery than someone who believes that Hesperus is distinct from Phosphorous. If a rational and competent cognizer can reasonably but uninformedly take one thing to be two, so too can a rational cognizer reasonably but uninformedly take two distinct things to be one. But if Jones, who suffers from a reverse Frege case, is no worse off than one who is subject to a straight Frege case, then when she makes such seemingly bizarre post-​ discovery statements as: (25) Mars is just Venus again.

she is certainly speaking falsely, but she is nonetheless speaking, and presumably intends to be speaking, English. And unless one is willing to say that knowing that Mars is distinct from Venus is required for full

that the “two” are in fact one. Famously, Frege wonders how such discoveries are possible, given that a statement to the effect that a thing is identical with itself would seem to be trivially true. In reverse Frege cases, things go the other way around. The believer starts out believing, this time correctly, that what are in fact two distinct things are two distinct things. But upon further investigation, she comes to mistakenly believe of the two distinct things that they are one. Reverse Frege cases bear a certain resemblance to Kripke’s (1979) “London”-​“Londres” case and his “Paderewski”-​“Paderewski” cases. My aim in examining reverse Frege is not quite the same as Kripke’s, though. I use such cases to draw a wedge between the referential commitments of the ascriber and those of the ascribee.

202  Referring to the World conceptual or linguistic competence, she is apparently doing so at least somewhat competently. But now suppose that Brown recognizes the nature of Jones’s confusion. And suppose that she wants to inform Black of something about Jones’s beliefs in a situation in which it is common ground between Black and Brown that Mars and Venus are distinct. Perhaps Jones has uttered the following: (26) I see that Venus is visible tonight.

And perhaps she has done so with evident intent of referring to the currently visible Venus rather than to the not yet visible Mars. It seems intuitively right to say that Jones has expressed a belief to the effect that Venus is currently visible. It is, after all, Venus that she sees. Moreover, on this occasion she correctly uses the name “Venus” to refer to the very object that she sees. The problem is that because Jones also takes that very object to be Mars, it also seems right to say—​or at least not wrong to say—​that Jones believes that Mars is visible tonight. After all, Jones would accept both the sentence “Venus is visible tonight” and the sentence “Mars is visible tonight.” Perhaps we can represent what Jones believes by (27): (27) Jones believes that Venus is visible and that Mars is visible.

But (27) is entirely silent about the character of Jones’s confused notions of Mars and Venus. (27) does not capture, and does not even purport to capture, the fact that by Jones’s notional lights Mars and Venus are one and the same planet. Just imagine that Brown does, but Black does not know that Jones takes Mars to be identical to Venus. An utterance of (27) would put Black in no position to infer that Jones takes Mars and Venus to be identical. Here again, Brown might resort to something like the elaborate circumlocution of a fulsomely de re ascription to fully depict the true character of Jones’s confused notions of Mars and Venus, without having to own the relevant confusion as her own, as in: (28) Jones believes of Venus, which she takes to be identical with Mars, that it is visible tonight. (29) Jones believes of Mars, which she takes to be identical with Venus, that it is visible tonight.

Representing Representations  203 One can easily imagine discourse situations in which one might prefer one of (28) or (29) over (27) as a way of reporting Jones’s belief, with the choice between them being driven largely by pragmatic considerations relating to what is foreground or background in the relevant discourse situation. Consider a slightly different scenario. In this scenario, Jones is even more clueless about the planets—​Mars in particular. Sometimes when she sees it, she takes it to be Venus. Other times, she takes it to be Jupiter. Now suppose that on appropriate occasions she utters (30) and then (31), each with the evident intent of referring to Mars: (30) My, how lovely Venus looks this evening. (31) My, how lovely Jupiter looks this evening.

How should we report the belief expressed by Jones? Our procedures so far may suggest (32) and (33) below: (32) Jones believes of Mars, which she takes to be Venus, that it looks lovely this evening. (33) Jones believes of Mars, which she takes to be Jupiter, that it looks lovely this evening.

These do get at something near the truth about Jones’s state of mind. But since Jones sometimes takes Mars to be Venus and sometimes takes it to be Jupiter, one may want to know more. One may want to know whether, in this very episode of believing, Jones is taking Mars to be Venus or is taking it to be Jupiter. This we can capture by expanding our ascriptions as follows: (34) Jones believes of Mars, which in this very episode of believing she takes to be Venus, that it looks lovely this evening. (35) Jones believes of Mars, which in this very episode of believing she takes to be Jupiter, that it looks lovely this evening.

Let us take brief stock of where we are in the overall argument. I began by drawing attention to the seamlessness of everyday transitions from thought and talk about worldly objects and their properties to thought and talk about beliefs about such objects. I  took it to be a mark against broadly Fregean views of ascriptions of attitudes about particulars that they make a prima facie mystery of that seamlessness. They do so by giving semantic pride of

204  Referring to the World place to notions, ideas, mental representations, modes of presentation, or the like in our talk about such beliefs. Especially when this view is married to the belief that there is no automatic way back up from reference to mode of presentation—​since referents will relate to modes in a one-​many fashion on any version of a Fregean approach—​a question is opened about how we manage to know the modes under which a believer cognizes the relevant objects. Now I have not argued that Fregeans have absolutely no resources for dispelling the prima facie mystery generated by their approach. In fact, it was partly by way of acknowledging that many have attempted to resolve the prima facie mystery within a broadly Fregean framework that I distinguished between the Fregean Thesis and the Fregean Mechanism for implementing the Fregean Thesis in the first place. It is the prima facie mystery that has led philosophers like Crimmins (1992), Schiffer (1977), Richard (1993) or Recanati (2010) to reject the Fregean Mechanism, while accepting the Fregean Thesis. Perhaps even Frege himself could be said to be alive to this worry. Perhaps that is why he argued that a determinate sense must be encoded in the meaning of each term, since otherwise it could be argued that his reference-​shifting mechanism would yield no determinate reference for embedded terms to denote. I have not argued that every conceivable mechanism for implementing the Fregean Thesis is bound to fail. What I have done is to take a fresh look at attitude ascriptions. The problem, I claim, lies not with this or that mechanism for implementing the Fregean Thesis, but with the Fregean Thesis itself. Once we take embedded expressions at face value, we not only regain or pre-​Fregean semantic innocence, but we also obviate any motivation to go searching for some non-​Fregean mechanism by which to implement the Fregean Thesis. The Fregean approach is motivated by misleading intuitions, mostly generated by considering attitude ascriptions in communicative isolation. It is as if such ascriptions are uttered by no one and are addressed to no one with no conversational end in view. But it is a mistake to divorce attitude statements from the communicative contexts that gives them point. I have taken some pains to rectify that mistake here.13 When we regard utterances 13 One could carry this line of reasoning further and argue that the real way to study attitude ascriptions is to study linguistic corpora. Harris and Potts (2009) draw just such a conclusion. They say, “We think that the investigative strategy of reporting basic intuitions about individual cases has run its course in this area. More and different evidence is needed. To this end, we present two human-​subjects experiments and some novel corpus work.” Even though I am a philosopher, rather than a linguist, and have not attempted to carry out either a human subject experiment or corpus work, I don’t entirely disagree with that thought. The bottom line is that it is high time that

Representing Representations  205 of propositional attitude statements as communicative acts, uttered against a shared background, with certain communicative ends in view, it becomes abundantly clear that semantic and communicative functions of expressions that occupy embedded argument places are not at all what Fregeans have imagined.14 Contrary to the Fregeans, the function of such expressions is neither in whole nor in part to semantically or pragmatically invoke the ascribee’s ways of cognizing the worldly objects about which she has beliefs. There is no reason whatsoever to believe that expressions either semantically refer to, quantify over, describe, or somehow pragmatically invoke the Fregean modes of presentations, or their cleaned-​up contemporary ilk, by which the ascribee putatively cognizes the worldly objects about which they have beliefs. In the unmarked case, embedded expressions are not ascribee-​centered at all. It should be abundantly clear that in the unmarked case, it is the ascriber and not the ascribee who derogates, and is represented as derogating, by the use of a derogatory referring term such as “that damned Kaplan.” But this is not just a fact about derogatory expressions. It is also true, for example, that something similar holds for the non-​derogatory complex demonstrative “that UCLA philosopher Kaplan.” In addition, it is clearly the ascriber and not the ascribee who, in the unmarked case, undertakes, and is represented as undertaking, an existential commitment in using an embedded definite description such as “the man in the doorway.” Finally, it is the ascriber rather than the ascribee who undertakes, and is represented as undertaking, a referential commitment in the use of an embedded name or complex demonstrative in embedded argument position. Even when Smith misuses “Mars” to refer to Venus, the ascriber cannot make “Mars” stand for Venus, or have the sense of “Mars, aka, Venus,” by embedding that term in a clause that purports to specify the notional contents of Smith’s beliefs Hardly anyone who has studied the behavior of embedded slurring and derogatory expressions has seen fit to deny their non-​displaceability. But the

philosophers of language stop examining the same hackneyed examples and stop considering attitude ascriptions in isolation from the discourse situations within which they are at home. 14 Hawthorne and Manley (2012) are two philosophers who seem to an extent to share this outlook. Their argument in opposition to what they call the spy argument against liberalism about singular thought makes fairly heavy appeal at various points to the conversational dynamics of belief ascriptions.

206  Referring to the World non-​displaceability of such expressions has widely been taken to be a peculiar feature of derogatory expressions, one in need of special explanation. As such, non-​displaceability has been taken to be the exception rather than the rule for embedded expressions and thus the marked case for embedded expressions. But I have been arguing, by means of a case-​by-​case slog through examples, that such behavior is much closer to being the rule rather than the exception. Just as an embedded use of the derogatory complex demonstrative “that damned Kaplan” would express only Jones’s and not Smith’s derogatory attitude toward Kaplan in (36) as uttered by Jones in addressing Black, so an embedded use of the non-​derogatory complex demonstrative “that famous UCLA philosopher Kaplan” would represents Jones’s and not Smith’s knowledge of Kaplan’s place of employment in (37) as uttered by Jones in addressing Black: (36) Smith believes that that damned Kaplan just got tenure. (37) Smith believes that that UCLA philosopher Kaplan just got tenure.

To appreciate that the complex demonstrative in (37) does not represent how the ascribee Smith is thinking of Kaplan but how the ascriber Jones and her interlocutor Black are thinking of Kaplan, imagine Jones producing (37) with the intention of making clear to Black which of two possible Kaplans—​one a UCLA philosopher, the other a Stanford historian—​Smith believes to have gotten tenure. Nor are complex demonstratives—​whether derogatory or not—​special in this regard. Suppose that Smith is hallucinating and mistakenly believes that there is a (unique) man in a (unique) doorway staring at her. In a situation in which it is mutually manifest to Jones and to Black that there is no such man, Jones could not felicitously use the description “the man in the doorway staring at her” as it occurs in (38) to report Smith’s belief to Black, at least not without some further stage setting: (38) Smith believes that the man in the door staring at her is about to jump her.

That’s because the description would misrepresent Jones’s existential commitments rather than correctly representing Smith’s. Notice too that this is not simply a matter of scope. Even if read in a narrow scope way, an utterance of (38) would be infelicitous.

Representing Representations  207 I do not deny that it is possible to shift the focus of an ascription from the ascriber and her interlocutors to the ascribee.15 We can always resort to the elaborate circumlocution of a fulsomely de re ascription, for example, especially when there is a mismatch between the commitments of the ascriber and those of the ascribee. Sometimes such circumlocution will not be necessary if context alone can do the trick. Though I have not dwelled on indefinites in this chapter, it is worth noting in passing that in certain contexts, going indefinite helps to directly shift the focus to the ascribee. For example, instead of uttering the problematic (38) to capture the content of Smith’s hallucinatory belief, Jones might resort to the unproblematic (39): (39) Smith believes that there is a man in a doorway staring at her and that he is about to jump her.

Here it is important that the apparently referring expression “he” does not occupy a discourse initial position. Rather it is anaphoric on an embedded indefinite. And though the indefinite itself is also not discourse initial, it seems to introduce what I call a notional frame.16 And “he” seems to be able to reach into that notional frame and pick up not an actual reference but a notional reference.17 Throughout this section, I  have focused mainly on the ascriber-​ centeredness of singular definite expressions occupying embedded argument places. But it is important not to lose sight of the fact that things are otherwise with embedded predicates. Embedded predications are always centered on the ascribee rather than the ascriber. Nor can they be shifted away from ascribee to ascriber in any conversational context, by any semantic or pragmatic mechanism. It might even be thought that this fact reflects the grain of truth in the Fregean Thesis. Indeed, it is precisely the ascribee-​centeredness of embedded predications that is responsible for the difference in subject matter between statements like (1)  and (2)  early in this chapter. This difference in subject matter follows from the fact that the semantic and communicative function

15 See Harris and Potts (2009) for examples relating to appositives and expressives in particular. They convincingly argue that shifting perspectives from ascriber to ascribee requires much pragmatic stage setting. See also Hom (2008), though Hom’s cases seem more equivocal. 16 Taylor (2002, 2003). See also Chapter 7. 17 Geach (1967) is the locus classicus. See also Guerts (1998), Chierchia (1995), and Hawthorne and Manley (2012).

208  Referring to the World of an embedded predication is to represent the predicative commitments of the ascribee. In using an embedded predicate, the ascriber ascribes such a commitment but does not herself express or undertake any predicative commitment of her own. In this, and this alone, I conjecture, lies the difference between statements about worldly objects and their properties and statements about our beliefs about such objects and their properties. But this gives us no reason to posit reference shifts a la Frege, unarticulated constituents a la Crimmins and Perry, Russellian annotated matrices a la Richard, free-​ enrichment a la Recanati, or any other peculiar semantic or pragmatic mechanism to explain the putatively peculiar behavior of embedded expressions.

3.  Substitution Puzzles Reconsidered In this section I revisit the substitution puzzles. I have claimed that garden variety attitude ascriptions of the sort that we make effortlessly on a daily basis have more in common with so-​called de re ascriptions than so-​called de dicto ascriptions. But if that is right, it may seem to remain something of a mystery why there should be a problem about substituting coreferring expressions one for another within the context of such attitude ascriptions in the first place. De re ascriptions are widely taken to be precisely those ascriptions that permit free substitution, while de dicto ascriptions are widely taken to be ascriptions that do not permit free substitution. Hence if ordinary attitude ascriptions typically do not permit free substitution, is this not grounds for thinking that they are de dicto and opaque rather than de re and transparent? I hope it is clear by now that this conclusion is far too hasty. I have already argued at several points throughout this book that even in supposedly “transparent” contexts, substitution of coreferring names is insufficient, on its own, to make the preservation of subject matter manifest, especially when the relevant names are merely coincidentally coreferential rather than explicitly coreferential. Even in supposedly transparent contexts, whenever subject matter is preserved, but not manifestly so, an imputation of distinctness of subject matter is typically generated. That is all that is going on, for example, in the sorts of cases first made famous by Jennifer Saul. Saul noticed, recall, that there are circumstances in which one would be prepared to assert (40) but not prepared to assert (41), despite the coreference of “Clark Kent” and “Superman.”

Representing Representations  209 (40) Clark Kent entered the phone booth and Superman came out. (41) Clark Kent entered the phone booth and Clark Kent came out.

To one reared in the Fregean tradition, it will be quite puzzling why this should be so, especially given that no sort of embedding is at issue here that might be thought to induce opacity and non-​innocence with it. But the correct lesson to be drawn from such examples is not the Fregean one at all. The correct lesson is that all cooperative conversation, whether involving attitude statements with embedded terms or no such statements, is governed by a (defeasible) directive constraining discourse participants to make the preservation of subject matter manifest. Such a constraint already predicts, without appeal to reference shifts and the like, that despite the coincidental coreference of, for example, “Hesperus” and “Phosphorous” or “Clark Kent” and “Superman” they cannot, in general, be substituted one for the other in what we might call a “dialectical significance preserving” manner, where dialectical significance has to do with significance for the stage-​wise evolution of a discourse, argument, or conversation. Now to say that substitution of coincidental coreferents fails to preserve dialectical significance is not to say that such substitution fails to preserve truth value. Preservation of truth value is one thing; manifest preservation of subject matter is something entirely different. Admittedly propositional attitude contexts constitute a still further challenge. That’s because in many so-​called transparent contexts, substitution of coincidentally coreferential expressions can be directly licensed when an identity sign is used to make manifest the coincidental coreference of two referentially independent designators. In such contexts, substitutions that would otherwise be dialectically problematic are rendered dialectically unproblematic. But this is not so with propositional attitude statements. In the context of such statements, even when referentially independent and coincidentally coreferential designators are linked via an explicit identity sign, substitution may still fail to preserve dialectical significance. For example, the inference from (42) and (43) to (44) following is not intuitively compelling, despite the fact that “Hesperus” and “Phosphorus” are linked via an explicit identity sign in (43): (42) Smith believes that Hesperus is rising (43) Hesperus is Phosphorus (44) Smith believes that Phosphorus is rising

210  Referring to the World If we want to resist the widespread, Frege-​inspired conventional wisdom that substitution failures show that embedding somehow makes the semantics of referring non-​innocent, we must provide an alternative explanation of the source and nature of substitution failures within attitude contexts and we must show that these failures are not grounded in semantics at all. It is on that task that I focus in this section. The key to my approach is rejecting another widely shared bit of conventional wisdom—​that our intuitions of the badness of such inferences are at bottom intuitions about truth-​value dependence and independence.18 I  shall argue that they are not. Such inferences really involve a kind of pragmatic infelicity; I shall argue that is ultimately traceable to the influence of what I call the default coreference constraint on propositional attitude ascriptions. I begin by introducing the notion of the coreference set of a given term for a given agent at a given time. If a is an agent, and n is a name in a’s lexicon at t, the coreference set of n for a at t is the set of expressions in a’s lexicon such that either (a) t is explicitly coreferential with n, or (b) if t is referentially independent of n, then t is in the coreference set of n for a just in case a accepts the sentence t = n. When two referentially independent expressions m and n are such that a accepts m = n at t, I will say that m and n are in-​the-​head-​ coreferential for a at t. In-​the-​head coreference is distinct from real-​world coreference. Expressions may be real-​world coreferential, without being in-​ the-​head coreferential. Moreover, expressions may be in-​the-​head coreferential, without being real-​world coreferential. Finally, it is important to stress that coreference sets are defined agent by agent and moment by moment. In particular, two referring expressions may be in-​the-​head coreferential for a given agent at a given time, but not in-​the-​head coreferential for either the same agent or some distinct agent at a distinct time. 18 Soames (1985, 1987a, 1987b, 2001), Salmon (1986, 1989a, 1989b, 1995) and Braun (1998) are the most dogged exceptions. Soames and Salmon apparently believe that ordinary speakers themselves mistake what are really intuitions about pragmatic infelicity for intuitions about truth and falsity. Braun apparently holds that although speakers do have truth value intuitions those intuitions are mistaken. My aim here is to offer a re-​diagnosis of our intuitions as intuitions about dialectical dependence/​independence. To that extent, I agree with Soames and Salmon and disagree with Braun. On the other hand, although I do think that philosophers of language have by and large misdiagnosed our intuitions, I do not claim to know how “the folk” understand those intuitions. Nor do I think it matters. I am trying to understand the intuitions themselves, and what drives them. I am not trying to explain the folk theory of those intuitions and feel no compulsion to respect that folk theory, whatever it is. The bedrock data before us is that the folk find some ascriptions acceptable in certain contexts, while finding others unacceptable. I am after a theory that explains such patterns of acceptance, not a theory which explains the folk explanation of those patterns of acceptance. Explaining the folk explanations might indeed be a very good thing to do. But it’s not the same thing as explaining the bedrock data themselves.

Representing Representations  211 Notice that in-​the-​head coreference is defined in terms of acceptance of identity sentences. Despite the intimate connection between acceptance and belief, acceptance qua attitude toward a sentence must be sharply distinguished from belief qua toward the proposition expressed by that very sentence. To believe the proposition expressed by a sentence is not ipso facto to accept that sentence. One who has no knowledge of English and its sentences can believe the proposition expressed by the English sentence “The cat is on the mat” even though she fails to accept that very sentence. Conversely, to accept a sentence is not ipso facto to believe the proposition expressed by that sentence. One can accept a sentence even if one does not know which proposition the sentence expresses. Now acceptance is itself perhaps a kind of belief. To accept a sentence S is to believe of S that it expresses some true proposition or other. But one can believe of a sentence that it expresses some true proposition or other, and thereby accept it, without knowing which proposition that sentence expresses. Suppose, for example, that Brown does not know who Smith and Jones are. Suppose further that Black utters the sentence “Smith loves Jones.” Assume that Brown takes Black at her word. She thereby comes to believe of the sentence “Smith loves Jones” that it expresses a truth, but she does not thereby come to believe that Smith loves Jones. To be sure, it can be a short step from accepting a sentence to believing if not the proposition expressed by that sentence, then various other propositions that are related to that sentence. If, for example, Brown merely recognizes some further grammatical and lexical facts about the sentence “Smith loves Jones” and its constituents, then even if she does not know who Smith and Jones are, there may be further propositions, closely connected to the accepted sentence, that Brown does comes to believe in coming to accept the sentence “Smith loves Jones.” If, for example, she recognizes that “Smith” and “Jones” are names and knows the meaning of “loves,” then in coming to accept this sentence, she thereby comes to believe the further and more articulated proposition that the referent of “Smith” loves the referent of “Jones.” That again, though, is still not the proposition expressed by “Smith loves Jones.”19 We might say that Black believes that Smith loves Jones via acceptance of the sentence “Smith loves Jones” if she accepts “Smith loves Jones” and knows which proposition it expresses. But we will not attempt to spell out at present just what it takes to know what proposition a sentence expresses. I take it to be a plausible (initial) hypothesis about the connection

19 It is, rather, one form of what Perry (2001) calls reflexive content.

212  Referring to the World between belief and acceptance for creatures like us that if P is a proposition such that a (explicitly) believes that P, there is some sentence S such that a believes P via acceptance of S.20 Armed with the notion of a coreference set, we can give an initial statement of the default coreference constraint on (one-​layer) belief ascriptions: Default Coreference Constraint: If a sentence of the form: A believes that . . . n . . .

is dialectically permissible for a player p in a dialectical setting D at t and m is in the coreference set of n for a at t, then a sentence of the form: a believes that . . . m . . . is dialectically permissible for p in D at t.

S is dialectically permissible for a player in the dialectical setting D at t, roughly, if given the common ground of D at t, the production of S by p in D at t would violate no norms of cooperativeness, clarity, coherence, relevance, or the like which jointly govern the players in D at t.21 The coreference constraint says, in effect, that (one-​layer) belief ascriptions are defeasibly dialectically sensitive to facts about ascribee coreference sets, rather than to facts about either ascriber coreference sets or even to facts about real world coreference. The fact that belief ascriptions are defeasibly sensitive to ascribee coreference sets explains why it is not in general dialectically permissible to move from (42) and (43) to (44). When attitude ascriptions are sensitive to

20 Perry (1980) offers an account of the relationship between belief and acceptance very much in the spirit of the account offered here. 21 Following Grice, I distinguish violating norms of cooperativeness, perspicuousness, coherence, and relevance from the flouting of such norms. Though one who flouts a norm may give the appearance of violating that norm, flouting a norm is not the same as either surreptitiously violating it or openly opting out of it. Flouting is also different from situations in which one must violate one or the other of two conflicting norms. Flouting a norm, according to Grice, is something that one does blatantly, with no intent to mislead, and where there is no apparent clash of conflicting norms. In so doing, one puts one’s audience in the position of having to reconcile the open appearance of a violation with the assumption that one is, in fact, respecting the relevant norm. Conversational implicatures are generated, according to Grice, by the attempted reconciliation of what is explicitly said, in apparent violation of the norms of cooperativeness, with the assumption that the cooperative principle is in fact being observed. I am not offering an alternative analysis of how conversational implicatures are generated here. My point is merely to stress that semantically equivalent expressions need not be dialectically equivalent—​perhaps because the utterance of one may, in a given context, generate a conversational implicature that an utterance of the other would not generate.

Representing Representations  213 facts about ascribee coreference sets, such ascriptions exhibit many of the hallmarks commonly associated with so-​called de dicto ascriptions. There are, however, dialectical settings in which the default sensitivity of (one layer) ascriptions to facts about ascribee coreference sets is overridden in favor of sensitivity to facts about the coreference sets that are elements of the common ground between speaker and hearer. In such dialectical settings, attitude ascriptions will exhibit many of the hallmarks of what are commonly called de re ascriptions. In such dialectical settings, whenever  m = n

is part of the common ground of D at t and a believes that . . . m . . .

is dialectically permissible in D at t, then a believes that . . . n . . .

is dialectically permissible in D at t. For illustrative purposes, consider the following scenario. Daniel Taylor, formerly a practicing Christian, decides to convert to Islam. In the course of his conversion, he adopts “Haazim Abdullah” as his legal name. Because he suspects that his devoutly Christian parents, Sam and Seretha, would be distressed by this turn of events, he informs them of neither his change of faith nor his change of name. He does, however, confide in his siblings, Robert and Diane, that he has changed his name, that he has converted to Islam, and that Sam and Seretha are unaware of his conversion. It is mutually manifest to Robert and Diane that they, but not Sam and Seretha, accept the following identity: (45) Haazim Abdullah = Daniel Taylor.

Some time goes by. Diane wishes to inform Robert that Seretha has still not figured out that Daniel—​that is, Haazim—​is no longer a practicing Christian. It is common ground between Diane and Robert that (a)  (45) holds; (b) that Seretha does not accept (45); and (c) that she does not accept (45) because there is no name N in Seretha’s private lexicon such that “Haazim Abdullah” belongs to the coreference set of N for Seretha. In the

214  Referring to the World imagined dialectical setting, the inference from (46) below to (47) seems perfectly acceptable: (46) Seretha believes that Daniel is still a Christian (47) Seretha believes that Haazim is still a Christian.

Because it is part of the common ground that “Haazim Abdullah” belongs to the coreference set of no name with which Seretha is competent, in the relevant dialectical setting, the inference from (46) to (47) generates no imputation that Seretha accepts (45). Because of what is common ground between Robert and Diane, an imputation that might otherwise be generated is simply forestalled. Similarly, because of the common ground of the relevant dialectical setting, the inference from (46) to (47) generates no imputation to the effect that (48) below is true: (48) Seretha accepts “Haazim is still a Christian.”

If I am right about the potential of facts about common ground coreference relations to forestall imputations of acceptance that might otherwise be generated, then there is a quite natural sense in which an occurrence of (47) in the sort of dialectical setting we have been imagining can be said to be dialectically governed by facts about common ground coreference sets rather than by facts about ascribee coreference sets. Consider a dialectical setting in which the default sensitivity to ascribee coreference sets is not overridden by any elements of the relevant common ground. Suppose that Seretha learns, by listening to the news on the radio, of the artistic achievements of one Haazim Abdullah, an Islamic poet of some renown. Suppose she does so without also coming to accept (45). Indeed, suppose that Seretha would explicitly reject (45). And suppose that Robert and Diane mutually know that Seretha has learned of Haazim Abdullah’s poetic achievements and that she has done so in a way that does not lead her to accept (45). Now consider the following ascriptions as they occur in a dialectical setting with a common ground of the sort just described: (49) Seretha believes that Haazim is a very fine poet (49’) Seretha believes that Haazim is not a very fine poet (50) Seretha believes that Daniel is not a very fine poet (50’) Seretha believes that Daniel is a very fine poet.

Representing Representations  215 In such a dialectical setting, (49) and (50) can be simultaneously dialectically permissible, while both (49') and (50') are dialectically impermissible, despite the fact that it is common ground between Robert and Diane that Daniel Taylor is Haazim Abdullah. In such a dialectical setting, I claim, an occurrence of (49) would generate an imputation to the effect that Seretha accepts “Haazim is a very fine poet” and an occurrence of (50) would generate an imputation that Seretha would accept “Daniel is not a very fine poet.” Since Seretha does accept the relevant sentences, (49) and (50) are unproblematic. On the other hand, an occurrence of (49') would generate the unacceptable imputation that Seretha would accept “Haazim is not a very fine poet.” And similarly for (50'). Hence neither (49') nor (50') is dialectically permissible. This is so because the relevant ascriptions are naturally interpreted as being dialectically governed by facts about ascribee coreference sets rather than by facts about common ground coreference sets. The illustrative dialectical scenarios just considered are not intended to represent the entire story about the pragmatics of attitude ascriptions. In particular, I have offered neither a deep explanation of just why the default coreference constraint is the apparent default nor a systematic account of which factors may serve, in particular discourse contexts, to override the default constraint. Nor will I attempt to do so here, since doing so would carry us very far afield.22 When the arguments of the current section are taken in conjunction with the arguments of the previous section, I do claim to have established that there is little, if any, reason to suppose that the factors governing the pragmatics of attitude ascriptions have little if anything to do with either the semantic peculiarities of embedded referring expressions or the semantic peculiarities of embedding contexts. They have more to do with the sorts of communicative acts we are performing in making attitude ascriptions in the first place. Thinking about what we are doing when we perform certain sorts of communicative acts lands us squarely in the domain of pragmatics rather than the domain of pure semantics. On my view, philosophers of language—​ especially those in the grip of the Fregean Thesis, including those who reject the Fregean mechanism for implementing the Fregean Thesis—​have not paid nearly enough attention to such pragmatic considerations in their study of attitude ascriptions.

22 See Taylor (2002a), (2002b), and Taylor (forthcoming-​a) for further discussion of attitude reports.

216  Referring to the World Though I do not aim to give a fully exhaustive and systematic account of the pragmatics of attitude ascriptions here, I cannot resist the temptation to close this section by noting that while the defeasible coreference constraint seems prima facie limited to singly embedded belief ascriptions and not to ascriptions involving multiple embeddings, there is, perhaps, a natural way to generalize that constraint, though with certain modifications, to ascriptions involving multiple levels of embedding. Begin by noting that the inference from (51) and (52) to (53) is not intuitively pragmatically compelling: (51) John believes that Mary believes that Superman can fly (52) John accepts “Superman = Clark Kent.” _​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​_​ (53) John believes that Mary believes that Clark Kent can fly.

It is worth considering just why this is so. Consider some further data. Imagine a dialectical setting in which Smith intends by an utterance of (51) to inform Jones of John’s beliefs about Mary’s beliefs about the abilities of Superman. Suppose that Smith and Jones mutually accept the sentence “Clark Kent = Superman,” and that, moreover, it is common ground between them that John accepts it as well. Even Smith and Jones’s mutual acceptance of “Clark Kent = Superman,” together with their mutual knowledge of John’s acceptance of (52), seems intuitively insufficient to render (53) an acceptable way for Smith to report to Jones John’s belief about Mary’s belief about Superman. But notice that (53) feels more acceptable when (54) below is common ground between Smith and Jones: (54) John believes that Mary accepts “Clark Kent = Superman.”

These data suggest that the dialectical permissibility of the inner-​most clause of a doubly embedded ascription is (defeasibly) sensitive neither to facts about common ground coreference sets nor to facts about the ascribee’s coreference sets, but to facts about the ascribee’s beliefs about coreference sets for what we might call the embedded ascribee. Consider a scenario in which facts about coreference sets for the embedded ascribee are common ground between the ascriber and the addressee. Suppose, for example, (55) below is, but (54) above is not, part of the common ground between Smith and Jones:

Representing Representations  217 (55) Mary accepts “Clark Kent = Superman.”

Where (55) is common ground between Smith and Jones, the default coreference constraint does seem to license a move from (56) as uttered by Smith to (57) as uttered by either Smith or Jones below: (56) Mary believes that Superman can fly. (57) Mary believes that Clark Kent can fly.

However, the standing of (55) as common ground between Smith and Jones still does not, it seems, render (53) permissible in light of (51) (and (52)). This fact lends additional weight to the hypothesis that it is facts about the ascribee’s beliefs about the coreference sets of the embedded ascriber that govern, perhaps defeasibly, the dialectical permissibility of the innermost clauses of an ascription containing a double embedding. Our hypothetical nested coreference constraint, as we might call it, entails that where n = m, discourse participants defeasibly lack entitlement to move from the utterance of an instance of the scheme (58) below to the utterance of an instance of the scheme (59) below, merely on the basis of facts about coreference sets for the embedded ascribee: (58) a believes that S believes that F(n) (59) a believes that S believes that F(m).

Because the ascriber need have no direct access to whatever governing constraints were operative in the “original” discourse situation, if there was one, or in a counterfactual discourse situation if no actual ascription has occurred, it is not at all surprising that something like the embedded coreference constraint should hold. In ascribing to another a belief about another’s beliefs, it is surely intuitively plausible that we should aim to be responsive to the words the outer ascribee would or did use to make her ascription to the embedded ascribee. If the outer ascribee is ignorant of facts about coreference sets for the embedded ascribee, there will be many discourse situations in which the ascriber’s ascriptions will be constrained to reflect the state of the outer ascribee’s ignorance. There may also be discourse situations in which a certain indifference to the constraints to which the outer ascribee was or would have been subject is called for.

218  Referring to the World Notice that the dialectical permissibility of the outermost that-​clause of an ascription containing a double embedding is itself (defeasibly) governed by facts about ascribee coreference sets. Suppose, for example, that Mary is sometimes called by the nickname “Cookie” and that it is common ground between Smith and Jones that John accepts the sentence “Mary is Cookie.” Imagine that Smith reports John’s beliefs about Mary’s beliefs to Jones via (51). In that case, (51) together with the relevant common ground seems sufficient to license: (58) John believes that Cookie believes that Superman can fly.

If, in addition, (54) is part of the common ground between Jones and Smith, then (59) below will be acceptable as well: (59) John believes that Cookie believes that Clark Kent can fly.

The provisional take-​home lesson is that coreference constraints, of one sort or another, govern even ascriptions containing multiple embeddings. The evidence we have so far considered suggests the hypothesis that for each level of embedding, a new but entirely predictable coreference constraint is defeasibly operative. For a singly embedded ascription, the default coreference constraint holds. For a doubly embedded ascription, the default coreference constraint applies to the outermost that-​clause, while the embedded coreference constraint applies to the innermost one. If I am right, then far from disconfirming the defeasible coreference constraint for ascriptions involving a single level of embedding, evidence from cases involving double embeddings lend additional credence to that hypothesis. Indeed, the default coreference constraint would seem to be the limiting case of an initially plausible, though perhaps unexpected, general hypothesis for ascriptions containing n embeddings, for arbitrary n.

4.  From De Re Ascriptions to De Re Attitudes In this section, I have mostly focused on ascriptions of beliefs, rather than on beliefs themselves. That is not entirely accidental. Some philosophers see the distinction between the de re and the de dicto as a distinction at the level of belief; but on my view, that distinction is best understood as a distinction

Representing Representations  219 at the level of ascriptions. Indeed, I have argued at length elsewhere, though I will not stop to repeat those arguments here, that a de re ascription and a corresponding de dicto ascription of a belief are often just distinct ways of partially characterizing the same total doxastic state of the ascribee.23 That is, the same total doxastic state of a cognizer may be partially characterizable by both a de re ascription and a de dicto ascription. This means that de re and de dicto ascriptions do not necessarily correspond to different kinds of beliefs with two different kinds of objects. It is just that a given doxastic state is sometimes best characterized in a de dicto manner and sometimes best characterized in a de re manner, where “best” is measured solely by our explanatory, evaluative, or communicative purposes. For example, in Taylor (2002) I show that even where a de dicto characterization of a believer’s total doxastic state is apt, such characterizations typically only partially characterize a doxastic reality that can be more fully and informatively characterized via a fulsomely de re ascription. Even if I am right, though, it does not follow that we can entirely escape worries about the probity of so-​called de re beliefs. Even granting that the de re/​de dicto distinction is a distinction best applied to ascriptions rather than to beliefs, standard philosophical worries about the coherence of the notion of a de re belief can be restated as worries about which, if any, aspects of a total doxastic state would suffice to render a de re or de dicto ascription true. There is, of course, a long tradition of doubting the coherence of the notion of a de re belief. Though Quine (1956) was perhaps the first to explicitly argue that no coherent sense can be made of the idea of a de re belief, he is not alone in his skepticism. Kaplan (2013) has claimed, for example, that “there is no natural, primitive and pure” notion of de re belief. And examples of such skepticism about the conceptual probity of the de re could easily be multiplied. The supposed problem with de re beliefs stem, I think, from what I have called the epistemic one-​sidedness of reference and objective representational content more generally throughout this book. All believing is no doubt mediated by representations in the sense that we do not relate to the objective representational contents of our beliefs directly but by entertaining representations that somehow encode or express such contents. Those mediating representations are one-​sided in the sense that a thinker can have two representations of the same object, without realizing they are 23 Bach (2010) makes a similar point. As he puts it, “The form of a belief report does not determine the type of belief being reported.” (45)

220  Referring to the World of the same object. It was partly to account for this one-​sidedness that Frege introduced the distinction between sense and reference. I have made clear at a number of different places in my overall argument that Frege—​mistakenly, on my view—​took one-​sidedness to be a feature of the very contents of our thoughts. That is why he argued that thought content was built out of senses rather than the references that sense determines. And it has seemed to many that de re belief must be the paradigm of beliefs that would not exhibit any such one-​sidedness, were they to exist. Witness Kaplan’s claim that there can be de re belief only if we are able to make “perfectly good sense of the claim that George IV has a belief about Sir Walter Scott independently of the way in which he is represented to George.” (Kaplan 2003, 2013) This is precisely a way of saying that if there were such things as de re beliefs, they would not exhibit one-​sidedness. Now it was precisely because he could not make sense of the idea of belief that was not in any sense one-​sided that Quine threw up his hands and gave up on the very idea of de re belief. To his credit, Kaplan (1969) did not follow Quine in throwing up his hands. But he did attempt a sort of rescue operation of de re belief, modeled on certain Russellian insights. De re beliefs could be rescued from the abyss of incoherence, Kaplan claimed, only by subjecting them to certain very stringent epistemic standards. One can have a de re belief about an object, he suggests, if and only if one can manage to get oneself in very close cognitive contact with the object. One has to achieve, as he put it, a certain degree of cognitive rapport with that object and thereby have the capacity to wield a very special sort of name for that object. According to Kaplan, one is en rapport with an object, roughly, if one has the sort of cognitive commerce with the object that renders (one’s use of) a name of that object vivid, where vividness has roughly to do with the fulsomeness (and accuracy?) of the descriptive contents one associates with the relevant name, and ofness has to do with the object playing the right sort of role in the genesis of an agent’s use of the relevant name. The idea seemed to be that if one is to have a bona fide de re belief about an object, one has to be able to cognize it as the same again, relatively independently of the way in which the object is presented. This notion of rapport naturally brings Russell’s notion of acquaintance to mind. Russell, recall, argued that one could directly refer to only those objects with which one is directly acquainted. Though Kaplan’s notion of rapport isn’t supposed to be quite so epistemically demanding as acquaintance, it is clearly intended to be a non-​trivial form of cognitive contact.

Representing Representations  221 Few contemporary philosophers share Quine’s utter despair about the intelligibility of the de re. What was once called the “new” theory of reference—​ an approach to which many contemporary philosophers of language are by now committed to varying degrees—​would seem to require that we accept rather than reject the intelligibility of de re cognitions of objects. Perhaps most proponents of that approach are convinced that some version or other of Kaplan’s attempted rescue of the de re, perhaps with epistemic standards watered down a bit to involve something short of acquaintance or rapport, will do the trick.24 But it is striking that Kaplan himself—​who is, after all, one of the founding fathers of the once new theory of reference—​still worries, to this very day, that his rescue operation remains incomplete. Indeed, he now seems close to an almost Quinean despair about the coherence of the de re. Deference to the wisdom and worries of our forefathers is often a wise course. But in this case, I suspect that the philosophical tradition that began with Frege has vastly overestimated the problems of the de re.25 And I suspect that the supposed problem to which Russellian acquaintance, Kaplanian rapport, and similarly stringent epistemic standards are all supposed to provide a solution isn’t, on my view, a problem at all. In particular, by marshalling the resources of this book and the two-​factor referentialism that it defends, I shall argue that having a de re belief about an object is less a matter of the tightness of our cognitive grip on that object than a matter of deploying a certain form of inner representation that is anchored to a really existing object. I start by clarifying the question to which the story I sketch is meant to provide a partial answer. I grant that a good theory of de re belief should answer the question of how objects as such become what I call de re thinkable. The problem of de re thinkability is the problem of explaining how our beliefs achieve a certain answerability to how things are by the objects themselves. When an object is de re thinkable, how things are by our beliefs in the way of truth or falsity depends “directly,” as we might say, on how things are by that object and its properties. By saying that the truth or falsity of a de re belief about an object depends “directly” on how things are by that object, I mean to point to the fact that our notions and conceptions of the objects are, 24 See, for example, Recanati (2010) for a defense of a weakened epistemic standard on de re belief. 25 I am not alone in the estimation that acquaintance theorists of de re belief have overestimated the epistemic hurdles in the way of de re belief. Though some sort of acquaintance condition is perhaps still the dominant view of what it takes to have a de re cognition, a growing minority of philosophers seem to imagine the possibility of de re belief without acquaintance. See Brandom (1994), Jeshion (2002), Jeshion (2010b), Bach (2010), Hawthorne and Manley (2012), Taylor (2010), and Crane (2013).

222  Referring to the World in one sense, irrelevant to the truth conditional contents of such beliefs. In episodes of de re believing, we undertake singular predicative commitments, commitments to the effect that a certain object has certain property (or that a tuple of objects stands in a certain relation). So, for example, in believing that Donald Trump is Putin’s favorite puppet, one undertakes a singular predicative commitment to the effect that a certain object, viz., Donald Trump, has a certain property, viz., the property of being the most favored of Putin’s puppets. And whether that belief is true or false depends entirely on how things are by Trump, in and of himself, and not at all on how one thinks about Trump. Whether one cognizes Trump as the greatest or worst American president ever is entirely irrelevant to determining whether one’s belief that Trump is Putin’s favorite puppet is true. To view a belief state as an undertaking of a commitment is to view belief states in quasi-​ normative terms. It is because belief states involve undertakings of commitments that such states and the cognizing-​agents whose states they are, are liable to normative assessments of various sorts—​ as, for example, true or false, rational or irrational. To be sure, if the sort of naturalism that I have undertaken to defend elsewhere is true, as I think it is, then there must also be a descriptive psychological story to tell about how belief states are partly constituted by a characteristic causal play of inner representations. Moreover, the psychological story about belief as involving a causal play of inner representations, and the quasi-​normative story of beliefs as undertakings of commitments, must ultimately be made to mesh somehow or other. Still, I  insist, despite my own naturalistic proclivities, that nothing but confusion and error results if we move too quickly from the quasi-​normative story about commitments undertaken to the causal cum psychological story about the play of inner representations. Indeed, premature transition from the normative to the psychological, or the psychological to the normative, is one path among others that may lead to the despairing but mistaken conclusion that de re belief is a secondary, more problematic, less natural form of believing than de dicto or notional or fully general belief. Understanding beliefs themselves as undertakings of commitments has consequences for our understanding of the communicative function of belief ascriptions. Belief ascriptions are, after all, assertions of a certain kind. And qua assertions, belief ascriptions have a communicative function, just as all assertions do. In effect, I have been arguing all along, throughout this chapter, that when we execute a gestalt shift and view belief ascriptions as constituting a distinguished class of communicative acts characterized by

Representing Representations  223 a distinguished communicative function, we see that their communicative function is not to mark inner pushes and pulls and certainly not solely and probably not primarily. Now it was once and perhaps still is widely believed that the primary job of belief ascriptions is to serve the aims of common-​ sense psychological explanations. Indeed, it because of the role of common-​ sense psychological explanations that many philosophers have taken de dicto ascriptions to be the paradigmatic form of a belief ascription. I do not mean to deny that ascriptions sometimes do play a role in common-​sense psychological explanations. Jones notices that Smith seems to be searching for something in the back yard. Black wonders why Smith is engaged in that behavior. Jones responds, “Smith is looking for the cat and believes that the cat is in the back yard.” That is perhaps an example common-​sense psychological explanation at its best. But it is seldom noticed by philosophers that we quite often perform attitude assertions not in service of common-​sense psychological explanation but in service of, say, increasing our knowledge. Black is not sure where the cat is. She thinks Jones may know. And so, she asks Jones. Jones herself has no direct knowledge of the cat’s whereabouts, but she knows what Smith thinks. She responds, “Smith says that the cat is in the backyard somewhere.” Black has thereby gained new knowledge of the potential whereabouts of the cat. But notice that in order to gain this new knowledge, it is of little relevance exactly how Smith was thinking of the cat, that is, via which one-​sided mode of presentation she was able to cognize the cat. Or consider a case of a different kind. We sometimes make assertions about what another believes not because we wish to gain or transmit knowledge, but because we want to raise doubts, outright refute them, or in some other way challenge their doxastic authority. “Ladies and gentlemen of the jury,” the prosecutor says, “the defense would have you believe that the defendant could not possibly have committed this crime, but the prosecution will show that the defendant’s protestations of innocence are not to be believed.” The point is just that attitude ascriptions play many different roles in our communicative lives. What binds all the roles together is that in making an assertion about what another believes, the fundamental thing that we are doing is ascribing a commitment to the ascribee. And we are doing so without necessarily endorsing or undertaking that commitment ourselves. Indeed, we sometimes take pains to distance ourselves from the ascribed commitments. And that is precisely what the circumlocution of fulsomely de re ascriptions enables us to do. Moreover, it is precisely the possibility of such a gap between commitments undertaken and commitments ascribed that makes de

224  Referring to the World re ascriptions to be not a secondary or problematic form of ascription, but the central and dominant form of ascription. The primary communicative function of belief ascriptions is to ascribe commitments undertaken while allowing for a certain distance between what is ascribed and what is undertaken, rather than to track inner pushes and pulls; this function tends to be obscured by the fact that sameness and difference of commitments undertaken does not map neatly onto sameness and differences of inner causal push and pull. And I suspect that failure to fully appreciate the significance of the difference between ascribing commitments and tracking inner pushes and pulls is one source of the tempting but mistaken conclusion that de re belief is therefore a secondary, more problematic, less natural notion of belief. This is so for at least three reasons. First, one who undertakes a commitment may fail to live up to the rational consequences or requirements of that undertaking. One may be rationally committed to adopting a certain further belief, given one’s other beliefs, and yet fail to adopt the relevant belief. The failure to undertake what one is rationally committed to believing may be thought to involve less than perfect rationality. But imperfect rationality is an ever-​present psychological reality for creatures like ourselves. And less than perfect rationality means that there will be no straightforward mapping from sameness or difference of commitments undertaken onto the causal to-​and-​fro of inner representations. And the same is true the other way around. A second, perhaps more fundamental, reason that sameness and difference of commitment undertaken does not map neatly onto the causal to-​ and-​fro of our representations rests on a fact deeply rooted in the epistemic one-​sidedness of all mental representations. Because of the epistemic one-​ sidedness of all mental representations, even a fully rational cognizer may undertake metaphysically conflicting or incompatible commitments, as I argued at some length in Chapter 5. We have already seen, for example, that without loss of rationality, Smith may simultaneously believe that Hesperus is rising, while also believing that Phosphorus is not rising. But this implies that if believing is a matter of staking out a commitment about an object as such, then Smith will have undertaken simultaneous commitments to one and the same object both having and (possibly) lacking one and the same property. Since there is no metaphysically possible world in which one and the same object can both have and lack the same property, there is no metaphysically possible world in which Smith’s commitments can be simultaneously made

Representing Representations  225 good. The fact that there is no such world would seem to reflect an incoherence of a deep, but hard to avoid kind on Smith’s part. Starting with Frege himself, there have always been those who take the very possibility that a rational cognizer can simultaneously both believe that Hesperus is rising, and either disbelieve that Phosphorus is rising or suspend judgment about the truth of that proposition, as sufficient reason to distinguish the potential thought content that Hesperus is rising from the potential thought content that Phosphorus is rising. In making such a distinction, even while perhaps conceding that the rising of Hesperus just is, and is as a matter of metaphysical necessity, the rising of Phosphorus again, such thinkers tacitly endorse a distinction between what we might call “worldly,” “metaphysical,” “referential,” or “wide content,” on the one hand, and what we might call “rational,” “epistemic,” “notional,” or “narrow” content, on the other. To a first approximation, the worldly or referential content of a belief is a matter of what predicative commitments are undertaken with respect to which actual existents are in the world. Rational or notional content, on the other hand, is a matter only of how things are by the cognizing subject’s own inner lights. Rational or notional content does satisfy, but metaphysical or referential content need not satisfy the following difference principle: If a rational cognizer simultaneously believes thought content C, and either disbelieves or suspends judgment about C' or believes not C', then C and C' are distinct thought contents.

Many thinkers, but most especially those who have been deeply influenced by Frege, take something akin to rational or notional content to be prior to or more fundamental than metaphysical or referential content. And this assessment of the priorities naturally leads many to believe in the priority of the de dicto over the de re. And this is so whether we are talking at the level of ascriptions or talking at the level of belief itself. Rational or notional contents are embraced as intrinsic and causally relevant. Metaphysical contents are dismissed as extrinsic and epiphenomenal. Rational belief contents have been thought to stand between the believer, on the one side, and the objects that the believer somehow indirectly cognizes via them, on the other. On this picture, it is as if the believer manages to have de re beliefs about the objects only by having de dicto beliefs not intrinsically and directly bound up with the objects. It is as if rational contents are not directly constituted out of the objects but at best out of the cognizing subject’s means of apprehending the

226  Referring to the World objects. 26 It is just such a conviction that motivates Kaplan’s worry that there may be no primitive, natural, or pure notion of de re belief. This Frege-​inspired approach to belief content is both venerable and ancient. But it seems to me to have gotten its priorities mostly wrong. Referential or worldly content is not in any sense posterior to rational content. Indeed, there are good reasons for doubting the very existence of an inner realm of intrinsic rational contents that somehow intervene between the cognizer and the objects with respect to which she undertakes predicative commitments. Certainly, nothing in our ordinary practices of ascribing beliefs suggests that we should see beliefs that way. Indeed, I have been at pains in this chapter to show that our cognitive hold on the inner world of notions, ideas, mental representations, modes of presentation, and the like is itself derivative of and dependent upon our ability to think and talk about the outer world of mind-​ independent objects. We cognize and represent inner mental contents by cognizing our thoughts in relation to the external world. The mind opens its inner representations to view by opening itself up to a largely shared, largely mind-​independent external world on which we “collectively triangulate” to use Davidson’s (1982) apt phrase. That is why we do not represent our thoughts and their contents by deploying, as Frege wrongly imagined, a set of special purpose linguistic devices, uniquely reserved for talking about inner thought constituents. We talk about thought and its contents simply by redeploying devices that are already available to us in describing and referring to the external world. Simply by possessing the power to say or to think, for example, that the cat is on the mat, we are a long way toward having the power to talk or think about the contents of a mind, to describe a mind as believing that or fearing that the cat is on the mat. 26 See, for example, Fodor (1987, 1984). For an early and now classic defense of narrow content see White (1982). For a treatment of narrow content as “notional” content, see Dennett (1982). For a series of daunting early attacks on the coherence of narrow content, see Tyler Burge (1979, 1982a, 1982b, 1986). Fodor, once the greatest advocate of the priority of narrow content officially renounces the need for narrow content in his 1993 publication. My own arguments against narrow content are contained in essays XI and XII of Taylor (2003). It should not be thought that narrow content is a dead letter. For one thing, Aydede (1997) makes a convincing case that there may be less to Fodor’s official “abandonment” of narrow content than meets the eye. Moreover, narrow content still has a number of able and ardent defenders. For a defense of a rather limited version of narrow content see Recanati (1993, 1994). For two more wholehearted recent defenses of the primacy of narrow content see Rey (1998) and Chalmers (2002). I take Frege himself to be the ultimate inspiration for the notion of narrow content, since it was he who most clearly located sense, and with it thought content, entirely on the side of the cognizing subject. There are, to be sure, early and forceful anticipations of this idea in the likes of Descartes, for example. Unlike Frege, however, Descartes really had no clue how to get mind and world back together again, once the world was stripped of any role of determining the contents of our thoughts.

Representing Representations  227 I do not mean to deny, as a behaviorist might, the distinction between the inner world of thought and the outer world to which thought and language are answerable. I claim only that the ability to cognize thought and its content is in some sense parasitic upon our ability to cognize the world—​in particular, to cognize ourselves and our inner representations in relation to an outer word. It is precisely the priority of our cognitive hold on the outer world—​and ourselves as standing in diverse cognitive relations to that outer world—​over our cognitive hold on the inner mental representations that is reflected in the relative priority of de re ascriptions of mental contents over de dicto ascriptions of mental contents. We talk about the fine-​grained contents of the mind, not by talking directly about the inner denizens of mental life—​ notions, ideas, representations, concepts, modes of presentations, or what have you—​but by talking about configurations of worldly objects, properties, and events in relation to our own diverse cognitive relations toward such configurations. That is why attitude ascriptions are designed to help us distinguish between commitments undertaken by the ascriber and commitments attributed to the ascribee. Such ascriptions help us to locate ascriber and ascribee as well as the interlocutors of the ascriber, both relative to one another and relative to a shared world. This should not be an unsettling or disturbing outcome. Nor does it give us any reason to doubt the coherence of either de re beliefs or de re belief ascriptions. The point is just that the contents of our beliefs are not pristine and unsullied by the world, inwardly safe from even the threat of deep incoherence, as Frege and his many philosophical descendants tend to believe. Belief content is a joint product of mind and world, with neither that which lies on the side of the subject nor that which lies on the side of the objects enjoying any peculiar priority over the other. Let me make clear what I am and am not claiming. To deny that there is an inner realm of pure de dicto content, a realm that is intrinsically rational and therefore inwardly secured from even the threat of incoherence, and to deny that any such pure de dicto realm is somehow prior to a realm of metaphysical or referential content—​that is, in a sense, sullied by our one-​sided engagements with the world—​is not to deny that there is a significant story to tell about the inner psychology of believing. Nor is it to deny that that story is best told in the idiom of inner representations.27 Again, the view gestured 27 Concluding that there is no inner representational story to tell about the psychology of thought on the basis of the non-​existence of narrow content is certainly a fallacy of some sort. I wish I had a

228  Referring to the World at here is meant to give off not a single whiff of behaviorism. Though de re believing is the undertaking of predicative commitments with respect to the objects themselves, the undertaking of such commitments is itself ultimately constituted by the deployment of inner representations of a certain sort in thought episodes. And like all mental representations in a finite rational mind, even those representations whose deployments constitute de re cognitions of objects are, in a sense, one-​sided. But to say that they are one-​sided is not to say that such representations enjoy purer, intrinsic, and therefore more epistemically secure rational contents that stand between the cognizer and the objects of her de re cognitions. Throughout this book I have argued that epistemic one-​sidedness is not a one-​sidedness at the level of thought content, as Frege and others mistakenly believed; rather, it is a merely syntactic one-​sidedness. As such, it is a one-​ sidedness at the level of form. It will help to recapitulate some of the main points of that argument in the current context. Doing so will help us to see more clearly why one-​sidedness is not any sort of threat to the conceptual probity of de re cognitions after all. First, I take episodes of de re believing to be partially constituted by the inner deployment of name-​like, indexical-​like, or demonstrative-​like mental representations in inner syntactic construction with predicate and verb-​like inner representations. The reader should recall that when I say that an inner mental representation is “name-​like,” I mean that it has, in the realm of thinking, syntactic and semantic roles similar in kind to the semantic and syntactic roles that are definitive of the public language category NAME.28 I argued at length in Chapter 4 that to be a name is to be an expression type N such that any two tokens of N are guaranteed to be coreferential. This, I claimed, is a linguistically universal fact that partially defines the linguistic category NAME. Co-​typical name tokens are explicitly coreferential, where, again, explicit coreference is sharply distinguished from coincidental coreference. Two name tokens that are not co-​typical can refer to the same object and thus be coreferential without being explicitly coreferential.

name for it. One prominent philosopher who seems to me to flirt with such a fallacy is Baker (1987, 1995). Another, more ambiguous case is Millikan (1993, 2000). 28 I do not mean to suggest that all types of natural language expression have language of thought correlates. For example, Richard Heck (2002) has argued persuasively that the second-​person pronoun has no language of thought correlate, since, roughly, our thoughts are never addressed to another. Addressing another, that is, is essentially a communicative act. So it is unsurprising that public languages have this pronoun as they are instruments of communication, but the language of thought, which is not such an instrument, does not contain a second-​person pronoun.

Representing Representations  229 For example, tokens of “Hesperus” and tokens of “Phosphorus” corefer but are not explicitly coreferential. And it is a correlative truth about names, a truth also partly definitive of the lexical-​syntactic character of names, that when m and n are distinct names, they are referentially independent. Recall that I have argued that there must be a class of mental representations that function as devices of explicit coreference in the de facto private language of thought. Without such devices, I claim, it would always be an open question for the individual cognizer whether, in thinking now of a particular o and now of a particular o', she has thought of two distinct objects or has thought of the same object twice. Though it may sometimes, perhaps even often, be an open question for a cognizer as to whether two of her thought episodes share a (putative) subject matter, it is surely not always an open question. I have claimed further that the ability to think token distinct thoughts that bear such relations of same-​purport to each other is a condition of the very possibility of the de re thinkability of objects. If no two thoughts purported to be about the same object, then in thinking any new thought, it would be inwardly as if one were always thinking about an object never previously cognized. It is even arguable that a mind in which no two thoughts bear relations of same-​purport to one another would altogether lack any cognitive hold on objects. Thus our ability to deploy in thought various devices of explicit coreference, devices such that to think with them again is to purport to think of the same object again, is a central source of our capacity for same-​ purporting thought. At the same time, I  have been careful to distinguish mere purport of thinking of the same again from success at thinking of the same again. Same-​ purporting thoughts need not be about any object at all. For example, I will make heavy weather in Chapter7 of the fact that Santa Claus-​thoughts one and all same-​purport with one another, but they are about no object.29 Nor is same-​purport the same as mere coreference. Two inner names may refer to the same object, and thereby both condition the de re thinkability of one and the same object again even if those names do not share inner referential 29 See Taylor (2014b) for more on fiction and empty names. In complete fairness, I should say Kant can plausibly be credited with some recognition of this fact. Witness in this connection his distinction between merely thinking an object and cognizing an object. In full-​blown cognition of an object, there must be both a given intuitive element and a formal conceptual element. In bare thought, devoid of intuitive content, we have, he claims, merely “empty concepts of objects, through which we cannot even judge whether the latter are possible or not—​mere forms of thought without objective validity.” Here Kant anticipates the possibility of same-​purport in the absence of reference to any object at all. Same-​purport in the absence of reference amounts to what I have called objectuality without objectivity, referential fitness without referential success.

230  Referring to the World purport. This is what happens when one thinks of Venus now via an inner “Hesperus” and now via an inner “Phosphorous.” Success requires that inner representations be “bound down” to outer objects. But nothing lying merely on the side of the cognizing subject can guarantee success on its own. For nothing solely belonging to subjectivity can guarantee that the subject’s name-​like representations are bound coherently down to outer objects. And this means that nothing merely on the side of the subject suffices to guarantee that its cognitions are de re cognitions. Nor can anything lying merely on the side of the subject guarantee that two singular representations that are bound to the same object again will be treated by the mind as devices for thinking of the same again. But this implies that nothing lying merely on the side of the subject can vouchsafe for the external coherence of such de re cognitions as the mind happens to enjoy. All that can be guaranteed on the side of the subject alone is that the subject’s inner representations be objectual or referentially fit. The mind cannot guarantee that those representations be objective or referentially successful. Recall that a representation is objectual or referentially fit if it is (syntactically) fit for the job of standing for an object. Expressions that are fit for the job of standing for an object are those that can well-​formedly flank the identity sign, that can well-​formedly occupy the argument places of verbs, and that can well-​formedly serve as links of various sorts in anaphoric chains of various sorts. Such are, on my view, antecedently poised to refer, prior to our encounters with the world. This inwardly determined property of referential fitness by which inner representations are rendered antecedently poised to refer must, of course, be sharply distinguished from referential success. Successful de re thinking happens when we deploy not merely objectual but fully objective representations in thought episodes. Because the objectivity of our representations, and thus the de re thinkability of objects, is not the business of the mind alone, there can be no purer, safer realm of pure de dicto thought contents, that is prior to and distinct from de re thought content. And that is why it should not be surprising or mysterious that de re ascriptions have a kind of priority over so-​called de dicto ascriptions. I do not mean to suggest that de re cognitions are cognitively unproblematic. The very fact that the mind alone cannot guarantee that its objectual or referentially fit representations are coherently bound down to outer objects—​a fact that gives us yet another reason for distinguishing the inner purport of coreference from actual coreference—​is the source of one great difficulty for our de re cognitions. Consider again the fact that a cognizer

Representing Representations  231 may cognitively encounter a particular object while mistaking it for another. I may, for example, cognitively encounter Joelle but mistake her for her twin sister Marie. In such a context, I may deploy an inner token of “Marie” in thinking about the woman I encounter. In that case, my thought will same-​ purport with many earlier thoughts about Marie. But there is also an intuitively clear sense in which my thought can be said to be about Joelle—​even if it is and purports to be about Marie as well. Though there is a sense in which my thought is about Joelle, it clearly does not same-​purport with my earlier thoughts about Joelle. I am in a divided mental state. I am confusedly thinking, via a tokening of an inner “Marie,” with respect to that very person now in front of me, who happens to be Joelle, that she is a promising young tennis player. I am, in effect, thinking of Joelle as Marie, thinking of Joelle with Marie-​purport. If my confused thought has at least as much claim to be about Joelle as it does to be about Marie, it follows that it is not necessarily and unambiguously the case that inwardly same-​purporting thoughts succeed in being purely and simply about one and the same external object. The correlative facts that an internal assurance of same-​purport does not yet constitute an external guarantee of coreference, and that actual coreference does not guarantee same-​purport, is a direct reflection of the syntactic one-​sidedness of the representations that mediate our de re attitudes. It is a reflection, that is, of the fact that the inner form and role of name-​like and other singular representations is insufficient to guarantee that when two such representations are bound down to the same outer object, they will ipso facto be syntactically and dynamically linked in our inner mental lives. Just because representations that are bound down to the same object again are not guaranteed to inwardly same-​purport and are therefore not guaranteed to be syntactically and dynamically linked in our inner mental lives, there is the ever-​present danger that even a fully rational mind may sometimes fall into external incoherence. Precisely the fact of this ever-​present danger leads Kaplan to despair over the purity and naturalness of de re belief. Indeed, it is precisely against such a possibility that his stringent epistemic standards are meant to safeguard de re belief. But Kaplan’s despair is misplaced. It should not lead us to following him in holding the possibility of de re cognition to such extraordinarily high epistemic standards. Even a confused or incoherent thought about an object may still be a thought about that very object. Begin by noting that otherwise referentially fit singular representations may lose their grip on the objects altogether. Consider Joelle again. Imagine that, entirely unbeknownst to me, she is one of quintuplets. Each time

232  Referring to the World I encounter one of her sisters, I token an inner “Joelle.” Now suppose that I mistakenly agglomerate all of the information I have about any of the sisters into one huge “Joelle” file. I think to myself, “My, that Joelle gets around.” I deploy my inner “Joelle” in a name-​like and fully objectual or referentially fit fashion. Each time I deploy “Joelle” in a thought episode, I thereby think with an inward purport of sameness again. That is, I thereby think as of the same object again. But of what object do I thereby purport to think as of the same again? Are my thoughts about Joelle? About one or the other of her sisters? Is it really determinate whether I am thinking of Joelle or one of her four sisters? Perhaps I think of now one sister as Joelle, now another as Joelle, and now yet another as Joelle. Perhaps I think of a mereological sum of Joelle and her sisters. Perhaps there is simply no fact of the matter about who, if anyone, I am thinking of. Perhaps, I do not succeed in having a de re cognition at all. A good theory of the ultimate source and nature of de re cognitions should ultimately answer such questions or at least say why, in the nature of the things, there can be no determinate answers to them. Moreover, any good theory of de re cognitions will have to accommodate the fact that nothing lying merely on the side of the cognizing subject can guarantee that when a thinker is presented with the same again, she will ipso facto recognize that she is presented with the same again. Correlatively, such a theory will have to accommodate the fact that nothing lying merely on the side of the subject guarantees that when a thinker inwardly purports to think of the same again, she necessarily and unambiguously succeeds in thinking of the same again. It is a consequence of these correlative facts, which I take to be a direct consequence of syntactic one-​sidedness, that there is an ever-​present possibility that entirely referentially fit or objectual representations, that are antecedently well-​poised to refer, may, in the end, be so incoherently and confusedly ordered in relation to outer objects that their inner deployment in thought episodes gives rise to no fully determinate de re cognitions. But even if we grant that enough external confusion and incoherence can cause otherwise inwardly fit representations to lose their hold on the objects, and even if we grant the ever-​present epistemic possibility that we have fallen into such confusion and incoherence, it simply does not follow from such considerations alone that our representations are actually so incoherently and confusedly ordered as to make de re cognitions impossible. But if the mere standing possibility of confusion and/​or incoherence in relation to outer objects in our de re cognitions does not suffice to undermine the standing of those cognitions as de re cognitions, then there is no reason to conclude that the

Representing Representations  233 mere one-​sidedness in any way threatens the purity and naturalness of de re belief. It would be surprising if it were otherwise. Thinking about an object is one thing. Thinking about that object coherently and unconfusedly is an entirely different matter. No form of incoherence is a good thing. The unfortunate epistemic predicament of finite cognizers like ourselves, however, is that even when we are as inwardly rational as we can be, we face the ever-​present possibility that we have fallen into incoherence. Some imagine that by retreating inward, to some safer realm, in which the rational powers of the mind over its inner representations is unhindered by the influence of the outer world, we can guarantee ourselves a kind of inner coherence. Such inward coherence, the further thought goes, will, at the very least, enable us to maintain clean hands in our cognitive encounters with the world. When we fail, for example, to recognize that the rising of Hesperus just is the rising of Phosphorus or that Tully just is Cicero again, the blame will lie with the world, or with the mind in relation to the world, but not with the mind itself. But the illusion that we can make such a clean cut between the rationally pristine inner contents of the mind and such contents as are sullied by our encounters with the world is, I think, little better than a comforting illusion. To retreat from the world in this way is not to withdraw into a realm of pure, unsullied inner mental content. It is to retreat from mental content itself. I will not argue that point further here, but I will say that I can find no reason to believe that either the mere possibility or even the actuality of a moderate degree of external incoherence is sufficient to undermine the standing of de re cognitions as de re cognitions. For one thing, in the communicative device of the fulsomely de re, we have a way of reporting such confused and incoherent states of mind that inoculates the reporter from having to share in or endorse the reported confusion and incoherence. But we do so not by retreating to an inner, safer, purer realm of purely de dicto content. Rather, we do so by triangulating on a shared world and our diverse relations to it. I concede that on the view of de re belief and de re ascriptions of belief that I have sketched in this essay, it will turn out that whatever suffices for the mere de re thinkability of an object will not suffice for the kind of tight cognitive grip that the likes of Russell or Kaplan apparently take to be criteriological of de re thinkability. But I insist that it was all along a mistake to set such high epistemic standards for the mere de re thinkability of objects to begin with. It may help to distinguish mere de re thinkability from what we might call, following Robert Brandom (1994), “epistemically strong” de re thinkability.

234  Referring to the World For the former, it suffices that our thoughts be determinately bound-​down to the objects, that our thoughts be answerable to how things are by the objects in a way that is independent of how those objects are presented to us. For the latter, it is necessary that we achieve the sort of tight cognitive grip on the objects that a Russell or a Kaplan was after. Epistemically strong de re attitudes would no doubt be very powerful things to have. They would enable one not merely to think about the objects but to recognize the objects that one thinks about when they are presented again, even if under different guises. Still, it seems to me crucial that we keep this useful and important distinction always in mind. When we fail to attend to the distinction between mere de re thinkability and epistemically strong de re thinkability, we are liable to the tempting, but mistaken, inference from the one-​sidedness of all mental representations to the conclusion that de re belief is somehow more problematic than some other more secure and purer notion of belief.

7 The Things We Do with Empty Names 1. Preliminaries A good theory of objective representational content ought to do justice to our quite frequent and impactful thought and talk about apparently non-​ existent objects. But thought and talk about the apparently non-​existent is not entirely unproblematic. Philosophers have exhibited many different attitudes toward such thought and talk—​from the squeamishness of Frege and Russell to the open and loving embrace of Brentano and Meinong. In one breath, Frege (1979) grudgingly admits that we do seem to think thoughts apparently about the likes of Santa Claus or Pegasus. But in nearly the next breath he dismisses such thoughts as merely “mock thoughts.” Russell (1905) was driven to the somewhat desperate departure of saying that expressions that superficially appear to be proper names, and thus to making putative reference to things that don’t exist, in fact contain no such expressions at all and really purport to make no such reference at all. Meinong (1960), by contrast, took expressions like “The round square” or “The golden mountain” at face value. Strikingly, he felt no need whatsoever to deny that such expressions purport to denote or refer to certain objects. And while he did not hesitate to acknowledge that the relevant objects don’t exactly “exist,” he did insist that they nonetheless do enjoy their own quite special and peculiar mode of being—​what he called “subsistence.”1 Meinong’s teacher, Brentano (1995), had already done him one better. He endorsed not just the view that some of the objects of our thought are merely “immanent” objects that don’t 1 A number of prominent contemporary philosophers happily embrace the idea of non-​existent objects. See, for example, Parsons (1980), Priest (2005), and Zalta (1983, 1988). Though I disagree, my explicit aim in this essay is not to argue against such views. Unlike Russell, for example, I do not claim that embrace of the non-​existent leads inevitably to contradiction. Indeed, I think it has by now been amply demonstrated that with the right choice of background logical principles, one can coherently maintain some form or other of Meinongianism. I do not reject Meinongianism, then, because I find it incoherent in the way that Russell alleged. I reject it because I see no positive basis for believing in any version of that doctrine. Rather, I hold that there is nothing that needs explaining about the behavior of putatively non-​referring expressions that requires or licenses us in positing non-​existent objects.

Referring to the World. Kenneth A. Taylor, Oxford University Press (2021). © Oxford University Press. DOI: 10.1093/oso/​9780195144741.003.0007

236  Referring to the World “really” exist, though they do exist for thought alone. He actually endorsed the stronger, more startling claim that every object of thought has a “merely” immanent existence as an object for thought but has no existence whatsoever “outside” of thought.2 I share neither the squeamishness of a Russell or a Frege nor the wholehearted embrace of the non-​existence of a Meinong or a Brentano. But I do afford a certain degree of respect and honor to the underlying philosophical proclivities that drove each to their respective departures. Frege and Russell, and all who follow in their squeamish footsteps, are, I think, correct to insist that our thought and talk about the non-​existent should not and need not be taken to be the thought and talk about a realm of peculiar entities with their own special mode of being. But it is a mistake to think that such thought and talk threatens to land us in contradiction, as Russell feared, or that such thought and talk should be dismissed as a mere mockery of genuine thought. Brentano and Meinong, and their many philosophical descendants, are entirely right to afford to such thought and talk a certain degree of seriousness. Indeed, though I shall argue that thought and talk about the non-​existent is literally about no things, I shall also insist that it may still have quite far-​reaching cognitive significance and may still exhibit sterling rational credentials. But accounting for the cognitive significance and rational credentials of such thought and talk does not require us to posit a peculiar realm of objects with a peculiar mode of being. On that score, I stand squarely with Quine, who once insisted that to pursue such a strategy is to lay ruin to the good old word “exist.” The key to making good on our attempt to do adequate justice—​but no more than that—​to both the embrace and the squeamishness is a trio of interlocking distinctions. The first of these distinctions is one on which I have already spilt a great deal of ink at various places in this book. It is

2 Thus Brentano (1995): Every mental phenomenon is characterized by what the Scholastics of the Middle Ages called the intentional (and also mental) inexistence of an object, and what we would call, although not in entirely unambiguous terms, the reference to a content, a direction upon an object (by which we are not to understand a reality . . .), or an immanent objectivity. Each one includes something as an object within itself, although not always in the same way. In presentation something is presented, in judgment something is affirmed or denied, in love [something is] loved, in hate [something] is hated, in desire something is desired, etc. This intentional inexistence is exclusively characteristic of mental phenomena. But see, Crane (2006) for a more subtle and nuanced reading. Fortunately, since Brentano scholarship is no part of my aim, it hardly matters for my purposes what Brentano finally actually thought.

The Things We Do with Empty Names  237 the distinction between merely objectual and fully objective linguistic and mental representations. The second is the distinction between what I  call non-​veridical and veridical language/​thought games. The third is the distinction between truth-​similitude and flat-​out or strict, literal truth. On the basis of this trio, I shall defend the view that (a) empty names and other referring terms are merely objectual representations rather than fully objective ones; (b) we play many non-​veridical language and thought games with such expressions; and (c)  while moves within such games do not always enjoy strict or literal truth or falsity, they often enjoy truth-​similitude. It is in their capacity to achieve truth-​similitude that the cognitive significance and rational credentials of non-​veridical language games played with merely objectual representations are most fully on display. My trio is thus indispensable for achieving a proper understanding of the pragmatic, cognitive, and rational significance of the many things we do with empty names and other merely objectual representations. For want of my trio, Russell was tempted by the mistaken conclusion that non-​referring names are not names at all. Brentano was lured into denying the mind’s ability to make any contact at all with independently real existents. Idealists more generally are led into thinking that objects are nothing but the shadows cast by certain sorts of representations. Platonists are misled into positing a realm of free-​standing mathematical objects whose nature is somehow entirely independent of our doings. None of these views is forced on us. I doubt that any can be sustained. My trio will help us see why we need not, after all, be so much as tempted by any of them.

2.  Empty Names and Fictional Discourse We begin by focusing on names explicitly drawn from fiction. Some are tempted to deny that names in fiction are genuinely empty names. Indeed, many philosophers, including some who endorse referentialism, claim that fictional names designate fictional objects. But I hope to show that my trio of distinctions frees us from any special pleading for either the existence or pseudo-​existence of fictional objects. Begin by considering the following three sentences: (1) Hamlet was the tormented prince of Denmark. (2) Hamlet is the main character in Hamlet. (3) Prince Charles has fewer reasons to be depressed than Hamlet did.

238  Referring to the World There is certainly an intuitive sense in which each of (1)–​(3) is, if not strictly literally true, then at least correctly assertable. A good theory of fictional discourse certainly ought be able to explain what we are doing when we utter any of (1)–​(3). And it ought to explain what the “correctness” of doing so could plausibly consist in. I have quite deliberately phrased our initial quandary about (1)—​(3) so as to allow for the possibility that literal truth and falsity are not at issue with respect to utterances of (1)–​(3). I do so in order to leave open the possibility that there is a principled distinction between what makes for the felt correct assertability (or pseudo-​assertability) of (1)–​(3) and what, if anything, would make for the strict, literal truth of (1)–​(3). Some philosophers will no doubt find this sort of forbearance overly cautious. They will insist that we have direct intuitions of truth and falsity in such cases. But I take no such thing to be obvious. I do concede that unless we are prepared to make some distinctions, we may find it difficult to resist the superficially tempting but ultimately mistaken inference from the felt correct assertability or pseudo-​assertability of (1)–​(3) to the conclusion that (1)–​(3) are in some sense “true.” That, of course, is precisely where my trio comes in. Once we are fully armed with my trio, we should be more able to appreciate that there is nothing in which the literal truth or falsity of an utterance of any of (1)–​ (3) could possibly consist. Taken strictly and literally, I claim that (1)–​(3) stake out no determinate claims at all about how the world stands. Indeed, I shall argue that they are devoid of any propositionally complete objective representational content. Because they are propositionally incomplete, they are not the kinds of things for which the question of strict, literal truth or falsity even arises. It is sometimes claimed that sentences like these express gappy propositions. I myself prefer to say that they express what is merely a proposition in waiting. Talk of “gappy propositions” suggests a thing that is a proposition, all right, just with “gaps” of a certain sort. I think it is better to think of such sentences as being “on the way” toward expressing propositions. And that is why I  prefer to talk of propositions in waiting. But disputes between me and those who prefer the idiom of gappy propositions are merely family squabbles in the overall scheme of things. I will say more about such matters shortly.3

3 A  fair number of referentialists seem to agree that sentences containing empty names fail to express a full-​blown proposition. There is some disagreement, however, over exactly what such sentences do express. Some referentialists hold that such sentences express so-​called “gappy”

The Things We Do with Empty Names  239 The claim that sentences like (1)–​(3) are propositionally incomplete may seem intuitively jarring. Perhaps it is. But I take it to be a more or less direct consequence of a pair of further theses—​one independent metaphysical thesis and one independent semantic thesis. Each of the further theses is independently plausible and is fully defensible—​perhaps without appeal to my trio. Still, once my trio is fully onboard, the two further theses and their joint consequence should all go down much more smoothly. The independent semantic thesis is that a name contributes nothing but its referent to the proposition expressed by any sentence in which the name occurs. The independent metaphysical thesis is the thesis that there are no non-​existent or fictional objects.4 I have already acknowledged that many able philosophers still lovingly and stubbornly embrace the non-​existent. Similarly, referentialism is still a topic of debate at least among philosophers of language. I doubt that the arguments of this chapter will suffice to weaken either opposition to referentialism among those still disposed to resist that doctrine, or opposition to the non-​existence claims among those still disposed to embrace the non-​existent. I will count my arguments successful enough if they are sufficient to give pause to those who have not yet taken the plunge into either semantic or metaphysical darkness and sufficient to signal to those committed to such darkness that the battle is still on and the opponent has not yet been defeated.

propositions. On my view, what such sentences express is not yet a proposition at all but is a propositional scheme or a proposition in waiting. Some thinkers hold that such entities can still have truth values. On my view, however, propositions in waiting (or “gappy” propositions) are not yet the sort of thing for which questions of truth or falsity meaningfully arise. I suspect that the family squabbles over such issues may be merely verbal. If they are, I still insist on my own formulation since I am most certain what is meant by that formulation. The literature on so-​called gappy propositions is by now extensive. For a representative sample, see Taylor (2003), Braun (1993, 2000), Adams and Detrich (2004), and Adams, Fuller, and Stecker (1997), among others. For a dissenting voice, see Everett, (2003). 4 Though fictional objects have certain affinities to the Meinongian non-​existent objects, belief in fictional objects can be seen as far less ontologically profligate than belief in the non-​existent more generally. One could reasonably believe that fictional objects are real, though secondary existents, somehow ushered into being by the very act of creating a work of fiction—​which is itself a real, though secondary existent. One need not believe that non-​existent golden mountains, nonetheless, subsist in order to believe, for example, that Sherlock Holmes has some sort of being—​real, actual being—​as a fictional character. See Kripke (2011), Fine (1982), Schiffer (2003), Salmon (1998), Searle (1979), Searle (19985) and Thomasson (1999) for views that endorse the reality of fictional characters. Though I reject such views, there is much that is plausible in them. I attempt to accommodate what is plausible in them by appeal to the concept of what I call merely notional objects. For a nice discussion of the pros and cons of realism versus anti-​realism about fictional characters, see Friend (2007).

240  Referring to the World A caveat and a concession are in order. When one creates a work of fiction, one does usher at least one new object—​the work itself—​into existence. The work is in one sense something brand new under the sun. It is what I earlier called a “secondary existent.”5 By a secondary existent, I mean a mind-​ dependent entity that exists because and only because we take it to exist. In such takings, we may confer a certain normative status and/​or function upon a configuration of more primary reality. This last qualification is important because it shows that secondary existents do not come to be ex nihilo or by a mere play of our representations. Beneath all secondary existents, there must ultimately sit certain more primary existents. At least relatively speaking, these more primary existents will count as preconfigured and preexisting. Neither novels nor nations nor tables and chairs float entirely free of collections of more primary existents. In order for nations to exist, there must already be people. In order for tables and chairs to exist, there must already be the relatively raw materials out of which tables and chairs are ultimately made. I insist that it is through, and only through, the conceptual and normative uptake of antecedently obtaining configurations of more primary existents that the mind is able to usher secondary existents of any sort into being. This last point is absolutely crucial. The bare existence of more primary existents is necessary, but is surely not sufficient, for the existence of secondary existents. Absent the power of the mind to confer status and function upon configurations of more primary existents through the deployment of certain concepts, such configurations would not yet constitute secondary existents. There would be no tables, nations, or novels without minds present to take certain configurations as tables, nations, or novels and thereby to make those configurations into such things. It is by and only by taking a configuration of primary existents as a table or a nation or a novel that we manage to make such things exist. It is also crucial to recall that when it comes to secondary existents, our concepts neither fit nor fail to fit antecedently existing realities. Rather, our concepts precede and make possible secondary existents that fall under those very concepts. We conceive of the possibility of a novel, and we set about to so arrange the world that certain configurations of primary existents count as novels. Concepts of kinds or classes of secondary existences are thereby “thrown out” toward the inward rush, without being extracted from it. 5 See Chapter 5 and also Taylor (2011, 2015a, and 2019) for more on the notion of a secondary entity.

The Things We Do with Empty Names  241 Secondary existents come in vast varieties. The social world in particular is replete with them. Our ontological permissiveness with respect to secondary existents may appear to suggest that it is a short and easy step from granting secondary existence to novels to also granting secondary existence to the fictional characters on whose doings a novel would appear to report. And that might lead one to conclude that names in fiction are not, after all, merely objectual expressions that are referentially fit, but not referentially successful. Names in fiction are referentially successful expressions that stand for or refer to fictional characters. And fictional characters are real, though merely secondary existents. If so, then fictional statements containing fictional names might well be thought to be literally true or false, as long as they correctly report on the fictional doings of fictional characters. But as ontologically profligate as I  am willing to be when it comes to secondary existences, I  deny existence of any sort to fictional characters. Putatively fictional characters are clearly not primary existents. And my resistance to counting them among the secondary is, I admit, partly due to a hankering if not exactly for sparse or desert landscapes, then at least to landscapes that are not arbitrarily overpopulated. That is, I can find no positive or principled basis for positing an ontology of fictional objects, even conceived as secondary. There is nothing about fictional discourse that needs explaining that the positing of fictional objects, even fictional objects conceived as secondary, would help us to explain. Or so I shall argue below.6 If there simply are no genuine objects—​primary or secondary—​for names and other singular referring expressions used in fiction and other non-​ veridical language thought games to refer to, and if all any singular referring expression ever contributes to determining either the proposition expressed 6 I do concede, however, that it is possible to introduce a series of “notional objects,” as I call them below. Notional objects encode certain purported relations of reference and co-​reference among names and other referential devices as they occur within works of fiction. Notional “objects” aren’t genuine objects in any robust metaphysical sense. They are, rather, entirely dispensable façon de parler. As such, talk of notional objects is not intended to add anything substantive to our understanding of the pragmatic and cognitive significance of fictional discourse. A survey of our ontological commitments which left off notional objects would have left nothing of genuine ontological significance out of the picture. There are no genuine objects, not even secondary ones, for fictional names to refer to or for fictional statements to be “about.” What really and truly exist in the case of fiction are certain representations, repeatable structures of representations, and certain secondary existents that are constituted as distinct, though secondary realities, through the collective exercise of our status-​conferring powers on those structures of representations. There are also various non-​ veridical language games grounded in our production and consumption of such representational arrays. That is already a lot. It is not, however, enough to generate genuine fictional “objects”—​or any other sort of non-​existent entity, for that matter. Even so, it should be enough, I argue below, to satisfy even the most ontologically voracious among us.

242  Referring to the World by or the truth conditions and/​or truth values of sentences in which it occurs is the object for which it stands, then it follows directly that when there is nothing for a name to refer to, there is no proposition for a sentence containing that name to express.7 Explaining how sentences containing empty or non-​referring names can, nonetheless, be correctly assertable or, as I prefer to say, “pseudo-​assertable”—​typically, but not always, in the context of some non-​veridical language game—​and also how such sentences can enjoy a kind of cognitive significance and rational warrant despite the fact that they fail to express determinate propositions, and thus fail to be even the kind of thing for which truth or falsity meaningfully arises, is the main burden of much of the remainder of this chapter.

3.  Empty Names as Merely Objectual Singular Representations I already appealed to the distinction between fully objective/​referentially successful representations and merely objectual/​referentially fit representations at many different points in the overall argument of this book. In returning to that distinction in the current context, it is worth recalling that from a structural-​syntactic point of view, there is nothing to distinguish referring names from non-​referring names. No narrowly linguistic transformation or operation sensitive to structure or form alone will be sensitive to whether a name has a referent or fails to have a referent. Names that fail to refer still exhibit all of the structural-​syntactic features that serve to constitute a name as a name. Whether referring or non-​referring, names still well-​formedly flank identity signs, still occupy argument places of verbs, and still anchor anaphoric chains. More important for our current purposes, empty names still function in both speech and in thought as devices of explicit coreference/​ same purport. It is in this role that we can most clearly see that name types 7 Many philosophers appear to think that we can effortlessly create fictional objects just by deploying the apparatus of singular reference in the context of fiction. Though I deny the existents of fictional objects, my talk of notional objects may appear to give credence to this view. But my notional objects are not intended to be objects in any deep or robust sense. On my view, notional objects aren’t really anything at all over and above the structure of representations in terms of which they are defined. Though I am willing to countenance notional objects if (and only if) they are understood in this way, I actually think that there is a better and more honest way of thinking about what we achieve when we deploy the apparatus of singular reference in the context of fiction. That way of thinking involves distinguishing between fully objective and merely objectual representations, between veridical and non-​veridical language games, and between truth and mere truth-​similitude.

The Things We Do with Empty Names  243 enjoy, in virtue of their structural-​syntactic roles, both a certain intrinsic purport to refer, and that numerically distinct tokens of the same name type enjoy an intrinsic purport to corefer with one another. And remember too, that tokens of distinct name types, whether referring or non-​referring, are referentially independent of names of any different type. From this much alone, we already have a basis for resisting Russell’s conclusion that empty or non-​referring name are not after all genuine names. But there is more. Qua merely objectual or referentially fit, a non-​referring or empty name is still in one sense a fully singular representation.8 To be sure, we do need to distinguish two distinct modes or types of singularity. Non-​ referring or empty names are singular in form but not singular in content. Referring names are singular in both form and content. Precisely their singularity of form, and the intrinsic singular referential purport that comes with singularity of form, will be essential to both explaining and explaining away widespread temptation to posit non-​existence or fictional objects for such names to refer to. It will also be central to explaining the cognitive, pragmatic, and communicative significance of the use of empty names. The crucial insight is that merely objectual singular representations that fail to refer are still, in a sense, made for talking about objects. And precisely because they are made for talking about objects, I shall argue below, our use of them can at times have a somewhat illusory feel that sentences containing such names have a subject matter. One of the main aims of the present chapter is to both diagnose and dispel such illusions of subject matter. But even after we have dispelled such illusions, empty names and the games we play with them will still stand as cognitively and communicatively significant. I do not mean to suggest that all uses of merely objectual representations must give rise to subject matter illusions. Within the context of truth-​seeking, veridical inquiry about the use of merely objectual, merely referentially fit singular representation may often be anticipatory or preparatory without generating any illusion of subject matter. Suppose it is conjectured that there is a yet to be confirmed planet orbiting the sun closer than Mercury does. In anticipation of its discovery, the not yet confirmed planet is dubbed “Vulcan.” It turns out, of course, that there is no such planet and that the anticipatory dubbing was, in fact, a failure. The sort of objectivity and singularity of

8 See Chapter 4 and Taylor (2010) and Taylor (2015b) for a fuller defense of this claim. See also Sainsbury (2005) for defense of a similar idea. See Brandom (1994) for an inferentialist defense of a similar claim.

244  Referring to the World content that are the fruits of referential success comes with referential success that is never achieved. Yet despite never achieving full-​blown referential success, “Vulcan” is still a name. And as such, it is still referentially fit. All it is missing is a determinate reference. As such, we may do with and for the name “Vulcan” many of the things that we do with and for referentially successful names. We may open a file, a Vulcan-​conception, ready to be written in. And in that file, we may record many facts about the history of our failed endeavors to confirm the existence of Vulcan. To be sure, our Vulcan-​conception will be an empty conception that applies to no real existent. Nonetheless, such anticipatory uses of the name “Vulcan” need not generate subject matter illusions. In fact, the name is liable to be marked an explicit referential failure, once it manifestly fails to achieve referential success. But there is another sort of setting in which we throw up merely objectual, referentially fit, but not referentially successful names in our thought and talk. In such settings, we are more likely to be prone to subject matter illusions. I have in mind non-​veridical language/​thought games. In such contexts, we are sometimes tempted to think that objects can be given to us through a mere play of representations. Precisely such illusions are responsible, I shall argue, for many misbegotten philosophical doctrines about our thought and talk as of the non-​existent. But more on this point in due course.

4.  Empty Names in Veridical and Non-​Veridical Language Games Turn now to the distinction between veridical or truth-​tracking language/​ thought games and non-​veridical language/​thought games. Veridical language games are paradigmatically played with representations that are presumptively fully objective. Though empty terms sometimes occur in such games, as in the case of the anticipatory inquiry cited above, such terms are not naturally at home therein. When an empty name occurs within a veridical language game but is still used as if it stood for a real existent, something has typically gone wrong somewhere, especially if it remains empty and never achieves full objectivity, Perhaps an existence conjecture has been disconfirmed but word of the disconfirmation has not spread. The point is that typically when we play veridical games, we are trying to stake out determinate claims about how the world is. We are trying, that is, to track the true, to make our claims ultimately match the way things actually go in the

The Things We Do with Empty Names  245 world. Non-​referring expressions are not comfortably at home in such games just because sentences that contain such expressions express no determinate propositions and thus stake out no determinate claims about how the world is—​at least not by directly encoding any propositions in virtue of their literal meanings. Of course, as soon one makes such claims, one is immediately confronted with examples like the following, which may seem to give the lie to the claim that sentences containing empty names stake out no determinate claims about how the world stands and so lack a determinate subject matter: (4) Santa Claus isn’t coming tonight (5) Santa Claus will come this year if you wish it hard enough.

An utterance of (4)  would seem to stake out a true-​seeming claim about how the world stands, while an utterance of (5) would seem to stake out a false-​seeming claim about how the world stands. A person who utters (4) or (5) need not be engaging in any sort of pretend, mythic, or fictional practice of the sort that often supports moves within non-​veridical language/​thought games. We may presume that a person who utters (4) or (5) intends to convey something that is strictly true and true about the world. Moreover, although the apparent truth of what is conveyed by (4) and apparent falsity of what is conveyed by (5) would seem to depend directly on the failure of “Santa” to refer, neither (4) nor (5) seems to convey a metalinguistic proposition about the word “Santa.”9 At first blush, these facts may appear to sit uncomfortably with my claim that sentences like (4) and (5) express no fully determinate proposition and stake out no determinate claim about how the world stands. But I need not and do not deny that sentences like (4) or (5) can be used to convey determinate propositions. Nor need I deny that those conveyed propositions are

9 For a sophisticated and subtle version of something like a metalinguistic approach (though not quite) of what makes sentences containing empty names true or false, see Perry (2001). See also Donnellan (1964), by which Perry’s much more developed approach was more or less directly inspired. Central to Perry’s account is the notion of what he calls “network content.” Though it is not entirely clear, Perry seems to be suggesting that network content functions as intentional content for true (and false?) negative existentials. This makes facts about networks not just the truth-​makers of what we convey via apparently true negative existentials but also the intentional content for such statements. This seems to me incorrect. While I think it is plausible that something like a network content plays a role in explaining why what we convey by a sentence containing a non-​referring name may be true or false, it seems to me implausible to say that the conveyed “intentional” content of such statements is, in any sense, “about” networks.

246  Referring to the World determinately either true or false—​and determinately either true or false of the real world rather than of some merely mythical, notional, or fictional world. The pressing question, though, is what such sentences convey and how they manage to do so. Since I reject the existence of non-​existent, fictional, or mythical objects and claim that empty names make no semantic contribution to sentences that contain them, it should come as no surprise that on my view whatever such sentences convey, it must ultimately be via pragmatics rather than semantics that they convey it. Specifically, I  claim that sentences containing empty names can be used to express full-​blown propositions, not through the mechanism of compositional semantics, but through the mechanism of “one-​and-​a-​half-​stage pragmatics.”10 Through the one-​and-​half-​stage pragmatic process of what I call “pseudo-​saturation,” a semantically incomplete, merely partially semantically valued sentence may be used to convey a certain true but “pragmatically external” proposition. Now, one-​and-​a-​half-​stage pragmatics externalities are typically generated on the journey up from not fully saturated sentence meanings to contextually determined propositional contents. And this may happen in two different ways:  either as contextually generated by-​products of successful saturation or when so-​called primary pragmatic processes like saturation fail to come off. Now the contents generated by one-​and-​a-​half stage pragmatics do not “saturate” the gaps in a mere proposition-​in-​waiting. Because such externalities are not themselves the direct consequence of semantically saturating unsaturated slots, they are, in a sense, less intimately associated with the sentence and its constituents than are the propositions that are the result of saturation. Such externalities are generated “on the side” and consequently do not result in propositional constituents of the sort semantically demanded by the context-​independent meaning of the relevant sentences. Because of their “external” character, they typically are neither entailed nor presupposed by fully saturated contents. On the other hand, such contents are, in a sense, more intimately associated with a sentence and its meaning than Gricean implicatures are. Gricean implicatures are generated post-​ propositionally—​that is, post-​saturation—​via so-​called “secondary pragmatic processes.”11 Since such processes come into play only after full-​blown propositional contents have been fixed via saturation, Gricean implicatures 10 I  first develop the notion of the one-​and-​a half-​stage pragmatic process in an essay that is reprinted in Taylor (2003) as essay 6. See also Taylor (2007) and Taylor (2012) on which the current chapter is largely based. 11 For the distinction between primary and secondary pragmatic processes, see Recanati (2004).

The Things We Do with Empty Names  247 are less intimately associated with the sentence and its meaning than my one-​ and-​a-​half stage externalities are. Let us consider some details of this proposal briefly. I claim that a speaker who utters a sentence like (4) above does not assert any fully determinate proposition. Rather, she “pseudo-​asserts,” as we might call it, a proposition in the propositional neighborhood generated by the following propositional scheme. not bears ‘Santa’)

In particular, she pseudo-​asserts something like the descriptive proposition that no jolly, white-​bearded, red-​suited fellow, who lives at the North Pole and delivers toys via a reindeer-​drawn sleigh is coming tonight. I say “something like” this descriptive proposition so as to allow for a certain vagueness and indeterminacy in what is pseudo-​asserted. There may be no definite p such that it is precise and determinate that one who utters (4) pseudo-​asserts p, rather than some other “nearby” proposition q. But subject to such vagueness and indeterminacy, an utterance of (4)  will express something truth evaluable. That means that there is, after all, a sense in which one who utters the sentence “Santa Claus isn’t coming tonight” does manage to “say” something truth evaluable. She pseudo-​asserts some proposition or other which happens in this case to be true. Similarly, if she were to utter (5) she would assert nothing truth evaluable, but she would pseudo-​assert something that happens to be false—​a proposition to the effect that if you wish it hard enough, a certain jolly, white-​bearded, red-​suited fellow, who lives at the North Pole and delivers toys, will (finally) come this year. Why just these propositions, or something vaguely like them? The answer turns on the fact that a speaker who utters (4) or (5) takes on what I call a double burden—​the semantic burden of making reference to a certain object and the communicative burden of making manifest to her interlocutors what object she intends to make reference to. But taking on semantic and communicative burdens in a speech situation does not ipso facto entail successfully and openly discharging them. The key insight behind the notion of one-​and-​ a-​half-​stage pragmatics is that manifest failures to discharge one’s burdens is often the stuff of which further communication is made. This makes one-​ and-​a-​half-​stage pragmatics very different from Gricean implicatures. With a Gricean implicature, we implicate one thing in the course of or by saying another thing. One-​and-​a-​half-​stage externalities, on the other hand, may

248  Referring to the World be generated when we fail (or otherwise refrain) from saying anything fully propositionally determinate. A speaker who uses the name “Santa Claus” in a sentence doesn’t say one thing and thereby implicate something different. Especially if she is attempting to say something true about how things actually are, rather than how things are in the context of some myth or pretense or fiction, she (manifestly) fails to (semantically) refer to any object at all and thereby fails to assert, at least strictly and literally, any proposition at all. She thereby fails to discharge the standard semantic burden that comes with the use of any name—​the burden of using it to refer to one definite object. Precisely because she fails to refer, she also fails to assert any determinate proposition. Her interlocutors, however, need make only the assumption that even in her failure she still intends to speak cooperatively. They need only recognize that she is using the name in accordance with the standards of the mythical practice within which it is at home in order to appreciate her communicative intentions. She intends to communicate, and intends that her interlocutors recognize that she intends to communicate, a pragmatically external proposition to the effect that no jolly red-​bearded fellow who lives at the North Pole and delivers toys on Christmas is coming tonight. But it is important to stress that the descriptive phrase “jolly, red-​bearded fellow . . .” is neither the meaning nor the reference determiner of the name “Santa.” Still, it will be a prominent element of the “Santa” file of anyone familiar with the myth and lore of Santa. Because of that prominence, it is highly available and salient to all interlocutors. And it will play a mediating and facilitating role in their attempts to understand what a speaker is doing in uttering sentences like (4) or (5). Next consider sentences like (3) above, which contain a mixture of the fictional and the non-​fictional. They merit similar one-​and-​a-​half-​stage pragmatic treatment. On my view there is no propositionally complete semantic content strictly asserted by an utterance of (3). But (3) does manage to convey something true—​something strictly and literally true—​about Prince Charles. Because Prince Charles is a real-​world existent, what (3) conveys cannot be just another authorized move in some non-​veridical or fictional language game. (3) appears to somehow express a proposition about Prince Charles, who had no role as a character in the Hamlet story. So, how does one who utters (3) manage to convey such a proposition, given our view that “Hamlet” has no reference and the fact that Prince Charles bears no particular relationship to the play Hamlet and has no role within that play? The key, once again, is the one-​and-​a-​half-​stage pragmatic process of pseudo-​saturation. Here

The Things We Do with Empty Names  249 pseudo-​saturation takes a mere propositional scheme—​a scheme that in some sense semantically calls for one kind of value—​in this case a person to be the referent of “Hamlet”—​and yields instead a proposition with a different sort of value—​in this case, something like a cardinality. This cardinality—​ let’s call it the Hamlet-​cardinality—​is determined by facts about the moves authorized by the interpretive game anchored in Shakespeare’s Hamlet. What (3) “says” is that the cardinality of the set of reasons that Prince Charles has to be depressed is smaller than the Hamlet-​cardinality. It says this not as a matter of its strict literal content but as a matter of its pseudo-​saturated content. If that is right, then we don’t need to posit fictional characters or any other object for “Hamlet” to refer to, to make sense of the intuition that (3) says something that is true—​really and truly, true, rather than just pretend true. And even though (3) is not an authorized move in a non-​veridical interpretive game anchored by Hamlet, the truth expressed by an utterance of (3) very much depends on what moves are authorized in such a game.12 It is worth making a brief digression from the main thread of my argument to make clear the exact nature of this proposal. Some have misunderstood my approach as suggesting that interlocutors are confused about what is being communicated in utterances of sentences like (4) or (5)—​that they mistake pragmatically conveyed contents for semantic contents. But I hope it is clear that it is no part of my conjecture that conversational interlocutors are in any way confused about either what is communicated by utterances of (4) or (5) or how what is communicated is communicated. Quite the opposite is true. The conjecture presupposes that knowledgeable and linguistically competent adult speakers of English will typically know, for example, that “Santa Claus” is a name, but a name that stands for no real existent. They can also be expected to know that names, at least when they do refer, are typically used to make assertions about the objects for which they stand. Just such knowledge is implicated in the generation of what I am calling one-​and-​a-​ half-​stage pragmatic externalities. When a speaker uses a name that stands for no object in an attempt to say something true about the world, informed interlocutors are perfectly capable of reasoning on the basis of this background knowledge to conclusions about what the speaker is “getting at”—​ even though they know that she has made no fully propositionally contentful claim about any real existent. 12 Though the details are different, my view has certain affinities with the views of Sainsbury (2005).

250  Referring to the World Though I am not prepared to say that speakers themselves are confused about such matters, too many theorists, it seems to me, may well be confused about how exactly to draw the line between semantics and pragmatics. Too often theorists suppose, without any real empirical or theoretical basis, that certain intuitions about the acceptability or unacceptability of utterances of sentences like (4) or (5) must be intuitions about the strict literal truth or falsity of the proposition “directly” expressed by any such utterance. But I suspect that such misdiagnoses of the nature and source of our common-​sense intuitions may well be sources of many philosophical errors. What is clearly needed are better, more theoretically and empirically well-​grounded tests and metrics by which to separate out which of our intuitions are tracking strict, literal truth and falsity and which of our intuitions are tracking other sorts of extra-​semantic or pragmatic felicities and infelicities. There are, of course, a fair number of extant, reasonably successful, reasonably well-​ grounded diagnostics that go at least partway toward helping us chart the interface between pragmatics and semantics. I have no objection, for example, to using cancellability as a diagnostic to help distinguish between what is asserted and what is merely implicated or to using persistence or lack of persistence under embedding to help us distinguish what is presupposed from what is asserted. Still, I  suspect, and have argued at length elsewhere, that the current kit of available diagnostic tools, even taken in their totality, are insufficient to solve either what I call the “demarcation problem” or the generation problem for pragmatics and semantics. By the demarcation problem, I  mean the problem of distinguishing “pragmatic externalities,” as I  call them, from ingredients of what might be called narrow semantic content within the totality of contents communicated by an utterance in any given speech situation. By the generation problem, I mean an account of how the pragmatic externalities are generated, processed, and recognized within a given speech situation. Now I suspect that the demarcation problem and the generation problem, though conceptually distinct, are deeply methodologically intertwined. To adequately demarcate pragmatic externalities from semantic contents, we must show how such externalities are generated on the basis of a combination of facts about syntax, meaning, context, and situation. But it is likely that a comprehensive theory of total utterance content will be only holistically testable against the total evidence. And this may mean that we will have few direct or local empirical diagnostics for disentangling semantics from pragmatics.

The Things We Do with Empty Names  251 To appreciate some of my reasons for suspecting this may be so, consider Francois Recanati’s (1993, 2004) distinction between primary and secondary pragmatic processes—​or, as I prefer to say, between “pre-​propositional” and “post-​propositional” pragmatics—​and some of the things he has said about it. I  have already leaned on this distinction to a degree both here and in Chapter 3 when discussing what I called “semantic partiality” in the course of my argument against jazz combo theorists. In his own original formulation, Recanati construed the distinction in a way closely tied to a view about linguistic processing. He defended what he called the “Availability Principle,” saying that according to this principle, “in deciding whether a pragmatically determined aspect of utterance meaning is part of what is said, that is, in making decisions concerning what is said, we should always try to preserve our pre-​theoretic intuitions on the matter.” (1993, 248) He conjectured that hearers are non-​inferentially aware of strict literal contents, and that, consequently, primary pragmatic processes are fast, unconscious, and modular. They are sometimes mandatory as in the case of saturation, but they are also sometimes optional as in the case of what he calls “free enrichment.” By contrast, he argued that hearers are inferentially aware of pragmatic externalities. And he insisted that in contrast to primary pragmatic processes, secondary pragmatic processes are slow, optional, conscious, and non-​modular. The payoff of this way of thinking was supposed to be that we have a systematic and tractable way of demarcating, at least in part, the semantic from the pragmatic, and of explaining how at least certain sorts of pragmatic externalities are generated. And this was all supposed to be grounded in a theory of linguistic processing. Recanati argued that if hearers are non-​inferentially aware of a content C then C is an ingredient of the strict literal content of U, while if hearers are inferentially aware of C, then C is a pragmatic externality of U. The problem, as I have argued at greater length elsewhere, is that although a distinction between primary or, as I prefer to say, pre-​propositional pragmatics, and secondary or, as I prefer to say, post-​propositional pragmatics, is well enough motivated, it cannot be tied to differences in processing in the way that Recanati imagined. The problem is that the hearer’s presumed “pre-​propositional knowledge”—​by which I mean what a hearer must know if she is to successfully cognize the contextually determined propositional content of an utterance of a semantically indeterminate sentence—​and the hearer’s presumed “post-​propositional knowledge”—​by which I mean what a hearer must know in order to cognize the post-​propositional pragmatic externalities of the speaker’s utterance—​are often of exactly the same general

252  Referring to the World character. But if that is right, it seems unlikely that there will be a principled distinction between the character of the hearer’s pre-​propositional processing and the hearer’s post-​propositional processing. If not, it follows that we cannot use anything like Recanati’s Availability Principle to help trace the boundaries of the interface between semantics and pragmatics, or even between pre-​propositional and post-​propositional pragmatic externalities. An example may help drive home the point. Consider the semantic and pragmatic differences between (6) and (7) below: (6) Have you had lunch? (7) Have you had sex?

Semantically, a typical or default utterance of (6) expresses a recency question. That is, it is much more likely to express the question of whether you have had lunch this afternoon rather than the question of whether you have ever had lunch in your life. Pragmatically, an utterance of (6) is likely to express an indirect invitation to lunch or perhaps an offer to make lunch or fetch lunch. By contrast, an utterance of (7) is certainly less likely to be taken to semantically express a recency question and less likely to pragmatically convey an indirect offer or invitation. The crucial point is that what each is likely to semantically express and what each is likely to pragmatically convey would seem to depend directly on mutual knowledge of shared sexual practices and shared eating practices. People tend to eat lunch daily and tend to ask casual acquaintances, business associates, and dear friends to join them for lunch. By contrast, most people do not have sex daily and when they do have it, they typically have sex with a much narrower range of partners. The minimal takeaway lesson from this rather lengthy digression is just that drawing the line between pragmatics and semantics is a tricky thing indeed. On my view, pragmatics happens everywhere. There is no single pragmatics/​semantics interface and thus no single diagnostic that can reliably solve both the demarcation problem and the generation problem. Pragmatics happens, for example, in the generation of the truth-​conditional contents of sentences like (6) and (7). It happens in the gap between sentence meaning and total utterance content as in one-​and-​a-​half-​stage pragmatics. It happens post-​propositionally as in the generation of conversational implicatures. Nor is that an exhaustive list of the way in which pragmatic factors operate. I have argued at length elsewhere, for example, that pragmatics also happens in what I call “contextually driven thematic expansion” via the process of what

The Things We Do with Empty Names  253 I call “pragmatic adjunction.” Pragmatics can even occur, as we have seen at various stages in the overall argument of this book, pre-​semantically when only syntax is at issue. Now my aim in this book has not been to systematically chart all the ways in which syntax, semantics, and pragmatics cooperate to generate the total communicated contents of our utterances. But it is important to understand that the pragmatic arguments of the current chapter are really just a small part of a more complex and more comprehensive account of the nature of the interface between semantics and pragmatics. It is against that backdrop that we end our digression and return to the main line of argument of the current chapter. Turn now to yet another sort of communicative pragmatics—​the cooperative play of non-​veridical language/​thought games. It is within such language/​ thought games that empty names are most naturally at home. Such games are structurally and dynamically similar to veridical language-​thought games. But when we play such games, we are engaged in a rather different sort of communicative enterprise. For one thing, such language/​thought games are paradigmatically played with merely objectual rather than fully objective representations. In the context of such games, the lack of successful reference to a real-​world existent is no mark of any sort of failure. In fact, names with real-​world reference are sometimes taken up within such games, but when they are, authorized moves with sentences containing that name are generally not required to track the facts about the relevant object. One might say that in the play of non-​veridical language/​thought games, a name with a real-​ world reference is used as if it were merely objectual—​that is, merely referentially fit and not referentially successful. This happens, for example, when one creates an entirely fictional narrative about a “real” person using that person’s actual name. In that case, there is no need to deny that one is referring to that very person. It is just that in the context of a non-​veridical language game, the governing concern that serves to determine which moves in the game are authorized is typically not strict literal worldly truth. When we play non-​veridical language games with singular terms that enjoy real-​world reference, we may be authorized by the rules of the game we are playing to make moves that enjoy game-​internal correctness but are strictly false of the actual world. More typically, though, the singular terms we deploy in non-​ veridical language games will be altogether devoid of real-​world reference. And sentences containing such expressions will fail to express any complete and determinate proposition. Consequently, game-​internally correct moves will be neither true nor false of the actual world.

254  Referring to the World It should be clear that many non-​veridical language games are grounded in explicit pretense. When we engage in explicit pretense, there is little basis for confusing what is merely game-​internal correctness with the worldly property of being strictly and literally true. To the extent that we are self-​ consciously and explicitly pretending, we typically know that we are. We know that in making “authorized” moves in the game, we are not making moves that purport to track how things objectively are by a realm of game-​ independent, real existents. This is not, however, to say that anything goes within such games. For even an explicit pretense may be subject to some more or less restricting constraints, generated by the governing concern that is operative within the pretense. Still, moves within a pretense enjoy characteristic degrees of freedom that are not enjoyed by moves in veridical games. One might even say that such freedom is the very hallmark of non-​veridical language games as such. That seems right for certain sorts of non-​veridical games. But that should not prevent us from noticing that moves within a non-​veridical game may be so tightly constrained by the governing concern and the rules operative within the game that they altogether lack the phenomenological feel of ­anything goes explicit pretense. In fact, moves within certain non-​veridical games can carry a high degree of rational compulsion. And rational compulsion is the very opposite of the sort of freedom from constraint that one might suppose is characteristic of explicit, self-​conscious, anything goes pretense. When non-​veridical language games lack the feel of pretense, and when permissible moves within such games present themselves to us as rationally required, that is precisely when the danger arises that we will confuse mere game internal correctness with objective, mind-​independent truth. In such cases, we are prone to believe that rational compulsion must somehow be rooted in forms of reasoning and inquiry that somehow make truth-​tracking contact with the objects themselves. The paradigm example of the kind of domain I have in mind is pure, as opposed to applied, mathematics. Moves in pure mathematics often enjoy a very high degree of felt rational compulsion. Indeed, given a choice of axioms and definitions, together with a background logic, moves in pure mathematics are often rationally inescapable. As a consequence of the high degree of felt rational compulsion enjoyed by its authorized moves, pure mathematics does not wear its standing as non-​veridical on its phenomenological sleeve. Phenomenologically speaking, pure mathematics may in fact seem to be the exact opposite of anything-​goes explicit pretense. This is not to

The Things We Do with Empty Names  255 deny that even in pure mathematics, there is a degree of freedom with respect to the choice of axioms and definitions, and perhaps even with respect to the background logic itself. But to whatever extent such matters are up for grabs, once they have been settled, permissible moves within mathematical games are often highly constrained. Of course, even non-​veridical games that involve explicit, self-​conscious pretense have their limiting constraints on permissible moves. But even so, in such games, permissible moves typically enjoy nothing like the degree of rational inescapability characteristic of moves within pure mathematics. This fact alone may lead one to suppose that in doing pure mathematics, we are doing something entirely different from what we are doing, pragmatically speaking, when we play the sort of non-​veridical language games that involve explicit and self-​conscious pretense. But to the extent that any such supposition rests on a comparison of the phenomenology of pure mathematics to the phenomenology of explicit, self-​conscious pretenses, it is mistaken. Indeed, I claim that despite the phenomenological differences between pure mathematics and explicit, self-​conscious pretense, the games we play when we engage in explicit, self-​conscious pretense and the games we play when we do pure mathematics are a species of a common genus. And there are, I shall argue, many reasons for believing and few for denying that when we engage in pure, as opposed to applied, mathematics, we are playing a non-​veridical language game, deploying merely objectual, rather than fully objective representations and that our governing concern is not a species of strict, literal truth but a species of truth-​similitude, a game-​internal right and wrong of the matter. It is important to distinguish, though, the truth-​relevant character of applied mathematics from the truth-​relevant character of pure mathematics. Applied mathematical statements of the form (8) 7 cups + 7 cups, yields 14 cups

express strict, literal truths about physical quantities. More generally, there are many mind-​independent truths about physical quantities and magnitudes. It can be strictly and literally true that there are 14 cups on the table, that E = MC2, that E = hv, and so on. The language of mathematics, as applied to physical quantities and magnitudes, enables us to state such truths. Indeed, I would argue that the language of mathematics, as applied to physical quantities and magnitudes, is well-​nigh indispensable for science. But

256  Referring to the World to grant the indispensability for science of mathematics as applied to physical quantities and magnitudes is not to grant the truth of statements of pure mathematics. Accepting the strict literal truth of (8) above does not require us to accept that (9) is strictly literally true as well: (9) 7 + 7 = 14.

Platonists will disagree. They will insist on the real existence of a mathematical realm distinct from the realm of physical reality. They imagine that this realm is populated with free-​standing, mind-​independent mathematical entities such as sets or numbers. I won’t argue the point at length here—​since doing so would carry us very far afield—​but I hold that there is neither a metaphysical nor an epistemic basis for believing in such entities. What clearly exist are numerals. And numerals are bone fide singular terms. As such, they enjoy intrinsic singular referential purport. But numerals are not referentially successful singular terms. They do not succeed in actually referring to anything at all. If that is right, then a statement like (9) expresses no determinate proposition but only a proposition in waiting. And propositions in waiting, I have already argued, are neither true nor false. They stake out no determinate claim about how the world is. But it should be clear by now that to deny that (9) is strictly, literally true is not to deny that it enjoy a species of truth-​similitude—​that is, a species of game-​internal correctness. We might put the point by saying of (9) that although it is “mathematically true,” it is not objectively true. This view amounts to a form of fictionalism about “pure” mathematics together with a form of realism about applied mathematics. Giving a full account of what I have in mind would require us to delve very deeply into matters both semantical and metaphysical, including not just the semantics of both pure and applied mathematical language but also both the metaphysics of physical quantities and magnitudes and the metaphysics of mathematical quantities and magnitude. That task would require a book-​length argument all of its own, considering issues beyond the purposes of the current volume. Very briefly, though, I will say that I maintain that although the language of pure mathematics makes no intrinsic reference to specifically physical quantities and magnitudes, nonetheless, through physicalistic (re)interpretation of the language of mathematics, mathematics is made applicable to the physical world. When interpreted so as to apply directly to the physical world, mathematics enables us to state many objective truths

The Things We Do with Empty Names  257 about the physical quantities and magnitudes. And I suspect that there may well be no other way of stating various truths about the quantitative structure of the natural world except by way of the language of mathematic as physicalistically interpreted. But acknowledging the practical indispensability of applied mathematics for purposes of total natural science does not directly require us to believe in a realm of free-​standing, purely mathematical entities that serve as the truth-​makers for statements of pure mathematics. Real physical quantities need not stand in real relations to putatively abstract mathematics in order for the statements of applied mathematics about physical quantities to be true. There can be strict truths about the number of planets, for example, without there being any numbers—​considered as free-​ standing mathematical entities—​to number the planets. Since I lack space to even begin to argue for these admittedly provocative claims in any detail, I will concede that one is entitled to raise some pressing and difficult questions about the package of views I have just sketched. For example, if, as I claim, there are no mind-​independent mathematical objects, with a mathematical nature wholly independent of us and our practices, by which we might somehow seek, by some epistemic methods or other, to constrain our mathematical moves, then it is fair to wonder how moves in pure mathematics possibly manage to be both tightly constrained and rationally compelling. It is also fair to ask whether the combination of realism about applied mathematics and fictionalism about pure mathematics is ultimately coherent. To be sure, we must distinguish in this connection between claims about the epistemic credentials of the methods of pure mathematics from claims about the ontology of mathematics. I do not deny that the methods of mathematical proof enjoy sterling epistemic credentials. But we should not allow those sterling credentials to bewitch us into ontological profligacy. Neither the epistemic credentials of pure mathematics nor even the profuse applicability of mathematics, as physicalistically interpreted, to the natural world directly entails that pure mathematics enables us to get at truths about a realm of mind-​independent objects. Though I  do not pretend to have conclusively settled these matters in this chapter or even in this book, by way of offering a down payment on the promissory notes I have taken out, I conjecture below that the temptation to posit a realm of free-​standing mathematical entities results from what I have called illusions of subject matter. Illusions of subject matter may, I suspect, be nearly inevitable outgrowths of certain kinds of non-​veridical language games. Such illusions are born of a number of confusions. For example,

258  Referring to the World there is a tendency to mistake merely notional objects, as I call them, for free-​standing real existents. And there is a tendency to conflate mere truth-​ similitude with genuine truth. I shall say more about the idea of a merely notional object and illusions of subject matter shortly. But let me turn first to a brief discussion of the distinction between truth and truth-​similitude.

5.  Truth versus Truth-​Similitude We begin by focusing more closely on what it is for a move within a non-​ veridical language game to be “authorized.” My focus here will not be on how, in detail, authorization works within this or that sort of non-​veridical language game. My aim is, rather, to clarify the very idea of an authorized move within a non-​veridical language game. First, a negative point. Making an authorized move in a non-​veridical game is not the same as asserting a strictly true proposition. When a move is authorized within a game, it does not thereby enjoy some peculiar species of the genus truth—​“truth” in a fiction, say. At least it does not do so merely in virtue of being authorized. On my view, truth is not properly construed as a genus of which there are many species. I say this even while granting that we do sometimes say such things as “It is true in the Holmes stories that . . .” or “It is true according to the Santa myth that . . .” Admittedly, in one deep and important respect our use of such expressions does resemble our use of genuine truth talk. Such expressions play within non-​veridical games dialogic roles quite similar to the dialogic role played by the genuine truth predicate in veridical games. In veridical games, “is true” functions as a device for claiming entitlements and attributing commitments. For example, when I say that it is true that snow is white, I thereby claim an entitlement to assert that snow is white. In a parallel fashion, the predicate “true in story S” functions as an entitlement-​claiming device with respect to assertion-​like moves within non-​veridical language games governed by the story S. It is because the dialogical function of the predicate “true in S” precisely mirrors the dialogical function of the genuine truth predicate “true,” though without sharing its metaphysical nature, that I count truth in a story as a species of truth-​similitude rather than a species of truth. More controversially, I  also take mathematical truth, as applied to statements of pure mathematics, to be a species of truth-​similitude rather than a species of genuine truth. Within mathematical practice, there is

The Things We Do with Empty Names  259 certainly a game-​internal right and wrong of the matter. Moreover, at the very least, the use of the mathematical truth predicate within the realm of pure mathematics functions as a device for claiming entitlements and attributing and expressing commitments to certain mathematical moves. But I insist on the further claim that because there are no free-​standing mathematical entities for mathematical terms to refer to, mathematical truth cannot be a species of strict literal truth. Like “true in the story,” “mathematically true” lacks the robust metaphysical nature, while sharing the dialogical character, of genuine truth. But I hope it is clear by now that I do not mean thereby to deny to pure mathematics deep cognitive significance or sterling epistemic credentials. Indeed, the methods of proof deployed in pure mathematics enjoy the highest of rational credentials. Moreover, once the language of mathematics is physicalistically interpreted and systematically applied to the natural world, mathematics is an all but indispensable cognitive tool for achieving the explanatory ambitions of total natural science. Now at superficial first glance, my distinction between truth and truth-​ similitude may appear to be something of a hostage to the debate between deflationists and inflationists about the metaphysical nature of truth. Deflationists regard truth as little more than a compliment that we pay to such statements as we are prepared to assert.13 So understood, truth would have no substantive metaphysical nature. We have said all there is to say about truth, according to the deflationist, when we have fully specified and explained the dialogical role of the truth predicate. If one is convinced that even the genuine truth predicate has no substantive metaphysical nature, then one may appear at first blush to have grounds for rejecting outright my distinction between truth and mere truth-​similitude. After all, my central claim is that it is in virtue of their distinct substantive metaphysical natures and not in virtue of distinct dialogical roles that truth is distinguished from mere truth-​similitude. But if all there is to truth talk is its dialogical role, then there would appear to be nothing to distinguish truth from mere truth-​similitude. Moreover, if the distinction between truth and mere truth-​ similitude goes, so, it may seem, does the distinction between veridical and non-​veridical language games. But this conclusion is premature. Suppose that the deflationist was to win the day with respect to the metaphysics of truth. We would still need some basis for distinguishing mere referential purport from full-​blown referential

13 This characterization of deflationism is due to Richard Rorty (1986).

260  Referring to the World success and thus of distinguishing merely objectual representations from fully objective ones. The deflationist has to tell a story that preserves the status of names like “Santa Claus” as genuine names and thus as would-​be referring devices—​ones that, while failing to actually refer, still purport to refer. And she has to distinguish such names from names like “Barack Obama”—​names that succeed in referring to some real existent. As long as we can make out such a distinction, whether on a deflationary basis or an inflationary one, it is a short step to our two further distinctions between veridical and non-​veridical language games, on the one hand, and truth and mere truth-​similitude on the other. To be sure, if the deflationist is correct about the nature of truth, then it may seem to follow that we will be unable to make out the distinction between mere referential purport and full-​blown referential success in robustly metaphysical terms. The difference between “Santa Claus” and Barack Obama would then have to be not that the one stands in a substantive metaphysical relation to a real existent while the other does not. On deflationary views of reference, referring involves no such relation. But even if the deflationist cannot coherently accept an inflationary theory of reference, that does not obviate her need for some way of distinguishing between mere referential purport and full-​blown referential success. It would just require that she make that distinction without appealing to a robustly metaphysical notion of reference. But this means that trio of distinctions on which my arguments turn does not stand or fall with deflationism. Now it is no burden of our own to make out the relevant distinctions on the behalf of the deflationist. If the deflationist cannot find a principled basis for our distinction within a deflationary framework, so much the worse for deflationism.14 14 It is important to distinguish deflationism about truth from deflationism about reference. Though deflationists about truth are often deflationists about reference as well, I take it to be logically possible to accept a deflationary theory of truth while endorsing an inflationary theory of reference. I admit, however, that there is something of an unsettled divide among philosophers on this point. Horwich (1998), Field (1994), and Brandom (1994) are all examples of deflationists about truth who also endorse deflationism about reference. See Boghossian (1990) for an explicit argument to the effect that a robust theory of reference already entails a robust theory of truth. If Boghossian is right, it would seem to follow that one cannot coherently combine a deflationary theory of truth with a robust theory of reference. Boghossian arguments are, however, primarily intended to undermine the coherence of non-​factualism (about content)—​the view that sentences containing certain sorts of problematic but meaningful expressions are not truth apt. Because I take sentences containing empty names to be neither true nor false, but devoid of truth value, there might seem to be a sense in which, on my view, sentences containing empty names fail to be truth apt. But I do not think that the sort of failure of truth-​aptness that results from reference failures falls prey to any argument like Boghossian’s—​since this sort of failure is perfectly consistent with both truth and reference being metaphysically substantial. For arguments to the effect that one can coherently defend a minimalist theory of truth and a non-​minimalist theory of truth-​aptness and thus, by extension, a

The Things We Do with Empty Names  261

6.  Truths about Fictions Distinguishing between truth and truth-​similitudes does not require that we deny that there are truths—​strict, literal truths—​about works of fiction. That Hamlet was a play, that it was written by Shakespeare, that it is about what it takes to live a distinctively human life in the face of resignation and despair, may be strictly literally true or false. The play itself is a real, though merely secondary, existent and there are many truths about it. As secondary, the existence of the play is in a sense derivative from an array of more primary existents on top of which the play sits. But, of course, absent the status-​conferring powers of the human mind, not even these primary existences would suffice to usher the play into existence. Now the more primary existents on top of which the existence of the play sits are themselves mere representations and repeatable structures of representations. The play exists because the mind confers a certain normative status—​the status of being a play—​upon that repeatable structure of representations. Once the status of being a play has been conferred upon a certain repeatable structure of representations, recurrences of that very structure—​in either written texts or performances on stage—​count as occurrences of that very play again. Because the play Hamlet is a real, though merely secondary, existent that has tractable and substantive identity conditions, it is the kind of thing that can have real properties and stand in real relations. Consequently, nothing bars us from making either true or false statements about the real though secondary existent that constitutes the play Hamlet. Earlier we raised the question of whether having already conceded the existence of the play as a kind of secondary existent, we shouldn’t also grant that in the writing of the play Hamlet, Shakespeare managed also to usher into being the universe of fictional entities that the play invites us to imagine? Just as the play exists as a real, though secondary existent, why not say that Hamlet himself—​the fictional crown prince of Denmark—​exists, as a real though secondary existent? Admittedly, a taste for relatively barren ontological landscapes may make us unwilling to extend our ontological permissiveness quite that far. But as I suggested earlier, the question is not really one of taste but of principle. The crucial question is whether there is any principled

non-​minimalist theory of reference, see F. Jackson, G. Oppy, and M. Smith (1994) and Holton (1993). Given the burdens I have and have not taken on in this chapter, I can afford to be neutral on these issues.

262  Referring to the World basis for counting the play Hamlet a real existent while denying real existence to the fictional Crowned Prince, whose comings and goings the play is putatively “about.” I approach this last question in what may at first seem a roundabout fashion. I  readily acknowledge the real existence of the names “Hamlet,” “Laertes,” and “Ophelia,” to list just a few of the names for Shakespeare’s cast of characters. Qua names, such fictional names are intrinsically indistinguishable from any other name. Such names enjoy intrinsic referential purport in the same way and for the same reasons that names in general do. Like names in general, fictional names in fictional contexts function as devices of explicit coreference in language and devices of same purport in thought. This last fact is crucial for understanding why the use of fictional names in fictional contexts should give rise to illusions of subject matter. Like names in general, names in fiction are in a sense “for” talking about objects. Especially when our deployment of a name is coherently regulated by some rationally compelling name-​using practice—​even where the regulating practice is non-​ veridical—​our use of names will still have the feel of object talk. But where names are purely fictional, the feeling that our talk is talk of actual objects is something of an illusion. There are typically no objects we are talking about when we deploy fictional names in fictional contexts. There is no reason to bloat our ontologies by suggesting that the mere writing of a play or novel can somehow or other usher such entities into existence.

7.  Notional Objects and Illusions of Subject Matter I have already acknowledged that nothing bars us from associating merely notional objects, as I have called them, with the use of a merely objectual or referentially fit, but not referentially successful, representation as it is used in the context of a non-​veridical language game. Notional objects are constructions out of coreference relations among such names. Because merely fit representations used in non-​veridical contexts do not actually succeed in referring to any real existent, there is a clear sense in which two such representations cannot be said to corefer either. But even empty names have intrinsic purport to refer. And if that is right, there is a sense in which tokens of the same non-​referring name again can enjoy an intrinsic purport of coreference, despite the fact that they actually neither refer nor corefer.

The Things We Do with Empty Names  263 We exploit this insight to construct a collection of merely notional objects. We proceed in two stages. First, we define, CRS (n), the S-​relative coreference set for a name n, where S is the closure of the set of assertion-​like moves that enjoy truth-​similitude in some non-​veridical language game G. S-​relative coreference sets encode both what we might call “intrinsic relations of purported coreference” and what we might call “extrinsic relations of purported coreference” among the singular representations deployed in S. Notional objects are, in effect, pleonastic reifications of such relations of purported coreference.15 CRS(n) will contain every name that either intrinsically or extrinsically purports to corefer with n in the scenario S. Since tokens of the same name again enjoy intrinsic purport of coreference, for any name n, and any scenario S, n will be a member of its own S-​relative coreference set. This guarantees that for any scenario S, CRS (n) is non-​empty. On the other hand, if m and n are distinct and therefore referentially independent names, then m ∊ CRS (n) just in case S |-​ ⌈m = n⌉. That is, two referentially independent names, m and n will be members of the same coreference set just in case the scenario S “licenses” the identity statement ⌈m = n⌉. When m and n are referentially independent—​and so do not enjoy intrinsic purport of coreference—​but the identity statement ⌈m = n⌉ is licensed by S, I will say that m and n enjoy extrinsic purport of coreference in S. S-​relative extrinsic coreference is both symmetric and transitive. That is, m ∊ CRS(n) just in case n ∊ CRS(m). And if m ∊ CRS(n) and n ∊ CRS(o), then m ∊ CRS(o). Finally, we note that scenario S may be incomplete in the sense that for distinct and therefore referentially independent names, m and n, S may leave it open whether ⌈m = n⌉ or ⌈~(m = n)⌉ . Again, notional “objects” are not genuine objects. They are nothing but pleonastic reifications of relations of intrinsic and extrinsic purport of coreference. We might say that an S-​relative notional object is that which the 15 My pleonastic reifications bear a certain family resemblance to the “pleonastic entities” of Schiffer (2003). Schiffer’s pleonastic entities, by contrast, are supposed to be free-​standing abstract entities, though entities with very minimal natures. Some such entities, like fictional characters, are created by human activity—​such as the “pretend” use of a name in the context of a fiction. More generally, pleonastic entities are generated by what Schiffer calls something from nothing transformations. On my view, by contrast, pleonastic reifications do not yield genuine, free-​standing real existents at all. The results of pleonastic reification are, rather, entirely dispensable façon de parler. They have no independent substantive nature that outstrips the construction by which they are generated. We have said all there is to say about them when we have given the rules by which they are constructed. Given that Schiffer says that his pleonastic entities have “minimal natures,” one might well wonder whether the difference between Schiffer’s views and my own are substantive or merely verbal. That is a question I lack space to consider here, but the less robustly ontological one takes Schiffer’s pleonastic entities to be, the closer they will be in spirit to my merely notional objects.

264  Referring to the World members of a given S-​relative coreference set one and all denote*. But denote* should not be understood as a genuine reference relation. Denote* does not relate words to real-​world existents. Denote* is what you get when you subtract out the real existents from one side of genuine reference relations. You get something like purport of reference and coreference. Similarly, a notional object is what you get when you suppose that the totality of purported reference and coreference relations suffices to give you an object. Notional objects are thus a kind of optional encoding and/​or score-​keeping device. They enable us, for example, to bind together fictional contents under file-​like structures linked to the use of fictional names. We may, if we like, identify fictional characters with such notional objects. But if we do, we need to take care not to confuse fictional characters with the genuine referents of fictional names. As I have already made clear, on my view fictional names typically have no referents at all.

8.  What Objects Are Not My notion of a notional object bears a certain rough affinity to this Fregean thought: If we are to use the sign a to signify an object, we must have a criterion for deciding in all cases whether b is the same as a, even if it is not always in our power to apply this criterion. (Frege 1893/​1974, x)

Frege is here suggesting that an identity statement expresses what is contained in a recognition judgment—​a judgment to the effect that one has been given the same object again. Frege’s further thought is apparently that we have succeeded in using a sign to designate a determinate object just in case we have fixed a significance for each identity statement in which a given singular term may occur. We thereby specify what it is for any two terms to (correctly) purport to stand for the same object. This approach is supposed to enable what the epistemic givenness of numbers could possibly consist in—​ despite the fact that we have, as Frege admits, neither “ideas” nor (sensory) “intuitions” of them. The thought seems to be that numbers, considered as real existents, are given to us through the use of singular terms. Indeed, though he is directly concerned here with numbers per se, he seems here to be endorsing a perfectly general claim about the very concept of an object

The Things We Do with Empty Names  265 as such, to the effect that the concept of an object as such is nothing but the concept of that which is given through the use of a singular term. Objects are, in effect, the shadows cast by the uses of singular terms, paradigmatically in identity statements.16 One finds a similar idea in recent work by Robert Brandom (1994). It is worth pausing over his approach in order to place in sharp relief a certain fundamental divide about the nature of objects and of their relations to objectual representations. Brandom seems to endorse a version of idealism according to which objects are nothing but constructions from or projections out of purported relations of reference and coreference among singular terms. But this approach to objects has, I shall argue, all the benefits of theft over honest toil. Brandom begins promisingly enough by appealing to something like our distinction between the mere purport of reference and/​or coreference and actual successful reference and/​or coreference. But he ends with a notion of existence and of objecthood that seems to me to strip both of any substantive metaphysical nature. For example, he says: It is not enough . . . to explain only successful reference. . . . [S]‌ingular terms are expressions that, in Quine’s useful phrase, “purport to refer to just one object.” The qualification expressed in this slogan by the use of “purport” has two different functions: to acknowledge the notorious possibility that a name or a definite description may fail in its referential bid . . . and to exclude accidentally singular expressions. (Brandom 1994, 360)

One is entitled to ask what is supposed to bridge the gap between purport and success, on Brandom’s view. His basic thought may seem reasonable enough. Using our own vocabulary rather than his, Brandom’s basic thought might be put this way: a referentially fit expression e is referentially successful if there exists an object o to which e purports to refer. That is to say, if e enjoys referential purport, and if what e purports to refer to actually exists, then e succeeds in referring to that very object. On Brandom’s approach, referential success is thus analyzed in terms of referential purport and existence. But as thus framed, there are two distinct sides to our question: What is it for an object to exist? What is it for an expression e and an object o to be so related that e purports to refer to o?

16 See Taylor (2019) for more on Frege’s understanding of objects.

266  Referring to the World We have already seen back in Chapter 1 and Chapter 3 that Brandom is deeply skeptical about the reference relation. Recall his claim that the reference relation is “a philosopher’s fiction, generated by grammatical misunderstandings.” But if so, it would seem to follow that there is also no such real word-​world relation as the relation of purporting to refer in which a term might stand to a worldly object. But if Brandom does reject the very idea of a metaphysically substantive purporting-​to-​refer-​to relation between words and things, the only real issue for him must be the issue of existence. If we can say what it is for an object to exist, we are done. We apparently have no further need to be puzzled about what it is for a given term to purport to refer to this object rather than that. According to Brandom, the key to explaining referential success without appeal to a substantive relation of either reference or purported reference is the notion of what he calls a “canonical designator.” A canonical designator is a designator the well-​formedness of which guarantees its referential success. The notion of a canonical designator is supposed to help explain what it is to be committed to the existence of a kind of object, according to Brandom. Roughly, to be committed to the existence of a kind of object is to treat a certain family of designators as canonical designators. Numerals, for example, are supposed to be canonical designators of the numbers. As he puts it, the issue of the success of their singular referential purport does not arise for expressions such as “121” and “161” in the same way that it does for expressions such as “the smallest natural number such that every larger one is the sum of distinct primes of the form 4n + 1.” It is to take a frankly inegalitarian approach to referential purport and its success. Numerals are semantically privileged ways of picking out numbers. By contrast to definite descriptions of numbers, the well-​formedness of numerals suffices for their referential success, guaranteeing that they pick out a corresponding object. Furthermore, distinct numerals are guaranteed (emphasis added) to correspond to distinct objects. According this privileged status to a class of singular terms is treating them as canonical designators of a kind of object. (Brandom, 1994, 431)

Now a family of canonical designators is supposed to define “a structured space of addresses” to which objects of the relevant sort can be assigned in some principled and rule-​governed way. To say that o exists is to say that

The Things We Do with Empty Names  267 there is some address in some structured space of addresses to which o may be assigned. Thus, to say that some numerical expression succeeds in referring—​to say that a number corresponding to it exists—​is to say that it has some address in the structured space mapped out by the successor numerals. (Brandom 1994, 441)

Now it is important for Brandom that different families of canonical designators define distinct structured spaces of addresses so that, for example, physical existence is defined over a different structure of addresses than numerical existence. As he puts it, To say that some physical object expression succeeds in referring, that the object it designates exists, is to say that it exists somewhere in space and time, that it occupies some spatiotemporal region. This is to say that it has some address in the structured space of spatiotemporal coordinates centered on the speaker. (Brandom 1994, 444)

There are also sets of fictional addresses defined by families of fictional designators: Fictional existence, existence in or according to a story, can be understood as having the same shape as that common to physical existence and to the various sorts of numerical existence. To say that in or according to Sherlock Holmes stories Holmes’s housekeeper exists (or to say that the expression “Holmes’s housekeeper” succeeds in referring to an individual) is to say that that expression is intersubstitutable with some singular term that actually appears in the story. . . . The singular terms that appear in the text that define the fictional context can be considered as the canonical designators. (Brandom 1994, 446)

So, by Brandom’s lights, “exists” turns out not to be a fully univocal expression. He does insist, however, that “different sorts of existence, or even senses of exists, have a structure in common that qualifies them [all] as notions of existence.” Brandom claims that there is “no question” whether the numeral “2” is referentially successful, and thus no question whether the number two exists.

268  Referring to the World By contrast, he does allow that the question of success does arise for a non-​ canonical designator like “the smallest natural number such that every larger one is the sum of distinct primes of the form 4n + 1.” Though such non-​ canonical designators are sometimes referentially successful, their referential success is not automatic. The referential success of a non-​canonical designator of an object can be established, Brandom claims, only by somehow deriving the truth of some non-​trivial identity statement linking it to one or another canonical designator. Thus, the referential success of the non-​ canonical designator “the smallest natural number such that every larger one is the sum of distinct primes of the form 4n + 1” consists in the truth of the following identity: the smallest natural number such that every larger one is the sum of distinct primes of the form 4n + 1 = 121.

Brandom’s approach enjoys one clear advantage over the approach of the early Frege cited above. It allows for the possibility of empty singular terms. An empty singular term is a non-​canonical designator that is not linked to any canonical designator via any true, non-​trivial identity statement. If one admits that there are singular terms through the use of which no existent is given, while insisting, with the early Frege, that the concept of an object is nothing but the concept of that which is given through the use of singular terms, that objects are nothing but shadows cast by singular terms, then Brandom’s account of referential success may seem nearly inevitable. Indeed, his “frankly inegalitarian approach to referential purport and its success” is apparently intended to reconcile the evident conflict between the thought that objects are nothing but shadows cast by singular terms and the thought that some bona fide singular terms fail to cast what we might call objective shadows. Not just any singular term will cast objective shadows. Only a special class of terms do so, viz., the canonical designators, together with those non-​canonical designators that can be linked to them in some systematic way. But it seems to me that Brandom’s approach to referential fitness and success enjoys the benefits of theft over honest toil. We are never told just what is supposed to justify treating a certain family of designators as canonical. We are told that there are certain disjoint families of canonical designators—​in fiction, in mathematics, in physical object talk. And we

The Things We Do with Empty Names  269 are told that each of these families defines a distinct sense of existence by defining a distinct system of “addresses” at which objects that enjoy the relevant sort of existence may be “located.” But unless such “address systems” are supposed to be free for the thinking up, there must be some principled constraint on their generation. Brandom offers no such constraint. In the absence of any such constraint, the notion of existence is drained of substantive content. At the most general level, Brandom can say that for an object to exist is for it to have an address in some structured system of addresses. But by his own admission, the “property” of having an address in the address system defined by the family of canonical designators for physical objects is a quite different “property” from the property of having an address in the address system defined by a collection of canonical designators for fictional objects. For example, Sherlock Holmes is supposed to occupy an address in a certain fictive address system. He is thereby supposed to enjoy a kind of existence. Since, however, Holmes occupies no address in the address system defined by the totality of spatiotemporal coordinates, there is another sort of existence that Holmes apparently fails to enjoy. Evidently, we cannot settle, once and for all, the question of whether Sherlock Holmes exists or fails to exist. We are apparently supposed to conclude that although Holmes does not “physically exist,” he does “exist in the Holmes story.” But the urgent question is whether existence in the Holmes story and physical existence are really supposed to be species of a common genus. If so, the relevant genus, whatever it is, would seem to be a motley collection indeed. For nothing holds the members of the species together except that each is correlative with some family or other of canonical designators. But because there is evidently nothing substantive to say about how, in general, that semantically privileged status is earned, there is evidently nothing substantive to say about the genus “existence.” Brandom is untroubled by this fact. As he puts it: On this relaxed account [of existence], there is no reason to boggle at claims that numbers or other abstract objects exist. One must insist only that a determinate sense have been given to such claims, by specifying the relevant class of canonical designators. (Brandom 1994, 449)

It is not my aim to decisively refute this relaxed understanding of existence here. It is enough for our purposes to point out that this understanding

270  Referring to the World stands in sharp contrast with the view implicit in this chapter and in this book. On my view, referentially successful expressions are referentially fit expressions that stand in a certain, admittedly not yet fully elucidated, real relation to an independently existing object. Where there is no such relation, there is no reference to any real existent. Apparently inspired by the early Frege, Brandom seems, by contrast, to hold that the very idea of an object is inextricably tied to the cognition of relations of purported identity and difference, as expressed by the regulated use of singular terms, paradigmatically in identity statements. This idea is not entirely devoid of philosophical merit. Indeed, I have argued throughout this book for a related, though distinct idea. On my view, it is necessary for the cognition of an object as an object that something extra-​representational be brought under singular representations such that through the deployment of those very representations we thereby think a totality of relations of purported identity and difference. But I insist, contrary to Brandom and with Kant, that we can think a totality of relations of purported identity and difference, in the absence of the extra-​representational given, without thereby being given any object that is any real existent. Thinking a totality of relations of purported identity and difference may suffice to give us a domain of merely notional objects, as I called them earlier, but it does not, on its own, suffice for the givenness of a genuine non-​notional object. Only non-​notional objects are real existents. And such objects are not given to us merely through such representation-​representation relations as enable us to think with purport of sameness and difference. If our words and/​or our thoughts are to reach beyond our notions and all the way out to really existent objects, something extra-​representational must be given as well. But it should be clear by now that, on my own view, to deny that genuine, non-​notional objects are given merely through representation-​representation relations is not to deny that such relations play a deep and ineliminable role in our cognition of objects as objects. By rendering our thought and talk objectual—​that is, inwardly fit for the job of referring to outer objects—​our antecedent readiness to deploy merely objectual representations in thinking with intrinsic purport of sameness and difference is what first prepares the way for the encounter with the extra-​representational given to result in full-​ blown cognition of objects as objects. Contrary to Brandom, the early Frege, and many other philosophers of a quasi-​idealist stripe, I deny that genuine objects can be given to us through the mere play of representations alone. That way lies only darkness.

The Things We Do with Empty Names  271

9.  Taking Stock It is time to recapitulate the philosophical benefits to be derived from my trio. Begin by revisiting our two historically influential philosophical casts of mind about the non-​existent. On the one hand, there is the embrace of the non-​existent exhibited by Meinong, Brentano, and many others. On the other, there is the squeamishness of Russell and Frege, a squeamishness that is widely felt even to the present day. I announced at the outset that there is something half right and half wrong both about the insistence and about the squeamishness. Take the squeamishness first. That squeamishness of Frege and Russell is, I think, partly explained by their tacit grasp of the fact that the primary and paradigmatic objects of thought are real existents. Contrary to Brentano, for example, Frege and Russell rightly held that we have little reason to deny that thought is primarily and paradigmatically directed toward real existents rather than at merely “immanent” objects that are objects for thought alone. By contrast, what Brentano correctly grasped, albeit through a glass dimly, is that a certain part of the business of determining the objective representational content of our thoughts and our words must lie entirely on the side of the thinking subject, with no contribution at all from the external world. The distinctions developed both in this chapter and throughout this book allow us to pay due respect to both the insistence and the squeamishness. Armed with the distinction between merely objectual and fully objective representations, for example, we can say more clearly what truth it was that Brentano dimply grasped. He dimly grasped that it is the business of factors lying entirely on the side of the subject to render our representations objectual and thereby to prepare the way for the mind to make contact with real existents. Thought must be objectual before it can possibly achieve full-​ blown objectivity. And Brentano dimly grasped that objectuality of thought is not the business of the external world at all. What Brentano failed to appreciate, however, is that no genuine object at all—​not even an intentionally inexistent one—​is yet given through a mere play of merely objectual representations. Frege and Russell, by contrast, correctly grasped that it is only when our representations achieve full-​blown objectivity that they are rendered semantically answerable to anything at all. Now if having states with determinate objective representational content is of the very essence of thought, then Frege and Russell can be seen as dimly realizing that that which is devoid of

272  Referring to the World full-​blown objective representational content is not yet thought in the fullest sense. That realization is precisely what drove Frege to dismiss thoughts as of non-​existent objects as merely “mock” thoughts. Though there is something right about the impulse the led Frege to make this dismissal, I hope to have demonstrated that what Frege dismissed as merely mock thoughts actually occupy a much grander place in our cognitive lives than he imagined. The key to understanding that place is to appreciate that many language-​thought games we play are non-​veridical. Again, though moves in such games often do not enjoy full propositional contents, fully determinate truth conditions, or truth values, nonetheless such moves may enjoy truth-​similitude. And despite enjoying mere truth-​similitude, moves in such games can have great cognitive significance. Indeed, non-​veridical games played with presumptively merely objectual linguistic representations are the stuff of which many shared imaginings are made. The interpretive games authorized by works of fiction represent one kind of shared imagining. But the capacity for shared imaginings is a distinctively human capacity, which lies at the very foundation of our capacity for distinctively human forms of cultural and social life. The referential apparatus plausibly shared by all human languages—​the whole apparatus of names, deictics, quantifiers, variables, and anaphora—​is one of the main sources of the capacity for shared imaginings. Although that apparatus is, in one sense, made for talking about actual existents, we can and do deploy that apparatus even in the absence of any actual existent. We do so precisely when we play non-​veridical language games with merely objectual representations. Our ability to deploy the referential apparatus of our language even in the absence of actual existents is part of the explanation of the very possibility of our capacity for shared imaginings. When we deploy the referential apparatus of our language in the absence of real existents, our linguistic play may enjoy the illusory feel of full-​fledged objectivity. Our moves may appear to have a sort of ethereal subject matter and thereby to track some peculiar species of truth rather than mere truth-​similitude. And this is due to precisely the fact that non-​ veridical games are typically played with merely objectual representations, and such representations are in a sense “designed” for talking about genuine objects. Moreover, the moves we make with such representations are often rationally warranted, at least relative to the governing concern operative in such games. There can even be a player-​independent—​though not a game-​ independent—​“right” and “wrong” of the matter within such games. It is

The Things We Do with Empty Names  273 when we mistake the player-​independent right and wrong of the matter for a game-​independent right and wrong on the matter, that we become liable to have subject matter illusions. Such illusions are only strengthened by the fact that even within non-​veridical language/​thought games, there may obtain a give and take of reasons and competing entitlements and commitments. One could not be accused of mere philosophical naiveté if one were to conclude on this basis that through the representational play of such games, a realm of objects was somehow being constituted. Indeed, one way to read idealism is as the view that genuine objects are nothing but projections or constructions out of mere relations of purported reference and coreference among our representations—​perhaps together with the added view that there is no principled basis for anything like our distinction between genuine truth and mere truth-​similitude. Throughout this chapter, and indeed this entire book, we have been at pains to distance ourselves from any such views. It takes more than a mere play of representations to constitute a world of objects.

10.  Believing with Singular Purport in the Non-​Existent We think about the mind-​independent world by deploying in thought inner representations that are bound to outer objects and properties. That is, it is only when our inner representations reach all the way out to objects and their properties that our thought and talk achieve objective representational content and thereby become more than a mere play of inner representations. Still, it must be admitted that some of our representations sometimes fail to make semantic content with outer reality. Consequently, some of our representational states are at best only partially contentful. Merely partial contents are not fully outwardly facing. This raises a question that we have touched upon lightly and have not yet addressed head-​on. Some people do seem to have sincere and genuine beliefs about non-​existent objects. And these beliefs seem to be beliefs about the external world and not just a play of inner representations. Moreover, some of those beliefs appear to be singular, at least in purport, if not in actuality. Consider, for example, the belief, or apparent belief, that Pegasus can fly. One who believes, or at least apparently believes, that Pegasus can fly would seem to undertake a singular worldly existential commitment, not merely to the existence of some flying horse or other but to a quite particular flying horse. In believing, or apparently believing, that Pegasus can fly, the believer seems apparently to have staked

274  Referring to the World out a claim about a particular horse and its capacities. If it turned out that, say, Secretariat could fly, and not just metaphorically but literally, that would not suffice to make the (apparent) belief that Pegasus can fly true. There is no such horse as Pegasus. Arguably, there could not be any such horse. If one were to enumerate all the horses that actually exist and even all the horses there could be, one would not find Pegasus among them. But that means that there is no actual or even possible horse about which one has a belief when one apparently believes with singular purport that Pegasus can fly. What are we to make of such beliefs? Given what has gone before, the most natural thing to say is that the belief that Pegasus can fly is singular in form but not in content. We might also add that because there is no such object as Pegasus, the belief that Pegasus can fly is not a belief with fully complete truth conditions or complete propositional content. And qua propositionally or truth conditionally incomplete, the apparent belief that Pegasus can fly stakes out no fully determinate claim to the effect that some particular horse can fly. It may purport to, but it does not succeed in so doing. This approach is right as far as it goes, but it does not go quite far enough. On my view, beliefs are by their very nature what we might call world-​involving attitudes. To have a belief is, in effect, to stake out a claim about how the world is. And herein lies the problem. What claim about how the world is has been staked out by one who believes that Pegasus can fly? It is tempting to say that no fully determinate claim has been staked out. But if that is right, and if it is also right that beliefs are essentially world-​ involving attitudes, does it not follow that the apparent belief that Pegasus can fly is not, after all, a genuine belief? But if the belief that Pegasus can fly is not a genuine belief, what exactly is it then and what exactly is its content? I am prepared to simply bite the bullet here. I am prepared to say that precisely because it is not fully world-​involving that Pegasus can fly is not the possible content of any genuine belief. Strictly speaking, that Pegasus can fly is incredible. The bullet I have just bitten may seem explosive, at least at first blush. But if we consider more carefully the world-​involving character of genuine belief, we will see that it is not so explosive after all. Recall that on my view, belief contents, including the contents of singular beliefs, are essentially individuated by the commitments undertaken in the adopting of those very beliefs. It is the existential, predicative, and referential commitments undertaken in the holding of a belief that jointly determine the worldly truth conditions or propositional contents of a belief. Where there are no such commitments

The Things We Do with Empty Names  275 undertaken, there is not yet a belief, at least not yet a full-​blown belief. In the case of singular existential commitments that are characteristic of singular beliefs, where there are no actual existents with respect to which the believer undertakes predicative commitments, there is not yet anything that can be properly characterized as a singular belief, at least not a singular belief with singular content. And this is just a way of saying that qua singular in purport, but not in content, that Pegasus can fly is not the possible content of any fully singular belief. But as it stands with its singularity of form but not of content, is the claim that Pegasus can fly a possible content of any state of mind at all? Perhaps. I certainly need not deny that partial or propositionally incomplete contents can be used to characterize certain attitudes and states of mind. Fantasy and imagination are perhaps two such attitudes. One can, I think, engage in what we might call Pegasus-​imaginings merely by ordering one’s Pegasus-​notions or conceptions in certain ways. But if the formally singular representations with which one fantasizes or imagines are not already bound down to worldly existents, then the inner deployment of those representations still will not yet amount to an undertaking of worldly commitments at all.17 Perhaps this is what separates essentially and fully world-​facing attitudes like believing from mere imaginings and the like. Indeed, I suspect, but will not argue in detail here, that the essentially and fully world-​facing character of believing rests ultimately on facts about the functional character of believing as opposed to imagining or fantasizing. It is because believing has the functional character that it has that in believing we stake out a claim on how things are in the world. Now someone disposed to defend the credibility that Pegasus can fly, while accepting my claim that beliefs are essentially and fully world-​facing, may want to insist that there is, after all, at least one real existent about which one who apparently believes that Pegasus can fly stakes out a claim—​the universe itself. One who apparently believes that Pegasus can fly stakes out a 17 Of course, dreams, fantasies, and even one’s wildest imaginings can come true. Does that mean that merely notional attitudes are, after all, truth evaluable? Yes and no. Yes, in the sense that we can sometimes compare notional contents with the world and ask whether there is or is not a match between world and content. So, for example, if I fantasize about marrying a lovely woman on earth and do marry a lovely woman, we may say that my fantasy has come true. But a failure to come true of a mere fantasy is no cognitive defect of either the fantasy or the fantasizer. Nor is the coming true of a fantasy a cognitive virtue of either the relevant state or the relevant cognizer. Notional states are not as such “aimed” at the world. In contrast with beliefs and desires, for example, there is no characteristic direction of fit between content and world for merely notional attitudes. Consequently, any worldly evaluation of a merely notional attitude is normatively otiose.

276  Referring to the World claim about the universe at large: that it contains a horse of a certain description that can fly. And this is as fully and essentially world-​facing as any belief could be. It is false that Pegasus can fly, on this view, just because the universe at large contains no flying horse of a certain description. But this line is utterly blind to the intrinsic singular purport that Pegasus can fly and blind to the fact that it does not do what it purports to do. To adopt this line is really just to repeat Russell’s mistake of thinking that empty singular representations are really just quantificational or general representations in disguise. It is true enough that any genuine belief, general or singular, stakes out a claim ultimately about the universe at large. But singular beliefs stake out such claims in a quite peculiar way. They stake out claims about the universe at large by staking out a claim about particular inhabitants of the universe. But there is no particular x such that one who apparently believes that Pegasus can fly thereby stakes out a claim about that very x. Despite its attempted or purported singularity, that Pegasus can fly doesn’t succeed in being singular in content. Still, the intrinsic singular purport of that Pegasus can fly needs explaining, not eliminating. To say that the attempted singularity of that Pegasus can fly fails is not to claim that that Pegasus can fly is really existentially general. After all, we have already seen that if some horse or other could fly, that would not suffice for the truth of that Pegasus can fly. Similarly, if more controversially, even though no horse can fly, that does not suffice for the falsity of that Pegasus can fly. It seems more apt to say that that Pegasus can fly is neither strictly literally true nor strictly literally false. That Pegasus can fly is not, in short, the sort of thing for which the question of literal truth or falsity even arises. And it is just because that Pegasus can fly doesn’t rise at all to the level of truth evaluability that I am willing to say that it is strictly speaking incredible.18 There are, to be sure, many other credible contents within we might call a “doxastic neighborhood” of that Pegasus can fly. One can strictly and falsely believe, for example, that there exists some winged horse or other that can fly.

18 Consider the following remark by Burge (2007, 48):

If A gullibly believes that Pegasus was a (real) horse, the demonstrative implicit in the name occurs anaphorically, perhaps without A’s realizing it, taking as antecedent some description, definite or not, in the repertoire of A or someone else. The name thus has the flavor of “that Pegasus (whichever one they are talking about).” Though I reject Burge’s claim that names contain implicit indexicals that can be anaphorically dominated by some antecedent, he has, I think, grasped part of the truth about names. A name that occurs within a purely notional frame will have a purely notional significance. It is tempting to say that such a name is used as if it refers to a particular object, without actually so doing. But again, I do not find saying this deeply linguistically illuminating.

The Things We Do with Empty Names  277 One can strictly and falsely believe that there was once some winged horse named “Pegasus.” And one can strictly and truly believe that the sentence “Pegasus can fly” fails to express anything fully truth evaluable. But all this shows is that the propositionally incomplete content that Pegasus can fly is surrounded by a neighborhood of fully truth conditional, fully propositional, and therefore fully credible contents. It would be a worthy enterprise to precisely characterize the doxastic neighborhood of that Pegasus can fly. I will not undertake that task in detail here. I will just say that the principles that determine the precise contours of doxastic neighborhoods are likely to be various. Something like John Perry’s (2001) approach to what he calls reflexive content comes to mind as one model of how to think about doxastic neighborhoods. On a version of Perry, views only slightly fine-​tuned to the case of worldly beliefs and their doxastic neighborhoods, many reflexive contents, as he calls them, would be located in the doxastic neighborhood of that Pegasus can fly without any of them being identical to it. But one can also imagine less “metarepresentational” approaches. One might, for example, appeal to something like my own distinction between concepts and conceptions, as developed in Chapter  4. Cognizers like us are such that she who uses a name or a name-​like mental representation typically opens a file—​a file ready for writing in that functions as a storehouse of information and misinformation. It would not be surprising if even in the case of an empty representation some of the information was fully and completely world-​facing and thus possible contents of further beliefs. Believing the world-​facing contents contained within a “Pegasus” file would not be tantamount to believing that Pegasus can fly. But it might be plausibly tantamount to believing things within the doxastic neighborhood of that Pegasus can fly. The central point here, however, is not so much to define in detail how doxastic neighborhoods might be generated but to distinguish the often incredible contents that may lie at the center of such a neighborhood from the more fully truth-​conditional and/​or propositional contents that occupy such a neighborhood. The point is that it should not be thought that in believing some world-​facing content or other via acceptance of a sentence or mental representation like “Pegasus can fly,” one thereby strictly, literally believes the content that Pegasus can fly. Strictly speaking, that incomplete “content” is not a possible content of any genuine belief. It is important to acknowledge that the temptation to misidentify the truth-​conditionally incomplete content that Pegasus can fly with some fully truth conditional content or other in its neighborhood has afflicted many

278  Referring to the World a philosopher. That temptation is a major source of something like what Perry (2001) has come to call subject matter fallacies. One commits a subject matter fallacy, according to Perry, when one supposes that the content of a statement or belief is wholly constituted by the conditions its truth puts on the subject matter of the statement or belief; that is, the conditions it puts on the objects the words designate or the ideas are of. (50)

Since I think that there is at least one sense in which it is correct to say that all there is to the content of a (singular) statement or belief is “the conditions it puts on the objects the words designate or that its ideas are of,” I wouldn’t put matters in quite the way that Perry does. But the intended spirit of Perry’s remarks here seems to me entirely correct. Speaking in my own vocabulary rather than his, I would say that one should always be careful not to misidentify belief in that which lies in a mere doxastic neighborhood of a failed content with belief in that failed content itself. Though failed contents are not strictly speaking believable, we can get “close” to believing them by believing one of their doxastic neighbors instead. From our perspective, Perry’s subject matter fallacies amount to mistaking belief in a doxastic neighbor of an incomplete content for belief in the real thing. Something like that mistake is quite widespread. It begins with the descriptivism explicit in Russell and plausibly implicit in Frege. It continues to the present day with those who take so-​called narrow contents to be the prior and fundamental form of belief content.

11.  Ascribing Belief in the Non-​Existent Again In the course of arguing for the priority of so-​called de re ascriptions over so-​called de dicto ascriptions, we briefly considered examples involving non-​ existence. We close out this chapter with a closer look at such ascriptions. It turns out that such ascriptions are quite subtle. I will argue that they lend additional weight to our claims about the priority of the de re in representing representations. Begin by recalling an example discussed briefly in passing in the previous chapter. Suppose Jones utters the following sentence and thereby expresses a certain belief:

The Things We Do with Empty Names  279 (10) The man in the doorway has a frightening look on his face.

Suppose, however, that there is no man with a frightening look and no actual doorway semantically implicated in Jones’s apparent belief. Suppose that Smith realizes that there is no man x and no doorway y such that Jones takes x to be in y. And suppose that Smith intends to inform Black of the apparent belief that Jones apparently expresses in uttering (10), and that she intends to do so while explicitly distancing herself from Jones’s bizarre existential commitments. Since no man and no doorway are semantically implicated in what appears to be Jones’s belief, it is not open to Smith to offer up such a man and such a doorway to Black for the cognizing preparatory to mounting what I called a fulsomely de re ascription back in Chapter 6. Nor will a straightforward embedding of the relevant description achieve the intended communicative effect. For consider: (11) Jones believes that the man in the doorway has a frightening look on his face.

A strict literal utterance of (11) would seem to express a commitment on Smith’s part to the existence of a man in a doorway, contrary to her own communicative intentions. As I pointed out earlier, if Smith uses an indefinite rather than a definite to characterize the belief apparently expressed by Jones in uttering (10), she would not thereby express a commitment to the existence of a doorway containing a man. Consider, for example: (12) Jones believes that there is a man in a doorway who has a frightening look on his face.

Indefinites give rise to puzzles of their own and I only hinted at such puzzles in the previous chapter. Let us take a close look now. First notice that although an indefinite like “a man” is not itself a referring expression, it can nonetheless serve to introduce a novel subject into a discourse. And in keeping with its subject introducing role, an indefinite can anchor an anaphoric chain all of whose links make reference to the newly introduced subject, as in (13) A man just came into the room. He is tall and handsome. His name is John.

280  Referring to the World A speaker who produces a discourse like (13) would normally thereby express a commitment of her own to the existence of a tall, handsome man named John. However, notice that no such existential commitment is expressed by an ascriber who uses an embedded indefinite in a belief ascription in the manner of (12). Nor can an unembedded pronoun have an embedded indefinite as its antecedent. And this fact suggests that an embedded indefinite does not serve a subject introducing role.19 Consider: (14) Jones believes that a man in a blue hat is at her door. He has a frightening look on his face.

Strikingly, there is no reading of (14) in which the pronoun “he” has the embedded indefinite “a man in a blue hat” as its antecedent. Interpreted anaphorically rather than deictically, the unembedded “he” must inherit its reference and/​or existential import from elsewhere. But the would-​be antecedent—​the embedded indefinite—​has neither reference nor existential import to pass on. To be sure, there is a charitable reading of (14) that seems roughly equivalent to (15) Jones believes that a man in a blue hat is at her door, and she believes that he has a frightening look on his face.

So construed, the indefinite “a man in a blue hat” does appear to function as the antecedent of the embedded “he.” The ability of an indefinite in one that-​ clause to function as the antecedent of a pronoun in a distinct that-​clause is a syntactic and semantic mystery worthy of serious theorizing. I suspect, but will not argue here, that this ability is the notional correlate of the subject-​ introducing role played by unembedded indefinites. That is to say, an indefinite that anchors an anaphoric chain within or across that-​clauses serves, in effect, to introduce what might be called a purely notional discourse subject. Each link in such an anaphoric chain will itself have a merely notional significance—​a notional significance shared with every other member of the relevant anaphoric chain. Thus, in



19 See Chierchia (1995), for a discussion.

The Things We Do with Empty Names  281 (16) Jones believes that a man in a blue hat is at her door. Smith thinks that he has a frightening look on his face,

the notional significance of the embedded “he” is evidently inherited from its antecedent “a man in a blue hat.” Call a linguistic context within which indefinites serve to introduce merely notional subjects a “notional frame.” That-​clauses appear at least sometimes to function as notional frames. Nor are they the only sorts of notional frames. Fictional narratives of the sort we have considered already are another and paradigmatic sort. Consider: (17) Once upon a time in a land far away, there lived a King. He was wise and benevolent. His name was Philo. King Philo had a beautiful daughter. Her name was Jasmine.

One who produces such a narrative characterizes what we might call a merely “notional world” in which a benevolent king named Philo has a beautiful daughter named Jasmine. That king has such a daughter within the notional world, but not within the real world. That is, in characterizing this merely notional world, one has clearly not undertaken a commitment to the real-​world existence of a such a king and such a princess. It is tempting to talk here of pretend existential commitments and pretend reference to particulars. I find such talk largely unobjectionable, but it does not add anything deeply illuminating to our understanding of the peculiar linguistic character and function of notional frames. Indeed, the direction of illumination is more likely to be the reverse. It is only by achieving an antecedent understanding of the peculiar linguistic character and function of notional frames that we are likely to achieve an understanding of what we are doing when we deploy the apparatus of singular reference in episodes of pretense. Notice that one who ascribes a belief to Jones using (15) does not thereby commit herself to the existence of a man in a blue hat with a frightening look. Both the embedded indefinite and its embedded anaphoric dependent function wholly notionally—​that is, solely to specify how things are in the notional world defined by Jones’s beliefs. They do so without thereby expressing worldly existential commitments of the ascriber’s own. It is worth contrasting (15) with (18):

282  Referring to the World (18) Jones believes that the man in the blue hat is at her door. He has a frightening look on his face.

There is no reading of (18) in which either the definite description “the man in the blue hat” or the pronoun anaphorically linked to it function wholly notionally. An ascriber who ascribed a belief to Jones via an utterance of (18) would thereby express only her own commitment to the existence of a man in a blue hat with a frightening look on his face. These last examples show that even when an ascribee uses a definite description attributively with reference to no particular object, a would-​be ascriber cannot, via mere embedding of the relevant description, escape expressing as her own the existential commitments normally conveyed by the use of that description. It may so far appear that where no actual existent is doxastically implicated in the ascribee’s belief, a would-​be ascriber cannot resort to the circumlocution of a fulsomely de re ascription to inoculate herself from the ascribee’s commitments. There is, however, another sort of inoculation available to the ascriber. She can inoculate herself from the ascribed commitments by deploying embedded indefinites, thereby going wholly notional in her ascriptions. By doing so, however, she represents the ascribed belief as entirely lacking in singular referential purport. It is as if the cost of representing the relevant mental state as a genuine, and not merely apparent, belief is that it must be represented as altogether lacking in even singular purport. One final qualification is in order. It turns out that when an embedded definite is anaphorically linked to an embedded indefinite with wholly notional significance, the definite may have wholly notional significance. Consider, for example, (19): (19) Jones believes that a man in a blue hat is at her door, and she believes that the man in the blue hat has a frightening look on his face.

At a minimum, with the definite anaphorically linked to the indefinite, (19) is awkward. (15), in which the definite is replaced by a pronoun, seems to be the preferred way of expressing what (19) awkwardly struggles to expresses. The awkwardness of (19) results, I think, from something like a “nearness” effect. Consider a discourse in which the indefinite that serves to introduce a notional subject is more distant from the anaphorically dependent definite as in this:

The Things We Do with Empty Names  283 (20) Smith: Jones believes that a man in a blue hat is at her door. Black: Is that why she looks so afraid? Smith: Yes. She believes that the man in the blue hat is about to beat her up.

The embedded definite seems perfectly acceptable here. Clearly, Smith does not commit herself to the existence of a man in a blue hat by her use of the embedded definite. Rather, her use of the embedded definite seems wholly notional. Such examples suggest that when a definite occupies the right sort of place in a chunk of discourse, it may enjoy what we might call “inherited,” as opposed to original, notional significance. It may even be that a standard way to make a so-​called de dicto ascription is by employing an embedded definite whose notional significance is wholly inherited from some explicit or implied embedded indefinite antecedent. If so, perhaps the de re/​de dicto distinction should really be viewed not as a distinction at the level of individual sentences, but as a distinction at the level of ordered chunks of discourse. Though there is much to be said for this thought, it does not, I shall argue, undercut my argument for the primacy of de re ascriptions, or my claim that fulsomely de re ascriptions are the preferred way of ascribing commitments to others that one does not share. After all, the use of an anaphorically dominated embedded definite description with wholly notional significance would seem to be conversationally permissible in only a limited range of conversational settings. Moreover, even when such a construction is permissible, the relevant ascription will typically only partially characterize a reality that would be more fully and informatively characterized via a fulsomely de re ascription. We might call the move of going wholly notional by deploying embedded indefinites, possibly together with anaphorically linked definite expressions within a (sequence of) belief ascription(s), “indefinitization.” Now recall our earlier scenario from Chapter 6, in which Jones mistakes the martini-​ drinking man for a gimlet-​drinking woman. Suppose that Smith intends to inform Black of Jones’s beliefs. Consider the following: (21) Jones believes that a woman in a corner drinking gimlets has had too much to drink.

With the indefinite noun phrase read as having narrow scope, (21) is arguably true. An utterance of (21) would clearly correctly partially specify the

284  Referring to the World notional world defined by Jones’s belief. The ascriber also avoids committing herself to the existence of a gimlet-​drinking woman by indefinitizing in this manner. Notice, however, that one who characterizes Jones’s belief by an utterance of (21) and leaves it at that in the imagined conversational setting, clearly violates something like the Gricean maxim of informativeness. (21) omits the crucial information that Jones believes what she believes on the basis of (mis)taking a certain man for a woman. This information is not ancillary or irrelevant but is of the very essence of Jones’s state of mind—​a state of mind better and more informatively represented by the fulsomely de re ascription (22) than by the indefinitized (21): (22) Jones believes of the martini-​drinking man in the corner, whom she mistakes for a gimlet-​drinking woman, that he has had too much to drink.

To see this, imagine that Smith utters (21) as a way of characterizing Jones’s belief and that Black attempts to act on the information conveyed by Smith’s utterance. We can imagine, for example, that as a consequence of Smith’s utterance Black is on the lookout for an inebriated woman ordering gimlets. Clearly, were Black to learn more about the facts of the matter—​that there is no woman drinking gimlets, that Jones has mistaken a martini-​drinking man for a gimlet-​drinking woman, and that Smith is herself entirely aware of these facts—​she could, with some justice, accuse Smith of misleading her. It is fair to say, therefore, that (21) is misleading in the imagined setting. It is misleading because an indefinitized ascription is a merely partial characterization of Jones’s episode of singular believing, a characterization that leaves out the crucial thing, viz., the truth-​relevant aspects of that episode of believing. The truth of Jones’s episode of singular believing depends on how things are by a certain man in a certain corner and not on whether there is some gimlet-​drinking woman or other in some corner or other. (21) is silent on these aspects of Jones’s singular belief episode. In effect, Smith has represented Jones’s episode of believing as a mere array of notions, or at best as an existentially general belief, and not as a singular belief semantically bound to an actual existent. And because she has left out the actual existent to which those notions are semantically bound down, she has not fully and correctly represented the worldly content of Jones’s belief.

The Things We Do with Empty Names  285 Finally, consider again the belief Jones expresses in uttering (10) above, with which we began. Since there would appear to be no real existents semantically implicated in Jones’s belief, it may appear that an indefinitized ascription is really our only option. But even here, despite appearances, there may in fact be a real existent implicated in Jones’s belief after all, especially to the extent that that belief is taken to be world-​facing. If that is right, then Jones’s state of mind might be thought to count as an episode of believing rather than an episode of fantasizing or imagining, after all. Even if there is no particular man and no particular doorway implicated in Jones’s belief, despite its purported singularity, there might just be a more or less determinate particular region of space that is semantically implicated in her belief. If that were so, then Jones’s episode of believing might after all have determinate worldly truth conditions. For consider more closely just what the truth or falsity of Jones’s belief might hang on. First, notice that it is not the case that any old man and any old doorway are even relevant to the truth or falsity of Jones’s belief. Even if there does exist a doorway containing a man with a frightening look in some region of space or other, that would not suffice for the truth of Jones’s belief unless that very doorway and that very man are implicated in Jones’s belief in the peculiar way that renders them truth-​conditionally relevant to Jones’s belief. On the other hand, unless there is some region or other of space that is truth-​conditionally implicated in Jones’s belief, even the fact that no region of space contains a man in a doorway would seem utterly powerless to render Jones’s belief false. Jones’s belief will be determinately true or false—​and true or false of the world, it seems—​just in case there is some more or less determinate particular region of space that Jones takes to contain a doorway with a frightening looking man in it. On this approach her belief would be true just in case that very region does contain such a doorway and such a man and it would be false otherwise. Moreover, it appears that if we wish to display Jones’s belief under its truth-​evaluable aspect, we cannot rest with the ascription of merely notional contents. We must ascribe singular referential content that is already semantically bound down to doxastically implicated existents. We must, in effect, “de-​indefinitize” our indefinitized ascription by going fulsomely de re. Indeed, it would appear that the following fulsomely de re ascription adequately captures both the referential and notional contents of Jones’s belief: (23) Jones believes of region R, which she takes to contain a doorway, that it also contains a man with a frightening look on his face.

Works Cited Adams, F., and Dietrich, L. A.. 2004. “What's in a(n Empty) Name?” Pacific Philosophical Quarterly 85: 125–​148. Adams, F., Fuller, G., and Stecker, R. 1992. “The Semantics of Thought.” Pacific Philosophical Quarterly 73: 375–​389. Adams, F., Fuller, G., and Stecker, R. 1997. “The Semantics of Fictional Names.” Pacific Philosophical Quarterly 78: 128–​148. Adams, F., and Stecker, R. 1994. “Vacuous Singular Terms.” Mind and Language 9: 387–​401. Anscombe, G. E. M. 1963. Intention. Oxford: Basil Blackwell. Asher, N. 2000. “Truth Conditional Discourse Semantics for Parentheticals.” Journal of Semantics 17(1): 31–​50. Bach, K. 1999. “The Myth of Conventional Implicature.” Linguistics and Philosophy 22(4): 367–​421. Bach, K. 2010. “Getting a Thing into a Thought.” In New Essays on Singular Thought, ­edited by R. Jeshion. Oxford: Oxford University Press. Baker, L. 1985. “Just What Do We Have in Mind?” Midwest Studies in Philosophy, 10: 25–​48. Baker, L. 1987. Saving Belief. Princeton: Princeton University Press. Bianchi, A., ed. 2015. On Reference. Oxford: Oxford University Press. Boer, S. 1975. “Proper Names as Predicates.” Philosophical Studies 27: 389–​400. Boghossian, P. A. 1990. “The Status of Content.” Philosophical Review 99(2): 157–​184. Brandom, R. 1994. Making It Explicit. Cambridge, MA: Harvard University Press. Brandom, R. 2000. Articulating Reasons: An Introduction to Inferentialism. Cambridge, MA: Harvard University Press. Braun, D. 1993. “Empty Names.” Nous 27: 449–​469. Braun, D. 2005. “Empty Names, Fictional Names, Mythical Names.” Nous 39: 596–​631. Brentano, F. 1995. Psychology from an Empirical Standpoint. Originally published in 1874; English edition edited by L. McAlister, London: Routledge and Kegan Paul 1973; reprinted with an introduction by P. Simons, London: Routledge 1995. Burge, T. 1973. “Reference and Proper Names.” Journal of Philosophy 70(14): 425–​439. Burge, T. 1977. “Belief De Re.” Journal of Philosophy 74: 338–​362. Burge, T. 1979. “Sinning against Frege.” Philosophical Review 88: 398–​432. Burge, T. 1982a. “Other Bodies.” In Thought and Object, edited by A. Woodfield. New York: Oxford University Press. Burge, T. 1982b. “Two Thought Experiments Reviewed.” Notre Dame Journal of Formal Logic 23: 284–​293. Burge, T. 1986. “Individualism and Psychology.” Philosophical Review 95: 3–​45. Cassirer, E. 1953. Substance and Function and Einstein’s Theory of Relativity. New York: Dover Chalmers, D. 2002. “Sense and Intention.” In Language and Mind, vol. 16 of Philosophical Perspectives, 135–​182. Oxford: Oxford University Press. Chierchia, G. 1995. Dynamics of Meaning: Anaphora, Presupposition, and the Theory of Grammar. Chicago: University of Chicago Press.

288  Works Cited Crane, T. 2006. “Brentano's Concept of Intentional Inexistence.” In The Austrian Contribution to Analytic Philosophy, edited by M. Textor. London: Routledge. Crane, T. 2013. The Objects of Thought. Oxford: Oxford University Press. Crimmins, M. 1992. Talk about Beliefs. Cambridge, MA: MIT Press. Crimmins, M. 1995. “Notional Specificity.” Mind and Language 10(4): 464–​477. Crimmins, M. 1998. “Hesperus and Phosphorus: Sense, Pretense, and Reference.” Philosophical Review 107: 1–​48. Crimmins, M., and J. Perry. 1989. “The Prince and the Phone Booth: Reporting Puzzling Beliefs.” Journal of Philosophy 86(12): 685–​711. Davidson, D. 1968. “On Saying That.” Synthese 19: 130–​146. Davidson, D. 1977. “Reality without Reference.” Dialectica 31: 247–​258. Donnellan, K. 1964. “Speaking of Nothing.” Philosophical Review 83: 3–​31. Everett, A. 2003. “Empty Names and ‘Gappy’ Propositions.” Philosophical Studies 116(1): 1–​36. Fara, D. G. 2011. “Names as Predicates.” Unpublished. Fara, D. G. 2015a. “ ‘Literal’ Uses of Proper Names.” In On Reference, edited by A. Bianchi, 251–​279. Oxford: Oxford University Press. Fara, D. G. 2015b. “Names Are Predicates.” Philosophical Review 124(1): 59–​117. Field, H. 1994. “Deflationist Views about Meaning and Content.” Mind 103: 249–​285. Fiengo, R., and R. May. 2006. Belief De Lingua. Cambridge, MA: MIT Press. Fine, K. A. 1982. “The Problem of Non-​Existents. I. Internalism,” Topoi 1: 97–​140. Fine, K. 2007. Semantic Relationalism. Malden, MA: Blackwell. Fodor, J. 1982. “Cognitive Science and the Twin Earth Problem.” Notre Dame Journal of Formal Logic 23: 98–​118. Fodor, J. 1994. The Elm and the Expert. Cambridge, MA: MIT Press. Frege, G. 1884/​1980. The Foundations of Arithmetic. Translated by J. L. Austin. Evanston: Northwestern University Press. Frege, G. 1887/​1997. Excerpts from Logic. In The Frege Reader, edited by M. Beaney. Oxford: Blackwell. Frege, G. 1977. Translations of the Philosophical Writings of Gottlöb Frege, edited by P. T. Geach and M. Black. Oxford: Blackwell. Frege, G. 1979a. Posthumous Writings, edited by H. Hermes, F. Kambartel, and F. Kaulbach. Chicago: University of Chicago Press. Frege, G. 1979b. “Seventeen Key Sentences on Logic.” In Posthumous Writings, edited by H. Hermes, F. Kambartel, and F. Kaulback, 174–​175. Translated by P. Long and R. White. Oxford: Blackwell. Friend, S. 2007. “Fictional Characters.” Philosophy Compass 2(2): 141–​156. Geach, P. T. 1967. “Intentional Identity.” Journal of Philosophy 64: 627–​632. Guerts, B. 1998. “Presuppositions and Anaphors in Attitude Contexts.” Linguistics and Philosophy 21: 545–​601. Harris, J., and C. Potts. 2009. “Perspective-​shifting with Appositives and Expressives.” Linguistics and Philosophy, 32(6): 523–​552. Hawthorne, J., and D. Manley. 2012. The Reference Book. Oxford: Oxford University Press. Heck, R. 2002. “Do Demonstratives Have Senses?” Philosopher’s Imprint 2(2): 1–​33. Holton, R. 1993. “Minimalisms about Truth.” In Themes from Wittgenstein, edited by B. Garrett and K. Mulligan. Working Papers in Philosophy, 4: 45–​61. Canberra: Australian National University. Hom, C. 2008. “The Semantics of Racial Epithets.” Journal of Philosophy 105(8): 416–​440. Hornsby, J. 2001. “Meaning and Uselessness: How to Think about Derogatory Words.” Midwest Studies in Philosophy 25(1): 128–​141.

Works Cited  289 Horwich, P. 1998. Meaning. Oxford: Clarendon Press. Jackson, F., Oppy, G., and Smith, M. 1994. “Minimalism and Truth Aptness.” Mind 103(411): 287–​302. Jeshion, R. 2002. “Acquaintanceless De Re Belief.” In Meaning and Truth, edited by J. Campbell, M. O’Rourke, and D. Shier. New York: Seven Bridges Press. Jeshion, R., ed. 2010a. New Essays on Singular Thought. Oxford: Oxford University Press. Jeshion, R. 2010b. “Singular Thought: Acquaintance, Semantic Instrumentalism and Cognitivism.” In New Essays on Singular Thought, edited by R. Jeshion. Oxford: Oxford University Press. Jeshion, R. 2014. “Referentialism and Predicativism about Names.” Erkenntnis 80: 363–​404. Jeshion, R. 2015a. “A Rejoinder to Fara’s ‘Literal’ Uses of Proper Names.” In On Reference, edited by A. Bianchi, 180–​194. Oxford: Oxford University Press. Jeshion, R, 2015b. “Names Not Predicates.” In On Reference, edited by A. Bianchi, 225–​250. Oxford: Oxford University Press. Kant, I. 1781/​1998. Critique of Pure Reason. Translated and edited by P. Guyer and A. W. Wood. Cambridge: Cambridge University Press, 1998. Kaplan, D. 1969. “Quantifying In.” In Words and Objections: Essays on the Work of W. V. Quine, edited by D. Davidson and J. Hintikka. Dordrecht: Reidel. Kaplan, D. 2003. “De Re Belief.” Unpublished. Kaplan, D. 2005. “The Meaning of ‘Ouch’ and ‘Oops.’ ” Unpublished. Kaplan, D. 2013. “De Re Belief.” In Presidential Addresses of the American Philosophical Association 1981–​1990, edited by R. Hull, 25–​37. Dordrecht: Kluwer Academic. Kripke, S. 1972. Naming and Necessity. Cambridge, MA: Harvard University Press. Kripke, S. 1977. “Speaker’s Reference and Semantic Reference.” In Midwest Studies in Philosophy, edited by P. A. French, T. E. Uehling Jr., and H. K. Wettstein, 255–​276. Vol. 2 in Studies in the Philosophy of Language. Morris: University of Minnesota. Kripke, S. 1979. “A Puzzle about Belief.” In Meaning and Use, edited by A. Margalit. Dordrecht: Reidel. Kripke, S. 2011. “Reference and Existence.” In Philosophical Troubles: Collected Papers, vol 1. Oxford: Oxford University Press. Lawlor, K. 2002. “Memory, Anaphora and Content Preservation.” Philosophical Studies 109: 97–​119. Matushansky, O. 2009. “On the Linguistic Complexities of Proper Names.” Linguistics and Philosophy 31(5): 573–​627. Meinong, A. 1960. “The Theory of Objects.” In Realism and the Background Phenomenology, edited by R. Chisholm. Glencoe, IL: Free Press. Millikan, R. 1993. White Queen Psychology and Other Essays for Alice. Cambridge, MA: MIT Press/​Bradford Books. Millikan, R. 2000. On Confused and Obscure Ideas: An Essay about Substance Concepts. Cambridge: Cambridge University Press. Parsons, T. 1980. Nonexistent Objects. New Haven, CT: Yale University Press. Perry, J. 1994. “Davidson’s Sentences and Wittgenstein’s Builders.” Proceedings of the American Philosophical Association. Reprinted in J. Perry, The Problem of the Essential Indexical on Other Essays. 2000. Stanford, CA: CSLI Publications. Perry, J. 2001/​2011. Reference and Reflexivity. Stanford, CA: CSLI Publications. Potts, C. 2005. The Logic of Conventional Implicatures. Oxford Studies in Theoretical Linguistics. Oxford: Oxford University Press. Potts, C. 2007. “The Expressive Dimension.” Theoretical Linguistics 33(2): 165–​197. Price, H. 2003. “Truth as a Convenient Fiction.” Journal of Philosophy 100: 167–​190.

290  Works Cited Priest, G. 2005. Towards Non-​Being: The Logic and Metaphysics of Intentionality. Oxford: Clarendon. Putnam, H. 1973. “Meaning and Reference.” Journal of Philosophy 70(19): 699–​711. Seventieth Annual Meeting of the American Philosophical Association Eastern Division, November 8, 1973. Quine, W. V. O. 1956. “Quantifiers and Propositional Attitudes.” Journal of Philosophy 53: 177–​187. Quine, W. V. O. 1960. “Variables Explained Away.” Proceedings of the American Philosophical Society, 104(3): 343–​347. Recanati, F. 1993. Direct Reference: From Language to Thought. Oxford: Blackwell. Recanati, F. 1994. “How Narrow Is Narrow Content?” Dialectica 48(3–​4): 209–​229. Recanati, F. 2000. Oratio Obliu, Oratio Recta. Cambridge, MA: MIT Press. Recanati, F. 2004. Literal Meaning. Cambridge: Cambridge University Press. Recanati, F. 2010. “Singular Thought: In Defence of Acquaintance.” In New Essays on Singular Thought, edited by R. Jeshion, 141–​188. Oxford: Oxford University Press. Richard, M. 1990. Propositional Attitudes: An Essay on Thoughts and How We Ascribe Them. Cambridge: Cambridge University Press. Richard, M. 2008. When Truth Gives Out. Oxford: Oxford University Press. Rorty, R. 1986. “Pragmatism, Davidson and Truth.” In Truth and Interpretation: Perspectives on the Philosophy of Donald Davidson, edited by E. LePore. Oxford: Basil Blackwell. Sainsbury, M. 2005. Reference without Referents. Oxford: Clarendon Press. Salmon, N. 1983. Frege’s Puzzle. Cambridge, MA: MIT Press. Salmon, N. 1998. “Nonexistence.” Nous 32(3): 277–​319. Salmon, N. 2002. “Mythical Objects.” In Meaning and Truth: Investigations in Philosophical Semantics. Edited by J. K. Campbell, M. O’Rourke, and D. Shier. New York: Seven Bridges. Saul, J. 1997. “Substitution and Simple Sentences.” Analysis 57(2): 102–​108. Saul, J. 2007. Simple Sentences, Substitution, and Intuitions. New York: Oxford University Press. Schiffer, S. 1977. “Naming and Knowing.” In Contemporary Perspectives in the Philosophy of Language, edited by P. A. French, et al. Minneapolis: University of Minnesota Press. Schiffer, S. 1995. “Descriptions, Indexicals, and Belief Reports: Some Dilemmas (But Not the Ones You Expect).” Mind 104: 107–​131. Schiffer, S. 2003. The Things We Mean. New York: Oxford University Press. Searle, J. R. 1979. “The Logical Status of Fictional Discourse.” In Contemporary Perspectives in the Philosophy of Language, edited by P. A. French et al., 233–​243. Minneapolis: University of Minneapolis Press, Searle, J. R. 1985. Expression and Meaning: Studies in the Theory of Speech Acts. Cambridge: Cambridge University Press. Soames, S. 1985. “Lost Innocence.” Linguistics and Philosophy 8: 59–​71. Soames, S. 1989. “Direct Reference and Propositional Attitudes.” In Themes from Kaplan, edited by J. Almog, J. Perry, and H. K. Wettstein. New York: Oxford University Press. Taschek, W. 1992. “Frege's Puzzle, Sense, and Information Content.” Mind 101(404): 767–​791. Taylor, K. A. 2002. “De Re and De Dicto: Against the Conventional Wisdom.” Philosophical Perspectives 16: 225–​265. Taylor, K. 2003. Reference and the Rational Mind. Stanford, CA: CSLI Publications.

Works Cited  291 Taylor, K. 2007a. “A Little Sensitivity Goes a Long Way.” In Context-​Sensitivity and Semantic Minimalism: New Essays on Semantics and Pragmatics, edited by G. Preyer and G. Peter. Oxford: Oxford University Press. Taylor, K. 2007b. “Misplaced Modification and the Illusion of Opacity.” In Situating Semantics: Essays on the Philosophy of John Perry, edited by M. O’Rourke and C. Washington, 215–​250. Cambridge, MA: MIT Press. Taylor, K. 2010. “On Singularity.” In New Essays on Singular Thought, edited by R. Jeshion. Oxford: Oxford University Press. Taylor, K. 2011. “Conceptual Relativism.” In A Companion to Relativism, edited by S. Hales. Oxford: Wiley Blackwell. Taylor, K. 2012. “Reference and Jazz Combo Theories of Meaning.” In Reference and Referring: Topics in Contemporary Philosophy, vol. 10, edited by J. K. Campbell, M. O’Rourke, and H. Silverstein. Cambridge, MA: MIT Press. Taylor, K. 2014a. “Names as Devices of Explicit Coreference.” Erkenntnis 80: 235–​262. Taylor, K. 2014b. “The Things We Do with Empty Names: Objectual Representations, Non-​Veridical Language Games, and Truth Similitude.” In Thinking and Talking about Nothing, edited by M. García-​Carpintero and G. Martí. Oxford: Oxford University Press. Taylor, K. 2015a. “How to Hume a Hegel-​Kant.” Philosophical Issues 25(1): 1–​ 40. Taylor, K. 2015b. “Names and Other Devices of Explicit Co-​reference.” Erkenntnis 80(2): 235–​262. Taylor, K. 2017. “Selfhood as Self Representation.” In Reference and Representation in Thought and Language, edited by M. de Ponte and K. Korta. Oxford: Oxford University Press. Taylor, K. 2019. Meaning Diminished: Toward Metaphysically Modest Semantics. Oxford: Oxford University Press. Thomasson, A. 1999. Fiction and Metaphysics. Cambridge: Cambridge University Press. Wettstein, H. K. 1986. “Has Semantics Rested on a Mistake?” Journal of Philosophy 83(4): 185–​209. Wittgenstein, L. 1953. Philosophical Investigations, edited by G. E. M Anscombe and R. Rhees. Translated by G. E. M. Anscombe. Oxford: Blackwell. Zalta, E. 1983. Abstract Objects: An Introduction to Axiomatic Metaphysics. Dordrecht: D. Reidel. Zalta, E. 1988. Intensional Logic and the Metaphysics of Intentionality. Cambridge, MA: MIT Press.

Index For the benefit of digital users, indexed terms that span two pages (e.g., 52–​53) may, on occasion, appear on only one of those pages.  Abdullah, Haazim, 215 act of thinking, 35, 145–​46 ambiguity, 69–​70, 107, 193–​94 ambiguity theorist, 107 anaphoric chain, 46, 80, 97, 100–​1, 106–​7, 230, 242–​43, 279, 280 anti-​psychologism,  147–​48 anti-​representationalism,  28 apperception, 49 Aristotle, 1, 163–​64   Berkeley, George, 1 binding theory, 110–​11 Brandom, Robert, 4–​5, 57, 62, 65–​66, 67, 68, 75, 233–​34, 270 Brentano, Franz, 15, 17, 18, 21–​22, 32, 236, 237, 271 Burge, Tyler, 104–​5, 119–​20,   causal/​informational theory,  159–​60 cause-​norm gap, 63–​64, 65 Cicero, 115–​16, 137–​38, 139, 140, 233 Claus, Santa, 1, 16, 17, 134–​35, 185, 229–​30, 235–​36, 245, 247–​48, 249,  259–​60 cognition, 13–​14, 22–​23, 29–​30, 33, 36, 49–​50, 124, 125–​26, 153–​55, 162, 167–​68, 170–​71, 172–​73, 221, 228–​29, 230, 233, 270 cognitive commerce, 26–​27, 41–​42, 220 Cognitive Criterion of Difference (CCD), 125 cognitive revolution, 11–​12 cognizer, 13–​14, 29, 34, 35, 36, 108–​9, 125, 148, 157–​58, 159, 164–​65, 167–​68, 173, 175, 201, 218–​19, 226, 227–​28, 229, 230–​31, 233, 277 co-​indexing, 110–​11,  137–​38

commonsense, 104, 143, 222–​24 communicative act, 180–​81, 204–​5, 215,  222–​24 communitarian, 156 competent rational cognition, 124, 125–​26 conation, 170–​71,  172–​73 conceptual analysis, 151 contentful, 7–​8, 11–​12, 14–​15, 52–​53, 60–​61, 62, 65, 73–​74, 75, 273–​74 conversational score, 190 Crimmins, Mark, 203–​4, 207–​8 Critique of Pure Reason, 49–​50   Davidson, Donald, 4–​5, 14–​15, 57, 68, 69, 90–​91, 182–​83,  226 deictic, 55–​56, 113–​14, 272, 280 DEICTIC, 113 demonstrative, 46, 96, 104–​5, 113, 114–​15, 120–​21, 174–​75, 198–​99, 205–​6,  228–​29 Dennet, Daniel Clement III, 13–​14 denote, 37, 90–​91, 127–​28, 135, 181–​82, 203–​4, 235–​36,  263–​64 derogation, 187, 188–​89, 191–​93, 198–​99, Descartes, René, 1, 13 designator, 102, 110–​11, 126, 127–​28, 131–​32, 133, 165–​66, 167–​68, 178–​79, 209,  266–​69 determinate extension, 152–​53 determination, 25, 34, 75, 86, 87–​88, 125, 135–​36, 144, 153, 156 determiners, 98–​99, 118 suppressed, 121 dialectical significance, 70–​71, 130–​31,  209 direct acquaintance, 36–​37, 42, 43–​44 direction of fit, 73 disambiguated, 93–​94, 104–​5,  106–​7

294 Index discursive community, 30–​31, 38, 78–​79, 85–​86, 147–​49, 156,  164–​65 disjoint reference, 110–​11 doxastic state, 219–​20 drainage thesis, 105–​6, 113 dualism, 13, 15, 17 Cartesian, 13, 172 mind-​body,  17 Dummett, Michael, 60–​61 dynamic priority, 6, 72, 74, 75, 77–​78, 79–​80, 84,  85–​86   embedded definite descriptions, 193, 196, 197–​98, 200–​1, 205, 212, 283, embedded predicates, 198–​99, 207–​8 empirical, 9–​10, 28, 107, 144–​45, 250 empiricism, 7, 27–​28, empty names, 75, 116–​17, 237, 243, 245–​46, 253, 262 en rapport, 41–​42, 220 epistemic, 25–​26, 31, 42, 43–​44, 48, 51–​52, 54–​55, 77–​78, 80, 89, 100–​1, 102, 134–​35, 136, 148–​49, 152, 171–​73, 174, 175, 220, 221, 225, 228–​29, 231, 233–​34, 256, 257, 258–​59, 264–​65 epistemology, 129 evening star, 33, 35, 36 expressing, 3, 6–​8, 10, 12–​13, 18 expressive commitments, 190–​91, 192, 193, 200 expressivists,  71–​72 external coherence, 49, 174, 230 external incoherence, 168–​69, 170, 231, 233 externalism, 45–​46, 156 externalities, 77, 246–​48, 249, 251–​52   falsity, 2, 6–​7, 9–​10, 17, 40, 53, 62, 64–​65, 72, 92–​93, 126, 178–​79, 221–​22, 236–​37, 238, 241–​42, 245, 250, 276, 285 Fara, Delia Graff, 104–​5, 119–​20 Fiengo, Robert, 136–​37, 138, 139–​40 Fodor, Jerry, 8–​9, 98, 99, 160, 161, 162 Foundations of Arithmetic, 50 free substitution, 208 Frege, Gottlob, 1, 8–​9, 17, 18–​19, 21–​22, 25–​26, 33–​34, 36–​37, 40–​41, 48–​49, 51–​52, 53, 54–​55, 59–​60, 61, 90–​91,

94–​95, 116, 128, 129, 136–​37, 139–​41, 142, 145–​46, 166–​67, 168–​69, 180, 182–​83, 201, 204–​5, 207–​8, 209, 210, 215, 219–​20, 221, 225, 227, 228–​29, 236, 264–​65, 268, 271–​72,  278, Fregean, 25–​26, 48, 89, 90–​91, 94–​95, 116, 129, 135–​37, 140–​41, 142, 166–​67, 180, 182–​83, 200–​1, 204–​5, 207–​8, 209, 215, 264 Fregean Mechanism, 182–​83, 200–​1, 204–​5,  215 Fregean Thesis, 182–​83, 204–​5, 207–​8,  215 mock proper names, 53 mock thoughts, 17, 21–​22, 53, 54, 235–​36,  271–​72 sense and reference, 33–​34, 53, 88, 125, 181–​82,  219–​20

  Grice, Paul, 77 Gricean, 77, 246–​48, 283–​84   Heidegger, Martin, 1 Hesperus, 101, 128, 130–​31, 132 Holmes, Sherlock, 1, 16, 134–​35, 258, 267,  268–​69 Homer, 162 Hume, David, 1 Husserl, Edmund, 1   idealism, 10, 16, 17, 48, 55–​56, 265, 272–​73 idealist, 9–​10, 15, 32, 237, 270 identity talk, 142–​43, 166 identity thought, 131–​32, 133, 135–​36, 166 idiom, 2, 8–​9, 90–​91, 227–​28, 238 illocutionary force, 73–​74 incomplete definite descriptions, 120–​21, 122, 123 independent existents, 6–​7, 16–​17 indexicals, 46, 96, 104–​5, 113, 114–​15,  228–​29 informativeness, 127–​28,  283–​84 intentionality, 15, 45–​46 type 1, 45–​46 type 2, 45–​46 intentionally inexistent, 17–​18, 21–​22, 271 internal coherence, 168–​69, 171–​73, 174

Index  295 internalist, 156, 171–​72, 173–​74 interpretationist,  14–​15 intersubjectivity, 69 intuition pump, 186–​87 invariants, 80, 86   Janus-​faced,  125–​26 jazz combo theory of reference, 30–​31, 57, 60–​61, 62, 63, 65–​66, 67, 68, 71, 74, 78, 79, 84, 85–​86, 87–​88, 251 Jocasta, 170–​71, 173 Jupiter, 203   Kant, Immanuel, 1, 18–​19, 25–​26, 28,  48–​49 Kantian, 25–​26, 28, 48, 153–​54 Kaplan, David, 42, 205–​6, 221, 225–​26, 231,  233–​34 Kaplanian, 42, 221 Kent, Clark, 91, 111, 169–​70, 209, 216, 217, 218, knowledge-​wh,  43 Kripke, Saul, 7, 29–​30, 36–​37, 93, 116,  193–​94   language game, 5, 54–​55, 61, 65–​66, 244–​45,  248–​49 non-​veridical, 55–​56, 76, 236–​37, 241–​42, 244–​45, 248–​49, 255, 258, 259–​60, 263,  272–​73 language of thought, 9, 109, 114–​15, 160, 175–​76,  229 lexicon, the, 81–​82 linguistic, 2, 5, 6, 7, 10, 11–​12, 21, 29, 31, 44–​45, 47, 54, 55–​56, 61, 62, 64, 66, 73–​74, 78–​79, 80, 83, 84–​85, 97, 98–​99, 101, 106, 107, 109, 110, 112, 113, 114, 116–​17, 118–​19, 127, 129, 130–​31, 137–​38, 139–​40, 147–​48, 183, 185, 188–​89, 196, 201–​2, 226, 228–​29, 236–​37, 242–​43, 245, 249, 251, 272–​73, 281 function, 59, 106 marked, 112, 137–​38 move, 5 significant, 5, 73–​74, 106, 116–​17 unit, 106 unmarked, 137

London, 94 Luthor, Lex, 170–​71, 172–​73   Mars, 202, 203, 205 materialism,  13–​14 materialist, 13, 14–​15 Matushansky, Ora, 104–​5 May, Robert, 136–​37, 138, 139–​40 mental representation, 44–​45, 160, 164–​65, 166, 169–​70, 177, 203–​4, 224–​25, 229, 233–​34, 236–​37, 277 mentalese, 9 metaphysical, 6, 13–​14, 15, 35–​36, 68, 70–​72, 81, 88, 90–​91, 146–​48, 151–​52, 157–​58, 161–​62, 173–​74, 225–​26, 227–​28, 239, 256–​57, 259–​60,  265 metaphysically, 4–​5, 13, 46–​47, 61–​62, 101, 168–​69, 224–​25, 266 possible world, 168–​69, 224–​25 metaphysics, 81–​82, 87–​88, 256–​57,  259–​60 metasemantic,  96–​97 mind-​independent, 2, 10, 18–​19, 22, 26–​27, 67–​68, 71–​72, 154–​55, 177, 226, 254, 255–​56, 257, 273–​74 morning star, 34, 35, 36 mutatis mutandis, 65 mutual understanding, 180–​81   nambiguous,  103–​5 NAME, 80, 98–​99, 101, 107, 113, 228–​29 non-​displaceability, 187, 188–​89,  205–​6 normative, 29–​30, 59, 63, 65–​66, 67, 146–​48, 168–​69, 171–​72, 173–​74, 222, 240, 261 notionally sensitive, 179–​80, 183   objective representational content, 2, 4–​5, 6–​8, 9–​10, 12–​13, 15, 17–​19, 21–​22, 25, 29, 30–​31, 45–​46, 48, 49, 57, 59–​60, 61–​62, 65–​66, 70–​71, 72, 78–​79, 85–​86, 87–​88, 97, 99–​100, 125–​26, 153–​54, 155–​56, 219–​20, 235–​36, 238, 271–​72,  273–​74 objectivity, 15, 21–​22, 49, 55–​56, 68, 69–​70, 71–​72, 78, 105, 230, 243–​45,  271–​73

296 Index objects of thought, 2, 9–​10, 17–​18, 21–​22, 41, 48–​49, 235–​36, 271 objectual, 18–​19, 21–​22, 25–​26, 46–​47, 50, 51–​53, 55–​56, 78, 79–​80, 83, 87–​88, 100–​1, 105, 108–​9, 154, 230–​31, 232–​33, 237, 241, 244, 253, 255, 259–​60, 262, 265, 270, 271, 272–​73 Oedipus, 168–​69, 173 opacity, 91, 142, 178–​79, 209 orthodoxy, 89, 102, 103, 151, 180   Pegasus, 17, 235–​36, 277–​78 Perry, John, 60–​61, 103–​5, 106–​7, 129, 130, 139–​40, 207–​8,  277–​78 philosophy, 2, 7–​8, 11–​12, 20–​21, 22, 42–​43, 48–​49, 89, 136, 144–​45, 151,  155–​56 Phosphorous, 128, 130–​31, 132–​33, 166, 168–​69, 201, 209, 229–​30 Plato, 1, 38–​39 Platonic,  145–​47 Platonism, 55–​56,  152–​53 Platonist, 152–​53, 237, 256 plenum, 172, 174, 175–​76 predicate, 3, 4–​5, 6–​7, 37–​38, 80, 104–​5, 118–​19, 120–​21, 123, 179–​80, 181–​82, 198–​99, 207–​8, 228–​29, 258, 259 name-​like, 118–​19, 120–​21,  123 predicative commitment, 183–​84, 185, 191–​92, 198–​99, 200, 207–​8, 221–​22, 225, 226, 227–​28, 274–​75 priority of the sentence, 4–​5, 6, 59–​60, 62, 75, 78, 79–​80, 84, 86 properties, 2, 4–​5, 6–​7, 9, 10, 13, 18–​19, 20, 25–​26, 28, 29–​30, 33, 35–​36, 38–​40, 45–​46, 95–​96, 98–​99, 107, 127–​28, 151, 157, 165, 178, 179–​80, 203–​4, 207–​8, 221–​22, 227, 261, 273–​74 propositional content, 59, 62, 63–​64, 65–​66, 67, 73–​74, 77, 86, 97, 126–​28, 129, 140, 246–​47, 251–​52, 271–​72, 274–​75,  277 propositions-​in-​waiting,  76 psychologism,  144–​45 psychology, 49, 81–​82, 140–​41, 142–​44, 145–​46, 149–​50, 151–​52, 153, 167–​68, 175–​76,  227–​28 psychological conception, 144–​45, 146–​47,  149

publicity, 146, 149, 153, 156 role,  175–​76 public language, 1, 2, 7–​8, 9, 110, 142–​43,  228–​29 Putnam, Hilary, 29, 148–​49, 161   QUANTIFIER, 80 Quine, Willard Van Orman, 14–​15, 18–​19, 48–​49, 90–​91, 221, 236, 265   Recanati, François, 77, 104–​5, 139–​40, 203–​4, 207–​8,  251–​52 reference relation, 4–​5, 6–​7, 57, 61–​62, 68, 137, 214, 262, 263–​64, 266 referentialism, 36, 41, 43–​44, 237, 239 fitness, 18, 19, 48, 52–​53, 78, 85–​86, 87–​88, 105, 108–​9, 230–​31, 232–​33, 241, 244, 253, 262, 265, 268–​70 independence, 93–​94, 102, 105–​6, 107, 110–​11, 112, 113, 120–​21, 127–​28, 131–​32, 136, 166, 167–​68, 170, 173, 209, 210, 228–​29, 242–​43, 263 purport, 53, 78–​79, 100–​1, 104–​7, 113, 135, 243, 256, 259–​60, 262, 266, 268, 282 referentialist, 29–​30, 76, 88, 96, 157–​58 semantic, 25–​26, 31 success, 19, 47, 52–​53, 78, 83, 84–​85, 87–​88, 100–​1, 105, 113, 116–​17, 230, 241, 242–​43, 244, 253, 256, 259–​60, 262, 265, 266, 267–​68, 269–​70 two-​factor, 3–​4, 18, 19, 45–​47, 87–​88, 89, 99–​100, 150, 164, 221 representational vehicles, 2, 99–​100, 129, 142–​43,  169–​70 Richard, Mark, 91–​92, 203–​4, 207–​8 Russell, Bertrand, 1, 18, 21–​22, 37–​38, 39, 42, 152, 207–​8, 221, 233–​34, 236, 237, 242–​43, 271–​72, 275–​76,  278 Russellian, 41, 42, 152, 207–​8, 221   saturation, 77, 246–​47, 248–​49, 251 Saul, Jennifer, 91, 208 Schiffer, Stephen, 203–​4 Searle, John, 15 secondary realities, 154 self-​negating,  171–​72 semantics, 75, 83, 88, 89, 94, 96, 97, 103–​4, 105, 123–​24, 126, 140, 142, 160,

Index  297 194, 210, 215, 245–​46, 250, 251–​53, 256–​57, answerability, 2, 6–​7, 8, 9–​10, 19–​20,  36–​37 bootstrapping, 75, 76, 79–​80, 82 composition, 3, 81–​82 contact, 18–​19, 21–​22, 40–​41, 71–​72, 78–​79,  116 content, 10, 25–​26, 28, 31, 83, 96–​97, 249, 250, 273–​74 holism, 79 posteriority, 72 presentationalism, 25–​26, 31, 35–​36, 88 presentationalist, 25–​26, 33 priority, 59–​60, 61–​62, 66, 72, 75, 77–​78, 79–​80, 84,  85–​86 priority of the constituent, 84, 86,  87–​88 property, 6 referentialism, 25–​26, 31, 43–​44 role, 6–​7, 39, 110, 116–​17, 125, 228–​29 value, 3, 75, 82, 92 sentence, 80, 98–​99 shared contact, 1 simpliciter,  106–​7 social-​dialectical, 67, 71 speech act, 5, 61–​62, 129, 190–​91 spelling, 96, 99, 103, 106, 110, 111–​13, 120, 123,  203–​4 subsentential,  73–​74 Superman, 91, 111, 169–​70, 209, 216, 217 syntactics, 10 category, 80, 98–​99 combination, 3, 81–​82 correlativity, 72, 79–​80, 81, 83, 84, 85–​86,  98–​99

landscape, 174, 175–​76 plasticity, 175 role, 25–​26, 84–​85, 106–​7, 228–​29,  242–​43 self-​negation,  174 syntax, 81, 83, 89, 95, 99–​100, 103–​4, 105, 106–​7, 110, 123–​24, 126, 140, 142–​43, 160, 166–​67, 194, 250, 252–​53 combinatorial, 98, 99–​100, 160 of singular reference, 94–​95   Taylor, Daniel, 213, 215 Taylor, Kiyoshi, 108 thought episode, 7, 26–​27, 108, 132–​33, 142–​43, 152, 153, 157–​58, 162–​63, 167–​68, 169–​70, 227–​28, 229, 230,  232–​33 tokening, 20–​21, 73–​74, 80, 86, 115–​17, 162, 165–​66, 167–​68, 169–​70,  230–​31 triad, 72, 84, 85–​86 triangulation,  69–​70 triviality,  127–​28 Trump, Donald J, 33, 39–​40, 221–​22 Twain, Mark, 90, 92 twin earth, 148–​49, 161   Universal Grammar hypothesis, 81–​82, 108, 110   vehicle of thought, 2, 3, 126–​27, 132,  142–​43 Venus, 101, 132, 165–​66, 202, 203, 205,  229–​30 VERB, 80, 98–​99   Wittgenstein, Ludwig, 1, 5, 79