Realism Antirealism and Epistemology

REALISM/ANTIREALISM AND EPISTEMOLOGY

Edited by Christopher B. Kulp

ROWMAN & LITTLFIELD PUBLISHERS, INC. Lanham· New York· Boulder· Oxford

ROWMAN & LITTLEFIELD PUBLISHERS, INC. Published in the United States of America by Rowman & Littlefield Publishers, Inc. 4720 Boston Way, Lanham, Maryland 20706 12 Hid's Copse Road Cummor Hill, Oxford OX2 9JJ, England Copyright © 1997 by Rowman & Littlefield Publishers, Inc. Chapter 2, "Does the Real World Exist?" by John R. Searle, originally published in The Construction a/Social Reality. © 1995 by Free Press. Reprinted by permission. Chapter 5, "Indeterminism and Antirealism," © 1997 by Donald Davidson. Chapter 7, "Realism, Antirealism, and Pragmatism," © 1997 by Richard Rorty.

All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, photocopying, recording, or otherwise, without the prior permission of the publisher. British Library Cataloguing in Publication Information Available Library of Congress Cata1oging-in-Publication Data

Realism/antirealism and epistemology / edited by Christopher B. Kulp. p. cm. - (Studies in epistemology and cognitive theory) Includes bibliographical references and index. ISBN 0-8476-8335-4 (alk. paper). - ISBN 0-8476-8336-2 (pbk. : alk. paper) 1. Realism. 2. Knowledge, theory of. 3. Justification (theory of knowledge). I. Kulp, Christopher B. II. Series. 96-34977 B835.R323 1997 CIP 149'.2---dc20 ISBN 0-8476-8335-4 (cloth: alk. paper) ISBN 0-8476-8336-2 (pbk. : alk. paper) Printed in the United States of America

8™The pa~t: u~~ in this publication meets the minimum requirements of American

NaUonal St~~'''t~r Informati,on,S,ciences-Permanence of Paper for Printed Library Mat~riaPs, 2f'9.f8-1984. " f

ANSr

To my parents, Lucille and Henry Kulp

Contents ix

Acknowledgments

1

In trod uction

1

Christopher B. Kulp

2

Does the Real World Exist?

15

John R. Searle

3

Realism and the Tasks of Epistemology

53

William P. Alston

4

Why the Theory of Knowledge Has to Be Realistic

95

Roderick M. Chisholm

5

Indeterminism and Antirealism

109

Donald Davidson

6

Pragmatism and Reasons for Belief

123

Gilbert Harman

7

Realism, Antirealism, and Pragmatism: Comments on Alston, Chisholm, Davidson, Harman, and Searle

149

Richard Rorty

Index

173

About the Contributors

177

vii

Acknowledgments

The papers in this volume stem from a conference, "The Implications of Realism and Antirealism for Epistemology," held at Santa Clara University in February 1992. I wish to thank the university for the funding and support that made the conference possible, and my friend and colleague Philip Kain for his considerable assistance in putting on the conference. I also wish to thank the contributors for their willingness to publish their papers, in many cases in substantially revised form, as well as the Free Press for permission to reprint John R. Searle's contribution from his book The Construction of Social Reality.

ix

Introduction

Christopher B. Kulp

here has been marked disagreement during the last few decades over just how we should understand the terms 'realism' and 'antirealism'. To what do the terms refer? This is no easy question. Some writers think there is something specific that makes a theory, proposition, or term X realist or antirealist. Others despair of being able to specify the property (or set of properties), or the necessary and jointly sufficient conditions that would qualify X as realist or antirealist. They think there is nothing common to all realisms and antirealisms. And among those who think there is something specific that makes X realist or antirealist, there is disagreement over what this is. All of this needs a bit of unraveling. Take the first group-those who think there is something specific for X to be realist. Some members of this group see realism as fundamentally an ontological thesis. They think that what is common to realisms is the ontological claim that some relevant thing Y exists. Conversely, they think that antirealisms imply that Y doesn't exist. To illustrate, consider naive realism in the theory of perception. According to naive realism, when person S sees a tree, what S sees-the object of S's visual perception-is a material object, a tree, which exists independently of S's perceptual act. Naive realism is a realist theory of perception in that it is committed to the "reality" of material objects. Material objects, as a class of mind-independent entities, are accorded positive ontological status. Contrast this with the version of phenomenalism advanced by John Stuart Mill in An Examination of Sir William Hamilton's Philosophy. 1 According to this theory, when S has a visual perception of a "tree," what S actually perceives is not a material object but a visual sensation. Material objects are not mind-independent entities but "permanent possibilities of

T

2

Realism/Antirealism and Epistemology

sensation"; and 5's "tree" is to be understood as the sensations 5 would have were 5 subject to certain conditions. Thus, unlike naIve realism, Mill's phenomenalism does not accord positive ontological status to material objects qua mind-independent entities. One might say it is antirealist with regard to material objects. We can go on in a similar vein with other instances of realism and antirealism. Realist/ antirealist debate in mathematics deals with issues such as whether the objects of mathematical inquiry-for example, numbers or sets-exist independently of mathematicians or are merely "constructed," the conceptual creations of human beings. In ethics, realists and antirealists dispute the existence of culture- or language-independent moral facts. In the philosophy of mind, they dispute the existence of mental states that are irreducible to physical states. Some philosophers, however, see the distinction between realism and antirealism not so much in terms of ontology as of truth. On one prominent version of this construal of realism, a proposition (statement, sentence) p is "realist" if and only if it says something about the way the world is. Thus, a proposition p of the form 'X is
3

Introduaion

understood in its normal way, as referring to a perceiver-independent material object, because it doesn't countenance the existence of perceiverindependent material objects. Material objects are not perceiver-independent entities; they are "permanent possibilities of sensation." Truth claims are made about sequences of sensations-"If 5 were to do such and such, 5 would have such and such a perceptual experience"-not about perceiving material objects. Thus, phenomenalism is antirealist. Similar things apply to other cases of realism and antirealism. Realist/ antirealist debate in the philosophy of mathematics deals with whether mathematical truth and falsehood are determined by a mindindependent world of numbers (or other mathematical entities) or by the characteristics of an elaborate conceptual scheme(s) developed by human beings for human purposes. Realists endorse the former, antirealists the latter. In ethical theory, debate deals with whether moral utterances can express moral propositions which are true or false depending on whether they correspond to moral facts-a realist position-or whether moral utterances are neither true nor false and therefore don't express moral propositions because there are no such things as moral facts-an antirealist (noncognitivist) position. Thus, among those who think there is something specific to realism and antirealism, there is divergence over what that something is: some think it is ontology, others truth and falsehood. But what of the second group I mentioned above, those who think that there is nothing definitive of realism and antirealism? There is divergence here as well. Some prefer not to speak of realism and antirealism per se, but of realisms and antirealisms-of different kinds or cases of realism and antirealism. Thus, in moral theory, for example, there is nothing unique to each and every moral proposition (or metaethical theory, or what have you) widely regarded as realist or antirealist. The necessary and jointly sufficient conditions requisite for a moral-X to be a realist or antirealist moral-X elude us, for the simple reason that there aren't any such conditions. 2 Others think there are different kinds of realism that may be graded by their respective "strengths," as determined by the degree or type of contrast with types of antirealism. 3 There are a variety of views, then, concerning what constitutes realism and antirealism. There is also variation among the contributors to this volume. Consider John Searle. Early in part 1 of his essay "Does the Real World Exist?" he tells us that although philosophers have used the term 'realism' in a variety of ways, what he means by the term is what he calls "external realism": "the view that there is a way that things are that is logically independent of all human representations" (20).4 By 'representations', Searle has ~..... '...~~.l

Ut~:

·r. "

:;f y

':"'H.

I'

4

Realism/Antirealism and Epistemology

in mind things like "perception, thought, language, beliefs, and desires as well as maps, diagrams, etc." (16). Antirealism denies that there is any such representation-independent reality. Contrary to what many believe, "realism is not a theory of truth, and it does not imply any theory of truth" (18). Nor does realism have "something epistemic about [it]" (19) or commit one "to the theory that there is one best vocabulary for describing reality, that reality itself must determine how it should be described" (19). As Searle sees it, external realism is an ontological theory, which "does not say how things are but only that there is a way that they are" (19). It is a theory that has the consequence that if we had never existed, if there had never been any representations ... most of the world would have remained unaffected. Except for the little corner of the world that is constituted or affected by our representations, the world would still have existed and have been exactly as it is now. (IS)

Clearly, then, Searle understands realism in ontological terms, not in terms of truth. Roderick Chisholm, in "Why the Theory of Knowledge Has to Be Realistic," adopts a view similar to Searle's. Chisholm says that he will use the terms 'realism' and 'antirealism' in line with Michael Dummett,5 who avers that clarity would be served by referring to realism regarding specific subject-matters. "Thus," says Chisholm, "most of us could be said to be realists about dogs, or the existence of dogs, and not realists concerning mermaids" (95). But this is "a question of zoology, not a question of philosophy" (95). Philosophical questions arise when questions are raised about the ontological status of kinds of entities: when it is asked "whether entities of a certain sort may be found 'among the constituents of reality'" (95). Unlike Dummett, however, Chisholm's primary concern is "not with language . .. [but] with those nonlinguistic questions we must be able to answer if we are to make any reasonable decisions about the meaning and reference of the language that we use" (95-96). These are questions of ontology. And raising ontological questions about "the ultimate constituents of reality" (96) has traditionally been known as "a theory of categories." Hence, Chisholm sees (philosophical) questions of realism in terms of ontological categories. William P. Alston, however, has much to say about the connection between realism and truth in his essay, "Realism and the Tasks of Epistemology." Alston terms the realism he is discussing "alethic realism," which he considers "primarily a doctrine about truth" (70), and which he

Introduction

5

takes to be the conjunction of the following two theses: 1. The senses of 'true' and 'false' in which such items as beliefs, statements, and propositions can be evaluated as true or false are what I will call realist senses of these terms. 2. It is important to determine the truth-value of such items in this sense (53).

The basic idea of this conception of truth, says Alston, "is that what determines whether a statement is true is whether what the statement says to be the case is the case" (53-54). Thus, if 5 states that grass is green, S's statement is true if and only if grass is green, and false otherwise. As Alston sees it, this is a realist conception of truth because what it takes to make a statement true on this conception is the actual obtaining of what is claimed to obtain in making that statement. If what is stated is that grass is green then it is grass's being green that is both necessary and sufficient for the truth of the statement. Nothing else is relevant to its truth value. This is a realist way of thinking of truth in that the truth maker is something that is objective vis-a-vis the truth bearer. It has to do with what the truth bearer is about, rather than with some "internal" or "intrinsic" feature of the truth bearer, such as its epistemic status, its place in a system of propositions, or the confidence with which it is held .... Truth has to do with the relation of a potential truth bearer to a reality beyond itself. (54)

Although I have portrayed Alston's discussion of realism in terms of its connection with truth - which is indeed the focus of his essay-I should be clear that he doesn't think alethic realism is the only legitimate form of realism. There are "departmental realisms"-realism about universals, realism about intentional psychological states, about values and moral obligation, etc. (60). But alethic realism is different from these realisms "in being unrestricted in scope. Alethic realism holds that all statements are usefully assessed for realist truth value" (60). A statement is true if and only if it says that things are the way they are. In certain respects Gilbert Harman, in his "Pragmatism and Reasons for Belief," sounds a theme similar to Alston. Realism, says Harman, holds that what makes an answer to a question correct is that the answer "represents things as they really are" (123). Thus: Consider the question of whether numbers exist. According to realism, the answer to this question depends on whether in reality numbers exist. If they do, the correct answer is yes; otherwise the correct answer is no. (123).

6

Realism/Antirealism and Epistemology

Antirealism, on the other hand, denies that correctness is a matter of representing things as they really are. There are, of course, different forms of antirealism, and at least one form of antirealism- or at least what is often considered a form of antirealism-pragmatism, denies that it even makes sense to suppose that numbers (or what have you) either "do or do not exist in reality" (123). This is because for the pragmatist, the question "Do numbers exist?" is, like other questions, at least in part a practical matter, the answer to which properly depends on "whether things work out better if one allows quantification over numbers in a theory one is developing" (123). As Harman sees it, however, pragmatism is better understood not as a version of antirealism but as rejecting the realist/ antirealist distinction, since for pragmatism "there is no sharp distinction between theoretical and practical questions" (123). Qua pragmatist, then (though he doesn't explicitly avow it here), Harman rejects the distinction between realism and antirealism. Donald Davidson also rejects the distinction in his "Indeterminacy and Antirealism," although it isn't as clear what he takes the putative realist/ antirealist distinction to be. Davidson says that antirealism is a manifestation of the irrepressible urge in western philosophy to insure that whatever is real can be known. Antirealism attempts to achieve this by reading out of existence whatever it decrees lies beyond the scope of human knowledge (109).

Indeed, by Davidson's lights "most reductive -isms ... [are] forms of antirealism: idealism, pragmatism, empiricism, materialism, behaviorism, verificationism" (109), because they all attempt to make reality comply with the requirements of their respective epistemologies. Thus, phenomenalism holds that "physical objects are nothing but permanent possibilities of sensation"; and idealism holds that ordinary objects like "tables and dinnerware" are real, "but exist only in our minds" (109). These and other antirealisms allow us to keep "the terminology of our old ontology as long as we agree to accept only what can be cobbled together out of entities we can know for certain" (109). On the other hand, Davidson thinks it advantageous to understand some antirealisms "as epistemic limitations on the concept of truth" (109). In such cases, if our epistemic faculties fall short of permitting us to determine the truth or falsity of a sentence, we may "rule that the sentence has no truth value, or that we should employ some reduced sense of truth" (109). But the result is the same as with the other construal of antirealism:

Introduction

7

"the real or the true is cut down to the size of a favored form of knowledge" (109-10). All this said, however, Davidson doesn't consider realism to be the alternative to antirealisms, for "one can reject one or another version of antirealism on the grounds that the arguments for it fail without having to endorse some vague position called realism" (110; my emphasis). Consider for instance the common construal of realism as the thesis that "there is something in or about the world" that somehow makes truth-bearers true independent of our "power to determine" their truth. Not only are there problems with "spelling out what 'power' and 'determine' mean in this context"(110); there are problems with even understanding the "prior claim" that there is something in or about the world that makes truth-bearers true. No one, thinks Davidson, has given a nontrivial explication of what it is for the world to make truth-vehicles true or false. Finally we come to Richard Rorty, who, in his "Realism, Antirealism, and Pragmatism: Comments on Alston, Chisholm, Davidson, Harman and Searle," is less explicit than the others about what he takes realism and antirealism to be. Nevertheless, it is clear that he wishes to reject the distinction. After avowing his pragmatism (149), Rorty says that he proposes to follow Colin McGinn and Alston in holding that realists aren't realists tout court but are realists relative to specific subject matters. And he goes on to say that "The terms 'realist' and antirealist' should be confined to people willing to take seriously the question 'Is there a matter of fact about X?' for some specified X" (151). This question, however, "neatly fuses" two "bad questions," the "epistemological question 'Can ... be known?' with the equally bad ontological question 'Is ... really real?'" (151), both of which should be dropped because they grow out of the scheme-content distinction, which also should be dropped. By Rorty's lights, then, we should quit talking about realism, and therefore about antirealism. Thus three of the contributors-Alston, Chisholm, and Searle-are full-blown realists; the other three-Davidson, Harman, and Rortydemur. Searle and Chisholm construe realism as an ontological thesis, while Alston, though not denying realism's ontological import, construes it (contra Searle) in terms of truth. Harman and Rorty, on the other hand, raise doubts about the realist/antirealist distinction's legitimacy on pragmatist grounds. And Davidson calls it into doubt because he thinks that no one has yet made nontrivial sense of the very concept of realism. But what of their broader views on the bearing of realism and antirealism on epistemology? Searle's contribution to this volume is, as he says,

8

Realism/Antirealism and Epistemology

"about realism" (15); and in "Part 1: Attacks on Realism," he is at pains to say what he takes (external) realism to be (this we have already discussed) and to examine what he judges the most persuasive objections to realism. He develops three arguments against realism: (1) the argument from conceptual relativity (roughly, that because there is any number of possible systems for representing reality, reality doesn't exist independently of our representations); (2) the verificationist argument (roughly, that if we assert, as does the external realist, that there is something beyond our experiences, what we assert is empirically vacuous and thus indefensible); and (3) the Ding an Sich argument (roughly, that externalism is unintelligible because any cognition occurs within a cognitive system, and it is impossible to escape this or some other cognitive system to determine whether the system properly represents the [putatively] externally real). Searle argues that each of these arguments is invalid and therefore fails to defeat realism. In "Part 2: Could There Be a Proof of External Realism?" Searle argues that external realism is properly understood not as an empirical theory but as a background presupposition of the intelligibility of a large part of our thought and language. Furthermore, he argues that external realism has no commitment to any particular kind of existents, for example, material objects. In defense of external realism, he offers "transcendental" arguments against what he considers the two most (relevantly) important forms of antirealism, "phenomenalist idealism" ("the view that all reality consists in conscious states" [40]), and "social constructionism" ("the view that reality is socially constructed" [40].) Against phenomenalist idealism, he argues that "for a large class of utterances, each individual utterance requires for its intelligibility a publicly accessible reality" (145; my emphasis)-the sort of "reality" asserted by external realism. Against social constructionism, he argues that there is a large class of utterances whose normal understanding pressupposes a representation-independent reality; and that a presupposition of socially constructed reality is a reality independent of social constructions. Searle does not claim, however, to have proved that external realism is true: he doubts that any nonquestion begging proof of realism is possible. Rather, he claims to have shown that much of our language presupposes external realism and that unless we are ready to give up this part of language-a decision with wide-ranging and profound consequences-we remain committed to realism. Alston also adopts a pro-realist stance, specifically, a pro-realist stance regarding alethic realism (the two theses of which I have exposited above). He develops what he considers a plausible minimalist account of

Introduction

9

realist truth, an account that embodies the "T-scheme," namely, "The proposition p is true iff p" (57)-and argues against epistemic conceptions of truth, especially Hilary Putnam's pragmatist theory. He also argues that Rorty is mistaken in his (former) view that it is appropriate to simply "turn one's back" on truth, that is, to hold that we shouldn't consider it important whether our beliefs, statements, or what have you are true. Alston devotes a substantial portion of his essay to its main theme, "the bearing on epistemology of the choice between alethic realism and its alternatives," (69), and finds the implications both legion and in favor of a realist conception of truth. Truth is important to the analysis of knowledge, to many accounts of internalist epistemic justification, to the distinctness of externalist theories of justification, and even to the very notion of epistemic justification; and antirealist conceptions of truth run into more problems than do realist conceptions. Alston's view in brief is: "Alternatives to alethic realism have nothing to recommend them in the field of epistemology" (91). Chisholm says that he wants to defend realism and oppose antirealism in epistemology and that he aims to do four things in his paper: (1) say what realism and antirealism are, (2) say what the theory of knowledge is, as traditionally conceived, and how it should be investigated, (3) argue that epistemology must be realistic, and (4) criticize antirealism in epistemology. I have already addressed (1). Regarding (2), he tells us: The theory of knowledge is that Socratic inquiry suggested by the following questions: (1) "What can I know?"; (2) "How can I distinguish beliefs that are reasonable for me from the beliefs that are not?"; (3) "How can I tell whether I have a good reason for a belief?"; and (4) "What can I do to replace unreasonable beliefs with reasonable beliefs about the same subject matter, and to replace beliefs that are less reasonable by beliefs that are more reasonable?" (97).

He goes on to say that there are three stages to this Socratic inquiry: "methodological doubt," "finding a foundation," and "the road back" (97). A substantial part of Chisholm's paper is an elaboration of these three stages of inquiry. As to goal 3 of his paper, he avers that we should be realist about substance (e.g., individual things), about the objects of perception (e.g., physical objects), about the objects of memory and about states of things. As regards goal 4, Chisholm considers three claims in defense of antirealism, claims (loosely) supporting Nietzsche's view (as Chisholm understands Nietzsche) that "all claims to ... knowledge are false, their objects are fictions, and they have no basis in reality and make no reference to reality" (105-6); and finds inadequate support for any of them.

10

Realism/Antirealism and Epistemology

He concludes that he has not found good positive reason for endorsing antirealism in epistemology. Davidson is also critical of antirealism. I/[A]ntirealisms," he says [are] sour-grape philosophies. Their motto is: if you can't grasp the grapes (in some approved sense), they weren't just sour; they were never there in the first place" (109). In this paper Davidson takes up antirealist arguments that the thesis of the indeterminacy of translation (or "interpretation," as he prefers to call it) provides grounds for doubting the reality of mental states and events and of proposition attitudes. Davidson endorses (and spends a good bit of time defending) the indeterminacy of interpretation and says that he I/would be aggrieved to find out that it entails some sort of antirealism" (110). But he does not think it does. He argues that the indeterminacy of interpretation "leaves the reality of the mental untouched" (113)-that sentences making assertions about mental states and propositional attitudes can be objectively true or false. Harman holds, as I noted earlier, that pragmatists should be chary of the whole realist-antirealist distinction, but he nonetheless considers it interesting that from an intuitive realist perspective, there is a clear distinction between reasons that bear on the truth of what is believed and reasons that bear on practical considerations for belief-between "purely epistemic reasons to believe something and other, more pragmatic nonepistemic reasons" (123). Harman's central concern is "with what pragmatism can say about this intuitive distinction" (124). Harman points out that pragmatists hold that there isn't a sharp distinction between epistemic and practical reasons for belief, but this, he thinks, leads to a problem. He agrees that there is the clear intuitive distinction of which realists speak, but he believes that pragmatists are right in accentuating the practical, goal-directed character of reasoning and inquiry. "How," he asks, "can the intuitively realist distinction between epistemic and nonepistemic reasons be accommodated while allowing for the insights of pragmatism?" (128). Perhaps epistemic reasons are connected with truth in a way that practical reasons are not. Harman considers this suggestion to be on the right track, although it leads to a variety of problems. He addresses them in the context of subjective probability theory, where much of his discussion centers on issues dealing with Temporal Dutch Book arguments (a Dutch Book is, roughly, a collection of bets such that were a person to bet on them, she would be guaranteed to loose; a Temporal Dutch Book is, roughly, a Dutch Book comprised of bets to be made over time), and the Rule of Conditionalization (roughly, a rule for updating one's probabilities for belief in light of new evidence). His analysis enables him, he believes, to explain the difference between epistemic and nonepistemic 1/ •

••

Introduction

II

reasons in terms of how one assigns one's subjective probabilities. He concludes that pragmatists can allow for the epistemic/nonepistemic distinction "without supposing that all practical reasons for belief are nonepistemic reasons" (144). Rorty begins his paper with a discussion of pragmatism'S perspective on the implications of realism and antirealism for epistemology, which, he avers, is to "discuss the consequences of an unhappy distinction for a dubious discipline" (149). We have seen that Rorty thinks pragmatism rejects the distinction between realism and antirealism; he also thinks that pragmatism wishes to get rid of the "scheme-content distinction, and thus the question of the adequacy of our minds, or our language, to reality" (149). With the rejection of the scheme-content distinction goes the rejection of both ontology (that is, "an account of content" [149]) and epistemology (that is, "an account of the relation between content and scheme" [149]). We need not concern ourselves with replying to the skeptic. Rorty then moves to a discussion of the contributors' papers. In Davidson he finds an ally-"I persist in trying to interpret Davidson as the most sophisticated and radical of the American pragmatists" (149)-but raises concerns when Davidson "flirts" (151) with the subject-object distinction, a distinction which Rorty eschews. Attendant to this flirtation are a variety of interconnected problems for Davidson, notably his willingness "to make invidious ontological distinctions" (151) and his view that there is more to indeterminacy of translation (or interpretation) than theory underdetermination. In Alston, however, Rorty sees more global problems. He questions, for example, the "realism" of Alston's conception of truth-compliance with the I-scheme isn't enough to make the theory realist: something like correspondence is needed, but Davidsonian objections to this are powerfuland he doubts that Alston has made, or could make, a persuasive case that truth can function as a "lodestar" for telling genuine justification-makers from impostors. Rorty also thinks that Searle has major problems. He points out that insofar as Searle's "external realism" is compatible with both the correspondence theory of truth and "a God's Eye View" (160), Searle is "in the position of defending a form of realism that ... no one has ever attacked" (160). But he also thinks that Searle employs two different senses of the term 'external realism', one that does, and one that does not, bear on truth and knowledge. And upon untangling the two senses of the term, Rorty finds Searle faced with unclarities, attacking views people don't actually hold, and in need of elaborating his position (which in some cases Rorty doubts could be done successfully). In addition, Rorty raises concerns

12

Realism/Antirealism and Epistemology

with the efficacy of Searle's appeal to "normal understanding" to defend realism. While Rorty grants that pragmatism is counterintuitive, he nonetheless considers pragmatism the best alternative. Chisholm, too, comes under Rorty's attack. Rorty notes, for example, that pragmatists don't think anyone has provided a clear sense of the term 'ultimate' in the phrase 'ultimate constituents of reality'. But the notion of "ultimate constituents of reality" is pivotal to Chisholm's conception of realism. He also claims that he doesn't understand what makes Chisholm's theory of categories "realistic," that he fails to see how individual substances are to count as ultimate constituents, and that he doesn't think there is a relevant connection between the sort of (self-refuting) skepticism Chisholm rejects and the rejection of realism. Harman's paper is more congenial to Rorty's views than the others, excepting (perhaps) Davidson's, but he nonetheless raises a number of concerns. Rorty says that he has "two sets of doubts about ... Harman's paper" (164). The first set clusters around "the seeming artificiality of his examples" (164), which Rorty thinks evidence a problematic conception of belief. The second set deals with Harman's effort "to help us pragmatists out by clarifying the distinction between 'practical' and 'epistemic' reasons for belief" (165). Rorty agrees that there is an intuitive distinction between the two, but thinks we should abandon the intuition. Moreover, he holds that Harman's proposal for how to distinguish between epistemic and practical reasons breaks down in "messy situations-those in which relevance is up for grabs" (167). Clearly, then, the contributors differ substantially over their understanding of realism and antirealism, over the legitimacy of the distinction, and over the distinction's implications for epistemology. Although they have not resolved the debate-how often does that happen when fundamental issues are at stake in philosophy?-they have surely deepened the discussion.

Notes 1. John Stuart Mill, An Examination of Sir William Hamilton's Philosophy and of the Principal Philosophical Questions Discussed in His Writings (Boston: William V. Spencer, 1866), vol. 1, chap. 11. 2. See, e.g., David Brink, Moral Realism and the Foundations of Ethics (New York: Cambridge University Press, 1989), chap. 2. Contrast A. C. Grayling, who argues in "Epistemology and Realism" (Proceedings of the Aristotelian Society 92 (1992): 47-65, esp. 55) that moral theory should not be considered either realist or antirealist.

Introduction

13

3. See, Peter Railton, "Moral Realism: Prospects and Problems," in Moral Knowledge, ed. Walter Sinnott-Armstrong and Mark Timmons (New York: Oxford University Press, 1996),49-81, esp. 51 ff. 4. Cf. his initial, more formal statement of external realism: "The world (or alternatively, reality or the universe) exists independently of our representations of it" (16). 5. Chisholm remarks, "It was Michael Dummett, I believe, who introduced the terminology of 'realism' and 'antirealism' into contemporary philosophical discussion" (95).

2 Does the Real World Exist?

John R. Searle

Part I: Attacks on Realism O far I have tried to analyze the nature and structure of those facts that, in a sense I attempted to explain, are dependent on human agreement or acceptance. The whole analysis presupposes a distinction between facts dependent on us and those that exist independently of us, a distinction I originally characterized as one between social and institutional facts on the one hand and brute facts on the other. It is now time to defend the contrast on which the analysis rests, to defend the idea that there is a reality that is totally independent of us. Furthermore, throughout the book I have been presupposing that in general our statements when true correspond to facts, and it is now also time to defend this presupposition. These defenses are made more pressing by the current philosophical scene in which it is common both to deny the existence of a reality independent of human representations and to deny that true statements correspond to facts. Part 1 is about realism; part 2 is about the correspondence theory of truth. A thorough discussion of these problems would require at least another book, but for present purposes I need at least a brief exposition of certain presuppositions behind our contemporary commonsense scientific world view because ... that worldview depends on these presuppositions ....

S

Some Presuppositions of Our Contemporary Worldview In order to understand what is at stake, we need to get some of the presuppositions of our world view out into the open, where we can have a From THE CONSTRUCTION OF SOCIAL REALITY by John R. Searle. Copyright © 1995 by John R. Searle. Reprinted with permission of the author and publisher, The Free Press, a Division of Simon and Schuster.

15

16

Realism/Antirealism and Epistemology

look at them. A formal feature of our world view is the distinction between objectivity and subjectivity. . . . In addition to the usual problems of vagueness and marginal cases-problems that are not serious-this distinction is systematically ambiguous between an epistemic and an ontological sense. In light of the distinction between epistemic objectivity / subjectivity and ontological objectivity/subjectivity, we can identify the following structural features of our world view. 1. The world (or alternatively, reality or the universe) exists indepen-

dently of our representations of it. This view I will call "external realism. I will refine its formulation later. 2. Human beings have a variety of interconnected ways of having access to and representing features of the world to themselves. These include perception, thought, language, beliefs, and desires as well as pictures, maps, diagrams, etc. Just to have a general term I will call these collectively "representations." A feature of representations so defined is that they all have intentionality, both intrinsic intentionality, as in beliefs and perceptions, and derived intentionality, as in maps and sentences. 3. Some of those representations, such as beliefs and statements, purport to be about and to represent how things are in reality. To the extent that they succeed or fail, they are said to be true or false, respectively. They are true if and only if they correspond to the facts in reality. This is (a version of) the correspondence theory of truth. 4. Systems of representation, such as vocabularies and conceptual schemes generally, are human creations, and to that extent arbitrary. It is possible to have any number of different systems of representations for representing the same reality. This thesis is called "conceptual relativity." Again, I will refine its formulation later. 5. Actual human efforts to get true representations of reality are influenced by all sorts of factors-cultural, economic, psychological, and so on. Complete epistemic objectivity is difficult, sometimes impossible, because actual investigations are always from a point of view, motivated by all sorts of personal factors, and within a certain cultural and historical context. 6. Having knowledge consists in having true representations for which we can give certain sorts of justification or evidence. Knowl-

Does the Real World Exist?

17

edge is thus by definition objective in the epistemic sense, because the criteria for knowledge are not arbitrary, and they are impersonal. Knowledge can be naturally classified by subject matter, but there is no special subject matter called "science" or "scientific knowledge." There is just knowledge, and "science" is a name we apply to areas where knowledge has become systematic, as in physics or chemistry. In light of the distinction between the epistemic and ontological senses of the objective/subjective distinction, we can say: Proposition 1 (external realism) is very close to the view that there is an ontologically objective reality. The two claims are not exactly equivalent, because the claim that there is a reality that is independent of representations (external realism) is not exactly equivalent to the claim that there is a reality completely independent of minds (ontological objectivity). The reason for this distinction is that some mental states, such as pains, are ontologically subjective, but they are not representations. They are representation independent but not mind independent. Ontological objectivity implies external realism, because mind independence implies representation independence. But not conversely. Pains, for example, can be representation independent without being mind independent. Proposition 2 implies that ontological subjectivity gives us epistemic access to all the reality to which we have access, whether ontologically subjective or objective, whether epistemically subjective or objective. Proposition 5 says epistemic objectivity is often hard to obtain; and proposition 6 says that if you have genuine knowledge, you have epistemic objectivity by

definition. I hope the reader finds these six propositions so obvious as to wonder why I am boring him or her with such platitudes, but I have to report that a great deal of confusion surrounds them. Propositions 1 and 3, realism and the correspondence theory, respectively, are often confused with each other; worse yet, they are both often supposed to have been refuted. Several philosophers think that proposition 4, conceptual relativity, creates a problem for realism; some think that it refutes it. Many philosophers think that proposition 3, the correspondence theory, has been independently refuted. Several literary theorists think that proposition 5 creates a problem for the very possibility of objective knowledge as stated in proposition 6, and perhaps even refutes realism as articulated by proposition l. So I fear there is nothing to do but slow down and go over at least some of these matters in low gear. Let us begin by asking,

18

Realism/Antirealism and Epistemology

What Is Realism? As a preliminary formulation, I have defined realism as the view that the world exists independently of our representations of it. This has the consequence that if we had never existed, if there had never been any representations-any statements, beliefs, perceptions, thoughts, etc.-most of the world would have remained unaffected. Except for the little corner of the world that is constituted or affected by our representations, the world would still have existed and would have been exactly the same as it is now. It has the further consequence that when we all die, and all our representations die with us, most features of the world will remain totally unaffected; they will go on exactly as before. For example, let us assume that there is a mountain in the Himalayas that I represent to myself and others as 'Mount Everest'. Mount Everest exists independently of how or whether I or anyone else ever represented it or anything else. Furthermore, there are many features of Mount Everest, for example, the sort of features that I represent if I make a statement such as "Mt. Everest has snow and ice near its summit," which would have remained totally unaffected if no one had ever represented them in any fashion and will not be affected by the demise of these or any other representations. One might put this point by saying that there are many language-independent features, facts, states of affairs, etc.; but I have put the point more generally in terms of "representations," because I want to note that the world exists independently not only of language, but also of thought, perception, belief, etc. The point is that, in large part, reality does not depend on intentionality in any form. In the history of philosophy the word 'realism' has been used with a wide variety of meanings. In the medieval sense, realism is the doctrine that universals have a real existence. Nowadays one hears talk of "modal realism," "ethical realism," "intentional realism," "mathematical realism," and so on. For the purposes of this discussion I am stipulating that 'external realism' and 'realism' ('ER' for short) name the view sketched in the previous paragraph. I use the metaphor of "external" to mark the fact that the view in question holds that reality exists outside of, or external to, our system of representation. Before examining arguments for and against realism we need to distinguish it from other views with which it is often identified. The first confusion is to suppose that realism is identical with or at least implies the correspondence theory of truth. But realism is not a theory of truth and it does not imply any theory of truth. Strictly speaking, realism is consistent with any theory of truth because it is a theory of ontology and not of the

Does the Real World Exist?

19

meaning of 'true.' It is not a semantic theory at all. It is thus possible to hold ER and deny the correspondence theory. 1 On a normal interpretation, the correspondence theory implies realism since it implies that there is a reality to which statements correspond if they are true; but realism does not by itself imply the correspondence theory, since it does not imply that 'truth' is the name of a relation of correspondence between statements and reality. Another misconception is to suppose that there is something epistemic about realism. Thus, for example, Hilary Putnam writes, "the whole content of Realism lies in the claim that it makes sense to think of a God's Eye View (or better a view from nowhere)."2 But that is not the content of realism as normally construed. On the contrary, the whole idea of a "view" is already epistemic and ER is not epistemic. It would be consistent with realism to suppose that any kind of "view" of reality is quite impossible. Indeed, on one interpretation, Kant's doctrine of things in themselves is a conception of a reality that is inaccessible to any "view." I realize that since the seventeenth century the most common arguments against realism have been epistemic-" all we can ever really know are our own sense data," that sort of thing-but the thesis under attack, realism, is not as it stands an epistemic thesis at all. I will have more to say later about the epistemic arguments against realism. A third mistake, also common, is to suppose that realism is committed to the theory that there is one best vocabulary for describing reality, that reality itself must determine how it should be described. But once again, ER as defined above has no such implication. The view that the world exists independently of our representations of it does not imply that there is a privileged vocabulary for describing it. It is consistent with ER to claim the thesis of conceptual relativity (proposition 4), that different and even incommensurable vocabularies can be constructed for describing different aspects of reality for our various different purposes. To summarize these points: Realism, as I am using the term, is not a theory of truth, it is not a theory of knowledge, and it is not a theory of language. If one insists on a pigeonhole, one could say that realism is an ontological theory: It says that there exists a reality totally independent of our representations. In the philosophical tradition there is a pervasive further ambiguity in the notion of realism that I need to expose and remove. Typically philosophers who discuss these issues treat them as if they concerned how the world is in fact. They think the issues between, say, realism and idealism are about the existence of matter or about objects in space and time. This is a very deep mistake. Properly understood, realism is not a thesis about

20

Realism/Antirealism and Epistemology

how the world is in fact. We could be totally mistaken about how the world is in every detail and realism could still be true. Realism is the view

that there is a way that things are that is logically independent of all human representations. Realism does not say how things are but only that there is a way that they are. And 'things' in the previous two sentences does not mean material objects or even objects. It is, like the 'it' in 'It is raining,' not a referring expression. It will seem presumptuous of me to claim that the issues do not concern specific claims about matter, and about material objects in space and time, if that is in fact what the disputants thought they were arguing about. But I hope to make it clear that the issues could not be about such specific claims. Realism could not be a theory asserting the existence of Mt. Everest, for example; because if it should turn out that Mt. Everest never existed, realism remains untouched. And what goes for Mt. Everest goes for material objects in general. But what if it should turn out that material objects do not exist or even that space and time do not exist? Well, in a sense it already has turned out that way, because we now think of material objects as collections of "particles" that are not themselves material objects but are best thought of as points of mass/energy; and absolute space and time have given way to sets of relations to coordinate systems. Not only is none of this inconsistent with realism; rather, as I will argue later, it all presupposes realism. It presupposes that there is a way things are that is independent of how we represent how things are. But let us continue with some science fiction thought experiments. Suppose it should turn out that physical reality is causally dependent on consciousness in such a way that with the last death of the last conscious agent all of physical reality blows up in a kind of Negative Big Bang. Would that still be consistent with external realism? It would, because the postulated dependence of matter on consciousness is a causal dependence like any other. When realism claims that reality exists independently of consciousness and of other forms of representation, no causal claim is made or implied. Rather, the claim is that reality is not logically constituted by representations, that there is no logical dependence. "But suppose it should turn out that the only things that exist or ever did exist are states of disembodied consciousness. Surely that would be inconsistent with realism and a vindication of idealism or at least some other version of antirealism." No, not necessarily. Realism does not say that the world had to turn out one way rather than another, but only that there is a way that it did turn out that is independent of our representations of it. Representations are one thing, the reality represented another, and this point is true even

Does the Real World Exist?

21

if it should turn out that the only actual reality is mental states. One way to see the difference between realism and antirealism is this: on the realist view if it turned out that only conscious states exist, then ships and shoes and sealing wax do not exist. But the claim that ships and shoes and sealing wax do not exist is a claim about external reality like any other. It presupposes realism as much as the claim that they do exist. On the antirealist view, such things, if they exist, are necessarily constituted by our representations, and they could not have existed independently of representations. For example, according to Berkeley, ships and shoes and sealing wax must be collections of states of consciousness. For the antirealist it is impossible that there should be a mind-independent reality. For the realist, even if there were no material objects in fact, there would still be a representation-independent reality, for the nonexistence of material objects would just be one feature of that representation-independent reality. The world could have been different, consistent with realism, but in fact it turned out that it contains material phenomena in space and time. (Alternative formulation: For the realist, it not only could have turned out that there are objects other than representations, but in fact it did turn out that way. For the antirealist it could not have turned out that there are representation-independent objects.) Strange as it may seem, realism has recently come under attack both in philosophy and in other disciplines. Thinkers as diverse as Michael Dummett, Nelson Goodman, Thomas Kuhn, Paul Feyerabend, Hilary Putnam, Richard Rorty, Jacques Derrida, Humberto Maturana, Francesco Varela, and Terry Winograd are often interpreted (not always correctly, I believe) as challenging our naIve assumption that there exists a reality totally independent of our representations of it. Some scientists have even claimed that modern physics is inconsistent with realism; thus J. R. Wheeler writes: The universe does not exist "out there" independent of us. We are inescapably involved in bringing about that which appears to be happening. We are not only observers, we are participators ... in making [the] past as well as the present and the future. 3

There are several disquieting things about all these attacks on realism. The first is that the arguments against our commonsense idea that there exists an independent reality are often vague and obscure. Sometimes no clearly stated arguments are even presented. Second, the alternative views, the views that are supposed to be presented in opposition to realism, are often

22

Realism/Antirealism and Epistemology

equally obscure and unclearly stated. Even among analytic philosophers many recent discussions of realism are symptomatic of the general looseness that has set in over the past couple of decades. What exactly are the propositions being asserted? What exactly are those denied? And what exactly are the arguments for both assertion and denial? You will look in vain for answers to these questions in most discussions of these matters. I think, furthermore, that this general carelessness is not accidental. It is somehow satisfying to our will to power to think that "we" make the world, that reality itself is but a social construct, alterable at will and subject to future changes as "we" see fit. Equally, it seems offensive that there should be an independent reality of brute facts-blind, uncomprehending, indifferent, and utterly unaffected by our concerns. And all of this is part of the general intellectual atmosphere that makes antirealist versions of "poststructuralism" such as deconstruction seem intellectually acceptable, even exciting. But once you state the claims and arguments of the antirealists out in the open, naked and undisguised, they tend to look fairly ridiculous. Hence the obscurity and even obscurantism of many (not all) of these discussions. So I have a problem. I said I would defend realism against the attacks made on it, but frankly I have trouble finding any powerful attacks that seem worth answering. Maturana rejects the idea of "an objective reality" in favor of the idea that nervous systems, as autopoietic systems, construct their own reality.4 The argument appears to be that since we have no conception of, and no access to, reality except through the social construction of realities in the "consensual domains," constructed by autopoietic systems, there is no reality existing independently of biological systems. Against this view I want to say, From the fact that our knowledge/conception/picture of reality is constructed by human brains in human interactions, it does not follow that the reality of which we have the knowledge/conception/picture is constructed by human brain in human interactions: It is just a non sequitur, a genetic fallacy, to infer from the collective neurophysiological causal explanation of our knowledge of the external world to the nonexistence of the external world. Winograd points out that the same sentence, for example, There is water in the refrigerator', can be used to make a false statement relative to one set of background interests, a true statement relative to another.s From this he concludes that reality does not exist independently of our representations. Once again, as with Maturana, the conclusion does not * There is, furthermore, a problem about the human brains and the human interactions themselves. Are they also supposed to be constructed by human interactions?

Does the Real World Exist?

23

follow. The interest-relativity of our representations of reality does not show that the reality represented is itself interest-relative. Like Maturana, Winograd tries to derive conclusions about reality from features of our representations of reality. Several "postmodernist" literary theorists have argued that because all knowledge is socially constructed and subject to all the arbitrariness and will to power of any social construction, that therefore realism is somehow threatened. As George Levine writes, "Antirealism, even literary antirealism, depends on a sense of the impossibility of unmediated knowledge."6 Derrida, as far as I can tell, does not have an argument. He simply declares that there is nothing outside of texts (Il n'y a pas de "hors texte"). And in any case, in a subsequent polemical response to some objections of mine, he apparently takes it all back: he says that all he meant by the apparently spectacular declaration that there is nothing outside of texts, is the banality that everything exists in some context or other!7 What is one to do then, in the face of an array of weak or even nonexistent arguments for a conclusion that seems preposterous? The strategy I will follow is to take what I think are the most powerful arguments against external realism and answer them. Suppose that I were convinced by antirealism; what in particular might have convinced me? Or if that seems too far-fetched, suppose the fate of humanity rested on my convincing someone else of antirealism, what arguments would I use? I will consider three arguments: the argument from conceptual relativty, the verificationist argument, and what I will call the Ding an sich argument.

The Argument from Conceptual Relativity against Realism The conceptual relativity argument is that proposition 4 above, conceptual relativity, refutes proposition 1, external realism. The idea of conceptual relativity is an old and, I believe, a correct one. Any system of classification or individuation of objects, any set of categories for describing the world, indeed, any system of representation at all, is conventional, and to that extent arbitrary. The world divides up the way we divide it, and if we are ever inclined to think that our present way of dividing it is the right one, or is somehow inevitable, we can always imagine alternative systems of classification. To illustrate this for yourself, take a piece of chalk and draw a line across a portion of the book in front of you and then onto the table and around in a circle and back to the book to connect. Now give a name to this new sort of object made of portions of surfaces of books plus tables delimited by a chalk line. Call this object a "klurg." We do not have a use for this concept in our language. But it would be easy to imagine a culture where klurgs are of great religious

24

Realism/Antirealism and Epistemology

significance, where they can be delineated only by sacred virgins working under water and their obliteration merits the death penalty. But if "klurg" is a new concept with previously unheard-of truth conditions, there is no limit to how many new concepts we can form. Because any true description of the world will always be made within some vocabulary, some system of concepts, conceptual relativity has the consequence that any true description is always made relative to some system of concepts that we have more or less arbitrarily selected for describing the world. So characterized, conceptual relativism seems completely true, indeed platitudinous. However, several philosophers have supposed that it is inconsistent with external realism, and consequently, that if we accept conceptual relativism, we are forced to deny realism. But if this claim were really true, we ought to be able to state the two theses precisely enough for the inconsistency to be quite obvious. Let external realism be the view that: ERl:

Reality exists independently of our representations of it.

Let the relevant thesis of conceptual relativism be the view that: CRl: All representations of reality are made relative to some more or less arbitrarily selected set of concepts. So stated, these two views do not even have the appearance of inconsistency. The first just says that there is something out there to be described. The second says that we have to select a set of concepts and a vocabulary to describe it. So why would anyone suppose that the second entails the negation of the first? The answer is that if we accept conceptual relativism, and try to conjoin it with realism, we appear to get inconsistencies. Consider the following example from Putnam. 8 Imagine that there is some part of the world that is like this:

Figure 2.1

Does the Real World Exist?

25

How many objects are there in this miniworld? Well, according to Carnap's system of arithmetic (and according to common sense), there are three; but according to Lesniewski and other Polish logicians, there are seven objects in this world, counted as follows: 1 =A 2=B 3=C 4=A+B 5 =A+C 6 =B+C 7 =A+B+C So how many objects are there really in the imagined world? Are there really three or really seven? There is no absolute answer to these questions. The only answers we can give are relative to the arbitrary choice of conceptual schemes. The same sentence, for example, 'There are exactly three objects in the world', will be true in one scheme, false in the other. The heart of the argument is that external realism leads to inconsistencies because it allows for inconsistent descriptions of the supposedly independently existing reality. The form in which the argument occurs in Goodman is that we characteristically make reality or, as Goodman would prefer to say, we "make worlds" by drawing certain boundaries rather than others. For example, Goodman says: Now as we thus make constellations by picking out and putting together certain stars rather than others, so we make stars by drawing certain boundaries rather than others. Nothing dictates whether the sky shall be marked off into constellations or other objects, we have to make what we find, be it the Great Dipper, Sirius, food, fuel, or a stereo system. 9

Goodman rejects realism and evades the inconsistencies by relativizing the facts described to a "world" that we make. Putnam says that instead of thinking that there is a mind-independent reality, we should say rather that "the mind and the world jointly make up the mind and the world."lD But are these supposed contradictions really a problem? About the miniworld example, a realist who was a convinced conceptual relativist would say that there really are three objects, as the criterion for counting objects has been set in the first system of classification, really seven as the criterion has been set in the second. And this answer removes the apparent contradiction, not by modifying or abandoning external realism but

26

Realism/Antirealism and Epistemology

by simply pointing out that the criterion for counting objects has been set in two different ways. Thus the same sentence, for example, 'There are exactly three objects in the world', can now be used to make two quite different and independent statements, one of which is true, one false. But the real world does not care how we describe it and it remains the same under the various different descriptions we give of it. Some of the examples of conceptual relativism given in the literature are more arcane and complicated than the ones I have given, but the principle they employ is the same, and I cannot see that anything is gained by the complexity. They all are designed to show that different conceptual systems will generate different and apparently inconsistent descriptions of the same "reality." As far as I can see there is nothing in any of them that is inconsistent with external realism. The appearance of inconsistency is an illusion and, on a natural interpretation of these views, there is no inconsistency whatever in accepting the most naive version of realism, and accepting any version at all of conceptual relativism.* Think of the relation of realism and conceptual relativism like this: Take a corner of the world, say, the Himalayas, and think of it as it was prior to the existence of any human beings. Now imagine that humans come along and represent the facts in various different ways. They have different vocabularies, different systems for making maps, different ways of counting one mountain, two mountains, the same mountain, etc. Next, imagine that eventually the humans all cease to exist. Now what happens to the existence of the Himalayas and all the facts about the Himalayas in the course of these vicissitudes? Absolutely nothing. Different descrip* What has gone wrong? In Putnam's case, a close look at the texts suggests that he is lumping at least two logically independent theses under his label, "Metaphysical Realism."

First:

Reality exists independently of our representations of it.

Second: There is one and only one correct conceptual scheme for describing reality. The first is what I have been calling external realism; let's call the second the theory of the "Privileged Conceptual Scheme" (PCS). Putnam sees correctly that CR refutes PCS. And since you can always refute a conjunction by refuting one conjunct, if metaphysical realism is the conjunction of ER and PCS, then metaphysical realism has been refuted. But you do not refute both conjuncts by refuting just one; so the falsity of PCS leaves ER standing untouched. Putnam's writings give the impression that he thinks that by refuting PCS he has refuted ER. Perhaps he does not think that the "refutation" touches ER, in which case a bald assertion in favor of ER would have been helpful to his readers. But he makes no such assertion; on the contrary, he endorses a view he calls "internal realism." I do not think there is a coherent position of "internal realism" that is halfway between external realism, as I have defined it, and out-and-out antirealism, which Putnam also claims to reject.

Does the Real World Exist?

27

tions of facts, objects, etc., came and went, but the facts, objects, etc., remained unaffected. (Does anyone really doubt this?) The fact that alternative conceptual schemes allow for different descriptions of the same reality and that there are no descriptions of reality outside all conceptual schemes has no bearing whatever on the truth of realism. But what about the possibility, raised by Goodman, of inconsistent descriptions made relative to different conceptual schemes? There is no substitute for going carefully through examples, so let us consider a case of how external realism deals with alternative vocabularies. Let us suppose that I am a complete, naIve external realist where weight, that is, the gravitational attraction of masses at the earth's surface, is concerned. I think that I really weigh 160 no matter what anybody thinks. But wait! I weigh 160 in pounds but only 73 in kilograms. So how much do I weigh really: 160 or 73? I hope it is obvious that both answers are correct, though each is incomplete. The appearance of inconsistency is only an appearance, because the claim that I weigh 160 in pounds is consistent with the claim that I weigh 73 in kilograms. External realism allows for an infinite number of true descriptions of the same reality made relative to different conceptual schemes. "What is my aim in philosophy? To teach you to turn disguised nonsense into obvious nonsense."ll It is disguised nonsense to say that conceptual relativism implies antirealism, obvious nonsense to say that I cannot, at the same time, weigh both 160 (in pounds) and 73 (in kilograms). Furthermore, if conceptual relativity is to be used as an argument against realism, it seems to presuppose realism, because it presupposes a language-independent reality that can be carved up or divided up in different ways, by different vocabularies. Think of the example of alternative arithmetics: Putnam points out that one way to describe the miniworld is to say there are three objects, another way is to say there are seven objects. But notice that this very claim presupposes something there to be described prior to the application of the description; otherwise there is no way we could even understand the example. And when Goodman writes, "We make stars by drawing certain boundaries rather than others," there is no way to understand that claim except by presupposing something there on which we can draw boundaries. Unless there is already a territory on which we can draw boundaries there is no possibility of drawing any boundaries. If we try to take these arguments as counting against ER, we commit a massive use-mention fallacy: From the fact that a description can only be made relative to a set of linguistic categories, it does not follow that the

28

Realism/Antirealism and Epistemology

facts/objects/states of affairs/etc., described can only exist relative to a set of categories. Conceptual relativism, properly understood, is an account of how we fix the application of our terms: What counts as a correct application of the term 'cat' or 'kilogram' or 'canyon' (or 'klurg') is up to us to decide and is to that extent arbitrary. But once we have fixed the meaning of

such terms in our vocabulary by arbitrary definitions, it is no longer a matter of any kind of relativism or arbitrariness whether representation-independent features of the world satisfy those definitions, because the features of the world that satisfy or fail to satisfy the definitions exist independently of those or any other definitions. We arbitrarily define the word 'cat' in such and such a way; and only relative to such and such definitions can we say, "That's a cat." But once we have made the definitions and once we have applied the concepts relative to the system of definitions, whether or not something satisfies our definition is no longer arbitrary or relative. That we use the word 'cat' the way we do is up to us; that there is an object that exists independently of that use, and satisfies that use, is a plain matter of (absolute, intrinsic, mind-independent) fact. Contrary to Goodman, we do not make "worlds"; we make descriptions that the actual world may fit or fail to fit. But all this implies that there is a reality that exists independently of our system of concepts. Without such a reality, there is nothing to apply the concept to. In order that we should get a version of conceptual relativism that is inconsistent with external realism, we would have to have a version that implies that the same statement (not the same sentence, but the same statement) could be true of the world in one conceptual system but false of the world in another conceptual system. I have not seen any example of this that is remotely plausible. The standard examples are something like this: Assume that we have different models for representing some domain of reality, say, Aristotelian physics vs. Newtonian physics or the Mercator projection of the earth's surface vs. a standard globe representation of the earth's surface. Now, on the Mercator projection, Greenland occupies a bigger area than Brazil, but on the globe, Greenland occupies a smaller area than Brazil. So don't we have here two models, both true of the same reality, but, in fact, inconsistent with each other? The answer is no. The Mercator projection is just inaccurate about the relative size of Brazil and Greenland. It is a well-known fact that certain models, for example, Aristotelian physics and the Mercator projection, are mistaken about or distort certain features of the world. All true statements about the world can consistently be affirmed together. Indeed, if they could not consistently be affirmed together, they could not all be true. Of course, we are always confronted with the prob-

Does the Real World Exist?

29

lems of vagueness, indeterminacy, family resemblance, open texture, contextual dependency, the incommensurability of theories, ambiguity, the idealization involved in theory construction, alternative interpretations, the underdetermination of theory by evidence, and all the rest of it. But these are features of our systems of representation, not of the representation-independent reality that some of these systems can be used, more [or] less adequately, to represent. Often the same sentence can be used to assert a truth in one conceptual scheme and a falsehood in another conceptual scheme. But this, as we have seen over and over, does not show a genuine inconsistency.

The Verificationist Argument Twentieth-century philosophy has been obsessed with language and meaning, and that is why it is perhaps inevitable that somebody would come up with the idea that nothing at all exists apart from language and meaning. Earlier centuries were obsessed with experience and knowledge, and correspondingly philosophers came up with the idea that there is no reality independent of experience and knowledge. In the history of Western philosophy since the seventeenth century, the most common argument against realism has been derived from epistemic considerations. I believe the basic philosophical motivation behind verificationists' arguments against realism is to try to eliminate the possibility of skepticism by removing the gulf between appearance and reality that makes skepticism possible in the first place. If reality consists in nothing but our experiences, if our experiences are somehow constitutive of reality, then the form of skepticism that says we can never get out of our experiences to the reality beyond is answered. This is a persistent urge in philosophy. Kant's transcendental idealism is a more sophisticated variant of it than one finds in Berkeley, and the same urge survives in the late twentieth century in the various efforts to analyze meaning in "public" terms, or even "behavioristically," so that there will be no private residue about which one might doubt that one had really understood what other people meant by the use of an expression. Even if I am right about this diagnosis, however, it doesn't answer the actual arguments, so I will now present what I take to be the most powerful verificationists' argument against realism. Here is how it goes: Ask yourself what you really know, I mean really know. Well, you might say you really know that you are sitting on a chair, that there is a desk in front of you, that you are looking at a computer screen.

30

Realism/Antirealism and Epistemalogy

But if you think about it, you will see that what you really know is that you are having certain experiences; so when you make these claims about a chair or a desk or a computer screen, either you are talking about your experiences or you are talking about something you don't really know. Furthermore, if you tried to talk about something other than experiences you would be talking about something that you couldn't know. If you ask how you know about the world, the answer has to be: from your experiences. But then you are faced with a dilemma. Either your claims to know simply report the content of your experiences or they go beyond those contents. If the former, then there is nothing known except the content of your experiences. If the latter, then you are making claims that you cannot validate, because all validation rests on experience, and you are ex hypothesi making claims that go beyond what you experience. For example, I claim to know there is a desk in front of me now. What does such a claim mean? Well, all I have direct knowledge of are these tactile and visual experiences, and all I-or anyone elsecould ever have direct knowledge of are more such experiences. So what does my original claim amount to? Either it amounts to the claim that there are actual and possible experiences ("sense data" in the twentieth-century jargon, "ideas" and "impressions" in that of the seventeenth and eighteenth centuries), or if something more is claimed, then it must be a claim about something totally unknowable and inaccessible to any investigation. Such a claim is empirically empty. The conclusion is obvious: experience is constitutive of reality. This argument occurs in several authors and the conclusion has been stated in a variety of vocabularies: Objects are collections of ideas (Berkeley). Objects are permanent possibilities of sensation (Mill). Empirical statements can be translated without residue into statements about sense data (twentieth-century phenomenalism). Berkeley summarizes this argument neatly when he says, "If matter did exist we could never know it, if it does not exist everything remains the same." There are, it seems to me, two strands to the argument. The first is, all we can ever perceive are our own experiences. Therefore, if there is supposed to be a reality beyond our experiences, it is unknowable, and ultimately unintelligible. The second is an extension of the first. It says the only basis we have for claims about the real world are our experiences. Therefore, if claims about the real world go beyond the content of our experiences, then ex hypothesi we are postulating something for which

Does the Real World Exist?

31

we can have no epistemic basis. I believe both strands are mistaken. Let us consider each in turn. It is indeed the case that whenever one consciously perceives anything, one has certain experiences. For example, for every visual perception there is a corresponding visual experience. In the formal mode of speech, to report "I see the table" implies "I am having a certain sort of visual experience." But from the fact that the visual experience is an essential component of the visual perception, it does not follow that the visual experience is that which is perceived. It does not follow, in short, that one does not have direct access to the real world whenever one employs one's perceptual apparatus to perceive it. Thus, for example, right now I see my desk in front of me. In such a case, I simply perceive the desk. In so doing, I have a perceptual experience, but the perceptual experience is neither the object of perception nor is it the evidence on the basis of which I conclude that there is a desk there. I do not "conclude" on the basis of "evidence" that there is a desk here; rather, I simply see it. So the first strand in the argument, namely, that all we have access to in perceptual experiences is the content of the perception itself, is mistaken. 12 I believe that the second is also mistaken. Let us grant, for the sake of argument, that the epistemic basis for my present claim that there is a desk in front of me is the existence of my present sense experiences and grant also that the claim that there is a table there-if it is understood in the commonsense, naIve realist way-states more than a mere summary of statements about my experiences. What follows? Does it follow that the claim that there is a table there states something unknowable, something that goes beyond any possible evidence or other epistemic basis? It does not follow. From the fact that the epistemic basis for my knowledge is my present experiences, it does not follow that all I can know are my experiences. On the contrary, the way we described the example was precisely as a case where my experiences give me access to something that is not itself an experience. It is a familiar point in philosophy that in general empirical claims go beyond the epistemic bases on which they are made. There would not be much point, for example, in making scientific hypotheses if they were just summaries of the available evidence. However, at this point, the defender of the antirealist position will want to say the following: In presenting these answers to the antirealist argument, you have tacitly presupposed that you are really perceiving mind-independent objects in the real world, but that is precisely what you are not

32

Realism/Antirealism and Epistemology

entitled to assume. The whole point of the argument is that you could be having exactly these experiences and there not be any desk there. But if that is the case, then it doesn't matter whether we think of the experiences as providing the "evidence" for your "conclusion" that there is a desk there. The point is that the only basis that you have for your confidence that there is a desk there is the presence of these sense-data, and if the desk is supposed to be something over and above the sense-data, they would not be sufficient to justify that confidence because you could be having exactly these experiences and be totally mistaken. The postulation of an external reality is essentially the postulation of something unknowable and ultimately unintelligible. What is the answer to this? In this discussion I am not trying to answer general skepticism. That is a set of questions that goes beyond the scope of this book. So for the sake of this argument let us just grant that I might be having exactly these experiential contents and be having a total hallucination. I might be subject to all the horrors of traditional epistemology: I might be a brain in a vat, I might be deceived by an evil demon, I might be dreaming, etc. But it does not follow that my claim that there is a desk in front of me is simply a summary of the experiences that prompt me to make the claim. That is, even if skepticism is right, and I am systematically mistaken, what I am mistaken about are the features of the real world. The possibility of being systematically mistaken about those features does not show that my claims about them are just summaries of statements about my sense experiences. These are ancient battlegrounds, and the landscape is much devastated by epistemic wars, but I believe the basic logical geography of the philosophical terrain is simple and readily discernible: The verificationist argument for antirealism is as follows: 1. All we have access to in perception are the contents of our own experiences. 2. The only epistemic basis we can have for claims about the external world are our perceptual experiences. Therefore, 3. The only reality we can meaningfully speak of is the reality of perceptual experiences.

Does the Real World Exist?

33

I have argued that statement 1 is false. We typically perceive objects and states of affairs in the world. Furthermore, I have argued that statement 2, though true, does not imply statement 3. It is a mistake to suppose that empirical claims are only meaningful to the extent that they are understood as summaries of their evidentiary or epistemic bases. Finally, I have claimed that the possibility of radical error, the possibility raised by skepticism, is irrelevant. Even if we are systematically mistaken, it la traditional skepticism, statement 3 does not follow. On the contrary, what we are mistaken about, if skepticism is right, is the real world. I realize that there are other versions of the verificationists' argument for antirealism, but I believe this version has been the one that has been the most pervasive in the empiricist tradition from the seventeenth century right through logical positivism. I also realize that in this entire tradition epistemology, in general, and the attempt to answer Cartesian skepticism in particular, were central to the entire philosophical enterprise. I regard these as mistakes. Epistemology has an important but certainly not a central place in the enterprise of philosophy. The deeper reason why epistemological considerations of the sort that I have been discussing could never provide a sound argument for antirealism is that in order even to state these considerations, we have to presuppose realism. I will return to this point in part 2. The Ding an Sich Argument

There is another argument against ER worth mentioning, the argument concerning things in themselves, the Ding an sich argument. It is hard to find an explicit version of this argument in contemporary philosophy, but it keeps coming up in the oral tradition. It is best to think of it as a combination of the argument from conceptual relativity and the argument from verification. Here is how it goes: When we deal with the world in perception, thought, inquiry, etc., we are always working from within some conceptual scheme. Even our so-called "experiences" are never directly of "reality" but are permeated by our concepts, and can refer ultimately only to other experiences. There is no Cod's-eye view from which we can survey the relations between our representations and the reality they are alleged to represent, to see whether they are really adequate to reality. There is no way we can see these relations from the side; rather, we are always inside our representations-our beliefs, experiences,

34

Realism/Antirealism and Epistemology

utterances, etc. Because we can't get outside the set of our representations to scrutinize reality directly, because there is no nonrepresentational standpoint from which we can survey the relations between representation and reality, and because there is not even the possibility of assessing the adequacy of our representations by measuring them against things in themselves, talk of a transcendent reality must be just so much nonsense. All the reality we can ever really get at, have access to, is the reality that is internal to our system of representations. Within the system there is a possibility of realism, internal realism, but the idea of a reality outside the system is as empty as Kant's notion of the Ding an sich, a thing in itself, beyond the grasp not only of our knowledge but of our language and thought. What external realism offers us is an unthinkable something, indescribable, inaccessible, unknowable, unspeakable, and ultimately nonsensical. The real problem with such a realism is not that it is false, but that it is ultimately unintelligible. What are we to make of this argument? Once again, if we try to state it as an explicit argument, with a set of premises and a conclusion, it is hard to see how the conclusion is supposed to follow.

Premise: Any cognitive state occurs as part of a set of cognitive states and within a cognitive system. From this premise it is supposed to follow that:

Conclusion 1: It is impossible to get outside of all cognitive states and systems to survey the relationships between them and the reality that they are used to cognize. And from this in turn it is supposed to follow that:

Conclusion 2: No cognition is ever of a reality that exists independently of cognition. It seems to me that, properly understood, conclusion 1 does indeed follow

from the premise. All representation occurs within a set of representations and within some representational system. Hence, any representation of the relation between the set of representational states and the representational system, on the one hand, and the reality represented, on the other, also occurs within some representational system. But so what? It simply does

Does the Real World Exist?

35

not follow from the fact that all cognition is within a cognitive system that no cognition is ever directly of a reality that exists independently of all cognition. Conclusion 2 just does not follow. Indeed, to suppose it does follow seems to be a mistake of the same form as the mistake committed by old-time idealism.

Diagnosis of the Problem I now want to offer a partial diagnosis of why it has become fashionable even among technically competent philosophers to attack realism, and to advance such feeble arguments against it. One of the oldest urges in Western philosophy is the urge to think that somehow or other truth and reality should coincide. That somehow or other, if there really were such things as truth and reality, as we normally think of them, then truth would have to provide an exact mirror of reality. The nature of reality itself would have to provide the exact structure of true statements. A classical statement of this position is in Wittgenstein's Tractatus,13 but I believe the idea is as old as Plato. When the philosopher despairs of achieving an exact isomorphism between the structure of reality and the structure of true representations, the temptation is to think that somehow or other our naIve notions of truth and reality have been discredited. But they have not been discredited. What has been discredited is a certain misconception of the relationship between truth and reality. There is a simple but deep reason why truth and reality cannot coincide in a way that many philosophers think that the naIve external realist is committed to their coincidence. The reason is this: All representation, and a fortiori all truthful representation, is always under certain aspects and not others. The aspectual character of all representations derives from such facts as that representation is always made from within a certain conceptual scheme and from a certain point of view. So, for example, if I describe the substance in front of me as water, the same piece of reality is represented as if I describe it as H 20. But, of course, I am representing the same stuff under a different aspect if I represent it as water than if I represent it as HzO. Strictly speaking, there is an indefinitely large number of different points of view, different aspects, and different conceptual systems under which anything can be represented. If that is right, and it surely is, then it will be impossible to get the coincidence between truth and reality after which so many traditional philosophers seem to hanker. Every representation has an aspectual shape. It represents its target under certain aspects and not others. In short, it is only from a point of view that we represent reality, but ontologically objective reality does not have a point of view.

36

Realism/Antirealism and Epistemology

Part 2: Could There Be a Proof of External Realism?

Realism as a Background Condition of Intelligibility I have said that certain standard arguments against realism are invalid. Are there any arguments to be given in its favor? There is something puzzling about demanding an argument to show that the world exists independently of our representations of it. I realize that Kant thought it a scandal that there was no such proof, and Moore thought he could give proof just by holding up his two hands. But, one feels, in the way that Kant posed his demand nothing could have satisfied it, and Moore's attempt to satisfy it somehow "misses the point." Yet, at the same time, one feels that one ought to satisfy Kant's demand, and that at some level Moore was surely right. He certainly did have two hands, and if he had two hands then the external world exists. Right? What is going on? We need to explain both our urge to prove external realism and our sense that any proof begs the question. The demand for a proof of external realism is a bit like the demands one used to hear in the 1960s for a proof of rationality-"What is your argument for rationality?"-in that the very posing of the challenge somehow presupposes what is challenged. Any attempt to provide an "argument" or "proof" already presupposes standards of rationality, because the applicability of those standards is constitutive of something's being an argument or proof. In a word, you can't prove rationality by argument because arguments already presuppose rationality. There are a number of such general frameworks where the demand to justify the framework from within the framework is always senseless and yet somehow seems incumbent upon us. Thus, although one can prove that a particular argument is valid or rational within the criteria of rationality and validity, one cannot prove within those criteria that rationality is rational or that validity is valid. Similarly, one can establish that a given sequence of words is a grammatical or ungrammatical English sentence, but one cannot establish that English as a language is grammatical or ungrammatical, because English sets the standard for grammaticality in English. The effort to establish external realism by some sort of "argument" would be analogous to one of these efforts. It would be as if one tried to establish that representation represents. One can show that this or that claim corresponds or fails to correspond to how things really are in the "external world," but one cannot in that way show that the claim that there is an external world corresponds to how things are in the external world, because any question of corresponding or failing to correspond to the external world already presup-

Does the Real World Exist?

37

poses the existence of an external world to which the claim corresponds or fails to correspond. External realism is thus not a thesis nor a hypothesis but the condition of having certain sorts of theses or hypotheses. You can see that something is wrong with this entire debate if you look at standard contemporary arguments in favor of external realism. A standard argument, perhaps the standard argument, for realism is that convergence in science provides a kind of empirical proof of realism. The idea is that because different scientific investigators working at different times and places come up with the same or similar results, the best explanation for their doing so is that there is an independently existing reality that causes them to converge on the same hypotheses and theories. The difficulty with this argument is that in our understanding of the possibility of there being such phenomena as either convergence or failure of convergence, we are already presupposing realism. In order for us even to raise the question whether scientific investigation does converge in the suggested fashion, we have to presuppose an independently existing reality of investigators engaging in investigations. These investigations either converge or fail to converge; that is, the entire discussion of convergence presupposes realism, because it presupposes that the statement 'Science converges', whether true or false, concerns a reality independent of that or any other statement. Another way to put this point: In areas where science fails to converge, for example, social psychology, our recognition of the failure provides exactly as much evidence for realism as our recognition of those areas in which it does converge, that is, it provides no evidence at all, since in recognizing something as either convergence or nonconvergence, we are already taking realism for granted. I realize that the convergence argument is often presented as an argument for the existence of unobservable entities postulated by scientific theories and not as a general argument for external realism. But then it is faced with a dilemma. If, on the one hand, the convergence argument is an argument to establish the existence of this or that type of unobservable entity, say, electrons, then the notion of convergence adds nothing whatever to the usual notions of evidence, verification, and truth. If the atomic theory that postulates electrons is confirmed in both my lab and yours, then that is further evidence that the theory is true, and if the theory entails that electrons exist, then we have good evidence that electrons exist. The notion of convergence adds nothing to this story. And the fact that we may have a number of such stories about different types of unobservable entities still gives us nothing more than a list of cases of scientific confirmation and disconfirmation. But if, on the other hand, the convergence argument is to be a genuine meta theory about the sociology of

38

Realism/Antirealism and Epistemology

scientific research, a theory to the effect that, as a matter of second-order empirical fact, scientists working at different times and places tend to produce convergent results, results that agree from one lab to another, and that this convergence is confirming evidence for realism, then it is subject to the objection I made earlier: In order that we can even consider the problem of convergence we have to presuppose realism. To explore this point further I want to ask, what is wrong with Moore's "proof"? Moore thought that by proving the existence of two or more things such as hands, sheets of paper, shoes, socks, etc., he would have proven the existence of "things outside of us" and ipso facto would have proven the existence of an "External World," because, as he says, "it will follow at once that there are some things to be met with in space."* On this view the relation between his premise and conclusion is a straight entailment relation: The proposition that I have two hands entails the proposition that the external world exists. The existence of the external world is a truth condition of the proposition that I have two hands in the same way that the existence of at least one hand is a truth condition of that proposition. If I have two hands then it follows immediately that "there are things to be met with in space." And he establishes the "premise" by demonstration. He simply makes a certain gesture and thereby "proves" the existence of his hands. But there is something fishy about this. Berkeley, for example, would have agreed that Moore had two hands but would have challenged the alleged entailment, so it looks as if Moore is "begging the question." Isn't the entailment precisely what is at issue? I suggest that there are at least two worrisome features of Moore's proof: The first is the assumption that ER is a truth condition like any * The crucial passage is this: "That is to say, if I can prove that there exist now both a sheet of paper and a human hand, I shall have proved that there are now 'things outside of us'; if I can prove that there exist now both a shoe and sock, I shall have proved that there are now 'things outside of us'; etc.; and similarly I shall have proved it, if I can prove that there exist now two sheets of paper, or two human hands, or two shoes, or two socks, etc. Obviously, then, there are thousands of different things such that, if, at any time, I can prove anyone of them, I shall have proved the existence of things outside of us. Cannot I prove any of these things? "It seems to me that, so far from its being true, as Kant declares to be his opinion, that there is only one possible proof of the existence of things outside of us, namely the one which he has given, I can now give a large number of different proofs, each of which is a perfectly rigorous proof; and that at many other times I have been in a position to give many others. I can prove now, for instance, that two human hands exist. How? By holding up my two hands, and saying, as I make a certain gesture with the right hand, 'here is one hand', and adding, as I make a certain gesture with the left, 'and here is another'. And if, by doing this, I have proved ipso facto the existence of external things, you will all see that I can also do it now in numbers of other ways: there is no need to multiply examples." G. E. Moore, "Proof of an External World,"in Philosophical Papers (London: George Allen & Unwin, 1959), 145-46.

Does the Real World Exist?

39

other, and the second is the related assumption that realism is a theory about external "objects" in "space." Against these views, the claims I want to make are: First, though there is no sharp dividing line between the two, we need to make a general distinction between truth conditions and conditions on intelligibility. There are conditions on the intelligibility of discourse, and indeed on the functioning of intentionality in general, that are not like paradigmatic cases of truth conditions. In the normal understanding of discourse we take these conditions for granted; and unless we took them for granted, we could not understand utterances the way we do or even have the intentional states with conditions of satisfaction that we have. In earlier writings I divided some of these conditions into a Network of beliefs and other intentional states, on the one hand, and a Background of abilities, capacities, etc., on the other. The claim I make here is that ER functions as a taken-for-granted part of the Background. Unless we take ER for granted, we cannot understand utterances the way we normally do. Furthermore we have to take ER for granted to engage in the sorts of discourse and thought that we have been engaging in. The presupposition of ER is thus a necessary presupposition for a large chunk of thought and language. We can't give it up as, for example, centuries ago we gave up our presupposition that the earth is flat. And the second response I want to make to Moore is that once we see that ER so construed is not an empirical thesis but rather a condition of intelligibility on having certain sorts of theses, then we can see that it has no special connection to the theory that there are "objects" in "space." As I said at the beginning of part 1, even if it should turn out that our notions of "objects" and "space" have to be radically revised, as in fact they have been revised by atomic theory and relativity theory, all the same, ER remains untouched. Carefully stated, external realism is the thesis that there is a way that things are that is independent of all representations of how things are. The thesis that there is a reality independent of our representations identifies not how things are in fact, but rather identifies a space of possibilities. Using a Wittgensteinian style of example, we can think of it this way. Suppose I say, "I have no money at all in my wallet." Now that utterance does not logically imply the existence of money. You cannot infer from -(3x)(money x & in my wallet x) (It is not the case that there is some x such

wallet.) to

that x is money and x is in my

40

Realism/Antirealism and Epistemology

(3x)(money x)

(There is some x such that x is money.) But all the same the original utterance only makes the kind of sense it does, we only understand it the way we do, against the presupposition of the existence of money. It has its sense against a space of possibilities of having money. In that sense, External Realism articulates a space of possibilities for a very large number of statements.

A "Transcendental" Argument for External Realism If these suggestions-that ER is a Background presupposition and not an empirical theory, and that it is purely formal without any specific content about, for example, objects in space-are right, the only argument we could give for ER would be a "transcendental" argument in one of Kant's many senses of that term: We assume that a certain condition holds, and then try to show the presuppositions of that condition. In order to do this, however, we have to make precise what the view is that we are arguing against. Antirealism is not a single doctrine but comes in different versions. For this discussion, the two most important are, first, the view that all reality consists in conscious states, and, second, the view that reality is socially constructed, that what we think of as "the real world" is just a bunch of things constructed by groups of people. To have labels, let us call the first view "phenomenalist idealism," and the second "social constructionism." There is a simple transcendental argument against phenomenalist idealism. I said that a transcendental argument is one that assumes a certain condition obtains and then tries to show the presuppositions of that condition. In this case, however, the "condition" has to do with our practices and the "presupposition" is what we, from our own first-person point of view, must presuppose when we engage in those practices. The condition is that we do in fact attempt to communicate with each other by making certain sorts of utterances in a public language and the presupposition is external realism. To spell this out a little bit more precisely: the assumption we are making is that there is a normal way of understanding utterances, and that when performing speech acts in a public language, speakers typically attempt to achieve normal understanding. The point we are attempting to show is that for a large class (to be specified further) a condition of intelligibility for the normal understanding of these utterances is that there is a way that things are that is independent of human representations. The consequence is that when we attempt to communicate to achieve

Does the Real World Exist?

41

normal understanding with these sorts of utterances we must presuppose external realism. Notice that we are not trying to prove the truth of external realism. I do not believe there could be a non-question-begging argument for ER. But we can show that when we engage in certain sorts of talk we presuppose external realism. To develop the argument I need to explain the notion of "normal understanding." For most speech acts there is a commonsense or normal understanding. Often this is given by disquotation; for example, the normal understanding of the utterance 'I have two hands' is that it asserts that the speaker has two hands. But wherever there is disquotation there must always be further ways of describing normal understanding. Thus in the normal understanding of 'I have two hands', for example, there must be a possible description of what a hand is. H you follow out the line of describing normal understanding, you soon reach conditions that are not truth conditions, at least not as usually construed. To see this, ask yourself what sorts of things we automatically take for granted when we understand Moore's claim, "I have two hands." ... There are lots of features of the Background that are not explicit in the semantic content of the sentence but that we automatically take for granted. For example, we take it for granted that Moore's hands stand in a certain relation to the rest of his body. We would understand the sentence quite differently if we understood it on analogy with the following: "I have two diamond necklaces and I keep them both in a bank vault in Switzerland and I have two hands and I keep them in the same bank vault." But where in the sentence does it say or imply that Moore's hands are not to be kept in a bank vault or even that they are attached to his body? This is one of the things that we simply take for granted. There is no limit to the number of such Background and Network presuppositions that we have to make in order to understand even such a simple utterance as Moore's. Thus, for example, suppose that we took it for granted that if Moore has two hands, they are attached to his body all right, but they are growing out of his left ear. Or perhaps that they are attached to his arms, but his body has shrunk to the size of a grain of sand, and his two hands have grown to be each as big as the Atlantic ocean. Again, suppose we assumed that if people have hands, they flash in and out of existence like an intermittent flashlight beam. With such crazy alterations in the Background, we would understand the sentence quite differently from the way we currently understand it. The point is that in our normal understanding we take a great deal for granted, but many of these conditions on our nor-

42

Realism/Antirealism and Epistemology

mal understanding cannot be thought of as truth conditions on the utterance without considerable distortion. These are the sorts of conditions that help us to fix the truth conditions of our utterances. They are not themselves part of those truth conditions. The claim I now want to try to substantiate is, External Realism is a Background presupposition on the normal understanding of a very large class of utterances. But it differs from many other Background presuppositions in that it is both pervasive and essential. It is pervasive in the sense that it applies to a very large class of utterances; it is essential in the sense that we cannot preserve normal understanding of these utterances without it. To see that it is pervasive, notice that it applies to a large range of quite different kinds of utterances such as Mt. Everest has snow and ice near the summit. My dog has fleas. Hydrogen atoms each contain one electron. To show that it is essential we need to remind ourselves that the sentences in question, as sentences of a public language, are assumed to be understood in the same way by any competent speaker and hearer. Normal understanding requires sameness of understanding by both speaker and hearer, and sameness of understanding in these cases requires that utterances of the referring expressions purport to make reference to a publicly accessible reality, to a reality that is ontologically objective. But the condition on public accessibility to the sorts of phenomena in these examples is that the way that things are does not depend on your or my representations. You and I can both understand the utterances above-about Mt. Everest, my dog, and hydrogen atoms-in the same way, because we take it for granted that the utterances are about a publicly accessible reality. And this point holds even if the particular references fail because of the nonexistence of the entities we are trying to refer to. Even if it turns out that Mt. Everest and hydrogen atoms had never existed, and I never had a dog, all the same, we still understand the utterances as depending for their normal intelligibility on the existence of an external reality. We almost want to say: "Even if no Mt. Everest, no hydrogen atoms and no Searle dog, all the same External Reality would still be such that: no Everest, no hydrogen atoms and no dog." But that is the wrong way to say it, because it makes it look as if each utterance contains a concealed reference to some special entity called "External Reality" with capital E and R; and that is precisely what we do not want to say. What we should say, rather, is this:

Does the Real World Exist?

43

A public language presupposes a public world in the sense that many (not all) utterances of a public language purport to make references to phenomena that are ontologically objective, and they ascribe such-and-such features to these phenomena. Now, in order that we should understand these utterances as having these truth conditions-the existence of these phenomena and the possession of these features-we have to take for granted that there is a way that the world is that is independent of our representations. But that requirement is precisely the requirement of external realism. And the consequence of this point for the present discussion is that efforts to communicate in a public language require that we presuppose a public world. And the sense of 'public' in question requires that the public reality exists independently of representations of that reality. The point is not that in understanding the utterance we have to presuppose the existence of specific objects of reference, such as Mt. Everest, hydrogen atoms, or dogs. No, the conditions of intelligibility are still preserved even if it should turn out that none of these ever existed. The existence of Mt. Everest is one of the truth conditions of the statement; but the existence of a way that things are in the world independently of our representations of them is not a truth condition but rather a condition of the form of intelligibility that such statements have. The point is not epistemic. It is about conditions of intelligibility and not conditions of knowledge, because the point applies whether or not our statements are known or unknown, and whether they are true or false, and even whether the objects purportedly referred to exist or not. The point is simply that when we understand an utterance of the sorts we have been considering, we understand it as presupposing a publicly accessible reality. There is another way to work up to the same conclusion. Any truth claim presupposes that there is a way that things are regarding the content of that claim. And this point holds as much for mathematical statements such as 2+2=4

or for statements about personal experiences such as I am in pain as it does for statements about mountains, dogs, and electrons. What is special about these latter sorts of statements is that they purport to make reference to publicly accessible phenomena, in these examples, publicly

44

Realism/Antirealism and Epistemology

accessible physical objects. But for such cases we presuppose not only that there is a way that things are that is independent of our representations, but that there is a way that things are in a publicly accessible, that is, ontologically objective, realm. But the presupposition of a mind-independent reality already contains the presupposition of a representation-independent reality, and that presupposition just is external realism. ER, so construed, is a purely formal constraint. It does not say how things are but only that there is a way that they are that is independent of our representations. The argument so far can be summarized in a series of steps: 1. The normal understanding of utterances in a public language

requires that the utterances be understandable in the same way by any competent speaker and hearer of the language. 2. A large class of utterances purports to make reference to phenomena that exist outside of, and independently of, the speaker, the hearer, and their representations, and indeed, in some cases, independently of all representations. 3. Features 1 and 2 require that we understand the utterances of many of these sentences as having truth conditions that are independent of our representations. By purporting to make reference to public phenomena, phenomena that are onto logically and not merely epistemically objective, we presuppose that the truth or falsity of the statements is fixed by how the world is, independently of how we represent it. 4. But that presupposition amounts to the claim that there is a way things are that is independent of our representations, and that claim is just (one version of) external realism. One last way-and perhaps the simplest way-to see this point is to use Brute Force: Put an explicit statement of the denial of the Background conditions into the speech act itself and see what happens. See, for example, how it contrasts with the denial of standard truth conditions. If I say: Mt. Everest has snow and ice near the summit and there is no snow on Mt. Everest, what I have said is self-contradictory, because the first clause entails the negation of the second. But if I say:

Does the Reol World Exist?

45

Mt. Everest has snow and ice near the summit, and external reality has never existed, what I say is literally puzzling. We do not know how to understand it in the normal way, because the second clause doesn't just contradict the first clause, but denies a condition that is taken for granted in the normal understanding of the first. Berkeley and other idealists recognized something very much like this point. Berkeley saw that it was a problem for his account that if each person refers only to his or her own ideas when speaking, then there is a question about how one succeeds in communicating with others. Berkeley's answer was that God guarantees successful communication. This, I believe both Berkeley and I would agree, is not a case of normal understanding in my sense. When I say 'snow is white' or 'my dog has fleas', I am not normally taken to be relying on God, since even an atheist can attempt to communicate in a public language. Berkeley saw that the price for abandoning external realism was an abandonment of normal understanding, and he was willing to pay the price. One objection to some of the current challenges to realism is that they want to abandon external realism without paying the price. The price of the abandonment of realism is the abandonment of normal understanding. If someone wishes to abandon normal understanding, he or she owes us an account of what sort of understanding is possible.

The Distinction between Brute Reality and Socially Constructed Reality My argument is not yet complete. The argument so far, if valid, is an answer to phenomenalist idealism but not to social constructionism. What it shows so far is that for a large class of utterances, each individual utterance requires for its intelligibility a publicly accessible reality. I have further characterized that reality as representation independent. But there is still an ambiguity. Talk of money and marriages is talk of a publicly accessible reality, and such phenomena are "representation independent" in the sense that this twenty-dollar bill or this marriage between Sam and Sally exists independently of your or my representations of it. After all, statements about money meet the conditions that there are facts independent of the speech act that make them satisfied or unsatisfied, for example, 'You owe me five dollars' presupposes an independently existing reality as much as does 'Mt. Everest has snow and ice near the summit'. But marriages and money, unlike mountains and atoms, do not exist independently of all representations, and this distinction needs to be made explicit in

46

Realism/Antirealism and Epistemology

the account. The argument so far might be interpreted to allow that all of reality is socially constructed in the way that, for example, money is socially constructed. Facts about money can be epistemically objective even if the existence of money is socially constructed, and, therefore, to that extent, ontologically subjective. To complete the argument we need to show that within the class of speech acts that refer to a reality beyond themselves there is a subclass whose normal understanding requires a reality independent of all representation. The simplest way to show that is to show that a socially constructed reality presupposes a reality independent of all social constructions, because there has to be something for the construction to be constructed out of. To construct money, property, and language, for example, there have to be the raw materials of bits of metal, paper, land, sounds, and marks, for example. And the raw materials cannot in turn be socially constructed without presupposing some even rawer materials out of which they are constructed, until eventually we reach a bedrock of brute physical phenomena independent of all representations. The ontological subjectivity of the socially constructed reality requires an ontologically objective reality out of which it is constructed. To the "transcendental argument" of the previous section-a public language presupposes a public world-we add a "transcendental argument" in this section-a socially constructed reality presupposes a nonsocially constructed reality. By this stage in the argument I hope the point is obvious. In a sense, one of the main aims of this book has been to spell it out. Because the logical form of the creation of socially constructed reality consists in iterations of the structure X counts as Y in C, the iterations must bottom out in an X element that is not itself an institutional construction. Otherwise you would get infinite regress or circularity. It is a logical consequence of the main argument of the book that you cannot have institutional facts without brute facts. ( To conclude the discussion of realism I would like also to show that there is a contrast between the conditions on our normal understanding of statements about brute physical facts and those about institutional facts. To show that there is a class of speech acts that presuppose for their intelligibility a reality beyond all representations, let us once again use "Brute Force" and observe the consequences of putting the counterfactual supposition of the denial of the condition into the representation itself. Consider, for example, the claims: 1. Mt. Everest has snow and ice near its summit,

Does the Real World Exist?

47

and its negation, 2. It is not the case that Mt. Everest has snow and ice near its summit. Speech acts of the sort exemplified by claims 1 and 2, so I will argue, purport to state facts that are "ontologically objective" and therefore "representation independent" in the sense that I have tried to explain. In this respect they differ from, for example, the claim 3. You owe me five dollars, and its negation, 4. It is not the case that you owe me five dollars. We can see the difference if we put the counterfactual supposition into the claims, as follows: A. In a world that is like ours, except that representations have never

existed in it, Mt. Everest has snow and ice near the summit, and B. In a world that is like ours, except that representations have never existed in it, it is not the case that Mt. Everest has snow and ice near the summit. Notice that in A and B, on our normal, naIve, intuitive understanding, the supposition of the antecedent does not affect our understanding of the whole statement, as is shown by the fact that the negation of the consequent leaves the status of this type of statement unaffected. The truth or falsity of both A and B depends entirely on the presence or absence of snow and ice near the summit of Mt. Everest, and the presence of snow and ice near the summit of Mt. Everest is in no way dependent on the existence of human or other sorts of representations. But contrast these cases with, C. In a world that is like ours, except that representations have never existed in it, you owe me five dollars,

and

48

Realism/Antirealism and Epistemology

D. In a world that is like ours, except that representations have never existed in it, it is not the case that you owe me five dollars. There is a crucial difference between A and B, on the one hand, and C and D, on the other. On our normal understanding, A and B are unaffected by the counterfactual supposition; our understanding is the same and their truth depends entirely on the existence of snow and ice at the summit of Mt. Everest. But C, as it stands, is puzzling and even self-defeating in the same way that 'There is snow on Mt. Everest and the external world has never existed' is self-defeating, for a condition of the possibility of your owing me money is the existence of certain human rules, practices, and institutions. And this is shown by the fact that if we negate the consequent in C so that we get D, if we could understand the result at all we would have to understand it as a trivial truth: There is no way that anyone could owe anyone anything in a world without representations. To say that you owe me money in a world in which no one ever said or thought anything would be like saying you got a base hit to left center field in the third game of the World Series in a world in which baseball never existed. To summarize, the claim that I am making is this: Any statement is a representation and therefore to be understood as a statement must be understood as a representation. Statements 1, 2, 3, and 4 all share this feature. But there is a difference between statements 1 and 2, on the one hand, and 3 and 4 on the other. 1 and 2 purport to represent mind-independent features of the world and therefore do not require the existence of representations in the world as part of the conditions of their normal intelligibility. Statements 3 and 4, on the other hand, purport to be about representation-dependent features of the world and therefore do require the existence of representations as part of the conditions of their normal intelligibility. You can see this by considering the normal understanding of sentences where 1, 2, 3, and 4 are embedded in sentences expressing a counterfactual supposition of the nonexistence of any representations, A, B, C and D. On our normal understanding, the truth value of 1 and 2 is unaffected; the truth value of 3 and 4 is affected decisively. On the supposition, 3 becomes self-defeating, almost self-contradictory; 4, if intelligible at all, becomes trivially true. Thus, on our normal understanding, statements about money require the existence of representations as part of their conditions of normal intelligibility. Statements about mountains are entirely free of any such requirement. The upshot then is that there is a contrast between the role of the presupposition of external realism and the presupposition of the existence of

Does the Real World Exist?

49

human representations in normal understanding. Normal understanding of talk of both money and mountains requires external realism, but normal understanding of talk of money presupposes the existence of representations in a way that normal understanding of mountains does not. Money is understood as socially constructed; mountains are not understood as socially constructed.

Strengths and Limitations of the Foregoing Arguments The aim of this part has been to show that our ordinary linguistic practices presuppose external realism, just as the aim of part 1 was to show that certain arguments against that presupposition do not work. Now I want to say what is and what is not proved by the "transcendental arguments" in this part. 1. I have not demonstrated that external realism is true. I have tried to show that it is presupposed by the use of very large sections of a public language. If you take yourself to be communicating with others in the normal way in the sort of speech acts I have given as examples, you are committed to external realism. I have not shown that there is a real world but only that you are committed to its existence when you talk to me or to anyone else. 2. An alternative is always solipsism, the view that my mental states are the only things that exist. I have not refuted solipsism; that is, I have not refuted solipsism for me. Only remember: Your solipsism is instantly refuted by me; mine-assuming that you exist-instantly refuted by you. 3. I have not shown that we all have a belief or are committed to a belief in realism. On the contrary, realism is part of the Background; and when functioning the Background is not a matter of any intentional states at all. One of the keys to understanding the Background is this: One can be committed to the truth of a proposition without having any beliefs, thoughts, assumptions, hypotheses, or other "propositional attitudes" regarding that proposition at all. 'Taking something for granted' need not name a psychological state. Pretheoretically we take external realism for granted, and for that reason it need not be a belief, but is prior to having beliefs. 4. There is nothing epistemic about the arguments. I am not saying that in order to know the truth of our claims we have to presuppose realism. My argument is completely independent of questions of knowledge or even of truth. On my account, falsehood stands as much in need of the real world as does truth. The claim, to repeat, is about conditions of intelligibility, not about conditions of knowledge.14 5. The arguments only apply to utterances for which there is a normal

50

Realism/Antirealism and Epistemology

understanding. Famously, there is no normal understanding of quantum mechanics or the set theoretical paradoxes. The struggles over the interpretation of quantum mechanics are, at least in part, an attempt to provide a normal understanding of these claims. Not every proposition about the world has a normal understanding. 6. There is nothing self-guaranteeing about normal understanding. Sometimes we are forced to revise our normal understanding because of new discoveries. This happened in the case of color statements. Pretheoretically we think of colors as intrinsic features of objects; but physics tells us that as far as color is concerned, the only intrinsic feature of an object is that it differentially scatters and absorbs the various wavelengths of light. These light/matter interactions are detected by our nervous system, producing the experiences that we interpret as color. In such a case, we replace one normal understanding with another. But notice that the replacement (and presumably correct) normal understanding presupposes ER as much as did the earlier (presumably mistaken) normal understanding. To put this point very crudely: the discovery that colors as such are not a part of the external world does not threaten our presupposition of the existence of the external world, because we still rely on the external world to give our backup account of the subjective illusion of color. Similar remarks could be made about, for example, solidity. The prospects of refuting ER by appealing to the history of science would appear to be doomed to failure, because the history is one of replacing a mistaken normal understanding, where an apparently ontologically objective phenomenon is shown to be really subjective, with an account given in terms of phenomena that supposedly really are objective. 7. If my argument is correct it should go part of the way toward explaining our embarrassment in the face of demands to prove the existence of the real world, and the inadequacy of existing proofs. I think it does this. Once we start talking to our interlocutors we have already presupposed the existence of the real world, and we are embarrassed to try to prove what our attempts at proof already presuppose. I conclude this chapter by answering the following question: Why does it matter? What difference does it make? After all, as Wittgenstein says somewhere, it is possible to construe these great debates between realism and antirealism, between idealism and materialism as just so many battle cries. The antirealist nonetheless takes his car to a mechanic to get it fixed and brushes his teeth, just as if he believed they were objects in the external world. So what difference does it make whether or not one says that one is a realist or an antirealist? I actually think that philosophical theories make a tremendous differ-

Does the Real World Exist?

51

ence to every aspect of our lives. In my observation, the rejection of realism, the denial of ontological objectivity, is an essential component of the attacks on epistemic objectivity, rationality, truth, and intelligence in contemporary intellectual life. It is no accident that the various theories of language, literature, and even education that try to undermine the traditional conceptions of truth, epistemic objectivity, and rationality rely heavily on arguments against external realism. The first step in combating irrationalism-not the only step but the first step-is a refutation of the arguments against external realism and a defense of external realism as a presupposition of large areas of discourse.

Notes 1. There is, furthermore, a problem about the human brains and the human interactions themselves. Are they also supposed to be constructed by human interactions? 2. H. Putnam, Realism with a Human Face (Cambridge: Harvard University Press, 1990), 23. 3. Quoted by N. Goodman, Of Mind and Other Matters (Cambridge: Harvard University Press, 1984), 36. 4. H. R. Maturana, F. J. Varela, Autopoieses and Cognition: The Realization of the Living (Dordrecht: D. Reidel, 1980). 5. Terry Winograd, "Three Responses to Situation Theory," Center for the Study of Language and Information, Report No. CSLl-87-106, 1987; and Terry Winograd and Fernando Flores, Understanding Computers and Cognition (Norwood, N.J.: Ablex, 1986), chap. 5. 6. G. Levine, "Looking for the Real: Epistemology in Science and Culture," in

Realism and Representation: Essays on the Problem of Realism in Relation to Science, Literature, and Culture, ed. G. Levine (Madison: University of Wisconsin Press, 1993), 13.

7. J. Derrida, Limited Inc. (Evanston, Ill: Northwestern University Press, 1988), 136. 8. Putnam, Realism with a Human Face, 96 ff. H. Putnam, The Many Faces of Realism (LaSalle, Ill.: Open Court, 1987), 18 ff. 9. Goodman, Of Mind and Other Matters, 36. 10. H. Putnam, Reason, Truth, and History (Cambridge: Cambridge University Press, 1981), xi. The phrase is repeated in Putnam, Many Faces of Realism, 1. 11. Ludwig Wittgenstein, Philosophical Investigations (Oxford: Basil Blackwell, 1953), pt. 1, para. 464 (my translation). 12. I apologize for the brevity of this discussion. I have discussed these same issues in greater detail in chap. 2 of Intentionality. For the best argument against the sense-datum theory, see J. L. Austin, Sense and Sensibilia (New York: Oxford University Press, 1962).

52

Realism/Antirealism and Epistemology

13. Ludwig Wittgenstein, Tractatus Logico-Philosophicus (London: Routledge & Kegan Paul, 1922). 14. Putnam, attacking realism, describes it as the view that "Truth is supposed to be radically nonepistemic." H. Putnam, Meaning and the Moral Sciences (London: Routledge & Kegan Paul, 1978), 125. But realism is the claim that reality is radically nonepistemic. And if it should turn out that the concept of "truth" is not radically nonepistemic, then we should simply have to get another concept that was, for we need a nonepistemic term to describe the correspondence between our statements and the radically nonepistemic real world.

3 Realism and the Tasks of Epistemology

William P. Alston

I. Alethic Realism

he proliferation of uses of the term 'realism' has kept pace with the recent increase of interest in the subject or, more accurately, with the recent explosion of interest in a variety of issues that go under that rubric. In this paper I want to discuss the bearing on epistemology of the position one takes with respect to realism in one sense of that term. It is no part of my purpose to launch a crusade for using 'realism' in this way. I have no objection whatever to people using the term in other senses; I frequently do so myself. (I do object to people using the term without attaching any definite sense to it and to using it in a confused way, but those are other matters.) The sense of the term in which this paper is concerned with realism is one in which it is primarily a doctrine about truth. Call this 'alethic realism'.1 We may think of alethic realism as a conjunction of two theses.

T

1. The senses of 'true' and 'false' in which such items as beliefs, state-

ments, and propositions can be evaluated as true or false are what I will call realist senses of these terms. 2. It is important to determine the truth-value of such items in this sense. The first step in filling out this formulation is to explain the realist conception of truth.

2. The Realist Conception of Truth

The basic intuitive idea of the realist conception of truth and falsity is that what determines whether a statement is true is whether what the

54

Realism/Antirealism and Epistemology

statement says to be the case is the case. The statement that sugar is sweet is true iff sugar is sweet; otherwise it is false. Sugar's being sweet is all it takes for that statement to be true, and it is fully sufficient to do the job. Why do I call this a realist conception of truth? It is because what it takes to make a statement true on this conception is the actual obtaining of what is claimed to obtain in making that statement. If what is stated is that grass is green, then it is grass's being green that is both necessary and sufficient for the truth of the statement. Nothing else is relevant to its truth value. This is a realist way of thinking of truth in that the truth maker is something that is objective vis-a-vis the truth bearer. It has to do with what the truth bearer is about, rather than with some "internal" or "intrinsic" feature of the truth bearer, such as its epistemic status, its place in a system of propositions, or the confidence with which it is held. It is not required for the truth of the statement that I, or anyone else, or any social group know that grass is green or be justified or rational in believing it. It is not required that science be destined, in that far-off divine event toward which inquiry moves, to arrive at the conclusion that grass is green. It is not required that it have been rendered probable by some body of empirical evidence. So long as grass is green, then what I said is true, whatever the epistemic or other status of that proposition, belief, or statement visa-vis any particular individual or community. Truth has to do with the relation of a potential truth bearer to a reality beyond itself. Nor is any of the other things I just mentioned sufficient for truth. It is obviously possible for an individual to believe rationally or justifiably that Sam is guilty of the murder even though it is not true that he is. The possibility of justifiably believing falsehoods is a truism of epistemology. But on the realist conception of truth, even the fact that a hypothesis is firmly accepted in a scientific community, however that community is specified (apart from being specified as having only true beliefs), is logically compatible with that hypothesis being false. So far I have conveyed the intuitive idea of a realist conception of truth. But providing a satisfactory formulation of this intuitive idea is far from simple. To see some of the problems involved, let's first consider what we should take to be bearers of truth-value. Truth-values are commonly ascribed to, inter alia, statements, beliefs, and propositions. In my view, propositions are the primary truth-value bearers. A little reflection will reveal that it is really propositions we have in mind when we speak of statements or beliefs as true. When I predicate truth of the statement that sugar is sweet or the belief that sugar is sweet, it is what is stated or what is believed, not the act of stating or the state of

Realism and the Tasks

of Epistemology

55

belief that is said to be true. 'Your state of belief that p is true' or 'Your act of stating that p is true' does not have any clear sense. And what is stated or what is believed is a proposition. When the proposition that p plays the role of the content of a statement or a belief, it is "the statement that p" or "the belief that p" in the sense of those terms in which they denote a truth-value bearer. Hence I will feel free in this essay to oscillate freely between speaking of propositions, statements, and beliefs as bearers of truth value. Note that I have avoided mentioning sentences, the current truth-value bearer of choice among English-speaking philosophers. I don't find it at all plausible or useful to think of sentences as true or false, but I won't be able to go thoroughly into my reasons for this. Suffice it to say that we can't take sentence types to be truth bearers if we want truth value to be a stable property of its bearers. Many sentences-for example, 'I'm hungry'-are sometimes used to say something true and sometimes to say something false. Taking sentence tokens as bearing truth-values escapes that objection, but we can explain what is meant by a sentence token's being true only by saying that in issuing it one makes a true statement. So we may as well stick with the more basic notion of a statement's being true. Now back to the problem of how to formulate the realist conception of truth. The initial intuitive formulation could, of course, be put as a contextual definition. 3. A statement is true =df What the statement says to be the case actually is the case. Otherwise it is false. 2 But there are problems with this. First, a statement doesn't literally say anything. It is speakers who say things in making statements. We can take account of this by replacing (3) with (4). 4. A statement is true = df If one made that statement, what one would thereby say to be the case actually is the case. Note that in putting this forward as a completely general account of what it is for a statement to be true, I am assuming an unrestrictedly general form for statements-namely, that they all have the form 'It is the case that p'. It is obvious that we do not always use that verbal formulation, but I believe that any statement could be so formulated without change of meaning. However, we could give a more natural account of the same general sort by taking every statement to predicate something of one or more items the statement is "about."

56

Realism/Antirealism and Epistemalogy

5. A statement is true =df" If one made that statement what one would thereby be talking about has the property one would thereby be predicating of it (them). It is obvious that not every statement has a simple subject-predicate form.

But I believe that all statements could be construed in such a way as to conform to (5). For example, we can understand quantified statements as predicating something of all the values of the variables involved. And we can understand singular hypotheticals like 'If he is invited he will come' as predicating a hypothetical property, 'will come if invited', of the person about whom the statement is made. However, these formulations are confined to statements, that is, to propositions functioning as the content of an act of stating. What about beliefs? And what about propositions themselves in abstraction from functioning as the content of statements and beliefs? Well, we can easily handle beliefs in parallel fashion. 6. A belief is true = di. What would be believed to be the case if this belief were held actually is the case. 7. A belief is true = df" What one who holds the belief thereby supposes to characterize what the belief is about really does characterize what the belief is about. As for propositions considered most generally, in abstraction from their involvement as contents of statements or beliefs, we can piggyback on the above formulations by considering what it would be for a statement or belief that had that proposition as content to be true. We can construct derivatives of this sort from any of (4)-(7). Here is one that derives from (4). It is obvious how to construct the others. 8. A proposition is true = df. If one were to make a statement with that proposition as content, what one would thereby say to be the case actually is the case. Note that these principles are to be understood not as material biconditionals-claiming only that the two sides have the same truth-value-but as necessarily true and, furthermore, since definitional, as analytically, conceptually true, true by virtue of the meanings of the terms (the content of the concepts expressed by the terms), including the term 'true'. A reader might feel that by this point I have provided her with enough,

Realism and the Tasks of Epistemology

57

and more than enough, alternative formulations from which to choose. But I am moved to extend this prodigality still further. I have two reasons for doing so. First, (4)-(8) all trade on one or another assumption about a general form into which all propositions can be put, and these assumptions can be questioned. Second, if propositions are the primary bearers of truth-values, it is less than fitting that the definition of truth for propositions be derivative from the definition of truth for statements and beliefs. Hence I would like to find a formulation that is superior in both these respects. A clue is found in the fact that all these principles have instances of the following sort. 9. The proposition (statement, belief) that salt contains sodium is true iff salt contains sodium. That is, for any declarative sentence s that expresses a proposition, we can form a true statement by replacing p with s in the following schema, which I will call the T-schema. 10. The proposition that p is true

iff p.

One might think that (10) is the formulation we seek. It is specifically about propositions directly, and it does not make any assumptions about the form of all propositions. But (10) has liabilities of its own. First, unlike (4)-(8), it does not give us a contextual definition of 'true'. Second, it is not a statement at all, much less an unqualifiedly general statement as to the conditions under which a proposition is true. It is merely a schema for statements. We could try to remedy the second deficiency by universally quantifying (10), thereby yielding: 11. (p) The proposition that p is true

iff p.

But this will count as a universal generalization only if it is understood in terms of substitutional rather than objectual quantification. For to obtain an instance, 'p' must be replaced by a declarative sentence, not by a singular referring term. I don't want to take on the burden of defending substitutional quantification, so I will, instead, move to the metalanguage and formulate a principle about (10). 12. Any instance of (10) is necessarily, analytically true by virtue of the meaning of, inter alia, 'true'.

58

Realism/Antirealism and Epistemology

In the sequel, I will sometimes speak of "accepting the T-schema," by which I will mean accepting (12). Although (12) is still not in the form of a definition of 'true', contextual or otherwise, it does provide a way of explaining a meaning of 'true', in the way any analytically true statement tells us something about the meanings of its constituent terms. Moreover, it does not just tell us something about the meaning of 'true'; it uniquely identifies the realist sense of the term. This can be seen from the fact that the set of sentences that (12) proclaims to be analytically true is just the set from which we would draw examples of what is said about truth by (4)-(8). Any of the sentences over which (12) ranges can be used to formulate examples of what (4)-(8) says to be (necessarily) necessary and sufficient conditions for the truth of a proposition, statement, or belief. For example, to illustrate the point that a statement is true iff what the statement says to be the case actually is the case, we would trot out such examples as that the statement that there is sugar in this bowl is true iff there is sugar in this bowl. It seems clear that what is said, in a certain way, by (4)-(8)is also said, in a different way, by (12) . It will not have escaped the reader's notice that (10) looks very much like Tarski's form (T): 13. X is true if, and only if, p.3 (Here 'p' is to be replaced by a declarative sentence and 'X' by a name of that sentence.) Nevertheless, there are fundamental differences as well. First, Tarski takes sentences, while I take propositions, as truth-value bearers. Second, I claim that the instances of my schema are analytically, necessarily true, while Tarski claims truth only for his "equivalences of form T." These differences stem from fundamental differences in our projects. Whereas Tarski set out to provide an explicit definition for a truth predicate for a formal language, I want to give an informal explanation of a truth predicate that I am not thinking of as applied to linguistic items at all. Principle 12 counts as what Horwich and others have termed a "minimalist" account of truth;4 it does the minimum to delimit effectively the concept of truth. It is my claim that it is sufficient for this. Anyone who sees that any instance of (10) is analytically true has a firm grasp of the realist conception of truth. That is not to say that it tells us everything we would like to know about the property of truth, the property picked out by this concept. For example, it is natural to think of realist truth in some such terms as "Whether a statement is true is a matter of whether it corresponds with the facts, whether it fits the world, whether there is a match between the statement and how things are .... " The connection of this

Realism and the Tasks of Epistemology

59

with (12) can be seen as follows. If it is true that the sun is shining if and only if the sun is shining, that could just as well be put by saying that it is the fact that the sun is shining that makes true the statement that the sun is shining. If that fact obtains, the statement is true; if it doesn't, the statement is false. Facts are truth-makers. But can the truth-making relationship be ultimate, or does it supervene on some more basic relation between statement and fact? It is plausible to choose the latter. Doesn't there have to be something like a "fit," a "match," a "correspondence" between a given statement and a given fact by virtue of which that fact is the one, out of all the facts that do or might obtain, that is the truth maker for that statement, the one the obtaining of which is necessary and sufficient for the truth of that statement?5 But even if this is so, we can zero in on the realist concept of truth, as I have just done in terms of the T-schema, without being committed to any particular view as to what this correspondence relationship is on which truth supervenes. For example, we are not committed to the view that a structural isomorphism between statement (proposition) and fact is what renders the latter a truth maker for the former. No doubt a full development of a realist theory of truth would involve some further details concerning the correspondence relation. But one need not wait for such developments to appreciate the truth of alethic realism as set out here. An incomplete theory may be the right one, for all that. When I speak in this essay of the realist conception of truth, you may think of it in terms of any of my formulations-(4)-(8) or (12). As I pointed out, it seems clear that they are all equivalent. I will be thinking mostly in terms of (12), but that is largely a matter of convenience. 6 The second clause of alethic realism is that it is important to assess propositions (statements, beliefs) for their truth value in the realist sense. Important for what? An adequate answer would run far beyond the bounds of this paper. It would encompass all the reasons, both theoretical and practical, that it is important for us to get it right rather than wrong about matters that concern us. Just for starters, to attain our goals, it is obviously important that we have a correct rather than an incorrect "map" of the things with which we must interact in certain ways in order to attain those goals. 7 If I need to get to Chicago to deliver a paper, I had jolly well better have true rather than false beliefs about the location of Chicago vis-a.-vis where I am at the moment and about what mode of transportation will get me there. Since my central issue in this paper is whether a concern with realist truth is important in doing epistemology, I will forgo any further discussion of this second component of alethic realism at this point.

60

Realism/Antirealism and Epistemology

3. Alethic and Other Realisms I have already acknowledged that what I am discussing here under the label "realism" is only one of the positions so called, and properly so called. For one thing, any claim that entities of a given description are real, exist, are really "out there," are to be reckoned among the "ultimate furniture of the universe," etc., is, reasonably enough, termed a form of realism. Thus we can have realism about universals, about values, about moral obligation, about intentional psychological states, about theoretical entities, about physical objects, and so on. Call these "departmental realisms." A departmental realism is susceptible of more than one contrary. Realism about physical objects would be denied by one who flatly denies that there are any such things, but also by one who acknowledges that there are tables and chairs but denies that they have the kind of "objective" status we commonsensically take them to have. An antirealist of this latter stripe might be a phenomenalist who says that tables and chairs are nothing but aggregates of actual and possible sensations (sensedata) or an idealist who takes them to be thoughts in the mind of the Absolute. Alethic realism is obviously distinct from departmental realisms in being of unrestricted scope. Alethic realism holds that all statements are usefully assessed for realist truth value. 8 In an excellent article, Colin McGinn writes: "Except in the vulgar sense, one is not a realist tout court; one is a realist with respect to some or other type of subject matter."9 To this I am tempted to react, as Thomas Reid did to Hume's contrast between the vulgar and philosophical opinions about the immediate objects of perception, by saying, "In this division, to my great humiliation, I find myself classed with the vulgar."l0 But even if one were to combine into a comprehensive metaphysical position all the departmental realisms to which one is committed, the result would still be a very different kettle of fish from alethic realism. For alethic realism is compatible with any metaphysical view whatever as to the status of entities of this or that type, including metaphysical views of maximal comprehensiveness. Whether physical objects enjoy their own mode of being independent of our experience of them or whether they exist only as patterns of actual or possible sense experiences, the statements we make about them can be meaningfully and usefully assessed for realist truth-value. The same judgment is to be made concerning the general metaphysical view most commonly called "realism," namely, the view that not everything that exists is dependent on mind. If I reject realism in this sense and adopt some form of idealism, I can continue to take my

. '.

Realism and the Tasks of Epistemology

61

statements about reality to be properly assessed as true or false in the realist sense. Even if I hold there to be nothing but the Absolute Spirit, I can still recognize that it is true that the Absolute Spirit thinks about matter in terms of the periodic table if and only if the Absolute Spirit thinks about matter in terms of the periodic tableY In short, alethic realism is independent of issues between metaphysical realisms and their contraries.

4. Alternatives to Alethic Realism: Epistemic Construals of Truth

I have been making a lot of fuss about the "realist conception of truth," as if it were some elaborate, complex theoretical conception that is properly a matter for philosophical debate and controversy. But in fact it is a miserable truism. What could be more obvious than that it is (necessarily) true that there are books in my study if and only if there are books in my study? Indeed, this point would hardly be worth mentioning, much less being made the subject of articles and books, were it not for the fact that it is so frequently and enthusiastically denied. If someone is going to deny this obvious truth, what additional or alternative requirements will be suggested in place of things being as the statement states? If we look at the antirealist literature of the nineteenth and twentieth centuries, we will find that the alternative is always something epistemicP What it takes to make a statement or a belief true is some epistemic status it enjoys, either for the individual subject or for some social group. That has been the constant refrain of the various coherence and pragmatist accounts of truth with which we have been bombarded for the last hundred years or so. Thus F. H. Bradley tells us that truth is "that which satisfies the intellect,"13 "an ideal expression of the Universe, at once coherent and comprehensive."14 In like vein Brand Blanshard holds that a proposition is true if it coheres with an all-comprehensive and fully articulated whole. IS From the pragmatist's side C. S. Peirce's well-known view is that "the opinion which is fated to be ultimately agreed to by all who investigate, is what we mean by the truth,"16 while William James writes that "true ideas are those that we can assimilate, validate, corroborate, and verify."17 John Dewey holds true ideas to be those that are instrumental to "an active reorganization of the given environment, a removal of some specific trouble or perplexity."IS More recently Hilary Putnam objects to "the most important consequence" of what he calls "metaphysical realism", that is, "that truth is supposed to be radically non-epistemic ... and so the theory that is 'ideal' from the point of view of operational utility, inner beauty and elegance, 'plausibility', 'simplicity', 'conservatism', etc. might

62

Realism/Antirealism and Epistemology

be false."19 Putnam's alternative to this is not to identify truth with rational acceptability tout court, but to hold that "truth is an idealization of rational acceptability. We speak as if there were such things as epistemically ideal conditions, and we call a statement 'true' if it would be justified under such conditions."2o Again, '''Truth', in an internalist view [Putnam's term for his alternative to "metaphysical realism"]' is some sort of (idealized) rational acceptability-some sort of ideal coherence of our beliefs with each other and with our experiences as those experiences are themselves represented in our belief system -and not correspondence with mind-independent or discourse-independent 'states of affairs'."21 And yet again, "a piece of knowledge (i.e., a 'true statement' [sic]) is a statement that a rational being would accept on sufficient experience of the kind that it is actually possible for beings with our nature to have."22 These versions differ in several ways. One of special importance has to do with whether the thinker identifies truth with some positive epistemic status of an ordinary, everyday sort, as both James and Dewey seem to do, or takes truth to consist in what would be positively assessed under highly idealized conditions, as Bradley, Blanshard, Peirce, and Putnam do in their several ways. The latter position is obviously more defensible. As I have pointed out, it is a truism of epistemology that false statements can be justified or rationally accepted if we employ everyday standards for these statuses. At a certain point, one can be justified in believing that Sam is the murderer, whereas actually it is Joe who did it. Newton was eminently rational in accepting his theory of gravitation, whereas now we know that it is not quite accurate. And so it goes. But the idealized versions are not so easily refuted. So as to have a reasonably definite target for discussion, I will concentrate on Putnam's view that a statement is true if it would be justified under epistemically ideal conditions. Putnam doesn't do much to spell out the idea of epistemically ideal conditions, but let's say that they would be conditions in which everything relevant to the epistemic assessment of a given proposition would be readily available. 23

5. Difficulties with Epistemic Accounts of Truth Although my primary aim in this paper is to explore the implications of a realist and an epistemic conception of truth for epistemology, I will postpone that endeavor long enough to give some indication of why I regard an epistemic conception of truth, even in its strongest form, as untenable. A full-dress presentation of these arguments can be found in chapter 7 of my Realist Conception of Truth. Here I will give a brief pre-

Realism and the Tasks

of Epistemology

63

sentation of the difficulty I consider to be most decisive. It will hardly come as a shock to Putnam that his epistemic view of truth is incompatible with a realist view. That, he would say, is his aim. But if I am correct in holding accepting the T-schema (that is, accepting [12]) is one way of accepting a realist conception of truth, then the fact that Putnam's view contradicts the realist account of truth turns out to be an Achilles' heel of his view rather than a strength. And leaving aside the question of whether (12) amounts to a realist conception of truth, it does seem that any epistemic account of truth contradicts (12), for anyone who claims that some epistemic condition is necessary and/or sufficient for the truth of the proposition that p would seem to be denying what is implied by an acceptance of the truth schema. If the fact that p is both necessary and sufficient for its being true that p, then there is no room for any further necessary or sufficient condition. Since the fact that p is sufficient, nothing further is necessary; and since it necessary, nothing that omits it is sufficient. But surely any position that implies a denial that it is (necessarily) true that sugar is sweet iff sugar is sweet is untenable. For how could that biconditional be false or even fail to be necessarily true? And if an epistemic account is committed to denying this, doesn't that denial amount to a death sentence? These considerations lead us to consider whether there is any way in which an epistemic theorist could reconcile her position with the T-schema. I believe that there are two ways that she might try. First, she might claim that something of an epistemic sort is necessary and/ or sufficient for what the T-schema represents as necessary and sufficient for truth. In that way the epistemic condition could be indirectly or derivatively necessary and/or sufficient for its being true that p, even though only the fact that p is directly or proximately necessary and sufficient. As an example of a derivatively necessary condition, consider the fact that it is both necessary and sufficient for my being your biological father that I impregnated the egg that developed into you. Nevertheless, it is also necessary for my being your father that I had sperm at the appropriate time, for that itself was necessary for my impregnating the egg in question. Here is an example of a derivatively sufficient condition. Though your heart's ceasing (permanently) to beat is sufficient for your death, my shooting you in the way I did was also sufficient for your death, since it was sufficient for your heart's ceasing (permanently) to beat. In like fashion, one might suggest that some epistemic condition, for example, its being justifiable in an ideal epistemic situation that sugar is sweet, is a necessary and/or sufficient condition for sugar's being sweet, and in this way is a necessary and/ or sufficient condition for its being true that

64

Realism/Antirealism and Epistemology

sugar is sweet. In order to get rid of the 'and/or', let's note that in order for the concept of truth to be the concept of justifiability in an ideal epistemic situation, the latter will have to be both necessary and sufficient for what is necessarily necessary and sufficient for truth. The concept of P cannot be the concept of Q unless being Q is necessary and sufficient for being P, and vice versa. But what are the prospects for such a relationship? It may seem plausible for this particular proposition that being justifiable in an ideal epistemic situation is both necessary and sufficient for sugar's being sweet. It does seem that if we think of an ideal epistemic situation as one in which all relevant evidence is readily available, it is reasonable to suppose that if sugar is sweet, the belief that sugar is sweet would be justifiable in such a situation. It would have everything going for it and nothing of any significance going against it. Thus ideal justifiability would seem to be necessary for the fact. How about sufficiency? That seems reasonable too. If the belief is justifiable in a situation where all relevant evidence is in, how could it fail to be the case that sugar is sweet? How could all relevant evidence point to it without its being so? But so far we have been considering only one proposition of a very special sort-a simple report of perception. What Putnam needs is the universal generalization-for any p the belief that p would be justifiable in an ideal epistemic situation if and only if p. And what are the chances of that? I think there are several reasons for rating those chances as very low. I will mention only one here. Can we assume that human cognitive powers suffice to envisage any true proposition? There are general considerations that strongly suggest a negative answer. Think of the limitations of our cognitive powerslimitations on our storage and retrieval capacity, on the amount of data we can process simultaneously, on the considerations we can hold together in our minds at one moment, on the complexity of propositions we are capable of grasping. Isn't it highly likely that there are facts that will forever lie beyond us just because of these limitations? And it is not just our finitude; there is also what we might call our "particularity." The cognitive design of human beings represents only one of a large multitude of possible designs for cognitive subjects, even for embodied cognitive subjects as finite as we, leaving out of account angels and God. It seems clear that there could be corporeal cognitive subjects with forms of sensory receptivity different from ours-sensitivity to different forms of physical energy. There could be subjects with different innate cognitive tendencies, propensities, and hard-wired beliefs and concepts. There could be subjects who reason in patterns different from those we employ.

Realism and the Tasks

of Epistemology

65

Given all this, shouldn't we take seriously the possibility that even if there is something wrong with the idea of facts that are in principle inaccessible to any cognitive subjects whatever (and I don't see any fatal flaw in this idea), it could still be that there are many facts accessible to cognizers with radically different hardware and software but totally inaccessible to us. And if this is the case, the fact that p could not be sufficient for a belief that p's being ideally justifiable. Where p is one of those propositions beyond our powers of envisagement, it could be the case that p without a belief that p being justifiable (by human beings) even in the most ideal epistemic situation. If no human being can entertain the proposition that p, no human being can justifiably believe that p in any situation whatever. But even if I am mistaken about this, and for any p, a belief that p would be ideally justifiable if and only if it is the case that p, that still doesn't give the epistemic theorist what she needs. The position she is trying to defend is that it is conceptually (semantically) necessary that it is the case that p if and only if a belief that p would be ideally justifiable. Anything less would not reconcile an epistemic conception of truth with the T-schema. Even if it were nonconceptually, nonanalytically true that it could not be the case that p if a belief that p were not justifiable in epistemically ideal circumstances and vice versa, it could still be the case that the T-schema is the only thing that lays down conceptually necessary and sufficient conditions for truth. Thus the present line of defense against the objection that is based on conflict with the T-schema would have to be that it is conceptually necessary and sufficient that it is the case that p if and only if a belief that p is ideally justifiable, and hence, by hypothetical syllogism from that and the T-schema, that it is conceptually necessary that it is true that p if and only if a belief that p is ideally justifiable. And the truth of the counterfactual-if it were not ideally justifiable that p, then it would not be the case that p-does not suffice for that conceptual linkage. In fact, wood would not burn unless there were oxygen in the immediate vicinity, but that does not show that having oxygen available is part of the (ordinary) concept of burning, or that part of what we are saying when we say that the log is burning is that oxygen is in the vicinity. A surface wouldn't look red in "normal" circumstances unless it was reflecting light with a frequency that falls within such-and-such a range (let's assume that something in this general ballpark is true); but that doesn't show that when we say the sweater looks red we are saying something about the frequency of light waves reflected from the sweater. Thus, even if we could show that the idealized justifiability of a belief that p was invariably counterfactually linked (as necessary and sufficient condition) to the fact that

66

Realism/Antirealism and Epistemology

p, that would not suffice to vindicate the position that the idealized justi-

fiability of a belief that p is conceptually necessary and sufficient for the truth of p, and hence that the former is part of the concept of the latter. And I fail to see any plausibility of such a conceptual linkage. What reason could there be for supposing that it is part of our concept of sugar's being sweet that a belief that sugar is sweet is ideally justifiable? So far as I am aware, none of the partisans of an epistemic account of truth does anything to support that supposition. And the simplest explanation of this lack is that there is nothing to be said along that line. There is, indeed, one condition under which there would be such a conceptuallinkage between the ideal justifiability of the belief that p (or some other epistemic status thereof) and p's being the case. That condition is that all propositions are to the effect that some statement, belief, or proposition is ideally justifiable (or has some other epistemic status). Indeed, if that condition were satisfied, the consistency of an epistemic conception of truth with the T-schema could be much more simply established. If what I am asserting whenever I make any assertion-for example, the one I make by saying liThe sun is shining" -is that some statement has a certain epistemic status, then, by the T-schema itself, a conceptually necessary and sufficient condition for my assertion's being true is that the statement in question does have that epistemic status. For that is what I am asserting. If the sun's shining (what I claim to obtain when I say liThe sun is shining") amounts to the fact that some proposition, belief, or statement enjoys a certain epistemic status, then the advocate of an epistemic account of truth can recognize that the statement that the sun is shining is true if and only if the sun is shining. If all assertions are assertions of epistemic status, then what it takes to make any assertion true is that the item in question does have the epistemic status in question. But this escape route is not likely to be attractive to Putnam. Surely it is as obvious as anything can be that most of our statements do not attribute epistemic status to anything, but make claims of quite different sorts. 'vVe say that it is raining, that Buffalo won the AFC championship, or that the Bohr model of the atom has been rejected. Indeed, the suggestion is worse than false; it cannot even be coherently thought. Consider the view that when I say liThe sun is shining," what I am asserting is that a certain statement is ideally justifiable. Well, what statement? The natural answer is, the statement that the sun is shining. And what statement is that? What am I asserting when I make that statement (that the sun is shining)? If the answer is that I am asserting that the statement that the sun is shining coheres with ... , then we still haven't said what statement the statement that the sun is shining is. And whatever other answer the

Realism and the Tasks

of Epistemology

67

epistemic theorist gives to the question "Which statement am I saying to be ideally justifiable?" the same problem will arise with respect to it. The general point should be clear. If we try to maintain that all statements are statements to the effect that some statement has a certain epistemic status, we have left ourselves without any resources for specifying what statement it is we are asserting, on a given occasion, to have a certain epistemic status. Nor would it help to proceed instead in terms of beliefs or propositions or anything else with propositional content. We would wind up in the same unpalatable impasse. I am not suggesting, of course, that one cannot assert that a particular statement has a certain epistemic status. The trouble comes with the claim that all statements have this form. That is what gives rise to the infinite regress.

6. Alternatives to Alethic Realism:Turning One's Back on Truth

Since alethic realism embodies two theses, opposing positions can be distinguished in terms of which thesis or theses they reject. We have been considering a view that rejects the first thesis and proposes an alternative conception of truth. But it is also possible to reject the second thesis but not the first, by denying that it is of any importance to consider whether beliefs, statements, or propositions are true. A prominent exponent of this move is Richard Rorty, at least a sometime Rorty, since, as his contribution to this volume makes explicit, he no longer views the matter in this way. But as of 1982 he was content to leave truth to those benighted reactionaries who still go in for that sort of thing, devoting himself (on this issue) to defending the claim that truth is really of no interest. The pragmatist agrees that if one wants to preserve the notion "correspondence with reality" then a physicalistic theory of reference is necessary-but he sees no point in preserving that notion. The pragmatist has no notion of truth which would enable him to make sense of the claim that if we achieved everything we ever hoped to achieve by making assertions we might still be making false assertions, failing to "correspond with something. "24 If the pragmatist is advised that he must not confuse the advisability of asserting 5 with the truth of 5, he will respond that the advice is question-begging. The question is precisely whether "the true" is more than what William James defined it as: "the name of whatever proves itself to be good in the way of belief, and good, too, for definite assignable reasons" .... To think that Truth is "out there" is, on their view, on all fours with the Platonic view that The Good is "out there" .... To think that we are "irrationalist" insofar

68

Realism/Antirealism and Epistemology

as it does not" gratify our souls to know /That though we perish, truth is so" is like thinking that we are "irrationalist" just insofar as it does not gratify our moral sense to think that The Moral Law shines resplendent over the noumenal world, regardless of the vicissitudes of spatio-temporallives. 25 When he [the pragmatist] asks himself about a given statement S, whether he "knows what has to be the case for it to be true" or merely knows "the conditions which we recognize as establishing the truth or falsity of statements of that class", he feels as helpless as when asked "Are you really in love, or merely inflamed by passion?" ... He refuses to take a stand-to provide an "analysis" of "s is true", for example, or to either assert or deny bivalence. He refuses to make a move in any of the games in which he is invited to take part. 26 What really needs debate between the pragmatist and the intuitive realist is not whether we have intuitions to the effect that "truth is more than assertibility" .... Of course we have such intuitions. How could we escape having them? We have been educated within an intellectual tradition built around such claims .... But it begs the question between pragmatist and realist to say that we must find a philosophical view which "captures" such intuitions. The pragmatist is urging that we do our best to stop having such intuitions, that we develop a new intellectual tradition. 27

These passages do not quite fit neatly into my characterization of Rorty as one who advocates turning one's back on (real) truth, rather than giving an antirealist account of it. For one thing, he obviously is not thinking of the realist conception of truth as I have been presenting it. Like other antirealists, he seeks to saddle his opponents with a certain notion of correspondence, and he makes free use of caricatures, such as "Truth being 'out there'." Moreover, at times, as in the second passage, he talks in terms of identifying truth with "the good in the way of belief" and the like. However, there are also passages that fit my portrayal, and for present purposes I will concentrate on that thread of the presentation. So let's consider the Rorty who confines his attack on alethic realism to alleging that no interest or importance attaches to applying the realist concept of truth to beliefs or statements. What should we say about that? Here I need only repeat my earlier remarks concerning the importance of assessing statements and beliefs for truth in the realist sense. It is important to me whether it is true that my wife loves me just because it is important to me whether my wife loves me. And if it is suggested that such interests as these could be expressed without using the concept of truth, the reply is that we do have occasion to generalize over propositions, statements, and beliefs with respect to truth-value. For example, in epistemology we need to make it explicit that one knows that p only if it

Realism and the Tasks

of Epistemology

69

is true that p. I fear this last consideration will not impress Rorty, who is, notoriously, on record as having announced the demise of epistemology. But, so far as I can see, his arguments, even if wholly successful, only show the failure of certain positions and certain programs in epistemology. The subject matter, the problems, the issues are still there, for all that; and hence philosophers will continue to investigate them, whatever proscriptions are posted.

7. The Tasks of Epistemology Let this suffice as a brief exposition of alethic realism, an exhibition of its plausibility, and a criticism of its putative alternatives. Much more could and should be said, but it is time to move on to the central theme of this essay-the bearing on epistemology of the choice between alethic realism and its alternatives. One finds a wide diversity of topics within epistemology, and the boundaries of the discipline are by no means well defined. But it is generally agreed that the consideration of the nature and conditions of knowledge and justified (rational) belief is at the heart of epistemology.28 I will consider some of the differences it makes to these topics what position one takes on the disputes over truth between realism and its opponents. As a background for that, we need to get clear as to where a consideration of truth enters into the investigation of knowledge and epistemic justification, as this is typically carried on in contemporary epistemology. As for the former, it is admitted on almost all hands, in twentieth-century English-language epistemology, that the truth of p is a necessary condition of knowing that p. If it is false that you are an English nobleman, then it is impossible for me to know that you are an English nobleman, whatever else is the case. How could it be otherwise? What sense can we make of the suggestion that even though there is no life on Mars, you or I know that there is life on Mars? There are derivative or ironical uses of 'know' that are compatible with falsity: "I just knew that he was going to be here. So where is he?" But it is clear, if anything is, that a fully serious attribution of knowledge that p requires the truth of p for its truth. The relation of truth to justification is both more complex and more controversial. On the one hand, it is generally agreed that it is conceptually possible for one to be justified (rational) in believing false propositions. On the other hand, it is generally thought that justification has some intimate relation with truth, though there is no agreement on just what this is. On what we may call a "truth-conducivity" position, it is a fundamental constraint on conditions of justification that the satisfaction of

70

Realism/Antirealism and Epistemology

those conditions guarantee that it is at least highly probable that the belief in question is true. Any candidate that does not meet this test is thereby disqualified. Thus, if someone claims that one is justified in believing that p if one has done everything one can to assure oneself that the belief is true, and if it can be shown that this does not guarantee a probability of truth for the belief, that claim must be rejected. If entailment does not constitute the general connection between justification and truth, what weaker tie might? Consider a case. Judy and Jack are taking a train on a fall evening. His view is obscured and he asks what color the foliage is. She might say, "Green", to which he might reply, "Can it still be green this far north?". A natural answer would be, "I can see it clearly, and it certainly looks green". It is significant that this answer both responds to a challenge of the truth of her statement and expresses a justification for her believing it. Imagine, moreover, that Jack perversely rejoins: "I know your seeing it clearly and its looking green to you justifies your belief, but is that relevant to its truth ?". What are we to make of this? ... [Hlis rejoinder seems unintelligible. I believe that it appears unintelligible precisely because of the conceptual connection between (epistemic) justification and truth: it seems to be at least partly constitutive of justification that, in some way, it counts toward truth. 29

S's believing that h on the basis of R is epistemically justified at t iff: S's believing that h on the basis of R is a reliable indication that hat t.30

The claim that justification is essentially truth-conducive is often supported by taking the basic aim of cognition to be the acquisition of true beliefs and the avoidance of false beliefs. [Epistemic justificationl has to do with a specifically epistemic dimension of evaluation. Beliefs can be evaluated in different ways. One may be more or less prudent, fortunate, or faithful in holding a certain belief. Epistemic justification is different from all that. Epistemic evaluation is undertaken from what we might call "the epistemic point of view". That point of view is defined by the aim at maximizing truth and minimizing falsity in a large body of beliefs .... [Alny concept of epistemic justification is a concept of some condition that is desirable or commendable from the standpoint of the aim at maximizing truth and minimizing falsity.31 [Elpistemic justification is essentially related to the so-called cognitive goal of truth, insofar as an individual belief is epistemically justified only if it is appropriately directed toward the goal of truth. More specifically, on the present conception, one is epistemically justified in believing a proposition only if one has good reason to believe it is true. To accept a proposition in the absence of good reason is to neglect the cognitive goal of truth. 32

Realism and the Tasks of Epistemology

71

Why should we, as cognitive beings, care whether our beliefs are epistemically justified? Why is such justification something to be sought and valued? ... [T]he following answer seems obviously correct. ... What makes us cognitive beings at all is our capacity for belief, and the goal of our distinctively cognitive endeavors is truth: we want our beliefs to correctly and accurately depict the world. If truth were somehow immediately and unproblematically accessible ... then the concept of justification would be of little significance .... But we have no such immediate and unproblematic access to truth, and it is for this reason that justification comes into the picture. The basic role of justification is that of a means to truth, a more directly attainable mediating link between our subjective starting point and our objective goal. ... If our standards of epistemic justification are appropriately chosen, bringing it about that our beliefs are epistemically justified will also tend to bring it about ... that they are true. If epistemic justification were not conducive to truth in this way, if finding epistemically justified beliefs did not substantially increase the likelihood of finding true ones, then epistemic justification would be irrelevant to our main cognitive goal and of dubious worth.33

On the contemporary scene this truth-conducivity view of justification is exemplified by reliability theory and other "externalist" positions, though it is by no means confined to them. 34 A reliabilist theory of justification, such as we find in Alvin Goldman, for example,35 holds that a belief is justified only if it is generated and/or sustained by a reliable process, one that would generate true beliefs in some (to be specified) percentage of cases over an appropriate spread. On any truth-conducivity concept of justification, we can't provide the most basic defense of a thesis as to what conditions are sufficient for the justification of beliefs of a certain type without considering whether the satisfaction of those conditions would make it likely that the belief in question is true. But by no means all theorists of epistemic justification embrace truthconducivity. Its rejection typically stems from an "internalist" orientation, according to which being justified in believing that p is a matter of how things seem from the subject's own perspective-where that perspective is usually identified with what one can ascertain just on reflection-rather than a matter of the objective likelihood of the belief's being true. [T]he concept of epistemic justification is ... internal and immediate . .. in that one can find out directly, by reflection, what one is justified in believing at any time. 36 The usual approach to the traditional questions of theory of knowledge is properly called "internal" or "internalistic". The internalist assumes that, merely by reflecting upon his own conscious state, he can formulate a set of epistemic principles that will enable him to find out, with respect to any

72

Realism/Antirealism and Epistemology possible belief he has, whether he is justified in having that belief. The epistemic principles that he formulates are principles that one may come upon and apply merely by sitting in one's armchair, so to speak, and without calling for any outside assistance. In a word, one need consider only one's own state of mind. 37 Everyone of every set of facts about S's position that minimally suffices to make S, at a given time, justified in being confident that p must be directly recognizable to S at that time. By 'directly recognizable' I mean this: if a certain fact obtains, then it is directly recognizable to S at a given time if and only if, provided that S at that time has the concept of that sort of fact, S needs at that time only to reflect clear-headedly on the question of whether or not that fact obtains in order to know that it does. 38

There is a tension between internalism and truth-conducivity just because the fact that a condition bestows an objective likelihood of truth on a belief is not something that one can ascertain just on reflection. Hence internalists typically reject truth-conducivity.39 According to this traditional conception of "internal" epistemic justification, there is no logical connection between epistemic justification and truth.4o Epistemic rationality is distinguished from other kinds of rationality by its truth-directed goal. The goal is for one now to believe those propositions that are true and now not to believe those propositions that are false. However, to say that the goal that helps distinguish epistemic rationality from other kinds of rationality is a truth-directed goal is not to say that truth is a prerequisite of epistemic rationality. In particular, it is not to say that it is impossible for what is epistemically rational to be false, and likewise it is not even to say that it is impossible for most of what is epistemically rational to be false. 41

If being justified in believing that p is not a matter of believing it in such a way that it is likely to be true, as internalists typically hold, then in what does justification consist? There are two main ideas on this among internalists. One is that being justified in believing that p is a matter of whether the subject is satisfying intellectual obligations (or, on a less objective version, whether the subject believes, would believe on sufficient reflection, or is justified(!) in believing, that intellectual obligations are being satisfied) in believing that p. I will call such a conception of justification (one that is in terms of some version of the "required-forbidden-permitted" triad) a deontological conception.

Realism and the Tasks

of Epistemology

73

One is justified in being confident that p if and only if it is not the case that one ought not to be confident that p; one could not be justly reproached for being confident that p.42 The rational belief is the belief which does not violate our noetic obligations. The rational belief is the belief which, by reference to our noetic obligations, is permitted .... Rationality consists in not violating those duties concerning one's believings. 43

Another idea is that a belief is justified if the subject has sufficient (adequate) reasons, grounds, or evidence for the belief, where the criterion of adequacy is not in terms of the objective likelihood of the truth of the belief but rather in terms of whether the subject believes, would believe on reflection, or is justified (!) in believing that the evidence renders the belief likely to be true. Doxastic attitude D toward proposition p is epistemically justified for 5 at t if and only if having D toward p fits the evidence 5 has at t.44

I go through these internalist accounts of justification in order to point out that here, too, truth plays an essential role in the understanding of justification, though a different role from the one it plays in truth-conducive conceptions. In the latter there was a straightforward conceptual connection between justification of the belief and the likelihood of its truth. Here the connection is more indirect. I have already hinted at it, but let me spell it out more explicitly. On the deontological version of internalism, truth comes in by virtue of the fact that a prominent place, among intellectual obligations, is typically given to the obligation to believe what is true and to avoid believing what is false, or to believe what in one's best judgment is true and to avoid believing what in one's best judgment is false. This is spelled out in the following passage from Chisholm, in which he is explaining his notion of 'more reasonable than', in terms of which he defines terms for various degrees of justification. The variables 'p' and 'q' range over "doxastic attitudes" toward propositions-accepting, rejecting, and "withholding" (neither accepting nor rejecting). We may assume that every person is subject to a purely intellectual requirement-that of trying his best to bring it about that, for every proposition h that he considers, he accepts h if and only if h is true. One might say that this is the person's responsibility or duty qua intellectual being .... One way, then, of re-expressing the locution lip is more reasonable than q for 5 at til is

74

Realism/Antirealism and Epistemology to say this: uS is so situated at t that his intellectual requirement, his responsibility as an intellectual being, is better fulfilled by p than by qU.4S

On this view we cannot think about whether S is justified in believing that p without employing the concept of truth. For S will be justified in her belief only if she is trying her best to believe what is true. Truth comes into the second internalist account in either of two ways. According to the first, adequate evidence, reasons, or grounds for a belief are those that satisfy the believer's intuitive sense of what is sufficient to establish the likelihood of the truth of the belief, whether it is an objectively adequate indication of truth or not. According to the second, given definitive expression in Foley's Theory of Epistemic Rationality, one is justified in believing that p (Foley says "rational" rather than "justified") iff one would believe on sufficient reflection that one has adequate reasons or grounds for supposing it to be true that p.46 On both versions truth enters into the conception of rationality (justification) but indirectly, as filtered, so to say, through the subject's conception of the situation, either her actual conception or the conception she would have if she reflected sufficiently. Thus, on all the accounts of justification I have been considering, we cannot raise and answer questions of whether various beliefs are justified without employing, or presupposing, the notion of truth. And since I take the realist concept to be the concept of truth, I feel free to assume, in the absence of explicit disclaimers, that this is the concept employed by the epistemologists in question. My argument here would be much simpler if I could claim, in good conscience, that the same holds for all the prominent, not obviously misguided ways of thinking about epistemic justification, but, alas, such is not the case. In the next section I will present some theories of justification that seek to get along without bringing in reference to truth.

8. Knowledge and Justification without Truth Thus far I have only been pointing out ways in which a realist concept of truth plays a role in epistemology as typically pursued. But, of course, we are also, and even more, interested in the de jure question. Is it essential for dealing successfully with epistemological issues that realist truth play a role? What differences would it make to the enterprise if we ignored truth, a la Rorty, or if we used an epistemic concept of truth, a la Putnam? Would the epistemologist be helped or hindered in her endeavors if these

Realism and the Tasks of Epistemology

75

changes were made? In this section and the next, I will consider the effects of turning one's back on truth, and in the following section, the effects of switching to an epistemic account of truth. In each section we can, again, usefully divide the discussion into two unequal parts, dealing respectively with knowledge and justification. Beginning with the former, it seems clear that if we forswear any attention to truth, we cannot preserve anything like the standard conception of knowledge, since we would not be able to distinguish knowledge from (mere) belief by the fact that knowledge entails truth. 'Know' is the preeminent success term" in epistemology; and the most basic kind of success it involves consists in truth. I really know that the document is in that safe when I am in a position to zero in on the true answer as to where the document is. That is not to say that knowledge is merely true belief. A belief could be true by accident. I might have a completely irrational conviction that the document is in that safe and, by luck, get it right. That isn't knowledge. But, on the other hand, if the document isn't there, it is impossible that I should know that it is. No power on earth, or in heaven, could enable me to know what isn't so, since no power can override conceptual necessities. 47 Thus, abandoning the concept of truth would lead to the impossibility of any consideration of knowledge. The very subject matter would have disappeared. To be sure, when we are dealing with a particular belief, we can avoid using the term 'true' in laying out what is required for its counting as knowledge. If the belief is that the safe is open, we can, instead of requiring that it be true that the safe is open, simply require that the safe is open. That follows from the equivalence of 'it is true that the safe is open' and 'the safe is open' that is guaranteed by the T-schema. But, of course, in epistemology we want to develop a general account of knowledge and justification, rather than dealing with each candidate separately. And for that we need the concept of truth. To give a general statement of the requirement, we have to say that the belief must be true. Or we have to say something tantamount to that, such as that the belief must correspond to an actual fact. As usual, the situation is more complicated with respect to justification. The externalist can no longer make it a constraint on sufficient conditions for the justification of a belief that the satisfaction of such conditions will guarantee the likelihood of truth for that belief. That would mean the elimination of externalism as a distinctive epistemological position. One might expect this result to be welcomed by internalists, however dismaying it might be to their externalist rivals. But since, as noted above, the aim at true belief plays an important role in many internalist epistemologies II

76

Realism/Antirealism and Epistemology

as well, those positions will no longer assume the same shape. Versions in terms of intellectual obligation will have to find some replacement for the master intellectual obligation to do what one can to see to it that one believes only what is true. Versions in terms of the subject's evidence, grounds, or reasons will have to find some criterion of adequacy for evidence other than that the subject believes, would believe on reflection, ... that the evidence renders the belief likely to be true. Is it possible for these positions to find suitable replacements that would enable them to preserve their distinctive contours? Fortunately, we are not left wholly to our own devices in attempting to answer this question. There are forms of an internalist theory of justification in which truth plays no role at all, forms that were deliberately left out of account in the above survey. Let's look at a couple of examples. Earlier I quoted Ginet as characterizing justified belief as belief that is permitted by the relevant intellectual norms or standards. It is belief for which one "could not be justly reproached." In developing this idea, Ginet, unlike Chisholm, does not say anything about a master obligation to try one's best to believe what is true and avoid believing what is false. No supreme intellectual standard of any sort is identified. Instead Ginet identifies particular conditions of permitted belief when discussing various modes of justification-inferential, perceptual, memorial. There is no suggestion that these are all derivable from a fundamental aim at truth and avoidance of falsity. The specific principles are represented as standing on their own feet. We find another truth-free account of epistemic justification in John Pollock's Knowledge and Justification. 48 There he puts forward the view that the content of our most basic concepts is given by "justification conditions." To learn the meaning of a concept is certainly not to learn its "definition". It is to learn how to use it, which is to learn how to make justifiable assertions involving it. Thus it seems to me inescapable that the meaning of a concept is determined by its justification conditions. (12) Just what is necessary before we can truly say of a person that he has learned the concept of a certain kind of thing, such as "red thing" or "bird"? There is perhaps a temptation to say that one must know what would count as making "x is a bird" true-one must know the truth conditions for "x is a bird". But it does not take much reflection to see that this is not the case. Although a philosopher or a lexicographer might (although I doubt it) be able to construct a definition of "bird" that would give us such a set of truth conditions, very few ordinary speakers of English would be able to do this .... But if in fact the child did not know how to ascribe the concept and its

Realism and the Tasks of Epistemology

77

complement justifiably (i.e., he did not know how to justifiably determine whether something was a bird), this would show that he had not learned how to identify birds and so does not have the concept. Conversely, when the child has learned to judge justifiably whether a thing is a bird (i.e., he has learned to ascribe the concept and its complement to things justifiably), we are satisfied that he knows how to identify birds and so has got the concept right-he knows what a bird is. (13-15)

In other words, to possess the concept of, for example, a bird is to have the ability to make justified attributions of it; it is to have a working knowledge of the "justification conditions" of the concept. But if that is what it is to have the concept, then what the concept is can be specified by laying out those justification conditions. Those conditions constitute the content of the concept. Thus, to say that a belief that X is a bird is justified by a perceptual presentation of such-and-such a kind is simply to spell out what is involved in the concept of a bird. What constitutes a justification of a certain concept attribution is given by the constitution of that concept and neither requires nor receives any further basis. Justification stands on its own feet, without any support from considerations of truth. Unfortunately, Pollock, unlike Ginet, neglects to tell us in Knowledge and Justification what he means by 'justification' as it applies to beliefs. Without some account of the sense of 'justification' in which, he avers, (many) concepts are made up of "justification" conditions, we have no basis for judging whether on his view justification is free of entanglements with truth. But in a later book, Contemporary Theories of Knowledge, he has more to say about the concept, and what he says pushes him over into the deontological camp.49 Like Ginet he identifies it with "epistemic permissibility. " A justified belief is one that it is "epistemically permissible" to hold. Epistemic justification is a normative notion. It pertains to what you should or should not believe. (7; see also 124-25)

In fairness to these authors, I should point out that they by no means reject the concept of truth altogether. On the contrary, they go along with the customary truth condition for knowledge, and Ginet makes use of the notion of truth in explaining his fourth condition for knowledge (over and above having a true justified belief), the requirement of what he calls "external conclusiveness." But they do eschew any reference to truth in the theory of epistemic justification, and it is this in which we are presently interested.

78

Realism/Antirealism and Epistemology

9. Problems with Truth-Free Justification

The identification of two theories of justification that do not make use of the notion of truth puts us in a position to discuss the question for the sake of which we embarked on this survey of views of justification-whether it is important, or even essential, that the concept of truth playa central role in accounts of epistemic justification. I will seek to show that it is. Before getting into that, I want to acknowledge a respect in which turning one's back on truth makes things easier for the epistemologist. The job is easier because the theorist is no longer subject to truth constraints in arguing that a belief is justified or that certain conditions are sufficient for justification. If we have externalist tendencies, we are no longer required to consider whether a putative justifier renders the belief in question likely to be true. If we are internalists, we don't have to determine, for example, whether a candidate for justified believing has satisfied obligations to believe what is true and avoid believing what is false. And that relieves us of a considerable burden. Indeed, it is often said, though not by the more careful thinkers, that it is impossible to show, ascertain, or be sure that a given belief is true. The grounds given for this startling claim typically have to do with the alleged impossibility of comparing our beliefs with "transcendent" or "mind-independent" reality "outside" our thought and experience. But this is hardly better than absurd. Given the equivalence of 'p' and 'it is true that p', if I am unable to determine whether the statement that there is a computer on my desk is true, I am unable to determine whether there is a computer on my desk. Thus, properly understood, the denial that we can determine truth values for our beliefs will be plausible only to a thoroughgoing skeptic, a beast found almost exclusively in the writings of the skeptic's opponents. But even if we steer clear of these untenable allegations, the point remains that if we don't have to check for truth or likelihood of truth or the satisfaction of obligations to pursue the truth, that is one less item on the agenda. But this advantage of ignoring truth, if such it be, is more than outweighed by crucial disadvantages of that policy. I believe that we can identify three issues on which the truth-free theories fare distinctly worse than their truth-connected (internalist and externalist) competitors. First there is the question of how to distinguish epistemic justification of belief from other sorts, for example, moral and prudential justification. As we have seen, a reference to the "epistemic point of view," defined by a concern for believing what is true and avoiding believing what is false, is an obvious way to do this. What distinguishes epistemic justification is that it is valuable for the pursuit of those goals. If we eschew that way of

Realism and the Tasks of Epistemology

79

making the distinction, how can we do it? Let's see how our two paradigm truth-free theorists seek to do it. Ginet recognizes that it can be permissible for one to believe that p (one can "have a justification" for believing that p) for other than epistemically relevant reasons. He mentions a case in which "S's strong desire that R should be his trustworthy friend, and S's reasons for having that desire ... may justify 5, in a perfectly good sense, in maintaining his confidence that R would not do such a thing [lift cash from the till]."50 Instead of separating epistemic justification of belief from the others by reference to the aim at truth, Ginet extrapolates from the above example by suggesting that epistemic justification is disinterested justification, which he explains as justification "that does not involve wanting it to be the case that p."51 But this fails to distinguish epistemic justification of belief from all others, for the moral justification of a belief that p may not involve wanting the proposition that p to be true. If I am justified in believing Smith innocent of a crime because I have promised to trust him, it mayor may not be that I want Smith to be innocent. Though Pollock recognizes the need to distinguish epistemic justification from other types, he does not make very explicit just how he draws the distinction. He speaks, as I do, of the epistemic point of view ("Thus I will think of epistemic justification as being concerned with questions of the form, 'When is it permissible [from an epistemological point of view] to believe P?"')52. But he doesn't spell out the distinctive character of this point of view. What I gather from his discussion as a whole is that what makes permissibility of belief epistemic is that it stems from a certain kind of source, namely, the norms we have internalized that govern "right reasoning." "Epistemic norms are supposed to guide us in reasoning and thereby in forming beliefs. The concept of epistemic justification can be explained by explaining the nature and origin of the epistemic norms that govern our reasoning. 53 I have called this 'the reason-guiding concept of epistemic justification'. "54 In supposing that this gives a sufficient answer to the question, "What distinguishes epistemic permissibility from prudential or moral permissibility?" Pollock seems to assume that we have not also internalized moral or prudential norms that govern reasoning! But this assumption is clearly false. How else could there be moral or prudential permissibility of belief except by way of conformity with appropriate moral or prudential norms? Pollock's way of marking off epistemic permissibility works no better than Ginet's. Another suggestion for distinguishing epistemic justification from other kinds without bringing in truth would be to think of it as conducivity to some other master goal of cognition. On this point I will follow a discussion

80

Realism/Antirealism and Epistemology

in chapter 5 of Goldman's Epistemology and Cognition. Goldman distinguishes five candidates for "justificationally valuable consequences": 1. verific consequences: believing truths, not believing falsehoods

2. coherence consequences: achieving coherence in one's belief (or creedal probability) corpus

3. explanatory consequences: believing propositions that explain other believed propositions

4. pragmatic consequences: realizing one's practical, nonintellectual goals

5. biological consequences: surviving, reproducing, propagating one's genes 55 For each of these we can consider whether we can bring out what is distinctive of epistemic justification in terms of its conducivity to that kind of consequence. Candidates 4 and 5 can be eliminated right away as insufficiently discriminative. It won't distinguish epistemic from prudential justification to point out that the former is valuable for the attainment of "practical," "nonintellectual," or biological goals. As for 2, though it is a distinctively intellectual desideratum, there are decisive reasons for denying that pure coherence suffices for epistemic justification, the main reason being that there is a potential infinity of incompatible but equally coherent systems of belief. That leaves 1 and 3. Candidate 1 explicitly involves the notion of truth, and it does not take much probing of candidate 3 to find that notion under the surface. What is it to explain something? That is a difficult question and one for which, in my opinion, no illuminating, generally applicable answer has been given. But I don't need to get into that to make the present point. In whatever way we spell out what is involved in rendering something explained, it will remain the case that an adequate, successful explanation involves citing something true as providing the explanation. No matter how a set of false statements is explanatorily related to the explanandum, the statements will not constitute the explanation of it. If my explanandum is the fact that I am getting no sound from my loudspeakers, their not being hooked up to a source of signals is related to this explanandum in the right sort of way to be an explanans. If that suggestion is true, it will explain the lack of sound-but only if. We haven't lit on what really explains the defect until we have specified some true proposition(s) that is (are) related to the explanandum in the right sort of way.

Realism and the Tasks

of Epistemology

81

Thus, this survey of candidates for the basic goal of cognition, by reference to which epistemic justification could be distinguished from other types, provides no hope of any satisfactory choice that does not involve the concept of truth. Doing without truth leaves us without any way of framing an adequate concept of epistemic justification. The second difficulty has to do with criteria for the assessment of principles of justification that lay down conditions of justification for one or another sort of belief. How do we tell which principles to adopt? So long as we are working with a truth-conducivity conception of justification we have something to go on. If a requirement for condition C's justifying belief B is that C renders B significantly likely to be true, we have a lodestar that will guide us in distinguishing between genuine justification makers and impostors. At least we do if, as seems to be the case, we can often tell when conditions do render a belief probably true. If a certain type of perceptual presentation is a strong indication that there is a maple tree in front of me, then such a presentation passes at least that truth-conducivity requirement for being a justifier of that belief. And if that is all it takes to be a justifier (perhaps together with the requirement, obviously satisfied here, of being cognitively accessible to the subject), we can without more ado judge that being perceptually appeared to in that way justifies the belief that one is face to face with a maple tree. If a certain kind of noise emanating from a certain house is a reliable sign that a party is going on therein, then by observing that noise, we are thereby (truth-conducively) justified in supposing that a party is going on. And so on. But how will we make a rational, principled choice between competing claims to justificatory efficacy if we abandon all reference to truth, likelihood of truth, probability of truth, reliable indication of truth, and the like? On what can we rely in determining what justifies what? Let's see what our truth-free justification theorists have to say about this. Here is Ginet's answer: Insofar as positions directly recognizable to a person can be objectively ranked as to how strong a belief in a given proposition, p, they justify that person in having-that is, insofar as we have a concept and practice of objective justification of degrees of belief-the ultimate authority for this ranking must be the concurring judgments of reasonable, experienced people who have the notion of and an interest in the practice of rational, objective justification of degrees of belief and who give the positions in question their thoughtful consideration. 56

But this amounts to an ultimate appeal to raw intuition, however ideal the subjects of the intuitions-"ultimate" because there is no provision for

82

Realism/Antirealism and Epistemology

any check on the accuracy or validity of those intuitions. They constitute the last word from which there is no further court of appeal. If that is the best we can do, we can learn to live with it. But since a truth-conducivity conception of justification gives us something better, we are ill advised to rest content with bare, unexamined intuition in this sphere. As for Pollock, his story as to how we tell whether we have the right principles of justification is that they are determined by the concepts involved in the beliefs in question. In section 6 I presented his exposition of this position in Knowledge and Justification. The account of the structure of concepts in Contemporary Theories of Knowledge (147-48) is more elaborate and sophisticated, but the application to the basis of justification is the same. Since concepts are constituted, at least in part, by "justification conditions," it is conceptually true that the belief that a tree is in front of me is justified by what the concept of a tree (in conjunction with the other concepts involved) lays down as sufficient for the justification of such a belief. This position fares no better than Ginet's. For one thing, there are reasons to doubt that our concepts are constituted, even in part, by justification conditions. If our concept of a tree were so constituted, how could that concept be possessed by subjects that have no inkling, even in a practical, know-how sense, of epistemic justification? Moreover if, as Pollock acknowledges in the later book, a given concept contains more than a set of justification conditions, we are faced with the question of why we should suppose that the satisfaction of the justification conditions involved guarantees the justification of the application of the rest of the concept to something. Let's say that part of the concept of tree is vegetable organism, and that the concept also contains a specification of perceptual experience that would justify taking a perceived object to be a tree. That raises the question as to why we should suppose that this perceptual condition justifies the attribution of the rest of the concept-being a vegetable organism. To be sure, by hypothesis, we find them stuck together in the same concept. But what guarantees, or even gives us reason to think, that they have been advisedly combined? Is this just another appeal to the status quo, along with Ginet's appeal to the judgment of "reasonable, experienced people"? Again, we would like a deeper, more impressive rationale than this. 57 The third issue concerns why we should be interested in justification. Earlier I quoted Bonjour as saying: If epistemic justification were not conducive to truth in this way, if finding epistemically justified beliefs did not substantially increase the likelihood of

Realism and the Tasks of Epistemology

83

finding true ones, then epistemic justification would be irrelevant to our main cognitive goal and of dubious worth. It is only if we have some reason for thinking that epistemic justification constitutes a path to truth that we as cognitive beings have any motive for preferring epistemically justified beliefs to epistemically unjustified ones. 58

Ginet's and Pollock's construals of justification fall under this stricture. Ginet says that we are epistemically justified in believing that p if we have a disinterested justification for p (where its being disinterested, remember, just amounts to its not depending on our wanting it to be the case that p). But why should it be so important to us whether we have a "disinterested" justification, or whether we are "disinterestedly" permitted to believe that p? In particular, why should we suppose that justification, so construed, is of any value or importance in the quest for knowledge? So long as no connection with truth or the probability thereof is built into the concept of justification, it is difficult to see why we should aim at our beliefs' being justified rather than the reverse if we are seeking truth and knowledge. Similarly, Pollock needs some explanation of why we should care whether our beliefs conform to norms that "govern right reasoning." What is "right" about the reasoning that conforms to those norms? Why should we suppose that it helps us to get knowledge or true belief? These defects stem from the thinness of what these theorists have provided by way of a concept of epistemic justification. It is a matter of permissibility of belief, not by reason of what we want (Ginet), or it is based on the norms of belief formation that we have in fact internalized (Pollock). With no more than this to go on, it is small wonder that we are left wondering how to distinguish epistemic justification from other modes, how to assess principles of justification, and why we should regard justified belief as a pearl of great price. These are grave disabilities indeed. If this is the result of shunning truth, we had better think twice before signing onto that enterprise.

10. Epistemic Accounts of Truth and Epistemology So much for turning one's back on truth. How about the implications for epistemology of adopting some epistemic conception of truth in terms of "justification" or "rational acceptability"? At first sight one may be tempted to give the same diagnosis here as the one I gave for the "ignore truth" position. If a belief's being true just amounts to its being justified, then doesn't the standard conception of knowledge once more boil down to

84

Realism/Antirealism and Epistemology

justified belief? Adding the truth condition would just be reiterating the justification condition. And don't we, pari passu, lose the truth constraint on justification, for that would just amount to a justification condition on justification? And even Yogi Berra might not be moved to say "If it ain't justified it ain't justified." But that would be a superficial diagnosis. Our brief survey of epistemic accounts of truth revealed that the more careful of them identify truth not with the kind of justification, coherence, or rational acceptability for which we ordinarily settle in the press of affairs, but rather, as Putnam says, with rational acceptability under "epistemically ideal conditions." A true statement is one that would be rationally acceptable for a subject that had access to all relevant information and considerations, in whose reasoning there is no defect, whose cognitive operations are free from all irrational influences, and so on. This is clearly not the same, either extensionally or intensionally, as meeting ordinary conditions for justification. On a Putnam-like view of truth, the class of true propositions and the class of propositions justifiably believed by any particular person or group will diverge drastically. Hence truth does not drop out as a separate condition or constraint. However, giving an epistemic construal of truth will make a difference. Let's look first at knowledge. Here the truth requirement is that the belief would be justifiable in ideal epistemic circumstances. Now whether this has the force the truth requirement is designed to have depends on whether an epistemic conception of truth succeeds in getting at the concept (nature) of truth, and I have already argued that it does not. However, that is not what concerns us here. Our present concern is how adopting this construal would affect the epistemology of knowledge. The most important result has to do with our access to the satisfaction of the requirement. Are we better or worse placed to determine whether a belief would be justifiable in ideal epistemic circumstances than we are to determine whether it is true on a realist conception of truth? Since this issue will be crucial for the impact of the epistemic construal of truth on justification as well, I will postpone its consideration until the end of this section. Once again, justification presents a much more complex picture. When considering the effects on epistemology of turning our backs on truth, I pointed out that this wipes out externalist positions, since it prevents them from using truth-conducivity as a central constraint on justification, a stance that is definitive of externalism. A parallel situation does not obtain here. If an epistemic construal gives us something recognizable as truth (as I am assuming for the purposes of this discussion), then the externalist can preserve the distinctive contours of her position by think-

Realism and the Tasks

of Epistemology

87

true. Hence in this discussion I will confine myself to the question of the extent to which we can ascertain truth-value on each of the rival conceptions of truth. On the realist conception, our ability to determine the truth value of a proposition is highly sensitive to content. Again, this is because of the Tschema that declares 'it is true that p' to be equivalent to 'p'. This guarantees that if and only if we can tell whether or not p, we can tell whether or not it is true that p. This means that for innumerable perceptual, introspective, and memorial cases, as well as for simple empirical generalizations and predictions, we are in a strong position to determine truth value. Just because we are all very good at determining whether there is a tree directly in front of one, we are equally good at determining whether it is true that there is a tree directly in front of US. 59 In other cases our powers of discrimination are less impressive. These include propositions about the remote past or about periods for which there is scanty documentary evidence, high-level theoretical propositions in science, philosophical and religious convictions, and so on. The main point of all this is that we have a very considerable capacity to ascertain truth-value in a wide range of cases, though this capacity varies with subject matter. With an epistemic conception of truth, the situation is radically different. Continuing to focus on Putnam's version as the most plausible, there are several points to be made. First, and most fundamentally, what we have to determine to ascertain whether a given belief would be justifiable in an ideal epistemic situation belongs to a category-the application of norms, standards, or evaluative principles to particular cases-that falls on the "hard to tell" side of the above distinction between different classes of propositions. Whether a particular action is legal, constitutional, or proper; whether a particular piece of legislation is in the interests of a majority of citizens; whether a decision of a church governing body is in accord with apostolic tradition; whether what Jim did or said constitutes sexual harassment-these are all notoriously controversial matters, about which it is difficult to come to any definitive resolution to which most interested parties will agree. And so it is with epistemic justification. Of course, there are many cases on which it is easy to get agreement. There are open and shut cases in police detection in which it is clear to any disinterested observer that the police are justified in taking Robinson to be guilty. And the overwhelming majority with no stake in the matter will agree that no one is justified in supposing that Elvis is alive and well in Patagonia. But there are many controversial cases as well. For confirmation of this claim, just pick up any treatise on epistemology. Am I justified in ordinary perceptual beliefs even though I am unable effectively to

Realism and the Tasks of Epistemology

85

ing of truth-conducivity in those terms. There will still be the controversy between externalists and internalists as to whether truth-conducivity is a necessary condition of justification. What about internalist accounts of justification? The ones that make no use of the concept of truth will obviously not be affected. But what about the others-Chisholm, Foley, Bonjour, Moser, and company? Would a switch to Putnam's conception of truth make an important difference to their positions and investigations, still leaving aside for the moment the question of differential accessibility of truth? Here it will be pertinent to consider the three respects in which we saw that theories of justification that completely eschew any reference to truth are adversely affected thereby, remembering that we are now looking at those issues with a different question in mind. First, there was the issue of how to distinguish epistemic justification from other sorts. Thinking now of internalist theories that make use of the concept of truth, does it make a difference to this issue what concept of truth is used? It certainly does. Look at what happens when one tries to appeal to the aim at truth to bring out what is distinctive of epistemic justification, using an epistemic conception of truth. The trouble is that now the goal an aim at which is supposed to bring out what makes epistemic justification epistemic is itself construed in terms of epistemic justification. For when Putnam construes truth as justifiability in an ideal epistemic situation, he means, of course, epistemic justifiability, not moral, prudential, or legal justifiability. And so the account becomes circular. We make use of the notion of epistemic justification in the explanation of what makes epistemic justification epistemic. But if we try to do the job in terms of the aim at realist truth, we run into no such difficulty. I now skip to the third of the issues considered in section 9, since the point to be made there is close to the one just made. Suppose we answer the question "Why should we want our beliefs to be epistemically justified?" by saying "Because that means they will most likely be true," and use the epistemic construal of truth in that answer. In that case we say that it is important to us that our beliefs be epistemically justified because such beliefs would be likely to be justified in an ideal epistemic situation! But if there is a question as to the value of epistemically justified beliefs, how does it help to point to the value of beliefs that are justified in ideal epistemic circumstances? Let's grant that in an epistemic value scale, being epistemically justified in ideal circumstances rates higher than being epistemically justified in ordinary circumstances by ordinary standards. But the original question had to do with the value of epistemic justification in general. And it doesn't help to answer that question to say that ordinary

88

Realism/Antirealism and Epistemology

defend those beliefs? Am I justified in generalizing from the golden retrievers I have observed that golden retrievers are generally affectionate, given that I have no reason to think that my sample is representative, even though it is? And so it goes. Questions about epistemic justification are often "essentially contested questions." So far I have been drawing on general features of epistemic justification, saying nothing as yet about the special and highly unusual questions of this type to which we must address ourselves if we are to determine truth value on Putnam's construal. To say that a proposition is true, in that sense, is to say that it would be justifiable under ideal epistemic conditions. And that kind of judgment of justification poses special and severe difficulties. I said earlier that I would take it that such conditions are ideal in that all relevant evidence is readily available. 60 The clinker here is "all." To what extent are we, presently or in the foreseeable future, in a position to determine what would come out as justified on the basis of all relevant considerations? Can we make an informed judgment on what would go into that totality, even vis-a.-vis a particular belief? And if not, how can we make a sufficiently informed judgment as to whether a given belief would be ideally justifiable? We are better placed to make these assessments for some propositions than for others. For what is observable we can, pace extreme theory-relativity-of-observation views, pretty well survey what is relevant to epistemic status-the character of the sensory experience, background beliefs that bear on the likelihood of what is putatively observed, relevant features of the condition and situation of the observer, the properties of the medium, and so on. With more theoretical matters, on the other hand, it would be rash indeed to suppose that we can specify everything that is relevant to justification. To take an extreme example, how about views as to the course of events in the first one-tenth of a second after the Big Bang? With a domain of theory in as much flux as this, we constantly encounter new relevant facts or hitherto unsuspected relevance of previously known facts. If you consider this example too recherche to be significant, consider the inability of physicists in the eighteenth and nineteenth centuries to realize the relevance for Newtonian mechanics of the considerations-empirical and theoretical-that led to the development of relativity theory and quantum mechanics. These, and many other examples, bring out the point that, at least with respect to many beliefs, judgments as to what is relevant to their assessment change and develop in the course of intellectual history. And it is a truism that we cannot anticipate changes in these respects. If we could, those changes would already have come about. I believe that this line of thought can be extended to cast cold water on the idea that we are generally in a good

86

Realism/Antirealism and Epistemology

justification is usually correlated with ideal justification. There is a kind of circularity here too. We argue for the value of epistemic justification by invoking the value of a special kind of epistemic justification. Another black mark for the epistemic theorist. No parallel point can be made for the second issue from the last section-how one assesses principles that lay down conditions of justification for particular kinds of beliefs. We saw that if we abjure truth altogether, we lose a key constraint on such principles and are left with an uncritical reliance on raw intuition. But construing truth epistemically (again, if acceptable on other grounds) will not take away the truth constraint. It will only reconstrue it. Thus the epistemologist who uses an epistemic conception of truth will still have a way of assessing principles of justification. He can consider whether the satisfaction of the conditions in question is a good indication that the belief would be justifiable in epistemically ideal circumstances. If he is in a worse position than his realist colleague, it is because it is more difficult for him than for the realist to determine whether the truth constraint is satisfied. This is a question that has surfaced repeatedly in this section. It is obviously of major importance for a comparative assessment of the value of realist and epistemic conceptions of truth for epistemology, and it is high time to confront it. In considering this issue, we can abstract from the differences between the role of truth in knowledge and its role in different accounts of justification. For knowledge that p there is a straight requirement that it be true that p. On all accounts of justification I am taking seriously, the requirement is less straightforward. It is either that the conditions of justification make truth likely; or that the subject has satisfied an obligation to do his best to believe true and only true propositions; or that the belief is based on evidence that the subject believes, or would believe on reflection, is sufficiently indicative of the truth of p; or .... I can abstract from these differences because in all cases it is important for the application of a theory that we be able to ascertain the truth-value of particular propositions. Think of views in which the role of truth is very indirect. How can we tell what an obligation to believe only the truth requires of us if we have little idea of how to determine when we are believing the truth? How can we make a judgment, on however much reflection, on whether the reasons we have for believing that p render it sufficiently likely that p is true if we are in a poor position for determining, in any instance, whether p is true? Whether truth, or the probability of truth, is a straight requirement for knowledge or justification or whether the concept of truth enters into the account only indirectly, it is of central importance for the enterprise that we have the capacity to determine (frequently) whether a belief is

Realism and the Tasks of Epistemology

89

position to make judgments as to what would and would not be justifiable under ideal epistemic circumstances, at least when we get away from the empirical and rational bases of our knowledge. And for those bases we are in a good position to determine truth in a realist sense. Hence an ideal justifiability conception of truth does not, to say the least, give us any edge over a realist conception in determining truth values. Moreover, if we go deeper into the matter, things look even bleaker for the epistemic theorist. Ignoring the difficulties concerning our present historical situation just aired, let's think more generally about what is involved in determining what is relevant to the epistemic assessment of a given proposition. The historical changes in judgments of relevance we have just noted suggest that such judgments are made in light of more or less extensive background considerations and that, depending on how those are set, we get different results concerning relevance. Given the theory of general relativity, what is relevant to determining the age of the universe is radically different from what is relevant to that issue given Newtonian physics. From a psychoanalytic perspective what is relevant for deciding on the etiology of Jim's obsession is quite different from what is relevant from a behavioristic perspective. There is no unique set of relevant considerations for a given statement. It all depends on the assumptions with which we are working. And how are those assumptions to be chosen for "ideal epistemic conditions"? They would have to be the assumptions that we are most justified in accepting. But most justified relative to what further background assumptions? An infinite regress looms. What can we do to rescue the notion of an ideal epistemic situation from this fatal relativity to background assumptions? Since we have just seen that the attempt to neutralize the relativity by choosing the best justified background assumptions in each case gives rise to a viciously infinite regress, I see no alternative to an appeal to coherence. Instead of trying to pick a unique set of background assumptions case by case, we consider what total set of assumptions would yield the most coherent system. I am not suggesting that we depend on coherence to do the whole job, as we would in a pure coherence epistemology. We can take observational, memory, and introspective beliefs, as well as the intuitively self-evident, to have an initial credibility apart from high-level background assumptions, though defeasible, in principle, by various considerations. This will give us a base of prima facie justified beliefs as an anchor for whatever larger system is constructed. We can then envisage that larger system as the one whose members best cohere with each other and with the base just specified, or with as much of the base as survives defeaters. One can reasonably hope that a unique system is thereby determined from which

90

Realism/Antirealism and Epistemology

the approved background beliefs can be drawn for assessing the justificatory status of any particular belief. This is only one of various possible ways of determining what background assumptions should be used in assessing the relevance of evidence vis-a.-vis a given belief. But I am convinced that any other plausible way of doing it would have to be at least this complicated. So let's work with the suggestion just broached. This indicates that one thing we have to do to identify the considerations that are relevant to the ideal epistemic assessment of a given belief is to determine what more or less theoretical propositions would figure in a total system whose members are most coherent with each other and with empirical (observational) and rational (self-evident) bases. Ideally this would involve spelling out the full constitution of such a system, and that is obviously far beyond our powers. But even something more modest that would still provide a basis for a sound judgment as to which of alternative sets of propositions would be in the maximally coherent system would seem to be a pipe dream. How could we make an informed judgment on this without a more complete grasp of the constituents of a maximally coherent system than we are able to attain? I take this consideration to put it beyond doubt that we are in a much worse position to determine Putnamian truth than we are to determine realist truth. Even with respect to observational and other low-level beliefs, since they are inherently defeasible, we would need a considerable grasp of the (allegedly unique) maximally coherent system to determine whether it contains sufficient defeaters for a given prima facie justified observation report. Thus even in those areas where the determination of realist truth-value is unproblematic, we are in a poor position to assess Putnamian truth-value. 61 And in other areas we are, at best, in no better position than is the realist. I conclude that switching from a realist to an epistemic conception of truth greatly worsens our capacity to determine the truth-values of beliefs. This adds one more item to the case against the employment of an epistemic conception of truth in epistemology. The irony of this situation hardly needs explicit notice. The very thing that has made an epistemic account of truth seem attractive to many-its promise to remove truth from the status of something inaccessibly transcendent by making it internal to our cognitive operations and hence easily within our grasp-turns out to be one of its most serious liabilities. What has happened is that in order to develop an epistemic conception that is not desperately implausible as an account of truth, we are forced to introduce a conception of epistemically ideal circumstances that is so far removed from our actual epistemic situation as to make it somewhere between very difficult and impossible for us to make informed judgments

Realism and the Tasks

of Epistemology

91

as to what would or would not be justifiable in such circumstances. The best laid plans of mice and men! To sum up this discussion of the bearing of the disputes over alethic realism on epistemology: We have found not only that epistemology, as commonly practiced, makes heavy use of a realist conception of truth, we have also found that if the epistemologist were either to eschew any consideration of truth or to switch to an epistemic construal of truth, his attempts to make progress on epistemological issues would be greatly worsened. Alternatives to alethic realism have nothing to recommend them in the field of epistemology.

Notes 1. From the Greek word for truth, 'aletheia'. 2. In all subsequent principles that give necessary and sufficient conditions for a proposition, statement, or belief to be true, the 'Otherwise it is false' will be tacitly understood. 3. Alfred Tarski, "The Semantic Conception of Truth," Philosophy and Phenomenological Research 4 (1944): 341-75. In the reprint of this article in Readings in Philosophical Analysis, ed. H. Feigl and W. Sellars (New York: Appleton-Century-Crofts, 1949), this formula is found on p. 60. 4. This is not to be confused with a "deflationary," "redundancy," "disquotational," or "speech act" account of truth, none of which regards truth as a "genuine" property of propositions or statements. I have borrowed from Paul Horwich (Truth [Oxford: Basil Blackwell, 1990], chap. 2) the idea of taking (12) as sufficient for identifying (what I call) a realist conception of truth. 5. I am indebted to Jonathan Bennett for this line of thought. 6. For a more extended discussion of ways of formulating a realist conception of truth, see chap. 1 of my Realist Conception of Truth (Ithaca: Cornell University Press, 1996). 7. Amazingly enough, this truism has been denied in Stephen Stitch, The Fragmentation of Reason (Cambridge: MIT Press, 1990), chap. 5. 8. One reader of this paper wondered if this was supposed to rule out, for example, noncognitivist interpretations of evaluative utterances. The answer is negative. I would say that the noncognitivist is not construing evaluative utterances as statements in the sense of 'statement' intended here. 9. Colin McGinn, "An A Priori Argument for Realism," Journal of Philosophy 76 (March 1979): 113-33. 10. Thomas Reid, Essays on the Intellectual Powers of Man (Cambridge: MIT Press, 1969), 221. 11. To be sure, many idealists have opted for some other account of truth, usually a coherence account. But I am only claiming that absolute idealism (as well as other forms of idealism) is compatible with alethic realism.

92

Realism/Antirealism and Epistemology

12. That is, this is the only kind of alternative that constitutes a serious attempt to explicate the concept of truth or makes a serious attempt to say what truth is. I am leaving out of account "defiationist," "disquotational," and other views that make no attempt to treat truth as a property of propositions. 13. F. H. Bradley, Essays on Truth and Reality (Oxford: Clarendon Press, 1914), 1. 14. Bradley, Truth and Reality, 223. 15. Brand Blanshard, The Nature of Thought (London: Allen & Unwin, 1939), 2:264. 16. C. S. Peirce, "How to Make Our Ideas Clear," in Collected Papers, ed. C. Hartshorne and P. Weiss (Cambridge: Harvard University Press, 1975), 268. 17. William James, Pragmatism (1907; reprint, Cambridge: Harvard University Press, 1975), 97. 18. John Dewey, Reconstruction in Philosophy (New York: Henry Holt, 1920), 156. 19. Hilary Putnam, "Realism and Reason," in Meaning and the Moral Sciences (London: Routledge & Kegan Paul, 1978), 125. 20. Hilary Putnam, Reason, Truth, and History (Cambridge: Cambridge University Press, 1981), 55. 21. Putnam, Reason, Truth, and History, 49-50. 22. Putnam, Reason, Truth, and History, 64. 23. In fairness to the actual human being, Hilary Putnam, I will point out that he no longer holds the views expressed in these quotations. It is the Putnam of the late 1970s and early 1980s who will be discussed here. 24. Richard Rorty, Consequences of Pragmatism (Minneapolis: University of Minnesota Press, 1982), xxiv. 25. Rorty, Consequences of Pragmatism, xxv-xxvi. 26. Rorty, Consequences of Pragmatism, xxviii. 27. Rorty, Consequences of Pragmatism, xxix-xxx. 28. 'Justification' is the term of choice for most contemporary epistemologists, though some (e.g., Richard Foley) prefer 'rationality', and Alvin Plantinga has decided to go with 'warrant'. The interrelation of all these is itself a tortuous and controversial matter, especially the relation of justified belief and rational belief. This short discussion will be conducted in terms of 'justified' and its cognates. 29. Robert Audi, The Structure ofJustification (Cambridge: Cambridge University Press, 1993),300-301. 30. Marshall Swain, Reasons and Knowledge (Ithaca: Cornell University Press, 1981),99. 31. William P. Alston, Epistemic Justification: Essays in the Theory of Knowledge (Ithaca: Cornell University Press, 1989), 83-84. 32. Paul K. Moser, Empirical Justification (Dordrecht: D. Reidel, 1985),4. 33. Laurence Bonjour, The Structure of Empirical Knowledge (Cambridge: Harvard University Press, 1985), 7-8. 34. For example, Laurence Bonjour, a vigorous opponent of reliabilism and other forms of externalism, is committed to a truth-conductivity position, as the last quotation indicates.

Realism and the Tasks

of Epistemology

93

35. See Alvin Goldman, Epistemology and Cognition (Cambridge: Harvard University Press, 1986), pt. 1. 36. Roderick M. Chisholm, Theory of Knowledge, 3d ed. (Englewood Cliffs, N.J.: Prentice-Hall, 1989), 7. 37. Chisholm, Theory of Knowledge, 76. 38. Carl Ginet, Knowledge, Perception, and Memory (Dordrecht: D. Reidel, 1975), 34. 39. But not invariably. Recall the earlier quotations from Bonjour and Moser, who are both internalists in the present sense and who take justification to be essentially truth conducive. 40. Chisholm, Theory of Knowledge, 76. 41. Richard Foley, The Theory of Epistemic Rationality (Cambridge: Harvard University Press, 1987), 155. See the whole chapter that begins on this page, "Epistemic Rationality and Truth." 42. Ginet, Knowledge, Perception, and Memory, 28. 43. Nicholas Wolterstorff, "Can Belief in God Be Rational If It Has No Foundations?" in Faith and Rationality: Reason and Belief in God, ed. Alvin Plantinga and Nicholas Wolterstorff (Notre Dame, Ind.: University of Notre Dame Press, 193), 144. 44. Richard Feldman and Earl Conee, "Evidentialism," in Human Knowledge: Classical and Contemporary Approaches, ed. Paul K. Moser and Arnold vander Nat (New York: Oxford University Press, 1987),334. 45. Roderick M. Chisholm, Theory of Knowledge, 2d ed. (Englewood Cliffs, N.J.: Prentice-Hall, 1977), 14. 46. This is not exactly Foley'S formulation (see chap. 1 of The Theory ofEpistemic Rationality), but I don't believe that I am misrepresenting him. 47. One may suggest that what distinguishes knowledge from mere belief is that the former involves conclusive justification. But in order for justification to be conclusive enough by itself to convert mere belief into knowledge, it has to be strong enough to guarantee the truth of the belief. And so the concept of truth still comes into the picture. 48. John Pollock, Knowledge and Justification (Princeton: Princeton University Press, 1974). 49. John Pollock, Contemporary Theories of Knowledge (Totowa, N.J.: Rowman & Littlefield, 1986). 50. Ginet, Knowledge, Perception, and Memory, 29. 51. Ginet, Knowledge, Perception, and Memory, 29. 52. Pollock, Contemporary Theories of Knowledge, 24. 53. Pollock uses 'reasoning' in a broad sense in which it includes basing beliefs on experience. In other words, 'reasoning' is used to cover any sort of belief formation. 54. Pollock, Contemporary Theories of Knowledge, 125. 55. Goldman, Epistemology and Cognition, 98. I don't mean to suggest that Goldman is concerned here with my current problem of distinguishing epistemic from other kinds of justification. He is, rather, engaged in finding a criterion for right

94

Realism/Antirealism and Epistemology

rules of justification. However, I find his list of "justificationally valuable consequences" useful for my purpose. Note too that one might well think that justification of belief is a central goal of cognition, one could hardly bring in that goal to explain what differentiates epistemic from other justification. For if the goal is to have the necessary relevance, it will have to be epistemic justification of which we are speaking. And saying that what is distinctive of epistemic justification is that it is valuable in attaining the goal of epistemic justification could scarcely be thought to throw light on what epistemic justification is. 56. Ginet, Knowledge, Perception, and Memory, 37. 57. For a more extended discussion of difficulties with Pollock's position, see my Epistemic Justification, 39-45. 58. Bonjour, Structure of Empirical Knowledge, 8. 59. In saying this, I am assuming an extreme skepticism to be untenable. I feel warranted in assuming that here, since an extreme skepticism would cut deeply into both of the rival positions on truth. 60. It will simplify matters if, instead of thinking in terms of one all-purpose ideal epistemic situation that would contain all the evidence that is relevant to any possible belief (!), we think in terms of different situations for each proposition. That is the way Putnam seems to construe the matter. See his Representation and Reality (Cambridge: MIT Press, 1989), 115. 61. To be sure, we could go piggyback on alethic realism. If I can show that there is a computer in my office (and hence that it is true, in a realist sense, that there is a computer in my office), I might try to use that as a basis for inferring that the proposition that there is a computer in my office would be rationally acceptable in ideal epistemic circumstances. But the epistemic account of truth was supposed to replace the realist account, not to be parasitic on it.

4 Why the Theory of Knowledge Has to Be Realistic

Roderick M. Chisholm

y principal role in this volume, at least as I see it, is to try to make sure that traditional theory of knowledge has a fair hearing. And this means that I shall defend realism and oppose antirealism in the theory of knowledge. And so what I shall say also involves questions of ontology. I will try to do four things. The first is to say what realism and antirealism are, using the terms 'realism' and 'antirealism' as they are now used in philosophy. The second is to say what the theory of knowledge is in its important traditional sense, and to suggest how it should be investigated. The third thing that I will do is to defend the thesis that the theory of knowledge must be "realistic." And the fourth and last will be a criticism' of antirealism in the theory of knowledge. It was Michael Dummett, I believe, who introduced the terminology of "realism" and "antirealism" into contemporary philosophical discussion. In order not to multiply terminologies beyond necessity, I will use the words 'realism' and 'antirealism' in the general way that he proposes. 'Realism', as he points out, has been used ambiguously in philosophy, sometimes to refer to a view about the existence of universals, and sometimes to refer to a view about the perception of external physical things. It would be clearer, as he suggests, to speak of "realism concerning a given subject-matter." Thus most of us could be said to be realists concerning dogs, or the existence of dogs, and not realists concerning mermaids, or the existence of mermaids. But the question of whether there are dogs, like the question of whether there are mermaids, is a question of zoology and not a question of philosophy. A philosophical dispute about realism, as Dummett writes, may be expressed by asking whether entities of a certain sort may be found "among the ultimate constituents of reality." I would go along with this too. But my concern, unlike Dummett's, is not primarily with language. I am concerned, rather, with those nonlinguistic questions we must be able to

M

95

96

Realism/Antirealism and Epistemology

answer if we are to make any reasonable decisions about the meaning and reference of the language that we use. I concede, however, that problems about "the ultimate constituents of reality" are inseparably tied to questions about language. But to say of entities of such and such a sort that they are not included in the ultimate furniture of the world is not to make an assertion about language. In this respect realism with respect to universals is like realism with respect to dogs or unicorns.

A Realistic Ontology: Attributes and States without Propositions

A theory about "the ultimate constituents of reality" is what has traditionally been called "a theory of categories." I shall, therefore, summarize the realistic theory of categories which is presupposed by the present discussion. I assume that there are four basic types of entity: states; contingent individuals; abstracta; and necessary substance. (Whether or not there is a necessary substance, the concept of such an entity should have a place in the theory of categories.) The theory of categories that I would defend involves four dichotomies-four ways of dividing things into exclusive and exhaustive subsets. The dichotomies are these: (1) things which are contingent and things which are noncontingent or necessary; (2) contingent things which are states and contingent things which are non-states or contingent individuals; (3) contingent individuals which are boundaries and contingent individuals which are non-boundaries or contingent substances; and (4) necessary things which are abstracta and that which is necessary substance. At the present time, many philosophers, and indeed most philosophers who write about language, assume that there is still another type of category-one that has no place in the present scheme. They assume that there are "propositions" to be found among the ultimate furniture of the universe. "Surely," they may reason, "this is the simplest approach. There are those entities that are propositions and everyone of them is either true or false. Therefore, if you believe one of them, then your belief is true if and only if the proposition that you believe is true, and otherwise your belief is false." What are these strange entities? Let us suppose that there is someone who believes, falsely, that there is a devil living on the planet Venus. This means, according to the propositional theory, that there really is an entity which could be described as "that there is a devil living on Venus" or as "there being a devil who lives on Venus." What is the makeup of this entity? Is it a kind of whole having an

Why the Theory of Knowledge Has to Be Realistic

97

actual devil as one of its parts or constituents and having Venus as another part or constituent? Would this mean that the diabolical constituent exists in some place other than Venus? Or, if there are no devils at all, then must we say that the proposition has a nonexistent devil as one of its constituents? Surely we need a very good reason before countenancing such entities. But I believe that no good reason is at hand.

What Is the Theory of Knowledge?

As I have said, I am discussing the fundamental sense of the theory of knowledge-its traditional sense. The theory of knowledge is that Socratic inquiry suggested by the following questions: (1) "What can I know?"; (2) "How can I distinguish beliefs that are reasonable for me from beliefs that are not?"; (3) "How can I tell whether I have a good reason for a belief?"; and (4) "What can I do to replace unreasonable beliefs with reasonable beliefs about the same subject matter, and to replace beliefs that are less reasonable with beliefs that are more reasonable?" Let us recall the first of Descartes's Meditations, which is entitled "Of the Things which may be Brought into the Sphere of the Doubtful." Descartes begins with this passage: It is now some years since I detected how many were the false beliefs that I

had from my earliest youth admitted to be true, and how doubtful was everything I had since constructed on this basis; and from that time I was convinced that I must once for all seriously undertake to rid myself of all the opinions which I had formerly accepted, and commence to build anew from the foundation, if I wanted to establish any firm and permanent structure in the sciences.

There are three different stages in this Socratic enterprise. We may call the first stage that of "methodological doubt," the second that of "finding a foundation," and the third that of "the road back." First we consider the concepts that are essential to the theory of knowledge. Then we consider each of the stages of that inquiry. Such questions do not interest many philosophers. But someone should take them seriously.

States as the Bearers of Epistemic Attributes

Those concepts that are strictly epistemic pertain to the justification of believing. We will make use of the locution, 'A is more reasonable for 5

98

Realism/Antirealism and Epistemology

than B', where the letters' A' and 'B' may be replaced by such expressions as 'Believing that there are stones' and 'Not believing that there are unicorns'. It is natural to interpret such expressions as designating states of the believer, but we will take them to refer to belief properties. Our ontology has no place for states which are such that nothing is in those states; but it does have a place for properties which are such that nothing exemplifies those properties. Our interpretation, then, enables us to say that, if believing that there are stones is more reasonable for 5 than believing that there are unicorns, then there is something which is more reasonable for 5 than believing that there are unicorns. We now turn to what may be called the epistemic hierarchy. This is a series of epistemic categories that define the levels of reasonableness that a belief may have for a particular person at a particular time. The faintest epistemic praise that we can give to a belief that a particular person has at a particular time is to say that the belief is probable for that person at that time; the highest praise is to say that it is certain for that person at that time; and there are three levels in between. The lowest positive level, then, is that of probability: D1

It is probable for 5 that p = df' Believing-that-p is more reasonable for 5 than believing-that-not-p

This concept of probability is the traditional one, going back to Cicero and to Greek skepticism. One of the most difficult epistemological problems of the general theory of probability is that of being dear about the relations between this epistemic concept of probability and the statistical concept of probability. The next three epistemic categories are these: D2

Believing-that-p is (epistemically) acceptable for 5 = df' Not-believing that p is not more reasonable for 5 than believing-that-p

D3

It is beyond reasonable doubt for 5 that p = dt' Believing-that-p is more reasonable for 5 than not-believing-that-p

D4

It is evident for 5 that p =

dt' (1) It is beyond reasonable doubt for 5 that p; and (2) resisting the temptation not to believe-that-p is more reasonable for 5 than not resisting that temptation

The sense of 'evident' here characterized is the one that is essential to the concept of knowing: if a subject 5 knows-that-p, then it is evident for 5 that p. To appreciate the difficulties of the theory of knowledge, one must real-

Why the Theory of Know/edge Has to Be Realistic

99

ize that "it is evident for 5 that p" does not imply "It is true that p." The highest category in the epistemic hierarchy is that of certainty: D5

p = df' (1) It is evident for 5 that p; and (2) there is no belief such that having that belief is more reasonable for 5 than believing-that-p

It is certain for 5 that

The categories just listed form a hierarchy in that each is included in the one listed just below it. We next consider what it is for one belief to provide evidence for another. In order to understand this concept, we make use of a relation that is considerably stronger than logical implication. It is that of doxastic entailment. D6

Believing-that-p entails believing-that-q = df' Believing-that-p is necessarily such that whoever exemplifies it and considers it exemplifies believing-that-q

If believing-that-p entails believing-that-q, then it is logically necessary that if p then q. But the converse of what we have just said is not true. It is logically necessary, for example, that, if there are stones, then 12 x 13 = 156. But believing that there are stones does not entail believing that 12 x 13 = 156. We shall now define the concept that may be expressed by saying "p makes q probable for the person 5," or "p confirms q for 5." In defining this, we follow a suggestion made by Carnap. Using the terminology of propositions, he had said that e confirms h provided that h is probable for anyone for whom e is "the total evidence." Without using the term 'proposition', we shall define the concept of total evidence in the following way:

D7

5's believing-that-p is equivalent to 5's acceptance of his total evidence = df' 5's believing-that-p entails every believing that is evident for 5 and entails no believing that is not evident for 5

D8

p-being-evident for 5 tends to make it probable for 5 that q = df' Believing-that-p is (1) evident for 5 and (2) necessarily such that, for every x, if believing-that-p is equivalent to x's total evidence, then it is probable for x that q

And now we move from tending-to-make-probable to making-probable. The latter concept enables us to say what it is for one believing to confirm another for a given believer.

100

09

Realism/Antirealism and Epistemology

p makes q probable (p confirms q) for 5 = df. (1) P tends to make q probable for 5; and (2) for every r, if p&r does not tend to make q probable for 5, then there is a w such that p&r&w does tend to make q probable for 5

Analogues of 08 and 09 may be formulated for the other epistemic categories here introduced-thus for making it acceptable, making it beyond reasonable doubt, and making it evident; one has only to replace 'probable' by the relevant epistemic concept. We turn now to the three stages of the theory of knowledge.

1. Methodological Doubt and Its Presuppositions The first of the three stages of the theory of knowledge is preliminary and negative-it is that of ridding oneself of everything that has been shown to be false or that appears to be doubtful. This stage may appeal to the antirealist. But it has realistic presuppositions. If we can ask and try to answer such questions as, "What do I know?" and "What is it reasonable for me to believe?" then we understand what it is to have a reasonable belief. And we can also understand what it is for one belief to be more reasonable than another. Making use of the latter concept, we can distinguish different degrees or levels of reasonableness that we have distinguished.

2. Finding a Foundation Traditionally, one has distinguished two types of epistemic certainty. The first is the a posteriori certainty that we can have with respect to some of our conscious states-some of those states that are either intentional or sensory. And the second is the a priori certainty exemplified by a subset of those beliefs that are necessarily true. An examination of these two types of certainty will make clear some of the respects in which traditional theory of knowledge may be said to be realistic. Examples of intentional states are: believing, thinking, considering, wondering, fearing, hoping, liking, disliking, and conceiving or grasping. If you are in such a state, then it is certain for you that you are in that state. Sensing, or being appeared to, is an example of a sensory state. If you are sensing, then there is a way of being appeared to which is such that it is certain for you that you are appeared to in that way. To be sure, from the fact that you are now sensing in the way in which Edward is now sensing, it does not follow that it is certain for you that you are now sensing

Why the Theory of Know/edge Has

to

Be Realistic

101

in the way that Edward is sensing. But we can say that you are sensing in a way which is such that (a) Edward is now sensing in that way and (b) it is certain for you that you are now sensing in that way. A priori certainty pertains to what have traditionally been called axioms. It is sometimes said that an axiom is "a proposition which is such that (1) it is true and (2) whoever understands it can see that it is true." We may put the point in our terminology by saying that the property of believing-that-p is an axiom provided it is necessarily such that, for every x, if x considers everything that is entailed by believing-that-p, then it is certain for 5 that p. I now list several epistemic principles. The first describes the type of foundation that is required by an adequate theory of knowledge. PI

(a) If 5 is the subject of an intentional state, then it is certain for 5 that he or she is in that state. (b) If 5 is being appeared to, then there is a way of being appeared to which is such that it is certain for 5 that he or she is being appeared to in that way. And (c) if believing-that-p is axiomatic and such that 5 considers everything that it entails, then it is certain for 5 that p

The principle tells us, with respect to certain types of states, that when you are in any such state, you can be certain that you are in that state.

3. The Road Back I will now set forth one way of dealing with the problems of what I have called "the road back." Otto Neurath has compared the situation that we find ourselves in to that of a sailor who needs to repair his ship at sea. If the sailor were to throw all the doubtful parts of the ship overboard, hoping thereby to build the ship anew, he might well be sunk. What he should do, obviously, is to make the most of what is at his disposal, however flimsy and inadequate that material may seem. We should begin, similarly, with the set of beliefs that we have (where else could we begin?) and then try to refine upon them and make the set more epistemically respectable. Having some faith in ourselves, then, we start out with our native common sense-with what we find ourselves believing. And we assume that the mere fact that we find ourselves inclined to believe one thing rather than another is itself a provisional-or prima facie-justification for believing the one thing rather than the other. "So far, so good," we decide.

102

Realism/Antirealism and Epistemology

Our second epistemic principle, then, concerns this provisional justification. It is a principle about the epistemic concept of probability: P2

If 5 is trying to reconstruct and improve 5' s set of beliefs, if 5 believes that p, and if 5 has not yet considered the logical relations of this belief to 5' s other beliefs, then it is probable for 5 that p

Our commonsensism is, to use C. S. Peirce's phrase, a "critical commonsensism." Although we start out with a mass of uncriticized beliefs, we will refine and improve it. One way of improving it is to sift it down, so to speak, and try to cast away the things that shouldn't be there. To take these further steps, we make use of the a priori knowledge already referred to. As rational beings, we can know, by contemplating various beliefs, that some of them logically imply others, some contradict others, some are such that they would inductively confirm others, and some are such that they would inductively disconfirm others. If, now, we find a pair of beliefs that contradict each other, we will reject at least one of them. Suppose we find ourselves believing, say (1) that all red-haired people are quick-tempered, (2) that Kevin is red-haired, and (3) that Kevin is not quick-tempered. Then we will give up at least one of these beliefs. If, after removing the contradictions that we find among our system of beliefs, we conclude the whole set that remains tends to disconfirm some of its members, then we will reject the ones that we believe to be disconfirmed. This sifting procedure enables us to find ways in which the provisional commonsense set of beliefs with which we began could be transformed into an epistemically more respectable set of beliefs. We next note how our foundation of certainty may be supplemented by a set of beliefs that are epistemically acceptable: P3

If 5 removes the contradictions that 5 believes that he or she finds in the set of beliefs that are probable for 5, and if 5 further removes the beliefs such that their negations are made probable by the conjunction of beliefs that are probable for 5, then the beliefs that remain are acceptable for 5

To move up a further step and to see how some of our beliefs about substances other than ourselves could be beyond reasonable doubt, we appeal to two further types of familiar experience. These experiencescall them, for short, "perceptual experiences" and "memory experiences" -provide us with what is essential for understanding perception

Why the Theory of Knowledge Has to Be Realistic

103

and memory. These are the experiences in terms of which we would eventually define two of the normative intentional concepts that are essential to the theory of knowledge. The first technical expression is an abbreviation of 'correctly taking':

5 correctly takes something to be F, if and only if: (1) there is an F that is appearing A to 5; (2) it is certain for 5 that he or she is being appeared A to; and (3) 5 believes that he or she is appeared A to by just one thing and that that thing is F We may put our principle about correctly taking" this way: II

P4

If (1) there is something that is appearing A to 5, if (2) it is certain for 5 that 5 is being appeared A to, and if (3) the belief that 5 is appeared A to by an F is acceptable for 5, then it is beyond reasonable doubt for 5 that he or she correctly takes something to be anF

The principle just formulated is not one telling us how the nature of the appearance may signify the nature of what it is that appears. It does not tell us how to "read off" the nature of the world from the nature of appearances. It tells us how an intentional act, which consists in making a judgment about the appearance, may signify something about what it is that appears. One may object: "But what if it is evident to you that most of the things that look like horses at the particular place where you happen to be are not horses? In such a case, the general principle would be false." The objection does not apply when we are setting out on the road back. We have not yet reached the stage where we can appeal to such evident beliefs. (The airplane pilot who is taking off does not have all the options that are available to him when he is air-borne and in flight.) And in conducting our Socratic inquiry, we are attempting to withhold our beliefs about what mayor may not be evident to us. What has been said about the relation between perceiving there to be a so-and-so and correctly taking may also be said, mutatis mutandis, about

recalling. We now introduce an abbreviation of our second technical expression-' recalling': 5 recalls being F, if and only if: (1) 5 was F; (2) 5 believes that 5 was F; and (3) believing that 5 was F is acceptable for 5

104

Realism/Antirealism and Epistemology

We may now move from the level of epistemic acceptability to that of

being beyond reasonable doubt: P5

If 5 recalls being F and if the belief that 5 was F is acceptable for 5, then it is beyond reasonable doubt for 5 that he or she was F

Finally, by appealing to the concept of concurrence, or mutual support, we may ascend to evident beliefs about contingent substances other than ourselves: P6

If 5 has a set of beliefs that are related by mutual support (any one of them is made probable by the conjunction of the others), and if each such belief is beyond reasonable doubt for 5, then the beliefs in the set that satisfy principles P4 and P5 are evident for 5

The Theory of Knowledge as Realistic

Let us now consider the realistic implications of what we have been saying. Consider any conscious property-say, sensing, judging, wondering, wishing, or hoping. What kinds of things could have such a property? If we can grasp the nature of such properties, and it is quite clear that we can, then we can see that they are properties that can be exemplified only by individual things or substances. Judging, wondering, wishing, hoping cannot possibly be properties of states of things, or properties of events. And they cannot be properties of abstract objects-such things as properties, numbers, and relations. You can hope for rain, but no state or event or number or property or relation can hope for rain. In other words, the fact that a given psychological property is exemplified-the fact, say, that the property of hoping for rain is exemplifiedlogically implies that there is an individual substance that has that property. This is a fact about the property itself: the property of hoping for rain is necessarily such that the only things that can have it are individual things-and analogously for other psychological properties. It was typical of British empiricism, particularly that of Hume, to suppose that consciousness is essentially sensible. The objects of consciousness were thought to be primarily such objects as sensations and their imagined or dreamed counterparts. Even if intentional phenomena are always accompanied by sensible or sensational phenomena, they are not themselves sensational or sensible phenomena. And the presence of cer-

Why the Theory of Know/edge Has to Be Realistic

105

tain intentional attitudes is at least as certain and indubitable for us as the presence of our sensations. If I make a certain judgment or ask myself a certain question, then I can know directly and immediately that I make that judgment or ask that question. (This is not to say, of course, that every intentional attitude may be the object of such certainty. Presumably there is a sense in which you may be said to like or to dislike a certain thing without realizing that you like or dislike that thing.) If I can know directly and immediately that I am making a certain judgment, then, I can know what it is to make such a judgment. And if I know what it is to make a judgment, then in making the judgment I can know directly and immediately that there is a certain individual thing-namely, the one who makes the judgment. And I, of course, am the one who makes my judgments and does my thinking. The same is true, obviously, of my other intentional activities-such activities as wondering, fearing, hoping, desiring, considering, liking and disliking. We may single out three different phases of this situation: (1) I can know that I hope for rain; (2) as a rational being, I can conceive what it is to hope for rain; and (3), in so doing, I can see that the only type of entity that can have the property of hoping for rain is an individual thing or substance. The epistemologist, then, should be a realist with respect to substance. And if what I have said about the experiences I have called "taking" and "recalling" is correct, the epistemologist should be a realist with respect to the objects of perception and memory. There are external physical objects which are perceived by us and there were states of ourselves and of other things which are remembered by us. The epistemologist will also be a realist with respect to states of things. He will assume that, for every x and every property P and every property Q, there is the state of x-being-P, if and only if x exemplifies P, and there is the state of x-being-both-P-and-Q, if and only if x exemplifies P and x exemplifies Q. Although he assumes that there are properties that are not exemplified, he need not assume that there are states which are such that nothing is in those states. Nor need he assume that there are those entities that are called "propositions" (sometimes "singular propositions"). So much, then, for the positive case for realism in the theory of knowledge.

The Case for Antirealism in the Theory of Knowledge

What is there to be said hr antirealism in the theory of knowledge? Nietzsche seems to have held that all claims to such knowledge are

106

Realism/Antirealism and Epistemology

false, their objects are fictions, and they have no basis in reality and make no reference to reality. How would one go about defending such a view? I shall consider three positive claims that may be considered grounds for antirealism. First, we found that when Descartes is speaking about the first and negative phase of his inquiry, there are two types of charge that he may direct upon a belief. The first, which is the more serious, is that the belief is a member of a set of beliefs which is such that most of its members are false. To what extent does this finding support the case of the antirealist? The antirealist may say: "There are good reasons, after all, for supposing that we cannot know the kinds of things that most people think they can know." In support of this contention, he may provide an impressive amount of information about human fallibility. Does the possession of such information mean that it is not reasonable for us to assume that we can answer our epistemological questions? If the antirealist has information about human fallibility, that is to say, if he knows something about it, then, obviously, he should take such information into account when he asks about what it is that he can know. Or, if he just has some slight evidence pointing to such a conclusion (say, one experiment here and another experiment there), then he should consider the knowledge that is involved in obtaining this slight evidence. He should consider, for example, the knowledge involved in drawing the conclusion that this experiment went this way and that that experiment went that way. The fact that he has such knowledge, if he does have it, makes clear that, despite the indications of fallibility, whatever these indications may be, we do have knowledge about ourselves and about the people and material objects around us. Of course, the antirealist may not be as cooperative as this. He may tell us various things about human fallibility but without professing to be justified in thinking that any of these things is really true. What should we say to him in this case? We may say, trivially, that we have ground for doubting only if we have ground for doubting. But how could the fact that there is something we are not justified in believing provide us with a ground for antirealism? I turn, then, to the second of the three considerations to which the antirealist may appeal. This is illustrated by the various versions of "the malicious demon argument" for skepticism with respect to the external world. One will point out to us that Descartes's supposition of a malicious demon is logically possible. This means that there is no logical contradiction involved in assuming that all my present experiences are caused by such an evil spirit. Therefore, the antirealist may conclude, it is not rea-

Why the Theory of Knowledge Has to Be Realistic

107

sonable for me to have the beliefs that I do have about the things around me. The type of antirealist we are now envisaging will say to us: "It is logically possible for one to have the kind of experiences that you are now having and yet to be deceived in your beliefs about the external world. Possibly you are just a brain in a vat of nutrients. Therefore you are not justified in believing, as you do, that you exist outside the vat of nutrients." Is his argument a good one? Not really. The conclusion does not follow from the premises. As Thomas Reid would say, "it taketh no great wit" to see that no logical contradiction is involved in affirming the premises and at the same time denying the conclusion. Does the antirealist have another move with respect to this type of argument? The obvious possibility would be for him to make explicit a third premise, one which will make his derivation formally valid. But what would this third premise be? And how would the antirealist defend it? We should ask him, of course, "What is your ground for affirming that philosophical premise? What justification do you have for thinking that your complex philosophical proposition is more reasonable for me than the belief that I am surrounded by familiar physical things?" It is difficult to think of what his answer might be. I mention, finally, a third type of consideration that may appeal to the antirealist. It is essential to realize that our epistemic concepts are normative concepts. They pertain to the justification of belief. They are like the concepts of moral philosophy in that they are used in formulating principles about what one is required-or not required-to do. Such principles have been called "normative supervenience principles." A simple form of ethical hedonism would be an example. One says: "If an act is such that performing it would result in more pleasure and less pain than would any other act that a person 5 could perform, then 5 ought to perform that act." The consequent of such a principle contains normative terms and the antecedent does not. In saying that the consequent "supervenes upon" the antecedent, one means, in part, that the principle is necessary. We may say in the simplest cases that necessarily anyone who satisfies the conditions of the antecedent is in the moral situation referred to by the consequent. The third form of antirealism involves the repudiation of such concepts. What are we to say to the philosopher who rejects the normative concepts of the theory of knowledge and contends that our normative epistemic statements are neither true nor false? For one thing we could ask him how he would deal with the traditional questions without such concepts. More important, however, we should ask him what his ground is for repudiating them. Once again, we should look to the reasons, such as they

108

Realism/Antirealism and Epistemology

are, for the doctrine that we are confronted with. And then we should try to find out, if we can, whether the realism that we want to affirm is more reasonable for us than is the doctrine that the antirealist would have us affirm. We should proceed, then, in the way we would proceed if we were told with respect to other types of statements that we affirm or deny that such statements are actually neither true nor false. How would we have dealt with the early positivistic contention that statements that are not about sense data are neither true nor false? And how do we deal with the contention that statements that are about sense data are neither true nor false? We treat them pretty much in the same way that we would deal with the contention that statements ostensibly about the past-or statements ostensibly about the future-are neither true nor false. Our question was: "Are there positive reasons for being antirealistic about the possibility of succeeding in the epistemic enterprise?" I feel entirely confident that the reasons we have found are not very good ones.

5 Indeterminism and Antirealism

Donald Davidson

ntirealism is a manifestation of the irrepressible urge in Western philosophy to ensure that whatever is real can be known. Antirealism attempts to achieve this by reading out of existence whatever it decrees lies beyond the scope of human knowledge. So Parmenides should be counted an early and extreme antirealist, since he declared that the globular, homogeneous, changeless One was all that was real, on the ground that it was the only possible object of knowledge. Plato must, of course, be counted an antirealist, since he held the physical world, and all in it, is less than real because it cannot be known. Most reductive "isms" I see as forms of antirealism: idealism, pragmatism, empiricism, materialism, behaviorism, verificationism. Each tries to trim reality down to fit within its epistemology. Each of these positions offers consolations: ordinary objects like tables and dinnerware are real, we are told, but exist only in the mind; physical objects are nothing but permanent possibilities of sensation; mental states are nothing but patterns of behavior; intentional phenomena are nothing but physical events and objects, etc. We are allowed the terminology of our old ontology as long as we agree to accept only what can be cobbled together out of entities or experiences we can for certain know. But these sops to skepticism should not deceive us; antirealisms remain sour-grape philosophies. Their motto is: if you can't grasp the grapes (in some approved sense), they aren't just sour, they were never there in the first place. For reasons I shall mention presently, some antirealisms are better expressed in terms of epistemic limitations on the concept of truth. Thus it may be held that when our epistemic faculties are deficient with respect to determining the truth or falsity of some sentence, we should rule that the sentence has no truth value, or that we should employ some reduced sense of truth. The outcome is the same: the real or the true is cut

A

109

110

Realism/Antirealism and Epistemology

down to the size of a favored form of knowledge. I do not see realism as the alternative to antirealisms. We can reject one or another version of antirealism on the grounds that the arguments for it fail without having to endorse some vague position called realism. One common characterization of realism is this: there is something in or about the world that makes our utterances or assertions or thoughts true when they are true, whether or not we have the power to determine their truth. Most criticisms of this formulation fasten, understandably, on the difficulty in spelling out what 'power' and 'determine' mean in this context. But I don't think we need to get in this deep, since I do not think we understand the prior claim that there is something in or about the world that makes our thoughts or assertions true when they are true. The fact is that no one has ever been able to say in a nontrivial way what sort of "thing" it is that makes a sentence (or other truth bearer) true. Some sentences or utterances or beliefs are true, and there are a lot of things in the world, but it explains nothing to say that the things "make" the truth vehicles true. The result is that realists are left holding the concept of truth, but can't explain how reality accounts for it. The failure of correspondence theories to give substance to the concept of truth at the same time demonstrates the vacuity of characterizations of realism in terms of correspondence. In this paper I want to discuss a special brand of antirealists, those who throw doubt on the "reality" of mental states and events in the light of the indeterminacy of translation or interpretation. I am equally concerned with those who argue that if we accept the indeterminacy of interpretation, then we must be in doubt about the status of propositional attitudes-or at least the status of attributions of propositional attitudes. Since I accept the thesis of the indeterminacy of interpretation, I would be aggrieved to find that it entails some sort of antirealism. But I do not think it does, as I shall try to explain. I neither think that indeterminism shows that propositional attitudes are less than fully real (whatever that may mean) nor that we must modify the concept of truth when we come to talk of propositional attitudes. In other words, many of our beliefs and statements about what people believe, intend, desire, and hope for are true, and they are true because people have those attitudes. Three considerations seem to stand in the way of accepting this thesis. The first is a form of scientism. The propositional attitudes do not seem suited to incorporation into a unified scientific view of the world. Thus Quine has said that "the essentially dramatic idiom of the propositional attitudes" has no place in serious science, l or, again, that our mentalistic vocabulary is "stubbornly at variance with scientific patterns." I think

Indeterminism and Antirealism

III

Quine is right in holding that the mentalistic vocabulary cannot be reduced to or incorporated into the vocabulary of physics, or of any of the other "hard" sciences. One reason for this irreducibility springs from the fact that any correct account of an agent's beliefs and other attitudes must take account of the normative element inherent in cognitive contents. One attitude has logical relations with others; these relations, though distorted and deranged by the limits of our powers, nevertheless serve to locate, and thus identify, the contents of our thoughts. When we treat the world as mindless, as in the natural sciences, nothing corresponds to this dimension of the mental. A less often noticed property that sets our talk of mental states apart from the vocabulary of the advanced sciences is dependence on causal concepts. Ordinary physical talk, like psychological talk, is full of causal concepts: the notion of a catalyst is as causal as the notion of an intentional action. The difference lies in the promise, intrinsic to physics but irrelevant to psychology, that the causal concept can, with time and research, be supplanted by an account of the mechanism which will explain what the clumsy causal notion merely finessed. The laws of physics may, if we please, be called causal. The point is that they do not employ causal concepts. The concentration of psychology on the causal role of reasons rules out any hope that the basic mental concepts can be fitted into a closed system of laws. There is one more (much debated) consideration which militates against the nomological or definitional reduction of mental concepts to those of physics: the fact that propositional attitudes and related events and states are in part identified in terms of their causal and other relations to events extraneous in time and place to the agent they characterize. Thus it is held, correctly in my opinion, that the history of an individual's learning and use of words and concepts is in central cases necessarily a factor in determining what the words mean and the contents of the concepts. I also think interpersonal communication plays a necessary role in the possibility and nature of thought. If externalisms of these sorts are indeed dominant and unavoidable features of the mental, the impossibility of incorporating psychology into a unified scientific theory of the world is clear. These remarks are no more than reminders of why many philosophers, including me, are convinced that psychology, as long as it is taken to include the concepts of intentional action, belief, perception, and the affective attitudes, cannot be made part of physics or any other "natural" science. The question here is not whether we are right in this opinion, but whether, if we are, we should question the ontological status of the propositional attitudes. Quine has said that he agrees with my token-token version of monism,

112

Realism/Antirealism and Epistemology

which holds that each object or event identified in mental terms is identical with an object or event identifiable in physical terms, though the classes staked out by the mental vocabulary cannot be equated by definition or strict laws with classes definable in the physical vocabulary. This position, which I call anomalous monism, comes close to reconciling most, though not all, of Quine's declarations about propositional attitudes. Anomalous monism makes sense of the claim that attitudes are dispositions to behave in certain ways, which are in turn physiological states, which finally are physical states, as well as the claim that intentional descriptions are not reducible to behavioral or physical descriptions and so are not suited to incorporation into any strict system of laws. The mental vocabulary is practical and indispensable, but it is not made for the most serious science. Anomalous monism does not suggest that mental events and states are merely projected by the attributor onto an agent; on the contrary, it holds that mental events are as real as physical events, being identical with them, and attributions of states are as objective. Quine's descriptions of attitude attributions as dramatic portrayals does not imply that there is nothing to portray. The fact that the mental vocabulary is not fit for inclusion in sciences like physics or physiology cannot in itself be taken as impugning the reality of the states, events, and objects it is used to describe. A perfected physics must comprehend every object and event, but this is an ontological and nomological requirement that defines the aim of physics; it says nothing about the interests that may demand other ways of characterizing things. I have not yet touched directly upon the issue of the indeterminacy of translation (or interpretation, as I prefer to call it). If the thesis of indeterminacy holds, doesn't this imply that propositional attitudes are less than real? Quine may be thought to hint as much when he says, "Brentano's thesis of the irreducibility of intentional idioms is of a piece with the thesis of indeterminacy of translation," since "[to] accept intentional usage at face value is ... to postulate translation relations as somehow objectively valid though indeterminate in principle relative to the totality of speech dispositions." 2 Doesn't this say that accepting the mental idiom means postulating something for which there is no empirical evidence-something that we therefore have no reason to believe exists? In any case, whether or not Quine's views entail that there is something unreal about the propositional attitudes, it is an idea common to a number of philosophers that indeterminism does undermine the reality of mental states. Fodor and others have thought this constitutes a reductio of indeterminism. Daniel Dennett, on the other hand, endorses indetermin-

Indeterminism and Antirealism

113

ism, but agrees that it subtracts from the reality of mental states. I argue against the inference. I see no way around indeterminism, but think it leaves the reality of the mental untouched. First, let me defend the passages from Quine that I just quoted, and that seem to aver that 'intentional usage' has no empirical content. There is a confusion here that may in part be Quine's fault, but it is easy to resolve. If by 'intentional usage' Quine means talk of meanings and propositions as hypostatized by philosophers, it is true that he rejects these as without empirical or explanatory content. But it does not follow that ordinary attributions of attitudes, including interpretations of speech, are empty. Indeed, how could they be if mental talk is, as Quine allows, something we cannot do without? There are, I think, two things that above all tempt us to suppose that there cannot be any indeterminacy about what we mean by our words or what we think. One is first person authority, the fact that there is a presumption that we know what we think in a way no one else can. The other is engendered by the semantics of the sentences that are used to attribute attitudes. Let me take the second point first, and let me stick to belief sentences for simplicity. It is obvious that there is a potential infinity in the number of beliefs that might be assigned to someone, and it is clear that there is a potential infinity of sentences available, for we can create a wellformed sentence by putting any of an infinity of sentences after the words 'Agnes believes that .. .' But the only viable semantic devices we have available to accommodate this situation involve taking 'believes' (or 'believes that') as a relational verb, and treating what follows as a singular term or description. The singular terms can't all be unstructured proper names; there aren't enough such. So if what follows 'believes' or 'believes that' is a singular term, a demonstrative device must be at work; otherwise we must have descriptions. But in either case, there must be an infinite store of entities to be picked out. Once we have picked the appropriate entity, though, we have, as attitude attributors, assigned a content to the belief. It is natural, perhaps, to suppose that the entity we refer to in order to specify the content of a belief is the object of the belief, an entity that the believer entertains or grasps, the thing the believer believes. Given the relational semantics of sentences about beliefs (and, of course, the other attitudes), and the natural assumption that the entity that identifies the belief is the object grasped by the believer, we are virtually forced to conclude that the entity designated must be unique. For if more than one entity would do equally well in specifying the same state of belief, which entity should we say the believer grasps, or how can we say that the believer knows what he believes?

114

Realism/Antirealism and Epistemology

This conundrum can be solved by giving up the idea of treating the grammatical objects in belief sentences as terms that name psychologically real objects, objects known to or entertained by or grasped by the believer. The only object required for the existence of a belief is a believer. Having a belief is not like having a favorite cat. It is being in a state, and being in a state does not require that there be an entity called a state that one is in. All that is necessary for the truth of an attitude attribution is that the predicate employed be true of the person with the attitude. Dummett deserves credit for characterizing antirealism in terms of truth rather than reference, in terms of sentences rather than names or descriptions. For the real issue about sentences that attribute mental states is not ontological; beliefs are not entities, nor do the "objects of belief" have to be objects. The real issue is whether or not attributions of attitudes are objectively true or false. 3 I see the attribution of attitudes as analogous in many ways to the measurement of various magnitudes. We can assign numbers to keep track of the sizes, weights, and speeds of objects, provided the objects exhibit a pattern of the appropriate kind. We do not suppose there are empirical entities called weights or sizes or speeds that objects have. As Carnap pointed out long ago, we should not think of the sentence 'This box weighs eight pounds' as identifying two entities called 'the weight of this box' and 'eight pounds', but rather as identifying the weight of the box in pounds with the number eight. Thus the ontology required consists of the objects that have weights, and the numbers. The numbers are not part of the weighty objects; they do not belong to the empirical world but to us, who need them in order to keep track of certain relations among objects. In the same way, the entities to which we relate thinkers when we attribute beliefs and other propositional attitudes to them are not in the thinkers-not even in their minds, or before their minds. Having a belief is just exemplifying a property, having a certain predicate true of one; but in order to have enough predicates for all the beliefs we may wish to distinguish, we must construct the predicates by using a relational verb and filling in one of the places with a reference to an object drawn from some infinite store. One such predicate is 'x believes that snow is white.' One virtue of this analogy is that it makes clear why different assignments of objects can capture all the relevant information about a situation without compromising the truth or "reality" of the situation. No one thinks the fact that we can register weight in either pounds or kilograms shows that there is something unreal about the weight of an object: different sets of numbers can be used to keep track of exactly the same facts. Suppose the fact is that A weighs the same as B; then the number assigned

Indeterminism and Antirealism

115

to measure the weights of A and B is the same number, whether we measure weight in pounds or kilos. Suppose the fact is that C weighs twice what D does; then the number assigned to measure the weight of C must be twice the number assigned to measure the weight of D, whether the weights are given in pounds or kilos. All that is necessary is that the entities we use to keep track of how much things weigh have a structure in which certain features of the objects weighed can be represented, and numbers can do this in an infinite number of different ways. Another virtue of the analogy is that it makes clear why Quine is justified when he insists, against Chomsky and others, that the indeterminacy of translation is distinct from the underdetermination of theories by all possible evidence. There is a question, to be sure, whether underdetermination in Quine's sense can occur, for it requires that there be empirically equivalent, but incompatible, theories. 4 Indeterminacy is not like this: the empirically equivalent theories it accepts as equally good for understanding an agent are not incompatible, any more than measurements of weight in pounds and kilos are incompatible theories of weight. What objects can we use to keep track of the attitudes? Clearly they must constitute approximately as complexly figured a field as the individual attitudes themselves. The most conspicuous features of the individual attitudes are their basically rational structures (if someone believes that everything is white, that person has a belief that entails that snow is white), and their relations to the world (the belief that snow is white is true if and only if snow is white).5 Entities that have these required properties are our sentences, and it is not clear that any other set of entities will do as well. In any case, there is a sense in which our standard attributions of attitude do use the sentences of the attributor to specify the contents of the attributed attitude. By this I mean no more than that we must use a sentence of our own after such words as 'Joan believes that ... ' to identify Joan's belief. Now we must ask which properties of the sentence are relevant to the identification of the belief, just as we ask which properties of numbers are essential to their role in recording weights. It is obvious that the property of having a truth value is not enough. The syntax would be more than enough if it were not for the fact that the same sentence can be used to mean quite different things, depending on the interpretation (the same sentence may belong to different languages or idiolects, or it may be ambiguous). Because of this difficulty, I have suggested that we take the actual utterance (or inscription) provided by the attributor as the object to which he is referring when to give the content of an attitude. So I proposed a paratactic account of attitude-attributing sentences, in which the 'that' of 'Joan believes that ... ' is

116

Realism/Antirealism and Epistemology

to be taken as a demonstrative referring to the utterance that follows. The utterance has a unique syntax, and can be taken to resolve the question of the appropriate language and idiolect, to have eliminated ambiguities, and to have provided the parameters needed to fix indexical references. As a rough equivalent of the paratactic account, we could think of the content sentence as enclosed in quotation marks, with the language and other features of the speaker and context understood. This second idea does better at capturing the intuition that attitude-attributing expressions are single sentences. It should not surprise us that what we can say and understand about the propositional attitudes of others should be what we can capture by matching up our own sentences (or utterances-I shall not bother to distinguish) to those attitudes. This does not mean, of course, that someone else may not think things I cannot; but then that is bound to be something I cannot express either. Neither do I want altogether to rule out propositional thoughts I have but cannot express. It is enough for my argument that when we can express a thought, our own or that of another, we must fall back on the basic device of representing it in the fabric of our sentences. Our sentences provide the only measure of the mental. What follows for indeterminacy? Everyone is apt to agree that sometimes quite distinct sentences of the same reporter may be used to attribute exactly the same thought, so some degree of indeterminacy comes automatically with the redundancy of language. But the indeterminacies of which Quine speaks, and most of which I accept, are another matter, for they suggest that sentences that no one would take to be even roughly synonymous may nevertheless be used to specify the same thought. For the sake of the record, let me say briefly where Quine and I differ and agree on these matters. Both of us take as the basic evidence for interpretation a person's attitudes towards various utterances and the circumstances that cause these attitudes. Quine takes certain patterns of assent and dissent to yield the interpretation of the truth functional sentential functions, and so do I; but I go on, as Quine does not, to use the same method to locate the devices for quantification and cross reference. Thus I do not, like Quine, see the internal structure of the simplest sentences as indeterminate. We both assume that the observed range of phenomena that prompt assent and dissent to certain sentences allows us to connect those sentences to events and objects in the world. As for sentences less directly tied to the readily observable, I think we can often do better than Quine believes, because I hold that we can use our ability to detect degrees of evidential support between sentences as a key to correct inter-

Indeterminism and Antirealism

117

pretation. There remain two important kinds of indeterminacy on which we agree: indeterminacy due to what Quine calls the inscrutability of reference, and indeterminacy that results from the blurring of the distinction between the analytic and the synthetic. The thesis of the inscrutability of reference says there is no way to tell which way of connecting words with things is the right way; if one way works, there will be countless others that do as well. From a technical point of view, this means that for the standard satisfaction relation (satisfaction is a sophisticated form of reference relation), we can substitute endless other relations without altering the truth conditions of any sentence or the logical relations among sentences. Since all the evidence for interpreting language must come at the sentential level (for only sentences have a use in communication), the result is that there can be no evidence that one of the satisfaction (or reference) relations is the right one. Here is an example. Suppose satisfaction relation 5 maps the word 'Rome' onto Rome, and the predicate 'is a city in Italy' onto cities in Italy. Then the truth definition will show that the sentence 'Rome is a city in Italy' is true if and only if Rome is a city in Italy. Now consider another satisfaction relation 5', which maps the word 'Rome' onto an area one hundred miles to the south of Rome, and the predicate 'is a city in Italy' onto areas on hundred miles south of cities in Italy. The truth definition will now say that the sentence 'Rome is a city in Italy' is true if and only if the area one hundred miles south of Rome is an area one hundred miles south of a city in Italy. The truth conditions are clearly equivalent. The thesis of the inscrutability of reference contends that there can be no evidence that 5 is any better than 5' for interpreting the sentence 'Rome is a city in Italy'. There is no telling what a sentence is "about," or what someone is thinking about. How can this be? Surely Rome and an area one hundred miles to its south are not the same entity. True enough; however, any differences we conventionally think of as differences in the reference of names or the extensions of predicates will be preserved by any correct satisfaction relation. The fact that there is no empirical difference between the interpretations yielded by 5 and by 5' doesn't entail that the person being interpreted can't tell the difference between Rome and an area one hundred miles to the south; but it does entail that there is no saying which one his word 'Rome' refers to. Correct interpretation keeps track of a complex pattern and locates particular sentences and attitudes within it. But the counters we use to represent this pattern, namely our sentences, can represent it in more than one way. If the two interpretations of someone's utterance of 'Rome is a city in

118

Realism/Antirealism and Epistemology

Italy' are equally correct, why is it that we feel that one is the "right" one? The complete answer is, I think, complex, but the simple answer is that we accept the standard, and in this case easiest, method of word-for-word translation. Of course, when the homophonic translation manual is one of the available manuals, we will use it; and in interpreting a language for which no homophonic manual is available, we will use the conventional, and for the most part shortest available, method. This does not mean that other interpretations are wrong: 5' provides as correct an interpretation of 'Rome is a city in Italy' as S. To someone who objects, "But 'Rome' doesn't mean 'area one hundred miles south of Rome,'" the right answer is: individual words don't have meanings. They have a role in determining the truth conditions of sentences, and this role is captured by 5' as surely as it is by s. Still, this answer leaves something out, which is the necessity, in assigning contents to an agent's speech or attitudes, of sticking to some one method of interpretation. It will falsify 'Rome is a city in Italy' to interpret it as true if and only if Rome is in an area one hundred miles south of a city in Italy; and it will destroy the entailment relations among sentences to use one method of interpretation for one sentence and another method for another sentence. Just as we must indicate whether the numbers we are using to measure temperature place the temperature on the Fahrenheit or the centigrade scale, so we must indicate (if we cannot assume) which method of interpretation we are using (5 or 5', for example). One of the conveniences of sticking to the same method is that we can leave the indication of the method tacit. (Somewhat similarly, if I say, "It must be seventy in this room," you will naturally understand me to be using the Fahrenheit scale.) There is a further objection which many people feel is fatal to the thesis of indeterminacy. The claim is that indeterminacy undermines first person authority. Such people want to say, "Maybe Quine's empirical criteria can't rule out the use of alternative methods of interpretation for others, but I know that I mean Rome by my word 'Rome' and not the area one hundred miles south of Rome." This remark is correct, but it doesn't conflict with indeterminism. The reason it is correct is that the speaker does not identify Rome and an area one hundred miles to the south of Rome. When a person is making such self-attributions, we know that words in the metalanguage play the same roles as those words do in the object language, for object language and metalanguage are one and the same. First person interpretations are necessarily tied to the homophonic translation manual (which is to say translation has no place here). It should not be concluded from the fact that a person is restricted to a unique way of interpreting himself (if this can be called interpretation; it

Indeterminism and Antirealism

119

would be better to say that aside from pathological cases, our way of interpreting others has no application to ourselves) that therefore his words have unique reference. All that the agent has discovered in our example is that his phrases 'Rome' and 'the area one hundred miles south of Rome' cannot be substituted for one another in his language, and this does not fix the reference of either expression. By the same token, the fact that the reference of his words is not fixed doesn't show that he doesn't know what he means or thinks. The different ways of interpreting his speech, or of representing his thoughts, mark no difference in the contents of his attitudes, so there is nothing he fails to know if he doesn't know which way of representing his thoughts is the right way; since there is no one right way, there is nothing more for him to know. Not all cases of indeterminacy depend on the possibility of systematically altering the satisfaction relation in ways that do not affect the truth conditions of sentences; there can be indeterminacies that affect truth values. These indeterminacies arise if one accepts (as I do) Quine's claim that there is no principled way to make a clear distinction between the analytic and the synthetic. Here it is easy to give examples. I find that I very often disagree with other people over whether to call the color of some object green or blue. The disagreement is consistent: there is a fairly definite range of cases where I say green and they say blue. We can account for this difference in two ways: it may be that I (or most other people) are wrong about the color of certain objects, or it may be that I don't use the words 'blue' and 'green' in quite the way others do. There may be no way to decide between these two accounts; by making compensatory adjustments elsewhere in one's interpretation of my sentences and beliefs, one can accommodate either story. But on one account certain of my pronouncements about colors are false, while on the other they are true. And on both accounts, I know what I think: I think the things I call green are green and the things I call blue are blue; neither account of what I mean and think threatens this first-person knowledge. Where the criticisms I have been discussing suggest that if one accepts indeterminism one is treating meanings and mental states as less than totally real, Daniel Dennett has refreshingly complained that on my view intentional states are too rea1. 6 Dennett sees clearly why my view, which compares rival theories of interpretation to different ways of representing length or temperature in the numbers, can't in itself call into question the reality of what is represented. But he thinks two different systems of belief attribution to an individual may differ substantially, even to the point of yielding different predictions of behavior, and yet nothing would establish that one system and not the other described the person's real beliefs.

120

Realism/Antirealism and Epistemology

Here I think the issue of prediction is something of a red herring. No system of attitude attribution, no matter how complete, yields any prediction of actions without a theory, and it is certainly possible to differ on predictive theories in psychology. It is clear that no plausible predictions will be forthcoming without a quantitative description that specifies degrees of belief (subjective probabilities) and relative strength of desires. But even supposing we could give a complete description of all the attitudes and their strengths for a given individual at a time, there is no reason to suppose that there exist strict laws to predict what the individual will do next-and good reasons to suppose there are no such laws. What sort of evidence can Dennett appeal to, then, to show that equally justified systems for attributing attitudes are incompatible? The fact that different systems result in different predictions proves nothing, since the same system can be used to support different predictions. Dennett's idea is that what is real is behavior, and intentional states are patterns in this behavior. The patterns aren't defined in terms of the behavior-they are perceived by an observer when the observer takes the "intentional stance." The value of the patterns is that they reduce a vastby us indescribably complex-physical situation to something we can grasp, and something on the basis of which we can make rough predictions. The patterns are in some (reduced) sense real, though abstract, and different people may perceive different patterns in the same behavioral field. Some of these patterns may do better at predicting or understanding some phenomena; other patterns do better at predicting or understanding other phenomena. All these patterns are "real," but when they are different there is no saying which represent the real attitudes of the targeted agent. It is not easy to see how to judge this suggestion, and harder to grasp how it relates to the issue of realism. Surely not every pattern one perceives in behavior is a propositional attitude. But is any? Patterns, Dennett tells us, are abstractions, like centers of gravity. But can they then be beliefs and desires? Beliefs and desires, we like to think, are states of a physical body which can have causal consequences; abstractions, I assume, have no causal relations. Forces don't act on centers of gravity but on the things that have centers of gravity, but forces certainly alter my beliefs and intentions. Do we perceive patterns? It seems to me we do not: what we perceive is something that has a certain pattern, and with luck (and the right stance) we may perceive that it has that pattern. So the question isn't whether patterns are real. Being no nominalist, I think patterns, like shapes and numbers, are as real as can be. But I do not see how the propositional attitudes of a person can be patterns. If we ask what exhibits

Indeterminism and Antirealism

121

the pattern, we can say it is the person, or we can say it is the observable behavior of the person. But in either case no issue concerning the ontological status of attitudes is at stake. If people or their behavior really do exhibit the patterns Dennett says they do, and for someone to have a propositional attitude is for that person to exhibit that pattern, then there is nothing ontologically or epistemically second grade about the attitudes. It seems to me Dennett has confused two issues. One is whether the attitudes are entities, and here I think the answer is no, unless you suppose that states are entities. Otherwise one should simply talk of people having attitudes, which means that certain predicates are true of them. The second issue is whether there is a correct answer to the question whether or not someone has a certain attitude. This I take to be not a question about vagueness or borderline cases but a question about whether there are objective grounds for choosing among conflicting hypotheses. Dennett has urged that the answer to the second question is that there are no such grounds, but I do not think he has given any reason to accept this answer. Are there objective grounds for choosing among conflicting hypotheses? Especially in this case we have to ask what makes grounds "objective." The only ultimate source 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, in my view, on the fact that two or more creatures are responding, more or less simultaneously, to input from a shared world, and from each other. We are apt to say that someone responds in "the same way" to, say, wolves. But of course, 'same' here means "similar." Our grounds for claiming that a person finds one wolf similar to another is the fact that the person responds in similar ways to wolves. This prompts the next question: what makes the reactions similar? The only answer is, someone else finds both wolves and the reactions of the first person similar. This of course only puts the basic question off once more. Nevertheless, it is this triangular nexus of causal relations involving the reactions of two (or more) creatures to each other and to shared stimuli in the world that supplies the conditions necessary for the concept of truth to have application. Without a second person there is, as Wittgenstein powerfully suggests, no basis for a judgment that a reaction is wrong or, therefore, right? This brings me to my last point. The analogy I proposed between measurement in the physical sciences and the assignment of contents to the words and thoughts of others is imperfect in an essential respect. In the

122

Realism/Antirealism and Epistemology

case of ordinary measurement, we use the numbers to keep track of the facts that interest us. In the case of the propositional attitudes we use our sentences. But there is this difference: we can mutually specify the properties of the numbers. The numbers, like the objects we apply them to, lie, as it were, halfway between ourselves and others. This is what it means to say they are objective, that they are objects. It cannot be this way with our sentences. You and I cannot come to agree on the relevant properties of our sentences as a preliminary to using them to interpret others, for the process of coming to such an agreement involves interpretation of the very sort we thought to prepare for. It makes no sense to ask for a common standard of interpretation, for mutual interpretation provides the only standard we have. We should not despair because we cannot provide a standard by which to judge the standard, a test for whether the standard meter bar is actually a meter long. Our conclusion should rather be: if our judgments of the propositional attitudes of others are not objective, no judgments are.

Notes 1. W. V. Quine, Word and Object (Cambridge: MIT Press, 1960),219. 2. Quine, Word and Object, 221. 3. There is, to be sure, the question whether meanings and propositions exist, but this I see as a matter of decision. If they exist, they are abstract entities and need no more than defining. Once defined, it remains to be shown whether they do useful explanatory or descriptive work. 4. Quine has changed his mind on the issue more than once. See W. V. Quine, Pursuit of Truth, rev. ed. (Cambridge: Harvard University Press, 1992),98 ff. 5. It should not be thought that by speaking of the relations between the attitudes and the world I am embracing a correspondence theory. A theory of truth of the sort I have in mind does depend on setting up a relation between certain words and objects ("satisfaction"), but it makes no use of objects to which sentences might correspond. 6. Daniel C. Dennett, "Real Patterns," Journal of Philosophy 88 (1991): 27-51. 7. For more on the subject of this and the concluding paragraphs, see my "Three Varieties of Knowledge," in A. J. Ayer: Memorial Essays, ed. A. Phillips Griffiths, Royal Institute of Philosophy Supplement no. 30 (Cambridge: Cambridge University Press, 1991), 153-66.

6 Pragmatism and Reasons for Belief

Gilbert Harman

Realism, Antirealism, Pragmatism

D ealism and antirealism with respect to a given question are compet.l'\....ing claims about what makes an answer to the question correct. Realism says that what makes an answer correct is that it represents things as they really are. Antirealism denies this. Consider the question of whether numbers exist. According to realism, the answer to this question depends on whether in reality numbers exist. If they do, the correct answer is yes; otherwise the correct answer is no. One form of antirealism about this question denies that it makes sense to suppose that numbers do or do not exist in reality and interprets the question as a practical question about whether things work out better if one allows quantification over numbers in a theory one is developing. Pragmatism supposes that all or almost all questions are at least in part practical questions. Since one answers many of these questions by adopting certain beliefs, pragmatism tends to emphasize practical reasons for belief. Pragmatism might be considered a form of antirealism, but it is probably better seen as rejecting the distinction between realism and antirealism on the grounds that there is no sharp distinction between theoretical and practical questions.1 From an intuitive realist point of view, there is a clear distinction between reasons to believe something that are relevant to the truth of what is to be believed and other reasons that indicate something to be gained through belief without making the belief any more likely to be true. One intuitively distinguishes between purely epistemic reasons to believe something and other, more pragmatic nonepistemic reasons. 123

124

Realism/Antirealism and Epistemology

My concern in this paper is with what pragmatism can say about this intuitive distinction.

Epistemic and Nonepistemic Reasons for Belief

Consider possible reasons for believing that cigarette smoking does not cause cancer. One reason might involve a statistical study of identical twins who differ in their smoking habits. If the life expectancy of the smokers was on average the same as the life expectancy of the nonsmokers, that would provide an epistemic reason to believe that cigarette smoking does not cause cancer. Another reason might be a practical one. An advertising agency who wants to obtain the RST Tobacco Company advertising account, knowing that RST will only use agencies who believe that cigarettes do not cause cancer, has a practical reason to believe that cigarettes do not cause cancer. Having that belief is a necessary condition of obtaining the RST advertising account. Clearly this is not an epistemic reason to believe that cigarettes do not cause cancer. It is a purely practical reason. The fact that having such a belief is needed to obtain the account is not evidence for the belief and does not make the belief more likely to be true. Similarly, with respect to whether or not to believe that capital punishment is a deterrent, statistical evidence about crime rates in states with and without capital punishment might provide an epistemic reason for a belief about this subject, whereas the consideration that one's friends will think badly of one unless one agrees with their position might very well constitute only a practical reason for a certain belief. If a friend has been charged with theft from a local store, the testimony of the store detective provides an epistemic reason to believe one's friend is guilty, whereas loyalty to one's friend provides a nonepistemic moral or pragmatic reason to believe one's friend is innocent. Suppose one is trying to persuade people to join the army to fight against the evil Rudu. One needs to convince people that the war will be short, something one does not oneself believe. One will be more persuasive if one can get oneself to believe that the war will be a short one. This gives one a reason to believe that the war will be short, but it is a purely pragmatic, nonepistemic reason. The fact that one will be more persuasive if one believes the war will be short does not make it more likely that the war will be short. In contrast, learning that Rudu is low on supplies would provide an epistemic reason to believe that the war will be short, since it would make it more likely that the war will be short.

Pragmatism and Reasons for Belief

125

Practical Reasons for Belief

It might be objected that one can have only practical reasons for an action

and that believing something is not an action. In this view, one can have practical reasons to do something that may lead one to have a belief, but one's practical reasons cannot be one's reasons for believing it. If one is asked one's reason for believing that cigarettes do not cause cancer, it makes sense to appeal to a statistical study of smokers and nonsmokers. It does not make sense to appeal to the fact that such a belief is needed if one is to obtain an advertising account. That might be one's reason for pretending to have the relevant belief or for undertaking activities that will bring about that belief, but it cannot be the reason for one's belief. In this view, belief is passive and practical reasons are always reasons for something active. 2 Opposed to this view is the observation that there are various situations in which one can decide to believe something. A friend gives her version of what happened; one decides to believe her; one decides to believe that she is telling the truth. 3 It is true that one can find oneself almost passively accepting certain beliefs. Sometimes a speaker is so persuasive that one cannot help believing him. But one can find oneself doing all sorts of things-stroking one's beard, brushing one's teeth when one intended merely to get a drink of water, and so forth. Finding oneself doing something does not distinguish coming to believe something from other actions. And coming to believe something is not always a matter of finding oneself believing something. What is it to believe something? At least two things are involved, both of which are things one can decide to do. Beliefs serve as the basis for other acts one may undertake, and beliefs end one's inquiry into an issue. One can certainly decide to accept a given proposition as a basis for action; that is, one can decide to act as if one believed it. Furthermore, one can decide to end one's inquiry into an issue and hold to a certain conclusion. Jonathan Bennett observes in effect that there is a difference between increasing one's strength of belief (subjective probability) and doing other things that may be involved in belief, such as ending inquiry.4 These are separable matters, since, for example, whether or not to end inquiry is not always completely determined by the subjective probabilities. Like Bennett, I am here concerned with reasons for increasing the strength of one's belief. So it is important to note that one can decide to increase one's strength of belief, inasmuch as one can decide that in further decisions one will treat a certain proposition as being highly probable. To do this is like deciding to adopt a particular working hypothesis to

126

Realism/Antirealism and Epistemology

tryout over the long haul, in depth. Such a decision is hard to distinguish from deciding to believe something.

Internal versus External Questions

Something like the distinction between epistemic and nonepistemic practical reasons plays an important role in Rudolf Carnap's distinction between internal and external questions. s Carnap argues that certain philosophical puzzles arise through not distinguishing issues to which epistemic reasons apply from issues to which practical reasons apply. A question like "Do numbers exist?" is confused with a question like "Do prime numbers greater than a thousand exist?" The first sort of question seems very "philosophical," pitting Platonists against conceptualists. The second sort of question seems to raise no serious philosophical issues, requiring only the briefest of calculations to answer it. Yet if there are prime numbers greater than a thousand, then there are numbers, so a positive answer to the nonphilosophical second question seems to imply a positive answer to the philosophical first question. To take another example, philosophers agonize over whether it is possible to know that there are any external objects, where, for example, a human hand would be an example of an external object. Yet it seems easy enough to determine whether a given individual has hands. G. E. Moore took this to imply that of course one knows that there are external objects. Holding up his hands, he would say, "Here's one and here's another." But many philosophers feel that Moore's approach is not relevant to answering the philosophical question. Carnap's response is to distinguish "internal questions" from "external questions." An external question concerns how philosophers are to talk about or otherwise represent a given subject matter. It is a question concerning choice of language. Carnap takes this to be a practical question. What system of representation best enables one to fulfill one's purposes? What allows the most efficient representation of what one wants to say? What most facilitates the sorts of calculations one needs to perform? Choice of language in this sense determines the methods to be used to settle issues within a given framework. In choosing a language of arithmetic in which one refers to numbers, one chooses a system of calculation that allows one to answer such questions as whether there are prime numbers greater than a thousand. When one chooses a language in which one refers to external objects like tables, chairs, and hands, one chooses a sys-

Pragmatism and Reasons for Belief

127

tern that enables one to find evidence for and against such propositions as that a certain person has two hands. The internal question presupposes an answer to the external question. It presupposes that one has already chosen a system of representation with rules for answering such internal questions. One has to answer the external question before one can intelligibly ask the internal question. Moore's way of answering external questions simply begs the point at issue. In Carnap's view, a system of representation determines what is to count as evidence, as an epistemic reason, but one cannot in the same way appeal to evidence to decide on a system of representation. Such a decision can be based only on practical or pragmatic reasons. Carnap is therefore an antirealist about external, philosophical questions and a kind of realist about internal questions. He is an external antirealist and an internal realist.

Pragmatism

Pragmatists deny that there is the sort of distinction between internal and external questions that Carnap describes and therefore deny the distinction between internal and external realism and antirealism. W. V. Quine famously responded to Carnap in the following way: I grant that one's hypothesis as to what there is, e.g., as to there being universals, is at bottom just as arbitrary or pragmatic a matter as one's adoption of a new brand of set theory or even a new system of bookkeeping .... But what impresses me more than it does Carnap is how well this whole attitude is suited also to the theoretical hypotheses of natural science itself, and how little basis there is for a distinction. 6

In Quine's view there is no principled distinction between choice of system of representation and choice of theory and there is no principled distinction between changing the system of representation and changing the views expressed in that system. There is no principled difference between keeping the same system of representation and changing one's view as expressed in that system and changing the system of representation by changing the meanings of representations. Choice of principles of theory is often dictated by how useful these principles will be in getting interesting results, simplifying calculation, and so forth. If the relevance of such considerations makes the choice a choice of system of representation rather than a choice of theory to be expressed in that system, then one is always choosing and changing the system of representation, just as one

128

Realism/Antirealism and Epistemology

is always choosing and changing the view expressed in that system? Quine is a pragmatist. In general, pragmatists deny that there is any sharp distinction between purely epistemic reasons to believe something and more practical reasons. Reasons to believe one thing rather than another are strongly connected with goals and interests. For one thing, goals and interests determine which questions there is a reason to be interested in answering. In addition, practical considerations lie behind a certain conservatism or momentum in reasoning. Initial beliefs provide the starting point for reasoning, and one tries to minimize the change in one's beliefs subject to other demands. It would be impractical to follow Descartes and abandon every belief for which one cannot provide an indubitable justification from premises one cannot doubt. 8 Furthermore, and this is important, when faced with inferring some explanation of the evidence, one finds oneself with infinitely many possible explanations, all of which account for the evidence. Without thinking much about it, one often immediately chooses the simplest of these hypotheses and ignores the others, where the simplest hypothesis is generally the hypothesis that would be easiest to use in order to answer the sorts of questions in which one is interested. The reasons for selecting the simplest hypothesis are entirely pragmatic. 9

Pragmatism and Purely Evidential Reasons

This leaves a problem. On the one hand, there is a clear intuitive basis for distinguishing purely epistemic reasons for believing something from other merely pragmatic reasons, such as obtaining the RST advertising account. There seems to be an intuitively clear distinction between reasons for belief that make the belief more likely to be true and reasons for belief that merely promise some practical benefit of belief without making the belief more likely to be true. On the other hand, pragmatism is surely correct in emphasizing the practical side of ordinary standards of evidence, the goal-directed character of reasoning, conservatism, the role of simplicity, and so forth. How can the intuitively realist distinction between epistemic and nonepistemic reasons be accommodated while allowing for the insights of pragmatism? Clearly, there is an accommodation only if the distinction between epistemic and nonepistemic reasons is not just a distinction between epistemic and practical considerations. Some practical considerations must be allowed to fall on the purely epistemic side of the distinction. But then what could distinguish epistemic from nonepistemic reasons?

Pragmatism and Reasons for Belief

129

Epistemic Reasons as Connected with Truth? It might be suggested that a reason is an epistemic reason for believing a particular proposition p if one is justified in believing that there is a con-

nection between the reason and the truth of p, where the relevant connection is explanatory in a wide sense that includes causal and logical connections.lO In this view, one can suppose that a statistical study is evidence that cigarette smoking does not cause cancer because one can conclude that cigarettes' not causing cancer is part of an explanation of the results of the study. One cannot reasonably take the opportunity to obtain the RST advertising account to be evidence that cigarette smoking does not cause cancer because (I am supposing) one cannot reasonably conclude that cigarette smoking's not causing cancer is any part of the explanation of the company's policy with respect to advertising agencies. How might this apply to simplicity as an epistemic factor? In any particular instance, many different hypotheses might explain the evidence. Now, it may seem that one can take simplicity to be a relevant epistemic consideration without having to suppose that the simplicity of a hypothesis is what explains why that hypothesis rather than some other hypothesis is what accounts for the data. For example, one can take simplicity to be a relevant epistemic consideration without supposing that statistics show that simpler hypotheses are more likely to be what account for the data than other, less simple hypotheses that are equally compatible with the same data. But people do think that the simplicity of a hypothesis has something to do with an explanation of the data, since people do suppose that, other things being equal, simpler hypotheses provide better explanations than complicated hypotheses do. So the suggested criterion could allow people to count the simplicity of a hypothesis to be an epistemic reason for accepting the hypothesis, since anyone can suppose that the simplicity of the hypothesis is part of what makes the truth of the hypothesis a good explanation of the evidence. The suggested criterion would not allow one to count the opportunity to obtain the RST account as an epistemic reason for believing that cigarette smoking does not cause cancer, because one cannot see this opportunity as part of an explanatory connection between the truth of that belief and one's evidence. So the suggested criterion would give an intuitively acceptable account of many cases. But, alas, it would also give incorrect results for certain cases. Suppose, for example, that the company requires agencies to believe in the relevant explanatory connection. That is, suppose that the company requires agencies to believe that cigarettes' not causing cancer is part of the

130

Realism/Antirealism and Epistemology

explanation of the company's policy of hiring only agencies who believe cigarette smoking does not cause cancer. Then the suggested criterion seems to imply that the offer provides an epistemic reason to believe that cigarettes do not cause cancer, which is the wrong result. In this case, there are practical reasons to conclude that, because cigarette smoking does not cause cancer, the company has the policy that gives one the practical reason. One is justified in believing that the required connection between truth and reasons holds, but the reasons are still merely practical, not epistemic. So this criterion seems to fail. H

Practical Sources of Epistemic Reasons? It is interesting that one common realist approach to explaining the distinction between epistemic and nonepistemic practical reasons for belief identifies an epistemic reason with a certain sort of practical reason! In this view, an epistemic reason to believe something is a practical reason to believe it deriving solely from an interest in believing what is true and not believing what is falseP Of course, people do not actually have this general desire. 13 Curiosity is more specialized. One wants to know I4 whether p, who did d, what things are F, and so forth. With respect to some limited range of options one wants to arrive at an answer p if and only if it is true that p. Typically, one wants to know whether p because finding out whether p will help one achieve some further goal, although one's goal may be simply to find out more about a certain subject. So, we might put the suggestion as follows. An epistemic reason to believe p is any practical reason to believe p that derives solely from the desire to know whether p is true, ignoring all other desires. Epistemic reasons are not just reasons that derive from a desire to believe what is true and avoid believing what is false. Perhaps one will be given a copy of the RST Encyclopedia only if one believes that cigarettes do not cause cancer. There are a lot of truths in that encyclopedia, so believing that cigarettes do not cause cancer would definitely promote one's desire to believe what is true and avoid believing what is false. But that will not count by the present criterion as an epistemic reason to believe that cigarettes do not cause cancer, because the reason in question here does not derive solely from a desire to know whether cigarettes cause cancer. What if one wants to read the encyclopedia to find out whether cigarettes cause cancer? One may have reason to believe that the correct answer is in the encyclopedia. However, one can obtain the encyclopedia

Pragmatism and Reasons for Belief

131

only if one already believes that cigarettes do not cause cancer! Here one's desire to have a true belief about whether cigarettes cause cancer provides one with a nonepistemic practical reason to believe that cigarettes do not cause cancer. So the criterion needs to be understood in a way that does not count this sort of reason as an epistemic reason. IS One idea is to require that the desire motivating the belief be a desire to have a true belief on this subject now.1 6 But what if the encyclopedia is right here, opened to the right page, and all that is needed is to glance down and read what it says, something one will be permitted to do only if one first has the belief that cigarettes do not cause cancer? Does one second make the difference between now and the future? Then how does one allow for the time it takes to form a belief? Might there be a case in which forming a certain belief would be a necessary condition for simultaneously learning the truth? Perhaps the criterion should be this: a reason is an epistemic reason to believe p if and only if it depends only on the desire that one's belief in this instance be true (or the desire that one's belief in this instance should constitute knowledge). The fact that the encyclopedia can be consulted only by someone with the appropriate belief does not give one an epistemic reason to form that belief, even if it gives one a reason that derives from one's desire to know the truth about this subject, since one's reason for forming that belief does not derive from the desire that that very belief be true (or be knowledge). The criterion is still not obviously adequate. It may turn out to be impossible to separate the factors leading to belief into those that do and those that do not depend only on the relevant desire for truth (or knowledge) in this case. It may turn out that everything is tainted by other desires or interests, so that no reasons count as epistemic by this criterion. Or there may turn out to be very few instances of epistemic reasons by this criterion. This may lead to a conflict with the original intuitions about the distinction. And other sorts of conflict are possible. For example, one natural way to develop this proposal assigns +1 units of epistemic value to believing any proposition p if P is true, -1 units of epistemic value to believing p if P is not true, and a units to failing to believe p whether p is true or false. Then the expected epistemic value of believing p = prob(p) - prob(not-p). This is greater than zero if, and only if, p is more probable than not. So this approach entails that one should believe all and only those propositions that are more likely to be true than falseY However, this consequence is bizarre. We might try a more complicated measure of epistemic value,18 but it is unclear how to do this in a way

132

Realism/Antirealism and Epistemology

that retains the intuitive appeal of the principles that lead to the bizarre consequence. Instead, I suggest that we simply stay with the idea we started with. An epistemic reason for believing something is a consideration that makes that belief more likely to be true. A nonepistemic reason for believing something is a consideration that provides a reason for belief without making the belief more likely to be true. But what is meant by 'likely to be true'?19

When Is a Conclusion "Likely to Be True"? It is often supposed that there are at least two kinds of likelihood or prob-

ability: probability as objective chance, which mayor may not be known; and probability as rational degree of belief. These can diverge, and it is degree of rational belief that is relevant here. Suppose one has a six-sided die of unknown constitution. It may be true in fact that the die is weighted to favor side six in the sense that if the die is rolled over and over again, six tends to come up about 75 percent of the time. Before one first rolls the die, one may know nothing about this. One's evidence about the die may not indicate that one side is favored over another. So one's rational degree of belief in the proposition that the die will come up six when one rolls it for the first time may equal one in six, even though the objective chance of the die's coming up six when one rolls it for the first time may equal three in four. Epistemic reasons have to do with rational degree of belief, not directly with objective chance. 2o One's evidence consists in what one learns by studying the die. By hypothesis, the evidence does not favor one side of the die over another. In the relevant sense, the evidence does not make six more likely than any of the other five outcomes. The evidence does not increase the epistemic likelihood of getting a six even though the objective chance is higher than with some other die. So if an epistemic reason is a consideration that makes a certain proposition more likely to be true, it is a consideration that raises the rational degree of belief one should attach to that proposition, not just a consideration that increases the objective chance of the proposition's being true. In other words, an epistemic reason is a consideration whose acceptance should rationally increase one's degree of belief in a proposition. Alas, this way of understanding an epistemic reason does not distinguish epistemic reasons from nonepistemic reasons. Any reason at all that makes it rational to increase one's degree of belief in a proposition is a rea-

Pragmatism and Reasons for Belief

133

son that raises the degree of belief one ought rationally to place in that proposition. Any reason, including the most blatant practical reason, will do. The fact that RST will accept only agencies who believe that cigarettes do not cause cancer may increase the degree to which it is rational for one to believe that cigarettes do not cause lung cancer. It is rational for one to increase one's degree of belief in that proposition inasmuch as doing so may enable one to acquire the RST advertising account. But something must have gone wrong here. This consideration (that RST will give its advertising only to someone who believes cigarettes do not cause lung cancer) is surely not a consideration that makes it more likely that cigarettes do not cause lung cancer. Increase in rational belief is not always increase in likelihood. How then is the relevant sort of likelihood to be understood? It is certainly not objective chance. In order to begin to answer this question, I turn to basic issues of subjective probability theory.

Coherence and Incoherence

Contemporary accounts of subjective probability do not provide necessary and sufficient conditions for determining the degree of rational belief one should place in a proposition. These accounts are instead concerned with the consistency or coherence of one's degrees of belief. For example, such accounts require that logically equivalent propositions be assigned the same degree of belief even though it is obvious that logically equivalent propositions often admit of different rational degrees of belief. It is inconsistent to assign different degrees of belief to logically equivalent propositions even though one may have every reason to do SO.21 As I have already mentioned, contemporary accounts relate subjective probability or degree of belief to certain practical considerations. Inconsistency in degrees of belief is represented by a certain sort of practical incoherence, or at least an incoherence in one's views about what it would be good for one to do or what transactions are fair. The relevant sort of incoherence holds, for example, if a person's preferences are intransitive. In this case, probabilities don't enter the picture; only preferences are relevant. Suppose Jack has the following preferences with respect to a painting by Alice Abbott, a novel by Barbara Baker, and a recording by Connie Cooper. (1) he would rather have Abbott's painting than Cooper's recording; (2) he would rather have Baker's novel than Abbott's painting; (3) he would rather have Cooper's recording than Baker's novel.

134

Realism/Antirealism and Epistemology

Suppose Jack has the painting by Abbott and suppose he also attaches value to money. He prefers having the Baker novel to having the Abbott painting. So he should be willing to pay some amount of money, say a dollar, for the opportunity of trading the Abbott painting for the Baker novel. But he also prefers having the Cooper recording to having the Baker novel, so he should be willing to pay some amount of money, say another dollar, for the opportunity to trade the Baker novel for the Cooper recording, if he should be in possession of the Baker novel. Finally, he also prefers having the Abbott painting to having the Cooper recording, so he should be willing to pay some amount of money, perhaps yet another dollar, for the opportunity to trade the Cooper recording for the Abbott painting, if he should ever be in possession of the Cooper recording. But this means he should now be willing to pay first one dollar, then a second, and then a third, simply to be able to return to his original situation in possession of the Abbott painting. This willingness to be a "money pump" for unscrupulous decision theorists illustrates the incoherence involved in having intransitive preferences of this sort. 22 But it is important to understand that the relevant incoherence does not consist in Jack's actually serving as a money pump in this case. Nor does there have to be any real threat that he will lose money in this way. His attitudes are incoherent in that they commit him to thinking that each of several transactions would be good for him to undertake and also thinking that it would not be good for him to undertake all of them. 23

Dutch Books

Similar considerations apply to subjective probability or degree of belief if, as is normally supposed, one's degree of belief in a proposition commits one to a view about what a fair bet would be concerning the truth of that proposition. If one's degree of belief in the proposition A is .6, then one is committed to taking it to be fair to offer odds of .6 to .4 (or 3 to 2) that A is true. Supposing that one values money (and that diminishing marginal utility of money24 is not a factor), one is committed to thinking the following bet is fair: if A is true, one gets four dollars; if A is not true, one loses six dollars. One's "expectation" with respect to this bet is determined by multiplying the gain or loss of each possible outcome by the probability of that outcome. So, one's expectation is .6 x $4 - .4 x $6 = $0. One is committed to supposing that the expectation is zero, which is to say one is committed to supposing that this is a fair bet. A book is a collection of bets of this sort. A Dutch Book from person P's

Pragmatism and Reasons for Belief

135

point of view is a collection of bets that P is committed to thinking would guarantee a loss for him- or herself. To take a trivial example, consider the following two bets: 1. If A is true, one receives four dollars; if A is not true, one loses six dollars. 2. If not A is true, one receives four dollars; if not A is not true, one loses six dollars.

If one takes both bets, one is guaranteed to lose two dollars. Either one wins four dollars from the first bet and loses six dollars from the second or one loses six dollars from the first bet and wins four dollars from the second. Since (by hypothesis) one values money, one is committed to thinking that this is not a fair set of bets. More generally, since a Dutch Book is a set of bets that one is committed to thinking guarantees that one ends up in a worse situation than one is now in, one is committed to thinking that no Dutch Book is a fair set of bets. Now suppose that one's subjective probabilities or degrees of belief commit one to thinking that each bet in a book of bets would be a fair bet and suppose that this book of bets is a Dutch Book from one's point of view. Then one's probabilities are incoherent; they commit one to thinking both that each individual bet in the book is fair and that the book of bets as a whole is not fair. H one's probabilities are not incoherent in this way, that is, if they are coherent, then one's probabilities must satisfy the usual axioms of probability theory.25 In the example just considered, the Dutch Book is made possible because the probabilities assigned to A and not A violate the axiom that says that the probability of A and the probability of not A should sum to exactly 1.0. The argument for the basic principles of probability is that if one's probabilities do not satisfy these axioms, they are incoherent.

Conditionalization and Temporal Dutch Books A similar argument can be given for the rule of conditionalization. 26 This is a rule about updating probabilities over time. This rule says that if one's evidence between now and time t can be expressed in the proposition e, where prior-prob(e) > 0, then one's subjective probabilities at time t should be in accord with the following principle:

136

Realism/Antirealism and Epistemology

b() new-pro p

prior-prob(p&e)

='--p~r--:i~or'---~pr~o--='b--:-(e--:-)-

The quotient on the right-hand side of this equation is called one's prior conditional probability of p given evidence e, and might be represented as cond-prob(p, e). The argument for conditionalization involves consideration of sets or books of bets that may be offered at various times. A temporal book of bets might include some bets to be made now and other bets to be made in the future if certain events should occur. A Temporal Dutch Book from one's point of view is a temporal book of bets with the property that one is committed to thinking that if one accepts all the bets if and when they are offered, one is guaranteed to lose. One's current commitments subject one to a Temporal Dutch Book if one is committed to supposing that each bet in the Temporal Dutch Book is or would be fair. If one's current commitments subject one to a Temporal Dutch Book, those commitments are incoherent, since they commit one to thinking that each of the bets is fair and also commit one to thinking that the book as a whole is not fair.27 For example, suppose one now takes the probability of a coin's coming up heads on the next toss to be .5 and one takes the probability of its coming up heads on both of the next two tosses to be .25. Suppose one also accepts a principle for changing one's probabilities that has the following consequence: if the coin comes up heads on the first toss, then change one's probability of its coming up heads on the second toss to .75 (perhaps because one takes the first head to be evidence that this is a two headed coin). Given these commitments, one is subject to thinking that all the bets in the following Temporal Dutch Book will be fair bets at the time they are made: A. A bet made now: If the coin comes up heads on the first toss, one receives forty-five dollars; otherwise one loses forty-five dollars.

B. A bet made now: If the coin comes up heads on both the first toss and second toss, one loses ninety dollars; otherwise one wins thirty dollars. C. A bet made after the first toss yields a heads (should that happen): If on the second toss, the coin comes up tails, one loses ninety dollars; otherwise one wins thirty dollars. If one accepts this book of bets, one is guaranteed to lose fifteen dollars, as should be clear from the following table:

137

Pragmatism and Reasans for Belief

Outcome of two fliEs Heads-Heads Heads-Tails Tails-Heads Tails-Tails

Result from bet A B C +$45 -$90 +$30 -$90 +$45 +$30 -$45 +$30 -$45 +$30

Total -$15 -$15 -$15 -$15

These commitments are incoherent. One is committed to thinking that it would be rational for one to be prepared to accept all these bets (since one is committed to thinking that they are or will be fair when they are made). One is also committed to thinking that it would be irrational to be prepared to accept all these bets because one would be guaranteed to lose at least ten dollars if one did so. (One is now committed to thinking that bet C will be fair when it is made because one is now committed to assigning a probability of .75 to getting a second heads, if the first toss comes up heads. So one is committed to thinking that odds of 3 to 1 would be fair at that point.) If one is committed to updating one's subjective probabilities in a way that diverges from conditionalization, then one is subject to a Temporal Dutch Book and one's commitments are incoherent in this way. In the example just given, the conditional probability of the coin's coming up heads on the second toss given that it comes up heads on the first toss equals the probability of getting heads on both tosses (.25) divided by the probability of getting heads on the first toss (.50), which equals .5. This is not the same as the probability that one is committed to assigning to getting heads on the second toss if heads comes up on the first toss, namely .75. This discrepancy makes one subject to the Temporal Dutch Book.

The Alleged Incoherence of Inference to the Best Explanation

Van Fraassen argues that any substantive commitment to using inference to the best explanation is incoherent. Given evidence e, inference to the best explanation considers possible explanations of e and either directly infers, or at least assigns higher probability to, the best of these explanations, where the criteria for best explanation might include the extent to which an explanation accommodates the data, conservatism, simplicity, and other factors. If one is committed to allowing such explanatory considerations to influence the way one updates one's probabilities beyond what is entailed by conditionalization, then one is subject to a Temporal

138

Realism/Antirealism and Epistemology

Dutch Book and one's overall commitments are incoherent. Van Fraassen concludes that it is necessary to reject any use of inference to the best explanation that goes beyond conditionalization. Inference to the best explanation is therefore either superfluous or incoherent, in van Fraassen's view. 28 As van Fraassen observes, the same argument applies to any ampliative rule, any rule that goes beyond conditionalization. For example, Goodman's theory of entrenchment assigns greater projectibility to hypotheses using predicates that have appeared in previous projections than to hypotheses that do not use such predicates. 29 If one is committed to allowing considerations of projectibility to influence the way one updates probabilities, given new evidence, in a way that goes beyond what conditionalization allows, then one is subject to a Temporal Dutch Book and one's commitments are incoherent.

Where Prior Probabilities Come From

Now, Goodman can accept van Fraassen's argument yet respond that there is still a role for entrenchment in determining prior probabilities. Considerations of entrenchment might affect a person's conditional probabilities, cond-prob(h,e). That is to say, considerations of entrenchment might affect the ratio between prior-prob(h&e) and prior-prob(e), for hypotheses with entrenched predicates. The effect on this ratio might itself be taken to be the result of conditionalization on evidence about entrenchment of predicates, call that evidence f Then

cond-prob(h, e&j) > cond-prob(h, e) Here, for example, h might be 'All emeralds are green', e might be evidence indicating that all emeralds examined for color have been found to be green (i.e., grue), andfmight be evidence that 'green' rather than 'grue' has been used in past projections. 3o This can seem strange, as if one were supposing that one would have made it more likely that all emeralds are grue if one had made predictions using 'grue' rather than 'green'. But the issue is degree of rational belief, not objective chance. There is no suggestion that the way one uses language to make predictions will have an effect on the objective frequencies with which nonlinguistic properties are correlated with each other. It might be suggested that, as one finds out what projections one is inclined to make, one finds out which hypotheses are more likely to be

Pragmatism and Reasons for Belief

139

true, given certain evidence. But that is not quite right. Even at the beginning, one is inclined to project 'green' rather than 'grue.' It is true even at the beginning that the probability that all emeralds are green, given that some have been observed to be green, is greater than the probability that all emeralds are grue, given that some have been observed to be grue. What one learns from patterns of past projection is what one's inductive principles are-not what principles one explicitly accepts, but what principles one implicitly follows. Initially, one just made inferences without having a theory about how one does it. As one sees what predicates one projects, one acquires information about one's prior probabilities, one's inductive principles.

Explanation and Prior Probabilities

In the same way, explanatory considerations, like simplicity, playa role in determining prior probabilities. Suppose one has two hypotheses, hand h', both of which accommodate the evidence e, but one takes h to be the better explanation because it is simpler. Suppose that this difference inclines one to infer h rather than h' or any of the infinitely many other competing hypotheses that could account for e. This does not mean that such explanatory considerations playa role in addition to conditionalization. Rather, the fact that one is inclined to infer h rather than h' shows that

cond-prob(h, e»> .5»cond-prob(h', e) This implies that

prob(h&e»>(prob(h'&e) in one's prior probability distribution. Here it may be useful to observe that the requirements of probabilistic coherence are consistency requirements, not directly requirements about inference. To take another example, if one believes both p and q, and sees that these propositions together imply r, that is not necessarily a reason to believe r. One may, for example, also believe not r. So maybe one has a reason to stop believing p or stop believing q. Nothing in logic or probability theory says that one has to stick with previous commitments. 31 When one realizes what those commitments entail, one is free to revise them, as far as anything that logic says. At best, one sees that something has to be done, but what is done may be to

140

Realism/Antirealism and Epistemology

change some of one's initial assumptions. It may also be useful to notice an ambiguity in the notion of "prior probability." The relevant meaning is not a probability arrived at ahead of time. Rather, what is meant is something mathematical: a probability distribution prior-prob that is related to one's current probability distribution prob via the formula of conditionalization:

prob(p)

prior-prob(p&e) prior-prob(e)

The distribution prior-prob can be a prior probability in the relevant sense without being a probability distribution one had ahead of time.

Belief and Degrees of Belief

To think of matters in terms of subjective probabilities is not to suppose that people have explicit degrees of belief that they use in deciding what to do and how to change their views in the light of new evidence. Nor is it even to suppose that degrees of belief are precisely determined. Degrees of belief are vaguely determined by a person's overall psychological or neurophysiological state, or perhaps by one or another part of that overall state. Something like a behavioristic account of degrees of belief may be right. Degrees of belief may be determined by a person's dispositions to make choices. Perhaps people have all-or-nothing beliefs and goals that are responsible for these behavioral dispositions. Degrees of belief can be determined by these dispositions without being functional aspects of a person's psychology.32

Nonepistemic Reasons and Temporal Dutch Books

One believes H that smoking cigarettes greatly increases a person's chances of getting cancer and dying early. One assigns a high probability to this conclusion. One thinks this probability would be unaffected by 0 receiving an offer from a tobacco company of a large amount of money if one would come to believe that smoking does not increase a person's risk of getting cancer. Let's suppose that prior-prob(H) = .9 and cond-prob(H,O) = .9, according to one's current probability distribution. Suppose one also thinks that if the tobacco company were to make one such an offer, one

Pragmatism and Reasans for Belief

141

would have a strong reason to form the belief that smoking is not bad for people. In other words, one thinks that if 0, one's prob(H) should become very low. But then one seems to be subject to a Temporal Dutch Book. 33 One seems to be committed to thinking that each bet in a Temporal Dutch Book should be accepted when offered, even though one is also committed to thinking one is guaranteed to lose. The point is tricky in the present case, because it is hard to bet on something like H. What counts as winning or losing the bet? So let us change the example slightly. Suppose a government study is being made and the question concerns how that study will come out. Suppose H is the proposition that the study will conclude that cigarettes cause cancer and that not H is the proposition that the study will not reach that conclusion. This study will come out a year from now and 0 is that the cigarette company offers one $100,000 to believe not H. Part of the relevant Temporal Dutch Book would be a bet on not O. Part would be a bet on O&H. Part would occur only if O. At that point one would be offered to bet on not H. Consider this last bet: at that point one would have shifted one's degree of belief in H to 0.1. So one would at that point be prepared to bet heavily against H, giving 9 to 1 odds! Now, when one is considering the offer of the tobacco company, should one conclude that one should change one's opinion of what a fair bet would be in this way? Should one really be prepared to give 9 to 1 odds against H? If one thought that someone would actually bet a large amount with one if one were to offer him such odds, then one now (before changing one's degree of belief) must conclude this offer from the tobacco company is not a good deal after all. One will get the $100,000 but almost certainly lose even more betting against H. In this case, whether one is willing to modify one's degrees of belief will depend on how one thinks one will have to act on one's modified degrees of belief after one makes the change. Here is a distinction between evidence and other sorts of reasons for believing something. Putting the point in another way, one will not want to view the sort of change of belief resulting from the tobacco company's offer as simple conditionalization. Envisioning the possibility of the offer ahead of time does not lead one to change one's prior probabilities in the way that envisioning getting certain sorts of evidence can lead one to change or firm up prior probabilities. So it looks as if epistemic reasons can be distinguished from other sorts of reasons by the following criterion. Suppose one would consider R to be a reason to believe H. If one's subjective probabilities are such that

142

Realism/Antirealism and Epistemology

prob(H&R) b(H) probeR) >pro then one is treating R as an evidential or epistemic reason. Otherwise one is treating R as a different sort of reason, possibly a purely pragmatic reason. It remains possible that some pragmatic reasons are also epistemic reasons. In particular, considerations of simplicity do seem to affect prior probabilities in the way that is characteristic of evidential reasons and not characteristic of the tobacco company's offer by the current test. This is so even if the only justification for this use of simplicity is a pragmatic justification to the effect that simpler hypotheses are easier to use.

Why Aren't Nonepistemic Reasons Incoherent?

This still leaves a problem. On the one hand, there is the Temporal Dutch Book argument that it is incoherent, and thus irrational, to be committed to a policy of updating degrees of belief in a way that does not coincide with conditionalization. On the other hand, there is the rationality of the policy of updating degrees of belief for the sorts of practical reasons that are involved in the RST example. Since it can be rational to be ready to increase one's degree of belief in a proposition simply in order to obtain the RST advertising account, something has to be wrong with the Temporal Dutch Book argument applied to this case. To see what is wrong, we need to distinguish two ways in which one might be committed to a policy p for updating one's degrees of belief given certain considerations e: (1) one is committed to thinking now that, given e the probability assignment according to p determines what bets would be fair; (2) one is committed to adopting a new probability assignment according to p given e for determining what bets one then thinks would be fair. These commitments are different. With respect to the RST account, one may have commitment 2 without having commitment 1. The Temporal Dutch Book argument is a good argument only with respect to commitment I, not with respect to commitment 2.34 A Temporal Dutch Book argument with respect to commitment 2 would overlook how the rationality of committing oneself to what one will do in hypothetical situations can depend on how likely such situations are and, more generally, on what is to be gained from undertaking such a commitment. To attach a high degree of belief to "cigarettes don't cause cancer" is

Pragmatism and Reasons for Belief

143

to commit oneself to accepting a bet that one antecedently expects would be ruinous. It can be rational to make such a commitment if one thinks it sufficiently unlikely that one will be called upon to keep the commitment. It can be rational to allow oneself to have commitments that subject one to a hypothetical Temporal Dutch Book if the chances are sufficiently low that one will have to accept the relevant bets.35 A similar point arises in discussions of deterrence: it can be rational to commit oneself to an act of retaliation that would be very costly by one's present lights if the expected gains from so committing oneself outweigh the expected loss from retaliation, discounted by the probability that retaliation will be required. 36

Explanation, Reasons, and Truth Again If, as I suppose,37 degrees of belief and probabilities are not themselves basic but derive from considerations of inference to the best explanation, then there ought to be a way of accounting for the distinction between epistemic and nonepistemic reasons that does not make such a heavy appeal to considerations about degrees of belief. I would like to be able to offer a nonprobabilistic account, but I have not found a fully satisfactory one. The best I can do is to return to the suggestion considered earlier: epistemic reasons involve an explanatory connection to the truth of what they are reasons for. A problem case is that in which RST requires of its advertising agencies that they do not just believe P, that cigarette smoking does not cause cancer, but also believe that the truth of P is part of the explanation of RST's policy with respect to agencies. The trouble is that this reason now does involve accepting a connection between the truth of P and the nonepistemic practical reasons that one has to believe that P. Here two possibilities need to be considered: (a) part of the reason for the policy is in fact that cigarettes don't cause cancer; and (b) their making the offer is explained entirely in some other way. The question is whether (a) is a better account than (b). Suppose that one's reason for concluding that (a) is a better explanation than (b) is that one will gain financially from so concluding. This makes (a) a better explanation in the sense of being more inferable, but it does not make (a) better as an explanation. It does not make (a) explain any better. On the other hand, the simplicity of a hypothesis may make an explanation better in the sense that the explanation explains better. Perhaps simpler explanations provide a better explanation in the sense that they provide a better understanding.

144

Realism/Antirealism and Epistemology

It may still be true that this is entirely a pragmatic matter. Simpler explanations may be easier to grasp in some way. Nevertheless, simplicity may play the relevant role in connecting a belief to the truth of what is believed, a role not played by expected monetary gain when RST is looking for an agency. So, maybe there is a way to salvage the idea that an epistemic reason for p must involve a relevant sort of explanatory connection between the truth of p and the reason. But this idea needs further development.

Conclusion

There is an intuitive realist distinction between epistemic and nonepistemic practical reasons for belief. This poses a problem for pragmatism, which holds that what are normally considered epistemic reasons often depend on practical considerations. Foundational considerations in the theory of subjective probability point to a solution to the problem. A commitment to inference to the best explanation does not lead to the incoherence of supposing that Temporal Dutch Books can be fair, because the commitment to inference to the best explanation helps to determine one's prior probability distribution (in the relevant sense of "prior probability"). Acceptance of nonepistemic reasons differs in not helping to determine one's prior probability distribution, and that allows a pragmatist to make the relevant distinction. I conclude that pragmatists can allow for a distinction between epistemic and nonepistemic reasons without supposing that all practical reasons for belief are nonepistemic reasons.

Notes Preparation of this essay was supported in part by a grant to Princeton University from the James S. McDonnell Foundation. 1. In his comments on an earlier version of this paper at a conference in Santa Clara, California, 28-29 February 1992, Richard Rorty persuasively argued that pragmatism rejects both realism and antirealism. 2. Louis P. Pojman, "Believing and Willing," Canadian Journal of Philosophy 15 (1985): 37-55, esp. 38--47. Pojman gives two arguments against supposing that one can normally believe something at will: a "phenomenological argument," which I dispute in this section, and an argument from "the logic of belief," namely, that it is incoherent to believe p and also believe that one believes p for reasons that do

Pragmatism and Reasons for Belief

145

not bear on the truth of p. Below I discuss the related idea that epistemic reasons have to be conceived as connected with the truth of what one believes. 3. Jack Meiland, "What Ought We to Believe, or the Ethics of Belief Revisited," American Philosophical Quarterly 17 (1980): 15-24. 4. Jonathan Bennett, "Why Is Belief Involuntary?" Analysis 50, no. 2 (March 1990): 87-107. In this article, Bennett begins with the intuition that there cannot be practical reasons simply to believe something-in contrast with practical reasons to get oneself to believe something. He then describes his various unsuccessful attempts to explain why that should be so. I believe that consideration of a range of cases undermines his intuition. 5. R. Carnap, "Empiricism, Semantics, and Ontology," Revue Internationale de Philosophie 4 (1950): 20-40. 6. W. V. Quine, "Carnap and Logical Truth," in The Ways of Paradox (Cambridge: Harvard University Press, 1966). 7. See also W. V. Quine, "Truth by Convention," in Ways of Paradox; W. v. Quine, "Two Dogmas of Empiricism," in From a Logical Point of View (Cambridge: Harvard University Press, 1953); W. V. Quine, Word and Object (Cambridge: MIT Press, 1960). I discuss some of the issues in "Quine on Meaning and Existence 1: The Death of Meaning," Review of Metaphysics 21 (1967): 124-51; "Meaning Holism Defended," Grazier Philosophische Studien 46 (1993): 163-71; and "Doubts about Conceptual Analysis," in Philosophy in Mind: The Place of Philosophy in the Study of Mind, ed. M. Michael and J. O'Leary-Hawthorne (Dordrecht: Kluwer, 1994). 8. Gilbert Harman, Change in View: Principles of Reasoning (Cambridge: MIT Press/Bradford Books, 1986); and "Rationality," in Invitation to Cognitive Science, ed. Dan Osherson, 2d ed. (Cambridge: MIT Press, 1995). 9. This sort of simplicity is relevant to all domains, because in any domain of inquiry it is necessary to reduce candidate theories to a small number. Whether some other sort of simplicity is a virtue may depend on considerations specific to a given domain, a point made by Elliott Sober in Reconstructing the Past (Cambridge: MIT Press, 1988). I offer more discussion in my paper, "Simplicity as a Pragmatic Criterion for Deciding What Hypotheses to Take Seriously," in Grue: The New Riddle of Induction, ed. Douglas Stalker (Peru, Ill.: Open Court, 1994), 153-71. 10. Alvin Goldman, "A Causal Theory of Knowing," Journal of Philosophy 64 (1967): 357-72; Gilbert Harman, "Knowledge and Explanation," in Thought (Princeton: Princeton University Press, 1973), 126-41. 11. However, I return to something like this idea at the end of the paper. 12. See, e.g., Richard Foley, The Theory of Epistemic Rationality (Cambridge: Harvard University Press, 1987). 13. Foley (Epistemic Rationality, 11.) incorrectly supposes that normal intrinsic curiosity involves this general desire. 14. Earl Conee ("The Truth Connection," Philosophy and Phenomenological Research 52 [1992]: 657-69) argues persuasively that the epistemic goal is knowledge rather than truth. But, again, normal curiosity is not the desire to know everything. 15. Conee ("Truth Connection," 664) gives a similar example.

146

Realism/Antirealism and Epistemology

16. Foley, Epistemic Rationality, 7-8. 17. C. G. Hempel, "Inductive Inconsistencies," in Aspects of Scientific Explanation and Other Essays in the Philosophy of Science (New York: Free Press, 1965). 18. See, e.g., Isaac Levi, Gambling with Truth (New York: Knopf, 1967). 19. Conee ("Truth Connection," 667) observes that a nonepistemic reason for believing something is a consideration that is not evidence for the truth of the proposition to be believed; an epistemic reason is a consideration that is evidence of the truth of that proposition. This raises the question of what is meant by "evidence of truth." I assume that, to a first approximation, evidence of the truth of a proposition is any consideration that makes the proposition more likely to be true. 20. A belief about the objective chance of p can provide an epistemic reason to believe p. But one has epistemic reasons to believe p without having beliefs about the objective chance of p. For many propositions p that one has reason to believe, it is not even clear what the objective probability of p might be. Suppose p is the proposition that cigarettes cause cancer, for example. 21. Discovery of a proof can change the probability that it is rational for one to assign to a proposition. I discuss the relevance of implication and inconsistency to inference and reasoning in Change in View, chap. 2. 22. Donald Davidson, J. J. C. McKinsey, and Patrick Suppes, "Outlines of a Formal Theory of Value, 1," Philosophy of Science 22 (1955): 146, attributing the point to Norman Kalkey of the RAND Corporation. 23. Brian Skyrms, "Higher-Order Degrees of Belief," in Prospects for Pragmatism, ed. D. H. Mellor (Cambridge: Cambridge University Press, 1980), 118-20; Brian Skyrms, "Coherence," in Scientific Inquiry in Philosophical Perspective, ed. Nicholas Rescher (Pittsburgh: University of Pittsburgh Press, 1987),225-42. 24. The marginal utility of money is diminishing if the value of an additional dollar declines as one's wealth increases. 25. Van Fraassen gives an elegant "two-minute" proof in Laws and Symmetry (Oxford: Oxford University Press, 1989), 159-60. 26. The argument is attributed to David Lews in Paul Teller, "Conditionalization and Observation," Synthese 26 (1973): 218-58. 27. This is not to say that it is always incoherent to expect to have degrees of belief that, together with one's present degrees of belief, would (if one acted on them) subject one to a Temporal Dutch Book. See David Christensen, "Clever Bookies and Coherent Beliefs," Philosophical Review 100 (1991): 229-47. Christensen takes this observation to refute the principle of reflection in Bas van Fraassen, "Belief and the Will," Journal of Philosophy 81 (1984): 235-56, and to cast doubt on Temporal Dutch Book arguments in general. Christensen bases his objection to reflection on van Fraassen's rough statement: "To satisfy the principle, the agent's present subjective probability for proposition A, on the supposition that his subjective probability for this proposition will equal r at some later time, must equal this same number r." ("Clever Bookies," 244). A better statement of reflection begins with the supposition that the agent is committed to a certain principle P for changing the probability distribution to use for determining what bets would be fair, in light of new evidence e. The agent further supposes that at a later time t the subjective probability for A determined by

Pragmatism and Reasons for Belief

147

P given the evidence up to that time will equal r. Reflection says that the agent's present subjective probability for A conditional on this supposition must also equal r. Christensen's observation is no objection to the Temporal Dutch Book argument for reflection, if the principle is understood in this way. I make a similar point about the correct formulation of conditionalization in the section below "Why Aren't Nonepistemic Reasons Incoherent?" 28. Van Fraassen, Laws and Symmetry, 160-76. Van Fraassen raises other objections to inference to the best explanation; I will not be concerned with these. 29. Nelson Goodman, Fact, Fiction, and Forecast, 4th ed. (Cambridge: Harvard University Press, 1983). 30. Goodman defines 'grue' in the following way. Select a future time F. Then something X is grue at a time T if and only if, either (1) X's color is observed at some time before F and X is green at T, or (2) X's color is not observed at some time before F and X is blue at T. A key point is that any emeralds so far observed to be green at the time of observation are also known to be grue at the time of observation. 31. Van Fraassen, "Belief and the Will," 235-56. Harman, Thought, 161-63; Harman, Change in View, chaps. 1-2. 32. In Change in View, chap. 3, I argue against taking degrees of belief to be functional aspects of human psychology. 33. Note: I say "seems to be." The issue is discussed further in the following section. 34. This is the same as what is said about reflection in note 27, above. 35. It can, of course, even be rational to be committed to an actual Dutch book, temporal or otherwise, if one's expected losses from the Dutch Book are not as great as the expected gains from the commitment. 36. Thomas Schelling, The Strategy of Conflict (Cambridge: Harvard University Press, 1960), chaps. 1-2; and Thomas Schelling, Arms and Influence (New Haven: Yale University Press, 1966), chap. 2. 37. See note 32, above.

7 Realism, Antirealism, and Pragmatism: Comments on Alston, Chisholm, Davidson, Harman, and Searle

Richard Rorty

oderick Chisholm begins his paper by saying that he assumes his principal role in this volume is "to try to make sure that traditional theory of knowledge has a fair hearing." I assume that my principal role is to make sure that pragmatism gets its innings. Pragmatists see a lot of the problems of modern philosophy as created by unnecessary dualisms (subject vs. object, subjective vs. objective, mind vs. world, etc.). They see epistemology as one of the weeds which sprout in the gaps created by such dualisms. So I take the title of this book-Realism/Antirealism and Epistemology-as an invitation to discuss the consequences of an unhappy distinction for a dubious discipline. At the beginning of his paper, Donald Davidson groups pragmatism together with idealism, verificationism, and other dubious and obsolete positions as a form of antirealism. He defines antirealism invidiously, as the attempt to "read out of existence whatever it decrees lies beyond the scope of human knowledge" and as "trying to trim reality down to fit its epistemology. " Since I persist in trying to interpret Davidson as the most sophisticated and radical of the American pragmatists, it suits my purposes to define pragmatism as the attempt to do something Davidson approves of: getting rid of the scheme-content distinction, and thus of the question of the adequacy of our minds, or our language, to reality. As I see it, pragmatists do not trim reality down to fit their epistemology. Rather, they refuse to have either an ontology (an account of content) or an epistemology (an account of the relation between content and scheme). If you forget about representational relations between mind and world and stick to causal relations between organisms and environments, as Davidson does, it is not clear that you will want, or can get, a theory of knowledge. 1 I think

R

149

150

Realism/Antirealism and Epistemology

Davidson comes pretty close to abandoning both ontological and epistemological questions-though, as I shall argue in the next section, not quite close enough. It is true that people become pragmatists for the same reason they become idealists or verificationists: they hope to frustrate the skeptic-the philosopher who says that there may be a lot that we cannot know. Idealists, verificationists, and pragmatists ask the skeptic the same rhetorical question: "Like what, for example?" They hope to force her into a dilemma: either she explains her chosen example of the unknowable in terms of the knowable (in which case it turns out to be knowable after all), or she confesses that it is ineffable (thus ceasing to be a skeptic and becoming a dogmatist).2 Idealism became unpopular because it added a constructive ontological suggestion ("reality is spiritual in nature") to this destructive rhetorical question. Verificationism made the same mistake when it asserted that everything is either a sense experience or a logical construction out of sense experiences. Pragmatists eschew such constructive suggestions and refuse to offer an ontological view. They stick to rhetorical questions. They back up the dilemma propounded above with the question: "What does the idea of seeking knowledge, or truth, add to that of collaborative attempts to achieve such concrete, easily recognizable goods as predictive power, reciprocal trust, or aesthetic bliss?" When they ask this second rhetorical question, pragmatists hope to elicit the answer. It adds nothing save the possibility of being first skeptical, and then epistemological. They hope thereby to block epistemological questions like "Can we know reality, and if so how?" and semantical successor questions like "Can we know when we are referring to reality, and if so how?"and "Is our language ever adequate to reality, and if so when?" They think nothing useful would be lost if philosophers were to stop asking such questions. Getting rid of such questions amounts to giving up on the idea of answering the skeptic by explaining how a scheme fits, or is adequate to, or is in some other way desirably in touch with, a content. Philosophers like Thomas Nagel and Colin McGinn, who are convinced that the skeptic's motives should be respected, think that this is (in Davidson's words) sour grapes; they see Davidson as throwing away content (the world, reality, truth, the goal of inquiry) in order to evade the thought that our schemes might be inadequate to it. So these philosophers typically view Davidson's abandonment of the scheme-content distinction as itself an attempt to trim reality down to size. This is why they call Davidson an antirealist and ignore his repeated protests that he rejects both antirealism and realism.

Realism, Antirealism, and Pragmatism

151

I propose that we follow McGinn (seconded by Alston at p. 60) in saying that "one is not a realist [nor an antirealist] tout court; one is a realist [or an antirealist] with respect to some or other type of subject matter." The terms 'realist' and 'antirealist' should be confined to people who are willing to take seriously the question "Is there a matter of fact about X?" for some specified X. That question neatly fuses the bad epistemological question "Can ... be known?" with the equally bad ontological question "Is ... really real?" Until we drop this fusion of bad old questions, we shall be haunted by the ghost of the scheme-content distinction: the distinction between what there is not a fact about (the subjective) and what there is a fact about (the objective). Pragmatists think that distinction should never be drawn.

Davidson on Objects, Objectivity, and the Indeterminacy of Translation

Davidson, however, flirts with this distinction in his paper for this volume. In its concluding paragraphs, after delighting us pragmatists by saying that "the only ultimate source of objectivity ... is intersubjectivity," he shocks us by making what seems to be an invidious distinction between the ontological status of numbers and of sentences. "Numbers," Davidson says, "like the objects we apply them to, lie, as it were, halfway between ourselves and others. This is what it means to say that they are objective, that they are objects." By contrast, he says, "It cannot be that way with our sentences"(122). He thus continues the Quinean line of thought according to which some things are less suited to be "objects" than others. Even though Davidson long ago waved aside Quine's preoccupation with the physicists' particles as "adventitious philosophical puritanism," and even though he says he is "no nominalist," he still seems willing to make invidious ontological distinctions. This willingness shows up again when he says that "The only object required for the existence of a belief is a believer ... being in a state does not require that there be an entity called a state that one is in" (4). Here we pragmatists want to know what practical difference it makes whether or not states are entities. To be sure, having a belief is different in many respects from having a cat, and both are different in many respects from having a lot of money, having progeny, having musical talent, and having Burmese citizenship. But does a term like 'object' or 'entity' help to explicate the relevant similarities and differences?3 Pragmatists think it simplest, and least ontological, to say that, since

152

Realism/Antirealism and Epistemology

objectivity is just intersubjectivity, anything you can talk about-that is, anything you can get some minimal amount of intersubjective agreement about-is, equally with anything else, both an entity and an object. Unicorns, citizenships, round squares, numbers, sentences, tables, quarks, propositional attitudes, and heights are, like everything else you care to mention, on a par when it comes to objectivity, factuality, and every other ontological virtue. (This, of course, is just a way of saying that there is no such thing as an ontological virtue. Paraphrasing what W. 5. Gilbert said about titles of honor, where everything's something, nothing's anything special.) Davidson's reason for saying that our sentences are not objects is a bit cryptic: You and I cannot come to agree on the relevant properties of our sentences as a preliminary to using them to interpret others, for the process of coming to such an agreement involves interpretation of the very sort we thought to prepare for. It makes no sense to ask for a common standard of interpretation, for mutual interpretation provides the only standard we have. (122)

Presumably the "relevant properties" in question are those that tell us which propositional attitudes the sentences express. If so, then Davidson's reason for saying that sentences, unlike numbers, are not out there, halfway between us and our interlocutors, can be paraphrased as follows: Although you and I can of course identify your and my utterances or inscriptions, as noises and sets of marks produced in particular circumstances and contexts, this is not an identification of them as sentences. For a sentence is something that can be held true, can express a belief. We cannot use marks and noises as reference points in our identification of propositional attitudes unless we have already identified which attitudes these sentences express. But if this paraphrase is right, then Davidson might equally well have said that marks and noises are out there, halfway between us and them, and that some of those marks and noises tum out to be sentences, also halfway between us and them. To say that some objects tum out to be this or that-tum out to have another description than the one we used at first-is just to say that they have been related to a different set of objects and thereby have been placed in a new context, fitted into another pattern. You cannot, however, identify or talk about an object without having already fitted it into some complex pattern, though any old complex pattern will do. So when Davidson says that "correct interpretation keeps track of a

Realism, Antirealism, and Pragmatism

153

complex pattern," we should respond that any and every description of anything as anything keeps track of a complex pattern. Even what Sellars called "natural-linguistic objects"-sentences described as marks and noises-are the natural-linguistic objects they are because, as Saussure pointed out, they have been placed within a complex pattern of phonic or graphical relations. 4 In short, relationality and objecthood are the result of having been placed within a complex pattern of relations to other objects. They go all the way down. You cannot use relationality to divide quasi-objects, pseudo-objects, fishy objects, from fine upstanding nonfishy objects. If you don't have any relations, you're nothing-or, at best, ineffable. 5 This amounts to saying: no describability, no objects. That dictum is the kernel of truth in what Davidson describes as the antirealists' "sour grapes" attitude: in Berkeley's claim that to be is to be perceived, James's that "reality is what it is known as," and Ayer's that we can't talk about X unless we can verify statements about X. These are three relatively awkward and misleading ways of making a more basic, better, claim: the requirements for objecthood and for know ability are no more and no less stringent than the requirements for describability and for effability. This is because all four sets of requirements are identical. Nothing effable is anything apart from its relations to other things, because you can't eff without describing, and you can't describe without relating. 6 I think that this pragmatic panrelationalism, and the consequent refusal to withhold full objecthood from any describable, should be congenial to Davidson. As far as I can see, his main reason for avoiding it is his conviction that Quine is right, and Chomsky wrong, about the" double" or "extra" indeterminacy of translation. Davidson shares Quine's conviction that there is something more to this indeterminacy than the ordinary underdetermination of theory by evidence. I do not share this conviction, and I doubt that it can be made compatible with Davidson's naturalization of semantics: his view that a theory of truth and meaning for a language is an empirical account of linguistic behavior. In his paper in this volume, Davidson says that the difference between indeterminacy of translation and undetermination of theory is that "the empirically equivalent theories it accepts for understanding an agent are not incompatible, any more than measurements of weight in pounds and kilos are incompatible theories of weight" (115). But we who follow Chomsky 7 should not accept Davidson's claim that measurement in pounds is compatible with measurement in kilos. Although it does indeed sound funny to describe these as "incompatible theories of weight," it does not sound so funny to describe them as

154

Realism/Antirealism and Epistemology

incompatible measurements. If you, on your side of the double-faced scale, keep calling out that the load weighs twenty, and I keep calling out that it weighs thirty-six, we are making prima facie incompatible statements. To be sure, all we have to do to straighten things out is to put each other's activities in context by relating those activities to the right things (i.e., the two faces of the scale) and then add, implicitly or explicitly, little phrases like 'in pounds' or 'in kilos' to the numbers we call out. But the same goes for a lot of other prima facie incompatible statements-for example, those made by Copernicans and Ptolemians. All we have to do to straighten some of these apparent incompatibilities out is to add little phrases like 'from a geocentric perspective' and 'from a heliocentric perspective.' More generally, by lavishly applying the maxim that whenever you meet a contradiction, you should make a distinction, you can render any two empirically equivalent theories compatible. 8 You can, if you really want to, treat them as stylistic variants on one another. But, it may be objected, in the one case the two parties are separated merely by alternative conventions, whereas in the other they disagree about the facts. That line of thought was enticing before Quine began attacking the conventional-empirical and language-fact distinctions, but it sounds fishy now. It should, I think, sound especially fishy to Davidson, the philosopher who generalized Quine's polemic into a criticism of all forms of the polymorphic scheme-content distinction. There is, I think, a way of saving the intuition to which Davidson is appealing without reinstating that latter distinction-though it is a way that lends no aid or comfort to the attempt to deprive states, or anything else, of the status of "objects." Instead of distinguishing language from fact, we can distinguish relative ease from relative difficulty of ironing out apparent disagreements by making distinctions. If just a few readily applicable distinctions will do the job, we shall say that the disagreement was merely apparent (and, if we have not read Quine, "merely verbal"). When it looks as if it will take a lot of distinctions to do the job, so many that it would be hard to keep them all in mind, we start saying that the disagreement is real (and, if we have not read Quine, "factual"). As the distinctions begin to pile up, we stop saying, "Do it either way, as you like," and start saying "We've got to figure out which way is right." Thus, in the case of interpretation, if only a few readily understandable transformations are required to convert translation manual A (in which 'Rome' means Rome) into translation manual B (in which 'Rome' means an area one hundred miles south of Rome), and conversely, we may say that we could get along well enough with either manual. But if a lot of hard-to-remember transformations are required, we shall say that the eas-

Realism, Antirealism, and Pragmatism

155

iest-to-use manual gets the language right and that the competitor is just weird. Pace Quine, getting something right makes just as much sense here as it does in the astrophysical case. In both cases, we should gloss 'right' not as 'true to the intrinsic nature of the objects in question' but as 'suitable tool for getting what we want'. Of course, despite my facile reference to geocentric and heliocentric perspectives (the easy way out that Osiander suggested in his introduction to Copernicus's book), you cannot really clear up the Ptolemy-Copernicus dispute that easily (except for a few pedagogical and navigational purposes).9 But all I need for my argument is the claim that a seamless continuum (of progressively increasing difficulty of reconciliation) stretches from the pounds-kilo case to the Ptolemy-Copernicus case. Different interpretations of quantum theory are nearer to the former, to the end where we say, "Oh, that's a merely philosophical difference. Who cares?" Different interpretations of the meaning of life (or of the Christian Scriptures) are closer to the latter, to the end where we insist that we have to get it right. This continuum of degrees of difficulty of reconciliation ignores traditional distinctions between convention and fact, modes of symbolization and empirical content, Natur- and Geisteswissenschaften. It does not lend itself to the suggestion that we can divide up inquiry into the part in which there are "objective grounds for choosing between competing hypotheses" and the part in which there are no such grounds. On the view I am suggesting, there is no temptation to defend, as Davidson does in his paper, the claim that attributions of propositional attitude are objective. This is because the term means no more than "intersubjective" and because nobody has ever denied that we largely agree about what justifies such attributions. As far as I can see, there is no conflict between abandoning the" double" indeterminacy of translation along the lines I am suggesting and Davidson's claim that "the concentration of psychology on the causal role of reasons rules out any hope that the basic mental concepts can be fitted into a closed system of laws" (111). There is, however, one passage in his section on anomalous monism that gives me pause: the suggestion that there is a promise "intrinsic to physics" that causal concepts such as "catalyst" can "with time and research, be supplanted by an account of the mechanism that will explain what the clumsy causal notion merely finessed." I am suspicious of this explaining-finessing distinction. Consider the biological, instead of the mental. Are the teleological notions employed by biologists clumsy causal notions, which physics promises to supplant? Are biologists who use these notions finessing

156

Realism/Antirealism and Epistemology

something that we are confident will eventually be explained? I think that although we all trust that biology will be unified with chemistry and physics in the way Carnap assumed it would,lO we should not say that, prior to this unification, we shall be finessing rather than explaining, being clumsy where we may hope to become graceful. We are, instead, explaining as well as we ever shall. The programmer's software explanation of why the computer is hung up in a loop is as graceful an explanation as we shall ever get (even though a tedious hardware explanation could be given, if anybody were fool enough to ask for it). The chemist's invocation of a catalyst to explain why the reaction took place is equally graceful and complete (even though we could explain it all over again at the molecular or atomic level, if anybody really wanted us to do so). Perhaps I am making too much of one of Davidson's incidental remarks. But I do so for the same reason that I have made a fuss about his use of 'object' and about his continuing sympathy with Quine's doctrine of an "extra" indeterminacy. The idea that the natural sciences, and especially physics, have some sort of ontological privilege goes very deep in Quine. It is clear that many of Davidson's principal differences from Quine stem from his rejection of this idea. ll But I cannot help feeling that there is still a little too much physicalism-and thus a little too much invidious ontologizing-left in Davidson's views and that those views would be more coherent if it were eliminated.

Alston on "Realist Truth"

Pragmatists base many of their arguments on the claim that it doesn't make any difference whether we think of ourselves as striving for truth or for justification, since the difference between the two searches could make no difference to what we actually do. The only way we have of detecting truth, after all, is to inquire into the justification of various truth candidates. So if Alston were right in saying that "a requirement for condition C's justifying belief B is that C renders B significantly likely to be true" and that "we have a lodestar that will guide us in distinguishing between genuine justification-makers and impostors" (81), pragmatism would be in big trouble. On a pragmatist conception, we have no way of distinguishing between genuine justification-makers and impostors, but only ways of distinguishing between what it would take to justify p to an audience A and to another audience A'. There is no such thing, for pragmatists, as justification tout court, precisely because if there were, we should have to have, per impossible, a lodestar of the sort Alston describes. Some audiences, of course, are

Realism, Antirealism, and Pragmatism

157

better or worse informed, more or less skeptical, than others. We can and do make invidious distinctions between audiences, in the manner suggested by Ginet (in a passage which Alston quotes): "the ultimate authority for this ranking must be the concurring judgments of reasonable, experienced people" (81). But what lodestar will guide us when we need to decide whether we ourselves are being reasonable, rather than gullible, in thinking a certain audience reasonable and experienced in a certain area of inquiry? Alston says that "a truth-conducivity conception of justification gives us something better" (81) than an appeal to intuition. But he does not do much to justify this claim about the function of his lodestar, nor to explain how we know a real lodestar from a pseudo-lodestar. Alston follows up his sentence about the lodestar with: At least we do if, as seems to be the case, we can often tell when conditions do render a belief probably true. If a certain type of perceptual presentation is a strong indication that there is a maple tree in front of me, then such a presentation passes at least that truth-conducivity requirement for being a justifier of that belief. (81)

But is the unanimous agreement of the learned men of 1200 on geocentrism, or of members of today's Royal Society on quantum mechanics, a strong indication that geocentrism, or quantum mechanics, is true? Does such agreement also pass the truth-conducivity requirement? Presumably not, but why not? Presumably because (as Alston says later) "for innumerable perceptual, introspective, and memorial cases, as well as for simple empirical generalizations and predictions, we are in a strong position to determine truth value" (87) whereas for more controversial matters we are not. The line between the relatively uncontroversial and the relatively controversial is the only one Alston offers as a way of demarcating areas of inquiry in which we have the help of a lodestar from areas in which, presumably, we do not. Alston does, however, seem to take seriously the claim that I may not be justified in my ordinary, noncontroversial, perceptual beliefs. "Questions about epistemic justification," he says, "are often 'essentially contested questions'" (88). But surely anybody who would contest the justifiability of my perceptually acquired beliefs about maple trees would also contest my claim to be in a strong position to determine the truth value of those beliefs. If we take one contestation seriously, why not the other? How can you be skeptical about justifiability without being equally skeptical about truthconducivity? How could we be, as Alston says we are, "in a much worse

158

Realism/Antirealism and Epistemology

position to determine Putnamian truth than we are to determine realist truth" (90)-in a worse position to decide on justifiabilty than on truth? Only if our tests for truth-value were distinct from our tests for justification would it be the case that, as Alston says, "switching from a realist to an epistemic conception of truth greatly worsens our capacity to determine the truth values of beliefs" (90). But what would be an example of such distinctness? To provide such an example, Alston would need to break out of the social practices that determine who is justified in believing what into a realm in which social practices and intuitions play no part-a realm in which the question of the question "Is this apparent truth-conducivity real?" somehow cannot arise. We would need to be shown a lodestar that - unlike our custom of taking perceptual reports of maple trees, but not of UFOs, at face value-gives us a kind of guidance that operates independently of our social practices (or else get an explanation of how the lodestar influences some social practices at some times, but not others at other times). While waiting to be shown this star, the pragmatist will continue to follow Sellars and Brandom in thinking that the sociological distinction between the relatively noncontroversial and the relatively controversial is of no help in detecting truth-conducivity. It is of no help, in demonstrating the existence of such a lodestar, to say that without one we shall be unable to distinguish between epistemic and prudential justification. Alston is quite right that if we turn our backs on truth we shall be left wondering "how to distinguish epistemic justification from other modes." But that is fine with pragmatists. They see the distinction between epistemic and prudential justification as bound up with the bad picture that held us captive and from which Davidson has helped to free us-the picture of two kinds of relations between our beliefs and reality: one merely causal and the other representational. Pragmatists do not agree with Alston that flit is obviously important for attainment of our goals that we have a correct rather than an incorrect 'map' of the things we must interact with" (59). For all the reasons that have led Davidson to say that nobody has been able to put any flesh on the metaphor of "correspondence" or on the claim (made by Alston, at p. 59) that "facts are truth makers," pragmatists think that we will never be able to explain why we interact successfully with things by the fact that we have accurately "mapped" them. Explanations of this sort of our practical successes are always "dormitive power" explanations. Just as there is no test of truth conducivity independent of a test for justifiability, so there is no test for successful mapping independent of successful coping.12 Alston says that "we can zero in on the realist concept of truth, as I have just done in terms of the T-schema, without being committed to any partic-

Realism,Antirealism, and Pragmatism

159

ular view as to what this correspondence relation is on which truth supervenes" (59). But the T-schema, unglossed by any notion of what Alston calls "something like a 'fit,' a 'match,' a 'correspondence' between statement and fact," is too uncontroversial to count as realist. It is only when we gloss the T-schema with some term like 'correspondence' or 'mapping' that Alston's claim that "Truth has to do with the relation of a potential truth bearer to a reality beyond itself' (54) begins to be controversial. Unglossed, it gives us, at most, the Davidsonian claim that an account of human behavior requires a theory that fills in the blanks in a lot of T-sentences. Glossed, it gives us the picture of a relation between a scheme (the potential truth bearers) and a content (reality) that mayor may not fit each other-the picture against which Davidson has leveled familiar objections. Davidsonians do not deny that we humans hold true what we do because of our relations to a reality "beyond" (in the sense of spatiotemporally distinct from) our beliefs. But they see the relevant relations as merely causal. So the question of accurate mapping never arises. To show that that question does arise, Alston would have to show us in more detail why the conception of truth that we get from seeing "that any instance of (10) is analytically true" gives us a "firm grasp of the realist conception of truth" (58; my emphasis). Alston, like Susan Haack,13 thinks that one can milk realism out of the T-schema while staying silent on the nature of correspondence. I follow Davidson in thinking that the T-schema is neutral on all the issues about which Alston and I disagree. Realism with correspondence seems to me unintelligible, and realism without correspondence unrecognizable as realism. More specifically, Alston thinks that anybody who agrees that "what it takes to make a statement true is the actual obtaining of what is claimed to obtain in making that statement" thereby holds a realist conception of truth. This seems to me like claiming that anybody who holds that what it takes to have a substance that puts people to sleep is to have a substance with genuine dormitive power thereby holds a realist conception of dormitivity. Until Alston comes up with distinct, independent tests for the justifiability of statements and for the actual obtaining of what is claimed to obtain in making those statements, I shall continue to think realism a pseudo-issue.

Searle on the Logical Independence of Reality and Representation

I turn now to John Searle's paper. Searle begins by saying that he will defend a view called "external realism," a view that is compatible with

160

Realism/Antirealism and Epistemology

giving up both the correspondence theory of truth and the meaningfulness of the idea of a God's-Eye View. 14 This puts him in the position of defending a form of realism that, I suspect, no one has ever attacked. If he were consistent in dismissing the topics of knowledge and truth as beside the point of a discussion of realism, it would be hard to bring Searle's paper into fruitful contact either with the other papers in this volume or with pragmatism. Fortunately, however, he deploys two quite distinct senses of the term 'external realism'-one irrelevant to knowledge and truth, the other relevant. The first sense of 'external realism' is defined by the notion of "logically constituted." Searle says that the external realist is claiming that "reality is not logically constituted by representations, that there is no logical dependence." So an external realist can be an idealist, in the sense of one who believes that all of reality consists of conscious states: "Representations are one thing, the reality represented another, and this point is true even if it should turn out that the only actual reality is representational states" (20-21). Searle goes on to say that "For the realist, it not only could have turned out that there are objects other than representations, but in fact it did turn out that way. For the antirealist, it could not have turned out that there are mind-independent objects" (21). This suggests that a "mind-[logically] independent object" is any object that is not a representation. If that suggestion is right, the antirealist, the one who believes that reality is logically constituted by representations, is somebody who thinks that anything that could possibly exist has to be a representation: that the very idea of a nonrepresentation is some sort of contradiction in terms or is somehow unintelligible. Berkeley may have come close to this view, but it is a long way from anything believed by the critics of realism whom Searle lists (Dummett, Goodman, Kuhn, Feyerabend, Putnam, Rorty, Derrida, Maturana, Varela, Winograd). What principally unites these people is Goodmanian suspicion of the idea that there is a Way the World Is. Searle's discussion of that idea is obscure. Having said, "The view that the world exists independently of our representations of it does not imply that there is a privileged vocabulary for describing it," he quickly goes on to add, "Realism is the view that there is a way that things are that is logically independent of all human representations. Realism does not say how things are but only that there is a way they are" (20). But now, with that introduction of "a way," it is no longer clear what is meant by 'logical independence'. Searle needs a definition of 'logical independence' that is neither (as most previous uses of the term have been) restricted to relations of entailment between propositions nor

Realism, Antirealism, and Pragmatism

161

restricted ad hoc for a single application. All he in fact gives us in his paper is the latter: he tells us that if you don't think it self-contradictory that there should be nonrepresentations in the universe, you thereby show that you believe that "reality is not logically constituted by representations." But this information is of no help whatever for understanding the phrase 'a way things are that is logically independent of representations'. Most of the people on Searle's list argue that a way a world is has to be a way a world is described. Undescribed worlds do not have ways they are. The people Searle wants to criticize deploy what he calls "the argument from conceptual relativity" to argue that the world does not divide itself up into bits, or what Searle calls "aspects," on its own. Many of them will agree with Putnam that elements of what we call "language" or "mind" penetrate so deeply into what we call "reality" that the very project of representing ourselves as being "mappers of something language-independent" is fatally compromised from the start."15

Many will also agree with Davidson that it is futile either to reject or accept the idea that the real and the true are "independent of our beliefs." The only evident positive sense we can make of this phrase, the only use that derives from the intentions of those who prize it, derives from the idea of correspondence, and this is an idea without content. 16

The point on which Davidson and Putnam concur with each other, and with most of the other people on Searle's list, is that causal independence is irrelevant to realism and that no other sense of 'independence' is in sight. If logical independence is to be substituted for causal independence, and if it is to be defined as a relation between items other than propositions, we are entitled to more information about this kind of independence than Searle gives us. Since Searle thinks the claim that there is no privileged vocabulary for describing reality has nothing to do with realism in the sense in which he wishes to defend it, he needs to answer the question, In what sense can a "way the world is" be "independent" of a description of the world? Some philosophers-notably Bernard Williams and David Lewiswould reply that some descriptions of the world are more closely related to the way the world is "in itself." They claim that certain vocabularies are

162

Realism/Antirealism and Epistemology

privileged. Searle rejects such claims and thinks of these philosophers as hankering after a "coincidence between truth and reality" (35) that can never be attained. He sees himself as occupying a middle ground between extremes: Williams and Lewis on the one hand and Davidson and Putnam on the other. On Searle's view, there is no way things are in themselves, but there is nevertheless enough of a representation-independent way the world is to make some systems of representation less, and others more, adequate to represent reality (27-28). I do not think that either his paper or the book of which it is part, The Social Construction of Reality, gives us enough help in understanding this intermediate position. When Searle says "some of these systems [of representation] can be used, more or less adequately, to represent" (29), people like Putnam and Davidson will protest that, if no vocabulary is privileged, no system of representation can be more or less adequate for this purpose than any other. I? The situation is not alleviated when we are told, "Carefully stated, external realism is the thesis that there is a way that things are that is independent of all representations of how things are" (39). Sometimes Searle seems to mean no more by this thesis than that a choice of descriptive vocabulary does not determine most of the truth values of the statements made in that vocabulary-a claim with which nobody disagrees. But most of the time he clearly means more than that. To explain what he does mean he would have to find a use for 'a way the world is' that is glossable by neither 'in itself' nor 'under a description'. My hunch is that no such sense can be found. There is a sense in which Searle is right that our "normal understanding" of much of what we say presupposes something like his external realism. But I also think that it is time to give up on this normal understanding. Searle agrees that normal understandings are mortal and gives an example of an earlier shift in normal understanding: the abandonment of the idea that colors are intrinsic properties of things as a result of a better scientific understanding of color perception. IS As I see it, the effect of Darwin's better understanding of the relation between humans and brutes on philosophers like Dewey, Davidson, and Dennett is analogous to the effect of the development of physiological optics on philosophers like Locke and Berkeley.I9 Just as the latter philosophers changed our normal understanding of color, so the former philosophers have been trying to change our normal understanding of truth and knowledge by getting us to drop questions like "subjective or objective?" and "representation or reality?" Such philosophers cannot appeal to intuition, as Searle does. They have to appeal to coherence, by claiming that it will be easier to make

Realism, Antirealism, and Pragmatism

163

things hang together if we give up a certain normal understanding than if we retain it. I have argued elsewhere that pragmatism is indeed counterintuitive, does indeed violate our normal understanding of our place in the universe-but is none the worse for that. 2o Searle ends his paper by saying that "philosophical theories make a tremendous difference to our lives" and that "the rejection of realism ... is an essential component of the attacks on epistemic objectivity, rationality, truth and intelligence in contemporary intellectual life" (50-51). He has expanded upon these concluding remarks in his "Rationality and Realism: What Difference Does it Make?"21 and I have replied to that paper in my "Does Academic Freedom have Philosophical Presuppositions?"22 So I shall not pursue these larger issues here.

Chisholm on the Ultimate Constituents of Reality

Roderick Chisholm opens his paper by agreeing with Michael Dummett that to explain what it is to be a realist about something, one needs the notion of "ultimate constituents of reality" (96). Pragmatists do not think that the philosophical tradition has given a clear sense to 'ultimate' in this phrase. Attempts to do so have, in recent decades, translated claims about the nonultimacy of XS into claims that talk of Xs is reducible to talk of Ys. But that sort of claim has become unpopular in the wake of Quine and Wittgenstein. Many philosophers now doubt that light is shed on philosophical disputes by attempts to specify necessary and sufficient conditions for the truth of statements. To develop Chisholm's point that questions of ontology are not questions of language, however, one would have to specify some way of detecting ultimacy other than the search for such conditions. I am not sure what this would be. Nor am I sure what makes the theory of categories that Chisholm puts forward in the section of his paper called" A Realistic Ontology" realistic (as he says on p. 96 that it is). Perhaps Chisholm's point is that any ontology that claims to have detected ultimacy is automatically realistic? If so, "realistic theory of categories" is pleonastic. After sketching his conception of the theory of knowledge, Chisholm considers "the realistic implications" of this conception (104). He says that "the epistemologist should be a realist with respect to substance" (105). But in the sense of 'realism' Chisholm is using, to be a realist with respect to X is to claim that X is among the ultimate constituents of the universe. It is easy to agree that one can hardly discuss knowledge without discussing psychological properties and can hardly do that without

164

Realism/Antirealism and Epistemology

presupposing "that there is an individual substance has that property" (104). But more would be needed to show that substances were ultimate constituents, and as I said above, I am not sure what this more would consist of. Perhaps the kind of argument against an ontology of events that one finds in Kant's "First Analogy of Experience" and again in Strawson's Individuals? When Chisholm turns to criticizing antirealism, he takes Nietzsche as an example of this view, but then he quickly turns to the Cartesian skeptic. Little more is heard of Nietzsche. But I take it that Nietzsche's relevance is that he may have been an example of the self-refuting sort of skeptic who claims to know that nothing can be known. One can agree that this sort of skeptic is, indeed, self-refuting, without seeing a connection between that sort of skepticism and a denial of realism. (Whether Nietzsche's skepticism actually was of this sort is a much disputed question.)23 Since Chisholm agrees with Dummett that realist-antirealist disputes are always "concerning a given subject matter"-whether something is or is not an ultimate constituent of reality-it is hard to fit the Cartesian skeptic, who denies the existence of knowledge, under this rubric. One can imagine such a skeptic saying that he entirely agrees with Chisholm about what the ultimate constituents of reality are and that his disagreements are confined to whether there are any instances of something-knowledgethat Chisholm would not claim was such an ultimate constituent. My own view is that it would be more revelatory of the structure of the controversies in recent philosophy in which the term 'realism' has figured to identify the antirealist not with the person "who is antirealistic about the possibility of succeeding in the epistemic enterprise" (108) but with the person who repudiates the attempt to find ultimate constituents of reality. I see no clear connection between the Cartesian skeptic and such figures as Goodman, Putnam, and Davidson. But these latter figures, with their denial that there is a Way the World Is, do seem to be denying that there is a point to formulating a theory of categories of the sort that Chisholm formulates. 24

Harman on the Epistemic-Prudential Distinction

I have two sets of doubts about Gilbert Harman's paper. The first is the seeming artificiality of his examples. When he says that I might be required to believe something in order to get an account, or to use an encyclopedia, I am inclined to protest that the most I could be required to do would be to act as if I believed this. Pragmatists who agree with Bain

Realism, Antirealism, and Pragmatism

165

that to have a belief is to have a habit of action can still distinguish the account executive who pretends to believe p from the one who really does, by reference to his actual and potential behavior outside the office. My intuition is that when I decide to believe in God, or in the innocence of my friend who has been accused of a vicious crime, or in the constancy of my lover, I am not so much acquiring a belief as exhibiting (or acquiring) faith, hope, and charity.25 When Harman suggests that belief is something you acquire passively, without decision, I tend to agree. When he counters this suggestion by saying that it is hard to distinguish deciding to believe something from adopting a hypothesis to "tryout over the long haul, in depth" (126), I am dubious. I see a big difference between finding myself believing p and finding myself willing to devote a lot of effort to test pout. I have a feeling, however, that at least some of the issues here are what pre-Quineans called 'merely verbal', and I do not want to insist on my intuitions over Harman's. So I shall rest content with saying that when we are active rather than passive, when we find ourselves trying hard not to think about something (e.g., of evidence of the absence of a benevolent and omniscient being, or of my friend's previous criminality, or of my lover's inconstancy), we seem less inclined to call what is going on "the acquisition of a belief." We grope around for other names (faith, trust, obsession, love) for our psychological state. I turn now to my second set of doubts, which concern Harman's attempt to help us pragmatists out by clarifying the distinction between "practical" and "epistemic" reasons for belief. Harman defines pragmatism as the denial that "there is any sharp distinction between purely epistemic reasons to believe something and more practical reasons." It should be noticed that whether this distinction is sharp is a different question from whether Davidson is right in saying that truth "rests ... on the affective attitudes." The latter claim says merely that you have to take desires, hopes, and fears into account in figuring out what people are saying, and thus in figuring out what T-sentences to believe. Such a view offers no support to the idea that you could, or should, abandon your belief that smoking causes cancer because of a belief that you will get money by doing so. Is there an interesting parallel between believing for the sake of the resulting money that smoking doesn't cause cancer and deserting Aristotle for Galileo for the sake of the resulting simplicity? Harman says that "the reasons for selecting the simplest hypothesis are entirely pragmatic." Many people have suggested that to grant this point and nevertheless to accept simplicity as a criterion of theory choice is to start rolling down a

166

Realism/Antirealism and Epistemology

slippery slope. At the bottom of that slope, one might have to agree with Harman that it might be rational to be ready to increase one's degree of belief in a proposition simply in order to obtain the RST advertising account. I suspect that the best reason to use simplicity in cosmological theory choice is that all the best cosmological theorists do, and that the best reason not to let our views about cancer be affected by the offer of the RST account is that only the worst account executives do. Saying that these are the best reasons in the neighborhood chimes with Goodman's claim that the rules of logic are descriptions of the inferences we and our friends typically accept, and with Kuhn's statement that there is no such thing as "the scientific method," only sedulous imitation of the scientists whom you want to grow up to be like. This replacement of epistemology with sociology and history contrasts with an idea shared by Alston and Harman: that one can evaluate rationality-where rationality is something like the ability to attain truth and avoid falsity-in a wholesale, philosophical sort of way. Pragmatists tend to eschew this idea, since a theory of rationality looks to us like another name for epistemology. So, of course, does the attempt to get beyond our intuitions and find an area-independent and context-independent distinction between epistemic and practical reasons. It is true that, as Harman says, "there seems to be an intuitively clear distinction between reasons for belief that make the belief more likely to be true and reasons for belief that merely promise some practical benefit of belief without making the belief more likely to be true" (128). But there are lots of seemingly intuitively clear distinctions that experience has shown cannot be backed up in any theoretically fruitful and interesting way-for example, the distinctions between scientific and unscientific approaches, kooks and geniuses, witches and non-witches, analytic and synthetic truths, and moral and prudential considerations. I do not think it likely that we will find an interesting area-independent or disciplineindependent way to tell in which of Harman's two classes a given reason belongs. Harman asks, "How can the intuitively realist distinction between epistemic and nonepistemic reasons be accommodated while allowing for the insights of pragmatism?" Pragmatists are short on insights but long on doubts about dualisms. They wonder about whether we need any sharper or more philosophical-sounding distinctions than the familiar, relatively noncontroversial, area-dependent and context-dependent distinctions between good reasons and bad reasons. The issue between Kuhn and his opponents, on a pragmatist view, is not about what sorts of con-

Realism, Antirealism, and Pragmatism

167

siderations it is rational to invoke when choosing between scientific theories. It is, or should be, about what sorts of considerations in fact get invoked within scientific communities when things get abnormal and messy (and in particular about what sorts of considerations led our ancestors to give up on Aristotle and go with Calileo). Once the historico-sociological task of spelling out these considerations has been accomplished in area after area, discipline after discipline, it is not clear to pragmatists that any philosophical task remains. Harman's concluding proposal about how to tell an epistemic from a nonepistemic reason seems fine for cases in which it is reasonable to ask someone whether Harman's formula on p. 139 holds for r. It works nicely for cases like a bribe offered by a tobacco company-cases in which relevance and irrelevance will strike people as intuitively clear. But I am not sure that it can be made to work in messy situations-those in which relevance is up for grabs, the sorts of cases that philosophers discuss when they argue about "rationality of theory choice." Do I really have a way of determining my prior probability distribution when r is something like "Belief in p may endanger my belief in the divine origin of the church" or "Belief in p will help my candidacy for membership in the Royal Society"? My hunch is that at least at times of scientific revolution, the situation is so confused that nobody has any clear sense of his or her prior probability distributions in the interesting cases. I suspect that sociologists of science like Shapin and Shaffer are right in thinking that it is fruitless at such times to try to disentangle the scientific from the political, the epistemic from the prudential.

Conclusion

Much of what I have said in these comments can be summed up in the claim, for which I have argued elsewhere,26 that truth is not a goal of inquiry. If this claim is right, then many of Alston's and Searle's arguments beg the crucial question. Further, it is less urgent to make the intuitive distinction that Harman is concerned to salvage. Again, the residual Quinean physicalism that I detect in Davidson can be viewed as a result of the traditional mistake of dividing up culture into the part that aims at truth-science, in a sense in which physics is the paradigm of scienceand the parts that aim at something else. That sort of physicalism seems to me the result of taking seriously the unhappy and unnecessary ontological quest for Chisholm's "ultimate constituents of reality" and of turning over authority to pronounce on

168

Realism/Antirealism and Epistemology

ultimacy from the philosopher to the physicist. 27 On my view, the results of all areas of inquiry-from microbiology to theology to literary criticism-are likely to be true, for the same uninspiring reasons that Davidson gives for thinking that most of anybody's beliefs are likely to be true. But we may nevertheless have doubts about the worth of one or another of these areas, and thus of those results. We may conclude that we no longer need any theologians, or literary critics, or microbiologists, just as we have decided in the past that we no longer needed any astrologers, monarchs, or torturers. Our decisions about which social practices, audiences, disciplines, and results to encourage and which to discourage are not made by reference to philosophical discoveries. What we call philosophical discoveries are, typically, retrospective justifications of such decisions. 28

Notes 1. John McDowell is right, I think, to emphasize the importance of Davidson's denial that we need "rational constraints on thinking and judging" and of his proposal that we "make do with nothing but causal constraints." Mind and World (Cambridge, Mass.: Harvard University Press), 1994,25. McDowell does not think that this strategy will work. I do. Michael Williams is right, I think, in saying that philosophers have adopted without argument "epistemic realism"-the view that there is a natural kind called "knowledge" to be studied. See his Unnatural Doubts (Oxford: Blackwell, 1993), for arguments against this view. 2. This dilemma is a version of what John Searle, in his paper, calls the "Ding an Sich argument." Another version is put forward in Donald Davidson, "The Very Idea of a Conceptual Scheme," in Inquiries into Truth and Interpretation (Oxford: Oxford University Press, 1984), 183-98. 3. Here I should admit that I too once made much of the distinction between objects and states. See my "Cartesian Epistemology and Changes in Ontology," in Contemporary American Philosophy, ed. John E. Smith (New York: Humanities Press, 1970),273-92. That paper was criticized by people who said that you could not solve any ontological problems by calling something a state rather than an object and that attempts to do so by Ryle and others (who talked of category mistakes and of the true logical form of sentences) were vain. I now think they were right. 4. This point that describability and being are coextensive and that sameness and difference are relative to descriptions is central to Wittgenstein's later thought. For a discussion of the importance of giving up the notion that sameness and difference may sometimes be "objective" and sometimes "subjective," see Barry Allen, Truth in Philosophy (Cambridge: Harvard University Press, 1993), esp. chap. 7 (on Wittgenstein).

Realism, Antirealism, and Pragmatism

169

5. Aristotle defined primary substance as what could get along without any relations to anything else. A familiar line of antisubstantialist thinking (common to Peirce, Whitehead, Nietzsche, and Heidegger) diagnoses the attempt to discover what Chisholm calls "ultimate constituents of reality" as the lingering residue of this unfortunate Aristotelian admiration for self-sufficiency. These antisubstantialists think Spinoza was right: if you define substance in terms of nonrelationality, there will turn out to be only one substance. This strikes them as a good reason to shrug off the notion of substance and to be suspicious of attempts to split up the universe along the lines of the two parts of subject-predicate sentences (e.g., by distinguishing objects from states). Substantialists who yearn for the start purity and self-sufficiency of the nonrelational and who are not satisfied by Spinoza's pantheism typically resort to ineffability. For, once you have defined God as existence a se, the only plausible path for theology other than pantheism is the via negativa. This was Hegel's point when he said that the concepts of Pure Being and Pure Nothing were pretty well indistinguishable. He went on to argue that, since Spinoza's atemporal One Substance was nothing without its temporal finite modes, and since in the dark night of Schelling's Absolute all cows were black, only a temporalized, dynamic, technicolor pantheism could do justice to God's plenitude. 6. This is just a way of rephrasing the challenge to the skeptic I sketched above: if you think something isn't knowable or isn't describable, either tell us more about it, so that we can see whether you are right (thereby refuting yourself) or don't (thereby appealing to ineffability and making it unnecessary to refute you). Nagel, who commands what he calls "the ambition of transcendence," thinks that an appeal to ineffability is not a concession that one has lost the dialectical game. Rather, it is a way of pointing out that Berkeley, James, Ayer, and Davidson are all playing the same unambitious, "sour grapes" sort of game. For a discussion of Nagel's view, see my "Holism, Intentionality, and the Ambition of Transcendence," in Dennett and His Critics: Demystifying Mind, ed. Bo Dahlbom (Oxford: Blackwell, 1993), 184-202. 7. I followed Chomsky'S lead in my "Indeterminacy of Translation and of Truth," Synthese 23 (March 1972): 443-62. This paper is now out of date, superseded by twenty-odd years of further discussion. But I still think I was right to argue that Quine's insistence on a "double" indeterminacy is puritanical physicalist prejudice. 8. For some advice from Davidson about how to do this, see W. V. Quine, The Pursuit of Truth (Cambridge: Harvard University Press, 1992),97-98. 9. Thomas Kuhn's Copernican Revolution (Cambridge: Harvard University Press, 1957) explains why Osiander's pragmatic ecumenicism could not succeed and tells the story of how heliocentrism required changes over the whole face of science and culture. 10. Note that this way does not require reduction, in the sense of discovering necessary and sufficient conditions for the truth of higher-level statements, phrased in lower-level vocabularies. 11. Consider, e.g., the reason Quine (Pursuit of Truth, 42) gives for refusing to "place the stimulus out where Davidson does"-to forget about impacts on nerve

170

Realism/Antirealism and Epistemology

endings and just talk about, e.g., tables and stars. He is, he says, "put off by the vagueness of the shared situations." I think that being put off by vagueness should be regarded as a residue of the old Aristotelian obsession with self-sufficiency. Davidson, it seems to me, can take vagueness in his stride. 12. I argue this point at some length in the concluding section of "Holism, Intentionality, and the Ambition of Transcendence." 13. See the exchange between Haack and myself in Rorty and the Pragmatists, ed. Herman Saatkamp (Nashville, Tenn.: Vanderbilt University Press, 1995), esp. 136-37; 148-49. 14. In chap. 9 of The Social Construction of Reality (New York: Free Press, 1995), Searle offers a defense of the correspondence theory of truth. This defense depends entirely on such claims as "We need a noun or noun phrase to name all those conditions that make the sentence true" and "We need a verb to name the variety of ways in which sentences, when true, relate to facts in a way that makes them true" (202-3). By presupposing that the right-hand side of T-sentences specifies "truth-makers," Searle begs all the questions that Davidson raises about the correspondence theory. Like Sellars, Davidson does not think that 'is true iff' relates linguistic to nonlinguistic items. Searle (along with Alston, Haack, and many others) assumes that it does do this. So he sees "realistic" implications in Tsentences that Davidson does not. The same standoff between Searle and Davidson is evident when, in an appendix to chap. 9, Searle argues against Davidson's assumption that all logically equivalent sentences name the same fact by saying that that assumption "violates our intuitions." The intuition that it violates is that we individuate facts according to which statements they make true (Social Construction of Reality, 224). Davidson, obviously, has no such intuition. 15. Hilary Putnam, Realism with a Human Face (Cambridge: Harvard University Press, 1990),28. 16. Donald Davidson, "The Structure and Content of Truth," Journal of Philosophy 87, no. 6 (June 1990): 305. 17. Davidson, but not Putnam, would refuse to accept the description of a language as a "system of representation"-but this difference is not important for present purposes. 18. My own view is that the important change brought about by modern physiological optics was to suggest that we could safely give up on the notion of "intrinsic property"-safely drop bad questions like "Are colors in things or in us?" "Are color properties intrinsic or relational" and "Is beauty in the eye of the beholder?" But Searle agrees with Locke and Berkeley in thinking the intrinsicrelational distinction of great philosophical importance. 19. I discuss the relation between Darwin, Dewey, and Davidson in "Dewey between Hegel and Darwin," in Rorty and the Pragmatists, 1-15. 20. See the concluding pages of my "Is Truth a Goal of Inquiry?: Davidson vs. Wright," Philosophical Quarterly 45, no. 180 (July 1995): 281-300. 21. DaedeIus 122 (1993): 55-83. 22. Academe 80 (1994): 52-63. 23. The best discussion of this interpretative question that I have come across

Realism, Antirealism, and Pragmatism

171

is Maudemarie Clark, Nietzsche on Truth and Philosophy (New York: Cambridge University Press, 1990). 24. Davidson has, in the past, put forward a theory of categories. See "The Method of Truth in Metaphysics," a 1977 essay reprinted in his Inquiries into Truth and Interpretation (Oxford: Oxford University Press, 1984), 199-215. But that essay answers the question, What kinds of things do we need to talk about in order to have a theory that explains how our language works? Davidson says that such questions are "the old questions of metaphysics in new dress" (Inquiries into Truth, 215). I am not sure that they are. Those old questions had many different motives. 25. I develop this point in "Religious Belief, Intellectual Responsibility, and Romance," in The Cambridge Companion to William James, ed. Ruth-Anna Putnam (Cambridge: Cambridge University Press, 1996). 26. See Rorty, "Is Truth a Goal of Inquiry?" 27. For a nice example of such deference, see Quine, Pursuit of Truth, 36. 28. I develop this last, Deweyan, point in "Philosophy and the Future," in

Rorty and the Pragmatists, 197-205.

Index

Alston, William P., 4-5, 7, 8-9, 11, 151, 156-59, 166, 167 anomalous monism, 112, 155 antirealism, 1-8, 10, 11, 12, 21-23, 95, 105-06, 109, 110, 123, 149, 160, 164 Aristotle, 165, 166 autopoetic system, 22 Ayer, A. J., 153 Bain, Alexander, 164-65 behaviorism, 6, 109 Bennett, Jonathan, 125 Berkeley, George, 21, 29, 30, 38, 45, 153, 160, 162 Blanshard, Brand, 61, 62 Bonjour, Laurence, 82-83, 85 Bradley, F. H., 61, 62 Brandom, Robert, 158 Carnap, Rudolf, 25, 99, 114, 126-27, 156 categories, 4, 12, 96, 163. See also ontology Chisholm, Roderick M., 4, 7, 9-10,12, 73-74,76,85,149,163,167-68 Chomsky, Noam, 115, 153 Cicero, 98 conditionalization, rule of, 10, 135-37, 139-40, 142 consciousness, 20-21 convergence, in science, 37-38 Copernicans, 154 Copernicus, 155 correspondence to facts, 15, 16, 58-59, 75, 158. See also truth

Darwin, Charles, 162 Davidson, Donald, 6-7, 10, 11, 149, 150,151-56,158-68 passim deconstruction, 22 Dennett, Daniel, 112-13, 119-22, 162 Derrida, Jacques, 21, 23, 160 Descartes, Rene, 97, 106, 128 Dewey, John, 61, 62, 162 Ding an sich argument (against realism), 8, 23, 33-35 doxastic entailment, 99 dualisms, 149, 166 Dummett, Michael, 4, 21, 95-96, 114, 160, 163, 164 Dutch Book, 10, 134-35. See also Temporal Dutch Book empiricism, 6, 104, 109 entrenchment, 138-39 epistemic conditions, ideal, 89-91 epistemic hierarchy, 98-100 explanation: inference to best, 137-38, 143, 144; and simplicity, 139, 142, 143-44, 165 externalism. See justification Feyerabend, Paul, 21, 160 Fodor, Jerry, 112 Foley, Richard, 74, 85 Galileo, 165, 166 Gilbert, W. S., 152 Ginet, Carl, 76, 77, 79, 81-82, 83,157

173

174

Index

Goldman, Alvin, 71, 80 Goodman, Nelson, 21, 25, 27, 28, 138-39,160, 164,166;theoryof entrenchment, 138-39 lIaack,Susan, 159 lIarman, Gilbert, 5-6, 7, 10-11, 12, 164-67 lIorwich, Paul, 58 lIume, David, 60, 104 idealism, 6, 20, 29, 35, 45, 50, 60, 109, 149, 150, 160; phenomenalist, 8, 40, 45 indeterminacy of translation or interpretation, 10, 11, 110, 112-22, 153, 155, 156 inscrutability of reference, 117-18 intentional acts, 103, 105, 111 intentional phenomena, 104, 109 intentional states, 5,120 intentionality, 16, 18, 39 internalism. See justification intersubjectivity, 152, 155 James, William, 61, 62, 153 justification: conditions of, 76-77, 82; deontological conception of, 72-73; epistemic, 69-74, 75, 85-91, 101, 156, 157, 158; and externalism, 71, 75, 78, 84-85; and internalism, 71-76, 78, 85; prudential, 158; and reliabilism, 71; and truth-conducivity, 81, 157; truth-free, 76, 78-83 Kant, Immanuel, 19,29,34,36,40, 164 knowledge: and antirealism, 105-8; having, 16-17; scientific, 17; theory of, 95, 97-108,149-50; and truth (see truth) Kuhn, Thomas, 21, 160, 166-67 Lesniewski, Stanislaw, 25 Levine, George, 23 Lewis, David, 161, 162 Locke, John, 162

materialism, 6, 109 Maturana, lIumberto, 21, 22-23, 160 McGinn, Colin, 7, 60, 150, 151 Mill, John Stuart, 1-2, 30 Moore, G. E., 36, 38-39, 41, 126, 127 Moser, Paul K., 85 Nagel, Thomas, 150 Neurath, Otto, 101 Newton, Isaac, 62 Newtonian mechanics, 88 Newtonian physics, 28, 89 Nietzsche, Friedrich, 9, 105-6, 164 normal understanding, 41---42, 45-50, 162 objectivity, 16, 17,35,151,152 ontology, 1---4,6,7,11,18,95-98,114, 149, 151, 163, 167 Parmenides, 109 Peirce, Charles S., 61, 62, 102 phenomenalism, 1-3, 6, 30, 60. See also idealism Plato, 35, 109 Platonists, 126 Pollock, John, 76, 77, 79, 82, 83 poststructuralism, 22 pragmatism, 5-6, 10, 11,61, 109, 123---44, 149-51, 156-58, 163, 164-66 probability, as rational degree of belief, 132; subjective, 10, 133-144, 167 Ptolemians, 154 Ptolemy, 155 Putnam, lIilary, 9,19,21,24,25,26,27, 51,52,61-67,74,84,85,87-9, 158-62 passim questions, internal vs. external, 126-27. See also Carnap Quine, W. v., 110-19, 127-28, 151-56, 163, 165, 167 rationality, proof of, 36 realism, 1-8, 12; alethic, 5, 8, 53-94; departmental, 60; and ethics, 2, 3;

Index

external, 3-4, 11, 15-52, 159-63; and mathematics, 2, 6, 123; metaphysical, 61; naive, 1-2,26,27,31; and philosophy of mind, 2 reality, brute vs. socially constructed, 45-49 reasons for belief: epistemic, 10, 11, 12, 123-26, 128, 129, 130-32, 140-41, 144,165,166; evidential, 128; nonepistemic, 10, 11, 123-26, 143, 166; practical, 10, 12, 165, 166; pragmatic,142 Reid, Thomas, 60, 107 relativity: argument from conceptual (against realism), 8, 23-29, 33; conceptual, 16, 17 representations, 4, 15-23,33-36,39, 43-49, 127-28, 149, 158, 160, 161 Rorty, Richard, 7, 9,11-12,21,67-69, 74, 160

Strawson, Peter, 164 subjectivity, 16, 17

Saussure, Ferdinand de, 153 scheme-content distinction, 11, 149, 150, 151, 154 scientific world view, 15-17 Searle, John R., 3-4, 7, 8, 11-12, 159-63, 167 Sellars, Wilfrid, 153, 158 sense data, 19,30,32 skepticism, 29, 32, 78, 106, 150, 157-58; Cartesian, 33, 164; Greek, 98 social constructionism, 8, 40, 45-49 Socratic inquiry, 9, 97, 103 solipsism, 49

universals, 5, 95-96

175

Tarski, Alfred, 58 Temporal Dutch Book, 10, 135-37, 140-43,144, 146-47n27. See also Dutch Book truth, 2, 5, 9, 10, 11, 35, 150, 156, 158, 165; bearers, 5, 7, 54-55, 58, 110, 159; coherence, 62; correspondence theory of, 16, 17, 18-19,58-59,62, 68, 110, 158, 160; epistemic account of, 62-67, 83-91, 158; and epistemic justification, 69-73; and epistemic reasons, 129-30; and knowledge, 75,86; turning one's back on, 9, 67-69,75-83,86,158. See also Tschema T-schema, 9, 11, 57-58, 63-66, 75, 87, 158-59

van Fraassen, Bas, 137-38 Varela, Francesco, 21, 160 verificationism, 6, 109, 149, 150 verificationist argument (against realism), 8, 23, 29-33 Wheeler, J. R., 21 Williams, Bernard, 161, 162 Winograd, Terry, 21, 22-23, 160 Wittgenstein, Ludwig, 35, 39, 50, 121, 163