Evidentialism and its Discontents

Evidentialism and its Discontents This page intentionally left blank Evidentialism and its Discontents EDITED BY T...

Author: Trent Dougherty

117 downloads 1377 Views 1MB Size Report

This content was uploaded by our users and we assume good faith they have the permission to share this book. If you own the copyright to this book and it is wrongfully on our website, we offer a simple DMCA procedure to remove your content from our site. Start by pressing the button below!

Report copyright / DMCA form

DOWNLOAD PDF

Evidentialism and its Discontents

This page intentionally left blank

Evidentialism and its Discontents EDITED BY

Trent Dougherty

1

3

Great Clarendon Street, Oxford ox2 6dp Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide in Oxford New York Auckland Cape Town Dar es Salaam Hong Kong Karachi Kuala Lumpur Madrid Melbourne Mexico City Nairobi New Delhi Shanghai Taipei Toronto With ofﬁces in Argentina Austria Brazil Chile Czech Republic France Greece Guatemala Hungary Italy Japan Poland Portugal Singapore South Korea Switzerland Thailand Turkey Ukraine Vietnam Oxford is a registered trade mark of Oxford University Press in the UK and in certain other countries Published in the United States by Oxford University Press Inc., New York # the several contributors 2011 The moral rights of the authors have been asserted Database right Oxford University Press (maker) First published 2011 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, or under terms agreed with the appropriate reprographics rights organization. Enquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above You must not circulate this book in any other binding or cover and you must impose the same condition on any acquirer British Library Cataloguing in Publication Data Data available Library of Congress Cataloging in Publication Data Data available Typeset by SPI Publisher Services, Pondicherry, India Printed in Great Britain on acid-free paper by MPG Books Group, Bodmin and King’s Lynn ISBN 978–0–19–956350–0 1 3 5 7 9 10 8 6 4 2

Acknowledgements I am grateful, ﬁrst, to my loving and supportive wife Sarah who has put up with my frenetic academic lifestyle for seventeen years this month. I am grateful to Rich and Earl for their friendship and great examples as scholars and epistemologists. I am grateful to the members of the University of Rochester graduate seminar in which Rich and Earl taught from the MS, especially Jon Matheson, Jason Rogers, and Bill Rowley. Their comments on the MS were helpful, and while I was at Rochester they were a continual source of epistemological inspiration. I am especially grateful to the members of the graduate seminar at Baylor University where I taught from the MS, especially Clifton Bryant, Ryan Byerly, Scott Cleveland, and Nate Jackson, whose summaries I consulted and adapted for some of the chapter introductions. I also beneﬁtted from the perspectives of Logan Gage, John Giannini, Kraig Martin, Ross Parker, Jonathan Rutledge, and Pete Younger. Stephen Margheim compiled the index and both he and Cameron Byrd did some proof reading. My amazing assistant Jordan Williams was invaluable at many stages. In the ﬁnal stages, my graduate assistant Chris Tweedt, by his uncanny efﬁciency, preserved any sense of sanity I started out with. Only by his efforts was I able to meet innumerable other deadlines. Finally, I wish to thank Baylor University and my colleagues for an ideal work environment. ‘Questioning Evidentialism’ by Keith DeRose here is based on his ‘Ought We to Follow our Evidence’ originally published in Philosophy and Phenomenological Research, 60/3 (May 2000), 697–706. ‘Evidentialism, Vice, and Virtue’ by Jason Baehr: originally published in Philosophy and Phenomenological Research, 78/3 (May 2000). Reprinted with kind permission of Philosophy and Phenomenological Research and Wiley/Blackwell. ‘Evidence’ by Richard Swinburne here is based on parts of his Epistemic Justiﬁcation, OUP 2001.

This page intentionally left blank

Contents Notes on Contributors

x

Introduction

1

Trent Dougherty

Part I. Evidentialism and Disagreement

15

1. Epistemological Egoism and Agent-Centered Norms

17

Michael Huemer 2. The Rational Signiﬁcance of Reﬂective Ascent

34

Jonathan Kvanvig 3. Evidentialism and the Paradox of Parity

55

Keith Lehrer

Part II. Virtue Critiques: Evidence and Inquiry

69

4. From Internalist Evidentialism to Virtue Responsibilism

71

Guy Axtell 5. Evidentialism, Vice, and Virtue

88

Jason Baehr

Part III. Evidentialism and Skepticism

103

6. Evidentialist Anti-skepticism

105

Matthias Steup 7. Evidentialism and the Great Pumpkin Objection

123

Michael Bergmann

Part IV. ‘Knowledge First’

135

8. Questioning Evidentialism

137

Keith DeRose 9. Improbable Knowing

Timothy Williamson

147

viii

CONTENTS

Part V. Concerns about the Internalist Component

165

10. Evidentialism about Knowledge

167

John Greco 11. Evidentialism and Truth

179

Richard Fumerton

Part VI. Evidence

193

12. Evidence

195

Richard Swinburne 13. Making it Evident: Evidence and Evidentness, Justiﬁcation, and Belief

207

Patrick Rysiew 14. In Defense of Propositionalism about Evidence

226

Trent Dougherty

Part VII. New Syntheses

233

15. Evidentialism, Internalism, Disjunctivism

235

Duncan Pritchard 16. Toward a Synthesis of Reliabilism and Evidentialism? Or: Evidentialism’s Troubles, Reliabilism’s Rescue Package

254

Alvin I. Goldman

Part VIII. Replies

281

17. Replies

283

Earl Conee and Richard Feldman ad Bergman ad Greco ad DeRose ad Pritchard ad Rysiew ad Swinburne ad Fumerton ad Huemer ad Kvanvig ad Goldman ad Lehrer ad Steup ad Baehr

283 285 287 289 292 294 296 298 299 302 306 308 310

CONTENTS

ad Axtell ad Williamson ad Dougherty

ix 314 316 321

Appendix A: Chart Summary of the Arguments

324

Appendix B: Bibliography of Conee and Feldman since the Publication of Evidentialism: Essays in Epistemology

330

Index

333

Notes on Contributors Guy Axtell (PhD University of Hawaii) is Assistant Professor of Philosophy at Radford University where he is the critical thinking coordinator of Radford University’s new Core A curriculum and is the editor of Knowledge, Belief, and Character: Readings in Virtue Epistemology (Rowman & Littleﬁeld 2000). Jason Baehr (PhD University of Washington-Seattle) is Associate Professor of Philosophy at Loyola Marymount University and is the author of The Inquiring Mind: On Intellectual Virtues and Virtue Epistemology (Oxford University Press 2011). Michael Bergmann (PhD Notre Dame) is Professor of Philosophy at Purdue University and is the author of Justiﬁcation Without Awareness (Oxford University Press 2006). Earl Conee (PhD Umass) is Professor of Philosophy at the University of Rochester and is an author of, among other things, Evidentialism: Essays in Epistemology (with R. Feldman, Oxford University Press 2004). Keith DeRose (PhD UCLA) is Allison Foundation Professor of Philosophy at Yale University and is the author of The Case for Contextualism: Knowledge, Skepticism and Context (Oxford University Press, 2009). Trent Dougherty (PhD Rochester) is Assistant Professor of Philosophy at Baylor University and is the editor of Evidentialism and Its Discontents (Oxford University Press, 2011). Richard Feldman (PhD Umass) is Professor of Philosophy at the University of Rochester where he is also Dean of the College of Arts and Sciences and is an author of, among other things, Evidentialism: Essays in Epistemology (with E. Conee, Oxford University Press 2004). Richard Fumerton (PhD Brown) is F. Wendell Miller Professor of Philosophy at the University of Iowa. His books include Realism and the Correspondence Theory of Truth (Rowman & Littleﬁeld 2002). Alvin Goldman (PhD Princeton) is Board of Governors Professor at Rutgers University. He is the author of, among other works, Epistemology and Cognition (Harvard University Press 1986) and Pathways to Knowledge (Oxford University Press 2002).

NOTES ON CONTRIBUTORS

xi

John Greco (PhD Brown) is Leonard and Elizabeth Eslick Chair in Philosophy at St Louis University. His books include Putting Skeptics in Their Place (Cambridge University Press 2000) and Achieving Knowledge (Cambridge University Press 2010). Michael Huemer (PhD Rutgers) is Associate Professor of Philosophy at the University of Colorado and is the author of Skepticism and the Veil of Perception (Rowman & Littleﬁeld 2001). Jonathan Kvanvig (PhD Notre Dame) is Distinguished Professor of Philosophy at Baylor University. His books include The Value of Knowledge and the Pursuit of Understanding (Cambridge University Press 2003). Keith Lehrer (PhD Brown) is the Regent’s Professor emeritus of Philosophy at the University of Arizona and currently Research Professor at the University of Miami in Florida. His books include Theory of Knowledge (Routledge 1990) and Self Trust: A Study of Reason, Knowledge and Autonomy (Oxford University Press 2007). Duncan Pritchard (PhD St Andrews) holds the Chair in Epistemology at the University of Edinburgh. His books include Epistemic Luck (Oxford University Press 2005) and The Nature and Value of Knowledge (with A. Haddock and A. Millar, Oxford University Press 2010). Patrick Rysiew (PhD University of Arizona) is Associate Professor of Philosophy at the University of Victoria and is the author of a number of articles in epistemology and Early Modern philosophy. Matthias Steup (PhD Brown) is Professor of Philosophy and Department Head at Purdue University and is the editor of Knowledge, Truth, and Duty: Essays on Epistemic Responsibility and the Ethics of Belief (Oxford University Press 2001). Richard Swinburne (Oxford) held the Nolloth Chair of the Philosophy of the Christian Religion at Oxford University from 1985 to 2002 and is the author of, among other things, Epistemic Justiﬁcation (Oxford University Press 2001). Timothy Williamson (D.Phil. Oxon.) is Wykeham Professor of Logic at Oxford University and is the author of Knowledge and Its Limits (Oxford University Press 2000).

This page intentionally left blank

Introduction Trent Dougherty

What follows is a brief account of some of the main lines of motivation—historical and conceptual—for evidentialism and their connections to chapters of this volume. Summaries of the chapters are relegated to the section introductions due to topical proximity. This introduction gets the theory on the table, in context, with motivation. It also ﬂags important research projects in epistemology related to evidence and evidentialism. At the end of their 1985 ‘Evidentialism,’—their ﬁrst piece explicitly on the subject— Earl Conee and Richard Feldman state, ‘Though we believe that the evidentialist approach to this matter that we have sketched is correct, the details must await another occasion’ (107). In the intervening years some details have been added by them (see the essays—new and collected—in Evidentialism (2004) and the Conee–Feldman bibliography in Appendix B) and by others, but the evidentialist account of epistemic justiﬁcation remains a bit sketchy. The purpose of this volume is to be an occasion for the ﬁlling in of some of the details. In what follows, we will consider both the sketch and some of the missing details, noting along the way where the issues are addressed in this volume. We begin with some conceptual and historical motivations for evidentialism.

Evidence is central to epistemology Evidence, it seems, is a central concern of epistemology. There are a number of reasons why this is so. First, consider this. If reliabilism were true and you wanted to know if the new health care bill was going to be good or bad, what would you do to ﬁnd out? If contextualism were true, and you wanted to know if a prospective neighborhood was safe, what would you do? If some kind of virtue epistemology were true, and you wanted to know whether diet soda caused cancer, what would you do? In all the cases, the answer is obvious: you’d seek out evidence. You might seek testimonial evidence, you might make simple observations, or you might even perform experiments, but in any event, you’d be seeking evidence. What’s interesting about this is that the

2

I N T RO D U C T I O N

importance of evidence is not tied to the importance of knowledge or any theory of it. For it seems that evidence is important for inquiry regardless of what theory of knowledge is true (maybe even if skepticism is true, for there still might be independent value in one’s beliefs ﬁtting one’s evidence). This is an important point, since, except for a recent surge, the trend in epistemology is to liberate many traditional concerns from the theory of knowledge proper, in recognition of their independent value. Yet there is an expected connection between knowledge and evidence if you shouldn’t believe contrary to your evidence (and most externalists admit the need for an evidential defeater condition), for it would be very strange if a theory of knowledge had the result that you could know something but you shouldn’t believe it. This suggests that in fact a theory of knowledge not only faces certain constraints concerning evidence but might even derive some of its importance from that connection. For if believing as one ought is important and knowledge entails believing as one ought, then knowing is one way to meet an important epistemic desideratum. Yet if believing as one ought requires—at least—believing what ﬁts one’s evidence, or perhaps seeking evidence rightly—then knowing will bear important entailments regarding evidence. (Several of the essays in this volume explore this theme, especially those of Axtell, Baehr, and DeRose). So consider this argument: The Knowledge Entails Evidence (KEE) argument 1. Knowing entails believing as one ought. 2. Believing as one ought entails believing what ﬁts ones evidence. 3. Thus knowing entails believing what ﬁts ones evidence. Some epistemologists—Greco and Goldman in this volume, perhaps Bergmann, too— wish to deny 3. But the argument is clearly valid, so this entails denying either 1 or 2. Denying 1 has a huge cost for the friend of the importance of knowledge, for if the two notions come apart, it seems knowledge is of the lesser epistemic value. Yet 2 is a very plausible claim. It doesn’t even say that believing what ﬁts the evidence is sufﬁcient for believing as one ought, only that it is necessary for it. Here’s another way to make essentially the same point. The Knowledge and Normativity reductio 1. Suppose for reductio that knowledge does not entail evidential justiﬁcation. 2. If 1, then possibly, S knows p though her evidence all points to not-p, even after perfectly virtuous inquiry. 3. If all S’s evidence points to not-p after perfectly virtuous inquiry, then S should believe not-p. 4. So if 1—modulo the seemingly undeniable 3—possibly, S knows p though S should believe not-p.

I N T RO D U C T I O N

3

If 4 were true, knowledge would have turned out to be far less valuable than we thought. Another way of motivating this concern is by considering the following conversation. A: S is believing as she ought. B: But S’s evidence doesn’t ﬁt what she believes. A: That’s OK, because she’s got X right. What could we ﬁll in the blank for X? X must be a very strong property, for it has to be sufﬁcient for believing as one ought even though one’s beliefs don’t ﬁt one’s own evidence. In this volume, Axtell, Baehr, and DeRose are interested in the role of inquiry in believing as one ought, but mostly to argue that it’s necessary for believing as one ought, not that it’s sufﬁcient. One might think that epistemic responsibility isn’t itself a proper subject of epistemology but rather ethics (see Dougherty 2011b for an argument for this). Others might think that ethics and epistemology overlap (see Zagzebski 1996 for a discussion of this). If either of these theses is correct, then if epistemic responsibility requires believing in accordance with one’s evidence, this shows that evidence may also have importance in ethics. Thus even if evidence were of no importance to epistemology—contrary to the KEE argument—its nature would still be of philosophical importance.

Locke and Hume’s dicta Certainly, evidence has a long tradition of importance to epistemology. Evidence has been central to epistemology for much of the time beginning in the Early Modern period. Locke gave us this afﬁrmation: The mind, if it will proceed rationally, ought to examine all the grounds of probability, and see how they make more or less, for or against any probable proposition, before it assents to or dissents from it, and upon a due balancing of the whole, reject or receive it, with a more or less ﬁrm assent, proportionably to the preponderance of the greater grounds of probability on one side or the other. (Essay Concerning Human Understanding, Book IV, Chap 15, Section 5)

Hume is a bit more economical in his dictum: A wise man, therefore, proportions his belief to the evidence. (Inquiry, ‘Of Miracles’)

This focus on evidence continues through the twentieth century with Russell (Problems of Philosophy) carrying the torch in the ﬁrst half, and Chisholm in the second half (Perceiving, Theory of Knowledge). There is an interesting commonality in these thinkers which I wish to draw attention to. Locke’s ideas, Hume’s impressions, Russell’s sense data, and Chisholm’s self-presenting states all have two things in common: they are internal and they are foundational. They are internal in the sense that they are minddependent entities, mental states, objects of consciousness, or however you want to put it. Furthermore, but less important I think, is that they are states to which one plausibly

4

I N T RO D U C T I O N

has privileged access. They are foundational in the sense that there is nothing beyond them to which we can appeal in the space of reasons. Even if—vis-a`-vis certain forms of semantic externalism or disjunctivism—mental states in illusory conditions are different in kind from mental states in veridical conditions (to which the illusory states correspond), there is still no way for the knowing subject to tell the difference. These mental states are therefore evidential bedrock; they are our cognitive home: we start from there and must always come back to them if questioned long enough. They are certain in Chisholm’s sense (p is certain for S =df ). For every q, believing p is more justiﬁed than withholding q, and believing p is at least as justiﬁed for S as is believing q (Theory of Knowledge, 3rd edn, 12). We simply have nothing closer to us, nothing more certain to us to which to appeal for reasons. These issues surface in several of the essays in this volume. They surface in consideration of how we are to deﬁne evidentialism (Pritchard), how we are to deﬁne evidence (Swinburne, Rysiew, Dougherty), and how we are to think about justiﬁcation in light of skeptical arguments (Pritchard, Fumerton, Steup). The central animus for this approach stretches all the way back through Aquinas to Aristotle who said ‘All knowledge begins in the senses’ (De Anima). So there is a strong historico-conceptual connection between evidence, internalism, and foundationalism. Most evidentialists have been internalist foundationalists, and this has seemed a natural and coherent ﬁt. Yet it is perhaps not inevitable that evidentialism or other views which give evidence its due be wed to this orbiting constellation of views. In the present volume Pritchard and Goldman attempt to develop evidentialism in ways which diverge from the tradition. (In this context it is important to draw attention to Conee’s ‘The Basic Nature of Epistemic Justiﬁcation,’ chapter 2 in Evidentialism: Essays in Epistemology. He suggests a view which appropriates aspects of internalism, externalism, foundationalism, and coherentism.)

Internalism, Empiricism, and Evidentialism Also historically associated with concern for evidence is deontology: the epistemic ought. One kind of epistemic ought concerns epistemic responsibility. It has been alleged by Plantinga (1993) that internalism has traditionally been motivated by deontology, though he notes that Conee and Feldman don’t follow this route (24). For Conee and Feldman, as well as many other evidentialists, internalism is not founded upon deontologism but rather evidentialism itself. If evidence consists in signs or indications of the way the world is, then ultimate evidence must come in the form of mental states with a certain kind of content. The phenomenal content of mental states is what gives a mental state the kind of character to indicate to a subject that something is thus-and-such. That is, internalism is derived from evidentialism because only mental states can serve as evidence in the relevant sense: ultimate indicators of the way the world is (‘ultimate’ in the sense that all other apparent correlations are derived from sense impressions with the appropriate indicative character). Our experiences—broadly construed—are the only window on the world we have. Or as Conee and Feldman put

I N T RO D U C T I O N

5

it, experience is ‘the interface of one’s mind and the rest of the world’ where ‘nothing available to introspection mediates between them and us’ (2008, 92). This is reminiscent of sense data theory. There is here a plausible line of argument leading to a union between evidentialism and a broad empiricism. The Justiﬁcation Entails Evidence argument 1. There is no justiﬁed belief without some phenomenal consciousness. Alternatively: All justiﬁed belief is at least partly explained by some aspect of phenomenal consciousness. 2. Phenomenal consciousness is what indicates to us what the world is like. 3. That which indicates to us what the world is like is evidence. 4. Thus there is no justiﬁed belief without evidence. Concerning 2, we’ve already mentioned the traditional notion of the ‘testimony of the senses.’ As Firth afﬁrms, ‘it has seemed undeniable to most empiricists that the character of my sense experience provides me with evidence’ (Firth, 1956). Non-empiricists might deny this, but they also might be nonplussed when asked for an example of just how we get proposed information about the world in another form. This traditional doctrine about the nature of empirical evidence, as I have so far described it, can be summed up in two propositions: (1) Sense experiences are evidence for perceptual judgments, and (2) Sense experiences are evidentially ultimate. That experiences (or maybe propositions describing them) are our ultimate evidence is a perfectly sensible thing to assert. There might be other senses of ‘evidence’ (Conee and Feldman mention the notion of ‘scientiﬁc evidence’ in their chapter ‘Evidence,’ in Smith’s 2008 collection of essays), but this empiricist equivalence—that evidence consists in and is limited to experience—is neither strained nor stretched. Now consider two people who have exactly the same phenomenal consciousness at all times. They are complete psychological duplicates experientially. It seems impossible for them to differ with respect to what they’re justiﬁed in believing. This gives us the following empiricist supervenience thesis. EST1 Necessarily, if S1 and S2 are total experiential duplicates at t, then for every proposition p and propositional attitude A, A is justiﬁed toward p for S1 at t iff A is justiﬁed toward p for S2 at t. But by the sensible empiricist equivalence of experience and evidence, we get this supervenience thesis. EST2 Necessarily, if S1 and S2 are total evidential duplicates at t, then for every proposition p and propositional attitude A, A is justiﬁed toward p for S1 at t iff A is justiﬁed toward p for S2 at t.

6

I N T RO D U C T I O N

But this seems equivalent to Conee and Feldman’s ES. ES The epistemic justiﬁcation of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at that time. The line of thought I’ve presented here—admittedly somewhat roughly—provides some support for thinking that empiricists—broadly construed—ought to be evidentialists, and vice versa. And note that I’ve characterized empiricism quite loosely as well. The key commitment seems to me to be the following explication of the slogan ‘all knowledge begins in the senses.’ Core Broad Empiricism (CBE) The only indications of how the world might be are ultimately derived from experiences of some kind (including memory impressions, apparent logical insights, introspection, and other traditional sources of evidence). The above discussion suggests that evidence should be a central concern in epistemology. No epistemological position has evidence more central to it than evidentialism. This is some reason to think that evidentialism is true. In fact, it might be that the biggest threat is not from arguments that it is false, but from concerns that it is trivially true until better developed. In the next section, we will look at several formulations of the thesis.

The platitude and its spellings out Conee and Feldman give both general characterizations and precise formulations of evidentialism. Evidentialism is in the ﬁrst instance an account of propositional justiﬁcation, but more of that later. Here are some general characterizations. What we call evidentialism is the view that the epistemic justiﬁcation of a belief is determined by the quality of the believer’s evidence for the belief. (Conee and Feldman, 2004, 83) Evidentialism is the view that epistemic justiﬁcation is a product of evidence. (Conee and Feldman, 2004, 83) The primary thesis holds that epistemic justiﬁcation is a function of evidence. (Feldman and Conee, 2005, 95)

Evidence ‘determines’ justiﬁcation, which is a ‘function’ or ‘product’ of it. Evidence ‘produces’ justiﬁcation. ‘Determines’ is, as I think the precise formulations will bear out, intended to convey ‘solely determines’. That is, justiﬁcation is a product of evidence and nothing else. If you want to know whether belief is justiﬁed in a proposition, the only facts you need to look at to determine an answer are the facts about what evidence a person has. As they say at one point, [A] whole body of evidence entirely settles which doxastic attitudes toward which propositions are epistemically justiﬁed in any possible circumstance. (Conee and Feldman, 2004, 101, Afterward)

I N T RO D U C T I O N

7

They also give a number of more precise formulations. Here are some of them. E S is justiﬁed in believing p at t iff S’s evidence at t on balance supports p. (Conee and Feldman, 2008, 83) EJ Doxastic attitude D toward proposition p is epistemically justiﬁed for S at t if and only if having D toward p ﬁts the evidence S has at t. (Conee and Feldman, 2004, 83) ES The epistemic justiﬁcation of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at the time. (Conee and Feldman, 2004, 101, Afterward)1

This latter version—ES—they call their ‘bedrock’ claim. Yet they go on to give another version which they call the ‘strong’ version of evidentialism (which they indicate is a strongly modalized ‘strengthening’ of E above). SE Necessarily, if S1 is justiﬁed in believing p, and E is the evidence that S1 has, then necessarily, 1) on balance E supports p, and 2) if E is the evidence that S2 has, then S2 is justiﬁed in believing p. (Conee and Feldman, 2008, 83)

In spite of its rather thorny syntax,2 it seems to express the same idea as the ‘bedrock’ supervenience thesis ES. Supervenience theses are not always informative, but the general characterizations make substantive claims that informatively reinforce the supervenience thesis. Furthermore, Conee and Feldman point out that ‘the general evidentialist perspective allows that there are contingent evidential relations which determine epistemic justiﬁcation. We endorse the strong supervenience thesis, but this claim is not entailed by the fundamental evidentialist thesis that evidence determines justiﬁcation’ (Feldman and Conee, 2005, 99). Having a number of statements of the central thesis on the table, in the next section we will look at some of the details. Conee and Feldman have admitted from the beginning that ‘There are difﬁcult questions about the concept of ﬁt, as well as about what it is for someone to have something as evidence, and of what kind of thing constitutes evidence’ (1985/2004, 84, n2). More recently, they have reiterated that the formalizations are merely ‘the bare sketch of a full theory of epistemic justiﬁcation. Among the things needed to develop the theory more fully are accounts of what evidence is, what it is for a person to have something as evidence, when a body of evidence supports a proposition, and what the basing relation is’ (2008, 84). In their 2008 ‘Evidence’, they address some of these issues, but this research project is bigger than a two-man project. It is the aim of this volume to shed further light on these issues at the center of the biggest research project in epistemic justiﬁcation. 1 A supervenience thesis is also stated in the introduction. There, the subvenient base is not the evidence itself, but ‘facts describing the evidence the person has.’ Conee and Feldman do not intend any signiﬁcant difference between them. I remain a bit uneasy about that. 2 The logical form of the formulation is this: ☐ (( J<S1,p> & H<S1,E>) ! ☐ (S<E,p> & ( H<S2,E> ! J<S2,p>))) where J is the justiﬁcation relation, H is the having-as-total-evidence relation, and S is the support relation.

8

I N T RO D U C T I O N

It is worth reﬂecting on the platitudinous nature of evidentialism in its general form and some of the most common challenges. That it is a person’s evidence which determines what it is justiﬁed for them to believe just sounds like a platitude. Conee and Feldman record that when they ﬁrst wrote on the subject explicitly in 1985 they were amazed that anyone would deny it (2004, 1; and this amazement has not faded much in the intervening quarter century). The early challenges (and, now, several of the late ones) have come from the idea that how one arrives at one’s evidence makes a difference to what one should believe (in this volume, Axtell, Baehr, and DeRose all raise these worries—ﬁrst raised prominently by Kornblith, 1983). Conee and Feldman have long insisted that this is a moral and pragmatic matter (for a defense, see Dougherty, 2011b). Other challenges to this apparent platitude come from examples of justiﬁed belief that seem to lack any evidence (Goldman and Greco both raise these issues in this volume). However, one way to frame these debates is not about whether evidentialism is true but rather how it is to be understood, how we are to understand the nature of evidence, having it, and its supporting a proposition or attitude (Dougherty, Goldman, Pritchard, Rysiew, and Swinburne all try to make some headway in this regard directly, and Huemer, Kvanvig, and Lehrer do so indirectly). My suggestion is that many of the apparent challenges to evidentialism are either peripheral to the focus of evidentialism—more about this in the sub-section concerning justiﬁcation—or are better framed as debates about the cluster of questions concerning the nature of evidence, the evidential support relation, and what it takes to instantiate it. Many of the objections can be reframed as arguments for speciﬁc understandings of these concepts.

Filling out the platitude We have looked at some historical and conceptual roots for a general evidentialism. We’ve looked at several versions of Conee and Feldman’s speciﬁcation of it. Finally, we’ll zoom in a bit more to look at some of the most important concepts in the theory. Even if one were to accept evidentialism as a truism, there would be much to do in ﬁlling out a substantive theory of epistemic justiﬁcation in evidentialist terms. As Feldman and Conee note, ‘A fully worked out version of evidentialism must spell out answers to some admittedly difﬁcult questions’ (2005, 99). We’ve seen several of those issues, all of which are addressed in some form or another in this volume, but one central concern is the precise nature not just of the analysans but of the analysandum. What is this justiﬁcation which is being analyzed? One issue must now be confronted. Conee and Feldman are explicit that they ‘do not offer EJ as an analysis’ (1985/2004, 83). What they have in mind here is that they are not offering a conceptual analysis. One could have, they allow, the concept of epistemic justiﬁcation without having all the concepts used in the account of it. Also, they don’t seem to be offering a reductive analysis. At least they never claim to. But there is little harm in continuing to speak of their account of justiﬁcation as an analysis of sorts

I N T RO D U C T I O N

9

as long as we keep these distinctions in mind. I want to look a bit more at the issue of reductionism in the analysis before moving on, in the ﬁnal section, to a closer look at the analysandum.

Reductionism Chisholm’s epistemology wasn’t reductive, because he took the concept more reasonable than or at times the notion of epistemic value as basic, undeﬁned notions which anchored his whole system (Chisholm, 1977, 7, 12ff ). However, it seems to me there is hope in Chisholm for a reductive account. Chisholm—who had a tremendous inﬂuence on Feldman in particular—had two kinds of epistemic principles (Chisholm, 1989, 61ff ). Formal epistemic principles connected normative principles to one another. For example ‘If something is self-evident, then it is beyond a reasonable doubt.’ Material epistemic principles connected descriptive facts to prescriptive facts. The descriptive facts over which Chisholm’s material epistemic principles were deﬁned were facts about a subject’s experiences. For example, ‘If it seems to S that there’s an a that’s F, then it’s certain for S that it seems to her that there’s an a that’s F.’ It is plausible that there are materials here for a reductive account of the normative language in the evidentialist principle. I think EJ above—Doxastic attitude D toward proposition p is epistemically justiﬁed for S at t if and only if having D toward p ﬁts the evidence S has at t—offers the best balance of breadth and informativeness of all the principles listed (in the Afterward to ‘Evidentialism’ in Evidentialism: Essays in Epistemology, Feldman and Conee discuss the relative merits of the different formulations). Both ‘ﬁts’ and ‘evidence’ seem to be normative terms. Both, however, can plausibly be deﬁned in non-normative terms. Conee and Feldman—like Chisholm and most empiricists—have taken experience as ‘ultimate’ in some way (a better term than ‘basic’ because an empiricist coherentist could still take experience to be the ﬁnal arbiter of justiﬁcation). So any locution referring to a subject’s evidence can be translated into a locution referring only to the subject’s experiences. Of course, the Problem of the Speckled Hen (see the references in Dougherty’s 2010 ‘Evidence’ in Oxford Bibliographies Online) shows that not all experiences count as evidence. Feldman has made some headway on this problem in his ‘The Justiﬁcation of Introspective Beliefs’ in Evidentialism: Essays in Epistemology (also titled ‘Foundational Justiﬁcation’ in Sosa and his Critics. See also Bonjour’s reply in Epistemic Justiﬁcation, 2003, 190ff ). Specifying just which experiences constitute evidence, even granting the thesis that experiences constitute ultimate evidence is one of the major research projects in evidentialism in particular and epistemology more broadly, since even if evidentialism were false, evidence would still be a very important concept. ‘Fits’ can also be given a non-normative reading. One relevant one is the concept of a match. A picture can match that which it pictures. It is, in this respect, a good picture. Yet there is no more to its matching than there being the right kind of function from regions of the picture to portions of the pictured object. In a similar way, our

10

I N T RO D U C T I O N

experiences can match beliefs in terms of their content (the question of the contentfulness of experience is a vexed one, and Conee and Feldman have not committed themselves explicitly to a view here. Still, their examples often ﬁt the paradigm I am currently discussing. The nature of experiential content is an important research project in evidentialism in particular and epistemology more generally). The experience as of a red mug on a table matches well the belief that there is a red mug on a table. The experience as of a cat on the roof matches well the belief that there’s a cat on the roof. This might seem like a narrow band of belief, but really it’s not. For our experiences are not limited to sensuous experiences. There is something it’s like to intuit that the Axiom of Choice is true or that causation is transitive or that a proof is valid. The belief that it is so matches well these experiences. We needn’t limit the scope to sensory states when Chisholm says ‘it is reasonable to interpret ways of being appeared to as being reports from the outside . . . one’s being appeared to in a certain way . . . one interprets as being a sign of some external fact’ (1989, 67). The ‘interprets’ here is to be taken loosely of course: it is not as if we typically even think about it. To borrow a phrase, our experiences are ‘seen as’ signs of extra-mental facts. On this model—which we might call a form of ‘Seeming Evidentialism’ with Conee’s ‘First Things First’ in Evidentialism: Essays in Epistemology—seemings that p provide reasons for believing that p. This is in line with a long tradition of commonsense epistemology. Chisholm explicitly called his view ‘commonsensism’ (1989, 63), Swinburne’s credulism (see esp. his 2001, 135ff, esp. 141) is in this vein, as is Huemer’s phenomenal conservatism (2001, esp. 99) (Huemer’s, Swinburne’s, and Rysiew’s essays herein are relevant to this topic, Rysiew explicitly discussing Thomas Reid). The commonsense maxim here is this: CSE If it seems to me that p, then I have a reason to believe p. We could say ‘epistemic reason’ or ‘item of evidence for’ if we wish to be more explicit. Now the reason you get for p from its seeming to you that p might be easily defeated by other considerations. Thinking in binary terms whether the attitude of believing best ﬁts p will be a matter of whether there is more reason to believe that p or not. How to weigh reasons is a vexed matter and an important research project. Thinking in graded terms, how much certainty you ought to have in p varies in proportion as your reasons to believe p outweigh the reasons against p (with withholding holding at the balance point). When we combined the matching notion of ‘ﬁts’ with the experiential model of evidence, we get a plausible and plausibly appropriately reductive account of justiﬁcation. The account of justiﬁcation is one in which the term ‘justiﬁcation’ acquires connotations along the lines of how the term is used when text on a page is perfectly aligned with the side of the page: it is a perfect ﬁt (this illustration was suggested by Conee). The question of the value of this notion is taken up in the responses, especially in Conee and Feldman’s response to Baehr and Axtell.

I N T RO D U C T I O N

11

Propositionalism, knowledge, and basing The ﬁnal detail to be considered concerns the distinction between propositional justiﬁcation and doxastic justiﬁcation. Evidentialism is centrally a theory about propositional justiﬁcation, and then other notions of justiﬁcation are deﬁned in terms of it. It is hard to deﬁne propositional justiﬁcation directly other than by saying that it is a relation of ﬁt between a proposition and a body of evidence. Propositions are the bearers of this relation. Given a body of evidence—or a description of an individual’s experiences—there will be a set of propositions that ﬁt it (and perhaps to various degrees). Thus we could think of propositional justiﬁcation as represented by a function from evidence sets to propositions, a set of ordered pairs with one slot for each proposition and another slot for each possible set of evidence (or, degree theoretically, as functions from these ordered pairs into the ﬁrst unit interval, ordered triples with a slot added for a degree measure). Clearly, inﬁnitely many propositions are justiﬁed on any body of evidence (many of them trivial logical consequences). So, clearly, one needn’t believe p for p to be justiﬁed for one to believe. Sometimes the phrase ‘S is justiﬁed in believing p’ is used instead of ‘p is justiﬁed for S’ even though the former is easily confused with ‘S justiﬁably (sometimes ‘justiﬁedly’ since the modal could be misleading) believes p.’ We can either agree to stick to p as the grammatical subject for propositional justiﬁcation or use only such clear phraseology as ‘S’s belief that p is justiﬁed’ for doxastic justiﬁcation which, unlike propositional justiﬁcation, entails belief. Note that we can’t say—when p is justiﬁed for S because of S’s evidence E—that if S were to form the belief B that p, B would be justiﬁed. For it might be that were S to form the belief B that p, B would be formed in a way that bears no relevant relation to E. If you stipulate that S’s coming to believe that p was suitably related to E, then you’ll have a case of doxastic justiﬁcation (or what Conee and Feldman call ‘wellfoundedness’ (1985/2004, 93). The relation in question is called the basing relation, and the formula for doxastic justiﬁcation is propositional justiﬁcation plus belief plus proper basing (not just propositional justiﬁcation plus belief ). An agent could then be assessed on the basis of what proportion of her beliefs are properly based. This would provide a uniﬁed account of the justiﬁcation of propositions, beliefs, and persons, and evidentialism is a theory ﬁrst and foremost of the root notion. This is one way in which, though narrow in focus, evidentialism can be of value in epistemology. (This is essentially the picture painted by Kvanvig and Menzel in ‘The Basic Nature of Justiﬁcation,’ 1990). The nature of the basing relation remains an important research project (see Korcz, 2010). This is especially so, since the basing relation is an important link connecting evidential justiﬁcation to knowledge, to which we will now turn to complete our opinionated survey of the role of evidence and evidentialism in epistemology.

12

I N T RO D U C T I O N

Knowledge and evidentialism If knowledge required infallibility, we’d have precious little knowledge, for there is precious little, if anything, about which we are infallible. This has led most epistemologists to be fallibilists about knowledge (for an opinionated yet historical survey of fallibilism, see Dougherty, 2011a). That means that it will be very difﬁcult, if not impossible, to deﬁne a single, uniﬁed property that transforms true belief into knowledge. Exercises of virtue and reliable processes can sometimes lead to false belief, so it looks like internalists and externalists must both be satisﬁed with deﬁning a root notion of knowledge and leaving the rest for Gettier. Thus we will bracket that kind of concern for the remainder of this brief discussion. The simplest evidentialist picture of knowledge is very simple indeed. The conceptual core is that when one forms a true belief because they were appropriately responsive to their evidence, then they know. Perceptual knowledge is true perceptual belief appropriately responsive to perceptual evidence, the ‘testimony of the senses.’ Memory knowledge is true belief appropriately responsive to memory impressions, what we might call the ‘testimony of memory.’ A priori knowledge is true belief about a priori matters appropriately responsive to apparent insights. One might just wish to quantify over basic faculties and endorse the following thesis. Faculty Evidentialism For any basic faculty F of S, if F says that p, then S has a reason to believe that p, and nothing is a reason to believe unless it bears a suitable connection to the deliverances of a basic faculty. One then has knowledge that p when the balance of one’s reasons is sufﬁciently heavily tipped in favor of the true belief that p, and the main reason one holds that belief is because of those reasons. By quantifying over faculties we both avoid the Problem of Scatter leveled against Chisholm (Sosa in BonJour and Sosa, 2003, 164–5), and provide enough generality to cover creatures—be they aliens or animals, mutants or marvels— who have faculties which function very differently from those of current, ordinary humans (who knows what faculties we might gain or might have lost?). And note that the same broad faculty types are capable of reporting to the subject quite different contents depending on the design of the subject. For example, a dog’s olfactory faculty may well deliver such reports as ‘This person is scared’ where the human olfactory faculty could be capable of no such thing. An expert’s visual faculty could deliver the report ‘That’s an elm’ where the novice’s could not. Though the expert and the novice might have the same sensuous experience, they wouldn’t have the same total experience, because something in the expert’s past experiences causes him to have a different experience in the present observation. When the expert hosts the exact same sensuous qualia there is an additional experience. The expert sees the object as an elm. This difference in total experiential/evidential proﬁle explains the difference in their justiﬁcation regarding the thing they both see and have the same visual experience of.

I N T RO D U C T I O N

13

This is a simple theory that accords well with paradigm cases of knowledge and common sense as well as explaining what’s going on in some difﬁcult cases. It is an evidentialist theory of knowledge in that evidence is the core notion and the evidentialist thesis about propositional epistemic justiﬁcation gives an account of the core conceptual content. As Conee and Feldman say about well-foundedness, ‘It is an evidentialist notion because its application depends on two matters of evidence—the evidence one has, and the evidence one uses in forming the attitude.’ (This picture is an explication and development of some brief remarks by Feldman in his great little introduction Epistemology (2003: 125)). Though Conee and Feldman think that their theory of epistemic justiﬁcation explicates one common usage of ‘reasonable’ which is independent of the theory of knowledge, that this notion is capable of being the kind of justiﬁcation suited for playing its own role in transforming true belief into knowledge puts certain constraints on how the theory is to be understood, especially its scope.

Conclusion At the end of their most recent account of some of the details of evidentialism, Conee and Feldman say ‘Much work remains to be done. We hope to have done enough in support of our preferred version of evidentialism to show that this further work is worth doing’ (2008, 104). As I mentioned above, there is indeed much work to be done, far more than two men can do in a lifetime. I hope this volume will show another generation of graduate students and junior scholars that, truly, this further work is worth doing. In similar fashion, they have recently said ‘Evidentialism has many virtues beyond the ones cited here, perhaps thousands of them. There is no need to try to list them all, but the virtues of evidentialism do need a bit of extolling. There are some tough problems involved in completing the theory, including the problems described above of giving complete accounts of perceptual and memorial evidence. Nevertheless, the virtues of evidentialism provide encouragement enough to continue seeking solutions to the remaining problems until they are found’ (Feldman and Conee, 2005, 108). I am hopeful that solutions can be found, and that, indeed, some of them can be found between the covers of this book.

References BonJour, Laurence and Ernest Sosa. 2003. Epistemic Justiﬁcation. Oxford: Blackwell. Chisholm, Roderick. 1957. Perceiving: A Philosophical Study. Ithaca: Cornell University Press. ——. 1977. Theory of Knowledge, 2nd edn. Harlow: Prentice Hall. ——. 1989. Theory of Knowledge, 3rd edn. Harlow: Prentice Hall. Conee, Earl and Richard Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Oxford University Press.

14

I N T RO D U C T I O N

——. 2008. ‘Evidence,’ in Quentin Smith (ed.), Epistemology: New Essays. Oxford: Oxford University Press. Dougherty, Trent. 2010. ‘Evidence,’ in Oxford Bibliographies Online. Oxford: Oxford University Press. . ——. 2011a. ‘Fallibilism,’ in The Routledge Companion to Epistemology. Abingdon: Routledge, 131–43. ——. 2011b. ‘Reducing Responsibility: An Evidentialist Account of Epistemic Blame,’ The European Journal of Philosophy forthcoming. Feldman, Richard. 2003. Epistemology. Upper Saddle River, NJ: Prentice Hall. —— and Earl Conee. 1985. ‘Evidentialism,’ Philosophical Studies, 48: 15–34. —— and ——. 2005. ‘Some Virtues of Evidentialism,’ Veritas, 50(4): 95–108. Firth, Roderick. 1956. ‘Ultimate Evidence,’ The Journal of Philosophy, 53/23: 736. Greco, John. 2004. Ernest Sosa and His Critics. Cambridge: Cambridge University Press. Huemer, Michael. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littleﬁeld. Korcz, Keith Allen. 2010. ‘The Epistemic Basing Relation’, Edward N. Zalta (ed.), The Stanford Encyclopedia of Philosophy (Summer), . Kornblith, Hilary. 1983. ‘Justiﬁed Belief and Epistemically Responsible Action.’ Philosophical Review, 92: 33–48. Kvanvig, Jonathan and Christopher Menzel. 1990. ‘The Basic Notion of Justiﬁcation,’ Philosophical Studies, 59: 235–61. Plantinga, Alvin. 1993. Warrant: The Current Debate. Oxford: Oxford University Press. Russsell, Bertrand. 1959 (Orig. 1912). Problems of Philosophy. Oxford: Oxford University Press. Sosa, Ernest. 1991. Knowledge in Perspective, Cambridge: Cambridge University Press. Swinburne, Richard. 2001. Epistemic Justiﬁcation. Oxford: Oxford University Press. Zagzebski, Linda. 1996. Virtues of the Mind. Cambridge: Cambridge University Press.

PART I

Evidentialism and Disagreement The three essays in this section deal with an emerging area of interest in epistemology, largely pioneered by Richard Feldman. His ‘Plantinga on Exclusivism’ (Faith and Philosophy 20 (2003): 95–119) was one of the ﬁrst statements by a major epistemologist directly addressing the question of the epistemological signiﬁcance of disagreement. This was followed in rapid succession by several other pieces. Feldman, Richard. 2005. ‘Respecting the Evidence,’ Philosophical Perspectives 19 (2005): 95–119. ——. 2006. ‘Epistemological Puzzles about Disagreement,’ in S. Hetherington (ed.), Epistemology Futures. New York: Oxford University Press. ——. 2007. ‘Reasonable Religious Disagreements,’ in L. Antony (ed.), Philosophers without Gods. Oxford: Oxford University Press.

In this time, the area has burgeoned, resulting in a volume edited by Feldman with Ted Warﬁeld, Disagreement (Oxford Universitiy Press, 2010). This volume contains an essay by Earl Conee, which might be a rare sighting of epistemological disagreement between the pair: ‘Rational Disagreement Defended.’ Michael Huemer’s aim in his essay is to defend a fully agent-centered epistemological perspective. An agent-centered epistemological perspective is opposed to an agentneutral view. Agent Neutrality: Necessarily, for any S, T, and C, if S’s satisfying condition C would confer prima facie justiﬁcation for S to believe that p, then S’s knowing that T satisﬁes C would confer equal prima facie justiﬁcation for S to believe that p. Agent Centeredness: Possibly, for some S, T, and C, S’s satisfying condition C would confer prima facie justiﬁcation for S to believe that p, but S’s knowing that T satisﬁes C would confer less prima facie justiﬁcation or no prima facie justiﬁcation for S to believe that p. Huemer claims that a consequence of his agent-centered view is that two subjects with perfect knowledge of one another’s epistemically relevant states who make no procedural errors in belief formation may still rationally disagree. They can rationally disagree

16

E VI D E NT IA L I SM A N D D I S AG REE M E N T

because each might assign different weight to the evidence of the other person than that other person may assign—evidence is not agent-neutral. Huemer also defends the stronger thesis that only one’s own mental states can provide justiﬁcation for one for believing anything. The fact that someone else has evidence that supports p by itself gives me no reason to believe that p. Huemer suggests that this doesn’t sit well with mentalist evidentialism as formulated, and that it favors his phenomenal conservatism. Part of the motivation for Huemer’s view comes from his noting analogies between ethics and epistemology, particularly between ethical egoism and agent-centered epistemology. Despite this analogy between ethical egoism and agent-centered epistemology, he argues there is an important asymmetry between ethics and epistemology. Jonathan Kvanvig argues that rational disagreement between epistemic peers is unproblematic. He does so in two steps. First, he urges that the contrary view faces a number of serious objections. Second, he proposes a plausible model of disagreement according to which rational disagreement between epistemic peers is sensible. Kvanvig focuses on the cognitive results of reﬂection and the sufﬁciency of justiﬁcation for closing inquiry. He raises some technical difﬁculties for some kinds of molliﬁcationism—a view bearing considerable similarity to the view defended by Feldman—and even levels charges of self-defeat. He suggests there is something other than evidence which goes into justiﬁcation: what you make of the evidence. Keith Lehrer argues that, after entering into the condition of parity concerning p, one can reasonably continue to accept that p. What it is to be in the condition of parity with respect to p is just this: that one disagrees with an interlocutor with whom one reasonably believes oneself to share all evidence concerning p except for the belief that p itself, and that one judges one’s interlocutor just as trustworthy as oneself with regard to evaluating the evidence bearing on p. What Lehrer calls the condition of parity is thus quite akin to what is called by others disagreement between epistemic peers. A key concept in Lehrer’s treatment is the idea that self-trust is the keystone of rationality. In order to enter the parity condition at all, one must, according to Lehrer, accept that one is reasonable in what one accepts. And once one realizes this, it is clear that one can reﬂectively take oneself to be reasonable in accepting p, and this latter fact will provide the evidential support needed to support p over not-p, making accepting p reasonable even after the condition of parity. Guy Axtell’s pragmatist critique of evidentialism at the beginning of the next section ends with a critique of the consequences of evidentialism for democracy, so it can be considered an honorary member of this section.

1 Epistemological Egoism and Agent-Centered Norms Michael Huemer

1.1 The idea of agent-centered norms In ethics, an agent-centered norm is one that requires agents to value their own prospective performance of some type of action differently from someone else’s performance of the same type of action.1 I use the term ‘value’ broadly here. For instance, it is widely believed that it is wrong to commit a murder, even if doing so would prevent someone else from committing two similar murders. Perhaps there is some number n such that one ought rather commit one murder than allow n murders to be committed by others— perhaps, for example, one would be justiﬁed in killing an innocent person if doing so would prevent one million innocent people from being killed in a similar manner. But most people will agree that, if there is such a number, it is greater than two. In that sense, most of us believe that one should assign greater negative weight to one’s own unjust killings than one assigns to unjust killings performed by others. From one’s own point of view, it matters not just what actions are performed, but who performs them. Does a similar notion apply in epistemology? An agent-centered epistemic norm would be an epistemological principle requiring agents to assign different evidential value to their own experiences or other epistemically relevant states from the value they should assign to the qualitatively similar states of someone else.2 Suppose that any epistemic agent who satisﬁes condition C thereby has prima facie justiﬁcation for believing p. If this source of justiﬁcation is governed by an agent-neutral norm, then an agent who merely knows that someone else satisﬁes C would thereby have equal prima facie grounds for believing p. If, on the other hand, the source of justiﬁcation is governed by an agent-centered norm, then such an agent might have less justiﬁcation or even no justiﬁcation at all for believing p, even prima facie. I shall call the view that all

1 See Schefﬂer’s (1982: 2–4) discussion of agent-centered restrictions and prerogatives, though Schefﬂer gives a somewhat different characterization from mine. 2 I refer to believers as ‘epistemic agents’ to emphasize the ethical analogy; however, my arguments herein do not turn on construing believing as an action in the ordinary sense.

18

E VI D E NT IA L I SM A N D D I S AG REE M E N T

norms of epistemic justiﬁcation are agent-neutral, ‘Agent Neutrality,’ and I shall call the contrary view ‘Agent Centeredness’: Agent Neutrality: Necessarily, for any S, T, and C, if S’s satisfying condition C would confer prima facie justiﬁcation for S to believe that P, then S’s knowing that T satisﬁes C would confer equal prima facie justiﬁcation for S to believe that p.3 Agent Centeredness: Possibly, for some S, T, and C, S’s satisfying condition C would confer prima facie justiﬁcation for S to believe that p, but S’s knowing that T satisﬁes C would confer less prima facie justiﬁcation or no prima facie justiﬁcation for S to believe that p.

Agent-centered norms must be distinguished from a kind of agent-relativity that is uncontroversial. Suppose you are in Paris when it starts to rain this afternoon. You acquire excellent grounds for believing that it is raining in Paris. But I, stuck in Denver with no communication from Paris, have no grounds for any particular opinion about the weather in Paris. This illustrates the uncontroversial sense in which justiﬁcation is agent-relative. How does the agent-centeredness with which I am concerned differ from the preceding case? In the preceding case, I fail to believe that it is raining in Paris, not because I attribute less evidential weight to your experiences than I do to my own, but simply because I am unaware of your experiences. If I knew that you were having an experience as of rain in Paris, I would justiﬁably accept that it was raining in Paris. Thus, no agent-centeredness is apparent in this case. Here is a sort of case in which agent-centeredness might seem appropriate, albeit still controversial. Assume that intellectual intuitions are a source of prima facie epistemic justiﬁcation. If I intuit that p, then in the absence of defeaters, I am to some degree justiﬁed in believing that p. But what if I merely know that someone else intuits that p? Suppose that, when considering the proposition, ‘Potential persons have a right to life,’ I have no intuition about it. But I know that you have the intuition that potential persons have a right to life. Do I thereby have as much justiﬁcation for believing that potential persons have a right to life as I would if the intuition were mine? I suspect that most people would say not. In any case, very few people would place as much conﬁdence in another person’s intuitions as they would place in their own. It therefore seems that most people hold an agent-centered view of intuitive justiﬁcation. We should set aside certain kinds of reasons for assigning less weight to other people’s reported experiences and other epistemically relevant states. Suppose that I attribute less weight to your reported observations than I do to my own observations, because I am less certain of what your actual observations were—after all, you might be lying, or I might be misunderstanding you. This would not evince any adherence on my part to

3 To make agent-neutrality initially plausible, I state the view in terms of prima facie justiﬁcation rather than ultima facie justiﬁcation. It is obviously false that, if I know you satisfy some sufﬁcient condition for being ultima facie justiﬁed in believing that p, then I am thereby ultima facie justiﬁed in believing that p myself—for I might have defeaters for p that I know you lack. For present purposes, prima facie justiﬁcation for believing p is understood simply as a condition that would render one justiﬁed in believing that p if one had no defeaters. It is not required that prima facie justiﬁcation be non-inferential.

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

19

an agent-centered norm. Consider an analogous case in ethics: suppose I am offered the chance to kill one innocent person to prevent two innocent people from being killed in an equally bad way. Suppose that my reason for declining is that I am not convinced that the other two killings will really occur. This rationale for passing up the opportunity does not reﬂect any agent-centered norm. Only if I forego the opportunity to kill when convinced that I really would prevent two similar killings do I manifest adherence to an agentcentered norm, for it is only in that case that I show that I attach greater negative weight to my own killing than I attach to someone else’s killing. To test whether we accept ethical agent-centeredness, then, we should stipulate that an agent knows for certain that, for example, two killings will occur unless he kills one person, and then ask whether we would consider the one killing justiﬁed. Similarly, to test whether we endorse epistemological agent-centeredness, we should stipulate that an agent knows for certain that another person is in a particular, epistemically relevant state, and then ask whether we would consider the agent to have the same degree of justiﬁcation (barring defeaters), for believing the same proposition, as the person who is actually in the state. Agent-centeredness should not be mistaken for a naive epistemological egotism. Again, an ethical analogy is instructive: the ethical egoist holds that one ought always to serve one’s own interests. But the ethical egoist is not an egotist: he does not take himself to be somehow special, or objectively more important than everyone else. Thus, the egoist does not say, ‘Everyone ought to serve my interests’; the egoist says, ‘Everyone ought to serve their own interests.’ Similarly, the agent-centrist in epistemology does not say, ‘Everyone ought to base their beliefs on my experiences’; the agent-centrist says, ‘Everyone ought to base their beliefs on their own experiences.’ This leads to a key, controversial implication of agent-centeredness, which differentiates agent-centered epistemological views from agent-neutral views. Suppose two subjects each have perfect (that is, complete and absolutely certain) knowledge of one another’s epistemically relevant states (sensory experiences, memories, intuitions, or whatever is relevant to what one is justiﬁed in believing). Suppose that neither party makes any procedural error in forming beliefs: for instance, neither party makes any oversights or incorrect inferences, neither party incorrectly weighs two pieces of evidence, and neither party accepts premises he is not justiﬁed in accepting. Both parties form their beliefs by the methods one ought to use in forming beliefs. Could these individuals still end up with unresolvable differences of opinion? On the agent-neutral view, both parties must fully agree with one another on all factual questions, for they have the same available evidence, and they weigh that evidence in the same way.4 On the agent-neutral view, though what evidence one has

4 More precisely, the two parties should have the same subjective probabilities. An agent-neutralist could allow that different agents may have different evidential thresholds for belief, as in Fantl and McGrath’s (2002) view. For instance, because being right about whether p matters more to Jon than to Mary, Jon might require more evidence before accepting p than Mary requires. Nevertheless, in the situation described in the text, Jon

20

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

often differs from one person to another, the epistemic force of a given piece of evidence is not itself agent-relative. That is, for any given piece of evidence, there is an objective degree to which that evidence supports a given conclusion, which should guide the thinking of any person who is aware of that evidence. But for the agentcentrist, two parties could rationally assign different weights to the same piece of evidence, depending on whose evidence it was, even though both parties were equally certain of the existence of that evidence. This has the consequence that the two parties described above might rationally disagree after all is said and done. My overall aim in what follows is to defend a fully agent-centered epistemological perspective. I begin by discussing some contemporary theories about the sources of epistemic justiﬁcation and considering which putative sources of justiﬁcation, if genuine, would be agent-centered. I go on to consider the main motivation behind the agent-neutral view, which I take to consist in a desire to avoid an implausible sort of egotism. I explain how an agent-centered epistemological theory can accommodate agent-neutralist intuitions by mimicking the behavior of an agent-neutral theory in common circumstances. I conclude by addressing the relationship between my agentcentered view and probabilistic approaches to epistemology.

1.2 Agent-neutrality and agent-centeredness in contemporary epistemological theories Which contemporary theories of epistemic justiﬁcation are agent-neutral, and which are agent-centered? Most theories could be held in either an agent-neutral or an agentcentered form, but in most cases, one of these forms of the theory would be more natural than the other. Reliabilism First, consider reliabilism as a theory of justiﬁcation. Roughly, this is the view that a belief is justiﬁed when it is the product of a reliable belief-forming mechanism.5 Now suppose you know that someone else has a reliable belief-forming mechanism that issues in the belief that p. If you are also aware of the truth of reliabilism, you could infer that the other party is justiﬁed in believing that p. But do you have justiﬁcation for believing p yourself, in virtue of what you know about the other person? The answer to this seems to be yes, at least prima facie. If you know that a reliable mechanism indicates that p, then, other things being equal, this is evidence for p, regardless of whether the mechanism belongs to your own cognitive faculties or someone else’s.

should take himself to have the same amount of evidence for p as Mary should take herself to have, and thus—assuming that one’s subjective probabilities should be determined by one’s evidence—Jon and Mary should agree on how likely p is. 5 Goldman 1979. For illustrative purposes, I discuss simpliﬁed versions of reliabilism and the coherence theory herein.

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

21

Knowing this, you would be prima facie justiﬁed in assigning a high degree of belief to p. An agent-centered version of reliabilism could be articulated, but it would lack plausibility. Coherentism Consider next the coherence theory of justiﬁcation. Roughly, this is the view that a belief is justiﬁed when it coheres with one’s belief system and that belief system is sufﬁciently coherent overall.6 Suppose you know that both you and Jon have highly coherent belief systems,7 and that p coheres sufﬁciently with Jon’s belief system. Are you thereby prima facie justiﬁed in believing p yourself? Each answer to this question has some plausibility. On the one hand, it might seem that the fact that p coheres with Jon’s belief system is irrelevant to you. You should accept p only if, and only because, p coheres with your own belief system. This parallels views one might take about norms relating to consistency and inferential support: if you know that someone else believes that p, then you can conclude that that person has reason not to accept propositions that obviously contradict p,8 and that person has reason to accept propositions that are obviously supported by p. But you do not thereby have reason not to contradict p, nor to accept propositions that p obviously supports. Because of the close connection between coherence and such relations as consistency and inferential support, these observations suggest that if coherence is a source of justiﬁcation, it is an agent-centered source. On the other hand, consider a popular account of why coherence is a source of justiﬁcation: some argue that it is unlikely that a large collection of (independently formed) beliefs should turn out to ﬁt together well, unless they were by and large correct.9 If this is the reason why one should accept one’s own coherent beliefs as by and large true, it seems that one has the same prima facie reason to accept someone else’s coherent beliefs as by and large true—it could equally well be said that their beliefs would be unlikely to ﬁt together well unless they were mostly accurate. I do not, however, claim that a coherentist must be an agent-neutralist. Rather, it seems to me unclear whether coherentists should be agent-neutralists or agent-centrists. Mentalist evidentialism Earl Conee and Richard Feldman defend the evidentialist view that whether one is justiﬁed in believing that p depends entirely on whether believing that p ﬁts one’s 6

BonJour 1985: 89–93; Lehrer 1974: 154. We must assume that your own belief system is coherent, since, on the coherence theory, you could not know anything, either about Jon’s belief system or about anything else, unless your own belief system was sufﬁciently coherent. 8 This claim is open to question. An agent-neutralist might argue that one has reason to avoid contradicting p only if one’s belief that p is justiﬁed, and that one has the same prima facie obligation not to contradict the justiﬁed beliefs of other people as well. A similar point applies to the alleged epistemic right to infer what follows from one’s present beliefs. 9 BonJour 1985: 147–8; Elgin 2005: 157. 7

22

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

evidence.10 In addition, they seem to regard one’s basic evidence or reasons for belief as consisting solely in propositions describing one’s own mental states.11 How do these views bear on the issue of agent centeredness? Suppose that everything one knows counts as evidence that one has: if one knows that p, one’s evidence includes p.12 And suppose that two people each have perfect knowledge of one another’s relevant mental states, and both reason ideally from their knowledge (they draw all and only the conclusions justiﬁed by their evidence). In this case, on Conee and Feldman’s view, it seems that these two individuals must have exactly the same propositional evidence. Each has as evidence (propositions describing) his own mental states, and each also has as evidence propositions describing the other’s mental states, simply by virtue of knowing the latter propositions. Since they are ideal reasoners, each also knows (and has as evidence) everything that can be rationally inferred from the facts about both individuals’ mental states. They have no other evidence, since neither has any basic evidence not included in the mental states of the two individuals. So it seems that they have the same evidence. Given evidentialism, we are under strong pressure to conclude that these two hypothetical individuals would have equal justiﬁcation for any proposition. This would amount to a rejection of agent-centeredness. There is one way to avoid this conclusion, consistent with mentalist evidentialism and with the view that all known propositions are evidence. Evidentialism holds that epistemic justiﬁcation supervenes on ﬁt with evidence. Agent-centeredness, therefore, might be maintained by taking the relation of ‘ﬁt’ to be agent-relative: perhaps the beliefs that ﬁt a given set of evidence may vary from one subject to another. Perhaps, for example, the fact that Mary has the intuition that p ﬁts with believing p if you are Mary, but does not ﬁt with believing p if you are someone else. This seems to be a plausible way of formulating agent-centeredness within an evidentialist framework. It is unlikely, however, that Conee and Feldman would take this view, since they seem to regard evidentialism as entailing that epistemic justiﬁcation supervenes, not just on what ﬁts with one’s evidence, but on what evidence one has.13 Phenomenal Conservatism At least one contemporary epistemological theory is more naturally held in an agentcentered than an agent-neutral form. This is the principle that I defend elsewhere under the name Phenomenal Conservatism (PC): if it seems to one that p, then in the 10

Conee and Feldman 2004: 83. Conee and Feldman argue that epistemic justiﬁcation supervenes on one’s mental states (2004: 56) and that ‘all ultimate evidence is experiential’ (2008: sect. IB, IC). They go on to propose that ‘Propositions to the effect that the evidence exists . . . can serve to state the reasons the evidence gives that adjudicate among hypotheses’ (2008: sect. IVB). I assume that the distinction between one’s evidence and the reasons that one’s evidence gives one is immaterial for present purposes. Feldman also speaks of one’s having propositions as evidence (Conee and Feldman 2004: 232). 12 Conee (p.c.) has expressed some sympathy with this view. 13 Conee and Feldman 2004: 101. 11

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

23

absence of defeaters, one thereby has some justiﬁcation for believing that p.14 One could hold an agent-neutral form of the doctrine: one could conjoin with PC the claim that, if one knows that it seems to someone else that p, one also thereby has some justiﬁcation, in the absence of defeaters, for accepting p. But it seems to me more natural to take an agent-centered view of the justiﬁcation provided by appearances.15 There are two reasons for this. First, one of the advantages of PC is its ability to account for the source of all epistemic justiﬁcation—whether one is considering perceptual beliefs, intuitive beliefs, memory beliefs, or introspective beliefs—using a single, simple principle. One might even comprehend inferential justiﬁcation under PC, conceiving inference as a process by which a conclusion is made to seem correct by virtue of its seeming to be supported by other propositions that one believes.16 Admittedly, to form a complete theory of justiﬁcation one would need to supplement PC with principles characterizing the modes of epistemic defeat. Nevertheless, PC can aspire to be a complete theory of what confers justiﬁcation. Suppose now that we introduce the following further thesis, which I will call ‘Altruistic Conservatism’: For any subjects S and T, and any proposition p, if S knows that it seems to T that p, then in the absence of defeaters, S thereby has some justiﬁcation for believing p. This thesis does not follow from PC, since you may know that it seems to someone else that p without its seeming to you that P. Nor is the thesis easily explained in terms of any traditionally recognized form of inference—‘It seems to T that p; therefore, p’ does not follow any well-known pattern of valid deductive reasoning or cogent inductive reasoning. Inference to the best explanation would be one’s best hope for explaining the truth of Altruistic Conservatism in terms of inferential justiﬁcation; defending that approach would, however, be a complex task. It is far from obvious why the truth of p should generally be presumed to be part of the best explanation, or supported by the best explanation, for the fact that it seems to someone that p.17 This is not to say that one could not defend an agent-neutral form of Phenomenal Conservatism, if one had grounds for respecting agent-neutrality. But the agent-neutral version of PC is less simple, and therefore less initially attractive, than an agent-centered form of the doctrine, because the agent-neutral view requires an extra principle of justiﬁcation. A second reason to prefer an agent-centered version of Phenomenal Conservatism derives from consideration of the motivations behind PC. One such motivation is the self-defeat argument. Very brieﬂy and roughly, this argument claims that competing 14

See my 2001: 99–115; 2006; and 2007. I use ‘appearance’ to refer to the mental state one has whereby it seems to one that something is the case. 16 Conee (Conee and Feldman 2004: 15) takes a view along these lines, discussing what he calls ‘seeming evidentialism.’ 17 It is not obvious on its face, for example, that ‘It seems to S that p because p is true’ is generally a simpler explanation than ‘It seems to S that p because p is false.’ Space limitations prevent further discussion here. 15

24

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

epistemological theories are self-defeating, because such theories are inevitably based on how things seem to the epistemologists who hold those theories. To be justiﬁed, a belief must be based on an adequate source of justiﬁcation. Therefore, if PC is false, then beliefs in competing theories are unjustiﬁed.18 Now, even if this argument supports Phenomenal Conservatism, it does not support Altruistic Conservatism. It may be self-defeating to deny that its seeming to one that p is a source of prima facie justiﬁcation for one to believe that p, but there is nothing self-defeating in denying that one’s knowledge that it seems to someone else that p is a source of prima facie justiﬁcation for one to believe that p. This is because, while one can scarcely avoid basing one’s beliefs on one’s own appearance states, one can easily avoid basing one’s beliefs on the appearances experienced by others. If, therefore, Phenomenal Conservatism is true, then the agent-centered approach is most likely correct. But some other prominent epistemological theories support agentneutrality. What can be said on behalf of agent-centeredness or agent-neutrality, apart from one’s adherence to one or another speciﬁc theory of the sources of epistemic justiﬁcation?

1.3 The case for agent-neutrality: Avoiding epistemological egotism On the face of it, there is a compelling thought behind agent-neutrality. Agentcenteredness seems to call for a kind of epistemological egotism, an a priori privileging of one’s own experiences merely because they are one’s own. Each agent seemingly must say, ‘My experiences, considered as such, are prima facie better indicators of reality than the experiences of others.’ Earlier, I tried to distinguish agent-centeredness from epistemological egotism, drawing an analogy with ethical egoism. The ethical egoist is not an egotist, because he does not hold that other people should serve his interests; he holds that everyone should serve their own interests. Similarly, the agent-centered epistemologist does not hold that other people should trust in his appearances; the agent-centered epistemologist holds that people should trust in their own appearances. But on second thought, this analogy overlooks an important asymmetry between ethics and epistemology. When the ethical egoist values his own happiness above that of others, he is not committed to holding that others are mistaken in valuing their happiness above his. The egoist can maintain that (some) value is agent-relative: my happiness is better for me, or from my point of view, than your happiness; but your happiness is better for you. More generally, I may rank outcome A above B, while you rank B above A, and both of us may be, not only justiﬁed in our rankings, but actually correct, in the sense that your ranking reﬂects what you should pursue, while mine

18

See my 2007.

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

25

reﬂects what I should pursue. There is no need for either of us to be mistaken, because you and I need not disagree about any factual question. An analogous view cannot be maintained in the epistemological case. If you and I disagree about whether p due to our differing appearance states, we may both be justiﬁed in our beliefs, but we cannot both be correct. Value may be agent-relative, and justiﬁcation may be agent-relative, but truth is absolute. I must take my own belief to be the one that is true, given that I continue to hold that belief. I must therefore hold that there is something special about me: I am the one whose appearances are calibrated with reality. And this can easily strike us as an improbable optimism.19 The issue is of more than just theoretical signiﬁcance. The following sort of situation seems to occur frequently in intellectual life: Two individuals are about equally intelligent, are about equally dedicated to ﬁnding the truth about some issue, and are aware of essentially the same evidence and arguments bearing on that issue—yet they come to very different conclusions. This sort of situation occurs with particular frequency in philosophical, religious, and political matters, though it can also occur in scientiﬁc matters. These disagreements may be due to such factors as differing intuitions, differing weights assigned to competing theoretical virtues, and differing assessments of the relative importance of various pieces of evidence. Some of these differences may be caused by non-rational biases afﬂicting one or both parties. In any case, it seems that, in the absence of differential information bearing on one’s own cognitive abilities and limitations relative to others—evidence showing oneself to be more competent or less subject to bias than others, for example—one should count oneself equally likely as another person to be in error in disputes of this kind. And counting oneself equally likely to be in error, one should assign equal weight to one’s own cognitively relevant experiences and those of another. 20 Agent-centered theories allow one to assign different (and presumably greater) weight to one’s own intuitions and judgments than one assigns to those of other people, even in the absence of any independent evidence for one’s own greater reliability. This seems to imply that each individual has justiﬁcation for taking himself to be a priori more likely to make correct judgments than other people. In one way, the agent-centered conception of justiﬁcation reduces the apparent egotism involved in sticking to one’s beliefs in the face of disagreement, since the agent-centered view allows that others may be equally justiﬁed and rational in their conﬂicting beliefs as one is in one’s own. But it leaves one with the perhaps less likely 19 Compare Hanson’s (2006) argument that individuals with differing prior probabilities must each regard themselves as more likely to have priors more closely calibrated with the truth. 20 Feldman (2006), Christensen (2007), and Elga (2007) defend this view, though Elga and Christensen sometimes seem to be making only the weaker claim that, if one has compelling evidence that others are equally reliable as oneself, then one should attach equal weight to their opinions as one attaches to one’s own opinions. I do not deny that claim; I deny only the stronger claim that, if one lacks evidence that one is more reliable than others, one should attach equal weight to their opinions. See my 2005 for an argument in favor of deferring to others in most cases of controversy, based on evidence that there are others who are more expert than oneself. See Kelly (2005) and Weatherson (2007) for criticisms of the ‘equal weight’ view.

26

E V I D E N T I A L I S M A N D D I S AG R E E M E N T

postulate that one’s own justiﬁed beliefs are a better indicator of the truth than the equally justiﬁed beliefs of others. In line with this objection, the agent-centered conception of justiﬁcation seems to conﬂict with our epistemological intuitions in certain cases. Sensory observation seems to be a source of justiﬁcation for external-world beliefs: if I seem to perceive that p, I have prima facie justiﬁcation for believing that p. This source of justiﬁcation seems to be agent-neutral: if I learn that another person saw, or seemed to see, a deer in the forest yesterday, then I ought to conﬁdently accept that there was a deer in the forest, just as I would if I seemingly saw the deer myself. Likewise, if I learn that Sally seemingly remembers getting a bicycle on her tenth birthday, I ought to believe that Sally got a bicycle on her tenth birthday, just as I should believe that I got a bicycle on my tenth birthday if I seemingly remember that event. It is unclear how the agentcentered view accounts for this. The agent-centrist need not be an extremist. He need not hold that your knowledge of others’ experiences gives you no justiﬁcation at all for believing what they should believe. The agent-centrist might hold only that, when you know that someone else has an experience that gives them justiﬁcation for believing that P, this knowledge gives you less prima facie justiﬁcation for believing that p than the justiﬁcation the other person has. Still, the above reﬂections call into question even this weak agent-centered view.

1.4 The epistemological egoist’s rejoinder My own view is fully agent-centered at the level of basic sources of evidence. I hold that only an individual’s own experiences provide that individual with any justiﬁcation for believing anything. If one has an appearance representing that p, one has at least some defeasible justiﬁcation for believing that p. This source of justiﬁcation is basic in the sense that one need have no separate justiﬁcation—either for believing p, or for believing that one’s appearances are likely to be reliable indicators of the truth, or for believing anything else at all—in order for one’s appearance state to provide this justiﬁcation for p. In contrast, if one merely knows that it seems to someone else that p, this by itself gives one no reason at all to believe that p.21 The emphasized ‘by itself’ hints at how I would seek to accommodate some of the intuitions cited in the previous section. Suppose you know that it seems to someone else that p, and you also have a justiﬁed background belief that other people’s appearances are generally reliable indicators of the truth. You then have (defeasible) inferential grounds for believing that p. This much nearly any epistemologist would grant, agent-centrist and agent-neutralist alike. And almost everyone in fact has the justiﬁed background belief that other people’s sensory experiences and memory experiences are reliable indicators, respectively, of facts about the physical world and about the past. This is why, 21 Thus, I take the extreme view to which, at the end of the last section, I said that agent-centrists need not be committed. See Wedgwood (2007: 260–3) for a similar defense of this position.

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

27

in the examples at the end of the previous section, almost everyone would in fact acquire strong grounds for believing (a) that there was a deer in the forest, upon learning that someone else seemingly saw a deer, and (b) that Sally got a bicycle on her tenth birthday, upon learning that Sally seemingly remembered that event. To that extent, agent-centrists can agree with agent-neutralists about those cases. Where agent-centrists and agent-neutralists disagree is in their accounts of the precise source of one’s epistemic justiﬁcation in the above cases. Agent-neutralists will trace one’s justiﬁcation to one’s awareness of another party’s satisfaction of some condition, where that condition is also the condition in virtue of which the other party himself possesses epistemic justiﬁcation. In contrast, the agent-centered view that I defend holds that one’s justiﬁcation for crediting another individual’s appearances in these cases depends upon one’s justiﬁcation for a certain background belief, where that belief need not be held by those who actually experience the appearances in question directly. One’s justiﬁcation for this background belief ultimately derives, in turn, from one’s own appearances—one must have had appearances that support the conclusion that the appearances of others are generally reliable. For example, one can compare what one sees in the immediate vicinity of other people with what those others themselves report seeing around them. One can observe others’ behavior to determine whether they appear to be successfully detecting the relevant physical objects in their surroundings. It is observations like this that justify one’s general background assumption that other people’s sensory perception is reliable. Thus, in my view, to draw justiﬁed conclusions from others’ experiences, one needs considerably more than just awareness of their justiﬁcation for belief. Nor need the agent-centrist be committed to epistemological egotism when it comes to intellectual disputes. He need not hold that epistemic fortune has smiled especially on him. Rather, the thinker following a set of agent-centered norms can inquire into whether epistemic fortune has smiled especially on him—relying, as in any inquiry, on his own appearances. It may seem to him unlikely that his appearances are signiﬁcantly better calibrated with reality than those of many equally intelligent and equally informed people. If so, that would give him grounds for doubting his own appearances when they conﬂict with those of many equally intelligent and equally informed people. This would give one an agent-centered rationale for respecting, to a large extent, the judgments championed by the agent-neutralist. I suggest that it is really this rationale on which the plausibility of the argument of the preceding section depends: if the agent-neutralist’s complaint about ‘epistemic egotism’ in Section 1.3 struck you as persuasive, this is probably because it seemed to you unlikely that you would be especially epistemically fortunate in the manner described. But if this seemed to you unlikely, then even an agent-centered Phenomenal Conservatism would counsel against your relying uncritically on your own intellectual appearances in the cases of controversy. Agent-centered and agent-neutral views cannot agree about all cases, of course. They must disagree about cases in which (a) an agent knows of another party’s

28

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

appearances, but (b) the agent has no information at all about the other party’s reliability (not even general background knowledge about the reliability of beings of that kind). In such a case, my agent-centered view holds that the agent would have no justiﬁcation, even prima facie, for believing what seemed true to the other party. But the agent-neutral view maintains that the agent would still have prima facie justiﬁcation for believing what seemed true to the other party (assuming that one’s own appearances generate prima facie justiﬁcation for belief). In other words, agent-centrists and agent-neutralists differ about the default attitude towards other beings: in the absence of evidence to the contrary, should one treat other beings’ appearances just like one’s own, or should one treat other beings’ appearances as merely another class of external phenomena that may or may not prove to have any correlation with other external phenomena? The agent-centered view strikes me as the more plausible in such cases. If I know that it seems to you that p, but it does not particularly seem to me, nor do I believe, that your appearances have any correlation with any reality beyond themselves, then why would I conclude that p is true? It is difﬁcult to imagine realistic situations in which that antecedent would hold, since we almost always believe, and justiﬁedly so, that others’ appearances are likely to have some correlation with reality. But in a situation in which you somehow have no information about how, if at all, another being’s appearances relate to reality, I see no reason for you to conclude, from a knowledge that it appears to them that p, that p is true. In contrast, I think that even when you have no background information about how, if at all, your own appearances relate to reality, you still have some (defeasible) justiﬁcation for believing that things are the way they appear to you. Why the asymmetry? Because you are not related to others’ mental states in the same way as you are related to your own. All of your beliefs are inevitably based upon your own mental states, in a sense in which they are never based upon the mental states of anyone else. Even when you form beliefs by accepting the testimony of others—perhaps the closest thing to basing your beliefs on someone else’s beliefs—there remains an important sense in which you really base your conclusions on your own mental states. Suppose Sara says that she met a talking bush yesterday. You believe her. Your belief in the talking bush is then based on your belief (or appearance) that Sara said she saw a talking bush, plus your belief (or appearance) roughly to the effect that Sara’s testimony is reliable. Because of your unique relation to your own appearances, a fundamental self-trust is necessary to avoid a debilitating skepticism.22 We must not require an epistemic agent to verify the reliability of his own appearances before trusting them, because the agent would have no way of verifying anything without trusting his appearances. But an agent can verify or confute the reliability of others’ appearances, by relying on his own

22

Lehrer (1997: 5) and Foley (2001: 19–20) make similar observations.

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

29

appearances, before trusting those others. I start out by trusting myself, but others must earn my trust. The agent-centered view allows one to take varying attitudes towards different kinds of appearances experienced by other people, depending on one’s background evidence about the reliability of these kinds of appearances. This may explain an otherwise puzzling phenomenon: most people tend to credit others’ perceptual observations and memories to a much greater degree than they credit others’ intuitions or intellectual judgments. If two individuals remember a particular event differently, each will likely give the other’s memory signiﬁcant credence and thereupon suspend belief about what really happened. But if two individuals have different philosophical intuitions, each is much more likely to simply go on assuming that his own intuition is correct. This might at ﬁrst appear to be a kind of inconsistency in our attitudes, as though we shift from an agent-neutral to an agent-centered epistemology for no apparent reason. But perhaps a more satisfying explanation is this: we have very extensive, compelling evidence for the reliability of other people’s perception and memory. It is quite easy to generate numerous examples of very speciﬁc details in respect of which other people have accurate memories and observations, as judged from the standpoint of one’s own memories and observations. Interpersonal differences in perceptual and memory beliefs tend to be relatively small and peripheral. In contrast, our evidence for the reliability of one another’s intuitions, particularly about philosophical matters, is much weaker. Large disagreements are more common, and even when individuals share intuitions, there typically are not nearly as many speciﬁc details to compare. Because of this, our attitude towards other people’s intuitions is closer to the default attitude of complete difﬁdence than is our attitude towards other people’s sensory observations and memories.23

1.5 Agent-centered phenomenal conservatism and probabilistic epistemology My view of the potential rationality of ongoing disagreement seems to ﬂy in the face of standard Bayesian models of rational belief. Space limitations preclude a full discussion here, but I cannot conclude without making at least some brief remarks regarding how my views expressed above relate to probabilistic epistemology. According to one celebrated result, Bayesian agents with different evidence but the same prior probabilities should be able to reach agreement in posterior probabilities simply by repeatedly exchanging information as to their degrees of belief in various propositions, including information as to how each agent updates his subjective

23 This is not to say that our attitudes are fully rational. I suspect that we tend to trust our own philosophical judgments too much because we fail to take full account of the defeaters we have for such judgments.

30

E VI D EN T IA L IS M A ND D I S AGR EE M E NT

probabilities in the light of the other’s reports.24 My own view seems to ﬂy in the face of this theorem, since I allow for rational, intractable disagreement, even for individuals with common prior probabilities. In fact, there is no need to invoke the theorem of the preceding paragraph, for my view ﬂies in the face of a much more trivial theorem of Bayesian epistemology. I am willing to stipulate a case in which two rational agents have the same prior probabilities and each knows exactly what evidence the other has. (All the work of Aumann’s theorem is directed at getting two agents to knowledge of one another’s evidence.) In such a case, I maintain, there could still be intractable rational disagreement between the two parties. Let us look at why this is impossible in the standard Bayesian view, and how it is possible in my own view. An ideal Bayesian agent’s ‘evidence’ has at least the following three characteristics: First, his evidence includes all those propositions for which he has acquired, non-inferential justiﬁcation. On the standard Bayesian view, once one’s prior probabilities have been ﬁxed, one can acquire justiﬁcation for believing some proposition, p, only if (a) one acquires p as a new piece of basic evidence, or (b) p receives a probability boost from conditionalization on some new piece of basic evidence. Since (b) represents a kind of inferential justiﬁcation, (a) is the only means of acquiring non-inferential justiﬁcation. Second, evidence is always true. If there is no turkey in front of you, then even if there appears to be one, your evidence by deﬁnition cannot include the proposition that there is a turkey in front of you; at best, it might include the proposition that there appears to be a turkey in front of you. Third, when E is part of your evidence, you are maximally justiﬁed in believing E. Thus, the Bayesian agent assigns probability 1 to his evidence. Given this view, it is obvious why agents with the same prior probabilities and perfect knowledge of one another’s evidence must agree completely. Once I learn that you have evidence E, I thereby know with certainty that E is true, since this follows from the meaning of ‘evidence.’ That you are also certain of E also follows from its being part of your evidence. So if you and I each know what evidence the other has, then we are both certain of all of the same evidence propositions. We have no other non-inferentially justiﬁed beliefs, and we update our beliefs by the same rule of Bayesian conditionalization. Assuming we start from the same prior probabilities, we obviously end up with the same subjective probabilities for every proposition. It is the traditional Bayesian conception of evidence that should be questioned. Not every proposition for which we have non-inferential justiﬁcation is one of which we can be absolutely certain, nor is every such proposition actually true. In particular, a person can become defeasibly, non-inferentially justiﬁed in believing E, simply by its appearing to that person that E, even if E is not in fact true. Richard Jeffrey has provided a way of incorporating the notion of uncertain evidence (propositions for which one 24 Aumann 1976. The theorem assumes that it is common knowledge that the two subjects are Bayesian reasoners with identical priors. Cowen and Hanson (2004) use the result to argue that most actual ongoing disagreements are to some degree dishonest.

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

31

acquires inconclusive, non-inferential justiﬁcation) into a Bayesian framework.25 In the traditional Bayesian view, if I know that you have evidence E, then both you and I must assign probability 1 to E. But once we allow for uncertain evidence, I may know that you have uncertain evidence E, which you are justiﬁed in believing to degree .9. My knowing this about you by itself places no constraints on the degree of belief I should assign to E. Recall that the constraint on my degree of belief in E provided by the traditional Bayesian view stemmed from the assumption that ‘evidence’ is by deﬁnition true—if, therefore, one knows for certain that E is in fact a piece of evidence, one must know for certain that E. Rejecting the assumption that evidence is always true, we are left with no constraint on my probability of E. I need not assign probability 1 to E, but nor need I assign probability .9 to E, for I did not acquire evidence E, not even uncertainly. I only acquired as evidence the proposition that you acquired evidence E. Jeffrey’s Bayesianism has nothing to say against this view of the situation, since it is not a theory of evidence. That is, Jeffrey’s Bayesianism is a theory of how one should modify degrees of belief in other propositions when one acquires evidence E with degree of certainty r; it is not a theory of the conditions under which E in fact becomes part of one’s evidence with certainty r. In particular, Jeffrey’s theory does not say whether, in learning that you acquired evidence E with certainty .9, I also acquired evidence E. On the agent-centered view, I need not acquire evidence E in such a situation, so I need not undergo the rational change in belief that you undergo. One could of course propose an alternative view. One could hold, for instance, that when you have the appearance that E, it is only the proposition that you have the appearance that E, and not E itself, that becomes part of your evidence. This would preserve the symmetry between one’s own appearances and the known appearances of others. This view, in my opinion, faces serious problems, including difﬁculties dealing with philosophical skepticism. But I cannot detail those problems here. Let it sufﬁce here to say that this is not my view. My view is that, upon having an appearance state, one acquires non-inferential, prima facie justiﬁcation for believing the propositional content of that state. (Of course, one may also, if one is self-aware, acquire justiﬁcation for believing that one has the appearance state.) Since I also believe that this is the only way in which one can acquire non-inferential justiﬁcation, I hold that individuals’ basic evidence consists entirely in the propositional contents of their own appearances. Bayesian theorems concerning agreement, then, fail to undermine a reasonable agent-centered epistemology, because those theorems depend crucially on a classical foundationalist assumption that very few epistemologists ﬁnd plausible any longer: the assumption that all justiﬁed belief rests on a foundation of true and indubitable evidence. Jeffreyan modiﬁcations to Bayesianism do not preserve the theorems concerning agreement, since they cannot guarantee that, in learning of someone else’s evidence, one must assign any credibility to that evidence.

25

Jeffrey 1983: ch. 11.

32

E VI D EN T IA L IS M A ND D I S AGR EE M E NT

I do not claim that the agent-centered conception of evidence accounts for all or even most actual disagreement, nor that all or most disagreement is actually rational. Differing priors and outright irrationality may well explain much of the ongoing disagreement that we observe in philosophy and other areas.26 The ﬁelds of inquiry suffering from the most disagreement are likely to be ones where multiple factors contribute to differences of opinion. The agent-centered view offers for our consideration just one more cause of intractable philosophical disagreement.27

References Aumann, Robert J. 1976. ‘Agreeing to Disagree,’ The Annals of Statistics, 4: 1236–9. BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. Christensen, David. 2007. ‘Epistemology of Disagreement: The Good News,’ Philosophical Review, 116: 187–217. Conee, Earl and Richard Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Clarendon. ——. 2008. ‘Evidence’ in Quentin Smith (ed.), Epistemology: New Essays. Oxford: Oxford University Press. Cowen, Tyler and Robin Hanson. 2004. ‘Are Disagreements Honest?’ unpublished MS, . Accessed 19 January 2011. Elga, Adam. 2007. ‘Reﬂection and Disagreement,’ Noûs, 41: 478–502. Elgin, Catherine. 2005. ‘Non-foundationalist Epistemology: Holism, Coherence, and Tenability,’ in Matthias Steup and Ernest Sosa (eds), Contemporary Debates in Epistemology, pp. 156–67. Malden, MA: Blackwell. Fantl, Jeremy and Matthew McGrath. 2002. ‘Evidence, Pragmatics, and Justiﬁcation,’ Philosophical Review, 111: 67–94. Feldman, Richard. 2006. ‘Epistemological Puzzles about Disagreement,’ in Stephen Hetherington (ed.), Epistemology Futures, pp. 216–36. Oxford: Oxford University Press. Foley, Richard. 2001. Intellectual Trust in Oneself and Others. Cambridge: Cambridge University Press. Goldman, Alvin. 1979. ‘What Is Justiﬁed Belief?’ in George Pappas (ed.), Justiﬁcation and Knowledge, pp. 1–23. Dordrecht, Holland: D. Reidel. Hanson, Robin. 2006. ‘Uncommon Priors Require Origin Disputes,’ Theory and Decision, 61: 318–28. Huemer, Michael. n.d. ‘Why People Are Irrational about Politics,’ unpublished ms. . ——. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman & Littleﬁeld. ——. 2005. ‘Is Critical Thinking Epistemically Responsible?’ Metaphilosophy, 36: 522–31.

26

Morris 1995; Huemer n.d. I would like to thank Robin Hanson and Trent Dougherty for helpful discussion of the issues in this paper. 27

E P I S T E M O L O G I C A L E G O I S M A N D A G E N T- C E N T E R E D N O R M S

33

——. 2006. ‘Phenomenal Conservatism and the Internalist Intuition,’ American Philosophical Quarterly, 43: 147–58. ——. 2007. ‘Compassionate Phenomenal Conservatism,’ Philosophy and Phenomenological Research, 74: 30–55. Jeffrey, Richard. 1983. The Logic of Decision, 2nd edn. Chicago, IL: University of Chicago Press. Kelly, Thomas. 2005. ‘The Epistemic Signiﬁcance of Disagreement,’ in John Hawthorne and Tamar Gendler Szabo (eds), Oxford Studies in Epistemology, vol. 1, pp. 167–96. Oxford: Oxford University Press. Lehrer, Keith. 1974. Knowledge. Oxford: Clarendon Press. ——. 1997. Self-trust: A Study of Reason, Knowledge, and Autonomy. Oxford: Clarendon Press. Morris, Stephen. 1995. ‘The Common Prior Assumption in Economic Theory,’ Economics and Philosophy, 11: 227–53. Neta, Ram. 2008. ‘In Defense of Disjunctivism,’ in Fiona MacPherson and Adrian Haddock (eds), Disjunctivism: Perception, Action, and Knowledge, pp. 311–29. Oxford: Oxford University Press. Schefﬂer, Samuel. 1982. The Rejection of Consequentialism. Oxford: Clarendon Press. Weatherson, Brian. 2007. ‘Disagreeing about Disagreement,’ unpublished ms., . Wedgwood, Ralph. 2007. The Nature of Normativity. Oxford: Clarendon Press.

2 The Rational Signiﬁcance of Reﬂective Ascent Jonathan Kvanvig

With saddles for pillows, staring up and contemplating the stars above, two cowboys converse on a warm summer night. Joe speaks ﬁrst. ‘What do you think happens when you die?’ Billy is silent and then responds. ‘I think when you die, it’s over. There’s nothing.’ Joe ruminates on a stem of grass, top-hand taciturn in demeanor. Finally, ‘I guess I don’t. I don’t think it’s over. But I don’t know.’ The conversation is intriguing philosophically if we take the assertions as accurately reﬂecting what internal states are present. Joe apparently believes things that he believes he doesn’t know, but more relevant for our present topic is that Joe and Billy disagree. Joe thinks that in some sense we survive death; Billy doesn’t. Some philosophers will feel the need to defend the rationality of Joe and Billy in some way. They will hold that the only way for both Joe and Billy to rationally believe what they do is for further factors to be present. Maybe Joe has some evidence that Billy doesn’t, or vice-versa; maybe one is a dolt and the other the next Einstein. For such philosophers, disagreement always delivers the scent of irrationality, a scent cleansed from the epistemic air only by the presence of factors such as those just noted. But, for such philosophers, once we control for a comparatively short list of ways to explain away the disagreement, epistemic irrationality is unavoidable. I doubt, however, that after reading the introductory paragraph, any scent of irrationality was detected. The scene is completely innocuous from an epistemic point of view. Describing the two as cowboys conveys some sense that there is no vast difference between the two in terms of intellectual competence, nor are cowboys noted for having a need to think of themselves as intellectually superior to their companions, and the subject matter is selected to block any tendency to think that one of them has special information about the subject that the other doesn’t. They disagree without rancor, and there is no tendency on our part to sniff to ﬁnd some purported irrationality in the story. There is nothing for the irrationality sleuthhounds to track here.

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

35

There is a caveat, though, since some philosophers will think that beliefs about such matters are all automatically irrational, and so might explain away the absence of an epistemic response to the story. If so, we can change the example, since the general point remains that even when we change the subject matter about which disputants disagree, sleuthhounds still have no scent to follow, whether the subject matter is the relatively sophisticated matters of global warming, voucher programs for private schools, capital punishment, or the more mundane matters of whether it will rain tomorrow, which schools and doctors and dentists are best for our children, or which sports franchises have the most illustrious pasts. Disputation and disagreement abound at every turn; suspicions of irrationality are rare. My goal here is to provide the epistemological underpinning to support this view of disagreement, the view according to which rational disagreement is unproblematic. I’ll begin with some problems for the alternative view. These problems show the need for a restricted view of the signiﬁcance of disagreement, and I will develop two models for such restrictions. We will see that though disagreement has some epistemic effect, it doesn’t have the sort of effects the irrationality hounds think they detect.

2.1 Molliﬁcationism and its discontents Let us call the position that views disagreement as a sure sign, albeit defeasible, of irrationality ‘Molliﬁcationism’. Molliﬁcationists see harmony as the equilibrium point in the theory of rationality. The rational expectation is that people will agree, at least under certain conditions, so when they don’t, perplexity arises that needs to be explained in one of three general directions: either there is some difference related to the possession of relevant information or the ability to process it, or somebody is being irrational. Molliﬁcationism faces serious problems, however. Notice, ﬁrst, what happens when one insists that agreement in attitude is the only natural point of equilibrium in the theory of rationality. Molliﬁcationists like to attend to cases of disagreement in which one person believes p and the other p, but this is only one kind of attitudinal distance between people. There are, at the coarse-grained level, four possibilities: believing, believing the opposite (disbelieving), withholding, and taking no attitude at all. And if we move to the ﬁne-grained level of degree of belief, there are as many possibilities as there are real numbers between 0 and 1. Molliﬁcationists hold that differences in attitude are always and everywhere rationally suspicious, and once we see the variety of attitudes possible, this view should look strange and bizarre. Even if we restrict ourselves to the coarse-grained level, the Molliﬁcationist has strange demands. If you and I disagree about p, it is not enough that each of us gives up our beliefs. If you give up your belief and come to withhold concerning p, but I give up any attitude at all because I don’t know what attitude ﬁts my evidence any longer, the Molliﬁcationist is still unhappy. We still disagree in attitude: you withhold and I take no attitude at all toward p. What are we poor creatures, desirous of being rational, to do now? One of us

36

E VI D E NT IA L I SM A N D D I S AG REE M E N T

must defer to the other, but we better not both do it at once! The same issue arises if you withhold and I believe. One must defer to the other, but which to which? The fundamental point here is one about the paralyzing effect that comes from thinking of sameness of attitude as the equilibrium point for rationality. In some cases, no movement to an intermediate stance is possible: all that is possible is for one of the two to adopt the stance of the other. There is a strong temptation at this point to appeal to the language of compromise to minimize the signiﬁcance of this difﬁculty. The response I’m imagining is one that insists that there is no special problem here, that what is highlighted above is just a matter of how compromises work in some extreme cases. In lots of cases, there is an intermediate position available, but in other cases there isn’t; in both kinds of cases, however, the demands of rationality still impose a requirement of compromise in the face of disagreement in attitude. The language of compromise is rhetorically useful here for the Molliﬁcationist, and it is important to see why it is inappropriate language to use to describe the view in question. To see why, consider ﬁrst how ordinary compromises occur and what they involve. Suppose two politicians disagree about policy. In the end, some resolution is necessary, since unending paralysis is intolerable (for whatever reason). So they compromise. Both think the result is less than ideal. To understand the result, we need to know not only the history of the process, what they used to think and why, but also what they presently think and why. They think that the compromise is best, in some sense, but also, in another sense, that it is not. Without some such internal conﬂict, we don’t yet understand the political process in question or the full nature of the accommodations that have been made by one or both parties. Such compromise requires some such internal conﬂict, whereas, had one of the two convinced the other of the correctness of his or her view, no such conﬂict would be present and no compromise would have occurred. Molliﬁcationism can be pleasingly put in terms of the language of compromise, involving the claim that when two cognizers disagree about some claim p, there is rational pressure for compromise. Such language is misleading, however, since situations of cognitive compromise are different from situations of full resolution of disagreement or conﬂict. When resolution occurs, the story we tell has present agreement as the outcome of past disagreement, as when a husband and wife start with different priorities for a daily schedule and resolve their differences by adjusting their priorities in light of the preferences of the other. Such accommodation in solving a coordination problem introduces new information about the preferences of others, and thus differs from cases of political compromise. When cognitive compromise occurs, we expect something akin to what we found in the political case. As the disputants become more aware that they are converging on a point where rationality compels them to abandon their beliefs (as Molliﬁcationists would have it), they may view the approaching event with consternation. They may view it as an intellectual loss to mollify their attitude, and this sense of loss will not leave once the convergence point

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

37

is reached. Just as in the case of political compromise, some present mark will remain in place to distinguish it from cases of simply becoming convinced that one is mistaken. It is not enough to note only that the two used to disagree and now no longer do, since that doesn’t distinguish compromise from mere change in view. It is difﬁcult to see what decent answer the Molliﬁcationist can provide here to explain the difference. After compromise, the two now agree about p: they both withhold judgment, let us suppose. How did this come about, in a way that makes compromise different from mere change in view? One might try to account for the difference by saying that they still assess the force of the evidence differently. But this can’t be the story if rationality requires compromise, since then X and Y still disagree: one’s assessment is that the evidence supports p and the other that it doesn’t. In order to be rational, the two will have to give up this view as well. If all we can say is that they used to disagree but no longer do, we have no difference at all between compromise and mere change in view. Perhaps, though, the Molliﬁcationist can go internal here. Each can hold that from their own point of view the claim in question appears true or not false, respectively, and that this appearance doesn’t disappear when compromise occurs. Putting the point in this way threatens to undermine Molliﬁcationism almost immediately, however, for rationality is perspectival in just the sort of way that, if the two points of view differ in the way imagined, then a difference in rational attitude is to be expected. What the Molliﬁcationist needs here is a distinction between total point of view (on which rationality supervenes) and some more partial point of view on which differing appearances remain in place. Such a distinction, however, is present in cases of fully resolved disagreement as well, however, since when you convince me that I’ve made a mistake, I can still see the aspects of the situation that led to my mistake. So the difﬁculty of distinguishing a situation of compromise from one involving a more ordinary resolution of disagreement through change in view remains, and I see no way to defend the use of the language of compromise here. So the language of compromise cannot be used to rhetorical advantage in support of Molliﬁcationism, and the difﬁculty concerning the requirement that attitude agreement is the equilibrium point in the theory of rationality remains. Molliﬁcationists might attempt to argue that this problem arises from our antiquated and coarse-grained psychology, one put in terms of the attitudes of belief, disbelief, and withholding. They might claim that a more ﬁne-grained psychology, appealing to degrees of belief or levels of conﬁdence, has the resources to avoid the problems raised. Such is not the case, however. Even on such a ﬁne-grained approach, there is no guarantee that attitude agreement is always possible. For there is quite a variety of possible psychologies to which our epistemic theory must be applicable. For truly gifted cognizers, a full range of attitudes, representable by the real number line from zero to one, is possible. But other possibilities abound. Some cognizers are not capable of credences, where a credence is represented in terms of a real number between zero and one: such cognizers have attitudes representable only in terms of

38

E VI D E NT IA L I SM A N D D I S AG REE M E N T

intervals of real numbers between zero and one. Call the point-like cognizers ‘credencers’ and the interval cognizers ‘intervalers’. Credencers and intervalers are always guaranteed to disagree in attitude. Even among credencers, however, there might be limitations on the range of credence states possible. For example, some possible credencers are capable of only eleven possible credence states: 0, .1, .2, .3, . . . 1. So when two credencers disagree by having adjacent attitudes (one has credence .1 and the other .2, or one has credence .5 and the other .6), there is no intermediate position of agreement available. Similar points can be made about conﬁdencers as well, and once we note the possibilities for cognitive architecture, it becomes clear that the move to a more ﬁne-grained psychology and epistemology is not going to help. In some cases of disagreement, an intermediate position can be found, but for other cases, no such intermediate point will be available. In such cases, the only options are acquiescence by one of the parties to the dispute or irresolvable disagreement automatically judged irrational. Moving from coarsegrained belief to ﬁne-grained degrees of belief offers no help on this issue. Molliﬁcationism faces other problems as well, one of which is its tendency toward self-defeat.1 To see the problem, begin with the hyperbolic thesis that whenever people disagree, they are both irrational. This view is obviously false, or so I believe anyway. This fact about what I believe together with the hyperbolic thesis in question implies that no one can rationally believe the hyperbolic thesis. It is a philosophical thesis undermined by a quite prosaic fact, the fact that some of us think that disagreement is at least as expected in matters philosophical as elsewhere. Moreover, consider the implications of this epistemological hyperbole. When tempted to such a view, a responsible cognizer should always check to see if others disagree. In cases of philosophical views, ﬁnding such evidence will always be quite easy. E-mail a few friends in philosophy and you will have it. But then not only the hyperbolic thesis itself cannot be rationally believed, neither can any disputed thesis in any part of philosophy or elsewhere. We should not label the horribly hyperbolic thesis ‘self-defeating’, however, at least not necessarily so. It isn’t, since it is possible that everyone agrees with the thesis. A nice analogy here is with two versions of the Liar Paradox. The direct version of the paradox occurs when we have a sentence such as ‘this sentence is false’. But there are indirect versions, such as when a shirt has the following sentences on front and back, respectively: ‘the sentence on the other side of this shirt is false’, and ‘the sentence on 1 I believe the ﬁrst printed version of such an argument against a version of Molliﬁcationism is in Plantinga’s 1995 piece entitled ‘Pluralism: A Defense of Religious Exclusivism’ (in Thomas Senor (ed.), The Rationality of Belief and the Plurality of Faith (Ithaca, NY: Cornell University Press, 1995)). Peter van Inwagen’s piece ‘Is It Wrong Everywhere, Always, and for Anyone to Believe Anything on Insufﬁcient Evidence?’ (In Jeff Jordan and Daniel Howard-Snyder (eds.), Faith, Freedom and Rationality (Savage, MD: Rowman and Littleﬁeld 1996). Reprinted in: E. Stump and M. J. Murray (eds.), Philosophy of Religion: the Big Questions (New York: Blackwell 1999)) contains such an argument as well, and was presented at the Chapel Hill Colloquium in 1993, but it wasn’t published until 1996. It is instructive to note that defenses of Molliﬁcationism since these pieces were published do not address the argument from self-defeat.

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

39

the other side of this shirt is true’. Here we have a kind of contingent undermining of each sentence. If we want to say that each sentence is a liar sentence, defeating its own truth, we will have to say the self-defeat is contingent rather than necessary as in the former case. Perhaps the language of self-defeat isn’t the right language here, but terminology isn’t really the point. What matters in both cases, semantic and epistemic, is that the view is problematic because of contingent factors that are in place. In the semantic case, it is the sentence itself that is defeated; in the epistemic case, it is belief in the hyperbolic thesis the rationality of which is defeated. In the epistemic case, what matters is that these contingent factors are so obvious and predictable that one needs a really serious epistemic blindspot not to take such a factor into account before endorsing the horribly hyperbolic position. We may want to say that the position is contingently self-defeating in the way the shirt sentences are, and I will adopt that terminology here, with the forewarning that contingent self-defeat may not be a form of self-defeat at all together with the additional proviso that the object of defeat is the truth of a sentence in one case and the rationality of belief in the other. What matters most is the clear point that the ﬂaw in question is debilitating to any theory endorsing such a claim. Contingent self-defeat doesn’t only apply to the hyperbolic position, however. For those who detect a scent of irrationality and try to conﬁrm it by additional factors, the same problem remains. If you think that scent and reality converge when the disputants are epistemic peers, the same problem remains; if you think such convergence occurs when epistemic peers share all and only the same direct evidence, the embarrassing problem is still present. I will not offer my own e-mail address for proof, but will now relativize a bit, in case I am not your peer or your knowledge of epistemology exceeds mine. The thesis may not be contingently self-defeating for absolutely everyone, since there may be an epistemologist who is either smarter than all the rest or who has better information than all the rest. So the claim could be believed by the smartest or best informed, but by no one else. For everyone else, though, the thesis is so likely to be contingently self-defeating that it is no better off than the horrible hyperbole itself. Perhaps we could put the point this way. There are too many assumptions needed for the amended hyperbole to be adopted. One assumption is that you are smarter and better informed than anyone who disagrees with you about the amended hyperbole. But there is another assumption that even the most arrogant among us will have trouble with. There will be lots of epistemologists in the future. Many of them will be really smart. Many of them will know lots more about epistemology than we do. And, right now, Molliﬁcationists should agree with the rest of us that some of these incredibly smart and well-informed epistemologists will think the amended hyperbole is false. But maybe I’m wrong. The point, however, is that it is an assumption needed for the amended hyperbole to avoid contingent self-defeat that I’m wrong. No one should be so intellectually arrogant as to commit themselves to the view that there will never be a brighter and better-informed disputant of the amended hyperbole.

40

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

Moreover, going down the hyperbolic path should seem, once we begin investigating it, a completely wrongheaded way to proceed intellectually. We want to get to the truth and avoid error. How to do so is often not clear, but it is easy to describe things at a very abstract level in terms of the concept of evidence or indications of truth or signs of truth. We seek indications of truth and falsity, and adjust our views to such signs. We want to be the kind of individual who looks for evidence of truth and follows the evidence where it leads. Nowhere in this story of the intellectual life does our neighbor enter in. When we want to know whether p is true, we don’t turn to surveys to ﬁnd if there are people who think p is false. Because we don’t, we don’t in turn try to sort these naysayers into more or less intelligent, more or less informed. At best, other people are repositories of the evidence we seek, not a source of basic evidence itself. That is how we behave when seeking the truth, and any account of the epistemic signiﬁcance of disagreement needs to take account of this practice. One might object that this description is woefully inadequate because it leaves out the role of testimony. In the course of ﬁnding the truth and avoiding error, our neighbor plays a key role as a source of information. This point should be acknowledged because others are repositories of evidence (or else there is a chain of testiﬁers leading back to such a repository). Once we learn, however, that the person speaking has no information about the matter beyond what we also have, that person’s word ceases to have the ordinary power of testimony, which the word of others has in virtue of our lack of information on the subject. We can no more ﬁnd out that the moon revolves around the earth by an opinion poll than we can that all ravens are black by looking at more sheets of white paper. The lesson is that we can easily supplement the story above about truth-seeking to accommodate a place for testimony without coming anywhere close to Molliﬁcationism. There are complications that would need to be addressed in a full discussion of the relationship between the epistemology of testimony and the epistemology of agreement and disagreement, but my point here is simple enough that we can bypass the complications at present. We are considering the self-defeating character of simple views on which disagreement undermines rationality. In response, I pointed out that we don’t engage in inquiry by taking surveys as seriously as such views would require, and the response to this simple point was that this simple point ignores the importance of testimony. My response is that we don’t need to place the kind of importance on opinion surveys that the self-defeating views imply in order to take testimony seriously in our epistemology. Reﬁning this point to make it precise would be an interesting detour here, but the point is so obvious that no such detour is needed to appreciate it. Here a note of caution is in order, however, about this self-defeat argument. I have been careful above to avoid characterizing this argument as an argument that Molliﬁcationism is false. The kind of self-defeat in question is not with respect to the truth of the view, but with respect to the rationality of endorsing it. The argument is not intended to show that Molliﬁcationism is false. It is only intended to embarrass defenders of the view. It embarrasses them because it shows that it is hardly possible

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

41

to rationally endorse it. The worst cases of such self-defeat are cases where it is impossible to rationally endorse the view. Molliﬁcationism isn’t in this predicament, but is still in a situation one shouldn’t be sanguine about. The facts that lead to the kind of self-defeat in question are fairly ubiquitous and obvious, so as a matter of contingent fact Molliﬁcationists should abandon their view.2 There is a more direct complaint as well to lodge against the idea that compromise is the default order of the day in the face of disagreement. The more direct complaint concerns what I will call cognitive self-alienation. To see what cognitive self-alienation involves, let me begin indirectly. Each of us has certain cognitive abilities and disabilities. Some of us are better at math, some are better at noticing small details, some are better at visual detection, etc. In addition to these differences in cognitive abilities, there is also the matter of our own view of ourselves as to the level of ability we have in a given domain. This perspective on self can lead us to demur on changing opinion in the face of disagreement and can also lead us to defer to others in certain circumstances, such as when we view them as in a better position on the matter. Whether to demur or defer is, in part, a matter of our perspective on ourselves. In between demurring and deferring is desisting in belief, which is perfectly sensible from each of our points of view when our view of ourselves falls between a view calling for one to demur or for one to defer. To defer and to desist both involve change in view, but all three responses are often to be understood in terms of the relationship between a ﬁrst-order response to disagreement and a metalevel perspective on oneself that makes sense of the various responses of demurring, deferring, or desisting. The notion of compromise championed by Molliﬁcationists, however, is different. Recall that we are supposing that compromise is required in such a way that points of view can differ in the way imagined between X and Y (so that from X’s point of view p is true and from Y’s p is true). Since resolution of disagreement has failed by ordinary means, each is aware of the presence of someone who disagrees, and neither has a perspective on self that removes the disagreement. We thus have individuals involved in a disagreement whose views of self are incompatible with desisting or deferring, but rather call for demurring on the issue of changing opinion. And yet, compromise is required, according to the Molliﬁcationist. Such a requirement, however, insists that the perspective on self taken by the two individuals be abandoned or ruled irrelevant. Neither is allowed to take into account a perspective on self that calls for demurring, no matter how they came to such a view of self, and instead must respond in a way that would occur quite naturally if their perspective on self supported a response of desisting or deferring. The Molliﬁcationist position thus requires cognitive self-alienation, where one has a view of oneself and one’s abilities that coheres fully with

2 Adam Elga, in ‘How to disagree about how to disagree,’ to appear in Richard Feldman and Ted Warﬁeld (eds), Disagreement (Oxford: Oxford University Press, 2010) responds to a version of the self-defeat argument. His response focuses on the issue of truth, however, and it is clear that the threat of self-defeat is not a threat to the truth of Molliﬁcationism.

42

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

all of one’s prior experiences and one’s total conception of things, and yet which cannot rationally play a role in what attitude is legitimate at the ﬁrst-order level. Instead of a happy union and coherence between ﬁrst-order belief and metalevel attitude toward self, we have alienation and hostility between these levels required by the theory of rationality in question. A theory that insists that rationality requires such cognitive self-alienation has a serious burden of explaining how the theory is appropriately sensitive to the perspectival character of rationality. So the problem of cognitive self-alienation is a problem of failing to honor appropriately the obvious perspectival character of rationality. The Molliﬁcationist emphasis on compromise requires discounting aspects of a perspective in determining what attitudes are rational. In any case of disagreement, part of what needs to be assessed is whether attitudes of deferring, desisting, or demurring are most appropriate. The mere fact that someone has one of the three attitudes doesn’t by itself make that attitude appropriate, but which attitude is appropriate should be a matter determined from the point of view of the individual in question. That is the lesson of the perspectival character of rationality, and the implication of it is that the Molliﬁcationist demand for compromise will often require cognitive self-alienation because the perspective of the person in question makes demurring the appropriate attitude when the Molliﬁcationist insists on desisting or deferring.

2.2 Testimony and disagreement All of the above might seem so obvious and compelling that one can begin to wonder how one could favor the view that disagreement imperils rationality at all. There are two concerns here that might give pause. First, there is a concern about levels confusion, thinking that the argument above concerning cognitive self-alienation is best addressed by carefully distinguishing between the rationality of belief and the rationality of meta-beliefs about the initial belief. Second, there is a legitimate worry from the fact that sometimes we do reasonably abandon a view because we ﬁnd others who disagree. Both issues need to be addressed in order to make sense of the signiﬁcance of disagreement. We shall also ﬁnd important implications in addressing these concerns, implications regarding the idea that there is no optionality in rational opinion and regarding the idea that rationality is a matter of one’s total evidence. We can begin this sorting task by considering how to understand the epistemic signiﬁcance of disagreement. Here we might model the epistemic effects of disagreement in two quite different ways. One way is to take the opinions of others as evidence regarding the target proposition.3 Another way is to treat the opinion of another as a

3 I believe it is fair to say that the evidence model is the default position in the literature on testimony and in the literature that endorses some version of what I am calling Molliﬁcationism. See, e.g., Richard Feldman, ‘Epistemological Puzzles About Disagreement,’ in Stephen Hetherington (ed.) Epistemology Futures (Oxford: Oxford University Press, 2006), 216–36; Richard Feldman, ‘Reasonable Religious Disagreements’, in Louise

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

43

way in which people typically signal that they have information that shows that the claim in question is true.4 On such a model, the word of another isn’t immediately and directly evidence for the claim in question, but rather (on the reasonable assumption that ordinary folk are typically reliable) it is evidence that there is good evidence regarding the claim in question. If we add to this model some account of how to detach the evidence operator, some account of the conditions under which the move from ‘there is evidence that there is evidence for p’ to ‘there is evidence for p’ is epistemically appropriate, then the model tells us conditions under which the opinion of another is evidence for the claim in question, though only indirectly so. The difference between these two models concerns whether the word of another is evidence or meta-evidence, and there are various subtypes for each. Some subtypes appear when we think of disagreement in terms of defeaters. In the theory of defeat in epistemology, the usual view is that of John Pollock according to which there are two fundamental types of defeaters: rebutters and undercutters.5 Since a rebutter of the evidential relation between A and B is just evidence against B, we get a distinguishable model only by treating disagreement in terms of undercutting defeaters. An undercutter for my evidence for p is a claim that supports the view that the evidence isn’t a reliable indicator of the truth of p. For example, suppose you know that if you seem to see a pink elephant, your eyes can’t be trusted. This information functions as an undercutting defeater for the usual evidential support relation between seeming states and related beliefs in the context in question. There is theoretical pressure to model testimony and disagreement in similar terms, so it would be theoretically awkward to adopt an evidence model for one and a metaevidence model for the other. Some may also argue that we get a more elegant model if we ignore undercutting defeaters, so that the word of another is either evidence for the claim in question or evidence against. For example, if the evidence model of testimony is accepted, such a requirement would force us to model disagreement in terms of rebutting defeaters (i.e. evidence against the claim in question). This second requirement, however, encounters difﬁculties when we think in terms of the meta-evidence model. To model testimony meta-evidentially is to hold that the word of another is ﬁrst and foremost evidence that there is evidence for the claim in question, so modeling disagreement as a metalevel rebutting defeater would be to insist that disagreement is

Antony (ed.), Philosophers Without Gods: Meditations on Atheism and the Secular Life (Oxford: Oxford University Press, 2007), 194–215; David Christensen, ‘Epistemology of Disagreement: the Good News’, The Philosophical Review, 116, 2 (2007), 187–217; Adam Elga, ‘Reﬂection and Disagreement,’ Noûs, 41, 3 (2007), 478–502; and Adam Elga, ‘How to disagree about how to disagree’, in Richard Feldman and Ted Warﬁeld (eds), Disagreement (Oxford: Oxford University Press, forthcoming). 4 For a defense of such a view, see Thomas Kelly, ‘The Epistemic Signiﬁcance of Disagreement’, in Tamar Szabo Gendler and John Hawthorne (eds), Oxford Studies in Epistemology, vol. 1 (Oxford: Oxford University Press, 2005), 167–96. See also Marc Moffett, ‘Reasonable Disagreement and Rational Group Inquiry’, Episteme, 4, 3 (October 2007), 352–67. 5 An early account of the distinction can be found in Knowledge and Justiﬁcation, (Princeton: Princeton University Press, 1975).

44

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

evidence that there is no evidence for the claim in question. Such a view is obviously too strong. At most, disagreement can only be meta-evidence that whatever evidence exists for the claim in question is not adequate to support rational belief in the claim in question. Once put in this way, however, it is clear that the best meta-evidence model will treat disagreement in terms of undercutting, rather than rebutting, defeaters, since an undercutter aims at undercutting the quality of the connection between evidence and what it defeasibly supports. The lesson, then, is that treating disagreement in terms of rebutting defeaters ﬁts best with the evidence model, while treating disagreement in terms of undercutting defeaters ﬁts best with the meta-evidence model. At ﬁrst pass, modeling the epistemic signiﬁcance of testimony and disagreement at the metalevel seems to have an advantage. First, when one hears an utterance of ‘Bismarck is west of Fargo’, the initial and basic epistemic signiﬁcance of such an experience is to make reasonable the claim, not that Bismarck is west of Fargo, but that someone is saying that this is so. The nature of the sensory experience in question is best understood in terms of the latter claim rather than the former one. Given ordinary circumstances, the utterance also supports the claim that the person saying this believes it as well, and also that what is said is true. Thus, in ordinary circumstances, such an utterance provides rational support for the claim that Bismarck is west of Fargo. But the latter claim, at ﬁrst glance at least, doesn’t receive rational support in the immediate and direct way that the claim that something is F is immediately and directly supported when one is appeared to F-ly. Such a metalevel view also ﬁts quite naturally with our attitudes toward devices built to gather data. Imagine that you want to know the range of light conditions shining on a particular window in your home, from sunup to sunset. You don’t have time to watch all day, and besides, if you look down to record an observation, you’ll miss some information. So you want to build a detection device that records the observations automatically. Of course, you don’t just construct the thing haphazardly and then take its word for the truth (the idea, let’s say, is to build a device that generates a continuous graph plotting light wavelengths against time). You construct it according to what you take to be a good design plan and then calibrate it to make sure that it is accurate. Prior to calibrating it, it didn’t provide undefeated evidence about the matter in question. Prior to calibrating it, you had insufﬁcient information to conclude that the machine was reliable. You need reason to trust it to use it to acquire information about the target of inquiry. Now, in general, failing to have any such meta-information is not itself a defeater of the conﬁrming power of ﬁrst-order information: I don’t have to have information that I’m reliable before acquiring reasonable beliefs on the basis of inquiry. Moreover, calibrating the machine doesn’t change its capacities with respect to my inquiry: it doesn’t somehow turn the machine into a ﬁrst-order evidence generator when it was only a second-order evidence generator before. Clearly, prior to calibration, the graph generated is evidence that this machine is responding differentially to different lighting conditions. We have evidence that the device is responding to

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

45

information about the target of inquiry; it is a relay device, relaying information about something I want to know about. The information relayed is, of course, evidence: it is the information relevant to the question of what to believe about the target of inquiry. The question is how best to understand this production of evidence, and as described, the best model is fundamentally in terms of meta-evidence: its readings are evidence about the evidence regarding the target of inquiry. Viewed in this way, we have a nice story to tell about the machine before and after calibration. Prior to calibrating the device, we are not in a position to detach the evidential operator: we don’t have a system of information that allows us to conclude that there is evidence for a given claim on the basis of the machine readings which give us evidence that there is such evidence. Prior to calibrating the device, we can’t rely on the machine in arriving at reasonable beliefs about the target of inquiry, but after calibration, such a possibility exists, and the difference between these two situations is just the difference between only having evidence that we have evidence for the claim in question and actually having evidence for that claim. I conclude then that the meta-evidence model of testimony and disagreement has an initial advantage over the evidence model. There is, however, a concern that might undermine this advantage. Such a picture may seem to conﬂict with the basic trust necessary for early learning and the rationality involved in it. Much of what we know depends on taking the word of others prior to having anything like the information generated by the calibration process described above for measuring devices. In being rational animals, we come equipped with default cognitive mechanisms and as learning progresses we come to adapt these default mechanisms in various ways, including the development of wariness about sources of information. So the description of calibrating a detection device is relevant once a certain level of sophistication is achieved, but it can’t be the basic description that applies to all testimony. This point about early learning appears to threaten the meta-evidence model in the following way. On the meta-evidence model, there are two stages needed to use testimony as a guide to belief. In the ﬁrst stage, detachment of the meta-evidence operator must be appropriate in order to get the information that there is evidence for the claim in question, and in the second stage, a second detachment of the remaining evidence operator must be permissible in order to arrive at the conclusion that the claim in question is true. On the evidence model, only one detachment step is needed. Once seen in this way, there is some pressure to view the meta-evidence model as an overly intellectual model of how learning by testimony occurs. Take your favorite three-year-old, and consider what it means to say that this small child is detaching an evidence operator twice over. Isn’t it all a bit silly to think of three-year-olds engaging in the practice of detaching evidence operators? No it’s not, once we are careful in our understanding of what epistemic operator detachment involves. We can see this point most clearly by considering the evidence model itself, on which only one instance of detachment occurs. On the evidence model, the child just hears what is said and detachment is just a matter of taking

46

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

the word of the person in question: it is to come to believe the claim that is asserted by the source. What legitimates this detachment is the absence of grounds for doubting what is communicated. The epistemic engine of rational belief operates on its own, independent of awareness or inference by cognizers. A simple perceptual case makes this point obvious. Consider the epistemic principle that licenses believing it is raining when it seems to be. The principle says something like ‘when it seems to you that p, and you have no grounds for doubting that p, it is reasonable to believe that p.’ There is an epistemic operator on the consequent of this conditional, and detaching it involves nothing more than believing p when the antecedent conditions are satisﬁed. One doesn’t reason ﬁrst to the conclusion that it is reasonable to believe p, and then detach in a further step of reasoning. But detachment must be legitimate, in spite of not being some step of reasoning encoded in one’s psychology. The meta-evidence model may still seem objectionable even given this point, since there must be a ﬁrst detachment involving the conclusion that there is evidence for the claim in question. Surely it is implausible to view this child as forming beliefs about evidence prior to forming beliefs about what is asserted. So the meta-evidence model still seems to over-intellectualize the process of belief formation on the basis of testimony, even when the language of detachment is understood to require no step of reasoning from operator-governed content to operator-free content. I think this criticism misconstrues the meta-evidence model, however. On this model, in order for testimony to make a belief rational, an intermediate explanatory step is required in which one moves from having evidence of evidence for p to having evidence for p. But the intermediate conclusion need not be present in the cognitive system in question in the form of belief. The most common approach to detachment rules begins from the idea that high probability, or a high degree of evidential support, is not sufﬁcient for rational belief, but it is in the absence of defeaters.6 Defenders of the meta-evidence model might transpose this idea into the present context by claiming that the presence of evidence that there is evidence for p is itself evidence for p when there are no defeaters present to undermine the connection between the claim about meta-evidence and the claim about evidence. Thus, when the small child hears a parent say, ‘There are cookies on the table’, evidence exists in the form of an experiential state in the child. This state, on the metalevel model, is initially only meta-evidence, but in the absence of defeat, it is also evidence; and once we reach the point that it is evidence, it makes the belief of the child rational so long as there are no defeaters to undermine the connection between it and p itself. Given this approach, two types of defeaters must be absent, one kind concerning the relationship between the existence of metaevidence and the existence of evidence itself, and the other kind concerning the relationship between the existence of evidence and the object of belief. 6 For discussion, see Timothy Williamson and Igor Douven, ‘Generalizing the Lottery Paradox’, British Journal for the Philosophy of Science, 57, 4 (2006), 755–79.

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

47

The key to avoiding the overintellectualization charge, then, is to refuse to treat the model in terms of inference structures that require an additional piece of information in the head of the child before allowing rational belief in the claim in question. Instead, the evidence claims themselves are cast in terms of properties of the state itself that is evidence. Thought of in this way, the experiential state in question has the property of being evidence for p always and everywhere on the evidence model, but it has this property on the meta-evidence model only when there are no other pieces of information internal to the agent in question that defeat the supports relation between the state in question and the claim that this state is evidence for the target proposition in question. This supports relation, however, is not a relation between an experience and some further belief or experience, but is simply a relation between the experience and a proposition. Once we adopt the meta-evidence model, we are in a better position to understand the differences between testimony and disagreement. Merely hearing ‘huh-uh’ from the back of the room after reporting a low opinion of Bush as President isn’t ﬁrst and foremost evidence that he’s a good President. It’s not always clear what counts toward being a good President, but what is clear is that merely having Joe Bozo (construing this name as a mere Millian tag, of course) in the back of the room on Bush’s side isn’t such evidence. Disagreement may give us pause about what we believe, and such pause is most naturally understood in terms of an initial role in terms of undercutting legitimate conﬁdence about the adequacy of our grounds for belief. In special cases, it may combine with other information to function in stronger ways, as when one acquiesces to greater expertise or more information. What is common to all cases of disagreement, however, is that such disagreement ﬁrst and foremost provides a metalevel defeater: in short, an undercutting defeater of some sort. Such a view of testimony and disagreement gives us resources for addressing arguments with Molliﬁcationist tendencies. Consider David Christensen’s discussion of such a case in which he and his friend are dividing up their respective shares of a bill at a restaurant, and they come up with different answers: Given that my friend and I are generally reliable thinkers who have studied the same evidence, the fact that we disagree in such a situation will be explained by the fact that at least one of us has made a mistake in the case. But intuitively, the explanation in terms of my friend’s mistake is no more reasonable than the explanation in terms of my mistake. And I should acknowledge this by moving my belief toward hers.7 The argument here begins by assuming that both parties are equally competent and have the same evidence (and, we may assume, both reasonably view the matter in this way as well), and concludes that both parties are required to abandon belief in the conclusions they reached. The key claim in the example involves the idea of explaining the disagreement. There are two ways in which the disagreement can be explained: 7 David Christensen, ‘Epistemology of Disagreement: The Good News,’ The Philosophical Review 116 (2007), 187–217.

48

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

either in terms of one person having made a mistake or in terms of the other person having made a mistake. The key claim here is that neither explanation is more reasonable to adopt than the other. The idea here, I take it, is that if your information includes parity of evidence and parity of competence regarding one with whom one knows that one disagrees, then change in view is required. The passage itself actually employs the stronger position that the mere fact of parity itself requires change in view. This stronger position should be avoided, however, since the facts may be as stated and yet it be rational to deny them (to say nothing of being unaware of the disagreement itself). So we should focus on the weaker position, that includes among your background evidence the parity in question and the disagreement itself. The example is well-chosen to make plausible the principle enunciated, that parity information plus knowledge of disagreement requires attitude adjustment. The example, however, does not require Molliﬁcationism, interpreted in terms of requiring compromise of attitude even when one’s view of self supports demurring. In this case, demurring is not the natural attitude assumed to be present in each party, and hence the case is not a case of intellectual compromise at all. The contrast between Christensen’s case and cases in which demurring is appropriate helps with the task of separating the theory of testimony from the theory of disagreement. The difference concerns the defeasible character of epistemic support. Testimony provides reasons for belief because, in the usual case, no internal defeaters are present concerning the evidentiary power of testimony. But in cases of disagreement, the default position is that the power of testimony is in conﬂict with the belief in question and whatever evidential support there is for it. The testimony in question is meta-evidence that threatens the relationship between whatever evidence one has and the conclusion drawn from it, and thus is an undercutting defeater of some sort. Undercutters can appear in various guises here, but the view I ﬁnd the most promising is that disagreement threatens the idea that enhancements of one’s own epistemic position with respect to the claim in question would still conﬁrm what one believes. In a context in which one has formed a rational belief on the basis of evidence available, the evidence must not only favor that belief over other attitudes, it must be evidence that is adequate for some given attitude such as belief. For evidence to be adequate evidence, it must not only conﬁrm the claim in question but it must do so in a way that makes reasonable closure of inquiry concerning the claim in question. Of course, it is not necessary to form a belief to this effect, but if the total information available to one doesn’t license closure of inquiry, then one shouldn’t believe the claim in question even if one has good evidence for it.8 This distinction between the quality of one’s evidence and the question of closure of inquiry gives us two different types of 8 For further discussion of this point, and application to the issue of norms of assertion, see my ‘Assertion, Knowledge, and Lotteries’, in Duncan Pritchard and Patrick Greenough (eds), Williamson on Knowledge (Oxford: Oxford University Press, 2008).

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

49

undercutters that might be present. One sort threatens the conﬁrmation relationship itself, and the other threatens the connection to closure of inquiry, and the approach that is quite intuitive here is that disagreement undercuts in the second way. Either way, however, cases of disagreement are cases in which we have conﬂicting meta-evidence, with disagreement (on my preferred approach) providing a defeater of the claim that one’s evidence is adequate evidence, evidence sufﬁciently telling to warrant concluding inquiry. It does so in a context, however, in which one’s total body of information also provides rebutting meta-evidence against the defeating power of disagreement, since we are assuming that until encountering the disagreement, the body of information in question made the belief in question rational, made closure of inquiry on the issue rational, and thus also provided warrant for the claim that anyone who disagrees must be mistaken. The result is that cases of disagreement are cases with conﬂicting defeaters, and some sorting is required to determine which defeat relationships take priority.9 Such cases are not limited to cases of disagreement, but arise quite generally when we confront new information that ﬂies in the face of rational opinion. What is distinctive about disagreement is that the conﬂict occurs at the metalevel, at a level that effects the rationality of the target proposition only indirectly. In the ordinary case of new information, everything is occurring at the object level itself, and the interaction between old information and new information is merely a matter of replacing one body of evidence with another. In cases, of disagreement, however, the conﬂict is occurring essentially at the metalevel, and metalevel information only affects objectlevel rationality when certain conditions are met. The issue here is related to a point made by William Alston, a point about levels confusion in epistemology. Alston argued that it is not a good idea to let the skeptic use second-order claims to undermine the epistemic status of ﬁrst-order claims.10 For example, when the skeptic asks how I know that it is raining, and I cite some evidence for it, the skeptic may question how I know that this evidence is a reliable indicator of rain. That’s an interesting question, but one that is irrelevant to whether I know that it is raining, according to Alston. It may be relevant to the question of whether I know that I know that it is raining, but if we hold that evidence needs to reliably indicate in order to generate knowledge, we do not also need to hold that we must know that evidence is a reliable indicator in order to know. Keeping clear of such levels of confusion helps avoid a very quick regress argument for skepticism, since the skeptic can always ask about the metalevel n+1 when we answer his or her query about level n.

9 The requirement of sorting helps explain why dogmatic rejection of those who disagree is not always rational, in spite of being in the position of reasonably believing that those who disagree must be wrong. To argue otherwise is to be guilty of a scope confusion. What is true is that it is reasonable for you to believe that if someone were to disagree, they would be wrong. To reach the dogmatic standpoint, one needs the narrow scope reading: that if someone were to disagree, it would be rational to conclude that they are wrong. 10 William P. Alston, ‘Levels Confusions in Epistemology’, Midwest Studies in Epistemology, 5 (1980), 135–50.

50

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

Once we are sensitive to confusions between the object level and various metalevels, we might conclude that metalevel information is never relevant to the rationality of belief. Nothing so strong is warranted by Alston’s arguments however. All that Alston’s arguments show is that metalevel claims are not automatically relevant to object level claims, and thus that the skeptical arguments that rely on such a maneuver should be rejected. Moreover, we already have good reason to reject the blanket prohibition against metalevel information ﬁltering down to the object level, since undercutting defeaters are themselves metalevel features and yet alter one’s body of evidence with respect to a target proposition when they are discovered by the individual in question. So the issue in question can’t be resolved just by noting that disagreement functions at the metalevel. Instead, we need a story about the conditions under which metalevel factors ﬁlter down and when they do not. Testimony itself, as I have argued, is best modeled at the metalevel, but in a way that allows testimony to ﬁlter down as well: when there are no defeaters in the neighborhood, epistemic operator detachment is licensed at the metalevel, with the result that testimonial information that is fundamentally meta-evidence is also derivatively evidence when no defeaters are present to prevent detachment of the operator. This point leaves us with a satisfying result that, even though we can’t automatically treat facts at the metalevel as having signiﬁcance for the lower level, there are circumstances in which signiﬁcance ﬁlters down—namely, when there are no defeaters in the story. That result still leaves disagreement, however, stuck at the metalevel. In cases of ordinary testimony meta-evidence becomes evidence through the absence of defeaters, and evidence secures rational belief because of the absence of defeaters at this level as well. But disagreement is metalevel evidence that often cannot ﬁlter down on this basis because of the problem of conﬂicting defeaters. In such cases, the only way to go is up: conﬂict is resolved at the next metalevel. It is here that one’s perspective of oneself as a cognitive being provides an arbiter in the form of an overrider in one direction or the other. That is the lesson of the discussion of deferring, demurring, and desisting connected with the problem of cognitive selfalienation for Molliﬁcationism. Which of these options is the rational one to adopt depends on the degree of self-trust that remains or arises in the face of the disagreement. As self-trust dissipates, desisting and deferring become more rational in virtue of the loss of an overrider for the defeating role that disagreement plays in rational cognitive life. But when self-trust remains, there is an overrider for the defeating role provided by disagreement by those recognized to be peers or even superiors in terms of amount of data and general ability to analyze the data. One might question what to say when the self-trust in question is itself clearly irrational. In such a case, there will be other elements in the system of information that imply this irrationality, and then, of course, the overriding power of the level of selftrust is itself overridden in turn. In the absence of such further information, however, there is nothing epistemically problematic in relying on the level of self-trust that is present to be the ﬁnal arbiter in cases of disagreement. Moreover, there has to be some arbiter, since the situation is one of conﬂicting defeaters, and some resolution is required.

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

51

The upshot of this account of the difference between testimony and disagreement is an insight into the degree to which rationality is a subjective and perspectival matter. In slogan form, the claim being defended concerns the rational signiﬁcance of reﬂective ascent. As a person loses trust in self as a cognitive being, rationality dissipates in the face of disagreement, and an adequate theory of rationality needs to honor this point. What is rational depends on one’s total perspective, and a theory that treats two perspectives as identical even though one perspective contains signiﬁcant self-trust that is not irrational and the other does not is a theory that doesn’t rightly treat the perspectival character of rationality. When self-trust in an area is lost, views that used to be reasonable are no longer reasonable, and retaining them unchanged, irrational. The role for self-trust here is limited. There is no claim to the effect that in every instance of rational belief, there is an attitude of self-trust that plays a role in explaining the rationality in question. The attitude of self-trust arises in cases of disagreement because of the nature of the case in question. When we encounter someone who disagrees, that information constitutes a defeater that affects the rationality of our belief. The claim defended here is that the reﬂective ascent prompted by cases of this sort can provide the resources for overriding the defeaters provided by disagreement. Knowing that disagreement is by someone with inferior information or inferior abilities to analyze information would be just such an overrider as well, but such overriders are not the only kind available. In addition, overriders are available on the basis of reﬂective ascent, where a cognizer turns reﬂective in the face of disagreement, and thereby generates attitudes and information suitable for overriding the defeating effects disagreement portends. Once having turned reﬂective about peer status, the question of self-trust in the domain in question arises automatically and thus becomes relevant. In non-reﬂective cases where the issue is not raised, nothing said here suggests that some attitude of self-trust plays a role in the story of rationality, though of course nothing said here is incompatible with that view either.11

2.3 Implications for evidentialism and optionalism These results about rational disagreement threaten two important positions. The ﬁrst concerns the degree to which optionality is found in the domain of epistemic rationality. The most restrictive view is one that proclaims that there is one and only one attitude toward any given claim rationally compatible with any body of total evidence.12 The other position threatened is evidentialism itself. In order for these two positions to be consistent with the role that reﬂective ascent plays in the story of 11 For a defense and explication of a view that requires a stronger role for self-trust, see Keith Lehrer, Self Trust: A Study of Reason, Knowledge and Autonomy (Oxford: Oxford University Press, 1997). 12 In ‘Reasonable Religious Disagreements’, Richard Feldman labels this position the Uniqueness Thesis. David Christensen, in ‘Epistemology of Disagreement: The Good News’, speaks of Rational Uniqueness, and Roger White, in ‘Epistemic Permissiveness’, Philosophical Perspectives, vol. 19: Epistemology (Oxford, Blackwell Publishers, 2005), 445–59, argues extensively for the view though he never fully endorses it.

52

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

rationality, reﬂective ascent will have to change one’s body of total evidence and change it in a way that is mandated by prior evidence. Begin with the latter worry, the worry about evidentialism itself. We can escape the worry by trivializing the view, by insisting that one’s total evidence is everything relevant to the rational status of one’s cognitive attitudes. Evidentialists, however, do not typically think of their view as an utterly trivial one, so suppose we do not wish to trivialize the truth of evidentialism in this way. If we do not, then there are difﬁculties accommodating the rational signiﬁcance of reﬂective ascent, precisely those difﬁculties noted above concerning the relationship between object level bodies of evidence and metalevel factors. Once we grant the mundane fallibilist point that meta-evidence isn’t evidence, we face the possibility that meta-evidence might affect the rationality of belief without itself being part of one’s object-level body of evidence regarding the content of that belief. These points thus leave the door slightly ajar for the claim that reﬂective ascent is relevant to rational status but not in virtue of changing one’s total body of evidence with respect to the target proposition. It is worth comparing the role of reﬂective ascent in the face of disagreement with undercutting defeaters. Undercutting defeaters are metalevel claims, threatening the relationship between evidence and belief, but add to the body of evidence in question in virtue of being a defeater: where e is evidence for p, an undercutter d is a defeater in virtue of the fact that e&d is not evidence for p. (Compare: a visual experience is evidence for the claim that a given object is a particular color; but that very same visual experience, combined with knowledge that lighting conditions are abnormal, provides no evidence for that object being that color.) It is thus clear that adding an undercutter of this sort to the story changes the total evidence with respect to the claim in question. That is, in the case of typical undercutters, a metalevel claim ﬁlters down into the body of evidence in question precisely because an undercutter is a defeater and hence alters the total evidence with respect to the claim in question. There are two important differences between typical undercutters and the role of reﬂective ascent in the face of disagreement. The ﬁrst is that the type of undercutter provided by disagreement is unique. It is a type of undercutter that threatens closure of inquiry, which is not the same thing as an undercutter that threatens the evidential support relation itself. As noted already, in cases of epistemically rational belief, the body of evidence in question needs to provide support both for the content of the belief and for the claim that further inquiry into the matter is not necessary (though, of course, there is no need for the individual in question to form any belief toward this claim or even to have considered it). In order for this type of undercutter to ﬁlter down to the body of total evidence, one will need to maintain that a defeater with respect to legitimate closure is also a defeater with respect to the target proposition itself. That is, where e is one’s total evidence for B(p), then if u is such that e&u fails to support the claim that there is no need for further inquiry concerning p, then e&u fails to support p. This claim, however, I think is false, and its falsity is revealed by the lottery paradox. In such a case, one’s background of the set-up of a fair lottery provides evidential support

T H E R AT I O N A L S I G N I F I C A N C E O F R E F L E C T I V E A S C E N T

53

for the claim that one’s ticket will lose (assuming a lottery of sufﬁcient size). But there is also information blocking the conclusion that no further inquiry is needed concerning whether one’s ticket will lose: that is why it is a mistake to give the ticket to me, and why people check the newspaper to see whether they won. The concern raised by this point thus goes as follows. Critics of evidentialism might claim that there is a levels confusion occurring when evidentialists count undercutting defeaters in the body of total evidence concerning a given claim. The response, for typical undercutters, is that such undercutters are defeaters of the support relation between the evidence and the claim in question. But the kind of undercutter provided by disagreement is not of this sort. It is an undercutter relevant to the rationality of belief, but some further argument would be needed to show that it is an undercutter of the support relation between the evidence and the claim in question, since lottery cases are examples in which an undercutter with respect to closure of inquiry is not a defeater of the support provided to the claim in question. Similar remarks apply to the reﬂective ascent information that resolves the conﬂict of defeaters that occurs when disagreement is introduced in a context of rational belief. Such information is relevant to the rationality of belief, but it is not obvious that it is legitimately included in one’s total body of evidence relevant to the target proposition. One way to put this point is that such reﬂective ascent information is not exactly evidence itself, but information relevant to how one should weigh the evidence one has. Such a result is a ﬁtting one as well, since it is something close to a truism that there is both the evidence and what we make of it. To the extent that what we make of the evidence we have is a function of a perspective with healthy doses of reﬂective ascent, to that extent there is pressure to think of rationality not solely in terms of one’s total body of evidence regarding a target proposition, but also in terms that are less sanguine to the claims of evidentialism. To say that rationality is a function of one’s evidence and what one makes of it is, in this way, to say something that goes beyond evidentialism itself. The role of reﬂective ascent also presents problems for highly restrictive views about epistemic rationality. Once we specify a given body of evidence and a given proposition in such a way that disagreement is part of the story, individual differences in how to weigh the available evidence in determining what level of self-trust to display seem inevitable. Restrictivists might insist that such individual differences must be eliminated because only one particular type of reﬂective ascent is rationally possible, but such a claim is implausible. Some egos are fragile and others are not; some people have a higher degree of self-conﬁdence than others. In the arena of action, it would be indefensible to maintain that there is some precise attitude toward risk that all must display or that there is some precise level of self-conﬁdence required of each person. In our intellectual lives, at the level of reﬂective ascent, such restrictions are equally implausible. Even when we engage in the most deliberative type of reﬂective ascent in attempting to sort out whether to defer, demur, or desist, the only hard data we have comes in the form of multiple reference classes. Is this case of disagreement one of mathematical competency? Or is it a case of disagreement arising because I voiced an

54

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

opinion in the presence of a contrarian? There is evidence relevant to this sorting, to be sure, but the process clearly involves a weighing of the evidence one possesses as well. What is needed is an evaluation of which reference classes are the most relevant for the particular issue at hand, and the idea that such evaluation can by done in an algorithmic fashion on the basis of the underlying evidence plus acceptable rules of inference is a position that needs a powerful argument on its behalf, given how doubtful such a claim is in itself. Such deliberation requires judgment, and though some such judgments are better than others, that is a far cry from saying that there is a uniquely acceptable position that must be adopted. A more realistic account of such cases is that in precisely the same overall evidential situation, some people are more intellectually timid and others are more intellectually self-conﬁdent, and that neither character trait is somehow irrational or epistemically inappropriate. As a result of these trait differences, reﬂective ascent plays out along different paths, with the result that a given belief is rational for one but not the other. The best explanation of the phenomenon in question is that the role that reﬂective ascent plays in the story of rationality provides a strong argument in favor of some optionality regarding rational belief.

2.4 Conclusion Three important conclusions thus can be drawn from our discussion. The ﬁrst is a lesson about rational disagreement, that it is possible even in the face of known sameness of data and competence. Careful attention to the way in which the epistemic signiﬁcance of disagreement is modeled reveals two other points as well. One is the pressure to expand the evidentialist motto, moving from the claim that rationality is simply a matter of total evidence to the enhanced claim that rationality is a matter not only of total evidence but also of legitimate weighings of it. This last remark, put intentionally as a plural, signals the way in which one’s total perspective, including whatever levels of reﬂective ascent have been achieved, play a role in the story of epistemic rationality. Such reﬂective ascent has boundaries of legitimacy, and it may even be possible in certain cases that one has evidence to compel some precise level of self-trust. Whether or not such cases are possible, however, they are outside the boundaries of rational expectation. Some people are more intellectually timid than others, and thereby lose self-trust more easily in the face of disagreement. Some are so timid that the loss of self-trust may be irrational, but the most plausible view is that the typical case involves something less than precise levels of self-trust implied by prior commitments or experiences. In this way, the epistemic signiﬁcance of reﬂective ascent also reveals a limited degree of optionality in the story of epistemically rational belief.

3 Evidentialism and the Paradox of Parity Keith Lehrer

Evidentialism afﬁrms that one is epistemically justiﬁed in accepting that p if and only if believing that p ﬁts the evidence one has for p, as Conee and Feldman (2004) suggest in the leading exposition of the theory. The notion of ﬁtting the evidence remains open for analysis and further articulation. The idea that justiﬁcation of belief is a matter of the relation of belief to evidence raises a number of questions I wish to consider in this chapter. The ﬁrst concerns the believer. Suppose that the believer, though in possession of evidence, is not trustworthy in the way he or she evaluates the evidence. The second concerns the transparency of the relation of belief to the evidence. To make evidentialism into an internalist account, which is the intention, it appears that the relationship of ﬁtting the evidence must be something transparent to a trustworthy subject. These are the concerns of this chapter. Chisholm (1966) earlier confronted the issue of transparency when he argued that there are special states, self-presenting states, being appeared to in some way, for example, such that if one is in such a state, then necessarily it is evident to one that one is in such a state. The objection is that a person might have no conception of being appeared to in that way, or, if one had such a conception, one might believe that one is not being appeared to in the way one is being appeared to, or more cautiously, one can never tell whether one is in such a special state. Now suppose a person holding such skeptical beliefs is in the special state. It seems altogether possible that it would not be evident for such a person that he is in the state. It might be replied that it is evident, that believing ﬁts the evidence, that the person does have evidence that he is in the special state, in spite of his skeptical beliefs. But should we say that a person is justiﬁed in believing that he is in a special state when he believes that he cannot tell whether he is in the state or not, or even, when he believes that he is not in the state? We do not raise such an objection as being fundamental against evidentialism, for the evidentialist need not embrace the sort of view of evidence considered. What is important, however, is that one recognize that any account of evidentialism as a theory of justiﬁcation must account for the fact that justiﬁcation depends on the epistemic competence, or, as

56

E VI D EN T IA L IS M A ND D I S AGR EE M E NT

Lehrer (1997) has suggested, the trustworthiness of the subject in evaluating the relationship of the evidence to what is believed.

3.1 The condition of parity The caveat about the trustworthiness of the subject becomes salient when one considers cases of disagreement between subjects who consider the other as trustworthy as themselves in the evaluation of evidence. Consider an example, from Van Inwagen (1996) of two philosophers arguing about compatiblism in the free will controversy. They look at the same evidence, the same arguments, and they exhaust philosophical discourse. Let us say that they have reached a point of dialectical equilibrium where each member to the dispute agrees that further discourse will not add any evidence. One believes p and the other the denial of p. Now we are asked to suppose that each party to the dispute recognizes that the other is as trustworthy philosophically and epistemically as the other. This is the condition of parity. Can either party to the dispute be justiﬁed in continuing to believe what he did while recognizing the condition of parity? Can either party even be reasonable to continue to believe what he did? There are important issues here about trustworthiness, justiﬁcation, and reasonableness. Let us consider the issue of reasonableness ﬁrst and then turn to the issue of justiﬁcation. There are distinctions to be drawn before we are going to be in a position to answer the questions arising from the situations of parity that are germane to evidentialism. The ﬁrst distinction is one between acceptance and belief (Lehrer 1997). A person may on reﬂection refuse to accept what he believes. Belief is like desire and arises without our deciding to believe. Belief may also remain when on reﬂection we would prefer not to believe. In short, though the will may have some inﬂuence on belief, we may ﬁnd ourselves believing things we have never decided to believe and continuing to believe things we would prefer not to believe. This independence of belief is, if sometimes regretted, an important constraint on cognitive processing. Thus, a person may ﬁnd that he continues to believe that p though he recognizes the condition of parity. Even if he reﬂects that, given parity, it would be more reasonable to suspend belief concerning p, he may well ﬁnd that he continues to believe that p. This is not to be regarded as irrelevant to questions of reasonableness and justiﬁcation. Doxastic resilience or ﬁxation has some force in ratiocination and the reasonableness of it. As Reid (1785) noted in reply to arguments of Hume, the irresistibility of perceptual belief in the external world confronted by speculative philosophical arguments for skepticism may be the basis for an argument that those perceptual beliefs are justiﬁed. However, confronted with belief that p, though recognizing the condition of parity, a person may reﬂect on whether he should accept what he believes. Acceptance and belief may separate. Separation is not uncommon. For example, a person may continue believing that a friend is faithful, but reﬂect that the evidence is to the contrary and accept that the friend is not. In such a case there are conﬂicting propositional attitudes

E V I D E N T I A L I S M A N D T H E PA R A D OX O F PA R I T Y

57

toward p, the belief that the friend is faithful, which is a result of resilience, and the acceptance that that person is not faithful, which is the result of the evidence. The person may consider his own belief unreasonable as he accepts the denial of it based on the evidence. We may think of belief as a state that arises without our bidding and may remain against our will. On the contrary, we may think of acceptance as an attitude of positive evaluation based on the evidence. So the ﬁrst distinction is between belief, which may not be responsive to ratiocination or evaluation, on the one side, and acceptance which is the positive attitude resulting from evidential evaluation, on the other. The evaluation of acceptance may involve conscious reﬂection and ratiocination or it may result from a principle of evaluation appropriate to reﬂection and ratiocination acting as a rule of thumb or an evaluation surrogate of the latter. Some perceptual beliefs and memory beliefs have the evidence of sense and memory to warrant the acceptance of them without reﬂection. Some beliefs are accompanied by evidence for them, many memory beliefs for example, in terms of the feelings of clarity and certainty.

3.2 Reasonableness and justiﬁcation The next distinction concerns the relation of reasonableness to evidence. There is an immediate distinction between strong and weak evidential restrictions on reasonableness. Consider the two following deﬁnitions: SR. It is reasonable for S to accept that p only if the evidence S possesses more strongly supports p than the denial of p. WR. It is reasonable for S to accept that p only if it is not the case that the evidence S possesses more strongly supports the denial of p than p. The second deﬁnition is latitudinarian and allows for the reasonableness of accepting anything whose denial is not supported or favored. However, WR is probably something that might be better be rejected in favor of: UR. It is not unreasonable for S to accept that p only if it is not the case that the evidence S possesses more strongly supports the denial of p than p. SR and UR allow for the possibility that there may be things that it is not reasonable to accept but also not unreasonable to accept. Let us take SR and UR together as R. It is clear that R is relevant to the parity issue and evidentialism. In the case of parity, it might not be reasonable for a person S to accept that p in the parity condition by R, but it would not be unreasonable for the person S to accept p in the parity situation by R. We might say the same thing about a weak notion of being justiﬁed in accepting p meaning something like that S has some justiﬁcation for accepting that p and of being unjustiﬁed in accepting that p meaning that S has some justiﬁcation for rejecting that p. Consider the following two conditions:

58

E VI D EN T IA L IS M A ND D I S AGR EE M E NT

J. S is justiﬁed in accepting that p only if the evidence S possesses more strongly supports p than the denial of p. U. S is not unjustiﬁed in accepting that p only if it is not the case that the evidence S possesses more strongly supports the denial of p than p. This pair of deﬁnitions allows that a person might not be justiﬁed in accepting that p on the evidence he possesses but, at the same time, not be unjustiﬁed in accepting that p on the evidence he possesses. Indeed, in the parity condition, this would be so. It must be acknowledged that this account of justiﬁcation does not have as much plausibility as the account of reasonableness and unreasonableness offered in R. For it is natural to so use the notion of justiﬁcation so that one is unjustiﬁed in accepting something that is not positively supported by the evidence. That is, it is natural to say that if a person is not justiﬁed in accepting something, then they are unjustiﬁed in accepting it, while it is not equally obligatory to say that if a person is not reasonable in accepting something, then he is unreasonable to accept it. The explanation is that there is a theory of reasonableness that allows that it is not unreasonable to accept something that the evidence permits you to accept even though it does not mandate this. For a discussion of such a theory see Lehrer et al. (1967). Reason may permit the acceptance of something it does not require. We shall continue discussing reasonableness but the remarks apply to the weak notion of justiﬁcation articulated above.

3.3 Evaluation and evidence: the acceptance solution The implications of this account of reasonableness formulated in R for the parity condition allows acceptance as not unreasonable. Suppose, moreover, that I accept that p in the parity condition to put the matter in the ﬁrst person. The question that now arises is whether my reﬂective evaluation and acceptance of p could in any way add to the evidence I possess. Suppose that upon considering the parity condition concerning p, I evaluate acceptance of p positively and accept p, nonetheless. Should my reﬂective evaluation itself count as evidence in support of p? Notice that if my evaluation of the matter and my acceptance can offer any positive support for p, no matter how little, then my evidence might support p more strongly than the denial of p. On this account, my mere acceptance of p after the recognition of parity might upset the parity. The result is that my evidence, including that acceptance, supports p more strongly than the denial of p. So condition SR allows for the reasonableness of accepting p after recognizing the parity condition. Acceptance trumps parity. The result would be the solution of the problem providing evidential weight of acceptance itself. There is an oddity to the solution but also to the rejection of it. The oddity might be put this way. It is strange to suppose that merely believing something can add to the evidence for believing it. The evidence for believing something, someone might contend, has to be independent of the state of believing it. Put another way, the mere fact that you believe something cannot be a reason for believing it. The reasons

E V I D E N T I A L I S M A N D T H E PA R A D OX O F PA R I T Y

59

for believing something must be something other than belief itself that supports it. Call this view that a belief cannot provide evidence for itself the independence constraint. No belief can support, provide evidence, or supply a reason for itself. The constraint contravenes supposing that believing something offers any support for believing it. The denial of the constraint might be called the self-support condition. Some have thought that the self-support condition is a feature of a foundation theory of reasonableness often favored by evidentialism. I suppose that Firth (1949) held this view and perhaps C. I. Lewis (1946) held it as well. However, the self-support condition may equally be a component in a coherence theory of reasonableness. The reason is that coherence theories of reasonableness may hold that the reasonableness of a belief is internal to a system of beliefs containing it, and therefore, indirectly to itself (Lehrer 1997, 2000). One might, of course, apply the independence constraint to reasonableness within a system requiring that reasonableness arises solely from the relationship of a belief to other beliefs in the system outside itself. Once one notes that the reasonableness of a belief on a coherence theory of reasonableness depends on a relation of a belief to a system of beliefs, however, it is hard to see why one should exclude the possibility that beliefs acquire some positive evidential status from being members of the system of beliefs including themselves. Let us return with these reﬂections before us to the distinction between acceptance and belief. Acceptance is positive evaluation in terms of the evidence. Belief may not have the backing of reﬂection and ratiocination. But in the parity example, acceptance, which involves positive evaluation in the face of the evidential parity, is most plausibly construed as the result of reﬂection and ratiocination that brings one to the recognition of parity. My positive acceptance is my judgment in favor of p upon recognizing parity in the evidence. One might protest that the judgment is unreasonable and should be given no weight. However, if R is taken as the condition of what is unreasonable, my judgment cannot be condemned as unreasonable even if not praised as reasonable either. It should, moreover, be allowed the courtesy of not being considered unreasonable. There is a kind of existential moment of decision, to accept or not to accept that p. To decide, I must rely, at this point, on my own judgment. I must decide. Suppose my judgment under recognizing parity favors p. In the existential moment, I accept that p. Can I supply any argument for the reasonableness of accepting that p? Moreover, can I do this in a way that respects the judgment of the other? Here is my reﬂection. To arrive at the condition of parity, I had to use my judgment and reach a verdict on the evidence. I had to accept that the other argued cogently in a certain manner and many things that he averred. I had, ﬁnally, to accept that his evidence and his arguments were the equal in plausibility to my own. Unless I accept these things, the problem does not arise. Moreover, it must be reasonable for me to accept these things, for if it is not reasonable my acceptance of them is no evidence. But what has made my acceptance of all the conditions that lead to parity reasonable? Part of what made them reasonable may be the merits internal to the evidence and reasonings themselves. But that is not the whole story. I must, after all, judge and reach a verdict

60

E V I D E N T I A L I S M A N D D I S AG R E E M E N T

on these matters. So I must place my conﬁdence in my judgment and, moreover, my judgment must be worthy of my conﬁdence. If I lacked conﬁdence in my judgment, I would not have accepted the claims that produce the parity condition. If I had conﬁdence but were not worthy of my conﬁdence, then my acceptance of the claims would not convert them to the evidence that creates the evidential parity for me. Parity does not drop from nowhere. It results from judgment and from the reasonableness of it. Where does the reasonableness of that come from?

3.4 Trustworthiness and reasonableness I have argued before that my reasonableness of judgment and acceptance come from my trustworthiness (Lehrer 1997); I drew the following argument, and now it draws me. 1. 2. 3. 4. 5. 6. 7.

I am worthy of my trust in pursuing the objectives of reason. I am trustworthy in pursuing the objectives of reason. I am trustworthy in pursuing the objectives of reason in what I judge and accept. I am reasonable in pursuing the objectives of reason in what I judge and accept. I am reasonable in what I judge and accept in pursuit of the objectives of reason. I am reasonable in what I judge and accept. I am reasonable in what I judge and accept concerning the condition of parity.

I do not claim that each of these steps follows deductively from the preceding, though some do. The argument from 1 to 2 seems deductive based on the verbal reformulation. The argument from 2 to 3 seems to be cogent, but it is defeasible reasoning rather than deductive. Step 2 describes a general capacity to be trustworthy, and step 3 describes a general application of the capacity. The argument from 3 to 4 is again defeasible because the inference from the capacity to be trustworthy to the capacity to be reasonable is defeasible, even when the former capacity is exercised in terms of the goals of reason. It is, nevertheless, cogent because being trustworthy in the service of reason results in being reasonable. The argument from 4 to 5 seems deductive as the latter is a reformulation of the former. The argument from 5 to 6 is defeasible because it is consistent, though implausible, to deny the unqualiﬁed reasonableness of pursuing the objectives of reason. However, the crux of what is relevant to us is the step from 6 to 7. Now again the argument is defeasible for the simple reason that 6 describes a general capacity of being reasonable while 7 describes a more speciﬁc application of the capacity. Nevertheless, the point remains that my reasonableness of accepting the condition of parity is explained by my reasonableness of judging and accepting. Someone unconvinced by the formulation of the argument can understand the need to explain why it is reasonable for me to accept the condition of parity. To explain the reasonableness of the latter acceptance, one must explain why it is reasonable for me to accept the conditions contained in the condition of parity. That is explained by my capacity to be

E V I D E N T I A L I S M A N D T H E PA R A D OX O F PA R I T Y

61

reasonable in how I evaluate the claims considered, and, therefore, in what I accept. Put the matter another way. Unless I suppose that I am reasonable in what I accept, I cannot suppose that I am reasonable to accept the parity condition. If it is not reasonable for me to accept the parity condition, then the problem about the parity of evidence does not arise.

3.5 The reasonableness of acceptance and parity Now someone might ask what difference it makes to the parity problem and evidentialism if I must suppose that I am reasonable in what I judge and accept. The answer is that, having reasonably accepted the parity condition, I now consider whether to accept p given my acceptance of parity. It might seem that the reasonableness of accepting parity in some way entails that I should accept neither p nor the denial of p. But suppose I accept R and consider it not unreasonable to accept p. I must decide for myself, because the evidence is balanced. However, my verdict on behalf of accepting p tilts the balance. I see the balance of evidence in the parity condition, but, at the same time, I see that the reasonableness of my accepting what I do is what explains my reasonableness in accepting parity. The conclusion is this. If I proceed to accept that p in pursuit of the objectives of reason in accepting that p, then I am committed to accepting that the evidence supports p more strongly than the denial of p, the reasonableness of accepting parity not withstanding. For, I am committed to the reasonableness of accepting what I do, including, most saliently, the acceptance of p. The problem is this: how can I square accepting that the evidence does not support accepting p more strongly than denial of p, while at the same time accepting p? The answer is that my scale of evidence is not independent of my capacity to be reasonable in how I judge. In fact, there is no escape from the existential moment in epistemology any more than in practice. I have to choose whether to accept p or not in the pursuit of the objectives of reason. I cannot escape judging in the matter. I may judge not to accept that p or the denial of p, as the parity condition is created to suggest, and suspend judgment. But it is up to me. I may, reﬂecting on the parity condition, judge that I would be making a mistake to suspend judgment. Suppose in the pursuit of the objectives of reason I judge that p. That is where reason, my reason, leads me. My judgment may tip the scale of evidence. Thus, if accepting parity, I judge and accept that p in pursuit of the goals of evidence, nevertheless, the balance is tilted and parity is lost. I accept that p. I am, by the argument above, reasonable in what I accept, so I am reasonable to accept that p. By the condition R, if it is reasonable to accept that p, then the evidence supports p more strongly than the denial of p. How did that come about if I accepted the parity condition? My acceptance of p is part of the evidence. The crux is that parity may be undermined by acceptance. I cannot consistently maintain parity when I continue to accept that p. Parity itself becomes a kind of illusion. In a moment of epistemic absentmindedness I forget to notice that my judgment is itself a source of evidence. I look at the situation as though my judgment

62

E VI D E N T I A L I S M A N D D I S AG R E E M E N T

did not weigh anything on the scale of evidence. But it does. One may be inclined to think of evidence as being something independent of judgment. That is to forget the role of judgment in the creation of evidence. The preceding may seem strange. One may ask how my merely accepting something can change what my evidence supports. The reply is that there is an oddity in the idea that my judgment can give no weight whatever to how evident something is for me. I may receive information and engage in ratiocination from and about something. Then I confront the decision of what I am to make of all of that. I cannot escape the decision and the responsibility of deciding where my information and ratiocination lead me. I take that as obvious.

3.6 The parity paradox An objection remains. Does not my acceptance of the parity condition amount to my acceptance that, given the parity between myself and my opponent, the evidence supports p and the denial of p equally? Was that not the problem? What the reﬂections show is that there is an instability in the acceptance of the condition of parity. Or put it another way. There is a paradox. Call it the parity paradox. As I notice that the reasonableness of my initial judgment of parity rests on the reasonableness of my judgment for me, I must, at the same time, acknowledge that my judgment that p will upset the parity for me. Suppose my judgment favors p contrary to my initial judgment of equal support of the evidence for p and the denial of p. Then my commitment to being reasonable commits me to rejecting the initial condition of parity, of equal evidential support for p and the denial of p. As I accept that p, I have to also accept, what reasonableness requires, that the evidence supports p more strongly than the denial of p. The existential moment makes or breaks parity for the reasonable person. Being reasonable destroys the paradox of accepting the condition of parity and accepting that p. A reasonable person accepting that p relinquishes acceptance of parity. One might reply that this is the problem of parity for evidentialism. One might afﬁrm that acceptance of parity given evidentialism, or an evidential account of reasonableness, commits one to the suspension of acceptance. The problem for such an account is that it ignores or forgets the role of acceptance in evidential support for a person. One might be inclined to say that you have to decide how strongly the evidence supports p ﬁrst and then decide whether to accept that p. But that is a mistake. The acceptance of p for a reasonable person and the evidential support for p for the person are mutually dependent. Indeed, in the kind of example that motivates acceptance of the parity condition, the scale of evidential support rests in the hands of judgment. The verdict of judgment of the reasonable person is at the same time and in the same decision an acceptance of p and acceptance that the evidence supports p more strongly than not p. A further question arises. In the case of parity, does my acceptance of p make it true that my evidence more strongly supports p than its denial? The answer is that acceptance of p explains why my evidence more strongly supports p than the denial

E V I D E N T I A L I S M A N D T H E PA R A D OX O F PA R I T Y

63

of p. Judgment is not all there is to support evidence, for judgment may err in any application, but, at the same time, the support of evidence is, in part, a matter of judgment. The reasonableness of a person is, as I have argued (Lehrer 1997, 2000) a keystone in the arch of evidence of the person. The keystone is supported by the stones of evidence, but it supports the arch itself. Judgment holds the arch of evidence together. The precise account of the metaphor of the keystone arch I have often defended is as follows. The principle of reasonableness of a person in accepting what he does, like the principle of trustworthiness, has the virtue of explaining why a person is reasonable in accepting the principle itself. If I am reasonable in what I accept, for example, and I accept that I am reasonable in accepting that I am reasonable in what I accept, then my reasonableness in accepting that I am reasonable in this way is explained by my reasonableness itself. There is here an explanatory loop. It is important to note that both the truth of the principle and the acceptance of the principle are components of the explanation. Indeed, it is neither the acceptance of the principle nor the truth of the principle alone that is explanatory. The explanation is the loop. I called this loop theory (Lehrer 2007). What I note here is the relationship of evidentialism and the parity problem to loop theory. The keystone is, in fact, an explanatory loop. Once the explanatory power of the principle of reasonableness is noted, the application to the principle SR, one of principles of R, becomes apparent. For the principle supports the reasonableness of accepting that evidence supports some claim, including the claim of the principle itself.

3.7 Reid and the ﬁrst ﬁrst principle The matter of the loop was ﬁrst clariﬁed by Reid (1785). Reid noted various principles of evidence which he called ﬁrst principles, ones concerning perception, memory, consciousness, testimony, and others. Evidence, Reid proposed, is the ground of belief, or, in our terminology, evidence is that which supports beliefs, including beliefs about general principles of evidential support. The most general ﬁrst principle, his principle seven, which I have called the ﬁrst ﬁrst principle (Lehrer 1998) is one afﬁrming that our faculties are not fallacious, or alternatively, that our faculties are trustworthy. Now to say that our faculties are not fallacious entails that the principles and convictions of those faculties, the ﬁrst principles of the faculties, are ones that we are reasonable or, according to Reid, justiﬁed in accepting. The ﬁrst ﬁrst principle is, of course, a ﬁrst principle, and one that we are reasonable and justiﬁed in accepting. However, the ﬁrst ﬁrst principle has a priority in the order of evidence over the other principles, Reid avers (1785), which is why I have called it the ﬁrst ﬁrst principle. For one might, as the argument presented concerning trustworthiness above exhibits, appeal to the ﬁrst ﬁrst principle as evidence for all the other ﬁrst principles of our faculties. It says that they are trustworthy and not fallacious after all. He also notes, ﬁnally, that as that principle vouches for all the ﬁrst principles it, therefore, vouches for

64

E VI D EN T IA L IS M A ND D I S AGR EE M E NT

itself as well as a ﬁrst principle. So if evidence is the support of belief, then the ﬁrst ﬁrst principle, since it supports the rest, is evidence supporting the rest. But then, it supports itself as well. Since it supports all ﬁrst principles, it is a ground for accepting them, and it is evidence for itself. One might inquire in what sense the ﬁrst ﬁrst principle is the ground of the rest. The answer is simply that the truth of that principle explains the truth of the others, and, moreover, the trustworthiness of the principle explains the trustworthiness of the others. So the support of evidence consists of the explanatory power of the evidence to explain why the things it supports are justiﬁed and, indeed, why they are true. It is interesting that Reid is claimed by foundationalists because of the reference to ﬁrst principles as well as by reliabilists (Goldman 1999) and warrant theorists (Plantinga 1993). However, it is clear that there is a coherentist component as well. The ﬁrst principles conﬁrm each other including the ﬁrst ﬁrst principle, he insists, and remarks that the ﬁrst principles hang together like links in a chain. One who lifts one link must be prepared to lift the rest. The doctrine of mutual support between a set of principles is a coherentist conception. What is suggested by Reid’s views innocent of current debates is that the appeal to various ‘isms’ in current epistemology is a mistake. The important question about a philosophical theory is not how it should be classiﬁed but whether it should be accepted. Evidentialism as deﬁned above leaves open how we are to analyze evidence and the support of evidence. Our point is that the commitment to evidentialism contained in the principle SR—It is reasonable for S to accept that p only if p is more strongly supported than the denial of p by the evidence of S—allows for the mutual dependence of reasonable acceptance and the support of evidence. That mutual dependence is explanatory as we have noted. So, if a person accepts p when the person is reasonable in what he accepts, that can be part of the explanation of why the evidence supports the acceptance of p more strongly than the denial of p. Part of the explanation lies in the evaluation and judgment of the person concerning the evidence.

3.8 Judgment and evidential support What is easy to ignore and forget is that the judgment may change the strength of evidential support. A person may consider a condition of parity, and in the moment of judgment, upset the parity with which he began. One may acknowledge the antecedent condition of parity, as in the philosophical example, acknowledging the epistemic merits of the position of the other, but then one must decide. If one judges on the basis of the condition of parity to suspend acceptance, that is reasonable enough. If, on the contrary, one judges in favor of one side against the other, parity is thereby relinquished. Judgment in favor of that side carries the implication that the evidence favors that side. The point is that there is a kind of autonomy in the life of reason which cannot be abrogated.

E V I D E N T I A L I S M A N D T H E PA R A D OX O F PA R I T Y

65

I add a reason about why we should not automatically relinquish our autonomy and adopt a rule to suspend judgment in the condition of parity. The reason is that evidence is partly a matter of cognitive skills. They are, after all, what makes us worthy of our trust as we seek to discern truth from error. So once the evidence is laid before us, the work of applying those cognitive skills, the skills of judgment, lies before us. Appealing to a simple rule to relinquish judgment under the condition of parity has all the advantages of sloth over honest toil. Why trouble? One reason, which we may accept, is that a diversity of cognitive tools leading to diversity in what we accept is more effective for reaching truth than expertise. Mill (1859) thought this, Scott Page (2007) showed that it is a mathematical theorem. If a social justiﬁcation for trusting individual and diverse judgment is wanted, there is the proof that it is socially efﬁcacious for reaching truth and avoiding error. So self-trust, on which acceptance must be based under the condition of parity, is the source diversity and leads to truth when socially aggregated. I leave the details of this issue to another occasion.

3.9 Parity and belief This leaves open some further questions. What about the issue of the relationship of parity to belief? We began considering belief, after all. We acknowledge a lack of autonomy with respect to belief. So what if we ﬁnd that in the condition of parity we continue to believe as we did before? If the theory of reasonable belief mirrors that of acceptance, then we must say that a person is reasonable to believe that p only if the evidence of the person more strongly supports p than the denial of p for the person. So what should we say if the person, who is refusing to accept that p because of the acceptance of the parity conditions, ﬁnds that he continues to believe it nonetheless? Should we conclude that the person is not reasonable to believe that p? Notice that we may say that the person is not unreasonable to believe that p, but that allows that he may not be reasonable in believing it either. The issue is complicated by the consideration of the principle that reasonableness depends on optionality. When a person has options and chooses one, it is appropriate to ask whether the person is reasonable in what he chooses. But suppose he has no choice. Is it equally appropriate to ask whether he is reasonable? We do ask, and we do judge. Why? There are two reasons for such judgment. First of all, we wish to make it clear to ourselves that it is not reasonable to accept what he believes. Secondly, the judgment that a person is not reasonable to believe what he does may have some inﬂuence on what the person comes to believe eventually. The judgment may be a trajectory for change. Much the same applies to ourselves. We may judge that what we believe is not reasonable and refuse to accept what we believe even though we know it will not immediately change what we believe. But it may inﬂuence what we come to believe and bring belief in line with reasonable acceptance. That is to be desired. Belief is the more durable state, after all, more tightly locked into memory and the springs of action and inference. So judgment that a belief is not reasonable, though it will not

66

E V I D E N T I A L I S M A N D D I S AG R E E M E N T

automatically alter belief, may lead to change. The present condition of belief may remain unchanged, but we contribute something to the dynamics. Moreover, belief has an inﬂuence on what we accept, and often it is important that it does. Consider perceptual belief. It provides a causal connection of our evaluation system to the world that lies beyond it. Such belief still leaves open the question of whether we should believe our senses in the particular case. But the pursuit of reason requires that we give defeasible credibility to the beliefs delivered by our senses. We must accord the appropriate respect to belief as we acknowledge the need to evaluate the deliverances of it before the bar of evidence. We take the credentials of belief into account as we judge whether to accept them. That is why and how the judgment that a belief is not reasonable, which we may now fail to relinquish however we judge, is important in the life of reason. Moreover, as we reason about other matters, those beliefs we judge not to be reasonable are ones we refuse to accept in our deliberations and ratiocination. Judging a belief not to be reasonable, we discount it on the scale of evidence.

3.10 Conclusion: evidentialism and justiﬁcation This leaves us with the evidentialist theory of justiﬁcation with which we began. We shifted the thesis of evidentialism from one concerning justiﬁed belief to one concerning reasonable acceptance. We deﬁned a weak notion of justiﬁcation that was equivalent to reasonableness in terms of two deﬁnitions of justiﬁed and not unjustiﬁed acceptance earlier. J. S is justiﬁed in accepting that p only if the evidence S possesses more strongly supports p than the denial of p. U. S is not unjustiﬁed in accepting that p only if it is not the case that the evidence S possesses more strongly supports the denial of p than p. Let us call the conjunction JA. Two reﬂections are important. First of all, JA may be thought to be too weak by those who think that the mere fact that the evidence favors p over the denial of p is, though a condition of justiﬁed acceptance of p, not sufﬁcient for justiﬁed acceptance required for a condition of knowledge. Stronger conditions must be satisﬁed for a condition of justiﬁcation that suits the needs of analyzing knowledge in terms of justiﬁcation. The latter would require, as I have argued (Lehrer 2000), that objections to accepting p can be met or answered by a person. However, this use of justiﬁcation in the analysis of knowledge, though backed by a tradition, may be too restrictive to capture ordinary intuitions in the matter. The notion of justiﬁcation captured in JA may be more useful in the current context provided that we keep in mind what we said earlier about the intended use of the terms. The use was intended to capture a weak notion of being justiﬁed in accepting p meaning something like that S has some justiﬁcation for accepting that p and a weak notion of being unjustiﬁed in accepting that p meaning that S has some justiﬁcation for rejecting that p. Justiﬁcation so

E V I D E N T I A L I S M A N D T H E PA R A D OX O F PA R I T Y

67

articulated would, like reasonableness, depend on judgment and the reasonableness thereof. It would be pointless to rerun all the arguments concerning reasonableness to the notion of justiﬁcation expressed in JA. The arguments concerning justiﬁcation would be the same as those offered above. In closing, it is worth noting that the notions of reasonableness and justiﬁcation offered in an account evidentialism, whether R or JA, are intended to be epistemic notions whose goal is the pursuit of theoretical reason. Reason divides, however, into a practical and theoretical sphere. What it is reasonable to accept in pursuit of the goals of one splits off from what is reasonable to pursue in pursuit of the other. The account of evidentialism might not accord with a notion of practical reasonableness and justiﬁcation. Of course, some pragmatists who identify truth with what it is useful to believe, James (1907/1981) most notably, would demure at the distinction in his rejection of evidentialism. We assume it here without argument. Our conclusion is this. In the epistemic sphere characterized by evidentialism, reasonable judgment and acceptance are acknowledged to depend on the support of evidence. However, the support of evidence and evidence itself depend on acceptance and the reasonableness of it. The problem and paradox of acceptance in the face of parity dissolves in the solvent of the verdicts of judgment. There is a loop in the solution, but the loop is explanatory as I have argued. The explanation of reasonableness and the support of evidence is contained in the loop. Evidentialism is tied together in the evidential loop of acceptance.1

References Chisholm, R. M. 1966. Theory of Knowledge. Englewood Cliffs: Prentice-Hall, 44. Conee, E. and Feldman, R. 2004. Evidentialism: Essays in Epistemology. Oxford: Clarendon Press. Firth, R. 1949. ‘Sense-Data and the Percept Theory,’ Mind, 58/232: 434–65. Goldman, A. I. 1999. Knowledge in a Social World. New York: Oxford University Press. James, W. 1907. Pragmatism: A New Name for Some Old Ways of Thinking. Indianapolis: Hackett Publishing, republished 1981. Lehrer, K. 1997. Self-Trust: A Study of Reason, Knowledge and Autonomy. Oxford: Clarendon Press, Oxford University. ——. 1998. ‘Reid, Hume and Common Sense,’ Reid Studies, 2/1: 15–26. ——. 2000. Theory of Knowledge, 2nd edn. Boulder: Westview Press. ——. 2007. ‘Knowledge, Art and Autonomy: Loop Theory,’ John Dewey Lecture, Proceedings and Addresses of the American Philosophical Association, 81/2: 121–36. ——, Roelofs, R., and Swain, M. 1967. ‘Reason and Evidence: An Unsolved Problem,’ Ratio, 38–48.

1 I wish to thank Nathan Ballantyne for calling my intention to the importance of the paradox of parity and for his editorial and philosophical assistance.

68

E V I D E N T I A L I S M A N D D I S AG R E E M E N T

Lewis, C. I. 1946. An Analysis of Knowledge and Evaluation. La Salle: Open Court Pub. Co. Mill, J. S. 1859. On Liberty. Page, S. E. 2007. The Difference. Princeton and Oxford: Princeton University Press. Plantinga, A. 1993. Warrant and Proper Function. New York: Oxford University Press. Reid, T. 1785. Essays on the Intellectual Powers of Man. Edinburgh: James Thin. Reprinted in The Works of Thomas Reid, D.D., 8th edn, ed. W. Hamilton (1895) Edinburgh: James Thin. Van Inwagen, P. 1996. ‘It Is Wrong, Everywhere, Always, and for Anyone, to Believe Anything upon Insufﬁcient Evidence,’ in J. Jordan and Howard-Snyder (eds), Faith, Freedom, and Rationality. Lanham, MA: Rowman and Littleﬁeld, 137–53.

PART II

Virtue Critiques: Evidence and Inquiry Evidentialism appears to be a normative theory. It says what one ought to believe. In Evidentialism Conee and Feldman say ‘We hold the general view that one epistemically ought to have the doxastic attitudes that ﬁt one’s evidence. We think that being epistemically obligatory is equivalent to being epistemically justiﬁed. There are in the literature two other sorts of view about epistemic obligations’ (2004: 88). Feldman has written a number of things concerning this—including ‘The Ethics of Belief’, reprinted in Evidentialism: Feldman, Richard. 1988. ‘Epistemic Obligations,’ in James Tomberlin (ed.) Philosophical Perspectives 2, Epistemology. Ridgeview Publishing Co., pp. 235–56. ——. 1988. ‘Subjective and Objective Justiﬁcation in Ethics and Epistemology,’ The Monist 71: 405–19. ——. 1998. ‘Epistemology and Ethics,’ Routledge Encyclopedia of Philosophy, edited by Peter Klein and Richard Foley. Routledge. ——. 2001. ‘Voluntary Belief and Epistemic Evaluation,’ in Matthias Steup (ed.) Knowledge, Truth, and Duty: Essays on Epistemic Justiﬁcation, Responsibility, and Virtue. Oxford: Oxford University Press, pp. 77–92. ——. 2002. ‘Epistemological Duties,’ in Paul Moser (ed.) The Oxford Handbook of Epistemology. Oxford: Oxford University Press, pp. 361–84. ——. 2008. ‘Modest Deontologism in Epistemology,’ Synthese, 161: 339–55. Evidentialism was in fact intended largely as a response to denials of evidentialism which stemmed from normative theories concerned with human limitations. Conee and Feldman think that practical concerns should be kept separate from epistemic concerns. Guy Axtell also offers a forceful call to unite theory and practice. He clearly thinks epistemic normativity should be tied to the process of inquiry. He also thinks evidentialist views are too narrow in a number of ways related to this. First, he argues that an adequate notion of epistemic normativity should be diachronic, not synchronic.

70

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

Second, he claims that evidentialists are committed to a form of epistemic value monism called veritism: that the only epistemic value is avoiding believing falsehoods. Finally, Axtell is writing from ﬁrmly within the Pragmatist camp, so he thinks that the social consequences of a theory pertain to its acceptabilitiy. He clearly thinks that evidentialism has bad consequences for democracy. Jason Baehr revives concerns with this aspect of evidentialism’s normative component. He offers cases which purport to show that though some proposition p ﬁts one’s evidence, one shouldn’t believe p. These cases all involve some kind of vice displayed in the subject whether that be because of lazy investigation or actual malicious intent. He suggests that evidentialists add a virtue component to their view in order to accommodate such cases. He worries that if they don’t, they risk focusing on a property that lacks any real value. Keith DeRose’s essay at the beginning of Part IV also raises concerns concerning inquiry which are part of what motivate his skepticism about evidentialism. Thus it is an honorary member of this section.

4 From Internalist Evidentialism to Virtue Responsibilism Guy Axtell

The question has not yet been generally accepted in the Academy—let alone any answer agreed upon—whether the twin concepts of ‘rationality’ and ‘reasonableness’ are not interdependent ideas, of comparable authority and philosophical interest. Indeed, it is not always recognized that the two ideas can be distinguished. Stephen Toulmin, Return to Reason (2001, 2)

4.1 Introduction Evidentialism as Earl Conee and Richard Feldman present it is a philosophy with distinct aspects or sides: Evidentialism as a conceptual analysis of epistemic justiﬁcation, and as a prescriptive ethics of belief—an account of what one ‘ought to believe’ under different epistemic circumstances. This chapter will focus especially on the evidentialist ethics of belief and an associated set of views in a growing literature usually dubbed the epistemology of disagreement. I ﬁnd myself a discontent with evidentialism in both of its main senses however, and will begin with an undercutting critique of evidentialism in its ﬁrst and primary sense. Effectively rebutting their evidentialist account of epistemic responsibility would serve to undercut Conee and Feldman’s ethics of belief because they are explicit in endorsing only cognitive and not also moral evidentialism; the two sides of the evidentialist research program are asymmetrically related, their epistemological account of what it means to be a successful epistemic agent being the whole of the philosophic underpinning they provide for the extension of evidentialism into an ethics of belief. It is cognitive evidentialism that the undercutting criticism of Part II is primarily concerned with. It is perhaps too usual a practice in epistemology today for writers to concern themselves directly only with matters such as these—only, that is, with the purported theoretical pedigree of an account—and to accept on that basis whatever consequences for actual practice it may turn out to have. Yet any proposed ethics of belief, I begin by noting, can have quite profound consequences on how we perceive and treat others

72

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

whose beliefs and/or attitudes differ from our own. Views about the ethics of belief affect our judgment of the reasonableness and even the sanity of people who differ from us in their beliefs and attitudes. Such judgments in turn have a sometimes deep impact on the great cultural debates of one’s day and age. Any philosophical ethic of belief, let us therefore say, proposes a kind of ‘accommodation’ between the individual and the community of inquirers, an accommodation that can and should be evaluated pragmatically. Adherence to the evidentialist accommodation is considered by many to be the very meaning of the Enlightenment Challenge to pre-modern modes of thinking. Yet as Feldman concedes with some reluctance in recent papers, his account renders reasonable disagreement among public evidence-sharing epistemic peers impossible, or nearly so. As Feldman puts it in ‘Reasonable Religious Disagreements’, ‘open and honest discussions seem to have the puzzling effect of making reasonable disagreement impossible’ (2007: 202). My discontent derives in part from the ‘Victorian’ austerity of the evidentialist ethic of belief, but also and more pertinently for this paper, from concern that Feldman’s puzzlement at what he concedes is an ‘unfortunate’ implication of his approach far understates the deleterious practical consequences of his claim. Essentially then, the critical claims of this chapter are that Conee and Feldman’s evidentialist philosophy has both ‘weak roots’ (Section 4.2) and ‘sour fruits’ (Section 4.3). A concern with pragmatic consequences will at points lead us away from epistemology proper and into a broader discussion of the importance of a sound ethic of belief as a philosopher’s tool of mediation in our ongoing and often volatile ‘culture wars’ over reason and faith, science and religion, etc. Thinkers can mean different things by ‘peer’ and ‘public-evidence sharing’ of course, and this may well be an area where Conee and Feldman are not in as close agreement as they are on the more general contours of the two sides or aspects of evidentialism (Conee 2010). In offering a critique of the practical consequences of the evidentialist accommodation in Section 4.3, I will focus primarily on Feldman’s recent papers (2006, 2007) which directly contest the possibility of reasonable disagreement. In the quote at the start of this chapter, Stephen Toulmin asks fellow philosophers if they should not be concerned to delineate the different extensions and normative meanings of ‘rationality’ and ‘reasonableness’. In his sweeping critique of intellectualism, Return to Reason, from which this passage is drawn, he calls upon us to correct ‘a current imbalance between our ideas of “rationality” and “reasonableness” ’ (2001: 12). The evidentialist approach not only doesn’t allow for this distinction, it indeed drives the reduction of the latter concept to an idiosyncratically internalist construal of the former. But with a different epistemological approach we are also likely to arrive at an ethics of belief with a substantially different character; with a different enough approach we may advantage ourselves to ‘recover reasonableness’ and ‘return to reason’. In order to answer this call, the concluding section (4.4) will brieﬂy suggest why what seems profoundly dissatisfying in the evidentialist epistemology and ethics of belief is avoidable when we begin instead from a virtue epistemology. What I propose

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

73

as an alternative responsibilist accommodation conceives the standards of intellectual responsibility or reasonableness pertinent to the ethics of belief to primarily derive from evaluation of intellectually virtuous/vicious motivations and habits of inquiry— diachronic or longitudinal ‘reasonableness’—rather than deriving from the narrow, technical sort of epistemic signiﬁcance they call ‘synchronic rationality’. In this way we recover the senses of reasonableness and responsibility that naturally beﬁt philosophic concern with a sound and civic ethic of belief, from the evidentialists’ reduction of intellectual reasonableness to that obviously responsibility-eschewing sense of it that ‘synchronic rationality’ (or evidential ﬁt) denotes.

4.2 Weak roots Internalist evidentialism, this section will argue, is highly incomplete as an account of epistemic responsibility: it offers too narrow a conception of the epistemic, of the epistemic goal, and of what it means to maximize epistemic value or to be a successful epistemic agent. The critique of internalist evidentialism in this section will be less severe than John Greco’s (2005) claim that internalism is false as a thesis about any interesting or important sort of epistemic valuation and any corresponding sort of epistemic normativity, yet substantially more robust than Jason Baehr’s claim that internalist evidentialism only needs mending with a ‘virtue patch’ (this volume). If internalism was ever a consensus view about epistemic justiﬁcation, this is no longer true in contemporary epistemology, and the centrality of doxastic justiﬁcation in non-internalist epistemologies, and in virtue epistemologies in particular, makes necessary a thorough reconceptualization of the relationship between doxastic justiﬁcation (concerned with belief and knowledge), and personal justiﬁcation (concerned with agency and responsibility). There does seem to be a near consensus view among internalists, however, ﬁrst conceding the need to take account of doxastic justiﬁcation in an analysis of knowledge, and secondly asserting the adequacy of addressing this need by describing the agent as doxastically justiﬁed whenever they possess normal evidential or propositional justiﬁcation that p, and ‘base’ their belief that p on the evidence which propositionally justiﬁes it. This way of bringing the basing relationship into the analysis of knowing through a back door will not do for the non-internalist, however, who may allow internalism about evidential justiﬁcation, but will not equate evidential justiﬁcation with epistemic justiﬁcation, nor accept any general internalist condition on knowing. The internalist still fails to give the importance of the etiology of belief its due, conceiving a justiﬁable belief and a belief in fact justiﬁed for an agent, as separated only by a synchronic, timeslice act of ‘believing on the basis’. But virtue epistemologists like John Turri object, arguing that ‘doxastic justiﬁcation cannot be understood simply as propositional justiﬁcation plus basing’ (2010: 13). Doxastic norms (norms concerning practices of belief-formation and maintenance) can be diachronic or synchronic; that the diachronic norms arise from cross-temporal

74

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

or longitudinal evaluations does not of itself undermine their status as genuinely epistemic. Indeed, if the relationship between knowledge-possession and the etiology of belief runs deeper than the internalist analysis allows for, then either propositional and doxastic justiﬁcation are independent sources of epistemic value (see Axtell and Olson 2009), or ‘we should explain propositional justiﬁcation in terms of doxastic justiﬁcation’ (Turri 2010, 1; compare Greco 2005). In either case the internalist fails to acknowledge what I want to insist is a robust inter-relationship between epistemic responsibility of the sort that results in doxastic justiﬁcation, and the etiology of belief. Epistemic rationality or personal justiﬁcation has two aspects, diachronic and synchronic, and while neither is directly truth-linked it is arguably the former that is the more salient preﬁguring cause of doxastic justiﬁcation and of knowledge in most cases. A proper understanding of epistemic responsibility therefore shows it to integrally involve knowledge-attributors in considerations of the etiology of an agent’s target belief (Greco 2005). ‘Poor utilization of even the best reasons for believing p will prevent you from justiﬁably believing or knowing that p. . . . The way in which the subject performs, the manner in which she makes use of her reasons fundamentally determines whether her belief is doxastically justiﬁed’ (Turri 2010: 10). So what happens when we don’t accept the internalist account of epistemic justiﬁcation qua evidential justiﬁcation in the ﬁrst place? What if we think with Michael Williams that the account presupposes a ‘Prior Grounding Model’ of justiﬁcation that is skepticism-inviting, and further that as Williams writes in ‘Responsibility and Reliability’ (2008), the debate over internalist and externalist views of justiﬁcation lingers on after it has ceased to be useful owing largely to a tendency on the one side, ‘to adopt an overly demanding, hyper-intellectualized conception of what epistemic responsibility demands’ (2008: 1)? In that case I think we’ll see the externalist turn in epistemology as undercutting or at least problematizing claims that function to privilege propositional justiﬁcation as conceptually more basic than doxastic justiﬁcation, and as the only epistemic good that diachronic responsibility— responsibility in inquiry—is good for. We will challenge and I think reject what I call the Well-foundedness Formula, that doxastic justiﬁcation is just synchronic propositional justiﬁcation ‘plus basing’. Conee and Feldman help justify the claim that only synchronic epistemic rationality—one’s present time-slice response to evidence—matters in epistemic evaluation of agent rationality and justiﬁcation, in part by insisting that across-time responsibility or irresponsibility in inquiry raises ‘moral or prudential questions rather than epistemic questions’ (2004: 178). The ‘too narrow objection’ to this view of the epistemic has previously been made by Alvin Plantinga, and Feldman’s reply toes a Chisholmian line, where anything except raw synchronic rationality, the rationality of evidential ﬁt, is ‘irrelevant to the central notion of epistemic justiﬁcation’ (1988: 249). Also irrelevant to this judgment are said to be the long-term epistemic consequences of adopting the belief, as well as any consideration of how the agent came to hold what they consider the evidence bearing on that target proposition, ‘whether by conscientious enquiry or by avoiding potentially

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

75

troublesome information’ (2004: 189–90). Evaluation of the agent’s inquiry-directed motivations and conduct across time is possible, but is irrelevant to what one ought to believe at any given moment; synchronic evaluation is the only genuinely epistemological concern, or the only one they allow to factor into their calculation of epistemic value maximization. I suggest that the claim that Conee and Feldman repeatedly make, that evidentialism provides no guidance about what an agent should do (2004: 189), but only what she should believe at any moment about any given proposition, sets up a classic false dichotomy, and that this teaches us more about the shortcomings of internalist evidentialism than about the irrelevance of inquiry-directed habits and activities to epistemic responsibility and epistemic justiﬁcation. Dichotomizing as the evidentialist wants to between the synchronic and diachronic rationality, in order to deﬁne only the former as arising from a general intellectual requirement and the latter as always having a non-epistemological source out of moral or prudential considerations, reﬂects a doubtful strategy. The virtue-theoretic response to this Chisholmian deontology, for the obvious reason it takes intellectually virtuous habits and abilities/competencies to determine doxastic justiﬁcation, provides what I think is strong support for the force of the ‘too narrow’ objection to Conee and Feldman’s account of what epistemic obligation and success consist in. In a ‘salience contextualist’ or ‘inquiry pragmatist’ version of virtue responsibilism, as one alternative, judgments about what constitutes an agent’s total evidence in respect of comprehensive religious, philosophic, and political beliefs are viewed as ‘entangled’ with qualitative or value-laden judgments about prudential, and even at times moral factors that affect the agent’s efforts at inquiry. ‘Intellectual responsibility’ should accordingly be acknowledged as the ‘thick’ concept common language takes it to be. In lieu of a cogent reduction plan, we should not assume that the sense or senses of responsibility that inform our ethics of belief are of a nature radically different than this. If there is a valid philosophical interest in synchronic epistemic justiﬁcation—and I have not denied that this is the case—it is not in a direct sense the interest in what epistemic justiﬁcation is, nor is it the interest in the sense or senses of responsibility that bear most directly on the ethics of belief. Especially when we attend to what we actually have meta-cognitive control over, our most fundamental doxastic responsibilities are diachronic; synchronic rationality is underwritten by responsibility in cultivating and exercising virtuous doxastic habits. Admittedly I offer no direct argument here for salience contextualism and for the entangled view—pragmatic/diachronic encroachment upon what the evidentialist understands as the purely epistemic (see Axtell 2010 for such a view). But the burden of proof does seem to me largely to fall on those whose claim it is that the reduction is possible: that we can always clearly factor out the purely epistemic considerations of ‘ﬁt’ (as a descriptive state of affairs) in attributions of knowledge, from agent/attributor interests and from evaluation of the agent’s own selected investigative methods and strategies. Perhaps when readers receive from Feldman his long-anticipated theory of

76

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

evidence we will see just how this reduction is supposed to proceed in application to cases like the theism/non-theism debate. But at present I would characterize Feldman’s account of these issues in the ‘Ethics of Belief ’ chapter of Evidentialism and in essays like ‘Epistemic Obligations’ (1988) and ‘Epistemic Duties’ (2002) to be carried largely by unargued assumption of a fact/value dichotomy, together with persuasive deﬁnition of key concepts like ‘epistemic responsibility’ and ‘reasonable belief ’, persuasive deﬁnition that would have us relocate these concepts quite a distance from their broader everyday connotations. But to go further into the substance of the matter, our ‘too narrow’ objection to their proposed restriction of epistemic value to maximizing synchronically rational belief can be framed as a challenge to the conception of epistemological axiology with which the evidentialists would have us begin. Conee and Feldman restrict the epistemic to the relation of beliefs to evidence. For them ‘the fundamental epistemic goal is just to have reasonable beliefs’, and taking this to be the case they further posit that ‘nothing about evidence gathering or the like follows as a means to that goal’ (2004: 188). This approach to justiﬁcation has deep roots in a Chisholmian conception of the tasks of epistemology. But virtue epistemologies contest that conception, routed in single-source epistemic value monism. They often posit that understanding, and not just the veritistic goals of true belief and knowledge or the internalist goal of maximizing synchronically rational beliefs, ranks highly in our epistemological axiology. If this rebuttal is sound, this has direct bearing on the cogency of Feldman’s claim that evaluation of the conduct of inquiry is a topic for a theory of practical reasoning but ‘irrelevant’ to a theory of epistemic justiﬁcation. In any conception of epistemological axiology that recognizes a more diverse range of epistemic goods, or that simply gives pride of place to understanding rather than to propositional knowledge within our cognitive economy, epistemic value will not be measured by synchronic rationality alone. Although we have here been largely limited to a critique of Feldman’s evidentialist epistemology, I have argued elsewhere more constructively for the genuine contributions to epistemic value of the intellectual virtues, illustrating how the virtues, as diachronic traits that bear upon the quality of our motivations and efforts at inquiry, preﬁgure doxastic justiﬁcation, and why they remain central to appraisals of agents that matter most to us, philosophically and scientiﬁcally (Axtell and Olson 2009). Arguments from the paucity of the evidentialist’s conception of epistemic axiology and from the positive contributions of diachronic rationality to epistemic value will also bear on Feldman’s and other philosophers’ endorsement of the Rational Uniqueness Thesis (hereafter RUT). This is the claim that given a body of evidence, there is only one reasonable doxastic attitude to take towards any proposition, where possible doxastic attitudes include believing, disbelieving, and suspending judgment. Once both degrees-of-belief and commitment-qualiﬁcation through epistemic humility are set aside, RUT becomes almost a foregone conclusion. In terms of ﬁt or doxastic attitude, there will always be one porridge that is too cold, another too hot, and a third that is just right.

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

77

With RUT and the role it has recently played in assertion of the impossibility of reasonable disagreement, we have to be cognizant of how its application already presupposes cognitive evidentialism. Thus the explicit defense of RUT in Evidentialism derives from thesis O2: that an agent ought always to have just exactly that attitude towards a proposition supported by his/her evidence at that particular moment. This in turn is best supported by V3: that being synchronically rational at every moment is uniformly what it is to constitute epistemic success and to maximize epistemic value (2004: 185, 258). But we have already given sufﬁcient reason to doubt V3, namely its dependence upon orthodox Chisholmian epistemic value monism. So to the extent that we adopt the richer conception of epistemological axiology that I argue elsewhere is a key lesson of recent work on the ‘value problem’ and value-driven epistemology (see Axtell 2008), then V3 can be counted false and no longer provides the needed support for O2. Thus RUT in turn, simply lacks the cogency with which much on the recent literature on epistemic disagreement invests it. This would be a good thing, as I’ll elaborate in Section 4.3. So from the present perspective it can be concluded that at two levels—that of how to maximize epistemic value when adopting an attitude towards a proposition, and that of how the ethics of belief is claimed to be governed only by consideration of the synchronic rationality of the agent’s doxastic attitudes—the evidentialist account looks quite doubtful. If this is the case then as we began by saying, it can’t provide the philosophic support that cognitive evidentialists rely upon it to provide in order to motivate evidentialism in its other main aspect—as a normative ethics of belief and/or a normative account of proper responses to peer disagreement/diversity.

4.3 Sour fruits Thus far I have followed an approach of undercutting evidentialism as an ethics of belief by criticizing its roots in the internalist evidentialist account of epistemic justiﬁcation. In this section we’ll ask whether the ethic of belief Feldman defends in ‘Reasonable Religious Disagreements’ and other recent papers is likely to be of value or disvalue as a tool of mediation in our present-day culture wars, including especially clashes of religious and secular culture. In the paper we are focusing on, Feldman reﬂects upon a course he recently taught in philosophy of religion, and attempts to show the falsity of ‘the tolerant and supportive view’ (hereafter TSV) he says a good many of his students had, both towards disagreement between believers and skeptics, and between (presumably moderate) adherents of different world faith traditions. He frames his argument around two questions: Q1 Can epistemic peers who have shared their evidence have reasonable disagreements?

78

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

Q2 Can epistemic peers who have shared their evidence reasonably maintain their own belief yet also think that the other party to the disagreement is also reasonable? (2007: 201). The burden Feldman places on his students to support TSV is not to argue just that religious beliefs can be held by ‘generally reasonable folks’ but that people are ‘epistemically reasonable with respect to their speciﬁc beliefs’. He argues that while reasonable disagreement (Q1), even with mutual acknowledgment (Q2), may appear initially plausible, we must ultimately come to negative answers to both questions. Epistemic peers who have shared their evidence cannot reasonably come to different conclusions: That is, ‘there cannot be reasonable disagreements of the sort I was investigating’, and ‘I cannot make good sense of the supportive and tolerant attitude my students displayed’ (2007: 213). I want to defend the students’ intuitions, while also exploring the possibility of turning Feldman’s or other RUT-defenders’ ‘unfortunate’ dismissal of reasonable disagreement back into a pragmatic test of the value or disvalue of his ethics of belief for mediating our most serious cultural debates. An old adage going back to Hume is that errors in religion are dangerous; those in philosophy only ridiculous. I’ll suggest that this is not so in the present case of using cognitive evidentialism plus RUT to undermine reasonable disagreement among evidence-sharing epistemic peers! My critique of the practical implications of accepting the position Feldman defends begins with William James’ characterization of ‘faith-tendencies as extremely active psychological forces constantly outstripping evidence’ (1979b [1911]: 112). The ‘faithladder’, thought James, was no logical chain of inferences, but neither is it usefully judged by the intellectualist standards of skeptical rationalism, that both demands establishing a threshold of sufﬁcient evidence for reasonable belief, and denies that that posited threshold is ever met. Moral evidentialists like William Clifford (1999 [1877]) are discontents as well, not of the evidentialist accommodation but of what Richard Rorty calls the Jeffersonian Compromise: mutual acquiescence to viewing religious beliefs as private in a sense of having little direct bearing on social/political order, but as still relevant to, and possibly essential for, individual perfection. Jefferson ‘set the tone for American liberal politics’ when he wrote in Notes on Virginia that ‘It does me no injury for my neighbor to say that there are twenty Gods or no God. It neither picks my pocket nor breaks my leg.’ Citizens of Jeffersonian democracy, the compromise holds, ‘can be as religious or irreligious as they please so long as they are not “fanatical”’ (Rorty 1991: 175). The problem is that making this important qualiﬁcation—‘so long as’—involves noticing the differences between radical and moderate forms of religious ﬁdeism in terms of the faith ventures they respectively authorize. Yet a cognitive evidentialist like Feldman can no more acknowledge a difference between the epistemic responsibility of moderate and radical ﬁdeists than can he acknowledge a difference between reasonable and synchronically rational belief. There is no more principled basis in his philosophy for the one distinction than for the other.

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

79

It is a practical and oftentimes a political decision explicitly to reject the Jeffersonian compromise, whether in the direction of hegemony of the ‘home religion’ or in that of an aggressive campaign of secularization or atheism. Conee and Feldman present themselves as holding the view ‘suggested by the words of Locke and Clifford that our epistemological duty is to believe as the evidence we have dictates’ (Feldman 2002: 364). But Locke and Clifford on closer examination are not nearly so closely aligned as this suggests. Clifford’s argument was basically that ‘Belief, that sacred faculty which prompts the decisions of our will, and knits into harmonious working all the compacted energies of our being, is ours not for ourselves, but for humanity’ (Clifford 1999: 74). Quite incompatibly with this claim, Locke’s conceptual pluralism is exempliﬁed in his Essay Concerning Human Understanding (1997 [1686]), where he writes that, ‘Since . . . it is unavoidable to the greatest part of men, if not all, to have several opinions, without certain and indubitable proofs of their truths, it would, methinks, become all men to maintain peace and the common ofﬁces of humanity and friendship in the diversity of opinions.’ There are quite difﬁcult issues that the epistemology of disagreement/diversity raises, and it is to Conee and Feldman’s credit that they helped bring these issues to centerstage in epistemology in recent years. But it is an ironic outcome if an approach Feldman describes as in part motivated by Locke’s philosophy should come around to problematize central values of Lockean reciprocal amity—the afﬁrmation of ‘common ofﬁces of humanity and friendship in the diversity of opinions’. Let us investigate further whether this might indeed be the case, by focusing more closely on Feldman’s denial of reasonable disagreement. The unfortunate fruits of Feldman’s evidentialist account of the epistemology of disagreement might be pointed out by noting certain literatures that RUT and its associated conception of reasonable disagreement appear to have negative implications for. I will survey these literatures brieﬂy ﬁrst, and then focus my strongest arguments around the third. The ﬁrst literature that evidentialism compromises is the work of philosophical and religious pluralists who explicitly base pluralism upon the ambiguity of total public evidence supporting religious and naturalistic ‘hypotheses’ (viewable as the ‘hard cores’ of very large scale alternative research programs). This includes leading proponents of religious pluralism (Basinger 2002; Bishop 2007), as well as philosophic pluralists who draw upon the same assumption (McKim 2001; Bernstein 2003). Religious pluralism is a theistic account, and I am not saying the evidentialist needs to accept its premises, but it is certainly a perspective that might be informing the perspective of some of Feldman’s students. For example, it is often possible to counter a leaning of intolerance or disrespect by a ‘moral case’ against salviﬁc exclusivism. ‘The idea that the most devout and ethical persons of non-Christian faiths might ﬁnd themselves in hell simply because they do not believe one of the essential tenets of Christianity seems difﬁcult to reconcile with God’s moral perfection’, and motivates inclusivist and pluralist sympathies (Himma 2002: 1). Now if we are epistemically tied to RUT, what Kenneth Himma calls this ‘moral high ground case for salviﬁc pluralism’ appears to be undercut, the former working against the latter. The constructive work that recognition of an epistemic situation of ‘religious ambiguity’ is supposed to serve in a pluralist conception

80

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

of religious truth and/or salvation falters, for the reason that the only entitled attitude to hold towards a proposition one acknowledges isn’t unambiguous in evidence is, according to Feldman, the attitude of strict suspension (as we’ll see more fully below). The second literature potentially problematized by Feldman’s position in the epistemology of disagreement is the theory of deliberative democracy, by which I mean a whole family of contemporary democratic theories that make central use of the Rawlsian thesis of reasonable pluralism among people with respect to their comprehensive religious, philosophical, moral, and political doctrines. Rawls describes reasonable pluralism not only as a ‘fact’ about a democratic polity but also as the ‘long-run outcome of the work of human reason under enduring free institutions’ (Rawls 1996: 129). This Rawlsian position called reasonable pluralism is a modern extension of Lockean reciprocal amity. That the burdens of judgment arguments are ones Rawls describes as epistemologically-driven indicates that deliberative democratists do make demands upon the epistemology of disagreement/diversity and the ethics of belief. So there is at least a deep tension here, one that Alan Hazlett (2009) puts in a quite useful way by deﬁning ‘Epistemic Liberalism’ (closely akin to the TSV claim of Feldman’s students) as a contrary to RUT, and then simply asking whether post-Rawlsian theories of political liberalism depend upon there being a connection between their political theories and Epistemic Liberalism. If so, then it appears no self-respecting deliberative democratist can be an evidentialist. One might object that while there is tension, there is no genuine contradiction here: Our commitments to toleration, as civil or political commitments are of a different order, and are not impugned even if the epistemological underpinnings Rawls thinks it important to give to reasonable pluralism turn out to be unsound. But for the sake of argument, Jean Jacques Rousseau considered our problem and wrote, ‘Those who distinguish civil from theological intolerance are, to my mind, mistaken. The two forms are inseparable. . . . Wherever theological intolerance is admitted, it must inevitably have some civil effect’ (1987 [1762]: 134). Evidentialists should not be ﬂippant about the impact of accidentally undermining epistemological arguments that much modern deliberative democracy appears to lean upon. The tension, as I think attention to our third literature shows, exists not just between RUT and these Rawlsian arguments, but between RUT and proper recognition of deliberative virtues such as ‘reciprocity’, ‘magnanimity’, ‘openness’, themselves. But these are difﬁcult questions that bear further study, as does the epistemology of liberal democracy more generally. The ﬁnal literature and the one I want to argue Feldman’s epistemology of disagreement/diversity most clearly compromises, is a literature focused around ‘friendly theism/atheism’, and religious toleration though epistemic humility (see Greco 2008; Kraft and Basinger 2008). These authors argue in favor of ‘friendly theism’ and ‘friendly atheism’ supported through epistemic humility, against the non-reciprocating ‘unfriendly’ versions of each that tend to predominate in the polemical discourse of our present-day ‘culture wars’ over reason and faith, science and religion. This is more straightforwardly a matter of prescribed attitude towards disagreement among peers, and is not

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

81

complicated by the distance between one’s epistemology and one’s political life. The clearest support I could give to my claim that Feldman’s ethics of belief is of disvalue in mediating these culture wars would be reasons to think that the combination of cognitive evidentialism and RUT in fact destabilizes the ‘friendly’ versions of both theism and atheism while empowering the alternate, ‘unfriendly’ versions of each. To show this let’s now return to look more directly at the reasons behind the negative answers Feldman provides to his own Q1 and Q2. The argument that Feldman gives for his skepticism about reasonable disagreement refers us directly back to RUT, and to that internalist account of epistemic justiﬁcation and value which we directly sought to undercut in the previous section. The crucial claim Feldman derives from it is that there are ‘really only two’ potentially philosophically respectable responses to serious peer disagreement, those which he terms the Hard Line model (hereafter HL), and the Modest Skeptical model (hereafter MS). Let’s consider these two evidentialist-approved ‘responses’ to peer disagreement, and the forced option Feldman alleges between them. HL: The Hard Line view The ﬁrst view or model that Feldman says may be the distinctly rational one to take towards your disagreement with another individual is the model that sanctions the response of, ‘I’m reasonable; you’re not’. Disagreement under conditions of shared public evidence indicates not just error, but indeed the epistemic irrationality or unreasonableness of one of the disputants. ‘The hard line says that the evidence they share really must support one view or the other, and the one whose belief ﬁts the evidence is the rational one’; hence ‘Either the theists or the atheists are rational, but not both. There can be no reasonable disagreements’ (2007: 211). The HL model is quintessential evidentialism, and while Feldman might not approve (Feldman describes himself as a ‘complacent atheist’, not an ‘aggressive’ one), a prominent example of the HL model put to work in popular culture wars is Sam Harris’s Letter to a Christian Nation, where at the outset he sets aside all Christian liberalism (as inconsistent non-sense), in order to pick out as his target audience just such people as really know what Christianity is: We agree, for instance, that if one of us is right, the other is wrong. The Bible is either the word of God, or it isn’t. Either Jesus offers humanity the one, true path to salvation, or he does not. We agree that to be a true Christian is to believe that all other faiths are mistaken, and profoundly so. (Harris 2006: 3).

Brieﬂy, the main problem with HL that we’ll return to is the one Peter van Inwagen (2010) worries over in ‘We’re Right; They’re Wrong’, that sanctioning this model/ attitude towards disagreement as the default view and as a kind of dictate of reason may sound good in theory, but in practice lends itself to rationalizing dismissals of the reasonableness of those we disagree with (and all the better with moderates summarily dismissed). With the sanction of the HL attitude behind us, do we not often come to

82

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

expect more intellectual virtue of others than we do of ourselves? Another worry is that Harris’ employment of an HL rhetorical strategy arguably encourages its Christian evidentialist mirror image, characterized by equally aggressive intelligent design theory and other examples of overt religious rationalism such as that by disinterested factweighing it is ‘overwhelmingly probable’ that Jesus was God incarnate (Swinburne 2003). Moreover, is it even possible to deﬁne concepts like religious ‘fundamentalism’ and religious ‘exclusivism’ cross-culturally—that is in terms of a shared manner or form of belief despite divergent belief content—without reference to an assumed HL attitude towards disagreement and towards outsiders to the home religion? MS: The Moderate Skeptical view While Feldman allows no principled distinction between strictly empirical debates and those over comprehensive doctrines, he does to his credit take note of his TSV students’ objections that the ‘ﬁt’ of the HL model will be harder to defend in cases in which people on both sides of a debate have been reﬂective and have openly discussed (shared) their reasoning and experiences. Perhaps, then, ‘In these moral, political, scientiﬁc, and religious disputes, it is implausible to think that one side is simply unreasonable’ (2007: 211). Feldman is not here conceding that he has personally become convinced to move off of his HL atheism, but in the event that he were to be convinced of it, the move is not ﬁrst towards a ‘friendly’ atheism though epistemic humility. Rather, he now claims, the Moderate Skeptical view becomes the singularly intellectually responsible and hence obligatory response to adopt. This second of the two evidentialist-approved ways of responding to peer disagreement is the one a person is intellectually obliged to conform to whenever there is recognition or even serious suspicion of vagueness or ambiguity of total public evidence on the claim in question. Since RUT tells us it is epistemically wrong to believe a proposition when one’s total evidence fails to decisively support that proposition, ‘the right thing for both of us to do is to suspend judgment’ (2007: 212). Because ‘the right thing’ here is presented as singular and obligatory, MS runs directly in conﬂict with literatures one and two in the philosophy of religion, noted above (while HL appears to conﬂict with all three). Suspending judgments we are given to believe is the only way to be epistemically ‘successful’ in the moment, where evidential ambiguity presents itself. So if ‘there are really only two’ options, HL and MS, because reasonable belief is just evidentially justiﬁed belief and there always either is or isn’t ‘sufﬁcient’ shared evidence for evidentially justiﬁed acceptance of whatever proposition is in question, then surely we have to answer ‘No’ to Q2, and then to Q1 as well. It may be that MS and not HL holds in the case of comprehensive doctrines, but TSV, the tolerant and supportive view, is nevertheless now exposed as a confused attempt to place sentiment and political correctness above plain reason. I want to suggest that we ‘return to reason’ not when we deny this logic, but when we actually stand it on its head: Rather than accepting the forced option he presents to his students between the HL and MS responses to deliberative disagreement, we should

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

83

straightforwardly re-describe this ‘forced option’ as thehorns of a dilemma that Feldman and other cognitive evidentialists impale themselves upon when they try to apply their epistemic principles to the real world of comprehensive philosophical, moral and religious doctrines. The Feldmanian over-extension of the RUT thesis places the evidentialist (and would if we allow, place all of us) in just those states of affairs often found precursory to war, when ‘Things fall apart; the centre cannot hold’; his ethic of belief evokes such circumstances where, as the poet W. B. Yeats bemusedly put it, ‘The best lack all conviction [MS], while the worst are full of passionate intensity’ [HL]. This Rational Uniqueness Dilemma, as we can term it, depicts the defenders of RUT and its close cousins as displaying a self-defeating intellectualism, one in which they ﬁnd themselves needing to embrace one or the other of two quite unwelcome consequences: (a) submitting to what seems the intellectual thanatos of applying the Hard Line’s logic of exclusivism across the board to disagreements over religious, philosophical, political, and moral ‘comprehensive’ doctrines (to the destabilization of Lockean reciprocal amity and ‘friendly’ theism/atheism) or (b) submitting themselves to the ‘spinelessness’ of strict suspension in all cases where evidential ambiguity or conceptual vagueness is apparent (as well as disappointing the Rawlsian expectation that a kind of conceptual pluralism is the natural outcome of an advancing liberal society). Consider more closely what it means to grab the ﬁrst horn. The unwelcome consequence here is essentially the same that McKim articulates when he writes that, ‘Advocates of large-scale systems of beliefs that include discrediting mechanisms are not in a position to appreciate the appeal to systems of belief they think to be discredited. They are not likely to be able to give them a fair or sympathetic hearing’ (2001: 152). The HL’s logic of exclusivism is one that liberalism’s discontents known well, one that allows easy discrediting of non-true-believers. Gutmann and Thompson argue in Why Deliberative Democracy that ‘Deliberation cannot make incompatible values compatible, but it can help participants recognize the moral merit in the opponent’s claims when those claims have merit’ (2004: 11). By contrast, prescribing HL as the proper response for disputants to take in disagreement among comprehensive doctrines just appears to crown as the height of reasonableness such judgments of the culpable failures of others that Rawls described as typically the result of ‘prejudice and bias, self and group interest, blindness and willfulness’ (Rawls 1993: 249). So yes, of course, we can follow the evidentialist’s and religious fundamentalist’s shared logic of exclusivism and just ‘bite the bullet’ of this ﬁrst horn; the question is whether we can do so without shooting ourselves in the head. But perhaps I am presumptuous in asserting this ‘we’. For the discontents of the Evidentialist Accommodation and of the Jeffersonian Compromise we are now seeing, are many and varied, and here make strange bedfellows. Part of what they have in common is that many seem not to really want the Jeffersonian middle to hold, except perhaps (as one would hope) in the thinnest, civic way, or until conditions ripen for their alternative view of the ‘new order’, whether religious, secularist, or secular humanist. To be a discontent as I intend it, however, is to be repelled by the whole state of our cultural debate over reason and faith dominated by just such

84

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

extremes of religious and skeptical rationalism. It is to be dismayed with the aggressive secularizers like Harris as well as with their Christian evidentialist mirror image. But grabbing the second horn, encompassing not increased epistemic humility about our belief but a change from full belief to suspension, is arguably even less appealing if it is indeed psychologically plausible to begin with (extricatable from strong voluntarist assumptions). ‘Do you have any convictions on controversial political, philosophical, or scientiﬁc matters? The equal weight view seems to say: kiss them goodbye’ (Elga 2007: 484). There are different forms of skepticism, and this one appears risk-averse in such a radical way as to be open to the charge of being unlivable. Instead of describing this ‘equal weight’ model as ‘moderate’, as Feldman does, its critics seem to me more apt in their description of it as demanding ‘dispiriting spinelessness’ (Sosa 2008) and/or ‘lack of selftrust’. ‘It is implausible that rationality requires such spinelessness’ (Elga 2007: 484). I think this equal weight view in the epistemology of disagreement results from disregard for the role of doubt in the history of social philosophy, especially in the early-modern period of Erasmus, Montaigne, and Locke where the theory of toleration was re-born. The nouveau Pyrrhonisme of Montaigne and Erasmus, Toulmin reminds us in Cosmopolis, ‘was no “negative dogmatism” which systematically refuses to accept everyone’s right to opinions arrived at by honest reﬂection on ﬁrst-hand experience’ (1992: 50). Doubt had to be a multi-edged sword for those who lived through the roughest centuries of intra-Christian warfare, serving the pragmatic aims of moderation and mediation against extremes on all sides. Feldman’s students or other supporters of TSV, I conclude, have plenty of resources to defend reasonable disagreement. In making our argument we have not impugned the importance of intellectual responsibility with respect to evidence. We have however applied our earlier critique of the internalist evidentialism’s claim to provide an adequate account of intellectual responsibility. If his students aren’t convinced of internalist evidentialism as an account of epistemic responsibility, they won’t accept his account of epistemic obligation and the forced option he presents it as issuing in. They won’t equate reasonable belief with synchronically rational beliefs, the diachronic activities of the agent set aside or bracketed. While they needn’t assume any kind of relativism or deny genuine disagreement, they will reject all attempts to apply RUT unqualiﬁedly to debates like those between theists, atheists, and agnostics. Accordingly they will neither conceive of the Hard Line view over such issues as the sin qua non of rationality, nor accept the claim that universal suspension of belief is the only reasonable response to the acknowledged ambiguity of total public evidence.

4.4 Conclusion: towards a responsibilist accommodation I have argued that the best-known version of evidentialism as an ethics of belief and position in the epistemology of disagreement/diversity has both weak roots and sour

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

85

fruits. The pragmatic critique of Section 4.3 was intended only to supplement the epistemological objections of Section 4.2. The dilemma just presented to Feldman is meant to illustrate just how far his evidentialist philosophy, purportedly initially motivated by important insights of John Locke, has come back around to problematize two key themes of Locke’s social philosophy: his conceptual pluralism and his defense of toleration (reciprocal amity) as a value and a virtue. The lack of any genuine positive epistemic invitations or entitlements to believe that are not merely obligations to believe, disbelieve, or suspend, has been another source of my own discontent with the evidentialist accommodation: its lack of recognition of what I think of as our need of and intellectual right to various ‘doxastic ventures’, philosophic, religious, or political; its evident lack of that ‘spirit of inner tolerance’ for another’s mental freedom, without which William James warned that our outer (social, political) tolerance is likely to be or to become unstable. Should the evidentialist accommodation between private persons and the collective community of inquiry also appear overly restrictive or ‘Victorian’ on this score, the resources of virtue theory and the prospects of an alternative responsibilist accommodation remain available to explore. On this alternative accommodation as I propose it, it is diachronic rather than synchronic evaluation of agents that most directly informs a sound and civic ethics of belief. Diachronic rationality’s contribution to epistemic value was the upshot of our epistemological critique, but has quite direct implications for our understanding of the ethics of belief as well. It implies that there may well be invitations of sorts—simply put, things an intellectually virtuous believer might believe, that are things not all virtuous believers in this self-same evidential situation necessarily would believe. But any right to be choosers of our risk is not on the virtue-theoretic view precritically given to us. We must earn our intellectual right to vary from the evidentialist norm of belief, though it comes far more readily in respect to weltanschaulich questions and beliefs as they arise in ﬁelds like politics, philosophy, and religion. We earn this right to maintain old or initiate new doxastic ventures—what Mill and James called our various ‘experiments in living’—through intellectually conscientious efforts at inquiry, and through an habituated sensitivity to potential defeaters, defeaters that enliven an agent to the kind or level of conﬁdence that they have in their belief, and to diachronic responsibilities to extend, revisit, or update the reﬂectively good reasons they could muster to support that conﬁdence.

References Axtell, G. 2008. ‘Expanding Epistemology: A Responsibilist Approach,’ Philosophical Papers, 37, 1: 51–88. ——. 2010. ‘Agency Ascriptions in Ethics and Epistemology,’ Metaphilosophy, 41, 1–2: 73–94. ——. and J. Adam Carter. 2008. ‘Just the Right Thickness: A Defense of Second-Wave Virtue Epistemologies,’ Philosophical Papers, 37, 3: 413–34.

86

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

——. and P. Olson. 2009. ‘Three Independent Factors in Epistemology,’ Contemporary Pragmatism, 6, 2: 89–109. Basinger, D. 2002. Religious Diversity: A Philosophical Assessment. Farnham, Surrey: Ashgate. Bernstein, R. 2003. ‘Pragmatism’s Common Faith,’ in S. Rosenbaum (ed.) Pragmatism and Religion. La Salle: Illinois University Press, 129–41. Bishop, J. 2007. Believing by Faith: An Essay in the Epistemology and Ethics of Religious Belief. Oxford: Oxford University Press. Chappell, T. 2009. ‘On Not Proportioning your Belief to the Evidence,’ (typescript). Clifford, W. K. 1999 [1877]. ‘The Ethics of Belief,’ reprinted in The Ethics of Belief and Other Essays. Amherst, NY: Prometheus Books. Conee, E. 2010. ‘Rational Disagreement Defended,’ in R. Feldman and T. Warﬁeld (eds), Disagreement. Oxford: Oxford University Press, 69–90. —— and Richard Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Clarendon. Elga, A. 2007. ‘Reﬂection and Disagreement,’ Noûs 41, 3: 478–502. Feldman, R. 1988. ‘Epistemic Obligations,’ Philosophical Perspectives 2: 235–56. ——. 2002. ‘Epistemological Duties,’ in Paul Moser (ed.), Oxford Handbook of Epistemology. Oxford: Oxford University Press, 362–84. ——. 2006. ‘Epistemological Puzzles about Disagreement,’ in S. Heatherington (ed.) Epistemological Futures. Oxford: Oxford University Press, 216–36. ——. 2007. ‘Reasonable Religious Disagreements,’ in L. Antony (ed.), Philosophers without Gods. Oxford: Oxford University Press, 194–214. Greco, J. 2005. ‘Justiﬁcation is not Internal,’ in M. Steup (ed.), Contemporary Debates in Epistemology. Oxford: Blackwell, 257–70. ——. 2008. ‘Friendly Theism,’ in J. Kraft and D. Basinger (eds), Religious Tolerance through Humility. Burlington, VT: Ashgate, 51–60. Gutmann, A. and Thompson, D. 2004. Why Deliberative Democracy? Princeton: Princeton University Press. Harris, S. 2006. Letter to a Christian Nation. New York: Alfred A. Knopf. Hazlett, A. 2009. ‘Faith and Liberalism.’ [typescript] Himma, K. 2002. ‘Finding a High Road: The Moral Case for Salviﬁc Pluralism,’ International Journal for Philosophy of Religion, 52: 1–33. James, J. 1979a. [1908] The Will to Believe and other Essays in Popular Philosophy. Cambridge: Harvard University Press. ——. 1979b [1911]. The Problems of Philosophy. Cambridge: Harvard University Press. Kraft, J. 2006. ‘Religious Tolerance through Religious Diversity and Epistemic Humility,’ Sophia 45, 2, 101–16. Kraft, J. and Basinger, D. 2008. Religious Tolerance through Humility. Farnham, UK: Ashgate. Locke, J. 1997 [1686]. Essay Concerning Human Understanding. London: Penguin Classics. McKim, R. 2001. Religious Ambiguity and Religious Diversity. Oxford: Oxford University Press. Pappas, G. 1994. ‘William James Virtuous Believer’ Transactions of the Charles Peirce Society, 30, 1: 77–110. Rawls, J. 1993. Political Liberalism. New York: Columbia University Press. Rawls J. 1996. Political Liberalism. New York: Columbia University Press. Rorty, R. 1991. Objectivity, Relativism, and Truth: Philosophical Papers, Volume 1. Cambridge: Cambridge University Press.

F RO M I N T E R N A L I S T E V I D E N T I A L I S M T O V I RT U E R E S P O N S I B I L I S M

87

Rousseau, Jean Jacques. 1987 [1762]. On the Social Contract. Indianapolis, IN: Hackett Publishers. Sosa, E. 2008. ‘The Epistemology of Disagreement,’ in A. Haddock, A. Millar, and D. Pritchard (eds), Social Epistemology. Oxford: Oxford University Press, 278–97. Swinburne, R. 2003. The Resurrection of God Incarnate. Oxford: Oxford University Press. Toulmin, S. 1992. Cosmopolis: The Hidden Agenda of Modernity. Chicago: University of Chicago Press. ——. 2001. Return to Reason. Cambridge, MA: Harvard University Press. Turri, J. 2010. ‘On the Relationship Between Propositional and Doxastic Justiﬁcation,’ Philosophy and Phenomenological Research, 80, 2: 312–26. Van Inwagen, P. 2010. ‘We’re Right. They’re Wrong,’ in R. Feldman and T. Warﬁeld (eds), Disagreement. Oxford: Oxford University Press, 10–28. Williams, M. 2008. ‘Responsibility and Reliability,’ Philosophical Papers, 37, 1: 1–26. Wood, J. 2006. ‘Intellectual Virtues and the Prospects of a Christian Epistemology,’ in M. Cherry (ed.), The Death of Metaphysics; the Death of Culture. Dordrecht: Springer, 41–69. Zagzebski, L. 2007. Philosophy of Religion. Oxford: Blackwell.

5 Evidentialism, Vice, and Virtue Jason Baehr

According to evidentialism, epistemic justiﬁcation is a function of the evidence one has in support of one’s beliefs. More precisely, evidentialists subscribe to (one variation or another of ) the following general principle: E. A person S is justiﬁed in believing a proposition p at time t if and only if S’s evidence at t supports p.1 The notion of ‘evidence’ employed here is rather broad: it includes (or potentially includes) supporting beliefs, experiential and introspective states, and rational insights.2 The relevant notion of ‘support’ is more difﬁcult to pinpoint. However, according to one intuitive and fairly standard account, a body of evidence e supports a proposition p just in case p is more probable than not in light of e.3 As this suggests, the kind of support in question—and epistemic justiﬁcation itself—is a matter of degree.4 Evidentialism has been subjected to a barrage of criticisms in recent years, many of which have been aimed at showing that its central condition is not necessary for justiﬁcation.5 I shall not address any of these objections here. Rather, my concern is with the sufﬁciency condition in (E). I begin by discussing several cases in which a belief apparently satisﬁes this condition but fails to instantiate one or more varieties of epistemic justiﬁcation presumably of interest to evidentialists. I go on to argue, however, that this does not warrant abandoning the thrust of evidentialism. Instead it calls for supplementing evidentialism with a certain constraint, according to which, 1 Though not all of them employ the term ‘evidentialism,’ defenders of the basic position indicated by (E) include Laurence BonJour (1985), Roderick Chisholm (1989), Earl Conee and Richard Feldman (2004), and Richard Swinburne (2001). 2 See chapters 4 and 9 of Conee and Feldman (2004) for relevant discussions. 3 See, e.g., Swinburne (2001: ch. 3). 4 While epistemic justiﬁcation arguably is an essential feature of knowledge, (E) should not be read as capturing that which, in addition to true belief, is necessary for knowledge. For knowledge may require a degree of evidential support not required by (E). Furthermore, the satisfaction of (E) does not rule out Gettier factors. Thus one can have a true belief that satisﬁes (E) but that falls short of knowledge. 5 See Plantinga (1993) for an assortment of these criticisms.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

89

under certain conditions, justiﬁcation requires intellectually virtuous agency. The discussion therefore brings to light a notable conceptual connection between one venerable and longstanding way of thinking about epistemic justiﬁcation and the more recent enterprise of ‘virtue epistemology’. Before getting started, an important methodological point is in order. There are at least two notably different ways of attempting to argue that the satisfaction of the evidentialist’s condition is insufﬁcient for justiﬁcation. The ﬁrst is predicated on the idea that there exists a single determinate and univocal concept of epistemic justiﬁcation and that disputes about the nature of justiﬁcation are disputes about this concept. Here, the strategy is to show that a belief can satisfy the evidentialist’s condition but fail to instantiate the concept in question. There are, however, serious problems with the idea that there is a single determinate and univocal concept of justiﬁcation. Indeed, as William Alston has argued, much of the debate in epistemology in recent years suggests that there are several such concepts or several ‘epistemic desiderata’.6 But if concepts of epistemic justiﬁcation are manifold, what would it mean to argue that the satisfaction of the evidentialist’s condition is not sufﬁcient for justiﬁcation? We can begin to see an answer to this question by noting that while there may be a rather wide and diverse variety of epistemic desiderata, it is plausible to think that evidentialists are interested in a certain limited subset of them—and perhaps just in a single desideratum. This is, at any rate, what I shall be assuming here. Accordingly, my concern is limited to what I shall call ‘evidentialist-relevant’ or ‘e-relevant’ varieties of justiﬁcation, which again are varieties the nature of which at least some evidentialist accounts of justiﬁcation presumably are attempting to capture.7 My immediate aim, then, is to show that a belief can satisfy the sufﬁciency condition in (E) while failing to instantiate one or more e-relevant varieties of justiﬁcation.

5.1 Problem cases Cases of defective inquiry My argument for the claim just noted centers around two kinds of cases. In cases of the ﬁrst sort, a belief satisﬁes (E) but only because the believer either fails to inquire at all or else inquires in a manner that is clearly defective.

6

Alston (1993). According to Alston, different camps of epistemologists (e.g. internalists and externalists) appear to be concerned with fundamentally different concepts of justiﬁcation. Alston compares the attempt to provide an analysis of justiﬁcation understood as a single, univocal concept to a ‘quixotic tilting at windmills’ (1993: 542). 7 I will not attempt to spell out the notion of e-relevant justiﬁcation in any detail here. However, given the close association between evidentialism and internalism (an association discussed in more detail below), e-relevant varieties of justiﬁcation are likely to be internalist in nature; and paradigmatically externalist varieties of justiﬁcation (e.g. crude reliabilism) are likely not to count as e-relevant.

90

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

Case 1. George epitomizes the vices of intellectual laziness, apathy, and obliviousness. He goes about his daily routine focusing only on the most immediate and practical of concerns: feeding himself, getting to work on time, doing his job in a minimally satisfactory way, paying the bills, etc. He lacks any natural curiosity and is almost entirely tuned out to the news of the day. Unsurprisingly, George has many beliefs he should not and fails to believe many things he should. In the former category is George’s belief that exposure to secondhand smoke poses no signiﬁcant health risks. Given his extremely narrow and practical focus, George is oblivious to all of the well-publicized research indicating the hazards of secondhand smoke. In fact, George actually has positive evidence in support of his belief. He recalls having learned from a reliable source some years ago that a considerable amount of research had been conducted concerning the effects of exposure to secondhand smoke and that this research had failed to establish any correlation between such exposure and any serious health problems. And as far as George knows, the research on this topic has not changed. Nor, we may suppose, does he have any reason to think that it might have changed.8 Case 2. Gerry holds the same belief as George and on roughly the same grounds. Therefore he too has positive evidence for thinking that secondhand smoke is benign. Unlike George, however, Gerry is not oblivious to the news of the day; in fact he is a reasonably inquisitive person and enjoys checking things for himself. The problem is that his inquiries tend to be insufﬁciently demanding and discriminating. They are prone to gullibility, carelessness, and hasty generalization.9 Upon hearing news reports afﬁrming the danger of secondhand smoke, Gerry decides to look further into the matter. The ﬁrst item he comes across happens to be a report published by an organization with major ﬁnancial ties to the tobacco industry. The report is aimed, not at a fair and balanced treatment of the issue, but rather at exposing any apparent weakness or grounds for doubt in the recently publicized research. To any reasonably intellectually rigorous or discriminating inquirer, the dubious nature of the report would be evident. But to Gerry it is not. And the result is that Gerry’s total evidence (which again includes his initial evidence for thinking that secondhand smoke is not a health threat) supports his belief.10

In each of the above cases, a person’s belief is well-supported by his evidence. As a result, the beliefs seem clearly to satisfy (E). The problem is that the reason these beliefs are well-supported traces back to certain defects on the part of the individuals who hold them. George has good evidence for his belief only on account of his intellectual ‘tunnel vision’. Gerry’s belief remains well-supported because of his undemanding and undiscriminating method of inquiry. In light of these shortcomings, the beliefs of

8 Whether anything like the foregoing is in fact true of the early stages of such research I am unsure. But surely things might have unfolded in this way (i.e. the presently well-known hazards of exposure to secondhand smoke might have taken some time to substantiate). The basic idea, which could be developed in connection with any number of other examples, is that George’s present evidence is sorely outdated and very different from what it would be if George were minimally inquiring or curious. (In presenting a version of this paper in Europe, for instance, I found the example of outmoded and irresponsible beliefs about global warming to be more convincing!) 9 We may stipulate that this is something of which Gerry is unaware. 10 For a similar kind of case, see Kornblith (1983). John Greco (2005) also raises a related worry for evidentialism.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

91

George and Gerry seem clearly to be unjustiﬁed. Finally, it is plausible to think that the variety of justiﬁcation in question is e-relevant. In response to cases of this general sort, Richard Feldman (2005; Conee and Feldman 2004) offers a defense of the sufﬁciency of (E) the upshot of which is that in fact there is no e-relevant variety of justiﬁcation according to which the beliefs of George and Gerry are unjustiﬁed. According to Feldman, an evidentialist theory of justiﬁcation is concerned strictly with the relation between a person’s evidence and her belief. It makes no difference whether this evidence is the result of (say) uncritical or hasty inquiry or whether the person has the evidence only on account of failing to inquire at all. While these factors may bear on the evaluation of the believer’s intellectual character or doxastic methods, Feldman claims, they have no bearing on the justiﬁcatory status of beliefs that result from such character or methods.11 I see no reason to deny that there may be some epistemic value in believing in accordance with one’s evidence per se—without any regard for whether this evidence is a result of defective inquiry. Beliefs of this sort might be said to involve a kind of logical coherence or consistency, which indeed is often regarded as an epistemic desideratum.12 Moreover, as Feldman reasonably queries, what other doxastic attitude could plausibly be required of agents like George or Gerry?13 Clearly it would be problematic, given their evidence, to suggest that either of them ought to believe that secondhand smoke is harmful—or even, for that matter, to suspend judgment on the matter (for again, they have reason to believe that it is not harmful). Thus there does appear to be a sense of ‘justiﬁcation’ according to which the beliefs of George and Gerry are justiﬁed. And there is little reason not to regard this as an e-relevant species of justiﬁcation. But this hardly puts (E) in the clear. For the fact that the beliefs of George and Gerry instantiate some e-relevant variety of justiﬁcation guarantees neither: (a) that there is not an additional variety of e-relevant justiﬁcation that these beliefs fail to instantiate; nor (b) that the variety of justiﬁcation they do instantiate is particularly worthy or signiﬁcant. Both of these possibilities merit further consideration. First consider (a). Despite whatever justiﬁcation the beliefs of George and Gerry may enjoy, there is indeed an additional intuitive and e-relevant sense in which these beliefs are unjustiﬁed. For while these beliefs are well-supported by the agents’ evidence, the evidence clearly is not what it should be. George, for instance, ought to have taken notice of some of the widely publicized research establishing a link between exposure to secondhand smoke and various diseases. Similarly, it ought to have occurred to Gerry to undertake a broader inquiry and to pay closer attention to the source of the relevant report. He too should be aware of the evidence against his belief. Since the beliefs of George and Gerry are based on evidence that ought to be other than it is,

11

See, e.g., Feldman (2005: 281); cf. Conee and Feldman (2004: 90, 101). See, e.g., Swinburne’s discussion of ‘synchronic justiﬁcation’ in 2001: ch. 1. See also Feldman (2005: 277–78). 13 2005: 282. See also Conee and Feldman (2004: ch. 4). 12

92

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

there is a straightforward sense in which they ought not believe as they do; that is, there is a straightforward sense in which their beliefs are unjustiﬁed. And again, the kind of justiﬁcation in question seems clearly to be e-relevant.14 One way of drawing further attention to this variety of justiﬁcation is to consider how we might evaluate certain actions of George or Gerry that are based on the beliefs in question. Suppose, for instance, that on the basis of his belief that secondhand smoke is benign, George proceeds to smoke on a regular basis in the company of his children. Surely we would condemn George’s behavior, despite the fact that from his own perspective he is doing no harm. And the reason is that he ought not to have had this perspective in the ﬁrst place; he ought not believe as he does. For again, evidence against this belief is abundant and readily available to him. He ought to be aware of this evidence and to believe in accordance with it. This strongly suggests that George’s present belief is, in a genuine and e-relevant sense, unjustiﬁed.15 Now consider (b). While the beliefs of George and Gerry may instantiate a certain epistemic desideratum, the signiﬁcance or worth of this desideratum is questionable. To see why, note that the evidence possessed by George and Gerry is in a certain substantial way defective or contaminated—and for reasons that trace back to these individuals’ own epistemic wrongdoing. Again, had either George or Gerry been even minimally attentive and discriminating in his thinking about the relevant subject matter, his perspective concerning the truth of the claim in question would have been 14 Trent Dougherty has suggested to me that the concept of justiﬁcation I am getting at here is strictly a deontological one, that evidentialism is not really a thesis about deontological justiﬁcation (hence that deontological justiﬁcation is not e-relevant), and that consequently my cases fail to pose a problem for their intended target. My response is as follows. First, while I have employed the language of ‘ought’ in characterizing what goes wrong in the cases in question, I do not think that the sort of justiﬁcation George and Gerry lack is merely the standard deontological variety. This will become clearer when I discuss point (b) momentarily. Secondly, if evidentialism really should be understood as limiting the scope of epistemic evaluation strictly to the present ‘ﬁt’ between a person’s belief and her evidence, with no regard whatsoever for how the person arrived at that evidence or why she possesses the evidence she does, then in fact it looks as though the beliefs of George and Gerry are justiﬁed from the standpoint of a deontological conception of justiﬁcation. For again, if we completely disregard the quality of their inquiry in connection with these beliefs, then given their evidence, they seem perfectly entitled to believe as they do. Third, I see no reason to think of ‘evidentialism’ as necessarily excluding deontological accounts of justiﬁcation. For instance, suppose someone maintains that a belief is justiﬁed if and only if it is supported by good evidence, but that this is because our sole intellectual duty is to believe in accordance with the evidence. Surely this could be regarded as a version of evidentialism (or as picking out an ‘e-relevant’ variety of justiﬁcation). Indeed, just such a possibility is at least suggested (if not endorsed) in Steup 1995 and 2001. 15 John Greco offers a similar assessment of some similar cases. He says that where ‘two persons arrive at the same internal [read: evidential] perspective, but . . . one does so in a way that is epistemically responsible, whereas the other does so on the basis of carelessness, thick-headedness, and stupidity,’ ‘[t]he two persons will not be alike in epistemic justiﬁcation’ (2005: 262). His conclusion is that ‘etiology matters’ when assessing the justiﬁcatory status of a belief (ibid.). While Greco’s immediate target here is internalism, his point applies equally to the formulation of evidentialism we are concerned with. It is also worth noting that nothing about this or the former assessment of these cases presupposes an objectionable doxastic voluntarism. While George and Gerry may lack direct control over their beliefs, presumably they have a sufﬁcient amount of indirect control. For presumably they have control over the cognitive behavior (e.g. the inquiry or lack thereof) that produces these beliefs. Thus there is no problem with saying that the beliefs are ‘unjustiﬁed’ or that George and Gerry ‘ought not’ to have these beliefs.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

93

very different and much more accurate. Things being what they are, however, why think that George or Gerry does particularly well from an epistemic standpoint to believe on the basis of his evidence? What is especially epistemically good or worthy about believing in accordance with a defective or contaminated evidence base, particularly when the defects in question are attributable to one’s own cognitive failure? It would seem not much. My suggestion is not that George or Gerry ought to believe, against his evidence, that environmental smoke is harmful. It is rather that the alternative has very little to recommend it.16 Cases like that of George and Gerry are analogous to what moral philosophers sometimes refer to as ‘tragic dilemmas’, which are situations in which a person is forced to choose ‘the lesser of two evils’ between ‘sin and sin’.17 Consider the case of Bertie, who, having squandered the semester partying with his friends and playing video games, is presently faced with the dilemma of cheating on one of his ﬁnal exams (his only hope for passing a course that he needs to graduate on time) or devastating his parents (who have sacriﬁced a great deal to pay for Bertie’s education and have planned a major family reunion in honor of Bertie’s graduation). On at least one plausible analysis of the situation (assuming, say, that these really are Bertie’s only options and that his family would be sufﬁciently devastated as a result of his failure to graduate), Bertie ought to cheat on the exam. Indeed we might say that, given the circumstances, this is the only ‘real’ or justiﬁable option.18 At the same time, however, this is not to say anything very positive about Bertie’s action. While it may be the ‘lesser of two evils’, it is an ‘evil’ nonetheless. Similarly, while George and Gerry in some sense do well to believe in accordance with the evidence they have (e.g. in doing so they avoid believing or withholding belief on no basis whatsoever), this does not entail anything very positive about them or their beliefs, for again, each believes in accordance with a defective evidence base, the very defects of which are a result of his own cognitive wrongdoing. The point is that while the beliefs of George and Gerry may instantiate a certain e-relevant concept of justiﬁcation, this variety of justiﬁcation is not a very signiﬁcant epistemic good. Their beliefs are, as it were, the lesser of two epistemic evils in the situation. We may conclude that it is possible for a belief to satisfy (E) while failing to instantiate any signiﬁcant e-relevant concept of justiﬁcation.19 16

This characterization of the problem with the relevant beliefs—which has largely been in terms of epistemic signiﬁcance or worth or value—underscores the point that the kind of justiﬁcation I am concerned with (the kind that George and Gerry lack) is not strictly a deontological one. That is, the point is not strictly that the beliefs in question involve a violation of their possessors’ epistemic duties, but also, and indeed primarily, that the inherent epistemic signiﬁcance or worth of these beliefs is minimal. 17 See, e.g., Hursthouse (1998) and Geach (1977). 18 If this seems implausible, I invite the reader to think of his or her own case in which (roughly) the cost of not performing some action x that would typically be morally wrong or unjustiﬁed is so great that performing x is in fact morally acceptable (even if not good, praiseworthy, or the like). 19 Greco (2005) makes a similar observation. The upshot of his discussion is that while the beliefs of George and Gerry might be ‘justiﬁed’ in some very weak sense, the relevant conception of justiﬁcation is objectionably abstract, uninteresting, and unimportant. In reference to this conception, he remarks that ‘ “time-slice” evaluations that abstract away from the formation of beliefs, their relation to the world, and the

94

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

Cases of defective ‘doxastic handling’ of evidence In the cases just discussed, the epistemic status of a belief is undermined on account of some prior mistake or defect on the part of the believer, and speciﬁcally, on account of whether or how the believer inquires at some point in the past relative to the belief in question. In a second set of cases indicating the insufﬁciency of (E), the epistemic status of a belief is affected by an occurrent mistake or defect on the part of the agent. Speciﬁcally, it is affected by the agent’s ‘doxastic handling’ of information that threatens to defeat or undermine her justiﬁcation, that is, by the way in which she treats or regards this information at the time of belief. In the ﬁrst case, the agent ignores or suppresses the potential defeater; in the second case she distorts or misrepresents it. Case 3. Like George and Gerry above, Daphne believes that exposure to secondhand smoke poses no serious health risks; she also has some positive evidence in support of this belief. However, she is neither intellectually lazy nor undiscriminating. Upon hearing about the relevant research, she does some looking into the matter and nearly all the information she comes across indicates that in fact environmental smoke is hazardous. Daphne’s problem is that she is a hypochondriac raised by two chain-smoking parents. Owing to extreme anxiety about her health, she cannot accept any of the relevant evidence; indeed, she quickly and conveniently (though genuinely) forgets about or suppresses it. The result is that, as far as she can tell, her evidence continues to support her belief.20 Case 4. Doris also believes with some positive evidence that secondhand smoke is benign. Upon hearing news reports to the contrary, she too engages in reasonably careful and discriminating inquiry on the matter and in doing so encounters a host of data that threaten to refute her belief. Like Daphne, Doris is unable to accept this data. But in Doris’s case, this is due to her own extremely strong attachment to smoking. Her habit of smoking wherever and whenever she wants represents one of few pleasures in her otherwise lonely and unhappy existence. Unlike Doris, Daphne’s cognitive constitution is such that she cannot simply ‘forget’ or suppress the relevant evidence. Instead she distorts or misrepresents certain critical aspects of it. The result is that from her standpoint, the case for thinking that environmental smoke is hazardous is weak and her belief remains well-supported.21

There can be little doubt that there is an e-relevant sense in which the beliefs of Daphne and Doris are unjustiﬁed.22 Daphne is suppressing evidence of which she has recently been made aware that casts major doubt on her belief. Doris, though not character of believers will not be very important.’ And he offers the following comparison with moral evaluation: ‘Neither do we care about whether some action A is right relative to S’s own moral norms, in abstraction from questions about how S did A, or why S did A, or whether S’s norms are themselves any good’ (2005: 267). 20 We may stipulate that she has no recollection of having encountered the counterevidence or having suppressed it. Thus from her standpoint, the totality of her evidence indeed supports her belief. 21 Again we can stipulate that Doris is unaware of having distorted or misrepresented the relevant data and thus that as far as she can tell, her belief is supported by her evidence. 22 Indeed, inasmuch as these cases can be shown to satisfy (E), they provide an even more striking illustration of the point, defended above, that a belief can satisfy (E) while failing to instantiate any epistemically signiﬁcant or worthy concept of justiﬁcation.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

95

exactly suppressing or ignoring such evidence, is distorting or misrepresenting key elements of it. Nevertheless, the beliefs of Daphne and Doris are well-supported from their respective standpoints. I take it that evidentialists are unlikely to deny that the beliefs of Daphne and Doris are unjustiﬁed. They may, however, attempt to argue that these beliefs fail to satisfy (E) and hence do not present a problem for their position. Speciﬁcally, an evidentialist might argue that what matters for justiﬁcation is not whether it seems to a person that her belief is well-supported by her evidence, but rather whether her belief really is well-supported. This objection draws attention to a certain ambiguity in (E). It suggests a distinction between the following two more precise formulations of the central evidentialist principle: E2. S is justiﬁed in believing p at t if and only if S’s evidence at t appears to S to support p. E3. S is justiﬁed in believing p at t if and only if S’s evidence at t in fact supports p. The suggestion, then, is that an evidentialist might respond to the cases of Daphne and Doris by claiming that evidentialism should be understood along the lines of (E3) rather than (E2), and that once it is, these cases cease to pose a problem for evidentialism. Principle (E3) does apparently provide the evidentialist with a way around the Daphne and Doris cases.23 For both Daphne and Doris are in some sense ‘in possession of ’ evidence that in fact tells decisively against their beliefs.24 The problem, in Daphne’s case, is that she is ignoring or suppressing this evidence; Doris, on the other hand, is distorting or misrepresenting it. But given that the evidence in question is in their possession, and that it actually tells against their beliefs, (E3) rules (plausibly) that these beliefs are unjustiﬁed. But (E3) is problematic on other grounds. Consider cases of belief that involve what might be referred to as ‘concealed’ evidential relations, which are relations that obtain between a person’s evidence base and one of her beliefs but that are extremely difﬁcult to discern, even from the standpoint of an entirely normal and well-functioning cognitive agent. Along these lines, Swinburne (2001) discusses a case in which a detective is in possession of a great deal of evidence regarding a certain murder. It follows from the detective’s evidence ‘by a complicated line of argument’ that a particular suspect is guilty. However, ‘the detective is overwhelmed by the sheer quantity of evidence and so has not seen the relevance of this piece of evidence.’25 Assuming, then, that concealed evidential relations are possible, (E3) entails that if a person’s evidence in such a case seems clearly to support a certain claim p, but on the 23 Note, however, that (E3) is still susceptible to the George and Gerry cases discussed earlier; for their beliefs, while intuitively unjustiﬁed, are in fact well-supported by their evidence. 24 On the assumption that the relevant evidence is not actually in their possession, matters are even worse for (E3). For in that case, the beliefs in question are well-supported by the relevant evidence and so turn out (implausibly) to be justiﬁed. 25 P. 154. Conee and Feldman (2001: 73) discuss a similar sort of case; as does BonJour (1998: 128).

96

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

basis of a ‘concealed’ entailment relation in fact supports a different claim q, this person is justiﬁed in believing q. But this is extremely implausible, for the person in question is entirely unaware of the fact that her evidence actually supports q.26 To see how an evidentialist might try to amend (E3) in light of this problem, note that most evidentialists also embrace internalism about epistemic justiﬁcation. According to one standard version of internalism,27 a person is justiﬁed in believing a given claim only if he has ‘direct and unproblematic access’ to any factors that justify this belief. If supplemented with an internalist condition, (E3) becomes: E4. S is justiﬁed in believing p at t if and only if S’s evidence at t in fact supports p and S has direct and unproblematic access to this fact at t. Principle (E4) allows the evidentialist to deal with one problematic aspect of the case just noted, since it entails (plausibly) that the person in question is not justiﬁed in believing q (the proposition supported by the concealed entailment relation).28 But a serious problem remains. For not only is it implausible to think that this person is justiﬁed in believing q, it is extremely plausible to think that she is justiﬁed in believing p (the proposition supported by ‘all appearances’). But if (E4) is correct, this person is not justiﬁed in believing p, for her evidence in fact supports q.29 One way around this difﬁculty would be to narrow the scope of the internalist element of (E4) such that justiﬁcation is said to be a function of the actual relation between a person’s belief and those aspects of the person’s evidence to which she has direct and unproblematic access. This would yield the following: E5. S is justiﬁed in believing p at t if and only if the aspects of S’s evidence to which S has direct and unproblematic access at t in fact support p.30

26 Perhaps the belief is justiﬁed in a certain robustly externalist sense. But such justiﬁcation presumably is not e-relevant; and it is e-relevant justiﬁcation that we are concerned with here. 27 See, e.g., BonJour (1992). ‘Mentalism’ is an alternative version of internalism according to which justifying factors are necessarily ‘internal’ in the sense of being ‘internal to the person’s mental life’ or ‘in the person’s mind’ Conee and Feldman (2004: 55). However, as I will note momentarily, an appeal to a mentalist version of internalism would be of little help in dealing with the case in question. 28 By contrast, if (E3) were supplemented with a mentalist internalist condition, it would still entail that the person in question is justiﬁed in believing q, for the support relation between the person’s evidence and her belief that q presumably would be ‘internal’ in the relevant sense. Thus an appeal to a mentalist version of internalism is unhelpful in the present context. 29 As this point suggests, ‘actual support’ views like (E3) and (E4) face a special challenge when it comes to embracing fallibilism about epistemic justiﬁcation. Few would deny the possibility of justiﬁed false belief. But to the extent that such fallibility includes cases in which a person’s grounds clearly seem to (but in fact do not) support her belief, actual support views run into trouble. 30 ‘Aspects’ should be understood to include either items of evidence (e.g. experiences, other beliefs, etc.) or any evidential relations between items of evidence and the believed proposition. Thus there is no need to stipulate further that S must have access to the fact that the aspects of S’s evidence to which S has access in fact support p, for as it stands (E5) can handle the sort of case that led to the adoption of an access-clause in the ﬁrst place. That is, given (E5), the person just discussed is not justiﬁed in believing q, for q is not supported by the aspects of her evidence to which she has direct and unproblematic access.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

97

Principle (E5) can handle both aspects of the case under consideration. It rules (plausibly) that the subject is not justiﬁed in believing q because, while q is supported by the totality of this person’s evidence, it is not supported by that part of her evidence to which she has direct and unproblematic access (i.e. the ‘unconcealed’ part). It also rules (plausibly) that the person is justiﬁed in believing p, for the part of this person’s evidence to which she has direct and unproblematic access does in fact support p. But while (E5) is an improvement over (E4) in one respect, it is vulnerable in a way that (E4) is not. For unlike (E4), (E5) generates the wrong result in the very cases that led us to distinguish variations of (E) in the ﬁrst place: viz. the Daphne and Doris cases. The details of these cases can easily be reﬁned such that Daphne and Doris lack the kind of access to the relevant counterevidence that is required by (E5), with the implausible result that their beliefs satisfy (E5) and thus are justiﬁed. We might imagine, for instance, that Doris is so attached to smoking at will that if she were to curtail her habit in any way (which she would feel compelled to do if she were honest with herself about the relevant evidence), her psyche would begin to unravel. Thus it would take extreme measures (e.g. hypnosis or therapy) to get her to confront this evidence honestly. On this rendering of the case, Doris presumably lacks anything like ‘direct and unproblematic’ access to that part of her evidence that she is distorting or misrepresenting. Therefore, since the evidence to which Doris does have the required kind of access in fact supports her belief, (E5) rules (implausibly) that her belief is justiﬁed.31 There is in fact a notable irony in such cases that further tells against the plausibility of (E5). It consists in the fact that the more defective the individuals in question are, the more likely they are to be justiﬁed according to (E5). For instance, the more Daphne suppresses or ignores the relevant counterevidence, the more problematic and less direct her access to this evidence will become, and thus the more likely it will be that her belief is justiﬁed according to (E5). This is problematic, of course, because it is precisely this self-deception that intuitively undermines the justiﬁcation of her belief in the ﬁrst place. We began this section by considering how two additional cases pose a problem for the sufﬁciency of (E). This led to a distinction between (E2) and (E3). According to the former, justiﬁcation requires mere ‘apparent support’ between one’s belief and one’s evidence, while according to the latter, it requires ‘actual support’. Problems with (E3) led to a consideration of two related principles, (E4) and (E5), both of which we also found susceptible to serious objections. My concern in the remainder of the paper will be with (E2). I shall argue that (E2) can be modiﬁed in a way that preserves the thrust of evidentialism and avoids the problems that plague these other formulations.

31 Clearly a similar story could be told about Daphne. It might, for instance, take the same sort of extreme measures to get Daphne to ‘recall’ the evidence she is suppressing. If so, she too lacks the kind of access to this evidence that is required by (E5) and thus is justiﬁed in believing as she does.

98

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

5.2 Modifying evidentialism We have examined two sets of cases in which a belief satisﬁes (E2) but fails to instantiate any interesting variety of e-relevant justiﬁcation. Thus (E2) clearly stands in need of modiﬁcation. To see what form this modiﬁcation should take, it will be helpful to look again at the various cognitive defects manifested in the cases in question. Doing so will provide an indication of what further, positive requirement should be added to (E2). The defects in question include the following: intellectual laziness, inattentiveness, lack of intellectual discrimination, gullibility, carelessness, disregard for the truth, ignoring and distorting counterevidence, self-deception, and the like. One striking feature of this list is that it consists entirely of intellectual vices, that is, of bad or defective traits of intellectual character. It is on account of an exercise of such traits that the individuals in Cases 1–4 above lack justiﬁcation for their beliefs. One strategy for amending (E2), then, would be to make the antidote to intellectual vice a necessary condition for justiﬁcation.32 This antidote is of course intellectual virtue. Intellectual virtues include excellences of intellectual character like carefulness and thoroughness in inquiry, inquisitiveness, attentiveness, fair-mindedness, open-mindedness, intellectual honesty, and intellectual integrity. Thus it might be thought that (E2) should be supplemented with an additional condition that makes an exercise of intellectual virtue a necessary condition for justiﬁcation. This would yield something like the following principle: E6. S is justiﬁed in believing p at t if and only if S’s evidence at t appears to S to support p and S exercises virtues of intellectual character in the formation of this belief. There is, however, at least one immediate and formidable problem with (E6): namely, that justiﬁed or known beliefs sometimes arise from the brute or default functioning of a person’s basic cognitive equipment (not from an exercise of any intellectual character virtues).33 Suppose, for example, that while working late one night in my well-lit study, the electricity suddenly and unexpectedly shuts off, immediately causing the room to go dark. In response, I automatically and without thinking form a belief to the effect that the room has grown dark. Intuitively, my belief is justiﬁed. It is also justiﬁed according to (E2), since my belief is (and appears to me to be) well-supported by my visual experience. However, it is not justiﬁed according to (E6), for there is little reason to think that I have exercised any virtues of intellectual character in the formation of 32 Another strategy would be to incorporate a purely negative condition according to which justiﬁcation requires not exercising any vices of intellectual character. One problem with this condition is that it sheds no light on what, positively, is required for justiﬁcation beyond the satisfaction of (E5). Moreover, as I will get to below, it may be possible for a belief to be justiﬁed from an e-relevant standpoint even though the person who holds the belief displays intellectual vices at the time the belief is formed. 33 See my (2006). An additional worry might be that the introduction of the second clause would make the ﬁrst clause obsolete. That is, it might be said that an intellectually virtuous person (characteristically) would accept a proposition only if this proposition were well-supported by his grounds. I will not attempt to resolve this issue here.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

99

this belief. Again, this belief, together with my evidence for it, is the product of the brute or mechanistic functioning of my cognitive apparatus. Virtuous agency is not involved in any signiﬁcant way.34 We have found, then, that a belief can be unjustiﬁed on account of the believer’s exhibiting vicious agency, but that it is implausible to make virtuous agency a necessary condition for justiﬁcation. If so, how can (E2) be modiﬁed—beyond the addition of a purely negative and unilluminating amendment to the effect that justiﬁcation requires an absence of vicious agency—such that it precludes the relevant manifestations of intellectual vice? The apparent solution is to modify (E2) in such a way that it requires something like an exercise of intellectual virtue, but only in cases like the ones discussed earlier (not in cases of passive or brute justiﬁcation). This can be done by supplementing (E2) with a certain proviso or constraint. To get an idea of what exactly this constraint might look like, we must examine more closely the difference between Cases 1–4 above, on the one hand, and cases of brute or passive justiﬁcation, on the other. As already suggested, the principal difference between the two concerns the role of personal agency in the formation of the relevant belief. In the former set of cases, the agency of the believer is involved; in the latter kind of case, it is not. Recall, for example, the case of Gerry. His agency bears immediately on the content of his evidence: his evidence is what it is largely because he has inquired in an undiscriminating and careless way. He then forms his belief on the basis of this evidence. Personal agency is also involved in the formation of the beliefs of Daphne and Doris. Here it bears most immediately, not on the content of their evidence, but rather on how they handle or regard this evidence. Doris, for instance, distorts or misrepresents critical elements of her evidence. She then forms her belief on the basis of the resulting evidential perspective.35 Thus, in Cases 1–4, personal agency makes a salient contribution to what we might refer to as the ‘evidential situation’ of the subject in question, meaning that it largely determines either the content of the subject’s evidence or how the subject handles or 34 It does not follow that my agency is completely idle. But as Linda Zagzebski (1999) has shown, any plausible virtue-based account of justiﬁcation or knowledge must posit a rather strong connection between virtuous agency and the relevant justiﬁed or known belief, such that a belief counts as justiﬁed, say, only if an exercise of intellectual virtue (or something like it) plays a salient causal role in forming or is the best explanation of the relevant belief. (Zagzebski focuses, more speciﬁcally, on the causal connection between virtuous agency and the truth of known beliefs.) Accordingly, in the present case, even if my agency were not completely idle, it clearly would not be the best explanation of why I form the relevant belief; again, the best and most salient explanation for this belief is simply the proper (and brute) functioning of my visual faculty. Again, see Baehr (2006) for more on this and related points. It is also worth noting the possibility that my agency might in some sense work against the formation of my belief in this case. Being under an imminent writing deadline, for instance, I might try to deny what has just occurred (i.e. that the lights have gone out); I might, to a certain extent, be incredulous. Here I would be attempting to resist the truth and so would be displaying a kind of intellectual vice. Nevertheless, there can be little doubt that I would be unable to resist the force of my own (brute) cognitive nature and would come to believe (and indeed to know—even from an evidentialist standpoint) that the lights have gone out. 35 A similar story could be told regarding the beliefs of George and Daphne. George’s agency is involved in the formation of his belief to the extent that it explains his intellectual laziness and obliviousness. And Daphne’s is involved to the extent that it is the cause of her ignoring or suppressing the relevant counterevidence.

100

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

regards this evidence.36 By contrast, in cases of brute or passive justiﬁcation, the believer’s agency does not contribute to his or her evidential situation. In the case of passive justiﬁcation discussed above, for instance, I acquire and confront the evidence concerning the lighting in the room simply as a result of the brute or natural functioning of my basic cognitive machinery. My evidential situation does not implicate or involve my agency in any signiﬁcant way. We are now in a position to see how (E2) should be amended: E*. S is justiﬁed in believing p at t if and only if S’s evidence at t appears to S to support p, provided that, if S’s agency makes a salient contribution to S’s evidential situation with respect to p, S functions, qua agent and relative to that contribution, in a manner consistent with intellectual virtue. Several elements of (E*) require further commentary. First, note that the virtue requirement in (E*) does not have universal application: it applies only to cases in which a person’s evidential situation involves or implicates her agency in the relevant sense. Thus (E*) is not proposing an additional necessary or deﬁning condition for justiﬁcation. Instead it is laying down a constraint regarding when or under what conditions a belief’s being supported by good evidence generates justiﬁcation. The idea is that while justiﬁcation is essentially a matter of the possession of good epistemic reasons, there are cases in which such reasons, in order to generate justiﬁcation, must be possessed against the ‘backdrop’, so to speak, of virtuous cognitive functioning. Secondly, the ‘qua agent’ qualiﬁcation in the ﬁnal clause of (E*) underscores the fact that while, in the cases in question, an agent’s brute cognitive machinery might be in good working order, and thus that the agent might be ‘functioning’ well or virtuously at one level, justiﬁcation requires virtuous agency. Thirdly, (E*) does not require that a believer actually be intellectually virtuous. It does not require, that is, that the virtuous conduct in question ﬂow from a ﬁxed or settled disposition on the part of the believer (as it would with one who is genuinely virtuous).37 Fourthly, (E*) does not demand that a believer manifest the height of intellectual virtue or that her intellectual conduct be maximally intellectually virtuous. It requires merely that she function in a manner consistent with intellectual virtue (or that she refrain from functioning in a way that a virtuous person characteristically would not). Two ﬁnal observations concerning (E*) are in order. First, (E*) generates the correct result in connection with Cases 1–4 above. None of the subjects in these cases turn out to be justiﬁed according to (E*), for in each case, while the person’s agency does make a salient contribution to his or her evidential situation, the person fails to function in a manner consistent with intellectual virtue. Moreover, (E*) explains why, in cases of

36

As indicated in note 34 above, it is important that personal agency make a salient (versus, say, a minimal or background) contribution to, or that it largely determine, the subject’s evidential situation. 37 Nor does it require certain other elements of genuine virtue: e.g. that the person in question take pleasure in her virtuous action. This feature of (E*) is important, since a version of evidentialism that required actual virtue possession would be unreasonably demanding.

E V I D E N T I A L I S M , V I C E , A N D V I RT U E

101

brute or passive justiﬁcation, a person can be justiﬁed absent any virtuous activity. For again, these are cases in which the subject’s agency does not bear on his or her evidential situation. Second, a commitment to (E*) does not require the repudiation of internalism, at least on the standard way of understanding this doctrine. Internalism is typically said to require that the factors which justify a belief be ‘internal’ in the relevant sense. While there are problems with regarding as ‘internal’ the fact of the matter as to whether a belief, say, is the product of virtuous agency,38 (E*) does not require the repudiation of internalism, for (as noted above) it does not maintain that virtuous functioning or activity is itself a deﬁning feature of justiﬁcation or a ‘justifying factor’. Again, the notion of virtue is intended to play a mere background or constraining role.

5.3 Conclusion We have seen that a plausible evidentialist account of epistemic justiﬁcation must be supplemented by a proviso or constraint which, when applicable, requires cognitive agents to function in a manner characteristic of intellectual virtue. Accordingly, the discussion reveals a notable connection between evidentialist accounts of justiﬁcation and the more recent development of ‘virtue epistemology’. Virtue epistemologists give the concept of intellectual virtue a central role in epistemological theorizing.39 According to some virtue epistemologists, an exercise of intellectual virtue is a deﬁning feature of justiﬁcation or knowledge.40 The conclusion of the present paper does not go this far. In fact, we have considered reasons for denying that virtuous motives and actions are a necessary or deﬁning feature of justiﬁcation or knowledge. Nonetheless, we have also seen that an adequate version of evidentialism must at least make use of the concept of intellectual virtue. The upshot is that evidentialists and virtue epistemologists are concerned with some common conceptual territory. If so, it stands to reason that these philosophers would beneﬁt from sustained and careful reﬂection on each other’s theories. This in turn is likely to lead to additional philosophically fruitful and interesting avenues of inquiry.41, 42

38

What exactly the problem is, and how serious it is, depends on whether one thinks of internalism along ‘access’ or ‘mentalist’ lines. Presumably we do not always have access to whether our justiﬁed beliefs have been formed in a virtuous way. And whether a belief has been formed thus is never ‘in the mind’ in the mentalist’s sense. 39 See Baehr (2008) for an overview of the ﬁeld. 40 Linda Zagzebski (1996), for instance, deﬁnes knowledge (roughly) as true belief produced by virtuous motives and actions. 41 For instance, it would be worth considering from the other direction, as it were, just how close the connection is between the notion of believing with good evidence and intellectually virtuous belief formation. It is reasonable to think that intellectually virtuous people tend to form beliefs on the basis of good grounds or evidence. But is belief in the absence of such grounds ever intellectually virtuous? Moreover, insofar as these are genuine (non-tautologous) questions, they presuppose, contrary to certain prima facie plausible ways of thinking about intellectual virtue, that intellectual virtue is not itself to be deﬁned in evidentialist terms (e.g. as a matter of ‘following the evidence’ in one’s inquiries and beliefs). This too is an issue that is likely to be of interest to evidentialists and virtue epistemologists alike. 42 I am very grateful to Laurence BonJour, Trent Dougherty, and Dan Speak for helpful comments and conversations on earlier versions of this paper.

102

V I RT U E C R I T I Q U E S : E V I D E N C E A N D I N Q U I RY

References Alston, William. 1993. ‘Epistemic Desiderata,’ Philosophy and Phenomenological Research, 53, 3: 527–51. Axtell, Guy. 1997. ‘Recent Work on Virtue Epistemology,’ American Philosophical Quarterly, 34, 1: 1–26. Baehr, Jason. 2006. ‘Character in Epistemology,’ Philosophical Studies, 128, 3: 479–514. ——. 2008. ‘Four Varieties of Character-Based Virtue Epistemology,’ Southern Journal of Philosophy, 46: 469–502. BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. ——. 1992. ‘Externalism/Internalism,’ in Ernest Sosa and Jonathan Dancy (eds), A Companion to Epistemology. Oxford: Blackwell. ——. 1998. In Defense of Pure Reason. Cambridge: Cambridge University Press. BonJour, Laurence and Ernest Sosa. 2003. Epistemic Justiﬁcation: Internalism vs. Externalism, Foundations vs. Virtues. Malden, MA: Blackwell Publishing. Chisholm, Roderick. 1989. Theory of Knowledge, 3rd edn. Englewood Cliffs, NJ: Prentice Hall. Conee, Earl and Richard Feldman. 2004. Evidentialism: Essay in Epistemology. Oxford: Oxford University Press. Feldman, Richard. 2005. ‘Justiﬁcation Is Internal,’ in Matthias Steup and Ernest Sosa (eds), Contemporary Debates in Epistemology, pp. 270–84. Malden, MA: Blackwell Publishing. Geach, Peter. 1977. The Virtues. Cambridge: Cambridge University Press. Goldman, Alvin. 2001. ‘Internalism Exposed,’ in Matthias Steup (ed.), Knowledge, Truth, and Duty, pp. 115–33. Oxford: Oxford University Press. Greco, John. 2005. ‘Justiﬁcation Is not Internal,’ in Matthias Steup and Ernest Sosa (eds), Contemporary Debates in Epistemology, pp. 257–69. Malden, MA: Blackwell Publishing. Hursthouse, Rosalind. 1999. On Virtue Ethics. Oxford: Oxford University Press. Kornblith, Hilary. 1983. ‘Justiﬁed Belief and Epistemically Responsible Action,’ Philosophical Review, 92, 1: 33–48. Plantinga, Alvin. 1993. Warrant: The Current Debate. Oxford: Oxford University Press. Steup, Matthias. 1995. An Introduction to Epistemology. Englewood Cliffs, NJ: Prentice Hall. —— (ed.) 2001. Knowledge, Truth, and Duty. Oxford: Oxford University Press. —— and Ernest Sosa (eds) 2005. Contemporary Debates in Epistemology. Malden, MA: Blackwell Publishing. Swinburne, Richard. 2001. Epistemic Justiﬁcation. Oxford: Clarendon Press. Zagzebski, Linda. 1996. Virtues of the Mind. Cambridge: Cambridge University Press. ——. 1999. ‘What is Knowledge?’ in John Greco and Ernest Sosa (eds), The Blackwell Companion to Epistemology, pp. 92–116. Oxford: Blackwell.

PART III

Evidentialism and Skepticism It should not be taken for granted that the most important part of an epistemology is what it has to say about skepticism. Nevertheless, failure to have something productive to say on the matter or, worse yet, making skepticism likely would be a great cost to a theory. In ‘First Things First’ and ‘Making Sense of Skepticism’ (the ﬁrst and last chapters of Evidentialism, and the only two chapters written just for that volume), Conee and Feldman address skepticism more directly than just about anywhere else. In both those essays, they show how the skeptical puzzle is connected with very fundamental and methodological aspects of epistemology. There is much to learn from considering the skeptical puzzle. Matthias Steup is more optimistic about an evidential response to skepticism, at least of the BIV variety (and the extension to at least certain other kinds of skepticism are fairly obvious). Steup presents an evidentialist response to the following standard skeptical argument: (1) If I know that I have hands, I know I’m not a disembodied brain in a vat (BIV) (2) I do not know that I am not a BIV. So, (3) I do not know that I have hands. Steup considers and rejects attempts to avoid the conclusion of this argument by denying the closure principle that motivates (1). He then presents an evidentialist case for denying (2)—we have great evidence that we are not BIVs, and that evidence is sufﬁcient for knowledge. He begins by listing two principles that might lead one to accept (2). The ﬁrst holds that if on some skeptical hypothesis (H) one’s evidence would be the same as it is now, then one has no evidence against (H). The second holds that if the same antecedent holds, then one does not have enough evidence to know that (H). Steup thinks these principles are false. We have, he says, excellent evidence that there are no round squares, and thus that no round square is deceiving me. But on the hypothesis that a round square was systematically deceiving me about the modal facts, I’d still have the same modal intuitions that give me such great evidence that there are no round squares.

104

EVIDENTIALISM AND SKEPTICISM

However, Steup says, I still have evidence that there are no round squares, and that evidence amounts to knowledge. Steup then presents some fairly unsurprising evidence that there are no BIVs— testimony of neurophysiologists, things of that nature. Notably lacking in his list of evidence against the existence of BIVs is his perceptual experience of his own hands. He seems to think that appealing to evidence of that sort against the argument in question would be question-begging. Steup spends the last part of the paper discussing the nature of question-begging and whether or not the argument he has presented is question-begging. He rejects a structural account of question-begging on which a response to a skeptical argument begs the question if it appeals to a premise that is threatened by the skeptical argument in question. In its place he tries to give us an evidentialist account of question-begging. On this account, an anti-skeptical argument that employs premise p, based on evidence E, begs the question against skeptical hypothesis H if E does not favor p over H. Michael Bergmann argues that the kind of response to skepticism Conee considers in ‘First things First’ which concerns the Problem of the Criterion fails to avoid the Great Pumpkin Objection (hereafter GPO). The GPO to a response to skepticism is that it can be mimicked by ridiculous views, which suggests that the response fails. Defenses against the GPO can argue either that ridiculous views cannot successfully mimic the original view, or that although the mimicry is successful, this doesn’t imply that the reasonable response is inadequate. Bergmann argues that Conee takes the unsuccessful mimicry approach. The ﬁrst GPO to evidentialism’s response to skepticism is that the evidentialist view is merely self-afﬁrming, and ridiculous views can be self-afﬁrming as well. Conee argues that SE is defended by reasons that seem to indicate the truth of SE, not merely by virtue of self-afﬁrmation. Further, Conee argues that there is no reason to think that the procedures endorsed by ridiculous views confer epistemic reasons. Bergmann argues, however, that proponents of ridiculous views could hold that the procedures they defend do confer epistemic reasons.

6 Evidentialist Anti-skepticism Matthias Steup

6.1 The BIV argument and how one might respond to it Epistemologists worry about not knowing they have hands. The worry arises from skeptical arguments such as the notorious Brain-in-the-Vat argument, which goes as follows:1 The BIV Argument (1) KH ! KBIV2 (2) KBIV (3) KH The ﬁrst premise can be seen as an implicit appeal to the closure principle, closure for short.3 If the underlying assumption is made explicit, the argument goes as follows: The BIV Closure Argument (1) (2) (3) (4)

[KH & K(H ! BIV)] ! KBIV K(H ! BIV) KBIV KH

When epistemologists worry that the conclusion of this argument might be true, their worries are fueled by the undeniable plausibility of the premises. The ﬁrst premise is an instance of closure, which is a principle with a degree of plausibility so high that some

If you are a brain in a vat, your brain was removed from your skull and is kept alive, ﬂoating in a vat. The nerve endings of your brain are stimulated in such a way that you have exactly the sort of experiences you would have if you had a normal body and were enjoying a normal life. 2 KH = I know I have hands, KBIV = I know I’m not a BIV. 3 According to the closure principle, we know what our knowledge entails if the entailments themselves are known. So if I know p, and I know that p entails q, then I know q. There are problems calling for a more cautious articulation of the principle, but these problems may be ignored in the present context. For a defense of closure and a discussion of these problems, see Hawthorne’s ‘The Case for Closure’ in Steup and Sosa (2005). 1

106

EVIDENTIALISM AND SKEPTICISM

consider it axiomatic (Cohen 1988, 1999). The second premise simply states a logical consequence of the BIV hypothesis and is thus completely safe. And the third premise exploits the undeniable difﬁculties of making a case against a hypothesis that’s deliberately designed to rob one of all possible evidence against it. So it looks like the BIV argument does give epistemologists plenty to worry about. To rebut the argument, which responses are there to choose from? Here are three:4 1. Non-epistemic entitlements, or knowledge without reasons. According to this response, it’s possible to know that skeptical hypotheses are false without having evidence or reasons against them (Stine 1976, 2004). 2. Closure denial. This response is associated with either Dretske’s conclusivereasons theory or Nozick’s sensitivity-based account of knowledge. 3. Evidentialist anti-skepticism. According to this option, appearance notwithstanding, we actually have evidence against the BIV hypothesis, and this evidence is strong enough for knowing that the hypothesis is false.5 Of these responses, I think the third is the best. In what follows, I will criticize the ﬁrst two and defend the third. I won’t have much to say about the non-epistemic entitlement response. According to that response, knowledge requires reasons except when it comes to knowing that skeptical hypotheses are false. So if I am to know I have hands, I have to have good reasons for thinking I have hands. But when it comes to knowing that I’m not a BIV, then I can enjoy such knowledge even though I have no reason at all for thinking that the BIV hypothesis is false. One problem with this approach is that it makes such knowledge mysterious. If knowledge requires reasons except when it comes to knowing that skeptical hypotheses are false, how can my knowledge of not being a BIV be explained? Another problem is arbitrariness. According to the non-epistemic entitlement response, knowledge requires reasons except when it comes to knowing the falsehood of skeptical hypotheses. What justiﬁes making this exception? What we would need for this response to be more attractive is a principled account of when reasons are necessary for knowledge and when they are not. In the absence of such an account, the non-epistemic entitlement suggestion must be rejected as arbitrary.

6.2 Dretske’s denial of closure response Next, let’s consider the response of denying closure, championed by Fred Dretske (1970, 1971).

4

I’m not mentioning contextualism because, except for one further footnote, it will lie outside the scope of this paper. 5 Evidentialism as an epistemological position is of course closely associated with the writings of Earl Conee and Richard Feldman. For a collection of their most important essays, see their Evidentialism. Essays in Epistemology (2004).

EVIDENTIALIST ANTI-SKEPTICISM

107

According to Dretske, your belief that p amounts to knowledge if, and only if, you have a conclusive reason for believing p. And your reason for p is conclusive just in case, if p were false, you wouldn’t have that reason. Your hand-like experiences give you a conclusive reason for believing that you have hands, for if you didn’t have hands you wouldn’t have those experiences. However, your hand-like experiences don’t give you a conclusive reason for thinking you are not a BIV. They do not because if you were a BIV, you would still have those experiences. The same applies to any other reason you might have for thinking you are not a BIV. The BIV hypothesis is designed to ensure that, no matter what anti-BIV reason we consider, you would still have it even if you were a BIV. So reasons for thinking you are not a BIV cannot be conclusive. And thus we get the outcome that, although you know you have hands, and you know that your having hands entails your not being a BIV, you don’t know that you are not a BIV. According to one familiar objection to this response, it engenders an abominable conjunction: I know I have hands, but I don’t know that I’m not a handless BIV (DeRose 1995). Later on I’ll mention a few more abominable conjunctions. For now, I want to press a different objection. To me, it seems a main ﬂaw with Dretske’s approach is that it restricts the scope of knowledge too much. To begin with, as a response to skepticism, Dretske’s theory is quite concessive. It concedes that we cannot know skeptical hypotheses to be false. The claimed beneﬁt of this concession is that ordinary knowledge—such as knowledge of one’s hands—is preserved. But, on Dretske’s theory, how much of ordinary knowledge is really secure? For example, do I now know, according to Dretske’s theory, that there is a desk in my ofﬁce on campus? I am now in my study at home, not in my ofﬁce on campus. Hence my reason for believing there’s a desk in my ofﬁce is not a present desk-like-experience but rather memories of past desk-like-experiences plus a bit of background knowledge: desks are rarely removed from faculty ofﬁces. But that reason—the combination of remembered perceptual experiences and general knowledge of desk-removal on campuses—is not conclusive: had my desk been removed or stolen, I would still have that reason. What, according to Dretske’s theory, I do know is that last time I looked there was a desk in my ofﬁce. But I do not know that there is now a desk in my ofﬁce. It seems to me, though, that this is something I know. That’s why I think Dretske’s theory is too restrictive. Here’s something else I think I know, but don’t know according to Dretske’s theory. I think I now know that there is not now an atomic bomb in my basement. My reason for believing that there isn’t an atomic bomb in my basement consists of a set of premises about atomic bombs: They are very difﬁcult to acquire. People who have them legally keep them within the conﬁnes of military installations. People who have them illegally are unlikely to put them in my basement. But of course these reasons are not conclusive. If due to fantastic coincidence, an atomic bomb had found its way into my basement, I would still have these reasons. So Dretske’s theory implies that I do not know that there is not now an atomic bomb in my basement. According to Dretske, if I want to know whether or not there is an atomic bomb in my basement,

108

EVIDENTIALISM AND SKEPTICISM

I would have to go and look. Until then, I remain in a state of ignorance. Once again, it seems to me Dretske’s account is too restrictive. Like Dretske, I would say that having knowledge requires having a reason. But unlike Dretske, I think that for a reason to give you knowledge, it need not be conclusive. That doesn’t mean any old reason can give you knowledge. There are reasons that, while they make believing p more reasonable than not believing p, are not strong enough to give you knowledge of p. My reasons for thinking that I will be alive one year from now are an example of that. How strong, then, must a reason for p be if that reason is to give one knowledge of p? According to the kind of anti-skeptical evidentialism defended here, a reason is good enough for knowledge if it eliminates all reasonable doubt. My reasons for thinking that there is a desk in my ofﬁce, and that there is not an atomic bomb in my basement, eliminate, it seems to me, all reasonable doubt. That’s why I think I know these things. Dretske thinks that knowledge of p always results from indicators that carry the information that p is true. Indicators are perceptual experiences, memories, as well as various forms of testimony: newspaper articles, books, and what people tell me. I would of course agree that such indicators can give us knowledge. The question is whether they are the only things that can give us knowledge. It seems to me in addition to such indicators, there is something else that can give us knowledge: reasoning applied to propositions expressing bits of common knowledge. But as the two examples I mentioned show, such reasoning need not be conclusive. Dretske would therefore deny that it can give us knowledge; I would insist that it can. In response to my view, I think Dretske would reply that, even if I’m that liberal about what it takes to know, I will still not be able to rebut the BIV argument. For no matter how liberal an account of knowledge is, it will not allow for knowledge of not being a BIV. Hence, according to Dretske, there is only one successful way of rebutting this argument: to give up closure, that is, to claim knowledge of one’s hands while admitting one can’t know one is not a BIV. I disagree with that, and I will now proceed to explain why.

6.3 Why we might think we don’t know we are not BIVs Many epistemologists think that the third premise of the BIV Closure argument is extremely plausible. The more one thinks about it, the more plausible it becomes. These epistemologists seem to be convinced that at least one of the following two theses is true: 1. The No-Evidence-At-All Thesis One has no evidence at all for thinking that one is not a BIV. Generalized: If according to a skeptical hypothesis, H, one’s evidence is the same as it is now, then one has no evidence at all against H.

EVIDENTIALIST ANTI-SKEPTICISM

109

2. The No-Evidence-That’s-Good-Enough-for-Knowledge Thesis One has some evidence for thinking one is not a BIV, but that evidence is not good enough for knowledge. Generalized: If according to a skeptical hypothesis, H, one’s evidence is the same as it is now, then one has no evidence that’s good enough for knowing H. The rationale for the No-Evidence-At-All thesis is obvious: A person who undergoes envatment does not undergo any change of her evidence. So after envatment, one has exactly the same evidence one used to have before envatment. Consequently, one cannot have any evidence for thinking one is not a BIV.6 Those who favor the Not-Good-Enough thesis assert that we have some reasons for thinking we are not envatted. For example, it would appear the technology for keeping a brain alive for an extended period does not yet exist. However, advocates of the Not-Good-Enough thesis would hasten to add that such evidence doesn’t do what evidence must do if it is to give us knowledge: protect us in a robust way against error. It does not so protect us because if we were BIVs we would still think that the technology needed for envatment is not available. The same holds for any evidential item that could be brought to bear against the BIV hypothesis. Dretske would say that such evidential items do not add up to conclusive reasons. Therefore, whatever antiBIV evidence we might have, it isn’t good enough for us to know that the BIV hypothesis is false.

6.4 Easter Bunny deception Both theses strike me as false. For an assessment of their plausibility, I recommend considering a different type of skeptical scenario, one in which one is envatted or otherwise deceived not by a mad scientist or some sinister state or rogue agency, but rather by the Easter Bunny. If one is a BIV, one thinks one has hands while in fact one does not. The same is going on if the Easter Bunny is the agent of deception. Being handless, and indeed being altogether disembodied, the Easter Bunny’s victim labors under the illusion of having a normal body. The Easter Bunny deception analog to the BIV argument, then, runs as follows: The Easter Bunny Deception Argument (1) (2) (3) (4)

[KH & K(H ! EBD)] ! KEBD7 K(H ! EBD) KEBD KH

Dretske employs such reasoning in ‘The Case Against Closure,’ in Steup and Sosa (2005: 23). EBD = Easter Bunny deception (I am deceived by the Easter Bunny into thinking I have hands (a body) when in fact I do not. KEBD = I know there is no Easter Bunny deception (I know that I’m not deceived by the Easter Bunny). 6 7

110

EVIDENTIALISM AND SKEPTICISM

How would one want to respond to this argument? I’m inclined to think most people take themselves to know that the Easter Bunny does not exist. I, in any case, take myself to know this. Assuming, then, we all think we know the Easter Bunny does not exist, we might wonder on the basis of what evidence we have acquired this bit of knowledge. Let me just mention one evidential item among many: having paws, bunnies lack the manual dexterity required for distributing and hiding eggs.8 Of course, if we think that knowledge requires truth-entailing evidence, then the paw argument is no good at all. But, obviously, if we take ourselves to know that the Easter Bunny does not exist, then we don’t think that knowledge requires truth-entailing evidence— which is to say we are endorsing fallibilism. According to fallibilist evidentialism, if evidence is to give us knowledge, it need not be truth-entailing, but it must eliminate all reasonable doubt. I take it our evidence leaves no reasonable doubt as to the Easter Bunny’s nonexistence. That’s why, according to fallibilist evidentialism, we know that the Easter Bunny does not exist. Let’s suppose, then, we agree that the following thesis is true: The Easter Bunny Nonexistence Thesis We have evidence for Easter Bunny nonexistence that’s good enough for knowledge. What’s interesting about this thesis is this: if you think it is true, and if you think your body of knowledge can be enlarged using deduction, then you should reject the NoEvidence-At-All and the Not-Good-Enough theses. Consider ﬁrst the general form of the No-Evidence-At-All thesis: The No-Evidence-At-All Thesis If, according to a skeptical hypothesis, H, one’s evidence is the same as it is now, one has no evidence at all against H. Here is why this thesis is false. In a world in which you are deceived by the Easter Bunny, your evidence is the same as it is in this world. In this world, you have excellent evidence for thinking that the Easter Bunny does not exist. Hence, in a world in which you are deceived by the Easter Bunny, you also have excellent evidence for thinking that the Easter Bunny does not exist. Next, imagine yourself the Easter Bunny’s hapless victim. It occurs to you that you can work on your evidence using deduction. You reason as follows: 8

Here are a few more reasons for thinking the Easter Bunny does not exist: The single Easter Bunny problem: one solitary Easter Bunny can’t distribute Easter eggs on a semi-global scale; the language and intelligence problem: the task of semi-global Easter egg distribution could not be pulled off by creatures who have the linguistic ability and level of intelligence characteristic of bunnies; the distributive interaction problem: in the course of Easter egg distribution, multiple encounters between humans and the Easter Bunny would be unavoidable, encounters that would clearly establish the Easter Bunny’s existence; the media problem: if the Easter Bunny did exist, we would expect both scientiﬁc and popular literature about the Easter Bunny. Finding still further reasons should be an easily accomplishable task.

EVIDENTIALIST ANTI-SKEPTICISM

111

The Easter Bunny Nondeception Closure Step The Easter Bunny does not exist. If so, I’m not deceived by the Easter Bunny. Therefore: I’m not deceived by the Easter Bunny. And so it turns out that even if you were deceived by the Easter Bunny, you would have an excellent argument for believing that your are not deceived by the Easter Bunny. So it looks like we may put forward the following thesis: The We-Have-Evidence Thesis We have excellent evidence for thinking that the Easter Bunny deception hypothesis is false even though, if that hypothesis were true, our evidence would be exactly the same as it is now. The We-Have-Evidence thesis strikes me as true. It seems to me, therefore, that the No-Evidence-At-All thesis is false. Next, consider the Not-Good-Enough thesis, which in its general form says the following: The No-Evidence-That’s-Good-Enough-for-Knowledge Thesis If according to a skeptical hypothesis, H, one’s evidence is the same as it is now, one has no evidence that’s good enough for knowing H. Let’s assume we are in agreement about the following: we have evidence good enough for knowing that the Easter Bunny doesn’t exist. Now each of us can perform a little deduction: If one knows that the Easter Bunny doesn’t exist, then one knows one is not deceived by the Easter Bunny. This looks like a rather safe step. It is not a complex and lengthy piece of reasoning that weakens the evidence to which it is applied. So if our anti-Easter-Bunny-existence evidence gives us knowledge of the Easter Bunny’s nonexistence, one should think that, performing the little deduction just mentioned, that very same evidence puts us in a position to know we are not deceived by the Easter Bunny. Let’s put this point in the form of another thesis: The We-Know-It Thesis We know that the Easter Bunny Deception Hypothesis is false even though, if that hypothesis were true, our evidence would be exactly the same as it is now. If the We-Know-It thesis is true, then the Not-Good-Enough thesis is false. Since the We-Know-It thesis is exceedingly plausible, it seems to me it supplies us with a good reason to reject the Not-Good-Enough thesis. The arguments for the We-Have-Evidence and the We-Know-It theses presuppose that we can expand our stock of knowledge using deduction. So they rely on closure. Dretske (2005) doesn’t hold that we can never use deduction to expand our stock of

112

EVIDENTIALISM AND SKEPTICISM

knowledge. However, he would deny that, using Easter Bunny nonexistence as our starting point, we can, using deduction, come to know that we are not victims of Easter Bunny deception. Deduction does not generate knowledge when it comes to consequences for which we lack conclusive reasons. That’s why, on Dretske’s view, closure fails. What’s nice about the deceptive Easter Bunny and other deceivers of the same ilk is the following: They illustrate just how painful it is to accept closure failure, at least when it comes to obviously non-existing agents of deception. For the position Dretske advocates implies conjunctions as abominable as the following: Closure Denial Costs You know that the Easter Bunny does not exist, but you don’t know that you are not deceived by the Easter Bunny. You know that Napoleon is dead, but you don’t know that you are not deceived by Napoleon. You know that dinosaurs are extinct, but you don’t know that you are not deceived by some dinosaur. As far as abominable conjunctions go, these are particularly egregious. It seems to me we should avoid them and agree that, if one knows the Easter Bunny doesn’t exist, then one also knows one is not deceived by the Easter Bunny. But then the following, rather effective response to the Easter Bunny Deception argument becomes available to us: The Easter Bunny Deception Counter-Argument (1) (2) (3) (4)

KEBE9 K(EBE ! EBD) [KEBE & K(EBE ! EBD)] ! KEBD KEBD

The short version of this argument goes like this: I know I am not deceived by the Easter Bunny because the Easter Bunny does not exist. The Easter Bunny nonexistence response to the Easter Bunny Deception argument is based on fallibilist evidentialism. According to this view, the standard we must meet to know is high, but not excessively high: we must be in possession of evidence that eliminates all reasonable doubt. There is no reasonable doubt about the Easter Bunny’s nonexistence. Nor is there reasonable doubt about the relevant entailment. Hence one knows that the Easter Bunny deception hypothesis is false.

6.5 Rebutting the BIV argument We have now reached the end of our detour through the territory of Easter Bunny skepticism. Next, let’s see whether BIV skepticism can be dealt with in an analogous fashion. So let’s consider the following reply to the BIV argument: 9

EBE = Easter Bunny existence; KEBE = I know the Easter Bunny does not exist.

EVIDENTIALIST ANTI-SKEPTICISM

113

The BIV Counter-Argument (1) (2) (3) (4)

KBIVE10 K(BIVE ! BIV) [KBIVE & K(BIVE ! BIV)] ! KBIV KBIV

The second premise seems rather innocuous. If BIVs don’t exist, then I’m not a BIV. That’s certainly beyond any reasonable doubt. So let’s move on to the ﬁrst premise. Is there any reasonable doubt as to the nonexistence of BIVs? Let’s dwell for a moment on the relevant evidence. Items that come to mind are the following: 1. Textbooks of neurophysiology don’t have a chapter entitled ‘Envatment’. 2. Departments of neurophysiology don’t offer courses entitled ‘Envatment 101’. 3. If you bother to call a renowned neurophysiologist or brain surgeon and ask whether envatment is possible, the answer is going to be ‘no’. 4. Essay collections for courses on applied ethics don’t have a chapter entitled ‘The ethics of envatment’. 5. No known episode of 60 Minutes has ever investigated let alone asserted the existence of BIVs. 6. There is no known case of someone ever having been sued for or found guilty of envatting a person. This list could go on. If one puts enough effort into it, it could ﬁll pages. Now, obviously even the collective force of such a list does not entail the nonexistence of BIVs. Nor does such a list give us conclusive reasons for the nonexistence of BIVs. But surely it eliminates all reasonable doubt about their nonexistence. The proposition that there are no BIVs, then, meets the standard of knowledge that fallibilist evidentialism endorses. We know that BIVs don’t exist, and hence, performing an easy deduction, we can know that we are not victims of envatment. This could change. Perhaps at some point in the future, envatment will be a widespread and much reported phenomenon. Then it’s going to be more difﬁcult to know one is not a BIV. We are not there yet, though. It’s like the difference between living in West-Lafayette, Indiana, and living in a crime infested metropolis. In WestLafayette, I know my car is where I parked it. In a crime infested metropolis, I might not. Likewise, in this world we know we are not envatted. In a world in which envatment is common place, it might be difﬁcult to know this. Let me review the salient points. I think the best response to BIV skepticism goes like this. We have excellent evidence for thinking that BIVs don’t exist, evidence that eliminates all reasonable doubt about BIV nonexistence. We have such evidence even though our overall evidence would be the same if we were BIVs. Therefore, the NoEvidence-At-All thesis is false. Using BIV nonexistence as a premise, an easy deduction 10

BIVE = BIV existence. KBIVE = I know that BIVs don’t exist.

114

EVIDENTIALISM AND SKEPTICISM

allows us to infer that, if BIVs don’t exist, we are not BIVs. Anti-BIV-existence evidence thus becomes anti-BIV-deception evidence. But is our anti-BIV-deception evidence good enough to know we are not BIVs? Since the deduction in question is simple and rationally compelling, there is no reason to deny that we know we are not BIVs by deducing it from the nonexistence of BIVs. We should, therefore, reject the Not-Good-enough thesis. If Dretske’s conclusive reasons requirement were true, the inference from BIV nonexistence to one’s not being a BIV would be blocked. However, Dretske’s conclusive reasons requirement is not obviously true. To the contrary, it is highly problematic. If we reject it, no obstacle remains to saying that we can know we are not BIVs by deducing it from the nonexistence of BIVs.11

6.6 Does evidentialist anti-skepticism beg the question? Some would argue that the BIV nonexistence argument begs the question. I do not think, however, that the charge of question begging sticks in any obvious way. I this section, I’ll argue that the charge doesn’t stick. Why, then, think that the BIV nonexistence argument begs the question? The reason, I take it, would be that the argument appeals to the premise that I know BIVs don’t exist. The skeptic would of course argue that, since I don’t know I’m not a BIV, I don’t know that BIVs don’t exist. So in advancing the BIV nonexistence response, I rely on a premise the truth of which the skeptic is not prepared to grant. However, the skeptic uses a premise the truth of which I am not prepared to grant, namely the premise that I don’t know I’m not a BIV. It is of course true that, if I really don’t know that I’m not a BIV, then I don’t know that BIVs don’t exist. But it is also true that, if I really know BIVs don’t exist, then I do know I’m not a BIV. This symmetrical structure reveals itself when we compare the two arguments in question: BIV Counter-Argument

Skeptical Reply

(1) KBIVE ! KBIV (2) KBIVE (3) KBIV

(1) KBIVE ! KBIV (2) KBIV (3) KBIVE

According to the anti-skeptical argument on the left, I know I’m not a BIV because I know BIVs don’t exist. According to the skeptical reply on the right, I don’t know

11

The reader might wonder how the response to skepticism advocated here is related to contextualism. According to contextualism, skepticism is correct in conversational contexts in which a high standard of knowledge is in operation, whereas anti-skepticism is correct in low standard contexts. Ignore the questions of whether contextualists are right about the way the standards of knowledge shift from one context to another and focus on this: according to contextualism, closure is true. Clearly, then, contextualists need to explain how, in low standard contexts, it is possible to know one is not a BIV, or at least to have the evidence needed for knowing one is not a BIV. Evidentialist anti-skepticism, as defended here, can be viewed as a strategy for providing that explanation.

EVIDENTIALIST ANTI-SKEPTICISM

115

BIVs don’t exist because I don’t know I’m not a BIV. The second premise of each of these arguments denies the conclusion of the other argument. What we are seeing is a clash between the skeptic’s modus tollens and the anti-skeptic’s modus ponens. In light of this symmetrical structure, why assume the BIV counter-argument begs the question? To get to the bottom of the issue, I suggest proceeding as follows: Since both arguments share the same ﬁrst premise, we should judge the respective merit of each argument by comparing the plausibility of each argument’s second premise. The issue of question begging, then, comes down to this: What is more plausible, the skeptic’s premise that I don’t know I’m not a BIV, or the anti-skeptical premise that I know BIVs don’t exist? To approach answering this question, let’s examine two more pairs of arguments, thus generating an anti-skeptical slippery slope. Consider ﬁrst how the skeptic would respond to the Easter Bunny deception counter-argument: The EB Deception Counter-Argument

Skeptical Reply

(1) KEBE ! KEBD (2) K EBE (3) KEBD

(1) KEBE ! KEBD (2) K EBD (3) KEBE

According to the argument against Easter Bunny deception, I know I’m not deceived by the Easter Bunny because I know the Easter Bunny does not exist. According to the skeptical response, I don’t know that the Easter Bunny does not exist because I don’t know that I’m not deceived by the Easter Bunny. To assess whether the Easter Bunnynonexistence response is question-begging, we need to focus on the second premise of each argument: What’s more reasonable of me to believe: (a) that I know the Easter Bunny doesn’t exist, or (b) that I fail to know that I’m not deceived by the Easter Bunny? The suggestion I should abandon (a) on the basis of (b) seems silly, precisely because, as already mentioned above, there is no reasonable doubt as to the Easter Bunny’s nonexistence. So my anti-Easter Bunny evidence gives me excellent reasons for preferring (a) over (b). Therefore, I can hardly be accused of question-begging when I argue I know I’m not deceived by the Easter Bunny because I know that the Easter Bunny does not exist. Next, consider a completely absurd skeptical argument, the round-square-deception argument. It goes like this: ‘If I know I have hands, I know I’m not deceived by a round square. But I don’t know that I’m not deceived by a round square. Therefore, I don’t know that I have hands.’ Now consider the following two arguments: The Round-Square Deception Counter Argument

Skeptical Reply

(1) KRSE ! KRSD (2) KRSE (3) KRSD

(1) KRSE ! KRSD (2) KRSD (3) KRSE

116

EVIDENTIALISM AND SKEPTICISM

I say I know I’m not deceived by a round square because I know round squares don’t exist. The skeptic replies I don’t know round squares don’t exist because I don’t know I’m not deceived by a round square. Once again, we need to focus on the second premise of each argument. What’s more reasonable of me to believe: (c) that I know round squares don’t exist, or (d) that I don’t know that I’m not deceived by a round square. Obviously, it is (c). The nonexistence of round squares cannot seriously be called into question. So when I prefer (c) to (d), there is no question for me to beg. Therefore, I do not beg the question when I argue I know I’m not deceived by a round square because I know that round squares don’t exist. It is difﬁcult to take round-square skepticism seriously. But when we consider BIV skepticism, it is perhaps not easy to see whether the skeptic or the anti-skeptic has the upper hand. To clear this up, we can employ the device of a slippery slope. At the end of the slippery slope, the light of reason shines brightly. Here, we are observing a skeptical argument that can only be classiﬁed as preposterous. Round-square skepticism is not a serious challenge. It is a mere play with words, utterly lacking in substance. Again: the nonexistence of round squares is not something that can reasonably be called into question.12 The round-square-nonexistence argument, therefore, does not qualify as an instance of begging the question. Now consider BIV skepticism, located at the other end of the slope. If BIV skepticism is a serious challenge and evidentialist anti-skepticism begs the question against it, then something must have changed on the way up during the transition from round-square skepticism to Easter Bunny skepticism to BIV skepticism. What might have changed? Perhaps it is the strength of the evidence. So perhaps the following diagnosis is true: 1. The evidence against the existence of round squares is as strong as it can get. There is no reasonable doubt as to the nonexistence of round squares. Therefore, round-square skepticism is unreasonable. The anti-skeptic does not beg the question against it. 2. The evidence against the Easter Bunny’s existence is not quite as strong, but still very strong, eliminating all reasonable doubt. Therefore, Easter Bunny skepticism is unreasonable. The anti-skeptic does not beg the question against it. 3. But when it comes to evidence against the existence of BIVs, we witness a catastrophic weakening. The nonexistence of BIVs is very much in doubt. BIV skepticism is, therefore, reasonable, and the anti-skeptic begs the question against it. While I ﬁnd myself in agreement with the ﬁrst two steps of this diagnosis, I must reject the third. I don’t think there is a catastrophic weakening of the evidence when it comes to the nonexistence of BIVs as opposed to the nonexistence of round squares or the nonexistence of the Easter Bunny. Rather, I think there is plenty of evidence making 12 Of course there might be good reasons to become suspicious of our modal intuitions. But round-square deception is not among them.

EVIDENTIALIST ANTI-SKEPTICISM

117

the assertion of BIV nonexistence reasonable.13 So my view is that, while the strength of the evidence diminishes with each step on the slippery slope, it does so only in very small increments. No changes occur that dramatically change the picture when we move from round-square skepticism to BIV skepticism. Hence I suggest replacing the third part of the diagnosis with: 3*. When it comes to evidence against the existence of BIVs, no signiﬁcant weakening has taken place. There is, therefore, no reasonable doubt as to the nonexistence of BIVs. BIV skepticism is, therefore, unreasonable, and the antiskeptic does not beg the question against it. Critics of evidentialist anti-skepticism who wish to oppose my line of reasoning are confronted with a dilemma. Consider the following thesis: The No-Dramatic-Change Thesis None of the steps of the slippery slope from round square to Easter Bunny to BIV skepticism involves a dramatic weakening of the evidence. Skeptics who wish to reject my slippery slope argument face a dilemma: for skeptics, neither accepting nor rejecting the No-Dramatic-Change thesis will be an easy option. If they accept the thesis, they must either abandon skepticism or embrace round-square skepticism. If, on the other hand, they reject the thesis, they incur the obligation to identify a change signiﬁcant enough to rehabilitate BIV skepticism while accepting the obvious failure of round-square skepticism. It is not easy to see, however, exactly where that change might occur.

6.7 Two objections It could be objected that a dramatic change does occur. It occurs at the transition from round-square skepticism to Easter Bunny skepticism. Skeptical hypotheses, if effective, must be logically consistent, representing genuine possibilities. Round-square deception does not meet this requirement, and thus is not suitable for the skeptic’s purposes. There are two problems with this objection. First, if we say that a dramatic change occurs at the transition from round-square to Easter Bunny skepticism, we assign, as a consequence, too much credit to the latter. It is of course true that, unlike a deceiving round square, a deceiving Easter Bunny is at least logically possible. But that doesn’t mean the Easter Bunny deception hypothesis needs to be taken seriously. Though round-square deception is even less plausible than Easter Bunny deception, certainly the difference is way too small to have the result of making Easter Bunny skepticism look good by way of comparison.

13 At least, there is plenty such evidence under the present circumstances. It’s easy to imagine circumstances where such evidence is absent.

118

EVIDENTIALISM AND SKEPTICISM

Second, we should bear in mind that we are, after all, liable to make mistakes about what’s logically possible and what’s not. Therefore, propositions that strike us as necessarily true are not immune to doubt just because of the modal status we attribute to them. In principle, it is possible for the skeptic to cast doubt on the reliability of our modal intuitions. Whether such doubts are to be taken seriously depends on the error possibilities the skeptic appeals to. If the skeptic were to present us with empirical data suggesting that our modal intuitions are in general unreliable, then we would be confronted with a serious worry. If, however, the skeptic presents us with the hypothesis that we are deceived by a round square into thinking that round squares don’t exist, then there is no reason to worry. The explanation of why there is no reason to worry is not the shift from empirical to logical implausibility but the obviousness of the implausibility of the error possibility in question. If a skeptical argument is to succeed, the skeptical hypothesis it is based on must itself be plausible or at least not be implausible. That is what matters. What does not matter is whether the skeptical hypothesis targets our empirical evidence or our modal intuitions. I therefore reject the claim that a dramatic change takes place at the transition from round-square to Easter Bunny skepticism. According to the second objection, I am too conﬁdent in my negative assessment of Easter Bunny and round-square skepticism. Why not let the direction of the argument go the other way? Why not, that is, argue as follows: ‘Since I don’t know I’m not a BIV, I don’t know that BIVs don’t exist. Since that’s an unassailable starting point, epistemological reﬂection tells us we must employ analogous reasoning when it comes to a deceiving Easter Bunny or a deceiving round square. Starting with the plausibility of BIV skepticism, I should conclude that, since I don’t know that I’m not deceived by the Easter Bunny, I don’t know that the Easter Bunny does not exist, and that, since I don’t know I’m not deceived by a round square, I don’t know that round squares don’t exist. So instead of undermining BIV skepticism, what the slippery slope does is illustrate how strong and far-reaching the skeptic’s position really is.’ Here is my response to this objection. BIV skepticism rests on the premise that I don’t know I’m not a BIV. This premise, unless the slide down the slippery slope comes to a stop somewhere, ultimately comes into tension with the anti-skeptical premise that I know round squares don’t exist. So looking at the starting point and the end point of the slippery slope, we are comparing (1) I don’t know I’m not a BIV, and (2) I know round squares don’t exist. According to the objection under consideration, it would be dogmatic to protect (2) by rejecting (1). To me, it seems the opposite is true. If (1) was at least as plausible as the nonexistence of round squares, the objection might have a point. But surely, the credibility of (1) rises nowhere near the certainty of round-square nonexistence. To the contrary, (1) is a problematic premise in need of support. What epistemologists have said in support of this premise is highly debatable. If knowledge requires infallible evidence, or if knowledge requires a conclusive reason, then it follows of course that one doesn’t know one is not a BIV. But none of these requirements are uncontroversial. Likewise, if the No-Evidence-At-All thesis or the Not-Good-Enough thesis were true, then, from an evidentialist point of view, it would be impossible to

EVIDENTIALIST ANTI-SKEPTICISM

119

know that one is not a BIV. I have argued, however, that both of these theses are false. According to the evidentialist approach I have been advocating, knowledge results from the possession of evidence that eliminates all reasonable doubt. From that point of view, asserting that one is unable to know one is not a BIV is not an innocuous starting point but rather a deeply problematic claim, one that, in light of the slippery slope (towards ridiculous consequences) it engenders, should be rejected.

6.8 What would be begging the question? I have argued that, if one argues for knowledge of one’s hands and of one’s not being a BIV using BIV nonexistence as a premise, one does not beg the question against BIV skepticism. In this section, I will consider an argument that I am inclined to think does beg the question: the I-have-hands response: I know I am not a handless BIV tricked into believing I have hands because I know I have hands. The challenge will be to explain why the BIV nonexistence response does not beg the question whereas the I-have-hands response does. Let’s begin by looking at an example that nicely illustrates the difference between begging and not begging the question. Suppose I believe about a table before me, on the basis of its looking red to me: (1) The table is red. Suppose further someone confronts me with the following skeptical alternative:14 (2) The table is white and illuminated by red light. Finally, suppose I respond to this challenge by saying I know the table isn’t white and illuminated by red light because I know it is red. That, I think, would be begging the question.15 But why? Here is an initial answer. Knowledge of the table’s redness is the very thing the hypothesis ‘The table is white and illuminated by red lights’ threatens. But one cannot defeat a skeptical hypothesis by appealing to the very knowledge the hypothesis is intended to threaten or call into doubt. That’s why appealing to knowledge of the table’s redness is not the right response to defeating the red light hypothesis. If I wish to avoid begging the question about the table’s color, what I need to do is defeat the skeptical hypothesis by citing a reason against that hypothesis that is not itself threatened or undermined by the hypothesis, such as: (3) There are no red lights present. If I don’t have evidence in support of (3), then the table’s looking red does not put me in a position to know it is red. But if, on the basis of such evidence, I come to know 14 The kind of skepticism here would of course be local. Its point would not be to call into question all of our empirical knowledge, but merely to undermine knowledge of the table’s color on a particular occasion. 15 I’m in agreement with Stewart Cohen (2002) on this.

120

EVIDENTIALISM AND SKEPTICISM

that (3) is true, then the red light hypothesis is defeated, and the table’s looking red gives me knowledge of its being red. Generalizing from this example, we might be inclined to think that what is sufﬁcient for rendering an argument question-begging is the attempt to defeat a skeptical hypothesis by appealing to premises knowledge of which the hypothesis is meant to threaten. So to avoid question-begging, one must attack the hypothesis using premises that the hypothesis fails to threaten or potentially undermine, such as, in our present example, a premise asserting the absence of red lights. However, such an account of question-begging is subject to counterexamples. Here are two arguments that do not beg the question: Easter Bunny Nonexistence: I am not deceived by the Easter Bunny into thinking that the Easter Bunny does not exist because the Easter Bunny does not exist. Round-Square Nonexistence: I am not deceived by a round square into thinking that round squares don’t exist because round squares don’t exist. If we think of question-begging as a solely structural matter—a matter of not appealing to any premises a skeptical hypothesis threatens—both of these arguments must be classiﬁed as question-begging. Knowledge of the Easter Bunny’s nonexistence is the very thing the Easter Bunny deception hypothesis is meant to undermine. Hence, on our present account, the argument must be classiﬁed as question-begging. Likewise, knowledge of the nonexistence of round squares is the very thing the round-square deception hypothesis is meant to threaten. So, we might think, we can’t defeat this hypothesis by appealing to the nonexistence of round squares. So if defeating a skeptical hypothesis using premises knowledge of which the hypothesis is meant to threaten is indeed sufﬁcient for rendering an argument question-begging, then, it would appear, both of the arguments displayed above are question-begging—when in fact they are not. How can the two arguments be vindicated? I suggest that we need to replace the solely structural account of question-begging considered above with an evidentialist one. According to this alternative account, for an argument to beg the question, there must be, relative to the evidence on which the argument is based, a question to be begged. But our evidential situation is such that there are no questions to be begged when it comes to the nonexistence of the Easter Bunny or the nonexistence of round squares. How can, in general terms, such an evidentialist account of question-begging be articulated? Here is a suggestion: Question Begging: An anti-skeptical argument, based on a premise P justiﬁed by evidence E, begs the question against a skeptical alternative, A, if and only if (i) A is intended to undermine P; (ii) E fails to favor P over A. Consider again the example of the red table. Here, the skeptical alternative is that the table is white and illuminated by red light, which threatens or potentially undermines the proposition that the table is red. If I argue against the alternative using ‘The table is red’ as a premise, E consists of a visual experience: the table’s looking red to me. This bit of

EVIDENTIALIST ANTI-SKEPTICISM

121

evidence, by itself, gives to me no reason to favor ‘The table is red’ over ‘The table is white and illuminated by red light’. In contrast, suppose I rebut the skeptical alternative by appealing to E+: the table looks red to me and it appears to me that there are no red lights present. Clearly, E+ gives me a reason to favor ‘The table is red’ over ‘The table is white and illuminated by red lights. Let us now consider the I-have-hands response that I claim does, unlike the BIVnonexistence response, beg the question against the BIV hypothesis. The I-have-hands response goes like this: I know I am not a handless BIV tricked into believing I have hands because—on the basis of hand-like experiences—I know I have hands. The evidence on which the asserted knowledge of my hands is based consists of handlike experiences. But that evidence by itself gives me no reason to favor ‘I have hands’ over ‘I am a handless BIV tricked into believing I have hands’. Therefore, the hands response to the BIV argument begs the question. Next, compare the hands response with the BIV nonexistence response: I know that I have hands and am not a handless BIV tricked into believing I have hands because I have hand-like experiences and plenty of evidence in support of BIV nonexistence. Now the evidence brought to bear against the skeptical alternative is not just hand-like experiences, but an enlarged body of evidence: hand-like experience plus the total body of evidence supporting BIV nonexistence. This enlarged body of evidence entitles me to favor ‘I have hands’ over ‘I am a handless BIV tricked into believing that I have hands’. Initially, it might appear as though the BIV-nonexistence response begs the question against the BIV hypothesis because the BIV hypothesis is meant to undermine (among other things) my knowledge of BIV nonexistence. However, according to the account of question-begging I have proposed, I beg the question only if the total body of evidence I am appealing to fails to favor the challenged proposition—‘I have hands’ in our present example—over the skeptical alternative. But this is not true of the argument we are considering. BIV-nonexistence evidence defeats the BIV hypothesis and, together with hand-like experiences, grounds knowledge of one’s hands. Therefore, the account I have proposed identiﬁes, as I believe correctly, the I-have-hands response as question-begging and the BIV-nonexistence response as non-question begging.16

16 I presented an earlier version of this paper as the Presidential Address at the 2008 annual meeting of the Central States Philosophical Association in St Paul. Still earlier versions were presented at Purdue University, St Cloud State University, the University of Minnesota, Morris, at the Bled Philosophy Conference in Slovenia, and the Bruce Russell Conference in Healdsburg, CA. I wish to thank the audiences on those occasions for helpful and stimulating discussions. For helpful comments and/or discussion, I am particularly indebted to Mylan Engel, Richard Fumerton, Adam Leite, Alastair Norcross, and Bruce Russell.

122

EVIDENTIALISM AND SKEPTICISM

References Cohen, Stewart, 1988. ‘How to be a Fallibilist,’ Philosophical Perspectives, 2: 91–123. ——. 1999. ‘Contextualism, Skepticism, and the Structure of Reasons,’ Philosophical Perspectives, 13: 57–90. ——. 2002. ‘Basic Knowledge and the Problem of Easy Knowledge,’ Philosophy and Phenomenological Research, 65: 309–29. Conee, Earl and Feldman, Richard. 2004. Evidentialism. Essays in Epistemology. Oxford: Clarendon Press. DeRose, Keith. 1995. ‘Solving the Skeptical Problem,’ The Philosophical Review, 104: 1–52. Dretske, Fred. 1970. ‘Epistemic Operators,’ The Journal of Philosophy, 67: 1007–23. ——. 1971. ‘Conclusive Reasons,’ Australasian Journal of Philosophy, 49: 1–22. ——. 2005. ‘The Case Against Closure,’ in Matthias Steup and Ernest Sosa (eds). Contemporary Debates in Epistemology, pp. 13–26. Malden, MA: Blackwell. Steup, Matthias and Sosa, Ernest (eds) 2005. Contemporary Debates in Epistemology. Malden, MA: Blackwell. Stine, Gail. 1976. ‘Skepticism, Relevant Alternatives, and Deductive Closure,’ Philosophical Studies, 29: 249–61. ——. 2004. ‘On Epistemic Entitlement: Warrant for Nothing (and Foundations for Free)?’ Aristotelian Society Supplement, 78: 167–212.

7 Evidentialism and the Great Pumpkin Objection Michael Bergmann

Evidentialism, as understood by its chief contemporary proponents (i.e. Richard Feldman and Earl Conee), is the view that epistemic justiﬁcation supervenes on the evidence one has1—whether that evidence is other beliefs or experiences or feelings. Earl Conee argues that, unlike other views, evidentialism has a response to skepticism about the external world that escapes what can be called ‘the Great Pumpkin Objection’.2 That objection says, of a response to skepticism, that those endorsing a silly view—such as the view that the gift-delivering Great Pumpkin rises each Halloween over the most sincere pumpkin patch—can offer exactly parallel responses to those who are skeptical of their Great Pumpkin beliefs.3 The obvious suggestion is that if one’s response to skepticism about the external world can be mimicked by those defending such a silly view, then that response to skepticism is inadequate. There are two main ways to respond to the Great Pumpkin Objection when it is applied to your favored response to some sort of skepticism. You could say that the allegedly parallel response offered by those defending belief in a silly view does not successfully mimic your favored response to skepticism. Or you could admit that it does successfully mimic your favored response to skepticism and yet deny that this implies

1 Conee and Feldman (2004: 1). Note that this use of the term ‘evidentialism’ is different from the way it is sometimes used in philosophy of religion to refer to the view that belief in God can be made rational or justiﬁed only on the basis of propositional evidence. In particular, as I’m using the term here (following Feldman and Conee), evidentialism does not say that justiﬁcation (for belief in God or anything else) requires propositional evidence—i.e., evidence in the form of other beliefs. See Conee and Feldman (2004: 2). 2 Conee doesn’t use this terminology though he considers this sort of objection. 3 This way of using the term ‘Great Pumpkin Objection’ was, so far as I know, introduced by Plantinga (1981). He uses it to describe an objection to Reformed Epistemology, the view that belief in God can be properly basic (i.e. justiﬁed noninferentially). The charge in that case was that if the Reformed Epistemologist’s response to religious skepticism were satisfactory, it could easily be mimicked by those defending silly views (the suggestion was that this shows that the Reformed Epistemologist’s response to religious skepticism is inadequate). Ernest Sosa (1997: section VI) considers a similar complaint about externalist responses to skepticism, although the example he uses has to do with crystal ball reading, not Great Pumpkin beliefs. Elsewhere I’ve called this sort of objection to such responses to skepticism the ‘Anything Goes’ objection. See Bergmann (2006: 229–33, 2008).

124

EVIDENTIALISM AND SKEPTICISM

that your response to skepticism is inadequate. The former we can call ‘the unsuccessful-mimicry response’ and the latter we can call ‘the successful-mimicry response’. As I understand him, Conee adopts the unsuccessful-mimicry response in defending evidentialism against the Great Pumpkin Objection, arguing that those defending their silly view against skeptical opposition to it cannot successfully mimic the evidentialist response to external world skepticism. I will argue that Conee is mistaken about this. In the ﬁrst section, I say more about what evidentialism is and explain the basic idea of its response to external world skepticism.4 Then, in the second section, I present Conee’s explanation for how this evidentialist response to skepticism avoids the Great Pumpkin Objection. In the third section I argue that Conee’s explanation fails. In the ﬁnal two sections I consider the implications this has for the adequacy of evidentialist responses to skepticism.

7.1 Evidentialism and skepticism As noted above, evidentialists say that justiﬁcation supervenes on the evidence one has. But that’s so only if they are speaking of propositional justiﬁcation. Propositional justiﬁcation is the justiﬁcation a proposition has for a person, whether or not the person believes that proposition for the right reasons (i.e. on the basis of the evidence that makes it propositionally justiﬁed). In fact, a proposition can be justiﬁed for a person by the evidence she has even if the person doesn’t believe the proposition at all. All that matters for propositional justiﬁcation, according to evidentialists, is what evidence the person has. It doesn’t matter what she believes or what she bases her belief on. But if we are speaking instead of doxastic justiﬁcation—the kind of justiﬁcation a belief has—then evidentialists will require that the belief is based on the evidence that makes its content propositionally justiﬁed for the person holding the belief. So evidentialists think of a belief’s doxastic justiﬁcation as something that applies to a belief (not to an unbelieved proposition) and as being determined by two things: the evidence one has and what that belief is based on. Henceforth, in this chapter, when I speak of justiﬁcation, I’ll have in mind doxastic justiﬁcation, not propositional justiﬁcation.5 In addition to saying that a belief’s justiﬁcation supervenes on the evidence one has and what the belief is based on, evidentialists add that a belief is justiﬁed when it epistemically ﬁts the evidence one has; it lacks justiﬁcation when it fails to epistemically ﬁt that evidence (henceforth when I say ‘ﬁt’, ‘ﬁttingness’, etc. I have in mind the epistemic variety). So a belief’s justiﬁcation depends on one’s evidence; more 4

I focus on a particular version of evidentialism because that is what Conee does in defending his view. In their 1985 article, Feldman and Conee deﬁne the evidentialist position on propositional justiﬁcation. But they go on in that article to speak of well-foundedness (their name for doxastic justiﬁcation) and say of it what I say here of doxastic justiﬁcation. Feldman makes clear in his (2004: 147–8) and his (2005: 274–5) that he understands this distinction between propositional and doxastic justiﬁcation in the way I’ve been describing it. 5

E V I D E N T I A L I S M A N D T H E G R E AT P U M P K I N O B J E C T I O N

125

speciﬁcally, it depends on whether one’s belief ﬁts and is properly based on one’s evidence.6 Moreover, evidence comes in degrees; and how justiﬁed one’s belief is depends on how well one’s belief ﬁts one’s evidence. If one’s evidence e provides a really good reason (or very strong evidence) for one’s belief B, then we say that B ﬁts e very well and is justiﬁed to a high degree. If e provides only a moderate reason for B, then we say B ﬁts e moderately well and is justiﬁed to only a moderate degree. In defending the evidentialist’s response to skepticism (which I’ll discuss below) against the Great Pumpkin Objection, Conee decides (2004: 15) to focus on a particular version of evidentialism—what he calls ‘Seeming Evidentialism’ or SE.7 Generic evidentialism says that justiﬁcation depends on evidence one has and it limits that to things of which one is aware (like beliefs, experiences, and feelings). SE goes further and limits evidence—of the sort that can confer justiﬁcation on S’s belief that p—to all and only those things of which S is aware that S is spontaneously inclined to regard as indicative of p’s truth.8 It is a substantive thesis of SE that these things one is inclined to regard as indicative of p’s truth are indicative of p’s truth—so they count as good evidence for p and good reasons for p. This isn’t to say that good evidence for p can’t be overridden (for example by much stronger evidence for ~p or by evidence for thinking that one’s evidence for p on a particular occasion is, contrary to how things seem, not indicative of p’s truth). Thus, SE is a version of evidentialism, as deﬁned above, that limits evidence for p to all and only things one is aware of that seem to indicate p’s truth.9 So how does SE respond to external world skepticism? According to SE, if the skeptic’s thesis or premises seem false to you or if the skeptic’s arguments don’t seem to you to support the skeptic’s conclusions, then you don’t have good evidence for skepticism. And if, in addition, it seems to you that the various components of the Standard View are correct, then you are reasonable in rejecting skepticism and accepting the Standard View. (The Standard View says—of most of the things you

6

To speak more carefully, S’s belief B is justiﬁed (according to evidentialists) if and only if S has some evidence e which is such that: (i) B is a ﬁtting response to e, (ii) B is based on e, and (iii) S has no more inclusive body of evidence e0 such that B is not a ﬁtting response to e0 . Clause (i) points out that a belief’s justiﬁcation depends on the evidence one has; clause (ii) emphasizes that justiﬁcation depends on what the belief is based on; and clause (iii) makes it clear that it is one’s total evidence that is relevant, not just some component of it. See Feldman and Conee (1985) where they make these points about well-foundedness (their name for doxastic justiﬁcation). 7 Conee himself doesn’t endorse SE in his (2004), though he does endorse evidentialism. 8 Suppose you feel (and are aware of) an inclination to believe p. If in response to this inclination you believe p, it’s natural to think that you are regarding that inclination as indicative of p’s truth. Is a proponent of SE committed to saying that whenever you feel inclined to believe p and believe p on the basis of that inclination, you have evidence for p? I’m not sure what Conee would say about this way of interpreting SE. At the very least, it must be that the inclination to believe p makes p seem true to you—i.e. that you regard this inclination as being indicative of p’s truth. 9 The way I’ve explained SE here might be only one version of SE. Another version might be one that says evidence for S’s belief that p includes all (rather than all and only) those things of which S is aware that S is spontaneously inclined to regard as indicative of p’s truth. See Conee’s (2004) discussion of SE. One serious worry about SE (which might explain why Conee doesn’t endorse it) is that it makes justiﬁcation far too easy to come by.

126

EVIDENTIALISM AND SKEPTICISM

believe via memory, perception, testimony, introspection, reasoning, and rational insight—that they are true and known by you.) After all, you have evidence for the Standard View and no outweighing countervidence for external world skepticism; and, as we’ve already noted, justiﬁcation depends on evidence. If things were different—if the skeptic’s thesis or premises and arguments did seem to you to indicate the truth of skepticism—then perhaps you wouldn’t be justiﬁed in holding the Standard View. But that isn’t how things are. Notice that SE does not say that a belief is justiﬁed for a person only if that person recognizes that SE is true and that her beliefs are justiﬁed according to SE. There is no such requirement. A person’s beliefs are justiﬁed if she’s inclined to regard what she’s aware of as indicating the things she believes. It may be that she is aware of things that seem to indicate to her that SE itself is true. This could make it the case that she’s also justiﬁed in believing SE. But even apart from considering or believing SE, her beliefs via memory, perception, testimony, etc. could be justiﬁed in the way SE says they are. So, according to SE, a person can be justiﬁed in rejecting skepticism and accepting the Standard View even if she has never considered SE. In short, one doesn’t have to assume SE in order to respond to skepticism in the way deemed appropriate by SE. 10

7.2 Trying to avoid the Great Pumpkin Objection Now that we have an idea of how SE responds to external world skepticism, let’s look at how Conee presents and replies to what I have called ‘the Great Pumpkin Objection’ to that response. He begins by addressing the point just made that one needn’t assume SE in order to give the SE response to skepticism: This point about not assuming a doctrine [such as SE] at the outset of epistemological theorizing may be granted. . . . Still, it may be replied, employing a procedure in which belief is counted as initially justiﬁed by evidentialism is as faulty a way to start as assuming a doctrine. This is ‘rule circularity’. It is mere self-afﬁrmation. What the employed procedure counts as justifying turns out to ‘justify’ the doctrine that it is justiﬁying. But this is merely ‘justiﬁcation’ by its own standards. In light of this, employing a procedure that is sound according to evidentialism is no better epistemically than employing tarot card reading and counter-induction on their own behalf. (Conee 2004: 19)

Conee also uses (2004: 20) the example of conjecturalism—the view that ‘intentionally guessing that a proposition is true is a good reason to believe that it is true’—and notes that it too might justify itself just as SE does. The point of this complaint is that the SE response to external world skepticism can be mimicked by those defending silly views, such as tarot card reading or counter-induction or conjecturalism, against skepticism about those views. It’s true that in some possible worlds, these silly views aren’t

10 The SE response to skepticism, as described in the last two paragraphs, is presented by Conee (2004: 16–19).

E V I D E N T I A L I S M A N D T H E G R E AT P U M P K I N O B J E C T I O N

127

self-afﬁrming. But the point is that there are possible worlds in which they are selfafﬁrming, and in those worlds, these sorts of defenses of such views would be exactly parallel to the response offered by SE. And, the complaint continues, given that these responses to skepticism about tarot card reading or counter-induction or conjecturalism are so obviously inadequate—even in worlds where the views are self-afﬁrming— the same goes for the SE response to skepticism about the external world. What response does Conee give to the Great Pumpkin Objection applied in this way to SE’s response to skepticism? In his initial response, he points out (2004: 19) that there is a ‘vital difference’ between SE and these silly views: SE is not defended by pointing out that it is self-afﬁrming; instead it is defended by evidence or good reason that seems to indicate that SE is true. And he claims that this is a good reason to accept SE. But, he goes on, ‘there is no good reason to think that a tarot card reading, or a counter-inductive inference, provides good reason for its outcome’ (2004: 19). I ﬁnd this initial response odd. Surely a supporter of tarot card reading (in a world where tarot card reading is self-afﬁrming) could count the tarot cards as providing good reasons just as the proponent of SE counts the fact that something seems indicative of p as a good reason for p. Of course, the proponent of SE could defend her treatment of ‘X seems to indicate p’ as a good reason for p by noting that ‘X seems to indicate p’ seems to her to be a good reason for p. But the tarot card reader could likewise defend her treatment of a tarot card reading that p as a good reason for p: she could say that she has a tarot card reading that a tarot card reading that p is a good reason for p. Later, in discussing conjecturalism, Conee concedes what I’ve just said against his initial response: Conjecturalists too can claim that they have a solution . . . that begs no questions, and is not merely self-afﬁrming. Conjecturalists can claim that some propositions about what we know are merely guessed, and thereby justiﬁed and known, simply by being guessed on purpose. . . . A conjecturalist can further claim that although conjecturalism is self-afﬁrming—he himself has guessed its truth— the resulting epistemic support arises not from the sheer fact of self-afﬁrmation, but rather from the good reason that guessing provides. (Conee 2004: 20)

So now we have a silly view mimicking not only SE’s response to external world skepticism but also the initial response Conee gave to the Great Pumpkin Objection. What does Conee have to say at this point in response to this further development of the Great Pumpkin Objection? The remaining problem with conjecturalism, according to Conee (2004: 20), is that ‘we have ample reason to doubt that intentional guesses can provide epistemic reasons’ and ‘intuitively, guesses are not epistemic reasons, no matter what’. He then summarizes his response to the Great Pumpkin Objection as follows: The crucial difference is that [seeming] evidentialism is supported by evidence, and evidence does give good reasons. Again, this is not mere proclamation, easily matched by any other approach. It is supported by reasons, according to our best understanding of reasons. (Conee 2004: 20)

128

EVIDENTIALISM AND SKEPTICISM

This is how Conee concludes his response to the Great Pumpkin Objection. He doesn’t consider any further responses that might be offered by proponents of the silly views. This leaves the impression that Conee thinks his account of the ‘crucial difference’ between the silly views and SE shows that the attempted mimicry fails.

7.3 Failing to avoid the Great Pumpkin Objection Proponents of the Great Pumpkin Objection to SE say that a silly view like conjecturalism can offer the same responses to skepticism about itself (and what is believed on the basis of intentional guessing) as SE makes to skepticism about itself and the Standard View. In the previous section we saw that Conee replies with the following four points: (i) We have ample reason to doubt that intentional guesses provide epistemic reasons. (ii) Intuitively, guesses are not epistemic reasons, no matter what. (iii) SE is supported by evidence and evidence gives good reasons, according to our best understanding of reasons. (iv) Responses (i)–(iii) are not mere proclamation easily matched by another approach. I assume that when Conee appeals to ‘our best understanding of reasons,’ he is not merely pointing out that we don’t know any serious defenders of conjecturalism. The problem remains that it’s possible for there to be serious defenders of conjecturalism and it’s possible for conjecturalism to be self-afﬁrming in all the ways it needs to be. Yet Conee thinks it would still be a silly view. Likewise, I assume Conee isn’t pointing out that conjecturalism isn’t in fact self-afﬁrming. For, once again, it’s possible for it to be self-afﬁrming and that’s all the Great Pumpkin Objection needs. But then what does Conee have in mind? It’s clear I think that the conjecturalist (in a world where conjecturalism is self-afﬁrming) could easily make claims parallel to (i)–(iii) from the previous paragraph, the claims Conee identiﬁes in support of SE. Point (iii) says that ‘evidence gives good reasons, according to our best understanding of reasons’. This could be understood in at least two ways: (a) ‘Having good evidence for p’ is pretty much equivalent in meaning to ‘having a good reason for p’. (b) Evidence for p of the sort SE says there is (i.e. something’s seeming to indicate p’s truth) gives good reasons for p according to our best understanding of reasons. I doubt that (a) is the claim Conee had in mind in stating point (iii), especially since conjecturalists could happily accept (a). I think the best way (in this context) to understand (iii) is to read it as claiming (b), which is a substantial thesis about good reasons—a thesis that is distinctive of SE. Conjecturalists, of course, have their own substantial thesis about good reasons:

E V I D E N T I A L I S M A N D T H E G R E AT P U M P K I N O B J E C T I O N

129

(b*) Intentional guessing gives good reasons according to our best understanding of reasons.11 Thus, conjecturalists could easily endorse the following claim in place of (iii): (iii*) Conjecturalism is supported by intentional guessing and intentional guessing gives good reasons, according to our best understanding of reasons. As for point (ii), it says that ‘intuitively’ things are thus and so. That’s like saying ‘it seems that’ things are thus and so. It is an appeal to the sort of thing SE says are good reasons: namely, how things seem (intuitively). But conjecturalism could easily appeal to what it says are good reasons—that is intentional guesses—in place of intuitive seemings. Thus, instead of (ii) the conjecturalist could endorse the ﬁrst or even both of the following two claims: (ii*) Intentional guessing tells us that intentional guesses are epistemic reasons. (ii**) Intentional guessing tells us that something’s seeming to indicate p is not an epistemic reason for p, no matter what.12 And just as point (i) follows naturally from points (ii) and (iii),13 so also the conjunction of (ii*) and (iii*) and the conjunction of (ii**) and (iii*) imply the following two points, respectively: (i*) We have ample reason to think that intentional guesses provide epistemic reasons. (i**) We have ample reason to doubt that something’s seeming to indicate p provides an epistemic reason for p. In short, there’s no obstacle to a conjecturalist claiming (i*)–(iii*) in support of conjecturalism—or even (i**), (ii**), and (iii*) which are critical of SE—in place of Conee’s (i)–(iii) offered on behalf of SE. But if it’s possible for a conjecturalist to easily claim (i*)–(iii*)—or (i**), (ii**), and (iii*)—in place of (i)–(iii), then Conee’s claim (iv) is false. For in that case, a silly alternative view (i.e. conjecturalism) could—contrary to (iv)—easily match proclamations (i)–(iii) made on behalf of SE.14 In fact, it looks like conjecturalism could (in a 11

Notice that conjecturalists could easily hold (b*) in conjunction with (a). The conjecturalists I have in mind would endorse (ii**) only if intentional guessing did in fact tell us that something’s seeming to indicate p is not an epistemic reason for p, no matter what. And they would endorse (ii*) only if intentional guessing did in fact tell us that guesses are good reasons. In some worlds, intentional guessing does tell us these things. 13 How does point (i) follow naturally from points (ii) and (iii)? (iii) says, inter alia, that evidence for p (of the sort SE says there is—namely, p’s seeming to be true) gives good reasons for p (here I’m interpreting (iii) in accordance with (b) above). (ii) says that it seems (intuitively) that guesses are not good reasons, no matter what. From this it follows that we have good reason for thinking that guesses aren’t good reasons—and so good reason for doubting that guesses provide good reasons, which is what (i) says. 14 (i) is supportive of SE and (ii)–(iii) endorse criticisms of conjecturalism; all three claims are SE-acceptable (i.e. acceptable from the perspective of SE supporters). One way of offering conjecturalist claims parallel to 12

130

EVIDENTIALISM AND SKEPTICISM

world where it is self-afﬁrming) easily mimic all of the claims that SE makes in responding to skepticism. But if this is so, then Conee’s way of avoiding the Great Pumpkin Objection fails. Conee wanted to respond to that objection by showing that SE’s response to skepticism cannot be easily mimicked by defenders of a silly view like conjecturalism. Unfortunately for Conee, the Great Pumpkin Objection is not so easily dismissed.

7.4 A better response to the Great Pumpkin Objection Do the points made in Section 7.3 imply that SE (and perhaps evidentialism more generally) has no adequate response to the Great Pumpkin Objection? No. For there’s another sort of response available to the proponent of SE besides Conee’s unsuccessfulmimicry response, namely, the successful-mimicry response. The supporter of SE can acknowledge what Conee mistakenly denies—that those defending silly views against skeptical objections can mimic exactly what SE says in response to skepticism about itself and the Standard View. And she can go on to argue that this acknowledged fact doesn’t imply that the SE response to skepticism is inadequate. As I’m thinking of the successful-mimicry response, the defender of SE can reply to the Great Pumpkin Objection as follows: ‘As already noted, my beliefs in the claims endorsed by the Standard View are justiﬁed in virtue of being based on things of which I’m aware that seem to indicate their truth. Likewise for my belief in SE itself: it too is justiﬁed in virtue of being based on what seems to indicate its truth. Moreover, many others who have never heard of SE also have justiﬁed beliefs in claims endorsed by the Standard View. Their beliefs in those claims are justiﬁed in the same way my beliefs in those claims are justiﬁed: their beliefs are based on what seems to indicate the truth of their beliefs. Justiﬁcation for beliefs depends on their being based on what seems to indicate their truth and this is so even if one doesn’t recognize that it is so. Silly views (like conjecturalism) can be self-afﬁrming in the way SE is self-afﬁrming. But silly views are wrong about what justiﬁcation supervenes on (I believe this on the basis of what seems to me to indicate that these silly views are wrong in this way). It’s true that silly views can mimic everything I’ve just said in defending themselves against skepticism about their views. And this means that I can’t show them, in a way that satisﬁes them, that they’re mistaken and I’m right. But that needn’t lead me to think SE is false or unjustiﬁed. SE is supported by the right sort of evidence after all (i.e. what seems to indicate its truth). Nor need it lead me to think conjecturalists are justiﬁed in their beliefs. Insofar as their beliefs are based only on intentional guesses and not on what seems to indicate the truth of their beliefs, they aren’t justiﬁed. And this is so even

(i)–(iii) is to do what (i*)–(iii*) do: offer conjecturalist-acceptable support for conjecturalism and for rejecting criticisms of conjecturalism. Another way of offering conjecturalist claims parallel to (i)–(iii) is to do what is done in (i**), (ii**), and (iii*): offer conjecturalist-acceptable support for conjectualism and endorse criticisms of SE.

E V I D E N T I A L I S M A N D T H E G R E AT P U M P K I N O B J E C T I O N

131

though I can’t show them, in a way they’ll ﬁnd convincing, that this is so—even though they can say things on behalf of their views that mimic exactly what I say on behalf of mine. It’s not what we can say on behalf of our beliefs using the methods and skills associated with philosophy that makes them justiﬁed. It’s whether they are based on what seems to indicate their truth.’ That, I believe, is a much better response to the Great Pumpkin Objection because it admits what’s true—namely, that silly views can successfully mimic SE’s response to skepticism. It won’t satisfy all who are skeptical of SE. And it doesn’t show supporters of silly views that they can’t say on behalf of their views things parallel to what the supporter of SE says on behalf of her view. But, as I’ve argued elsewhere, comments of this sort apply to all nonskeptical positions that aren’t ruled out as implausible on other grounds.15 And that’s a reason for thinking that it’s not a ﬂaw in one’s position for such comments to apply to it.

7.5 A worry for evidentialism Unfortunately, the successful-mimicry response to the Great Pumpkin Objection laid out in the previous section does not ﬁt well with the evidentialist’s sympathies toward internalism and toward certain objections to externalism. To see this, it will be helpful to compare SE’s successful-mimicry response to the Great Pumpkin Objection with an externalist response to the Great Pumpkin Objection. SE says that a belief is justiﬁed if it satisﬁes the SE-condition—that is being based on what seems to indicate the belief ’s truth. An externalist view such as a proper function view says that a belief is justiﬁed if it satisﬁes the PF-condition—that is being formed in accord with proper function.16 Consider now how the SE reply to the Great Pumpkin Objection compares with the proper functionalist reply to that objection. Each will endorse the following ﬁve claims: 1. Beliefs are justiﬁed in virtue of satisfying the relevant justiﬁcation-conferring condition (according to SE, this will be the SE-condition; according to proper functionalism, this will be the PF-condition). There’s no need for the believer to know what the relevant condition is or that it’s satisﬁed. It’s enough for justiﬁcation that it is satisﬁed by the subject’s beliefs. 2. Belief in an epistemological view (be it SE or proper functionalism) is justiﬁed only if that belief itself satisﬁes the relevant condition (the SE-condition according to SE, the PF-condition according to proper functionalism). 3. Silly views can be self-afﬁrming just as SE or proper functionalism can be self-afﬁrming. But silly views aren’t justiﬁed if they don’t satisfy the relevant

15

See Bergmann (2006, 2008). I defend a proper function analysis of justiﬁcation in Bergmann (2004, 2006: ch. 5). The view I defend is a little more complicated than the view described here but this view will sufﬁce for illustrative purposes. 16

132

EVIDENTIALISM AND SKEPTICISM

condition (the SE-condition according to SE, the PF-condition according to proper functionalism). 4. Silly views like conjecturalism can mimic SE or proper functionalism, saying things parallel to each of the above remarks in its own defense, which means conjecturalists can’t be shown, in a way that satisﬁes them, that they’re mistaken. 5. Despite this, conjecturalism isn’t a reasonable view in the way SE or proper functionalism is, for although conjecturalism is self-afﬁrming, it doesn’t satisfy the relevant justiﬁcation-conferring conditions (the SE-condition according to SE or the PF-condition according to proper functionalism). Notice that the key point made here by both SE and the proper functionalist in response to the Great Pumpkin Objection is that the silly view (conjecturalism) may be as selfafﬁrming and as plausible from its proponents’ own perspective as the correct view (whether that is SE or proper functionalism), but the silly view is not as reasonable as the correct view because it doesn’t satisfy the condition that is in fact justiﬁcation-conferring (the SE-condition according to SE or the PF-condition according to proper functionalists). In short, the reasonability of your view depends not on whether it is, from your own perspective, plausible or conﬁrmed; rather, it depends on whether your view satisﬁes the conditions that are in fact justiﬁcation-conferring. But when stated in that way, this sort of response to the Great Pumpkin Objection (whether offered by SE or by the proper functionalist) looks a lot like an externalist response. At the very least we can say this: it has precisely those features that make internalists unhappy with the externalist response to the Great Pumpkin Objection. It focuses on the actual epistemic facts of the matter—what justiﬁcation in fact supervenes on and whether that supervenience base is in fact present—regardless of whether the believer’s view is, from her own perspective, plausible and conﬁrmed, regardless of what the believer thinks the epistemic facts are. Should an evidentialist be worried by this? Maybe not. Maybe an evidentialist can avoid both internalism and externalism.17 Or maybe an evidentialist needn’t buy into the internalist’s dissatisfaction with the externalist response to the Great Pumpkin Objection. I suspect though that, for the reasons mentioned at the end of the previous paragraph, evidentialists won’t be content with the successful-mimicry response to the Great Pumpkin Objection laid out in Section 7.4. But this should give them a reason to worry about evidentialism itself. For the alternative to the successful-mimicry response sketched in Section 7.4 will be something like Conee’s unsuccessful-mimicry response described in Section 7.2, a response that is unsatisfactory for the reasons laid out in Section 7.3. Ultimately then, the Great Pumpkin Objection is more worrisome for evidentialists than it is for externalists. For unlike evidentialists, externalists can at least be comfortable with the successful-mimicry response.18

17 18

See Bergmann (2006: 55–7) for how a view on justiﬁcation can be neither internalist nor externalist. Thanks to Trent Dougherty for comments on an earlier draft.

E V I D E N T I A L I S M A N D T H E G R E AT P U M P K I N O B J E C T I O N

133

References Bergmann, Michael. 2004. ‘Externalist Justiﬁcation Without Reliability,’ Philosophical Issues: Epistemology, 14: 35–60. ——. 2006. Justiﬁcation without Awareness. New York: Oxford University Press. ——. 2008. ‘Externalist Responses to Skepticism’ in John Greco (ed.), Oxford Handbook of Skepticism, pp. 504–32. New York: Oxford University Press. Conee, Earl. 2004. ‘First Things First’ in Conee and Feldman (2004: 11–36). Conee, Earl and Richard Feldman. 2004. Evidentialism: Essays in Epistemology. New York: Oxford University Press. Feldman, Richard and Earl Conee. 1985. ‘Evidentialism,’ Philosophical Studies, 48: 15–34. Plantinga, Alvin. 1981. ‘Is Belief in God Properly Basic?’ Noûs, 15: 41–52. Sosa, Ernest. 1997. ‘Reﬂective Knowledge in the Best Circles,’ The Journal of Philosophy, 94: 410–30.

This page intentionally left blank

PART IV

‘Knowledge First’ For Conee and Feldman as for most evidentialists, the notion of evidence is an anchor concept. It anchors the concept of epistemic justiﬁcation which anchors the concept of knowledge. Timothy Williamson has inaugurated a counter-revolution under the banner ‘Knowledge First.’ The ﬁrst chapter in this section, by Keith DeRose, questions evidentialism in part with this challenging question: Why should evidence ground the deﬁnition of ‘epistemically ought’ rather than knowledge? He offers cases similar to ones from Axtell and Baer in the previous section (indeed, similar to those of Kornblith’s challenge in ‘Evidentialism’ in the ﬁrst place) where it seems to him that—at least in a permissible sense, if not a preferable sense—we ought not say that S ought to believe p even though p ﬁts S’s evidence, because were S to believe p she would fail to know p. Timothy Williamson describes an example which he takes to establish the possibility of knowledge in the presence of the extreme improbability to one that one knows. Prima facie, it seems such an example doesn’t sit well with evidentialism, for how could one’s (total) evidence support knowledge level justiﬁcation for p, yet at the same time support the same level of justiﬁcation for the proposition that you don’t know. John Greco’s essay at the beginning of the next part can be an honorary member of this part, as it also argues that knowledge is a better epistemic starting point than justiﬁcation.

This page intentionally left blank

8 Questioning Evidentialism Keith DeRose

My focus will be on Richard Feldman’s evidentialist thesis: O2. For any person S, time t, and proposition p, if S has any doxastic attitude at all toward p at t, then S epistemically ought to have the attitude toward p supported by S’s evidence at t. Specifying what attitudes one epistemically ought to have in various cases, Feldman explains: If a person is going to adopt any attitude toward a proposition, then that person ought to believe it if his current evidence supports it, disbelieve it if his current evidence is against it, and suspend judgment about it if his evidence is neutral (or close to neutral). (Feldman 2000: 679; Conee and Feldman 2004: 178)1

I will propose some potential counter-examples to test this evidentialist thesis. My main intention in presenting the ‘counter-examples’ is to better understand Feldman’s evidentialism, and evidentialism in general. How are we to understand what our evidence is, how it works, and how are we to understand the phrase ‘epistemically ought to believe’ such that evidentialism might make sense as a plausible thesis in light of the examples? Of course, we may decide that there’s no such way to understand evidentialism—that it just isn’t a plausible thesis. I must admit that my suspicions lean in that direction. But the potential counter-examples are put forward, not in a refutational spirit (though I have nothing against good refutations in philosophy), but as an invitation to evidentialists and potential evidentialists to reﬁne and/or explain their thesis in light of the at least apparent problems that the examples highlight. Though O2 is a thesis concerning what we epistemically ought to believe, similar issues arise about evidentialism concerning what subjects are epistemically justiﬁed in believing, so the worries about evidentialism I raise here may have more general application. 1 Feldman’s formulation of O2 mistakenly contains these words at the end of O2’s antecedent: ‘and S’s evidence at t supports p.’ Feldman conﬁrms (p.c.) that these words should be omitted.

138

‘ K N OW L E D G E

FIRST’

8.1 ‘Epistemic ought’ The ﬁrst example is mainly intended to point out some possible obscurity in the phrase ‘epistemically ought’ that plays such a key role in Feldman’s evidentialist thesis. The example probably needn’t be described in great detail to bring the problem to light. Suppose that Henry ﬁrmly believes that p—it doesn’t matter much what p is—and has excellent evidence for p (evidence that’s strong enough to adequately support the ﬁrm and conﬁdent attitude Henry has adopted toward p). Suppose further that Henry doesn’t possess evidence against p, so the attitude toward p that ﬁts all the evidence Henry possesses is the conﬁdent belief that p which Henry in fact holds. But suppose that Henry doesn’t believe p on the basis of the excellent evidence for it that he possesses. Indeed, Henry hasn’t even considered p in the light of this excellent evidence, and the fact that he possesses good evidence for p is no part of the explanation for why Henry believes that p. Rather, Henry believes that p on the basis of some other beliefs of his that he considers to be good evidence for p, but which in fact constitute absolutely lousy evidence for p. Henry has no business believing p at all on the basis of the evidence he is in fact using, much less believing it as ﬁrmly as he does. If you want some details—maybe your intuitions are helped out by ﬁlling in the story a bit— suppose that p is a proposition to the effect that someone that Henry hates very much has done some horrible thing. It’s Henry’s hatred and resulting desire to believe his opponent has done this horrible thing that causes him to think of his lousy evidence as strong evidence and to thereby believe that p. Well, what’s your intuition as to what Henry epistemically ought and ought not to believe and about how ﬁrmly Henry ought or ought not to believe? Mine is that Henry ought not to believe p at all, much less believe it to the degree that he does. If this is correct, then Feldman’s evidentialist thesis, at least as I understand it, gets this case wrong. It says that Henry epistemically ought to have the attitude (belief, disbelief, suspension of judgment; or degree of belief) that ﬁts the evidence he has. And we’re supposing that a ﬁrm belief that p is the attitude that ﬁts the evidence Henry has. I suspect that, presented with the question of what Henry epistemically ought to believe, most will either react as I do, or else will respond by (very wisely) saying something along the lines of, ‘Well, it all depends on what you mean by “epistemically ought”.’ There certainly is a very good sense of ‘ought’ in which Henry ought not to believe that p. This is the sense that my intuition responded to. And this sense I think has a good claim to the title of ‘epistemic’. But there’s also some sense, I suppose, in which Henry ought to believe that p. To the question, ‘In the matter of p vs not-p, what would Henry believe if he were to take all the evidence he possesses and process it correctly?’, the answer is, ‘p’. And I suppose there’s some admissible way of understanding the question, ‘What ought Henry to believe?’, such that it’s asking roughly the same thing, and should get the same answer. After all, ‘ought’ is a very semantically ﬂexible word.

QUESTIONING EVIDENTIALISM

139

Now, given that there is a good sense in which Henry ought not to believe that p and a good sense in which he ought to believe that p, how are we to understand and answer the question, ‘What ought Henry to believe?’, when we are explicitly instructed to understand the ‘ought’ as an ‘epistemic ought’? I see a couple of different ways of understanding the qualiﬁer ‘epistemic’ here. First, it can be (however vaguely) directing us to understand the ‘ought’ in the way that’s most tightly connected to the concept of knowledge. It was this understanding that led me to respond that Henry epistemically ought not to believe that p. (Well, it was this understanding plus the fact that the sense of ‘ought’ which yielded a Feldman-friendly answer seemed a bit strained to me.) Here I was led by the intuition that should p turn out to be true, Henry would not know that p, and that his failure to know would be tightly bound up with fact that he was not, in this sense of ‘ought’, believing as he ought. That’s what led me to think that the sense of ‘ought’ in which Henry ought not to believe that p would be the more likely candidate for the title of ‘epistemic’ ought.2 But there’s another way to understand ‘epistemically ought’. As I admitted, there does seem to be some admissible understanding ‘What ought Henry to believe?’ in which the answer is: ‘p’. Perhaps I should simply understand ‘epistemically ought’ as picking out that sense of ‘ought’? But in that case, given my tenuous grasp of that sense of ‘ought’, this comes to little more than: ‘Understand ‘ought’s in such a way as to make O2 come out true.’ And it’s no surprise, then, that on this understanding, O2 does come out true. This problem can be made more acute by adding to our example. Suppose then that there is some evidence Henry very easily could have, and should have, gathered, but that he negligently never encountered. This would have been very strong evidence against p. So strong that, despite the excellent evidence Henry possesses in favor of p, this negative evidence that Henry should have gathered would have completely outweighed the positive evidence he actually possesses, such that disbelief of p would have been the attitude that would have best suited Henry’s evidence, had he gathered this negative evidence. (We want it to remain the case that the evidence Henry actually possesses still favors p, so don’t imagine this case such that Henry has good reason to think that the evidence he neglected to gather would have counted against p. This would raise the suspicion that, whatever these good reasons for thinking the evidence will be unfavorable to p are, they probably already also constitute some evidence against p itself.) This addition reinforces my sense that Henry ought not to believe that p. Not only is Henry in fact basing his belief on lousy evidence (which is still part of our story), but,

2

Of course, the mere fact that Henry doesn’t know even where he meets the condition of believing as he ought, in the sense of ‘ought’ that Feldman needs, does not entail that meeting that condition isn’t necessary for knowledge. But if Henry’s failure to know is due to the fact that he’s not believing as he ought, in some sense of ‘ought’, that gives us some good reason to think that it’s some such sense of ‘ought’—some sense in which Henry is not believing as he ought—that deserves the name ‘epistemic’.

140 ‘ K N OW L E D G E

FIRST’

even if he had brought all the evidence he had to bear on the issue, he still would have ended up not believing in p if he had also been responsible in his gathering of evidence. Now the main reason I react negatively to the question ‘Ought Henry to believe that p?’ is still the old reason that the evidence that Henry is actually basing his belief upon is lousy evidence for p. The new addition to our story is there to show just how special is the sense of ‘ought’ which yields the answer that Feldman wants. I still think there is some admissible sense of ‘ought’ in which Henry ought to believe that p. This sense directs us to consider whether Henry’s belief ﬁts the relevant evidence. When we ask whether the relevant evidence includes evidence that Henry doesn’t possess, but easily could have, and should have, possessed, we’re instructed: ‘No, only look at Henry’s actual evidence; evidence that Henry doesn’t actually have doesn’t count.’ Given that instruction, one might be inclined to suspect that we’ll also consider irrelevant evidence that Henry possesses, but doesn’t use. After all, what credit is it to Henry that he possesses excellent evidence, if he didn’t use it? But, to our surprise, when we ask: ‘Should we consider as relevant evidence Henry possesses, but never uses?’, we’re told: ‘Yes, by all means. It doesn’t matter that he doesn’t use it.’ A counter-example to O2? Well, I think I can still ﬁnd an admissible sense of ‘ought’ in which Henry ought to believe that p, and if that’s the sense that’s supposed to be picked out by the term ‘epistemically ought’, I guess O2 isn’t refuted. Indeed, elsewhere (Feldman and Conee 1985), as I should in fairness point out, Feldman considers problems not unlike the one I’m raising, not in relation to the question of what we ‘epistemically ought’ to believe, but to the closely related question of what we’re ‘epistemically justiﬁed’ in believing. He and his cohort, Earl Conee, respond by registering their sense that what is ‘epistemically’ justiﬁed is just what their evidentialist thesis says is epistemically justiﬁed, and sometimes by locating other concepts, distinct from their ‘epistemic’ ones, that might be leading folks’ intuitions astray. This all left me uneasy. And that’s why I here present this type of problem with a different focus: that of trying to illustrate just how special is the sense of ‘ought’ it appears we must end up with to save O2. I guess the challenge to Feldman is not to defend that there’s some admissible sense of ‘ought’ that’s friendly to his thesis—I’m happy enough to grant that, given just how semantically ﬂexible ‘ought’ seems to me—but to tell us why that very particular sense is epistemically important. Here are three questions: 1. Relative to the evidence he actually brought to bear on the issue, ought Henry to believe that p? 2. Relative to all the evidence that he has, ought Henry to believe that p? 3. Relative to the evidence available to him (the evidence he has plus what evidence he easily could have, and should have had), ought Henry to believe that p? In our story, the answers to questions 1 and 3 seem to be ‘no’, while the answer to 2 is ‘yes’. The general question,

QUESTIONING EVIDENTIALISM

141

G. Ought Henry to believe that p?, it seems, could be asking something along the lines of any of 1–3. Apparently, where the ‘ought’ is an ‘epistemic’ ought, we’re to understand it along the lines of question 2. But what’s so special about that question? Is it in some way that I can’t see especially tightly connected with knowledge and/or other important epistemic concepts? A reason for pessimism on this front, as I’ve already indicated, is our strong sense that Henry doesn’t know in our cases, and the feeling that his failure to know seems tightly tied up with the thought that he’s not believing as he ought in some very relevant sense of ‘ought’—which must not be Feldman’s sense, because Henry is believing as he ought in the sense his thesis requires. Is there any other way of understanding the phrase ‘epistemically ought’—a way on which Feldman’s thesis is plausible, but not trivial? Which gives us a handle on what we’re being asked when we’re asked what someone ‘epistemically ought’ to believe that amounts to more than just ‘understand it so as to make O2 true’, but on which Feldman’s answer seems true, or at least plausible, nonetheless? I see two hints in Feldman’s paper. I don’t think either pans out. First, at places, Feldman writes as if the senses of ‘ought’ in which our Henry ought not to believe that p are long-range, down-the-road senses, whereas the ‘epistemic’ sense in which Henry ought to believe that p, and on which O2 is true, speciﬁes what Henry ought to believe right now. This comes up in Feldman’s answer to the problem of ‘Evidence One Should Have Had’.3 If there’s evidence one could have and should have possessed, this may indicate such long-term conclusions as that one ought to go gather this evidence, or ought to adopt different policies of evidence gathering. These may generate various ‘oughts’ about one’s beliefs, Feldman admits—though he suspects that these will be moral or prudential, and not epistemic, oughts (2000: 689–90; Conee and Feldman 2004: 189). At any rate, Feldman claims that his thesis gives the right answer to the questions, ‘What should I believe in the meantime?’, ‘What should I believe until I have a chance (or the courage) to look at that new evidence?’. He writes, ‘It’s that natural and central question to which evidentialism provides a good answer’ (2000: 688; Conee and Feldman 2004: 188). This would indeed make Feldman’s sense of ‘ought’ quite important. But the other senses of ‘ought’—the senses on which Henry ought not to believe that p, some of which seem tightly connected to the thought that Henry doesn’t know that p even if p is true, and so seem pretty ‘epistemic’ to me—also seem to specify things about the status of Henry’s belief right now. (To make it explicit, in questions 1 and 3 above, add the words ‘right now’ in the appropriate places.) That there is important evidence against p that Henry easily could have, and should have, possessed, while it may indicate various things that Henry should do in the future, it also seems to ground a legitimate, and potentially important, sense in which, because he ought to have already gathered that evidence, Henry right now ought not to believe that 3

See Feldman (2000: section DI, and also D2, 687–90; Conee and Feldman 2004: 186–9).

142 ‘ K N OW L E D G E

FIRST’

p. At any rate, the much more pressing problem I have with Feldman’s O2 involves not the evidence that Henry should have possessed, but the problem that the evidence that Henry actually bases his belief on is in fact lousy evidence for p. This indicates an important sense in which Henry ought not to believe that p which most deﬁnitely indicates a problem with Henry’s believing right now, and not just down the road. Note that the person who thinks it’s question 1 above that’s the ‘epistemically’ important one could try (with about an equal lack of success, to my thinking) to do to Feldman’s question (question 2) what Feldman does to question 3, arguing: ‘To be sure, in some (perhaps prudential or moral) sense, down the road, Henry ought to bring more of his evidence to bear on the issue. It’s this down-the-road issue that question 2 addresses. But there’s still the question of what Henry ought to believe in the meantime, what he should believe until he has the chance or the courage to consider p in light of the evidence he hasn’t yet considered in relation to the issue. It’s question 1 that addresses the important issue of what Henry ought to believe right now.’ This would be unfair to question 2, because though question 2 may well be relevant to issues about what Henry ought to do down the road (like bring more evidence to bear on the issue), it is also tied to a perfectly good sense in which Henry right now ought to believe that p. But question 3 can likewise be defended, against Feldman, as specifying a sense in which right now Henry ought not to believe that p. The second hint we get comes from Feldman’s account of ‘epistemic ought’ being tied to a certain important role we play—the role of a believer. Feldman writes: [F]orming beliefs is something people do. That is, we form beliefs in response to our experiences of the world. Anyone engaged in this activity ought to do it right. In my view, what they ought to do is follow their evidence (rather than wishes or fears). I suggest that epistemic oughts are of this sort—they describe the right way to play a certain role. . . . They are based on what’s good performance. (2000: 676; Conee and Feldman 2004: 175)

This perhaps can help give us another handle on what’s intended by the qualiﬁer ‘epistemic’ in the title ‘epistemic ought’: To believe what you ought to believe (and to the degree that you ought to), in this sense, is to perform well in ﬁlling the role of a believer. If Feldman’s thesis were correct about the ‘ought’ that speciﬁed how we are to ﬁll that important role of ours, it would certainly be a very important result. But if this is what Feldman means, then his evidentialist thesis is surely false. For Henry’s performance is poor, in almost every way I can think of. He’s doing it (believing), but not doing it right. He’s doing a lousy job of gathering evidence, and a lousy job of processing the evidence he does have. His belief seems irrational and unjustiﬁed, and even if it happens to be true, it certainly isn’t a piece of knowledge. He’s not doing a good job of ﬁlling the role of a believer. Yet, at least as I understand it, he satisﬁes Feldman’s evidentialist thesis for believing what he ought. Perhaps Feldman would say that although Henry did perform poorly, he still ended up believing what he would have believed if he had performed well in ﬁlling the role of a believer, and is in that way believing what he ought to. But this is true only on a very

QUESTIONING EVIDENTIALISM

143

specialized understanding of what it is to ﬁll the role of a believer well. To ‘do it right’, to do a good job of ﬁlling the role of a believer, or of a reasonable believer, it seems one should both gather and process evidence well. And if Henry had both gathered and processed evidence reasonably, he would not believe that p. So Henry did not perform well in ﬁlling the role of a believer, nor did he end up believing what he would have believed if he had performed well in that role, despite the fact that he satisﬁes Feldman’s account of believing what he ought. It’s not the (more general) role of a believer that Feldman’s account seems to answer to, but the (more specialized) role of an evidence-processor. Henry also performed poorly in ﬁlling that more specialized role. (Recall that the evidence Henry actually used didn’t support his belief, and it was only through bad processing of the evidence that he used that Henry ended up believing that p.) But Henry did end up believing what he would have believed if he had taken all the evidence he had and processed it well. Perhaps it’s in that very specialized sense that he’s believing what he ‘ought’. But if what we mean by believing what we ‘epistemically’ ought to believe is the specialized notion of believing what we would have believed if we had taken all the evidence we actually have (no matter how unreasonable we were in getting just that evidence) and then processed it well, we seem to be back to just understanding ‘ought’ so as to make O2 come out true, and then it’s no surprise that O2 gets that very specialized notion right.

8.2 Evidence and skeptical hypotheses We meet another type of at least apparent counter-example to evidentialism when we face these questions: Ought we to believe that skeptical hypotheses are false? And how ﬁrmly ought we to believe this? For instance, ought I to believe that I’m not a bodiless brain-in-a-vat who’s been electrochemically stimulated to have precisely those sensory experiences that I’ve had? I think I ought to believe this hypothesis is false, and to so believe very ﬁrmly and conﬁdently indeed. One who really didn’t ﬁrmly believe such an hypothesis to be false of himself would seem to me to be epistemically defective, perhaps even deranged, and is certainly not believing as he ought. (We may be able to cook up some very special circumstances in which this wouldn’t be true of someone— circumstances in which some believer ought to be agnostic about whether he’s a BIV, or perhaps even ought to believe that he is a BIV. But these would have to be fairly weird circumstances, I think.) Now, in the matter of whether or not I’m a BIV, what attitude best ﬁts my evidence? My initial inclination is to say that I really don’t have any effective evidence one way or the other. If this inclination is right, Feldman’s evidentialist thesis would again be in trouble.4

4 Feldman and Conee have some interesting evidentialist thoughts about skepticism in their ‘Making Sense of Skepticism,’ in Conee and Feldman (2004: 277–306). Their focus, however, is on arguing that their form of evidentialism can make sense of the threat that skepticism at least seems to pose for our ordinary external world beliefs, and not primarily on the issue I’m raising here: that our beliefs that skeptical hypotheses are false seem to

144 ‘ K N OW L E D G E

FIRST’

But this is only an initial inclination, and there’s plenty of room for maneuvering here for the evidentialist—mostly having to do with what our evidence is and how it works to justify beliefs. So this type of example is put forward mainly with the hope of extracting from the evidentialist some help in understanding such matters. One possible move for the evidentialist to make is to claim that I know such things as that I have hands, which provide for me excellent evidence for belief that I’m not a BIV, since BIVs don’t have hands. After all, in other settings, it can often seem that I can cite the fact that I have hands as evidence for various conclusions. And if the fact that I have hands is evidence that I possess, what’s to prevent me from citing it as evidence for the conclusion that I’m not a BIV? The problem here is that it seems—or at least seems to me—that I don’t have one simple body of evidence that constitutes what evidence I have. Rather, I have evidence for various conclusions, and what my evidence is seems to depend on what issue is being discussed. Thus, suppose that the Cubs beat the Braves yesterday, and that I heard on the radio that ‘The Cubs beat the Braves’, in a list of results that were read during the sports segment of the news, and that that radio report is my only source of information for that result. Here, that the Cubs beat the Braves is evidence I have, or at least seem to have, and can properly cite, if the issue being discussed is whether the Cubs have beaten any good teams this year. If, for instance, you claimed that the Cubs haven’t beaten any good teams this year, I seem to be in a position to argue, ‘Well, they beat the Braves yesterday. And the Braves are a good team.’ However, suppose that you got it into your head to instead discuss the issue of whether the sports segment I heard on the radio was accurate in the scores it reported. That the Cubs beat the Braves seems not to be evidence I have, or at least seems not to be evidence I can properly use, for the purposes of discussing that issue. If I were to argue, ‘Well, the report certainly was accurate, because, according to it, the Cubs beat the Braves, and they did, and it was reported that the Marlins beat the Cardinals, and they did . . . ’, you wouldn’t be very impressed if you knew that in each case, my only source for the ‘and they did’ part was the very radio report whose accuracy was presently in question! That the Cubs beat the Braves, and the like, seems not to be evidence I have, or at least not to be evidence I’m allowed to use, in these circumstances (I have no other sources of information for the results of yesterday’s games), where the issue is whether the radio report I heard was accurate. This evidence, which I seem to have when it’s another be beliefs we ought to hold to an extent that far exceeds the extent that these beliefs are supported by our evidence. Though their focus (at least in the above-mentioned paper) does not match our current focus, Feldman and Conee do make it clear that they think our beliefs to the effect that skeptical hypotheses are false are justiﬁed enough to constitute knowledge (see esp. 2004: 300), and that these are beliefs that we epistemically ought to ﬁrmly hold, because these beliefs are strongly enough supported by our evidence to make such a ﬁrm belief be the attitude that ﬁts our evidence. But if you, like me, ﬁnd it prima facie implausible to suppose these beliefs are supported by our evidence to the extent needed, Feldman and Conee don’t offer much help in overcoming these doubts: While it’s clear that they accept something you ﬁnd implausible, they don’t offer much by way of why you should not ﬁnd it so implausible—in large part, no doubt, because this issue is not the focus of their paper.

QUESTIONING EVIDENTIALISM

145

issue that’s in question, seems to be undermined for the purpose of establishing that the radio report was accurate. In a similar way, that I have hands seems to be undermined as a piece of evidence if the issue under discussion is whether or not I’m a BIV. If so, what evidence do I have for my belief that I’m not a BIV? Again, my inclination is to say that I have no effective evidence for that belief. That I’m not a BIV seems to be a belief that I ought to hold, and ought to hold very ﬁrmly, despite the fact that the opposite position—that I am a BIV—seems to ﬁt my evidence—at least what evidence it is proper to use in discussing this issue—every bit as well as does my actual belief. Perhaps the evidentialist can argue, following Jonathan Vogel (1990), that though our evidence is severely limited in the context of discussing whether skeptical hypotheses obtain—limited to something like sensory appearances—still this highly limited evidence does, after all, support the ‘real world hypothesis’ over the various skeptical hypotheses, perhaps because the real world hypothesis somehow provides a better explanation for the sensory appearances. I have my doubts about whether the real world hypothesis really does ﬁt our evidence better than do the various skeptical hypotheses, but it certainly seems to me that even if the RWH does eke out a narrow victory here, it won’t ﬁt the evidence better by a wide enough margin to justify the degree to which I think rationality not only allows but requires us to believe the RWH over the various skeptical hypotheses. Well, that’s how it seems to me. Presumably matters seem different to the evidentialist. My hope is that the evidentialist will explain what our evidence is and how it works in such a way as to make their evidentialism plausible in the face of these examples. Since these issues involving evidence are, or should be, important to all epistemologists, whether they’re evidentialists or not, this would be a great service. My suspicion, however, is that the best understanding of the concept of evidence and how it relates to other important epistemic concepts will be a non-evidentialist understanding according to which such things as that I’m not a BIV are things I know, ought to believe (and ought to believe very ﬁrmly), and am justiﬁed in believing, despite the fact that they don’t ﬁt my evidence any better than (or, at least, don’t ﬁt it much better than) do their opposites, and an understanding according to which the degree to which I ought to believe the Real World Hypothesis, as opposed to its various skeptical competitors far exceeds the degree to which it ﬁts my evidence better than do those competitors. Indeed, I’ve long suspected (following the insightful lead of Thomas Reid) that the poor unfortunate who really did believe the RWH over its rivals only to the extent that it ﬁt his evidence better than its rivals would probably serve well as a great (and wonderfully instructive) example of someone who is not believing as he ought.5

5 This paper is a revision and updating of my ‘Ought We to Follow Our Evidence?,’ Philosophy and Phenomenological Research 60 (2000): 697–706.

146 ‘ K N OW L E D G E

FIRST’

References Conee, Earl and Richard Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Oxford University Press. Feldman, Richard. 2000. ‘The Ethics of Belief,’ Philosophy and Phenomenological Research, 60: 667–95; reprinted in Earl Conee and Richard Feldman. 2004. Evidentialism: Essays in Epistemology, pp. 166–95. Oxford: Oxford University Press. —— and Earl Conee. 1985. ‘Evidentialism,’ Philosophical Studies, 48: 15–34; reprinted, with an added ‘Afterword,’ in Earl Conee and Richard Feldman. 2004. Evidentialism: Essays in Epistemology, pp. 83–107. Oxford: Oxford University Press. Vogel, Jonathan. 1990. ‘Cartesian Skepticism and Inference to the Best Explanation,’ Journal of Philosophy, 87, 11: 658–66.

9 Improbable Knowing Timothy Williamson

How much can we turn the screw on counter-examples to the KK principle? The principle, also sometimes called ‘positive introspection’, says that if one knows that p, one knows that one knows that p. It is widely, although not universally, acknowledged that the KK principle is false, and not just for the boring reason that one can know that p without having formed the belief that one knows that p. One can know that p, and believe that one knows that p, without knowing that one knows that p, because one is not in a strong enough epistemic position to know that one knows that p (Williamson 2000: 114–30). One’s epistemic position is weaker with respect to the fact that one knows that p than it is with respect to the fact that p. But how much weaker can one’s epistemic position be with respect to the fact that one knows that p than it is with respect to the fact that one knows that p? Of course, if one does know that p, one’s epistemic position with respect to the fact that one knows that p cannot be so bad that one positively knows that one does not know that p, as opposed to merely failing to know that one knows that p, for knowing is factive: one can know only what is indeed the case. But we can ask a more graded question in terms of evidential probabilities. Given that one knows that p, how low can the probability, on one’s own present evidence, that one knows that p sink? It will be argued here that it can sink arbitrarily close to 0. At the limit, the probability on one’s evidence that p can be 1 while the probability on one’s evidence that one knows that p is 0. The difference between the probabilities can be as large as probabilistic differences can go. One’s evidence can be radically misleading about one’s own present epistemic position. If the rationality of an action depends on one’s epistemic position, then one’s evidence can be radically misleading about the rationality of the various actions available to one. Such phenomena will be used to cast light on some epistemological puzzles.

9.1 It is useful to explore the issues within a framework taken from the standard possible worlds semantics for epistemic logic, introduced by Hintikka (1962). Such a formal

148 ‘ K N OW L E D G E

FIRST’

framework keeps us honest, by making it straightforward to check whether our descriptions of examples are consistent and what their consequences are, and by facilitating the identiﬁcation of structurally appropriate models. Of course, we must also consider whether the mathematical models we use are realistic on their intended epistemic interpretation in the relevant respects. It will be argued below that the respects in which they are idealized are consistent with the uses to which the models are here being put. We recall some basic features of possible worlds models for epistemic logic. For present purposes, we can make two convenient simpliﬁcations. First, we need only consider one agent at a time. Although important analogues of the failure of the KK principle arise for the interpersonal case too (Williamson 2000: 131–4), we can ignore such complications here. Second, we can ignore the strictly semantic aspect of possible world models, by discussing propositions rather than sentences; the resultant structures are frames. Given those simpliﬁcations, a frame is just an ordered pair <W, R>, where W is a set and R a binary relation on W (a set of ordered pairs of members of W ). Informally, we think of W as the set of relevant worlds, and R as a relation of epistemic possibility between them, where a world x is epistemically possible in a world w (Rwx) if and only if, for all one knows in w, one is in x, in other words, whatever one knows in w is true in x (where ‘one’ refers to the relevant agent and the present tense to the relevant time). Correspondingly, we think of the subsets of W as propositions; a proposition p W is true in a world w if and only if w 2 p. Obviously the conjunction of two propositions is their set-theoretic intersection, the negation of a proposition is its complement in W, and so on. For p W , we deﬁne: Kp ¼ fw 2 W: 8x 2 W ðRwx ! x 2 pÞg Informally, Kp is to be the proposition that one knows p. On its intended reading, the deﬁnition of K presupposes that one knows something if and only if it is true in all epistemic possibilities for one, that is, in all worlds in which whatever one knows is true. This involves assumptions about the agent’s logical omniscience, in two ways. First, any treatment of K as a function from propositions to propositions automatically requires that if p is the same proposition as q then Kp is the same proposition as Kq. Since propositions are being treated as sets of worlds, this means that if p is true in the same worlds as q, then Kp is true in the same worlds as Kp. Moreover, since truth in a world respects the usual truth-functions—a conjunction is true in a world if and only if all its conjuncts are true in that world, the negation of a proposition is true in a world if and only if the original proposition is not true in that world, and so on—truth-functionally equivalent propositions are identical, so both are known or neither is; in particular, every truth-functional tautology is known if any is. That ﬁrst form of logical omniscience is independent of the speciﬁcs of the right-hand side of the deﬁnition. The second form depends on those speciﬁcs, but unlike the ﬁrst applies whenever premises p 1, . . . , pn entail a conclusion q(p1\ . . . \pn q), even if the

I M P RO B A B L E K N OW I N G

149

conclusion is not identical with any of the premises; it says that, in those circumstances, knowing the premises entails knowing the conclusion (Kp1 \ . . . \Kpn Kq), for if each premise is true in all worlds in which whatever one knows is true, the conclusion is also true in all worlds in which whatever one knows is true. On most views of propositional attitudes, logical omniscience is an extreme idealization. Someone can know that p ! p without knowing that ((p ! q) ! p) ! p, even though both formulas are tautologies. For present purposes it is a harmless one, for if the total evidence of a logically omniscient agent can be radically misleading about their epistemic position, our own abject failure of logical omniscience will not save us from the same fate. We will return to this issue towards the end of the paper. One effect of full logical omniscience is that for each world w there is the strongest proposition known by the agent in w, R(w), in the sense that it is known in w and entails every proposition that is known in w. We can deﬁne R(w) = {x 2 W : Rwx}. Then a proposition p is known in w if and only if p follows from R(w); more formally, w 2 Kp if and only if RðxÞ p, by deﬁnition of K. One constraint on the epistemic possibility relation will be maintained throughout. R is reﬂexive; every world is epistemically possible in itself, because knowledge entails truth: whatever one knows in a world is true in that world. Consequently, w 2 RðwÞ for any world w, and Kp p for any proposition p. Now recall the formal structure of counter-examples to the KK principle in such frames for epistemic logic. Here is a toy example. W is a three-member set {x, y, z}: x —————————— y ——————————z

In the diagram, R holds between worlds just in case they are identical or neighbours; thus R is both reﬂexive and symmetric. R is not transitive, because Rxy and Ryz but not Rxz. The strongest things known in each world are these: R(x) = {x, y}; R(y) = {x, y, z}; R(z) = {y, z}. In effect, if one is at one of the endpoints, what one knows is that one is not at the other endpoint; if one is at the midpoint, one knows nothing nontrivial about one’s position. Now let p = {x, y}. Then Kp = {x}: one knows p in x because p is true in all worlds epistemically possible in x; one does not know p in y because p is false in z, which is epistemically possible in y. Consequently KKp = K{x} = {}: one does not know Kp in x because Kp is false in y, which is epistemically possible in x. Thus the KK principle fails in x, because Kp is true and KKp false there. As is well known, the non-transitivity of R is necessary and sufﬁcient for a frame to contain a counter-example to the KK principle. For consider any frame <W, R>. Suppose that R is non-transitive. Thus for some x, y, z in W, Rxy and Ryz but not Rxz. By deﬁnition, KR(x) is true in x. KR(x) is not true in y, because Ryz and R(x) is not true in z (since not Rxz). Therefore KKR(x) is not true in x, because Rxy. Thus the KK principle fails in x. Conversely, suppose that there is a counter-example to the KK principle in <W, R>, say in x 2 W . Thus for some p W , Kp is true in x and KKp false in x. By the latter, for some y 2 W , Rxy and Kp is false in y, so for some z 2 W , Ryz and p is false in z. But not Rxz, otherwise Kp is false in x, contrary to hypothesis. Thus R is non-transitive.

150 ‘ K N OW L E D G E

FIRST’

Of course, the existence of such non-transitive frames for epistemic logic does not by itself establish that there are counter-examples to the KK principle on its intended interpretation, for it remains to be shown that these mathematical structures represent genuinely possible epistemic situations. Before we turn to such matters, however, we must ﬁrst enhance the frames with probabilistic structure, so that we can model issues about the probability on one’s evidence that one knows something.

9.2 In adding probabilities to a frame <W, R>, the account of evidential probability in Williamson (2000: 209–37) will be followed. We start with a prior distribution Probprior over propositions. Thus we can take a probabilistic epistemic frame to be an ordered triple <W, R, Probprior>, where W and R are as before and Probprior is a probability distribution deﬁned over subsets of W. In the frames considered in detail below, Probprior always takes the particularly simple form of a uniform distribution over the subsets of a ﬁnite set W. Thus every world has the same probabilistic weight as every other world; where |p| is the cardinality of p W , Probprior(p) = |p|/|W|. It is not suggested that non-uniform or inﬁnite probability distributions are in any way illegitimate. However, if a nonuniform distribution were used to illustrate the epistemic phenomena in question, they might look like artefacts of gerrymandering. Similarly, if W were inﬁnite, the phenomena might look like paradoxes of inﬁnity, given the complications of probability distributions over inﬁnite sets. It is therefore best to use a uniform prior distribution over a ﬁnite space where possible. For such uniform prior distributions, every nonempty subset of W has nonzero probability. We can therefore unproblematically deﬁne prior conditional probabilities by ratios in the usual way: Probprior(p|q) = Probprior(p\q)/Probprior(q) whenever q is nonempty and so Probprior(q) > 0. The evidential probability of a proposition in a world w is identiﬁed with its probability conditional on one’s total evidence in w. One’s total evidence in w can in turn be identiﬁed with the total content of what one knows in w (Williamson 2000: 184–208). In a frame <W, R>, the total content of what one knows in w is just R(w). Since w 2 RðwÞ, R(w) is always nonempty, so probabilities conditional on R(w) are always well-deﬁned. So if Probw(p) is the evidential probability in w of a proposition p: Probw ðpÞ ¼ Probprior ðpjRðwÞÞ ¼ Probprior ðp \ RðwÞÞ=Probprior ðRðwÞÞ Thus in ﬁnite uniform frames, the evidential probability in w of p is simply the proportion of epistemically possible worlds in w in which p is true. We can locate propositions about evidential probabilities in the frame. For instance, the proposition [Pr(p) = c] that the evidential probability of p is the real number c is simply {w 2 W : Probw(p) = c}, and similarly for inequalities involving evidential

I M P RO B A B L E K N OW I N G

151

probabilities. Thus propositions about evidential probabilities will themselves have evidential probabilities. Let <W, R> be the three-world toy example from the previous section. As before, p is {x, y}, so Kp is {x}. Let Probprior be the uniform distribution for W, so Probprior({x}) = Probprior({y}) = Probprior({z}) = 1/3. Hence Prx(Kp) = ½, since Kp is true in just one of the two worlds that are epistemically possible in x. Thus in x, even though one knows p, the probability on one’s evidence that one knows p is no more than 50–50. To say that the probability on one’s evidence that one knows p is just ½ is already to say something much worse about the epistemic status for one of the proposition that one knows p than merely to say that one does not know that one knows p. Let <W, R, Probprior> be any probabilistic epistemic frame where W is ﬁnite, R is reﬂexive and Probprior is uniform. Suppose that the KK principle fails in the frame. Thus for some w 2 W and p W , Kp is true in w while KKp is false in w. Thus for some x 2 RðwÞ, Kp is false in x. Since Probprior({x}) > 0, Probw(Kp) < 1. Thus wherever the KK principle fails in such models, one knows something although the probability on one’s evidence that one knows it is less than 1. By contrast, if the KK principle holds in a frame, if Kp is true in w then KKp is true in w, so RðwÞ Kp, so Probw(Kp) = 1: whenever one knows something, the probability on one’s evidence that one knows it is 1. Indeed, in such models knowing p is equivalent to its having probability 1 on one’s evidence; Kp = [Pr(p)=1]. Thus the KK principle is equivalent to the principle that if the evidential probability of p is 1, then the evidential probability that the evidential probability of p is 1 is itself 1. In frames where W is ﬁnite, R is reﬂexive and Probprior is uniform, how low can Probw(Kp) be when Kp is true in w? When one knows something, how low can the probability that one knows it be on one’s evidence? The probability can be any rational number whatsoever strictly between 0 and 1, so it can be arbitrarily close to 0. To see this, let m/n be any rational number such that 0 < m/n < 1, where m and n are positive integers, so 0 < m < n. We construct a model of the required kind with some worlds in which Kp is true while the evidential probability of Kp is m/n. The idea is simple: take the three-world toy model used to illustrate the failure of the KK principle in section 1, multiply the world x by m, yielding m mutually indiscernible copies, and the world y by nm, yielding nm mutually indiscernible copies. Thus W = {x1, . . . , xm, ym+1, . . . , yn, z}. R is reﬂexive and symmetric; for any i, i*, j, j* where 1 i, i* m < j, j* n: Rxixi*, Rxiyj, Ryjyj*, and Ryjz, but not Rxiz; diagrammatically: x1 : xi —————————— : xm

ym+1 : yj —————————— z : yn

If p = {x1, . . . , xm, ym+1, . . . , yn}, for 1 i m, R(xi) = p. Thus Kp = {x1, . . . , xm}. Consequently, while in xi Kp is true, it is true in only m of the n epistemically possible

152 ‘ K N OW L E D G E

FIRST’

worlds. Since Probprior is uniform, Probxi(Kp) = m/n, as required. In particular, for m = 1, Probxi(Kp) = 1/n, which goes to 0 as n goes to inﬁnity. By using non-uniform prior probability distributions or inﬁnite sets of worlds we could construct similar models in which Kp is true while actually having evidential probability 0, but such reﬁnements are unnecessary here.1 In the model just illustrated, m/n is the highest evidential probability that Kp attains anywhere, since Probyj(Kp) = m/(n+1) and Probz(Kp) = 0. Thus the proposition [Pr(Kp)m/n] is true at every world in the model. Consequently, so is the proposition Kk[Pr(Kp)m/n], where k is a natural number and Kk means k iterations of K (thus K 0q is q and Kk+1q is KkKq). In other words, knowing p is compatible not just with the probability on one’s evidence that one knows p being close to 0, but even with knowing, and knowing that one knows, and knowing that one knows that one knows, . . . that the probability on one’s evidence that one knows p is close to 0. We have added evidential probabilities to epistemic models in a way that embodies several strong assumptions. In particular, one’s total evidence was equated with the total content of one’s knowledge, and probabilities on that evidence were calculated by conditionalizing a prior probability distribution on it. These assumptions are defensible (Williamson 2000), but can of course be challenged. However, someone who denied that they always hold would not be thereby committed to rejecting their present applications. For they are being used to argue that a speciﬁc phenomenon can occur, not that it always occurs. That requires that the relevant models can be instantiated by genuine epistemic phenomena, not that all genuine epistemic phenomena have the structure of those models. Indeed, the assumptions at issue should make it harder to construct models with the target phenomenon, which involves a sort of tension between knowing and the evidential probability of knowing. For what is most distinctive about the present approach is the intimate connection it postulates between evidential probabilities and knowledge; the assumptions cramp attempts to arrange the tension between them. By contrast, an approach that allowed more independence between evidential probabilities and knowledge would have much more scope to arrange the tension, by varying the evidential dimension independently of the knowledge dimension or vice versa. Similarly, allowing non-uniform prior probability distribution or inﬁnite sets of worlds would give far more scope for arranging odd probabilistic phenomena, for example by giving special weight to ‘bad’ worlds. If the target phenomenon occurs even under the unhelpful restrictive conditions postulated by the present approach to evidential probabilities, it is robust. A less restrictive approach could hardly rule out the models already constructed. The challenge to opponents is to motivate an approach that is more restrictive in some relevant way. So far, however, we have been working at the level of formal models, without any positive argument that they represent genuinely possible epistemic situations. We now 1 For an argument that epistemically possible propositions can have probability 0, even when inﬁnitesimal probabilities are allowed, see Williamson (2007b).

I M P RO B A B L E K N OW I N G

153

turn to that task, and provide a much more realistic, only slightly schematized description of a mundane type of epistemic situation that exempliﬁes the target phenomenon.

9.3 Imagine a plain, unmarked circular dial with a single pointer that rotates in imperceptibly short discrete jerks, a modernist clock-face designed with an eye to the appearance rather than the reality of functional efﬁciency. The hand can point at any one of n equally spaced unmarked positions on the perimeter of the dial. We measure distances between positions by the minimum number of steps needed to go from one to the other (clockwise or anti-clockwise). Number the positions 0, . . . , n1, where 0 is 12 o’clock. For simplicity, we individuate ‘worlds’ (mutually exclusive and jointly exhaustive relevant circumstances) just by the position of the hand. Thus there are exactly n worlds w0, . . . , wn1, where in wi the hand points at position i. We measure distances between worlds by the corresponding distances between positions. An interval of positions corresponds to a time interval; it is a nonempty proper subset of the set of positions such that the hand goes through every position in the set without going through any position not in the set (intervals are ‘connected’). An endpoint of an interval is a member next to a non-member. Any interval with at least two members has two endpoints. An interval with an odd number of members has a unique midpoint, equidistant from its endpoints. Given the natural one–one correspondence between worlds and positions, the terms ‘interval’, ‘endpoint’, and ‘midpoint’ can be applied just as well to sets of worlds as to sets of positions. Now imagine that you are looking at the dial from a ﬁxed point of view equidistant from all points on the perimeter, and that you have no independent information about the hand’s position. You can make some discriminations between positions, and correspondingly between worlds, but the difference between neighbouring positions is well below your threshold of discrimination. We may assume that your capacity to discriminate between positions depends only on their relative distance; thus if world w is at least as close to world x as world y is to world z, then you can discriminate w from x only if you can also discriminate y from z. Consequently, if you are in fact in world w, the worlds that for all you know you are in (the epistemically possible worlds) are those at most h steps from w, for some natural number h; h is greater than 0, otherwise your discrimination would be perfect. We can regard h as the width of the margin for error you require in order to know something in the model (Williamson 2000: 125–34). More formally, let R be the epistemic accessibility relation; then for all worlds w and x, Rwx if and only if the distance between w and x is at most h. Thus R has both reﬂective and rotational symmetry. For any world w, R(w) (the set of worlds epistemically possible in w) is not the whole of w, otherwise you could not make any discriminations at all. Thus R(w) is an interval with w as its midpoint and 2h+1 members. Consequently, R(w) is known only at w. For suppose that R(w) is known at a world x. Since R(x) is the strongest proposition known at x, R(x) R(w). But R(x)

154 ‘ K N OW L E D G E

FIRST’

and R(w) have the same ﬁnite number of members, 2h+1. Thus R(x) = R(w). So the midpoint of R(x) is the midpoint of R(w); that is, x = w. Hence KR(w) is true in w and in no other world. Now add evidential probabilities to the model as above, with a uniform prior distribution, Probprior. Since KR(w) is true in just one of the 2h+1 worlds in R(w), its evidential probability in w, Probw(KR(w)), is 1/(2h+1). By increasing the number of positions round the dial while keeping your discriminatory capacities ﬁxed, we can increase h without limit, and thereby make the evidential probability in w that one knows R(w) as small as desired, even though R(w) is in fact known in w. As in Section 9.2, the evidential probability in the model that the proposition is known is not only small, but known to be small, and known to be known to be small, and known to be known to be known to be small, and. . . . For since KR(w) is true in only one world, and for any world x R(x) has 2h+1 members, Probx(KR(w)) is always at most 1/(2h+1). Thus the proposition [Pr(KR(w))1/(2h+1)] is true in every world in the model. Consequently, the proposition K k[Pr(KR(w))1/(2h+1)] is also true in every world. In other words, one can have any number of iterations of knowledge that the probability of R(w) is at most 1/(2h+1). One cannot avoid these structural results by tightening the conditions for knowledge, short of complete scepticism. For reducing the range of truths known amounts in this setting to increasing the margin for error h. But, given the symmetry of the situation, the argument holds for any positive margin for error—unless h is made so large that R(w) is the whole of W, which is in effect to say that one learns nothing by looking at the dial. Even denying the equation of evidence with knowledge would make very little difference to the argument. It would presumably involve postulating one margin for error h for knowledge and a distinct margin for error h* for evidence: the worlds compatible with the total content of one’s evidence in w would be those within a distance h* of w; h* is nonzero too, for more than one position is compatible with one’s evidence. That would not affect the argument that KR(w) is true in no world except w. Hence the probability on one’s evidence in w of KR(w) would be 1/(2h*+1). By increasing the number of positions for the dial, one can make h* arbitrarily high, and therefore the probability on one’s evidence in w that one knows R(w) arbitrarily low. The foregoing results should still hold on reasonable variations in the prior probability distribution R(w) that make it slightly non-uniform, for KR(w) will still be true only in w, and so its probability (the probability that one is in w) will still be low in w and a fortiori everywhere else too. Similarly, making the allowable space of positions for the hand continuous rather than discrete should not make much difference. One would also expect the target phenomenon to often arise in comparable ways when the epistemic accessibility relation R takes different forms, for example by being linear or multi-dimensional. Nor do psychologically more realistic descriptions of knowledge seem to raise the probability on one’s evidence that one knows the strongest relevant proposition one can know, when one does in fact know it. Thus the target epistemic phenomenon seems robust.

I M P RO B A B L E K N OW I N G

155

Reﬂection suggests a generalization of the example. One key structural feature of the model is this: (*) For all worlds w, x: R(x) R(w) only if x = w. That is, shifting from one world to another (as from w to x) always opens up new epistemic possibilities as well as closing down old ones. Some worlds are close enough to x to be epistemically possible in x but not close enough to w to be epistemically possible in w. This is a plausible feature of real-life examples of inexact knowledge. In fact, this more limited feature may do: (**) Some world w is such that for all worlds x: R(x) R(w) only if x = w. For, given (**), w is still the only world in which R(w) is known, so the evidential probability of KR(w) will tend to be small in w, given the subject’s imperfect powers of discrimination, even though KR(w) is always true in w. Note that if at least one verifying world w for (**) has R to a world other than itself then R is non-transitive. For suppose that Rwx, x 6¼ w and R(x) R(w) only if x = w. Then for some world y, Rxy but not Rwy, so transitivity fails. The existence of natural structural generalizations such as (*) and (**) provides some further conﬁrmation of the robustness of the phenomenon of knowing that is highly improbable on the subject’s own evidence.

9.4 One restrictive feature of the model in Section 9.3 is that the width of the margin for error required for knowledge is in effect treated as beyond doubt, since it is built into the structure of the model. More speciﬁcally, since the model has only one world in which the clock hand has a given position, worlds can differ over what positions are epistemically possible for the hand only by differing over which position it in fact has. Yet it is plausible that there is inexactness in our knowledge of the width of the margin for error in addition to the inexactness in our knowledge of the position of the hand. If so, then in more realistic models the worlds epistemically possible in a given world w will include some in which the margin for error differs slightly from that in w, while the position of the hand is the same. In particular, in w a world x is epistemically possible in which the margin for error is slightly less than in w. In such cases we may have R(x) R(w) even though x 6¼ w. Pictorially: a sphere may contain a sphere of slightly smaller radius whose centre is a slight distance from the centre of the ﬁrst sphere. Then whatever is known in w is also known in x. In such cases, (*) and even (**) may fail.2

2 On an epistemic account of vagueness, such variable margins for error yield distinctive forms of higherorder vagueness. Williamson (1999: 136–8) argues that if the ‘clearly’ operator for vagueness obeys the analogue of the B (for ‘Brouwersche’) axiom p ! K¬K¬p (which corresponds to the condition of symmetry on R) then any formula with second-order vagueness has nth-order vagueness for every n > 2, but does not

156

‘ K N OW L E D G E

FIRST’

To construct models with a variable margin for error is not hard. But doing so without making ad hoc choices is harder. In effect, one must specify higher-order margins for error distinct from the ﬁrst-order margins for error. There is no obvious non-arbitrary way of determining the relation between the widths of the margins at different orders. By contrast with the simpler case in Section 9.3, it is not clear which models one should be considering. Nevertheless, in a setting with variable margins for error, one can still give an informal argument for a conclusion similar to that already reached in the case of constant margins. Let H(w) be the strongest proposition known in w about the position of the hand (or whatever other non-epistemic fact is relevant). Thus H(w) may be true at worlds other than w; its truth-value remains constant across worlds where the position of the hand is the same, even if the epistemic facts differ. Let h be the ﬁrstorder margin for error (the one relevant to knowledge of the position of the hand) in w. Thus H(w) is true in exactly those worlds where the distance of the hand position from that in w is at most h. Let ME<w be true in just those worlds in which the ﬁrst-order margin of error is less than h, ME>w be true in just those worlds in which the ﬁrstorder margin for error is greater than h, and ME=w be true in just those worlds in which the ﬁrst-order margin for error is equal to h. These three possibilities are mutually exclusive and jointly exhaustive. Therefore, by deﬁnition of conditional probability: Probw ðKHðwÞÞ ¼ Probw ðKH ðwÞjME<w Þ:Probw ðME<w Þþ Probw ðKHðwÞjME¼w Þ:Probw ðME¼w Þþ Probw ðKHðwÞjME>w Þ:Probw ðME>w Þ

ð1Þ

In any world x in ME>w some world is epistemically possible in which H(w) is false, because the ﬁrst-order margin for error in x is some k > h, and a sphere of radius k cannot be contained in a sphere of radius h. Thus ME>w is incompatible with KH(w), so Probw(KH(w) | ME>w) = 0. Consequently, (1) simpliﬁes to: Probw ðKHðwÞÞ ¼ Probw ðKH ðwÞjME<w Þ:Probw ðME<w Þþ Probw ðKHðwÞjME ¼w Þ:Probw ðME ¼w Þ

ð2Þ

Since Probw(KH(w) | ME<w) 1, (2) yields: Probw ðKHðwÞÞ Probw ðME <w Þ þ Probw ðKHðwÞjME ¼w Þ:Probw ðME ¼w Þ

ð3Þ

For simplicity, we may reasonably assume that Probw(ME<w) = Probw(ME>w), that is, that the ﬁrst-order margin for error is equally likely to be less or greater than its actual value. Since Probw(ME<w) + Probw(ME=w) + Probw(ME>w) = 1, Probw(ME<w) = (1Probw(ME=w))/2. Therefore, by (3):

endorse the B axiom for ‘clearly’. In response, Mahtani (2008) uses variable margins for error to argue against the B axiom for ‘clearly’ and suggests that they allow vagueness to cut out at any order. Dorr (2008) provides a formal model in which he proves Mahtani’s suggestion to hold. These arguments all have analogues for the overtly epistemic case.

I M P RO B A B L E K N OW I N G

157

Probw ðKHðwÞÞ ð1 Probw ðME ¼w ÞÞ=2 þProbw ðKHðwÞjME ¼w Þ:Probw ðME ¼w Þ

ð4Þ

But Probw(KH(w) | ME=w) is in effect the probability of KH(w) in the case considered previously of a constant margin for error (h). From that case we have at the very least that Probw(KH(w) | ME=w) < ½. Consequently, by (4): Probw ðKHðwÞÞ ð1 Probw ðME ¼w ÞÞ=2 þ Probw ðME ¼w Þ=2 ¼ 1=2

ð5Þ

In other words, although you in fact know H(w) in w, it is no more probable than not on your evidence in w that you know H(w). Even if we slightly relax the simplifying assumption that Probw(ME<w) = Probw(ME>w), the probability on the evidence in w that H(w) is known will not rise signiﬁcantly above evens. Indeed, the probability may well be close to zero. For if the width of the ﬁrst-order margin for error varies only slightly (as a proportion of h) over the worlds epistemically possible in w, then Probw(KH(w) | ME<w) will be close to Probw(KH(w) | ME=w). Therefore, by (2), Probw(KH(w)) will be at most only slightly greater than Probw(KH(w) | ME=w).Probw(ME<w) + Probw(KH(w) | ME=w).Probw (ME=w) = Probw(KH(w) | ME=w).Probw(MEw). But, as noted above, Probw(KH(w) | ME=w) is in effect the probability of KH(w) in the case already considered of a constant margin for error. That probability goes to zero as the number of hand positions increases. Hence Probw(KH(w)) may well be close to 0 even when the width of the margin for error varies. But even without that stronger conclusion, the result of the informal argument is enough for present purposes. Uncertainty about the width of the margin for error does not undermine the possibility of knowing something without its being probable on one’s evidence that one knows it.

9.5 Examples of the kind considered in previous sections make trouble for accounts of propositional justiﬁcation on which a sufﬁcient condition for having such justiﬁcation to believe p is that the probability of p on one’s evidence exceeds a threshold less than 1. For in such cases one would then have propositional justiﬁcation to believe the Moorean conjunction p & ¬Kp. Consider a world w in which one knows p but Probw(Kp), the probability on one’s evidence in w that one knows p, is less than 1/n, so Probw(¬Kp) > (n1)/n. Since one knows p, Probw(p) = 1. By elementary probability theory, it follows that Probw(p & ¬Kp) > (n1)/n. Thus by letting n go high enough, we can ﬁnd a case in which the probability on one’s evidence of the Moorean conjunction exceeds the given threshold and the supposedly sufﬁcient condition for justiﬁcation is met. But it is not plausible that one can have justiﬁcation to believe a conjunction of the form of ‘It’s raining and I don’t know that it’s raining’. However, the phenomenon at issue does not merely raise problems for particular philosophical views. More importantly, it raises problems for knowing and acting

158

‘ K N OW L E D G E

FIRST’

subjects as such, not just for those knowing and acting subjects who choose to theorize about knowing and acting subjects. Suppose, for instance, that it is rational for Hamlet to perform some action A if he knows p and not rational for him to perform A otherwise (given the relevant background circumstances, such as his other attitudes), and that those facts about rationality are clear to him.3 Suppose also that although Hamlet knows p, it is almost certain on his evidence that he does not know p. Should Hamlet do A? Since he knows p, it is rational for him to do A. However, since it is almost certain on his evidence that he does not know p, it is almost certain on his evidence that it is not rational for him to do A. It is therefore very tempting to say that after all it is not rational for him to do A. But to say that would be to contradict the conditions of the example. Thus there is pressure to say instead that the example cannot really arise, that if it is almost certain on one’s evidence that one does not know p then one does not really know p. We have already seen that such pressure should be resisted, since it will not stop short of skepticism. Nevertheless, its effect when we consider particular cases may be to make us withdraw true knowledge ascriptions, under the false impression that they have been defeated by the negative evidence. Of course, we have no general difﬁculty with the idea that a claim may be true even though it is almost certain on someone’s evidence that it is false—for example, when the claim states the actual outcome of a long sequence of unobserved coin tosses. What we ﬁnd harder to accept is the possibility of the same combination when the claim ascribes knowledge to the very person whose evidence is in question at that very time. Their strong evidence that they do not know p seems incompatible with whatever sort of reliance on p is mandated by their knowing p. But the apparent incompatibility is an illusion, which can seriously distort our assessment of particular ascriptions of knowledge and with it our epistemological theorizing. The phenomenon that we ﬁnd hard to accept is an extreme case of anti-luminosity, the failure of non-trivial states to satisfy the constraint that whenever one is in them one is in a position to know that one is in them (Williamson 2000: 93–113). In evidential terms, when luminosity fails one is in a state S even though it is not certain on one’s evidence that one is in S. In the present case, S is the state of knowing p. But the phenomenon at issue is stronger than a mere failure of luminosity, since it involves being in S even though it is almost certain on one’s evidence that one is not in S. If we do resist the skeptical pressure, and acknowledge the possibility of the phenomenon, then we must regard Hamlet’s problem above as a genuine problem for him, an instance of a more general practical problem for agents acting on inexact knowledge, not as something to be dissolved by epistemological redescription. The solution is clear from the conditions of the example: it is rational for Hamlet to do A if and only if he knows p, and he does know p, so it is rational for him to do A. The trouble is that he has strongly misleading evidence about those facts. But that does not 3 For a recent discussion of the relation between knowledge and reasons for action see Hawthorne and Stanley (2008).

I M P RO B A B L E K N OW I N G

159

mean that they are not facts after all; it just means that he has a good excuse—not justiﬁcation—for not doing A.4 We must check these general comments by examining a speciﬁc case in more detail. That is the business of the following ﬁnal section.

9.6 Competent deduction is a way of extending our knowledge, as in mathematical reasoning. A natural formulation of the underlying principle is this form of multipremise closure: MPC. If one believes a conclusion on the basis of competent deduction from premises each of which one knows, one knows the conclusion. For present purposes we can leave the phrase ‘on the basis of competent deduction’ vague, and assume that something in the spirit of MPC is correct.5 MPC faces a strong challenge from cases of reasoning from many premises, each with a small independent risk of error, where those risks combine into a large risk of error for the conclusion. Such situations arise most simply when the conclusion is the conjunction of the premises, since then the conclusion entails the premises and any error in a premise involves an error in the conclusion; the deduction of the conclusion from the premises also takes a particularly elementary form. Suppose that each premise is true, so the conclusion is true too. We may further suppose that, according to ordinary standards, one knows each premise. For example, each premise may state a separate matter of particular historical fact that one has carefully checked. To deny that one knows a given premise would look like extreme skepticism. From those propositions one has competently deduced their conjunction and believes it on that basis. By MPC, one thereby knows the conclusion. Nevertheless, much past experience may show that there is a small non-zero rate of error for such carefully checked historical claims, on the basis of which it is almost inevitable that the long conjunction contains several false conjuncts. Thus one’s belief in the conjunction may seem to have too strong a risk of error to constitute knowledge. This is a variant of the Preface Paradox, adapted to knowledge. On a fallibilist conception of knowledge, it seems, the risk of error for any one premise may be within the threshold for knowledge, while the risk of error for the conclusion is outside the threshold; thus MPC fails. By contrast, on an infallibilist conception, knowledge requires that there be no risk of error, rather than at most a small risk, so MPC may hold; but such infallibilism seems to lead to scepticism. Alternatively, one might try to preserve MPC by postulating some sort of variation

4 5

For a critique of internalism about justiﬁcation see Williamson (2007a). Such a principle is called ‘intuitive closure’ at Williamson (2000: 117–18).

160

‘ K N OW L E D G E

FIRST’

across contexts of utterance in the reference of ‘know’, or some loss of knowledge in the premises through the very act of deducing the conclusion.6 Once we recognize the phenomenon of knowing when it is almost certain on one’s evidence that one fails to know, we can see a possible diagnosis of the problem cases for MPC which allows us to keep MPC while neither falling into scepticism nor postulating truth-conditions for ‘knowledge’-ascriptions of any non-standard kind. One does indeed know each premise, without knowing that one knows it. Since one believes the conclusion on the basis of competent deduction from the premises, by MPC one also knows the conclusion, although without knowing that one knows it. For each premise, it is very probable on one’s evidence that one knows it. However, it is very improbable on one’s evidence that one knows every premise. Given that one knows the conclusion (the conjunction) only if one knows every premise, it is very improbable on one’s evidence that one knows the conclusion. Since we are tempted to conceive knowing as at least somewhat luminous, we are tempted to deny that one knows the conclusion.7 We can give a formal epistemic model of that description. As noted in Section 9.1, such a model automatically validates not just MPC but logical omniscience, an unrealistically strong version of multi-premise closure that does not even require the agent to have carried out the relevant deductions, competently or otherwise. As a positive argument for MPC, it might well be accused of both begging the question and proving too much. However, its role here is different. It is being used to defuse an objection to MPC, by showing that even on the assumption of the strongest possible version of multi-premise closure, one would predict the occurrence of epistemic phenomena that it is very tempting to misinterpret along the lines of the objection as counter-examples to multi-premise closure. In particular, if we treat ascriptions of knowledge as defeated by a low probability of knowing, but not by a high probability short of 1 of knowing, on the subject’s evidence, then we shall tend to judge that the subject knows each conjunct without knowing the conjunction, even though the conditions for MPC are satisﬁed; we are deceived by the false appearance of a counter-example to MPC.8

6

For discussion of MPC in relation to the Preface Paradox see Hawthorne (2004: 46–50, 154, 182–3). Similar cases involving future chances are used in Hawthorne and Lasonen-Aarnio (2009) against the safety conception of knowledge in Williamson (2000); for a reply see Williamson (2009). 7 A similar problem arises for single-premise closure principles when one competently carries out a long chain of single-premise deductive steps, each with a small epistemic probability of inferential error (in the multi-premise case, for simplicity, one’s deductive competence is treated as beyond doubt); see LasonenAarnio (2008) for discussion. Here is a parallel account of that case. One knows the premise without knowing that one knows it. For each deductive step, one carries it out competently without knowing that one does so. By single-premise closure, one knows the conclusion, without knowing that one knows it. For each deductive step, it is very probable on one’s evidence that one carries it out competently. However, it is very improbable on one’s evidence that one carries out every deductive step competently. Since it is granted that one knows the conclusion only if one carries out every deductive step competently, it is very improbable on one’s evidence that one knows the conclusion. 8 The usual form of epistemic modelling is not appropriate for treating possible errors in deductive reasoning, since logical omniscience suppresses the dependence of inferential knowledge on correct inferential processes.

I M P RO B A B L E K N OW I N G

161

Here are the details of such a model.9 For worlds we use n-tuples of numbers drawn from the set {0, 1, . . . , 2k}, where n is the number of premises (conjuncts) and k is a large natural number. Thus there are (2k+1)n worlds. The n components of a world represent its locations on n independent dimensions of a state space. The ith dimension is the one relevant to the ith premise pi. Let the ith component of the n-tuple w be wi, and the world just like w except that its ith component is m be w[i|m], so w[i|m]i = m and w[i|m]j = wj if i 6¼ j. Let pi be true in w if and only if wi > 0. Let x be epistemically possible in w (Rwx) if and only if for all i, |wi–xi| k, that is, w and x do not differ by ‘too much’ in any of their respective components. In effect, a margin for error is applied to each of the n dimensions separately. The relation R is obviously reﬂexive and symmetric. We can easily check that for any world w, pi is known (Kpi is true) in w if and only if wi > k.10 Similarly, we can check that pi is known to be known (KKpi is true) in w if and only if wi > 2k; hence pi is not known to be known (KKpi is not true) in any world in this model. In particular, in the world <2k, . . . , 2k>, each premise pi is known and none is known to be known. As usual, the prior probability distribution is uniform: for any world w, Probprior({w}) = 1/(2k+1)n. We must check that the model makes the n dimensions probabilistically independent. For any given i, a proposition q is i-based if and only if for all worlds x and y, if xi = yi then q is true in x if and only if q is true in y (1 i n). That is, whether an i-based proposition is true in a world depends only on the ith component of that world. In particular, pi is an i-based proposition. Obviously, the negation of any i-based proposition is also i-based, as is any conjunction of i-based propositions. We can also prove that whenever q is an i-based proposition, so is Kq.11 Thus Kpi and KKpi are also i-based propositions. Then we can prove that whenever for each i qi is an i-based proposition, q1, . . . , qn are mutually probabilistically independent on the evidence in any world w, in the usual sense that the probability (on the evidence in w) of their conjunction is the product of the probabilities (on the evidence in w) of the conjuncts.12 Although a model could have been constructed in which the evidence at some worlds establishes epistemic interdependences between the different dimensions, 9

The technical details are taken from Williamson (2009). Proof: Suppose that wi > k. If Rwx then |wi–xi| k, so xi > 0, so x 2 pi. Thus w 2 Kpi. Conversely, suppose that wi k. Then Rww[i|0], for |wi–w[i|0]i| = |wi–0| = wi k and if i 6¼ j then |wj–w[i|0]j| = 0; but w[i|0] 2 = pi because w[i|0]i = 0, so w 2 = Kpi. 11 Proof: Suppose that q is i-based and xi = yi. Suppose also that x 2 = Kq. Then for some z, Rxz and z 2 = q. But then Ryy[i|zi], for |yi–y[i|zi]i| = |yi–zi| = |xi–zi| (because xi = yi) k (because Rxz), and if i 6¼ j then |yj–y[i|zi]j| = 0. Moreover, y[i|zi] 2 = q because z 2 = q, q is i-based and y[i|zi]i = zi. Hence y 2 = Kq. Thus if y 2 Kq then x 2 Kq. By parity of reasoning the converse holds too. 12 Proof: Set #(i, q, w) = {j: 0 j 2k, w[i|j] 2 q and |wi–j| k} for any w 2 W, q W, 1 i n. For each i, let qi be i-based. Let \qi = q1 \ . . . \ qn. For w 2 W, R(w) \ \qi = {x: 8 i, xi 2 #(i, qi, w)}, since for each i and x 2 W, x 2 qi iff w[i|xi] 2 qi since qi is i-based. Since Probprior is uniform, Probw(\qi) = |{R(w) \ \qi|/ |R(w)| for w 2 W. But |R(w) \ \qi| = |{x: 8 i, xi 2 #(i, qi, w)}| = |#(1, q1, w)| . . . |#(n, qn, w)|. By the special case of this equation in which each qi is replaced by W (which is trivially i-based for any i), |R(w)| = |#(1, W, w)| . . . |#(n, W, w)|. Consequently: 10

Probw(\qi) = (|#(1, q1, w)| . . . |#(n, qn, w)|)/( |#(1, W, w)| . . . |#(n, W, w)|).

162 ‘ K N OW L E D G E

FIRST’

for present purposes we can do without such complications. In particular, p1, . . . , pn are mutually probabilistically independent on the evidence in any world, as are Kp1, . . . , Kpn. But, on the evidence in the world <2k, . . . , 2k>, for any given i, the probability that pi is known is k/(k+1).13 By probabilistic independence, the probability of the conjunction Kp1 \ . . . \ Kpn is (k/(k+1))n. That is the probability that each conjunct is known. But, by the logical omniscience built into the model, knowing a conjunction (K(p1 \ . . . \ pn)) is equivalent to knowing each conjunct. Thus the probability on the evidence in <2k, . . . , 2k> that the conjunction p1\ . . . \ pn is known is also (k/(k+1))n. For ﬁxed k, this probability becomes arbitrarily close to 0 as n becomes arbitrarily large. Thus, for suitable k and n, the world <2k, . . . , 2k> exempliﬁes just the situation informally sketched: for each conjunct one knows it without knowing that one knows it, and it is almost but not quite certain on one’s evidence that one knows the conjunct; one also knows the conjunction without knowing that one knows it, and it is almost but not quite certain on one’s evidence that one does not know the conjunction. Of course, in some examples one’s epistemic position with respect to each conjunct is better: one not only knows it but knows that one knows it. If one also knows the relevant closure principle, and knows that one satisﬁes the conditions for its application, one may even know that one knows the conjunction. Consequently, the probability on one’s evidence that one knows the conjunction is 1. However, the previous pattern may still be repeated at a higher level of iterations of knowledge. For example, for each conjunct one knows that one knows it without knowing that one knows that one knows it, and it is almost but not quite certain on one’s evidence that one knows that one knows the conjunct; one also knows that one knows the conjunction without knowing that one knows that one knows it, and it is almost but not quite certain on one’s evidence that one does not know that one knows the conjunction. To adapt the previous model to this case, we can simply expand the set of worlds by using n-tuples of numbers from the set {0, 1, . . . , 3k} rather than {0, 1, . . . , 2k}, leaving the deﬁnitions of the epistemic possibility relation R and the truth-conditions of the pi unchanged (so pi is true in w if and only if wi > 0); then <3k, . . . , 3k> is a world of the required type. More generally, if one uses as worlds n-tuples of numbers from the set {0, 1, . . . , hk}, leaving the other features of the model unchanged, then will be a world at which one has h–1 but not h iterations of knowledge of each conjunct, and it is almost but not quite certain on one’s evidence that one has h–1 iterations of knowledge of the conjunct; one also has h–1 but not h iterations of knowledge of the conjunction,

For any given i, consider another special case in which qj is replaced by W whenever i 6¼ j. Since n–1 of the ratios cancel out, Prw(qi) = |#(i, qi, w)|/|#(i, W, w)|. Therefore Probw(\qi) = Probw(q1) . . . Probw(qn), as required. 13 Proof: We have already established that x 2 Kpi iff xi > k. Thus, in the notation of the previous footnote, #(i, Kpi, <2k, . . . , 2k>) = {j: k < j 2k}, so |#(i, Kpi, <2k, . . . , 2k>)| = k, while #(i, W, <2k, . . . , 2k>) = {j: k j 2k}, so |#(i, W, <2k, . . . , 2k>)| = k+1. By the formula for Probw(pi) in the previous footnote, Prob<2k, . . . ,2k>(Kpi) = k/(k+1).

I M P RO B A B L E K N OW I N G

163

and it is almost but not quite certain on one’s evidence that one does not have h–1 iterations of knowledge of the conjunction.14, 15 Many other variations can be played on the same theme. The general idea is this. Suppose that the epistemic status E satisﬁes an appropriate principle of multi-premise closure. In some situations, one attains E with respect to each conjunct, without knowing that one does so (this is possible by the anti-luminosity argument). By multi-premise closure, one also attains status E with respect to the conjunction, without knowing that one does so. Then for each conjunct it may be almost certain on one’s evidence that one attains E with respect to it, even though it is almost certain on one’s evidence that one does not attain E with respect to the conjunction. If we treat ascriptions of E as defeated by a low probability of E, but not by a high probability short of 1 of E, on the subject’s evidence, then we shall tend to judge that the subject attains E with respect to each conjunct but not with respect to the conjunction, even though the conditions for multiple-premise closure principle are satisﬁed; we are deceived by the false appearance of a counter-example to the multi-premise closure principle.

9.7 The considerations of this paper raise a more general question. Knowledge claims are often thought to be defeated by various sorts of misleading evidence. In how many cases is the correct account that the subject knows, even though it is almost certain on the subject’s evidence at the time that they do not know? That is left as an open question.16 14

A similar generalization to higher iterations of knowledge is possible for the case of multiple risks of inferential error in a single-premise deduction. One has at least n iterations of knowledge of the premise. For each deductive step, one has n1 but not n iterations of knowledge that one carried it out competently. By single-premise closure and plausible background assumptions, one has n but not n+1 iterations of knowledge of the conclusion. For each deductive step, it is very probable on one’s evidence that one has at least n1 iterations of knowledge that one carried it out competently. However, it is very improbable on one’s evidence that one has at least n1 iterations of knowledge that one carried out every deductive step competently. Since it is granted that one has at least n iterations of knowledge of the conclusion only if one has at least n1 iterations of knowledge that one carried out every deductive step competently, it is very improbable on one’s evidence that one has at least n iterations of knowledge of the conclusion. 15 See Williamson (2008) for more discussion of the structure and semantics of higher-order evidential probabilities. The phenomenon discussed in the text involves the apparent loss of only one iteration of knowledge between premises and conclusion. However, the apparent absence of a given number of iterations of knowledge can cause doubts about all lower numbers of iterations, by a domino effect, since lack of knowledge that one has n+1 iterations implies lack of warrant to assert that one has n iterations (Williamson 2005: 233–4). 16 The ideas in this paper were prompted by discussion with John Hawthorne. Earlier versions of the material were presented in classes and at discussion groups at Oxford, at an Arche´ workshop on Basic Knowledge at St Andrews, where Philip Ebert was the commentator, at a workshop on epistemology at the University of Wisconsin at Madison, where Jim Pryor was the commentator, at graduate conferences at University College London and the University of Miami, at a workshop on mathematical methods in philosophy in Banff (Alberta), at a symposium for Jaakko Hintikka in Copenhagen, and at colloquia in the philosophy departments of the University of Nottingham and the University of Texas at Austin. I thank all those who have contributed to the paper with their comments and questions.

164

‘ K N OW L E D G E

FIRST’

References Dorr, C. 2008. ‘How vagueness could cut out at any order’. MS. Hawthorne, J. 2004. Knowledge and Lotteries. Oxford: Clarendon Press. —— and Lasonen-Aarnio, M. 2009. ‘Knowledge and objective chance’, in P. Greenough and D. Pritchard (eds), Williamson on Knowledge. Oxford: Oxford University Press. —— and Stanley, J. 2008. ‘Knowledge and action’. The Journal of Philosophy 105: 571–90. Hintikka, J. 1962. Knowledge and Belief. Ithaca, NY: Cornell University Press. Lasonen-Aarnio, M. 2008. ‘Single premise deduction and risk’. Philosophical Studies 141: 157–73. Mahtani, A. 2008. ‘Can vagueness cut out at any order?’ Australasian Journal of Philosophy 86: 499–508. Williamson, T. 1999. ‘On the structure of higher-order vagueness’. Mind 108: 127–43. ——. 2000. Knowledge and its Limits. Oxford: Oxford University Press. ——. 2005. ‘Contextualism, subject-sensitive invariantism and knowledge of knowledge’. The Philosophical Quarterly 55: 213–35. ——. 2007a. ‘On being justiﬁed in one’s head’. In M. Timmons, J. Greco, and A. Mele (eds), Rationality and the Good: Critical Essays on the Ethics and Epistemology of Robert Audi. Oxford: Oxford University Press. ——. 2007b. ‘How probable is an inﬁnite sequence of heads?’ Analysis 67: 173–80. ——. 2008. ‘Why epistemology can’t be operationalized.’ In Q. Smith (ed.), Epistemology: New Philosophical Essays. Oxford: Oxford University Press. ——. 2009. ‘Reply to John Hawthorne and Maria Lasonen-Aarnio’, in P. Greenough and D. Pritchard (eds), Williamson on Knowledge. Oxford: Oxford University Press.

PART V

Concerns about the Internalist Component Technically, evidentialism is perfectly compatible with externalism. However, most who have found evidentialism attractive have also found internalism attractive (for, I think, reasons touched upon in the introduction). Internalism has been defended by Conee and Feldman in ‘Internalism Defended’, in Evidentialism. Also, though evidentialism is ﬁrst and foremost a theory of epistemic justiﬁcation, many epistemologists— Conee and Feldman included—think that knowledge requires epistemic justiﬁcation, among other things. The thesis that knowledge entails evidential justiﬁcation we can call ‘evidentialism about knowledge’. Greco, in addition to offering counter-examples to the thesis that knowledge requires good evidence, puts forward a dilemma for evidentialism about knowledge. Speciﬁcally, any version of evidentialism must be either internalist or externalist about what counts as good evidence. But internalist versions of evidentialism are false and externalist versions are unmotivated. More exactly, once we adopt externalism about what counts as good evidence, we lose any motivation for thinking that all knowledge requires evidence. On the contrary, it seems that basing one's beliefs on good evidence is just one way to acquire knowledge among many. Richard Fumerton notes some topography on the logical map that might concern fallibilist internalists like Conee and Feldman. One must be either an infallibilist or a fallibilist. Infallibilism has been common in the internalist tradition, but tends to lead to skepticism. Then one must be either an internalist or externalist. Externalists typically—though not necessarily—assert a strong connection between justiﬁcation and truth, which seems to be an advantage in their view. Fallibilist internalists, though, face a special problem. To maintain supervenience, they can either assert that propositions describing evidential connections are contingent propositions describing the subject states of a believer or that they are necessary truths trivially entailed by propositions describing the subjective states of believers. The former alternative entails

166

CONCERNS ABOUT THE INTERNALIST COMPONENT

radical subjectivism, but the latter alternative—that actually taken by Conee and Feldman—severs an important conceptual connection between justiﬁcation and truth. That is on this view it could turn out that a large proportion or perhaps even all of one’s justiﬁed beliefs are false. So if one wants such a connection—and it seems to be a good-making feature of an epistemological theory—then one is stuck with a choice between externalism and infallibilist internalism. Finally, he suggests that it obfuscates this discussion to say that all justiﬁed beliefs are epistemically likely because epistemic probability is deﬁned in terms of evidence.

10 Evidentialism about Knowledge John Greco

In this paper I argue against a position that I call ‘evidentialism about knowledge’.1 Section 10.1 articulates the position as a supervenience thesis: roughly, that the facts about epistemic normativity supervene on the facts about evidence. A corollary of this thesis is that all knowledge requires good evidence. Section 10.2 raises some initial problems for the position. The remainder of the paper puts forward a dilemma for evidentialism about knowledge. Speciﬁcally, any version of evidentialism must be either internalist or externalist about what counts as good evidence. But internalist versions of evidentialism are false and externalist versions are unmotivated. More exactly, once we adopt externalism about what counts as good evidence, we lose any motivation for thinking that all knowledge requires evidence. On the contrary, it seems that basing one’s beliefs on good evidence is just one way among many to acquire knowledge.

10.1 Evidentialism about knowledge It is now commonplace to recognize a variety of epistemic values or merits and for philosophers to specify interest in understanding one or another of these.2 Let us use the term ‘epistemic normativity’ to refer to the full normative status required for knowledge. That is, we use the term to designate the normative (or evaluative) dimension of knowledge, whatever that may be. Alternatively, we may use the terms ‘knowledge-relevant normative status’ or ‘k-normative status.’ Evidentialism about knowledge, as I will understand it here, is a thesis about knowledge-relevant normative status. Put very roughly, the thesis is that knowledgerelevant normative status is a function of what evidence one has and/or uses. Let’s call

1 The arguments in this paper are adapted from my Achieving Knowledge (Cambridge University Press, 2010), especially chapters 3 and 4. 2 For example, see Alston (1993).

168

CONCERNS ABOUT THE INTERNALIST COMPONENT

this kind of evidentialism ‘k-evidentialism’. We can state the thesis more carefully as a supervenience thesis: KE. The facts about k-normative status supervene on the facts about evidence. Among contemporary philosophers, Earl Conee and Richard Feldman come closest to explicitly endorsing k-evidentialism. First, Conee and Feldman famously endorse evidentialism about ‘epistemic justiﬁcation,’ where the latter notion marks an important part of the full normative status required for knowledge. Here are some typical passages: As we understand it, evidentialism is a view about the conditions under which a person is epistemically justiﬁed in having some doxastic attitude toward a proposition. It holds that this sort of epistemic fact is determined entirely by the person’s evidence. In its fundamental form, then, evidentialism is a supervenience thesis according to which facts about whether or not a person is justiﬁed in believing a proposition supervene on facts describing the evidence that the person has. (Conee and Feldman 2004: 204)

Again, Our bedrock view is a supervenience thesis. Justiﬁcation strongly supervenes on evidence. More precisely, a whole body of evidence entirely settles which doxastic attitudes toward which propositions are epistemically justiﬁed in any possible circumstances. That is, ES. The epistemic justiﬁcation of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at the time. (Conee and Feldman 2004: 204)

According to Conee and Feldman, epistemic justiﬁcation is only one necessary condition on knowledge. Knowledge also requires that one’s belief be ‘well founded’. Roughly, the idea is that knowledge must be based on one’s evidence as well. Here is what they say about well-foundedness: WF. S’s doxastic attitude D at t toward proposition p is well-founded if and only if (i) having D toward p is justiﬁed for S at t; and (ii) S has D toward p on the basis of some body of evidence e, such that (a) S has e as evidence at t; (b) having D toward p ﬁts e; and (c) there is no more inclusive body of evidence e0 had by S at t such that having D toward p does not ﬁt e0 . (Conee and Feldman 2004: 93) Notice, however, that Conee and Feldman understand the notion of well-foundedness in terms of ‘the facts about one’s evidence’, at least broadly construed. On the assumption that justiﬁcation and well-foundedness are the only normative conditions that Conee and Feldman place on knowledge, their position amounts to a version

E V I D E N T I A L I S M A B O U T K N OW L E D G E

169

of k-evidentialism. That is, they think that the facts about k-normative status are exhausted by the facts about one’s evidence.3 We may get clearer about evidentialism by getting clearer about the notion of evidence that it employs. What is evidence supposed to be? First, it would seem that evidence must be some sort of representational state. If evidence is to support a belief, if it is to act as a ground for believing, then it must have some appropriate content to do the grounding. Let’s characterize this sort of content as weakly as possible and call it ‘representational’ content. Secondly, evidence must be accessible to the knower in at least the following sense: it must be available for use in reasoning and inference. That is, evidence is supposed to play a functional role in our cognitive activity, but to do so it must be available to the knower in some sense appropriate to that functional role. Finally, evidence is an indication of truth. Thus Conee writes, Any epistemic support a person has for a proposition is some sort of indication to the person that the proposition is true. Indications of the truth of a proposition are evidence for it. (Conee and Feldman 2004: 252) . . . all evidence for a proposition, however weak, is some indication that the proposition is true. Thus, the sort of justiﬁcation that is constituted by evidence always bears on the truth of what is justiﬁed. (Conee and Feldman 2004: 253)

Accordingly, we can derive the following corollaries from KE: KE-1. S’s belief that p has k-normative status at t only if S’s evidence at t sufﬁciently supports p. KE-2. S’s belief that p has k-normative status at t only if S’s evidence at t is a sufﬁcient indication that p is true. And since knowledge requires k-normative status, we can derive the following corollaries about knowledge speciﬁcally: KE-3. S knows p at t only if S’s evidence at t sufﬁciently supports p. KE-4. S knows p at t only if S’s evidence at t is a sufﬁcient indication that p is true. Let us stipulate that KE through KE-4 are deﬁning theses of k-evidentialism, or evidentialism about k-normative status. As noted above, Conee and Feldman come close to endorsing KE. But in any case, I expect that KE has wide appeal, and certainly many philosophers would endorse KE-1 through KE-4. In the remainder of this chapter I will argue that k-evidentialism is false, however. That is, I will argue that all of KE through KE-4 are false. In the next section I raise some initial problems for the position. In the remainder of the chapter I pursue the dilemma described above.

3 Conee and Feldman recognize that an additional condition on knowledge is necessary for addressing Gettier problems, but it is not clear whether this would be a further condition on k-normative status as well. In what follows I will ignore that possibility.

170

CONCERNS ABOUT THE INTERNALIST COMPONENT

10.2 Two problems for evidentialism about knowledge First, the position suffers from a psychological plausibility problem. Even if some cases of knowledge seem to ﬁt the evidentialist model well, it is a stretch to extend the position to other cases, such as cases involving memory knowledge. What is my evidence, for example, that I ate eggs for breakfast, or that my car is parked in the driveway? There might be some vague phenomenology involved in remembering these things. My memory beliefs might be accompanied by a characteristic sort of feeling—a kind of conﬁdence, or perhaps a kind of attraction. But it is a stretch to think that these states are evidence—to think that these states ‘support’ my memory beliefs, or that my beliefs are evidentially based on such states.4 Neither should we think about my evidence here in terms of meta-beliefs; for example, beliefs about my memory beliefs and about their source in reliable memory. Again, it is psychologically implausible that all memory knowledge has its basis in this sort of evidence.5 Similar points apply to introspective beliefs and to beliefs about simple a priori truths. No doubt I have abilities to know such truths. But it is implausible to think of such abilities as involving reasons or evidence in the sense intended here. A second problem for evidentialism is that the position is undermined by contemporary cognitive science. Recent empirical studies make it doubtful that paradigm cases of knowledge, such as perceptual knowledge, memory knowledge, and inductive knowledge, can be understood entirely in terms of person-level representational states, as evidence is understood to be.6 First, the reliability of perception might very well depend on sub-personal inputs that cannot be appropriately understood as evidence. Blind-sight and other kinds of ‘non-conscious’ perception offer examples of this possibility. The relevant thought is that such perceptual faculties are reliable, but that their reliability is not entirely a function of person-level representations, such as perceptual appearances or ‘seemings’. Rather, reliability here depends on sub-personal inputs, so that no person-level representation, and therefore nothing counting as evidence, is itself ‘a sufﬁcient indication that p is true’. Secondly, the reliability of cognition might very well depend on processing that operates on a sub-personal or even non-representational level. Various connectionist models illustrate these possibilities. The relevant thought here is that such processing is reliable, but that its reliability is not entirely a function of operations on person-level representational contents, such as evidence and inference rules are understood to involve. If this is the case, then the reliability of such processing cannot be understood in evidentialist terms, such as evidential ‘ﬁt’ and ‘support’. 4

For an argument along these lines, see Plantinga (1993: 57–64). For an extended argument to this conclusion, see my ‘How to Preserve Your Virtue while Losing Your Perspective,’ Greco (2004). 6 See my ‘Virtues and Rules in Epistemology,’ in Greco (2001, included as chapter 2 of Achieving Knowledge). See also Horgan and Tienson (1996) and Henderson and Horgan (2000). 5

E V I D E N T I A L I S M A B O U T K N OW L E D G E

171

In any case, these are empirical questions, depending on contingent facts about human cognition. But then evidentialism is in trouble as a theory about knowledge and knowledge-relevant normativity. According to evidentialism, the facts about k-normative status entirely depend on the facts about evidence. According to our best cognitive science, the reliability of our cognitive systems might not depend just on the facts about evidence. But then evidentialism is inadequate to capture an important dimension of k-normative status. To be clear, the point is not that some non-evaluative or non-normative dimension of knowledge cannot be understood in evidentialist terms. For example, the point is not merely that some of the causal mechanisms involved in knowledge-producing cognition cannot be captured in evidentialist terms. That much is obvious, and no evidentialist would deny it. The relevant point, rather, is that evidentialism fails as a theory of knowledge-relevant normativity. That is, evidentialism fails to capture normative differences between knowing and not knowing. The reasoning here can be summarized as follows. 1. One’s belief that p has k-normative status only if one’s belief that p is reliably formed. Put differently, the facts about k-normative status partly supervene on the facts about reliability. 2. Whether one’s belief is reliably formed is not entirely a function of one’s relevant evidence and/or relevant evidential relations. Put differently, the facts about reliability do not (entirely) supervene on the facts about evidence. Therefore, 3. K-normative status is not entirely a function of one’s relevant evidence and/or evidential relations. The facts about k-normative status do not (entirely) supervene on the facts about evidence. I take it that premise 1 is relatively uncontroversial. That premise is not about subjective justiﬁcation, or epistemic responsibility, or some other epistemic merit or value falling short of k-normative status. Rather, it is about the full normative status required for knowledge, and it is relatively uncontroversial that (a) knowledge requires reliability and (b) reliability is normative. That is, it is epistemically better that one’s beliefs be reliably formed than not. A failure by way of reliability constitutes an epistemic lack or ﬂaw, and is therefore one way that a belief can fall short of the full normative status required for knowledge. Premise 2 is supported by the references to cognitive science above, which demonstrate (at least) the possibility that reliability in human cognition is not a function of evidence and evidential relations such as ‘ﬁt’ and ‘support’. But then evidentialism about k-normative status is false. The facts about k-normative status do not supervene on the facts about evidence.

172

CONCERNS ABOUT THE INTERNALIST COMPONENT

10.3 A dilemma for k-evidentialism In the remainder of the chapter I will develop a different argument against k-evidentialism. The argument takes the form of a dilemma: Any version of evidentialism must be either internalist or externalist about what counts as good evidence. More generally, any version of evidentialism must be either internalist or externalist about k-normative status. But internalist versions of k-evidentialism are false and externalist versions of k-evidentialism are unmotivated. More exactly, once we adopt externalism about k-normative status, we lose our motivation for thinking that all knowledge requires evidence. On the contrary, it seems that basing one’s beliefs on good evidence is just one way among many to acquire knowledge. Internalism and evidentialism Let’s deﬁne internalism about knowledge-relevant normative status as follows. KI. The facts about k-normative status supervene on facts that are internal to the knower. We get different versions of KI depending on how the notion of ‘internal to the knower’ is understood. One important version of KI is Privileged Access Internalism. Let’s say that S has ‘privileged access’ to the fact that p just in case S can know that p ‘by reﬂection alone’, and let’s understand this latter notion as follows: S can know that p ‘by reﬂection alone’ just in case S can know that p by virtue of introspection, or by a priori intuition, or by reasoning from truths that are known in these ways. Privileged Access Internalism about epistemic normativity can now be deﬁned as follows: I-PA. The facts about k-normative status supervene on facts to which S has privileged access. A different way to understand the notion of ‘internal to the knower’ is due to Conee and Feldman.7 According to their ‘mentalism’, a state of affairs F is internal to S just in case F constitutes part of S’s mental life. For example, that S has some particular perceptual experience counts as internal on this understanding, since how things appear perceptually to S is part of S’s mental life in the relevant sense. Also, any belief or other representation that S has about how things are would be internal on this understanding, since a person’s beliefs and other representations are also part of her mental life. Here is what Conee and Feldman say about their internalism about justiﬁcation. . . . internalism is the view that a person’s beliefs are justiﬁed only by things that are internal to the person’s mental life. We shall call this version of internalism ‘mentalism.’ A mentalist theory may assert that justiﬁcation is determined entirely by occurrent mental factors, or by dispositional ones

7 See, e.g., Conee and Feldman (2001); Feldman and Conee (1985). Both of these articles are reprinted in Conee and Feldman (2004). See also their ‘Introduction’ in that volume.

E V I D E N T I A L I S M A B O U T K N OW L E D G E

173

as well. As long as the things that are said to contribute to justiﬁcation are in the person’s mind, the view qualiﬁes as a version of mentalism. (Feldman and Conee 1985: 55)

We are here interested in internalism about knowledge-relevant normative status rather than justiﬁcation. That is, we are interested in internalism about the full normative status required for knowledge. Accordingly, we may deﬁne ‘Mental State Internalism’ as follows: I-MS. The facts about k-normative status supervene on facts about a person’s mental life. The ﬁrst thing to note about KI (understood as either I-PA or I-MS) is that it is clearly implausible. The main reason for this is that knowledge requires modally strong relations to the world, but such relations clearly do not supervene on what is internal to the knower. Philosophers disagree on which modal relations are necessary for knowledge. Many think that knowledge requires reliability. Some think that it requires some other causal connection as well or instead. Some think that what is required is safety or sensitivity, and there are variations on all of these positions. But everyone thinks that some such relation is required. In the post-Gettier world, no one thinks that knowledge can be understood as merely internal justiﬁcation and truth. But then the facts that are ‘internal to the knower’ are not sufﬁcient to ﬁx the facts about knowledge-relevant normative status. Internalism about k-normative status is false. Here is the argument put more formally, taking Privileged Access Internalism as the target. 1. Some of the necessary conditions for k-normative status are modally strong in nature. 2. But the states of affairs to which S has privileged access are not sufﬁcient to ﬁx the relevant modal facts. There are possible worlds where the facts to which S has privileged access are the same but the relevant modal facts are different. Therefore, 3. The states of affairs to which S has privileged access are not sufﬁcient to ﬁx the facts about epistemic normativity. K-normative status does not supervene on facts to which S has privileged access. An analogous argument, of course, could be run against Mental State Internalism. In fact, we may assume that any version of knowledge internalism will fall to an argument along these lines. That is because, on any reasonable interpretation of ‘internal to the knower’, the relevant modal facts will not entirely supervene on facts that are internal to the knower. Another way to see the point is to note an important difference between k-evidentialism and the sort of internalism being considered here. Both involve supervenience theses: KE says that k-normative status entirely supervenes on the facts

174

CONCERNS ABOUT THE INTERNALIST COMPONENT

about one’s evidence, broadly construed, whereas KI says that k-normative status entirely supervenes on what is internal to the knower. But notice that the supervenience base speciﬁed by KE is much broader than that speciﬁed by KI. For example, the facts about one’s evidence, broadly construed, go beyond the facts about one’s mental life. Presumably, the former will include facts about how one’s evidence is related to the world, thereby making KE a far more plausible thesis than KI. The anti-internalist claims in this section are therefore relatively uncontroversial, insofar as they are claims about knowledge-relevant normative status, or the full normative status required for knowledge. Internalisms that are defended in the literature are usually about some different kind of normative status—either a kind not required for knowledge at all, or a kind marking some dimension of k-normative status. Insofar as knowledge requires some right relation to the world, and insofar as this relation must be modally stronger than what can be guaranteed by facts that are ‘internal to the knower’, it should be unsurprising that internalism about k-normative status turns out to be implausible. Evidentialism wedded to internalism about k-normative status is implausible, because internalism about k-normative status is implausible. But now we may ask: What happens when evidentialism is wedded to externalism about k-normative status? Here I want to argue that evidentialism loses its motivation. Once we adopt externalism about k-normative status, we lose any motivation for thinking that all knowledge requires evidence. That will be the argument of the next section. Externalism and evidentialism We have seen that internalism about k-normative status is implausible. Why so? Because it is clear that knowledge requires some modally strong relation to the world, and that this relation itself falls within the epistemically normative domain. If a belief fails to be so related to the world, we think that this constitutes an epistemic lack, that is a lack which makes the belief fall short of knowledge, even if true. It is important to note here that the evidentialist about k-normative status can agree with this much. That is, the evidentialist can agree that k-normative status involves some modally strong relation to the world, but she will think that good evidence grounds exactly that relation. For example, one might be an evidentialist about knowledge and a reliabilist about what makes for knowledge-producing evidence. One’s position would then be that knowledge requires good evidence, and that good evidence is reliable evidence. Basing one’s belief on good evidence thereby insures the modal relation that knowledge requires. Alternatively, one might be a safety theorist or a sensitivity theorist about evidence: good evidence is safe (or sensitive) evidence, and basing one’s belief on such insures the relevant modally strong relation to the world. But now consider: If one is a reliabilist (or a safety theorist, or a sensitivity theorist) about evidence, why should evidence itself be so important? The natural thought is that reliability (or safety, or sensitivity) is important, and that evidence is merely a means for achieving the relevant relation. If one’s belief could be made reliable (or safe or

E V I D E N T I A L I S M A B O U T K N OW L E D G E

175

sensitive) in some other way, this would serve the relevant normative demands of knowledge just as well. Again, once we adopt externalism about knowledge-relevant normative status— once we are reliabilists, or causal theorists, or safety theorists about such status—it is hard to see why evidence itself should be so important. At most, it would seem, evidence is required for knowledge because it is required for reliability, or safety, or some other modally strong relation to the world. But why think even that is the case? Certainly it is not a necessary truth that, for example, reliable cognition must employ reliable evidence. We can imagine cognition that reliably relates belief to the world, but not by means of reliable evidence. At best, it is a contingent fact about human cognition that reliable cognition always employs reliable evidence. But the contingent facts here are dubious, and in any case are insufﬁcient to ground a theory of epistemic normativity. The evidentialist about knowledge wants to say that necessarily all knowledge requires truth-indicating evidence. KE through KE-4 are intended to have the sort of modal strength that is appropriate to a philosophical theory, and so a merely contingent requirement on knowledge will not serve evidentialist purposes.

10.4 A different motivation for k-evidentialism? Evidentialism about knowledge is not well motivated by an externalist theory of knowledge-relevant normative status. Might there be some other sound motivation? Two possibilities come to mind. The ﬁrst is that evidentialism is required for a ‘satisfying’ reply to skepticism. The second is that evidentialism is motivated by the role played by knowledge in social discourse. Replying to skepticism What would it take to have a ‘satisfying’ reply to skepticism? One might think that it would take having a satisfying answer to the skeptical question, ‘How do you know?’, where this in turn would require having adequate reasons for one’s belief in every instance. That is, one might think that having a reply to skepticism requires always being able to give good reasons for one’s belief, and that this entails an evidentialist condition on knowledge. But can we be externalists about good evidence and still hope to give a satisfying reply to skepticism? The skeptic in our imagined dialectic, of course, will demand not only that we show reasons, but that we show that our reasons are good ones. For example, the skeptic will demand that we show that our reasons are reliable. And this, in turn, will give rise to Humean and Pyrrhonean problems that are all too familiar. I will not rehearse those problems now (see instead Greco 1999, 2000). Sufﬁce it to say that externalists escape the problems by rejecting the demands straight away. Internalists typically hold out hope for meeting the demands, but that is irrelevant in the present context. For here we are working within an externalist framework, where the prospects for a ‘satisfying’ reply to skepticism have already been given up. That is,

176

CONCERNS ABOUT THE INTERNALIST COMPONENT

we are operating in a context where the skeptic’s completely general demands to show good reasons are rejected from the start. The need to satisfy such demands, then, cannot here generate a motivation for evidentialism. The social function of knowledge A number of contemporary epistemologists endorse a broadly Sellarsian theme: that knowledge properly so-called involves abilities to articulate and give one’s reasons, and to defend one’s knowledge and reasons against relevant objections (e.g. Lehrer 2000). Here the concern is not primarily with skepticism. Rather, the idea is that knowledge has a role in science and in practical reasoning, and that this role requires social abilities involving articulation, reason-giving, and defense against objections that appropriately arise in relevant contexts. Hence knowledge must be grounded on available reasons, as evidentialism requires.8 Thus Keith Lehrer writes: It is fundamental to . . . human knowledge . . . that it is inextricably woven into reasoning, justiﬁcation, conﬁrmation, and refutation. It is required both for the ratiocination of theoretical speculation in science and practical sagacity in everyday life. To do science—to engage in experimental inquiry and scientiﬁc ratiocination—one must be able to tell whether one has correct information or not. . . . Engaging in law or commerce requires the same sort of knowledge, which may be used as the premises of critical reﬂection or claimed as the prizes of informed reasoning. (Lehrer 2000: 6)

This all sounds reasonable enough, but it again raises a question about psychological plausibility. Namely, is it plausible that we typically have the sort of evidence that Lehrer requires? That is, do we always have it, in every case of knowledge? Lehrer’s answer to this concern is to understand the relevant evidence in terms of acceptances, and to understand acceptances in terms of their functional role in reasoning, conﬁrmation, etc. (Lehrer 1989 and 2000: esp. 39–40). Hence, for S to accept that p does not require that S explicitly judges that p. Rather, it requires only that S act in certain ways when in certain circumstances. Lehrer writes, a person may be said to have information he cannot easily describe and to employ such information in various ways. For example, suppose I know the shortest route from Rochester to Buffalo, though I cannot tell you the name of the highway. Moreover, imagine that I am not very good at giving directions, so I cannot tell you how to get from Rochester to Buffalo. . . . That I make the trip successfully on many occasions shows that I have the required information, which I might ﬁnd difﬁcult to articulate, about the route from Rochester to Buffalo. Similarly, my reliability in accepting that I see when I do, or even in accepting that I feel or think when I do, depends on my ability to employ information, which I might ﬁnd difﬁcult to articulate, about seeing, feeling, and thinking. (Lehrer 2000: 74–5)

8 This is not a line of argument that would be endorsed by Conee and Feldman, who explicitly reject the idea that evidence must be publicly available. (See, e.g., Conee and Feldman 2004: 2–3.)

E V I D E N T I A L I S M A B O U T K N OW L E D G E

177

The problem is now this: The present motivation for evidentialism is that knowledge requires evidence for the purposes of reasoning, justiﬁcation, conﬁrmation, and refutation, where these are understood as social activities played out among our epistemic peers. But now, it turns out, the evidence that supposedly grounds knowledge need only be implicit and inarticulate, and therefore need not be available for use in these social activities. Sellarsian themes regarding the role of knowledge in social discourse do not motivate an evidentialist condition on knowledge after all. What about ‘internal discourse’ rather than social discourse? Might knowledge require evidence because knowledge requires reason giving and the like in foro interno? Perhaps this is the better way to interpret Lehrer’s thought that reasons are required for ‘science and practical sagacity’. But here we depend on an analogy between giving reasons to others in a social context and using reasons oneself in one’s own thinking. Insofar as one’s ‘reasons’ are only implicit and unarticulated, however, the analogy breaks down and so the requirement for reasons (or evidence) properly so-called is undermined. The close connection between knowledge and giving reasons, whether in social contexts or in foro interno, does not generate a general evidentialist requirement.

10.5 Conclusion In conclusion, the prospects for evidentialism about knowledge look bleak. First, both introspection and cognitive science suggest that not all knowledge ﬁts well into an evidentialist model. Second, the evidentialist faces a dilemma regarding the nature of knowledge-relevant normative status: internalist versions of evidentialism are false because internalism is false, and externalist versions of evidentialism are unmotivated. Finally, neither concerns about answering skepticism nor the close connection between knowledge and reason-giving motivate a general evidentialist requirement on knowledge.9

References Alston, William. 1993. ‘Epistemic Desiderata,’ Philosophy and Phenomenological Research, 53, 3: 527–51. Conee, Earl and Richard Feldman. 2001. ‘Internalism Defended,’ in Hilary Kornblith (ed.), Epistemology: Internalism and Externalism. Oxford: Blackwell Publishers. —— and ——. 2004. Evidentialism: Essays in Epistemology. Oxford: Oxford University Press. Feldman, Richard and Earl Conee. 1985. ‘Evidentialism,’ Philosophical Studies, 48: 15–34.

9 I would like to thank audiences at Baylor University, Purdue University, Washington University, and the Central States Philosophical Association for comments and suggestions that improved the paper. Thanks especially to Michael Bergmann, Trent Dougherty, Mylan Engel, John Heil, Jon Kvanvig, Bruce Russell, and Matthias Steup.

178

CONCERNS ABOUT THE INTERNALIST COMPONENT

Greco, John. 1999. ‘Agent Reliabilism,’ in James Tomberlin (ed.), Philosophical Perspectives, 13, Epistemology. Atascadero, CA: Ridgeview Press. ——. 2000. Putting Skeptics in Their Place. New York: Cambridge University Press. ——. 2001. ‘Virtues and Rules in Epistemology,’ in Abrol Fairweather and Linda Zagzebski (eds), Virtue Epistemology: Essays on Epistemic Virtue and Responsibility. Oxford: Oxford University Press. ——. 2004. ‘How to Preserve Your Virtue while Losing Your Perspective,’ in John Greco (ed.), Sosa and his Critics. Oxford: Blackwell. Henderson, David and Terence Horgan. 2000. ‘Iceberg Epistemology,’ Philosophy and Phenomenological Research, 61, 3: 497–535. Horgan, Terence and John Tienson. 1996. Connectionism and the Philosophy of Psychology. Cambridge, MA: MIT Press. Lehrer, Keith. 1989. ‘Replies’ in John Bender (ed.), The Current State of the Coherence Theory. Dordrecht: Kluwer Academic Publishers. ——. 2000. Theory of Knowledge, 2nd edn. Boulder: Westview Press. Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford University Press.

11 Evidentialism and Truth Richard Fumerton

One might well begin a discussion of epistemic justiﬁcation by wondering why a philosopher should be particularly interested in the concept. And perhaps the most natural answer is that justiﬁcation is important because it has some sort of connection to truth. After all, as cognitive agents our primary concern is to arrive at the truth concerning matters about which we are curious. And if having justiﬁed beliefs makes it no more likely that we have true beliefs, why would we care whether or not we have justiﬁed beliefs? And there lies a superﬁcially powerful argument for various versions of externalism. The externalist will often proclaim that only externalist accounts of justiﬁcation manage to secure the crucial connection between having justiﬁed beliefs and having beliefs that are (mostly) true. If this claim is correct, versions of evidentialism of the sort defended by Conee and Feldman might face a powerful objection. Ironically, this very feature of externalism that its proponents claim as a strength is also one if its great vulnerabilities. Having too strong a connection to truth is a doubleedged sword. First, it is surely a datum that there can be epistemically justiﬁed false beliefs. I can have powerful evidence that your name is ‘Fred’ when it isn’t, that I was born on 7 October when I wasn’t, that I will arrive safely in Chicago tomorrow when I won’t, and so on. If these justiﬁed beliefs can be false, then why can’t most or, even all of my justiﬁed beliefs be false? An afﬁrmative answer to the question posed in the consequent of the above conditional doesn’t follow, of course, from its antecedent. Such a presupposition commits the fallacy of composition. But the fallacy of composition isn’t one of those charges that one should just wave around in dismissing an argument. There is, to be sure, such a fallacy. One can’t infer that my car weighs under 200 pounds, because each of its components weigh under 200 pounds. One can’t infer that it is possible for (p and not-p) to be true because it is possible for p to be true and it is possible for not-p to be true. But on the other hand there does seem to be an intimate connection between all of the components of my car being made of metal and my car being made of metal. And if p and q are logically independent propositions, and it is possible for p to be true and it is possible for q to be true, then surely it is possible for p and q to be true. One needs an explanation of why the world won’t allow most justiﬁed beliefs to be false even if it allows each individual justiﬁed belief to be false.

180

CONCERNS ABOUT THE INTERNALIST COMPONENT

Additionally, most evidentialists (and, more generally, most internalists) will argue against many forms of externalism by describing hypothetical situations in which it seems to them obvious that there can be subjects enjoying perfectly rational, justiﬁed beliefs despite the fact that through deception or bad luck they are doomed to massive error. The so-called ‘new evil demon problem’ is a thorn in the side of most externalists. After seeing movies like Total Recall or The Matrix it is even easier now than it has ever been to imagine victims of massive hallucination. Computers have replaced evil demons, but the plot is essentially the same. In The Matrix the subjects undergoing vivid hallucination surely have reached perfectly reasonable conclusions about their whereabouts even though such conclusions are nearly always false. The new evil demon problem is most acute for reliabilists—philosophers who try to deﬁne justiﬁed beliefs as reliably produced beliefs, and who understand, in turn, reliablyproduced beliefs as beliefs that are usually true relative to the way in which they were formed (at least when the input beliefs are true).1 But variations on the problem of massive deception will face just about any version of externalism. Michael Bergmann, for example, eschews reliabilism for the view that beliefs are epistemically justiﬁed when (in part) they are produced by processes that are functioning as they were originally designed to function (2006: 133). The process in question need only be reliable when operating in the environment for which it was designed. Bergmann can, then, allow that one can move into a demon world and still have justiﬁed beliefs provided that our dispositions to respond to sensation with external-world beliefs were designed for a non-demon environment. It doesn’t matter how long one is deceived, or how often one falls into error—the justiﬁcatory status of the beliefs remains positive because the mechanism that produces them is still conforming to its design plan—a design plan that was reliable in the environment for which it was designed. But a view like Bergmann’s might be accused of having the worst of two worlds. It fails to secure a tight connection between having justiﬁed beliefs and having mostly true beliefs, and it is still vulnerable to a variation on the evil demon argument. The fundamental intuition driving the aforementioned objection to externalism is the thought that cognitive agents who are massively deceived in the way envisioned will still have justiﬁed beliefs. And we can easily imagine a world designed by a demon to deceive people. Why would our intuition be any less strong that the victims designed and chosen for deception nevertheless still have justiﬁed beliefs? In what follows I want to explore more carefully the implications of various views for the connection between justiﬁcation and truth, and address the more important question of what one should take the relevant connection to be.

1 As we’ll discuss in more detail below, the reliabilist will almost certainly distinguish between noninferential and inferential justiﬁcation, and further will struggle to deﬁne more precisely (and more plausibly) the critical notion of reliability.

E V I D E N T I A L I S M A N D T RU T H

181

11.1 Some terminological issues This chapter focuses on evidentialism and its discontents. You will already have noticed that I have been playing rather loose with the term ‘evidentialism’, often using it almost interchangeably with the term ‘internalism’. I’ve always been a bit uncomfortable with the term ‘evidentialism’ and its corresponding slogan that it is only one’s evidence that justiﬁes one’s beliefs. Conee and Feldman, the philosophers most closely associated with the label, are clearly sympathetic to a version of foundationalism according to which there are noninferentially justiﬁed beliefs. To be sure, any plausible version of foundationalism will still hold that there is some feature of a noninferentially justiﬁed belief that gives it that special epistemic status, but it is often a bit strained to characterize the feature in question as evidence possessed by the believer. As I understand the current debate, self-proclaimed evidentialists are certainly on the side of internalism when it is understood a certain way. Unfortunately, as many have illustrated in some detail, there is no one universally accepted way of understanding the internalism/externalism debate in epistemology. Painting with a broad stroke, we should distinguish at least the following versions of internalism (and corresponding versions of externalism): Access internalism The access internalist is committed to the view that something J can justify S in believing p only if S has actual or potential access to the fact that J is a justiﬁer of S’s belief that p. (Where we distinguish one’s possessing evidence and the fact that the evidence makes likely p for S, I am using J to refer to the conjunctive fact that S possesses the evidence and that it is does make likely p for S.) The access in question is often taken to be direct access in the form of introspection coupled with, perhaps, a priori knowledge of evidential connections. Almost all versions of access internalism ﬂirt with vicious regress and the evidentialists we are considering here reject the view. Mentalism or internal state internalism Conee and Feldman introduce the term ‘mentalism’ to refer to the version of internalism they accept (2001). The heart of the view is that the epistemic status of a belief supervenes exclusively on the mental states of the believer. The strongest version of the view holds that the relevant epistemic properties of a belief at a time supervene exclusively on the believer’s mental states at that time. A weaker version of mentalism might allow that one’s past mental states can still play an epistemic role in justifying present beliefs. There are two potentially misleading features of this initial characterization of the view. The ﬁrst is that it seems to leave out of consideration the critical evidential connections between justiﬁcation and what it justiﬁes. But as we will see, Conee and Feldman could take evidential connections between the relevant mental states and what they justify to hold necessarily. On such a view, because a necessary truth is

182

CONCERNS ABOUT THE INTERNALIST COMPONENT

entailed by everything, there is a sense in which if mentalism is true, the epistemic status of a subject’s belief will still be entailed by the mental states themselves. The second complication concerns the ongoing internalism/externalism controversy concerning the content of mental states themselves. It is almost the received view now (though it is, I am convinced, mistaken) that the content of intentional states is ontologically ﬁxed in part by features of the world external to the subject in that intentional state. If a ‘mentalist’ in epistemology is a content externalist, then one must worry that the ‘internalism’ espoused by the mentalist might be an externalism in disguise. It is for that reason that I prefer the less elegant locution ‘internal state internalism’ to refer to the kind of view that I think Conee and Feldman have in mind by ‘mentalism’. For these purposes we can understand S’s internal states as S’s of possession states of affairs that can obtain even when no spatio-temporal object other than S exists. Inferential internalism In what follows I’ll also make reference to a view that I call inferential internalism. This is a view about what is necessary for inferential justiﬁcation. The inferential internalist claims that for S to be inferentially justiﬁed in believing p on the basis of E, S must be aware that E makes likely p (where E’s entailing p can be viewed as the upper limit of E’s making likely p). Because the inferential internalist emphasizes the importance of access to evidential connections the view can easily be confused with access internalism. But it is quite distinct. Inferential internalists may deny that one needs access to noninferential justiﬁcation in order to possess it, and, indeed, may deny that one needs access to inferential justiﬁcation in order to possess it. The view requires only access to the connection between one’s premises and one’s conclusion in order for one to acquire inferential justiﬁcation for believing the conclusion on the basis of justiﬁed belief in the premises. With these distinctions in place, let us return to questions concerning the connection between having justiﬁed beliefs and having mostly true beliefs.

11.2 Noninferential justiﬁcation and true belief: internalist and externalist perspectives There is considerable irony in the fact that most contemporary externalists stress the importance of securing a connection between justiﬁcation and truth, while their internalist critics often attack externalism by denying the relevant connection. Descartes, the patron saint of many internalists, was clearly obsessed with ﬁnding an internal feature of beliefs that distinguishes the true from the false. His search for the foundations of an ideal system of knowledge just was the search for beliefs supported in such a way that the support precludes the possibility of error. If we think of the foundations he wanted as noninferentially justiﬁed beliefs, then he seemed to want being noninferentially justiﬁed to entail being true.

E V I D E N T I A L I S M A N D T RU T H

183

It is not only the desire for certainty that has led some philosophers to seek infallible foundations. Some philosophers (Lewis 1946; McGrew 1999) have argued that there must be epistemically infallible beliefs if there is to be any weaker sort of justiﬁcation. The gist of the claim is that a claim about epistemic probability is always implicitly relativized to evidence. So p can be probable for S only relative to some other proposition E1. If E1 is itself merely probable it will of necessity be probable only relative to some other proposition E2, and so on ad inﬁnitum. The whole point of foundationalism is to avoid such regresses, so to ﬁnd a foundation that accomplishes its goal we are going to need to ﬁnd propositions that are not merely probable for us. The argument has never seemed to me plausible. As we shall see shortly, there are versions of internalist foundationalism that seek to explain fallible foundational justiﬁcation without relativizing the merely probable foundation to some other known or justiﬁably believed proposition. There are different ways of embracing an infallibilist account of foundationally justiﬁed belief. On one approach, the search is for a belief the mere existence of which guarantees its truth. The paradigm for Descartes is, of course, your belief that you exist. On another approach, however, the search is for a kind of justiﬁcation one might possess where the justiﬁcation guarantees the truth of what is believed. So on my view, for example, the paradigm of noninferential justiﬁcation consists in direct acquaintance with a correspondence between thought (the bearer of truth value) and the fact that is the truth-maker for that thought. Since direct acquaintance is a real relation that guarantees the existence of its relata, it is trivially true that one’s being acquainted with the fact that p entails that p is true. Furthermore, the entailment is nontrivial in the sense that the truthmaker is a constituent of the justiﬁcation. This last point helps deal with the technical problem that all necessary truths are (again, trivially) entailed by everything—it allows us to introduce the concept of an epistemically relevant entailment. Internalists aren’t the only philosophers trying to build truthmakers into the justiﬁcation for foundationally justiﬁed belief. Armstrong’s (1973) externalist paradigm of foundational knowledge is a belief caused by the fact that makes it true. If we extend the concept of justiﬁcation to this property of a belief, then, again, the justiﬁcation for the belief will (trivially) entail the truth of the belief. So the internalist/externalist controversy doesn’t necessarily divide philosophers over the connection between noninferential justiﬁcation and truth. Each side can maintain a very tight connection—each side can identify noninferential justiﬁcation with infallible justiﬁcation. Furthermore, just as both internalists and externalists might identify noninferential justiﬁcation with justiﬁcation that guarantees true belief, so also each side may have the resources to weaken the connection between noninferential justiﬁcation and truth. Most obviously, perhaps, the reliabilist can allow for noninferential justiﬁcation that is no stronger than the weakest of inferential justiﬁcation. On standard versions of reliabilism a belief is noninferentially justiﬁed when it is produced by a reliable process where either the input to the process is something other than a belief (e.g. a sensation),

184

CONCERNS ABOUT THE INTERNALIST COMPONENT

or where the input includes belief states, but where the epistemic status of the input belief states is irrelevant to the epistemic status of the ‘output’ belief (e.g. introspection of what one believes). In Goldman’s classic original presentation of reliabilism, he described the processes that yield noninferentially justiﬁed beliefs as unconditionally reliable belief-independent processes. The process is described as unconditionally reliable because its reliability isn’t conditional on the input beliefs being true. On the crudest characterization of the view, an unconditionally reliable process is simply one that produces mostly true beliefs. A process can be reliable in this sense if just over half of the beliefs it produces are true. But we still appear to have a tight conceptual connection between having noninferentially justiﬁed beliefs and having mostly true beliefs. Appearances can be deceiving, however. As Goldman and others would be quick to admit, everything depends on one’s characterization of reliablility. As I indicated above, the crudest version of the view deﬁnes an unconditionally reliable process as one that produces mostly true beliefs. Such a view secures the connection between having noninferentially justiﬁed beliefs and having mostly true beliefs, but it also faces the same objections encountered by equally crude characterizations of probability deﬁned in terms of relative frequency. On the crudest relative frequency conception of probability, the probability of a’s being G relative to its being F is simply a function of the actual percentage of F’s that are G. But on such a view, if there is only one silver coin (with a head and a tail) and it is tossed only once, then the probability of its coming up heads relative to its being a silver coin that was tossed will be either 1 or 0. And that’s an absurd consequence of the view. In the same way, the reliabilist must allow for the possibility of a belief-independent process that kicks in only once producing either a true or false belief. It would be equally absurd to suppose that the process in question was, for that reason, either completely reliable or completely unreliable. The moral seems to be that one must abandon actual frequencies in deﬁning both the relative frequency conception of probability, and the critical concept of reliability needed to analyze the base clause in a reliabilist’s recursive analysis of justiﬁcation. There are, of course, alternatives. Most obviously one can turn to counterfactuals. What makes a belief-producing process reliable (whether or not it ever actually processes any input) is that it would probably produce mostly true beliefs if it were used indeﬁnitely many times, where the percentage in question would continue to converge on some particular percentage the more the process were employed. Turning to counterfactuals takes care of some of the most obvious counterexamples, but still, of course, faces the new evil demon problem. The existence of the unrelenting demon makes it true not only that the actual frequency with which perceptually-based beliefs are true is very low—it also makes it true that beliefs produced that way would continue to be false no matter how many times the process were employed. But in addition, counterfactual reliability severs a necessary connection between having noninferentially justiﬁed beliefs and having mostly true beliefs. Just as one can toss a fair coin 20 times and have it turn up heads all 20 times, so also on this conception of

E V I D E N T I A L I S M A N D T RU T H

185

justiﬁcation one can, in principle, have indeﬁnitely many justiﬁed false beliefs in the actual world. As we noted earlier, Bergmann’s proper-functioning account of epistemic justiﬁcation doesn’t avoid variations on the demon problem and, importantly for the present discussion, fails to preserve not only an actual connection between justiﬁcation and truth, but also fails to preserve even a counterfactual connection between justiﬁcation and truth. And so it goes for other attempts to avoid counter-examples. Goldman himself once tried to deﬁne epistemic justiﬁcation in terms of ‘normal worlds’ reliability (1986). Roughly, normal worlds are worlds in which things are fundamentally as they are believed to be in this world. Reliability is deﬁned in terms of truth ratio of output beliefs in normal worlds. But given that this might not be a normal world, there is obviously no necessary connection on this account between having justiﬁed beliefs and having mostly true beliefs, or even having justiﬁed beliefs and having mostly true beliefs in ‘close’ possible worlds. David Henderson and Terrence Horgan (2007) have toyed with the idea of introducing the notion of transglobal reliability (a kind of super reliability) in deﬁning a concept of epistemic justiﬁcation. A process is transglobally reliable if it yields mostly true beliefs in most possible worlds. Again, justiﬁcation deﬁned in terms of transglobal reliability allows for the possibility that most actual justiﬁed beliefs are false. It even allows for the possibility that most justiﬁed beliefs in close possible worlds are false. I suppose the idea is that one can still take comfort in the fact that it is unlikely (in a frequency sense) that the actual world or its close counterparts are worlds in which one has mostly false justiﬁed beliefs. But as Henderson and Horgan know full well, the very notion of transglobal reliability is problematic. For any belief-producing process that is only contingently (actually or counterfactually) reliable there are inﬁnitely many possible worlds in which the process is (actually or counterfactually) reliable and inﬁnitely many possible worlds in which that same process is (actually or counterfactually) unreliable. It’s hard to understand the sense in which any such process could be reliable in ‘most’ possible worlds. To be honest, however, I actually agree that there is intuitive sense we can make of the critical concept upon which Henderson and Horgan rely. Suppose, for example, that someone is asked to randomly select a number. Intuitively, there are more possible worlds in which the number selected is not a prime number than there are possible worlds in which the number selected is prime, and that is the case even though there are inﬁnitely many possible worlds in which the number selected is prime, and inﬁnitely many possible worlds in which the number selected is not prime. But I suspect that our intuitive understanding of this sort of talk is parasitic upon our prior understanding of epistemic probability. We believe that it is unlikely relative to our evidence that a person will select a prime number and that is all we mean by saying that there are ‘more’ possible worlds in which that doesn’t happen. If I’m right about this, it is not even clear that justiﬁcation deﬁned in terms of transglobal reliability is an alternative to evidentialism (as deﬁned above). It is interesting to note that both

186

CONCERNS ABOUT THE INTERNALIST COMPONENT

Horgan and Henderson and the evidentialists are going to end up committed to the claim that claims of epistemic probability are necessary truths. I have argued that paradigmatic externalists can attempt to understand noninferential justiﬁcation without building the truth conditions for a belief into the having of noninferential justiﬁcation. But I have also argued that plausible attempts to do so run the danger of severing the connection between having noninferentially justiﬁed beliefs and having mostly true beliefs. But a similar problem will probably affect the internalist who moves towards a ‘modest’ foundationalism that allows for fallible noninferntial justiﬁcation. So I have argued, for example, that one might allow that one can have a noninferentially justiﬁed false belief that p when one is directly acquainted with the thought that p and a ‘quasi-correspondence’ between that thought and a fact very similar to the fact that p. At the very least it seems to me plausible to suppose that one can recognize that beliefs under such circumstances enjoy more justiﬁcation than other obviously absurd beliefs.2 If one allows that direct acquaintance with a fact other than the truthmaker for a belief can yield noninferential justiﬁcation, it is hard to see how one could put an a priori limit on the number of such noninferentially justiﬁed beliefs. And once again, one runs the danger of severing a conceptual connection between having noninferentially justiﬁed beliefs and having mostly true beliefs. To be sure, the acquaintance theorist might argue that there is still a connection between having noninferentially justiﬁed beliefs and having beliefs that are at least ‘almost’ true, but being almost true isn’t as good as being true. Of course, the attempt to understand noninferential justiﬁcation in terms of direct acquaintance is only one version of evidentialism (internal state internalism). Conee himself (2004) seems to be sympathetic to a version of epistemic conservatism. Like Mike Huemer (2001), Conee appears to think that the mere fact that it seems to one as if p can (noninferentially) justify one in believing p. And if this ‘seeming’ is an intentional state that carries no guarantee of truth, there seems to be no reason why there can’t be indeﬁnitely many noninferentially justiﬁed, but false beliefs. Indeed, more generally, any version of mentalism that allows a mental state to confer noninferential justiﬁcation on a belief when the mental state does not include as a constitutent the truthmaker for the belief will, I believe, be committed to the view that one can have mostly false noninferentially justiﬁed beliefs.3 So where are we? If we want a strong connection between noninferential justiﬁcation and truth, who has the upper hand—the externalist or the evidentialist (internal state internalist)? We have seen that there are versions of both internalism and externalism that secure the tightest connection possible between noninferential justiﬁcation and truth. We have also seen that there are also versions of both views that allow for fallible noninferential justiﬁcation and that even plausible reliabilism allows for the 2

See also Fales (1996). And this includes a plethora of mentalist accounts of noninferential justiﬁcation. Chisholm’s foundationalism, for example, seems to allow for the possibility of indeﬁnitely many noninferentially justiﬁed beliefs. 3

E V I D E N T I A L I S M A N D T RU T H

187

possibility that most noninferentially justiﬁed beliefs are false. It is not clear that either side in this dispute is better positioned to develop an intimate conceptual connection between having noninferentially justiﬁed beliefs and having mostly true beliefs. Of course, any philosopher who has introduced the possibility of weak (fallible) noninferential justiﬁcation can introduce a technical notion of epistemic probability. One can argue that a noninferentially justiﬁed belief is, by deﬁnition, a belief that is epistemically likely to be true. So even if one’s noninferentially justiﬁed beliefs are mostly false, they can still be epistemically likely. But the technical concept of epistemic likelihood just obfuscates the issue we are discussing here. We were concerned with the question of who has the upper hand when it comes to securing a connection between having epistemically justiﬁed beliefs and having mostly true beliefs. If our beliefs can be ‘likely’ to be true while most of them are false, we have still lost the relevant connection between epistemic justiﬁcation and truth.

11.3 Inferential justiﬁcation and truth: internalist and externalist perspectives While traditional (internalist) foundationalism can attempt to secure a tight connection between noninferential justiﬁcation and truth, it will have a much more difﬁcult time getting the connection between inferential justiﬁcation and truth. At least this is so if the foundationalist allows for the possibility of fallible inferential justiﬁcation. And any philosopher who doesn’t allow for the possibility of fallible inferential justiﬁcation will inevitably reach a relatively radical skeptical conclusion. Earlier, I talked about inferential internalism—the view that one can be inferentially justiﬁed in believing p on the basis of E only if one can be aware that there is an evidential connection between E and p. The view isn’t very popular these days (though it was once almost presupposed by most epistemologists, certainly those who took the challenge of skepticism seriously). But if one accepts the principle of inferential justiﬁcation, one can avoid a vicious regress only if one allows that there are propositions describing evidential connections that can be known a priori. In short one would need to embrace a Keynesian notion of epistemic probability according to which claims that one proposition makes probable another are synthetic necessary truths knowable a priori. The only way that one can construe a probability claim as a necessary truth knowable a priori is to strip it of empirical content. And if one strips the claim of empirical content one severs the conceptual connection between having inferentially justiﬁed beliefs based on evidence and having inferentially justiﬁed beliefs that are mostly true.4 Although the evidentialists/mentalists might well reject the principle of inferential justiﬁcation as requiring too much for inferential justiﬁcation, they too are likely to end

4

For further discussion of this issue, see Fumerton (1995: ch. 7) and Russell (1948).

188

CONCERNS ABOUT THE INTERNALIST COMPONENT

up committed to the view that propositions describing evidential connections are necessary truths. We saw earlier that the evidentialist is committed to the view that the epistemic justiﬁcation one possesses for one’s beliefs strongly supervenes on one’s (internal) mental states. The mentalist or internal state internalist’s claim covers both noninferential and inferential justiﬁcation. The internal state internalist will presumably hold that the existence of inferential justiﬁcation requires the existence of the relevant evidential connections. Consequently, the internal state internalist must secure a necessary connection between one’s internal states and the existence of these necessary connections. There are only two ways truths describing evidential connections could be entailed by truths about internal states. The ﬁrst is the highly implausible view that propositions describing evidential connections just are, somehow, descriptions of the subjective states of individuals. One might hold, for example, that E is evidence for S to believe p when S takes E to make probable p (where the relevant notion of probability would need to be independently deﬁned), or when S believes p as a result of believing E (where the causal connection itself supervenes strongly on internal states). Such views require a radical relativism and subjectivism about what constitutes evidence for what and most evidentialists have little interest in defending a view so radical. As we noted earlier, the other way that propositions describing evidential connections could be entailed by truths describing internal states is if such propositions are themselves necessary truths of the sort needed by the inferential internalist. On such a view, truths describing evidential connections would trivially be entailed by truths describing the internal states of believers (indeed would trivially be entailed by all truths). It is precisely because the entailment is trivial that one could still maintain, as the mentalist does, that both noninferential and inferential justiﬁcation strongly supervene on the internal states of believers. The point can be summarized easily. The existence of appropriate probabilistic connections (or on the inferential internalist’s view, justiﬁed belief in the existence of such connections) is a necessary condition for one’s possessing inferential justiﬁcation. The truth of such propositions is either contingent or necessary. If such propositions are contingent and their content goes beyond a description of the subjective states of believers then the mentalist’s thesis is false. But only the most radical versions of subjectivism and relativism would hold that propositions describing evidential connections are contingent propositions describing the subjective states of a believer. The only other alternative consistent with mentalism is the view that propositions describing evidential connections are necessary truths trivially entailed by propositions describing the subjective states of believers. But a view that strips propositions describing evidential connections of empirical content severs a necessary connection between having inferentially justiﬁed beliefs and having mostly true beliefs. At ﬁrst glance, then, an externalist, in particular, a reliabilist seems to have the upper hand when it comes to the possibility of securing a connection between having inferentially justiﬁed beliefs and having mostly true beliefs. But as we saw earlier,

E V I D E N T I A L I S M A N D T RU T H

189

appearances can be deceiving. Consider again the externalist view most explicitly concerned with establishing a connection between justiﬁcation and truth—process reliabilism. As we saw earlier, the reliabilist cannot plausibly identify the reliability of a noninferential belief-forming process with the actual frequency with which it generates true beliefs. The same, of course, is true when it comes to deﬁning the reliability of what Goldman (1979) once called a belief-dependent, conditionally reliable process. Turning to counterfactual reliability, proper function, normal worlds, or transglobal reliability severs the connection between having inferentially justiﬁed beliefs and having mostly true beliefs. Again, both sides can muddy the waters by appealing to a technical concept of epistemic probability deﬁned in terms of epistemic justiﬁcation. In one perfectly clear sense, p is epistemically probable for S when S is justiﬁed in believing p relative to S’s epistemic situation. The connection between epistemic justiﬁcation and truth is then deﬁned in terms of epistemic probability. An epistemically justiﬁed belief is, by deﬁnition, one that is epistemically likely to be true. There is therefore a connection between having justiﬁed beliefs and having beliefs that are likely (epistemically likely) to be true. But, as we noted earlier, this just obfuscates the answer to the question in which we are here interested. We want to know what the connection is between having justiﬁed beliefs and having mostly true beliefs. And the fact is that there is no plausible account of corrigible epistemic justiﬁcation on which there will be a conceptual connection between having justiﬁed beliefs and having mostly true beliefs. There might be a trivial connection between having justiﬁed beliefs and having beliefs that are epistemically likely to be true—but the claim is trivial precisely because talk of justiﬁed beliefs and talk of beliefs likely to be true end up being alternative ways of saying the same thing. So the fear is that neither internalists nor externalists can secure a robust connection between having justiﬁed beliefs and having mostly true beliefs.

11.4 Conclusion If that’s true, let’s return to the question with which I began. Why should the philosopher be interested in the concept of epistemic justiﬁcation? There are at least the following four possible answers. 1. One might argue that possessing epistemic justiﬁcation is simply intrinsically good. When we ask why something is good the context is usually one in which it is presumed that the thing is good, if at all, only for its consequences. But not all things are instrumentally good. There is nothing to stop a philosopher from arguing that a belief is intrinsically valuable just in virtue of its being epistemically justiﬁed (on one’s favored account of epistemic justiﬁcation). Argument tends to break down relatively quickly when it comes to debate over what kinds of things are intrinsically valuable. Moore thought that friendships are intrinsically valuable. Others would argue that

190

CONCERNS ABOUT THE INTERNALIST COMPONENT

friendships are valuable only for the pleasure they give. To settle the controversy we would need an account of both the metaphysics and epistemology of value judgments, a project that takes us well beyond the scope of the present paper. 2. Nothing in what I argued above shows that there isn't a contingent connection between having justiﬁed beliefs and having mostly true beliefs. And that is true on virtually every half-way plausible account of epistemic justiﬁcation. Whether we identify the epistemic status of a belief with internal properties of the believer, coherence of beliefs, epistemic virtues of a believer, proper functioning of beliefproducing mechanisms, or some sort of counterfactual reliability of the process that produced it, the (one and only) actual world might be cooperating so as to ensure a strong statistical correlation between a belief’s being epistemically justiﬁed and that belief’s being true. Indeed, I daresay almost every philosopher proposing an account of epistemic justiﬁcation believes (truly or falsely, justiﬁably or unjustiﬁably) that most justiﬁed beliefs are true. If we deﬁne the instrumental value of something in terms of its actual consequences or its statistically likely consequences, then most philosophers will believe that epistemic justiﬁcation has instrumental value because of its contingent connection with truth. To be sure, some philosophers have questioned the instrumental value of having true beliefs, but while one can certainly point out speciﬁc cases in which it would have been better to have a false belief than a true belief, one can give countless examples of situations in which one is better off with true beliefs. While believing that a justiﬁed belief has instrumental value doesn’t give it instrumental value, we would at least have an explanation of why philosophers care about epistemic justiﬁcation. 3. I have argued elsewhere (1990) that it isn’t plausible to deﬁne what one ought to do in terms of actual or statistically probable consequences. It has always seemed more plausible to me to understand the relevance of consequences in deciding what one ought to do (including what one ought to be interested in) in terms of the epistemic probability of consequences relative to the evidential position of the actor. What one ought to do is a function of the expected utility of an action, where that is a function of the value and disvalue of possible consequences of the action adjusted for the epistemic likelihood of them occurring. Now I argued earlier that there is a perfectly clear sense of epistemic probability that can be deﬁned in terms of epistemic justiﬁcation. I also pointed out that we might well need to distinguish clearly a belief’s being epistemically likely from its having a high statistical probability of being true. Indeed, I argued that virtually any plausible account of epistemic probability will allow for the possibility that most epistemically probable propositions are false. But if what I have said above is plausible, that still allows us to explain why we ought to be interested in epistemic justiﬁcation by appealing to that concept of epistemic probability. Provided that it is epistemically likely that having true beliefs will result from having justiﬁed beliefs, it will also follow on my account of rational action that we ought to be concerned with having epistemically justiﬁed

E V I D E N T I A L I S M A N D T RU T H

191

beliefs. At least it will follow provided that the expected utility of having true beliefs is high. 4. Finally, and one might, of course, come to the perhaps surpising conclusion that our concept of epistemic justiﬁcation is such that the epistemic status of a belief isn’t very important. That wouldn’t make the meta-epistemological task of coming up with the correct analysis of justiﬁcation any less important, for one would only be in a position to reach that surprising conclusion after one successfully completes one’s analysis of our epistemic concepts.

References Armstrong, David. 1973. Belief, Truth and Knowledge. Cambridge: Cambridge University Press. Bergmann, Michael A. 2006. Justiﬁcation without Awareness . Oxford: Clarendon Press. Conee, Earl. 2004. ‘First Things First,’ in Conee and Feldman (2004: 11–36). —— and Richard Feldman. 2001. ‘Internalism Defended’ in Hilary Kornblith (ed.), Epistemology: Internalism and Externalism. Oxford: Blackwell Publishers; reprinted in Conee and Feldman (2004). ——, and ——. 2004. Evidentialism. Oxford: Oxford University Press. Fales, Evan. 1996. A Defense of the Given. Lanham, MD: Rowman and Littleﬁeld. Fumerton, Richard. 1990. Reason and Morality: A Defense of the Egocentric Perspective. Ithaca, NY: Cornell University Press. ——. 1995. Metaepistemology and Skepticism. Lanham, MD:Rowman and Littleﬁeld. Goldman, Alvin. 1979. ‘What is Justiﬁed Belief?’, in George Papps (ed.), Justiﬁcation and Knowledge. Dordrecht: Kluwer, 1–23. ——. 1986. Epistemology and Cognition. Cambridge, MA: Harvard University Press. Henderson, David and Terry Horgan. 2007. ‘The Ins and Outs of Transglobal Reliabilism, in Sanford Goldberg (ed.), Internalism and Externalism in Semantics and Epistemology, Oxford: Oxford University Press. Huemer, Mike. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littleﬁeld. Lewis, C. I. 1946. An Analysis of Knowledge and Valuation, Chicago, IL: Open Court. McGrew, Timothy. 1999. ‘A Defense of Classical Foundationalism,’ in Louis Pojman (ed.), The Theory of Knowledge. Stamford, CT: Wadsworth, 224–35. Russell, Bertrand, 1948. Human Knowledge: Its Scope and Limits. London: Simon and Schuster.

This page intentionally left blank

PART VI

Evidence Just as ‘Evidence’ is a large part of ‘Evidentialism’, so evidence is a big part of evidentialism. Indeed, it is just a schema—though perhaps an important schema— without a substantive theory of evidence. In most of their discussions of justiﬁcation, Conee and Feldman show a clear preference for the theory that evidence consists in feelings or experiences. (See especially ‘First Things First’, ‘Internalism Defended’, and ‘Having Evidence’ in Evidentialism, and more recently ‘Evidence’ in Epistemology: New Essays, ed. Quintin Smith, Oxford University Press, 2010.) Richard Swinburne argues for a doxastic theory of evidence and of having it. That is, evidence consists in beliefs and having p as evidence consists in having the basic belief that p. More generally, someone’s evidence consists of the propositions which they are inclined to believe, each having a subjective probability proportional to their inclination to believe it; those propositions to which a subject ascribes a subjective probability greater than a half are their basic-beliefs. But in Swinburne’s view some basic-beliefs may be overruled (i.e. rendered no longer justiﬁably believable) by propositions rendered probable by other basic beliefs (or more generally by other basic propositions). Patrick Rysiew also proposes a theory of evidence. He draws heavily on Thomas Reid. He emphasizes the functional roles evidence is typically said to play. He lists four roles in which the notion of evidence has been prominently cast. After noting some noted tensions among them he offers an explanation of how adopting a Reidian theory of evidence—which he explicates in some detail—can bring unity to the notion of evidence, relieving the tension felt between its roles. Dougherty takes up Williamson’s argument that evidence must consist in propositions which Conee and Feldman reply to in ‘Evidence’. He argues that experiences matter evidentially because they are where all evidence comes from, but that the evidence itself is propositional. Pritchard’s paper at the beginning of the next part is an honorary member of this part, since it has much to say about the nature of evidence from a disjunctivist perspective.

This page intentionally left blank

12 Evidence Richard Swinburne

I seek in this chapter to investigate how best to understand the notion of ‘evidence’ in an internalist theory of synchronic justiﬁcation, taking as my main example of the latter Conee and Feldman’s theory of ‘evidentialism’, as expounded in their collection Evidentialism.1 (I refer to this work throughout as ‘CF’.) But before concentrating on the notion of evidence itself, I need brieﬂy to analyse the nature of such a theory and the other elements involved therein.

12.1 I distinguish a synchronically justiﬁed belief as one which has good grounds at the time when it is believed, from a diachronically justiﬁed belief as one which is not merely synchronically justiﬁed but the result of proper practices in the past (for example, results from adequate investigation of the issue). Like most philosophical theories of justiﬁcation, Conee and Feldman’s theory2 is a theory of synchronic justiﬁcation. Alston has argued that, because the various philosophical theories of ‘justiﬁcation’ are concerned to analyze different concepts, we should dispense with the very notion of a theory of justiﬁcation, and talk instead of different ‘epistemic desiderata’ which beliefs possess. But he recognizes (Alston 2006: 51) that ‘truth-conducive’ desiderata are ‘clearly the most basic’; and so I shall understand epistemic justiﬁcation as truth-conducive justiﬁcation, and I shall understand by a belief being (synchronically) justiﬁed at a time t as it being at t likely to be true (which I take to be the same as ‘probably true’ or ‘more probable than not’). The common framework for all truth-conducive theories of synchronic justiﬁcation is as follows: a token belief is justiﬁed insofar as it is ‘based on’ or has grounds which make it probably true; and better justiﬁed, the more probable they make it. This framework has been ﬁlled out in many different ways according to whether ‘based on’ 1 2

Much of this chapter is based on my book Epistemic Justiﬁcation (Swinburne, 2001), especially ch 5. ‘Evidentialism is best seen as a theory about synchronic rationality’—Conee and Feldman (2004: 189).

196

EVIDENCE

is required or ‘has’ is enough, and on how ‘grounds’, ‘based on’, and ‘probably’ are understood. Most writers appeal to intuitions about ‘justiﬁcation’ which—they claim—suggest that their theory is the true theory of epistemic justiﬁcation, and rival theories are false. Conee for example claims that certain examples ‘make it reasonable to conclude that a belief is epistemically justiﬁed only when the believer has cognitive access to evidence that supports the truth of the belief’ (CF: 49). I suggest, however, that while many of us use the word ‘justiﬁed’ in a sense which involves this much of the time, many of us do not use it in such a sense at least some of the time. Only by supposing that there are different senses of ‘justiﬁed’, can we account for the longstanding clash of intuitions between externalists and internalists. The many different theories of ‘justiﬁcation’ and in particular of synchronic truth-conducive justiﬁcation, all have some claim to explicate some ordinary-language use of ‘justiﬁed’. There are nevertheless, I suggest, two criteria which must be satisﬁed by any theory which has a good claim to be a theory of justiﬁcation. The ﬁrst criterion is simply that the theory can be spelled out in a coherent way. For example, to spell out reliabilism, the most common form of externalism, in a coherent way requires providing a solution to ‘the generality problem’ (see CF: ch. 6). The second criterion is that the theory must not have the consequence that most of our obviously true beliefs are quite unjustiﬁed. If a theory has the consequence that we are not justiﬁed in believing that the world has existed for more than ﬁve minutes, that the earth is round, or that my head is not made of sawdust, then it must be ruled out. For these beliefs are obviously true—more obviously true than any theory of epistemology could be—and so (in any sense of ‘justiﬁed’ which explicates ordinary use) well-justiﬁed. We shall come in due course to a theory which fails this criterion. The main question to ask about any theory of justiﬁcation which satisﬁes these criteria is whether the concept of justiﬁcation which it provides is one useful in providing philosophical insight (e.g. into the extent of the ability of humans to discover truth) or practical help (e.g. in acquiring more true and less false beliefs). An internalist theory is a ‘mentalist theory, one that holds that a person’s beliefs are justiﬁed only by things that are internal to the person’s mental life’ (CF: 55). A theory is then externalist insofar as it holds that a person’s beliefs are justiﬁed (at least in part) by things external to a person’s mental life. On a natural understanding of the mental, it is that to which the believer has privileged access by introspection. Conee and Feldman distinguish ‘mentalism’ from what they call ‘accessibilism’ which holds that beliefs are justiﬁed ‘by things to which the person has some special sort of access’. They then go on to claim that ‘philosophers have not separated mentalism from accessibilism because they have tacitly assumed that the extensions of the two do not differ in any signiﬁcant way’ (CF: 55). But as the special sort of access is presumably the privileged access which one has to one’s own mental life, the two theories are necessarily logically equivalent. (My access to my own mental life is privileged in that whatever ways others may have of ﬁnding out about my thoughts, etc., I can also use—I can study my behaviour on a

EVIDENCE

197

ﬁlm or my brain states via instruments and mirrors. But I have a way which only I can use—I learn about my mental life by experiencing it.) On this understanding of internalism, a believer’s grounds for some belief, naturally described as his ‘evidence’, may include his or her other beliefs (and what I shall call ‘inclinations to belief’), sensations, occurrent thoughts, desires, and purposes. I will come to the issue of just which of these should be included in due course. An externalist’s grounds for the justiﬁcation of a belief will typically include the token process of the belief’s production—the sense-organ-to-brain-state causal chain, or the more extended nearby-external-state-of-affairs-to-brain-state causal chain; and facts about the proportion of true beliefs produced by a certain type process to which the token process belongs. It may also include other things such as the token process being a type ‘designed’ (in some sense) to produce true beliefs.3 A belief being ‘based’ on its grounds is most naturally understood as the belief being ‘caused’ (or, more precisely ‘causally sustained’) by (some of) those grounds; and if a theory involves ‘being based on’ in this sense, it would seem to be in this respect an externalist theory. I distinguish three basic kinds of probability: physical, statistical, and inductive. Physical probability is a measure of the extent to which nature has a deterministic propensity towards bringing forth events. Statistical probability is a measure of the proportion of events of one type in some class of events of another type. The class may be a class in the actual world, or a class in a possible world—for example, the proportion of heads in a series of tosses of this coin if we were to toss it indeﬁnitely often; or a class in a world in which the laws of nature are very different from our laws. And ﬁnally there is inductive probability, which is a measure of the extent to which one proposition makes another one ‘probably’ (or ‘likely to be’) true. There is more than one kind of inductive probability. There is subjective probability which is the probability of one proposition on another by the criteria of inductive probability used by a certain person or group. This measure is person-relative, and has as many variants as there are people or groups. But certainly in some cases there are (at least within rough limits) true person-independent criteria for determining the value of the inductive probability of one proposition on another. I call this latter objective kind of inductive probability, ‘logical probability’; it has the value which would be ascribed to it on the basis of correct criteria by a logically omniscient being who could see all the consequences of these criteria. Any objectively true person-independent relation of inductive probability (e.g. that q makes p probable) which depends on contingent circumstances (e.g. r) can be represented as a necessary relation of logical probability (e.g. that (q and r) make p probable). Almost everyone is prepared to allow that there is an objectively true value of logical probability of p on q (often symbolized ‘P(p|q)’) where q entails p (that is, 1) or entails 3 One of Plantinga’s (1993) criteria for a belief being ‘warranted’ (that is, having a particular kind of justiﬁcation) is that the belief is produced by a ‘creator’, ‘according to a design plan aimed at the production of true beliefs’ (Plantinga 1993: 194).

198

EVIDENCE

not-p (that is, 0). And most writers are happy with the idea that the probability of one proposition on another proposition sometimes entails (or puts logical limits, or limits of probabilistic coherence,4 on) the probability which can be possessed by some third proposition on some fourth proposition. And they are also happy with there being logical relations between assertions of statistical and logical probability—for example that the logical probability of a token belief being true on the evidence that it is produced by a process which produces a proportion ç of true beliefs (that is, the statistical probability of a true belief being produced by that process is ç) is ç—which is roughly the principle which David Lewis (1986: 20) called ‘the Principal Principle’. But what many writers deny is that there are what I shall call ‘wide criteria’ of logical probability, knowable a priori, which allow us to ascribe an objectively true value to any non-probabilistic proposition p (e.g. an explanatory hypothesis) on any other nonprobabilistic proposition q (e.g. a proposition reporting observations). That probability may be either an absolute value, for example 2/3, or a relative value, for example greater than the probability of some particular proposition, or a very vague value, for example ‘very improbable’. The consequences of denying that there are such criteria is, however, that any scientiﬁc hypothesis logically compatible with the data is—objectively—just as likely to be true as any other one. Our preference on data available up to the present for General Relativity over a theory which claims that General Relativity holds until the galaxies reach a certain average distance apart (which distance they will reach tomorrow) after which it will be replaced by a universal law of gravitational repulsion, would be merely a subjective preference. The only way to save the view that science reaches true results about which theories are probably true and which are probably false is to acknowledge wide criteria of logical probability.5 Since humans are unable (at least at present) to work out all the consequences of their criteria of logical probability (e.g. for whether the axioms of arithmetic entail Goldbach’s conjecture—that every even number is the sum of two prime numbers— or whether they entail its negation, and so whether the probability of Goldbach’s conjecture on the axioms is 1 or 0) there seems room also for a third kind of inductive probability. It does look as if Goldbach’s conjecture would be probable in some objective sense on the axioms, together with evidence that it had been found to be true of the ﬁrst trillion numbers and that an otherwise totally truthful and brilliant mathematician claimed to have proved it. Yet the kind of probability designated by that sense cannot be the sense of logical probability, since the logical probability of the conjecture on the axioms (and so on the axioms plus any other proposition) is either 1 or 0. Loosely, this third kind of probability is probability relative to the inferential abilities of a certain person or group; in my example the group is the human race of the

4 A set of probability judgments is coherent if it is not inconsistent with the normal axioms of the probability calculus. For different accounts of why probability judgments should so conform (including the standard ‘Dutch book’ argument), see Howson and Urbach (1993: ch. 5). 5 For my own account of what these criteria are see my Epistemic Justiﬁcation (2001: ch. 4).

EVIDENCE

199

second decade of the twenty-ﬁrst century. To make the notion more precise, we must deﬁne it as the logical probability of one proposition on another insofar as this can be determined by a certain limited number of inferential steps following certain rules of inference. I call this kind of inductive probability ‘epistemic probability’. (Alas, there is a lack of standard terminology in this area of philosophy, and so different writers use ‘epistemic probability’ in different senses.) Epistemic probability is rule-and-steprelative, and so will have many variants.6 The normal form of externalism is reliabilism. A reliabilist’s grounds of ‘the proportion of true beliefs produced by a certain type process’ are grounds of statistical probability. In virtue of the Principal Principle these grounds give a corresponding logical probability to a token of the type being true. An internalist’s probability on evidence must be a species of inductive probability. A believer’s criteria of subjective probability are clearly internally accessible. In so far as there are true criteria of logical or epistemic probability, they are a priori criteria; and derivable by reﬂection on thought experiments, and so also internally accessible. Some writers give an entirely subjective account of internalist justiﬁcation, that a belief is justiﬁed if it is rendered probable by the subject’s own criteria. For example, Richard Foley (1993: 79) thinks that we cannot have more than ‘egocentric rationality’ which ‘requires that we have beliefs that are to our own deep intellectual satisfaction’, that is, in my terminology ‘are subjectively probable’. While an internalist theory in which the ‘probability’ is subjective yields one kind of justiﬁcation, and plausibly one worth having, we also need an internalist theory in which the ‘probability’ is of an objective kind, and so either logical or epistemic probability, which will allow us to distinguish beliefs which are really supported by their evidence from those that are not. For almost all of us believe that science achieves (objectively) probably true results, and it couldn’t do that unless there are such objective criteria of probability. Conee and Feldman, state their theory of ‘evidentialism’ in one place as the theory that: Doxastic attitude D toward proposition p is epistemically justiﬁed for S at t if and only if having D towards p ﬁts the evidence S has at t. (CF: 83). So someone’s belief is justiﬁed at a time iff it ﬁts the evidence they have at that time. Elsewhere they construe a belief ‘ﬁtting’ the evidence as the evidence making it ‘epistemically probable’ that the belief is true (CF: 100); but they acknowledge a difﬁculty in clarifying the notion of ‘epistemic probability’ (CF: 100 and 305). ‘Fitting the evidence’ seems however to be regarded by Conee and Feldman as an objective notion, and so naturally to be understood either in terms of ‘logical probability’ or in terms of a variant of ‘epistemic probability’ (in my sense). A belief may then be justiﬁed by being rendered probable by correct criteria, or by correct criteria applied with a 6 For ﬁrst work on systematizing ‘epistemic probability’ (in my sense) see the description of Daniel Garber’s ‘local Bayesianism’ and Richard Jeffrey’s pioneer suggestions in my Epistemic Justiﬁcation (2001: Additional Note E).

200

EVIDENCE

limitation on the rules and steps allowed (and perhaps one obvious limited way in which the probability should be measured is the least limited one the believer can utilize). Ordinary usage could, I suggest, provide plenty of examples of beliefs being said to be ‘justiﬁed’ in virtue of being made probable by evidence in either of these ways.

12.2 Having analyzed the nature of an internalist theory and the other elements involved in it, I come at last to the issue of what the internalist should regard as evidence. In nonphilosophical discussion ‘evidence’ consists of publicly accessible states of affairs, either states actually observed and studied by investigators or ones readily available for such study. But in the context of an internalist theory, it consists only of states accessible to introspection. But which states? Bonjour has recently revived the classical empiricist project7 of arguing that our internal sensory states form the evidence which renders probable our beliefs about the physical world. A basic belief is, for Bonjour, one for which there is ‘an internally available reason why it is likely to be true that does not depend on any further belief or other cognitive state’ (Bonjour 2003: 69). This reason is the content of our sensory ‘experiences’, which he regards as sense-data; these include patterns of colours in the visual ﬁeld, and tactual feels, noises, smells, and tastes. The fact that a belief seems to characterize accurately the content of that experience (e.g. as a red triangular shape) in my visual ﬁeld ‘provides an entirely adequate basis for thinking that the description is correct’ (Bonjour 2003: 74). Bonjour goes on to argue that the ways in which sense-data change in line with our kinaesthetic sensations (of apparently changing position) is then best explained by supposing that the sense-data are caused by the physical objects which we ordinarily believe that we perceive. That is, he claims in effect, the (logically) most probable explanation of the sense-data each of us is currently experiencing is that they are caused by physical objects. All of that might be correct,8 but the only beliefs which would be rendered probable in this way would be each believer’s beliefs about the physical objects which he or she is currently perceiving. It would not justify memory beliefs or beliefs derived from testimony (what other people tell us). For our apparent memories are not in general memories of the sense-data we had yesterday, but of the objects we perceived yesterday (when we cannot recall the sense data which accompanied those perceptions). And much of what we believe about the world is derived from our apparent memories of the testimony of others; and these apparent memories of what other people told us are not normally apparent memories of the sounds 7 Bonjour himself refers to Locke, Berkeley, Broad, and Price as contributors to this project. See Bonjour 2003: 86. 8 It is of course open to the objection that it assumes ‘the myth of the given’, that is that any description of the contents of our mental life will be in terms of its normal cause in or some other relation to the public world, and so already presupposes various features of the public world and thus cannot be used without circularity as evidence for these features. See Sellars (1963). Bonjour (2003) responds to this objection in various places.

EVIDENCE

201

they uttered but of the propositions they expressed (when we cannot even recall the words, let alone the sounds by which they expressed them). Our current sense-data taken on their own are simply too narrow a base to render inductively probable (on virtually anyone’s criteria of subjective probability, let alone—I suggest—on the criteria of logical or epistemic probability) beliefs of memory and testimony. Also most of our very general common-sense beliefs which seem to us obviously true—that the world has existed for more than ﬁve minutes, that the earth is round, that my head is not made of sawdust, etc.—as well as our ordinary beliefs about geography and history, are basic-beliefs in the sense that we do not now hold them because they are rendered probable by any other beliefs (for we cannot now recall how we acquired them), let alone by any sense-data. For these reasons a theory which claims (as Bonjour’s theory seems to claim) that our sole foundational evidence which renders probable our beliefs is our sense-data fails to satisfy my second criterion above for a possible theory of justiﬁcation. So we need a wider base than our present sensory experiences if our normal beliefs about the world are to be justiﬁed. Our other mental states include our purposes (what we are trying to achieve), our desires, and our occurrent thoughts; but there is no obvious way of extrapolating from these to the probable state of the physical world. The only way to get from our mental life to the world is to start from our beliefs and inclinations to belief; or rather from those ones which we hold not because they are made probable by other beliefs but because—it seems to us—they are forced upon us by how things are in the world (whether or not we have a sensory awareness of what it is about the world which forces the beliefs upon us). A person starts with inclinations of different strengths to believe various propositions, insofar as he ﬁnds himself with them (because they seem to him forced on him by the world), and not in so far as they are rendered (subjectively on his criteria) probable by other of his beliefs. These are what I shall call the subject’s basic propositions. The strength of an inclination to believe a proposition may be measured by the (subjective) probability which the subject would ascribe to that proposition (if he had the concept of probability) on the evidence merely of his having that inclination; those basic propositions to which he ascribes a probability on this evidence greater than half are his basic-beliefs. Most philosophers seem to ignore the evidential role of inclinations which are not of sufﬁcient strength to constitute beliefs. But such inclinations to believe, when combined with other inclinations or beliefs, can render (on most people’s criteria of subjective criteria, and—in my view—on the criteria of logical probability) some belief probable which would not otherwise be probable. If, as I watch the cars passing my window, I catch glimpses of several cars each of which it looks as if it might be red, although more probably it is black, and on the basis of each glimpse I ascribe to the proposition that the car in question was red the probability of ⅓, these inclinations together plausibly make it probable overall (more probable than not), in the absence of other evidence, that on at least one occasion a red car passed my window. (In writing in future of our ‘basic-beliefs’ I should be taken—where it is appropriate—to include our basic ‘inclinations to believe’, but I shall not always repeat this phrase each time I mention ‘basic-beliefs’.) Philosophers also sometimes ignore the

202

EVIDENCE

point that very strong basic-beliefs give much greater probability to other beliefs than do weaker basic-beliefs. Our ordinary basic-beliefs include beliefs about our present mental states, beliefs about what we are currently perceiving, beliefs about necessary truths (such as that ‘2+2=4’), and memory beliefs, both particular beliefs (e.g. that I saw a hedgehog yesterday), and general beliefs (e.g. that Marks and Spencer sell good quality food), and beliefs about what others have told us. They also include beliefs which we believe that‘everyone knows’, for example that the earth is spherical and billons of years old (beliefs which we have clearly acquired as a result of teaching or reading or listening to conversations, but the source of which we may not now be able to recall). In starting from his or her actual basic-beliefs a subject initially holds that they are (subjectively) probably true to the degree to which they seem to him or her to be probably true in the absence of other relevant evidence in the form of other beliefs. But which basic propositions are we right to take as evidence? I shall understand by a rightly basic proposition a basic proposition to which the believer ascribes that degree of logical probability which is the same as its subjective probability. Thus if he ascribes to some basic proposition a subjective probability of ‘as probable as not’, it is rightly basic if its logical probability on the evidence that he ascribes to it that degree of probability, is ½. My answer to the question is then that all basic propositions about any subject matter are rightly basic;9 and so all basic-beliefs are rightly basic. Our evidence is the way things seem to us; it is from that that we must start to build our worldview. This is the Principle of Credulity; the rational person is the credulous person; he is right to believe everything he believes as strongly as he believes it until it is rendered improbable by something else he believes. The Principle of Credulity says that the subject should start from where he or she is—doxastically. The empiricist tradition has sought to claim that only certain kinds of basic-belief are rightly basic. Plantinga (1983: 55–9) interpreted the medievals as claiming that only propositions which are self-evident or evident to the senses are ‘properly basic’; and he interpreted the early modern empiricist tradition as claiming that only incorrigible propositions (the extension of which were propositions about sense-data) are ‘properly basic’. He combined these two views into a view which he called ‘Classical Foundationalism’ such that ‘A proposition p is properly basic for a person S if p is either selfevident to S or incorrigible for S or evident to the senses of S’. (‘The senses’ means the ﬁve senses.) But, as Plantinga pointed out, this is still too narrow a base to give us our ordinary beliefs. It will only justify our beliefs about presently perceived physical objects, not our beliefs about the past, nor much else including the very general beliefs whose source we cannot now recall. And once all that is included I cannot see any a

9 My use of ‘rightly basic’ is similar to Plantinga’s use of ‘properly basic’ (see his 1983: 55–9). But I use a different phrase in order not to be thought to endorse Plantinga’s view that a belief being properly basic may arise from it having non-doxastic grounds (mental grounds other than beliefs) of a right kind. (See Plantinga 1983: 78–82.)

EVIDENCE

203

priori reason for denying evidential status to any logically contingent basic proposition. We do not normally need a priori principles to rule out ‘wild’ basic-beliefs (beliefs that one has seen fairies, or knows the future from an astrological chart, etc.), since a posteriori criteria will normally rule these out. We almost all have theories rendered (subjectively and—I suggest—logically) probable by other basic propositions (including what ‘everybody knows’ about how the world works) which rule out ‘wild’ basic propositions from being overall probable. If someone believes that he has seen the Great Pumpkin return at Halloween, then this belief is normally rendered improbable by a lot of other evidence (in the form of his or her basic-beliefs or what others have told him or her about their observations). People are obviously right to take into account background evidence where it differs quite a bit from the background evidence we currently have—for example, when Aristotelian physics was well established, people were right to regard apparent change in the region of the heavens beyond the moon as very probably an illusion (which is why no supernovas were reported in the West before 1500). It is the simplest extrapolation from this to allow that people are always right to amend their basic-beliefs to ﬁt background evidence10—unless there are too many strong basic-beliefs inconsistent with that evidence. In that case the theory has to change. It is also the simplest extrapolation from all this, that if there is no background evidence which rules a basic-belief in or out, then it is probably true—but not nearly as probable as it would be if it ﬁtted our background evidence. It follows that if someone has what we would consider a ‘wild’ basic belief (e.g. a hunch that he is being abducted by alien space travelers), and has no background evidence which renders it improbable, then it is still on balance probable, and so on an internalist theory he is justiﬁed in believing it. And this consequence is surely correct, since it can hardly be an a priori truth that there are no fairies, or that astrology is false. So we should include our basic-beliefs as part of our evidence. Should we include the other elements of the mental life as well—our purposes, desires, occurrent thoughts, and sensations? No, because believers cannot use these things as evidence for anything else, except in virtue of what they believe (true or false) about them. The internalist should construe a subject’s evidence as his basic propositions, that is mainly his basic-beliefs; and only his basic propositions.11 I am writing ‘basic-belief’ rather than ‘basic belief’, as is normal, in order to draw attention to the point that a basic-belief may ultimately be regarded by a subject as improbable in the light of further evidence, and then it will not be a belief of the subject at all. An extreme case of this is amending a basic-belief in the light of itself. A basic-belief which entails that it is true in any logically possible circumstances will have a logical probability of 1 on any evidence or none; and one which entails that it is true in no possible circumstances will have a probability of 0. A subject who amends in accord with correct criteria will therefore regard a belief of the former kind as certain, and no longer believe the latter. 11 Conee and Feldman typically construe one’s evidence as consisting of ‘experiences’. Conee holds that ‘experiences need not either consist of beliefs or be a topic of beliefs’ (CF: 43), which entails that things other than beliefs (presumably sense-data) can function as evidence. He also holds that ‘it would not be most reasonable to assume that only sensory events are foundational’ (CF: 43, n16). He allows that ‘sensory awareness’ might consist of ‘apprehending certain propositions’ (CF: 40, n6), albeit ones of a very narrow kind, and such apprehension would presumbably involve belief. 10

204

EVIDENCE

Beliefs, like desires and unlike sensations or occurrent thoughts, are continuing mental states. We have them while we are not thinking about them. Feldman however seems to favour the view (CF: 232) that ‘S has p available as evidence at t iff S is currently thinking of p’. But—contrary to Feldman—we should regard all our beliefs, and for a similar reason all our basic propositions and hence our basic-beliefs, as available evidence, since we are (subconsciously) ‘aware of’ them (they are not merely ‘accessible’) even when we are not currently thinking about them. This is shown by the fact that, when we are not thinking about them, they still inﬂuence which other beliefs we form and which actions we do; and if asked why we formed a certain belief or did a certain action, we often cite beliefs which we did not consciously think about while forming the former belief or initiating the action. That I have the belief that there is food in the larder is shown by my going to the larder when I want some food, even if it never ‘crosses my mind’ that there is food in the larder. And so more generally. Any ‘belief’ which had no affect on a person’s actions or other beliefs would not be a belief of theirs at all; and if it does affect their actions and so (at least subconsciously) they are aware of it, it is available to them as evidence and so we can assess the propriety of their response to it. One way in which a belief guides our actions is that it determines what we say spontaneously when we seek to answer questions honestly. (The spontaneity of an answer is very strong evidence—but no more than that—that the belief is already there, before the question is asked. Feldman is, however, right to reject the view that anything would be evidence now just because we could recall it easily, since we can sometimes easily recall what we have currently forgotten.) Feldman ends his discussion of this issue (CF: 239) by showing some sympathy for a wider understanding of which beliefs can constitute evidence. He allows that we have ‘operative’ background beliefs,‘beliefs that are playing an active role in sustaining one’s current state’ of the beliefs currently being thought about. But, I am suggesting, the grounds for attributing available ‘operative’ beliefs to people are much wider than that. 12 Everything which I have just written about beliefs applies for similar reasons to all basic propositions and so to basic-beliefs (even when they don’t become beliefs.)

12 Another restriction on the beliefs which can count on evidence has been proposed by Williamson: ‘knowledge, and only knowledge, constitutes evidence’ (2000: 185). Williamson acknowledges (with qualiﬁcations arising from what he regards as the unclarity of the notion of the belief) that knowledge entails belief; and since clearly knowing p entails that p is true, this has the consequence that only true beliefs can constitute evidence. And of course something else—which Plantinga calls ‘warrant’ is required for a true belief to constitute knowledge, though Williamson does not think that this can be analysed. But our ‘evidence’ is for Williamson a public matter; and since the truth of a proposition about the physical world depends on how things are in that world and that is something about which we can always be mistaken, it follows that ‘we are not always in a position to know what our evidence is’ (2000: 190). That makes Williamson’s theory (in this respect) externalist, and so beyond the scope of this paper. Nevertheless Williamson’s theory still has the problem of accounting for beliefs which are made (in some inductive sense) very probable by conjunctions of pieces of uncertain evidence—as in my example of the cars. In such cases Williamson will need to say that it is an item of knowledge that certain propositions have certain probabilities, e.g. that there was a probability of ⅓ for each of the cars that it was red. But to allow that sort of knowledge to function as evidence seems contrary to the spirit of Williamson’s project—see his (2000: 193).

EVIDENCE

205

So I suggest that Conee and Feldman’s ‘evidentialism’ should be spelled out by understanding a subject’s ‘evidence’ as the collection of the subject’s basic-beliefs (and other basic propositions), while their ‘ﬁtting the evidence’ should be understood either as ‘made logically probable by the evidence’ or as ‘made epistemically probable [on a certain variant of “epistemic probability” in my sense] by the evidence’. Clearly, on the assumption that there are wide criteria of logical probability and that one of these criteria is that basic propositions are probable insofar as they are believed, it is an instrumentally good thing for a person to have a belief B rendered logically probable by their evidence. For this very fact means that it is probable to a signiﬁcant degree (varying with the probability of the evidence and the probability that it confers on B) that B is true—and true beliefs are a good thing to have. And it would be the next best thing to have a belief rendered epistemically probable (on any variant of this, and in particular on the least limited variant of this concept the subject can utilize) by the evidence. But it would be better if the believer’s belief is ‘based on’ the evidence, and Conee and Feldman deﬁne a second concept (additional to the concept of being ‘justiﬁed’) of a doxastic attitude being ‘well-founded’ (CF: 93) which involves both the belief being justiﬁed and also being based on the justifying evidence. They see the ‘epistemic merit’ of a belief being based on the evidence as the believer ‘appreciating’ the evidence (CF: 93, n22). This presumably amounts to the believer recognizing the probabilistic force of the evidence. That is an internalist notion, and ‘well-foundedness’ so understood remains an internalist concept, and surely has an intrinsic value greater than that of mere justiﬁcation. For it involves the believer being aware of the application of a very important a priori concept of objective inductive probability. But it seems clear (despite CF: 93, n22) that one can in this sense ‘appreciate’ the evidence for a belief even if that belief is not caused by that appreciation; it might be caused by totally irrational factors. Yet it would be intrinsically more valuable if the belief is caused by the subject recognizing its probabilistic force. For then the believer does not merely have a true belief about inductive probability, but is guided by it in his formation of other beliefs, and that makes for signiﬁcantly greater sensitivity to the a priori. Thereby the believer manifests that scientiﬁc rationality which contributes to the intrinsic worth of humans (and to a lesser extent, animals). On this understanding ‘well-foundedness’ involves an externalist element. In these various senses of being ‘epistemically justiﬁed’ and being ‘well-founded’ which can be developed from Conee and Feldman’s account, it is good to have beliefs which have these features.

References Alston, William P. 1989. Epistemic Justiﬁcation. Ithaca, NY: Cornell University Press. ——. 2006. Beyond ‘Justiﬁcation’: Dimensions of Epistemic Evaluation. Ithaca: Cornell University Press.

206

EVIDENCE

Bonjour, L. 2003. ‘A Version of Internalist Foundationalism’ in L. Bonjour and E. Sosa, Epistemic Justiﬁcation. Oxford: Blackwell. Conee, Earl and Richard Feldman, 2004. Evidentialism. Oxford: Clarendon Press. (I refer to this work throughout the present paper as ‘CF’.) Foley, Richard, 1993. Working Without a Net. Oxford: Oxford University Press. Howson, C. and P. Urbach. 1993. Scientiﬁc Reasoning, 2nd edn. Chicago: Open Court. Lewis, David. 1986. ‘A Subjectivist’s Guide to Objective Chance’ in D. Lewis, Philosophical Papers Vol. II. Oxford: Oxford University Press. Plantinga, Alvin I. 1983. ‘Reason and Belief in God’ in A. Plantinga and N. Wolterstorff (eds), Faith and Rationality. Notre Dame, IN: Notre Dame University Press. ——. 1993. Warrant and Proper Function. Oxford: Oxford University Press. Sellars, Wilfrid. 1963. ‘Empiricism and the Philosophy of Mind’, ch 5 of W. Sellars, Science, Perception, and Reality. London: Routledge and Kegan Paul. Swinburne, Richard. 2001. Epistemic Justiﬁcation. Oxford: Clarendon Press. Williamson, Timothy. 2000. Knowledge and its Limits. Oxford: Oxford University Press.

13 Making it Evident: Evidence and Evidentness, Justiﬁcation, and Belief Patrick Rysiew

‘We give the name of evidence to whatever is the ground of belief.’ (Reid IP II 20, W 328a1) ‘A proposition is epistemically justiﬁed to someone when it is evident to that person that the proposition is true.’ (Conee 1992: 252) ‘ . . . a central function of evidence is to make evident that which would not be so in its absence.’ (Kelly 2006: }3)

13.1 Introduction As usually stated, evidentialism constitutes the outline of a theory: supposing that justiﬁcation of the sort required for knowing is solely a function of one’s evidence,2 a well-developed evidentialist theory will tell us what sorts of things constitute one’s evidence, what it is to have evidence, what it is for evidence to support a given belief or proposition, and so on. But whether or not evidentialism is true such questions merit attention. For whether or not evidentialism is true evidence remains a central epistemological concept. Answering questions concerning evidence, however, is made more difﬁcult by the fact that it is often asked to play roles in addition to that (/those) featured by evidentialism. And while one would like (other things being equal) to provide a uniﬁed account of a given concept or phenomenon, as Thomas Kelly (2006) has

1 References to Reid will be given by the Chapter/Essay and Section numbers in Reid’s Inquiry into the Human Mind (IHM), his Essays on the Intellectual Powers (IP). These will be followed by ‘W’ and the corresponding page numbers in the Hamilton edition of Reid’s Works (8th edn, 1895), with ‘a’ referring to the right-hand side of the page and ‘b’ to the left-hand side. 2 This statement of the evidentialist thesis is of course an approximation: one’s belief must be properly based on one’s evidence. Such over-simpliﬁcation is, however, harmless enough for present purposes.

208

EVIDENCE

documented, some of the roles that evidence has been thought to play stand in apparent tension to each other; the result is that it’s not clear whether a single theory can ﬁt everything we want to say about evidence. The present chapter offers partial and tentative answers to questions like those posed above, while attempting too to move towards a resolution of the apparent tensions Kelly describes. The discussion is structured as follows. First, I rehearse some of the main lines of Kelly’s discussion, outlining the prominent theoretical roles evidence has been asked to play and the apparent tensions in which they stand. I then present a conception of evidence—one found in Thomas Reid—which might enable some progress: Reid’s views on evidence represent an alternative to usual ways of thinking about it. It has, on its own, certain attractive features (inter alia, it brings together certain evaluative-normative and psychological-descriptive notions, and represents an attractive combination of ‘internalist’ and ‘externalist’ ideas). And further, Reid’s conception of evidence gives us a way of preserving a unity among the various roles evidence has been asked to play. Elements of the Reidian view can be detected in some of the writings, both separate and together, of Feldman and Conee; and it is certainly compatible with the basic evidentialist claim. By the same token, however, the view developed here is neutral as to the truth of evidentialism. Apart from its having the virtues just mentioned, its more immediate interest lies in its offering at least the beginnings of a fuller account of evidence—something, again, that should be of interest whatever one thinks of evidentialism proper.

13.2 Evidence—some roles and tensions As various people have noted, philosophers’ use of the term ‘evidence’ doesn’t neatly match how it’s used in non-philosophical domains.3 But that’s not surprising: theorists often use terms in somewhat technical, non-ordinary ways. More troubling is the fact that there’s not much agreement within philosophy as to what evidence is: it has been variously said to consist in one’s ‘sense-data’ (certain empiricists), ‘observation statements’ (positivists), what one knows (Williamson 2000b), the ‘information a person has to go on’ (Feldman 2003: 45) or whatever indicates to us that the proposition is true (Conee 2011), one’s non-factive, phenomenal states (certain epistemic internalists), whatever states or processes, etc., lead in some suitaby reliable way, to a belief (e.g., Greco 1999, 2002). And so on (Kelly 2006). But while there is much philosophical controversy over just what evidence is, there is at least greater agreement about what it does—that is, there are certain key roles that evidence has generally been thought to play. In particular, we can—again, following Kelly—see at least four roles that evidence has typically been asked to play by those interested in the notion.

3 What follows in this section is a highly compressed version of Kelly’s very nice discussion, to which readers are directed for greater detail.

MAKING IT EVIDENT

209

1. Evidence as that which justiﬁes belief Again, ‘evidentialists’ regard the justiﬁedness of a belief as entirely a function of how well it ﬁts one’s evidence. But even those who don’t explicitly advertise themselves as evidentialists can plausibly be seen to regard evidence as central to justiﬁcation. Thus, for example, even if coherence is supposed to be the central epistemic notion, that with which a given belief coheres is plausibly understood to be one’s evidence in favor of it. Indeed, even those whom it is natural to see as enemies of evidentialism seem at times to be admitting evidentialist considerations in through the back door. Thus, for example, even reliabilists—who try to understand notions like justiﬁcation without resorting to notions like evidence—seem at times to be forced into allowing its centrality.4,5 2. Rational thinkers respect their evidence This idea is closely related to the previous one—and not surprisingly. ‘What else,’ one might ask, ‘would rational belief be?’ And: ‘Isn’t that just what evidence is—the sort of thing which rational believers respect?’ Here too, it is testament to the intuitive strength—or, perhaps, the truistic nature—of the relevant idea that even those whose epistemological views are rather heterodox acknowledge it. Thus, historically, we ﬁnd not only the likes of Locke and Clifford, but Hume as well, endorsing it.6 But this same idea is just as current in contemporary theorizing too—where, particularly among ‘internalists’, ‘rational’ and ‘justiﬁed’ are often used interchangeably to describe epistemically good beliefs (/believings), and ‘justiﬁcation’, ‘reasons’, and ‘evidence’ are very often used interchangeably in specifying what makes a given belief epistemically good (cf. Rysiew 2008: }2). But here too, even those squarely within the externalist camp can regard (2) as past controversy. Thus, Williamson: ‘Rational thinkers respect their evidence. Properly understood, that is a platitude’ (2000a: 613). 3. Evidence as a guide to truth: evidence as sign, symptom, or mark It is natural to think that evidence ‘gets its point’, so to speak, from the fact that we very rarely, if ever, grasp truths in a completely direct, unmediated way (see Kelly 2006: }3;

4 Note that I don’t mean to be endorsing this line of criticism against reliabilism, or necessarily to endorse any of the various things that have been said in support of thinking of evidence in this or that way. In the present Section, the goal is just to articulate and brieﬂy illustrate the relevant claims about evidence and its alleged roles. 5 Thus, consider the analogue of the ‘no defeaters’ requirement in Goldman’s (1979) version of reliabilism: a reliably produced belief may fail to be justiﬁed if the subject fails to avail themselves of some alternative available reliable processes. Which such processes should one make use of? ‘What I think we should have in mind here are such additional processes as calling previously acquired evidence to mind, assessing the implications of that evidence, etc.’ 6 ‘[T]he mind, if it will proceed rationally, ought to examine all the grounds of probability, and see how they make more or less for or against any proposition, before it assents to or dissents from it.’ (Locke 1690/1959, Book IV, Chapter XV, p. 366 of Fraser); ‘[I]t is wrong always, everywhere, and for anyone, to believe anything upon insufﬁcient evidence’ (Clifford 1877/1901, Section I); ‘A wise man . . . proportions his belief to the evidence’ (Hume 1748/1987, ‘Of Miracles’, Part I).

210

EVIDENCE

cf. Kelly 2008: 942). If we could always just see, so to speak, the truth—if things were just obvious or apparent—we’d have no need of evidence. (Presumably, God doesn’t need evidence.) But as we’re at best only very rarely in that position, we need some guide to the truth; and evidence is thought to play that kind of mediating role: ‘a central function of evidence is to make evident that which would not be so in its absence’ (Kelly 2008: 942). ‘Perhaps the root notion of evidence,’ then, ‘is that of something which serves as a reliable sign, symptom, or mark of that which it is evidence of’ (Kelly 2006: }3). On this ‘reliable indicator’ conception of evidence, all sorts of things may fall under its extension—beliefs, experiences, states of affairs, and so on. So long as there is some objective fact-indicating relation between X (/its occurrence) and some proposition or state of affairs, the former may be said to be evidence of the latter. 4. Objectivity, publicity, and intersubjectivity: evidence as neutral arbiter Lastly, the concept of evidence seems naturally connected to that of objectivity, and related concepts. For instance, it’s natural to think of an inquirer as objective just in case their beliefs are based on evidence, as opposed to being the product of prejudices, preferences, interests, and so on. And two such objective inquirers who share the same evidence can, other things equal, be expected to converge in their opinions. By the same token, among inquirers who disagree, evidence can play the role of neutral arbiter, a consideration of which might settle the question. (For it to play this role, of course, the evidence must be available to all relevant parties.) It’s not hard to see how the various demands placed upon the concept of evidence give rise to some apparent tensions. For instance, (3) suggests that something can be evidence, even evidence which one possesses, even though one is not aware of its evidentiary status or signiﬁcance. (To take an example of Kelly’s, one might see that a patient has Koplik spots without knowing that they are a sign of measles.) But, the thinking goes, it’s not clear how such evidence could be relevant to the justiﬁcation of one’s belief (1), or something which one, qua rational thinker, must always be expected to respect (2). If evidence is to be something that guides rational agents and justiﬁes their beliefs, it’s natural to suppose that it must be within such agents’ ken. Thus: ‘the evidence a person has at a given time consists of all the information that the person has to go on at that time’ (Feldman 2003: 45); one’s reasons (evidence) are ‘things that the person can, at least in the typical case, describe to someone else and cite in support of the belief’ (2005: 273; cf. Conee 1988: 48–9). So, for example, while we might sometimes talk as though the bare reliable indicator were itself the evidence, that is misleading: While we might ordinarily say that your reason for thinking that the tree is a maple is that its leaves are a particular shape, the fact that the leaves are that shape is not part of your evidence. What you are going on in judging the tree to be a maple is your belief that it has leaves of a particular shape, and perhaps ultimately you are going on how the tree looks to you (your perceptual experience). These are internal, mental states you are in. (Feldman 2005: 273)

MAKING IT EVIDENT

211

So, according to Feldman (/and Conee), (1) and (2) favor mentalism about justiﬁers (Conee and Feldman 2001)—the idea that justiﬁers consist solely of one’s beliefs, experiences, apparent memories, and other mental states—whereas there is no such restriction upon the reliable indicators of (3). In fact, thinking along these same lines, a number of philosophers7 have found it plausible to suppose that the role of evidence as justiﬁer, and as that which rational agents respect, means that it can’t be an inherently truth-linked notion. Here, the device of a demon world is often invoked: whether or not one is demonically deceived, one can do a better/worse job of respecting one’s evidence, being epistemically responsible, etc., and so of arriving at justiﬁed beliefs; so de facto reliability can’t enter into determining what evidence one has or whether one’s beliefs are justiﬁed (rational, etc.). According to some, this same line of thinking leads inexorably to the view that evidence is best understood in terms of one’s speciﬁcally phenomenal states (see Williamson 2000a, 2000b; Kelly 2006, 2008). But even short of that result, as we’ve seen, such thinking leads away from the conception of evidence required by (3); further, in so doing, it moves towards thinking of evidence in terms blatantly ill-suited to its playing the role of neutral arbiter (4)—that is, the sort of stable, public items that multiple persons might grasp, observe, consider, and so on. In light of such results, it’s not clear how much theoretical unity we can reasonably expect. Kelly writes: Both in and outside of philosophy, the concept of evidence has often been called upon to ﬁll a number of distinct roles. Although some of these roles are complementary, others stand in at least some measure of tension with one another. . . . [I]t is far from obvious that any one thing could play all of the diverse roles that evidence has at various times been expected to play. Different theories about the nature of evidence might thus naturally emerge from different emphases on the competing demands that have been placed on the concept. (Kelly 2006: Intro.)

There are in fact two issues here: whether a single kind of thing is suited to play the various roles evidence has been thought to play; and whether we’re likely to arrive at a uniﬁed theory (a single concept) of evidence. I am doubtful about the ﬁrst but hopeful about the second: I doubt that there’s a single type of thing that falls under the extension of ‘evidence’; but there may well still be an over-arching or unifying role that these things all play, one that suits evidence to fulﬁll all the various demands (1–4) that have been placed upon it. In fact, I think that Kelly comes close to identifying that common, unifying function when he ventures that the ‘root notion of evidence’ is that of a ‘reliable sign, symptom, or mark’ (2006: }3). Close, because, rightly understood, the ‘indicator’ conception of evidence does not in fact prevent evidence’s fulﬁlling its other roles, including that featured by evidentialism.

7 Cohen (1984) gives a forceful presentation; apropos of evidentialism, see e.g. Feldman (2003: 94–5); Conee and Feldman (2001: 60–1).

212

EVIDENCE

13.3 Reid on evidence (and self-evidence) On the question of what sorts of things count as evidence, Reid is clearly a pluralist: he says that ‘[w]e give the name of evidence to whatever is the ground of belief ’ (IP II 20, W 328a); and, he thinks, there are different types or sources of evidence: there is the evidence of sense, of memory, of consciousness, of axioms, of reasoning, and so on (IP IV 20, W 328a).8 Various kinds of experiences (perceptual, memorial, introspective), arguments, testimony, the judgment of recognized authorities, the marks or signs by which we distinguish between kinds of things, a person’s past actions, various ‘signs’ of another’s mind and/or character (gestures, facial expression, etc.), observed connections in the world—these are all things which Reid seems to count as evidence.9 In some cases, the evidence prompts belief due merely to some natural principle of our constitution (e.g. the ‘original perception’ whereby a given sensation serves as a ‘sign’ of hardness—see IHM 6 20, W 182bff.; IP II 21, W 330bff.); in others, it requires the right kind of experience (e.g. in ‘acquired perception’, the original sensation or something perceived comes to serve as a sign of something else—hearing a coach passing, or seeing the sphericity of a ball [ibid.]). So too—moving away from perception—in some cases it requires the right kind of education, training, and/or specialized knowledge (to return to Kelly’s example, inferring from the presence of Koplik spots that the patient has measles). What do all of these things have in common? They seem to me to agree only in this, that they are all ﬁtted by Nature to produce belief in the human mind, some of them in the highest degree, which we call certainty, others in various degrees according to circumstances. (IP II 20, W 328b)

Nor does certainty attach only to demonstrative reasoning, according to Reid. Demonstrative and probabilistic reasoning differ only in these two respects: the former concerns necessary truths, the latter contingent truths (IP VII 3, W 481b);10 and probable evidence comes in degrees. But certainty, the ‘highest degree’ of evidence and belief (IP II 20, W 328b), is not reserved for necessary truths: That there is such a city as Rome, I am as certain as of any proposition in Euclid . . . (IP VII 3, W 482b) [M]y senses give me as immediate conviction of what they testify, as my understanding gives of what is commonly called an axiom. (IP II 20, W 329a)

Notice: Reid says that ‘[w]e give the name of evidence to whatever is the ground of belief’. He says too: the various forms of evidence are all ‘ﬁtted by Nature to produce

8

For further discussion of Reid’s views on evidence and self-evidence, see Rysiew (2005). The discussion is not always explicit and systematic. E.g., in comparing ‘the evidence of sense’ with that of reasoning and consciousness, Reid doesn’t come right out and say just what the evidence of sense is. Some of the relevant passages include: IP II 20, W 328aff.; VII 3, W 481bff; VI 5, W 441aff. 10 A slight complication is that Reid allows that there can be probabilistic evidence of necessary truths (ibid.). Also, in terms of the latter, Reid clearly at times has in mind analytic truths—e.g., IP II 20, W 330a. 9

MAKING IT EVIDENT

213

belief in the human mind’. Is ‘ground’, then, synonymous with ‘cause’? No. For one thing, Reid freely mixes talk of evidence with talk of ‘just’, ‘good’, and ‘reasonable’ belief: All men of common understanding agree, that each [kind] of evidence may afford just ground of belief. (IP II 20, W 328a) All good evidence is commonly called reasonable evidence, and very justly, because it ought to govern our belief as reasonable creatures. (Ibid.)11

Such passages suggest that evidence is ‘what makes us justiﬁed in our beliefs’ (Lehrer 1989: 114; emphasis added). Further, while Reid does tie evidence very closely to the production of belief— ‘when we see evidence, it is impossible not to judge’ (IP VI 2, W 415a), he says— passages like the following also show that it’s implausible that Reid wants to deﬁne it as what leads to, even compels, assent: . . . such is the constitution of the human mind, that evidence discerned by us, forces a corresponding degree of assent. And a man who perfectly understood a just syllogism, without believing that the conclusion follows from the premises, would be a greater monster than a man born without hands or feet. (IP VI 5, W 448a)

But if such a monster is a coherent possibility it can’t be that evidence is deﬁned as what forces assent (Van Cleve 1999: 18). So if ‘evidence’ is ‘whatever is the ground of belief’, it can’t be that ground is a purely psychological notion. However, as we’ve seen, it’s implausible too to think that ‘ground’ is a purely normative/evaluative notion. What to say? As I’ve argued elsewhere (Rysiew 2005), Reid’s way out of this impasse is to reject the idea that we have to choose between the psychological-descriptive and the normative-evaluative readings of ‘ground’—hence, of ‘evidence’. How might that go? It’s rather natural, and certainly within the philosophical tradition, to think of ‘evidence’ in argumentational terms: evidence is that which bears the right kind of logical or probabilistic relation to some claim/belief (cf. Greco 1999: 273 and passim; 2002: 562; Alston 1989: 41); this view encourages thinking of evidence in sentential or propositional terms since, one might think, only such entities as these have the right kind of structure to serve as premises in an argument (cf. Kelly 2008: 941). But this isn’t the only truth-linked notion of evidence. For instance, there is a reliabilist conception (cf. Greco 2002: 562; Alston 1985: 437): if a given belief forming process is reliable, then beliefs produced thereby are ipso facto justiﬁed, and may serve as evidence for other believed propositions; alternately, one’s evidence for a given

11 Other examples: ‘To believe without evidence is a weakness which every man is concerned to avoid’ (IP II 20, W 328a); ‘I shall take it for granted that the evidence of sense, when the proper circumstances concur, is good evidence, and a just ground of belief ’ (ibid., W 328b).

214

EVIDENCE

proposition may consist in whatever experiences, etc., lead in some reliable process to a belief that that proposition is true. Each of these conceptions of evidence faces problems. The former seems to ﬁt badly with the idea that experiences are evidence, since they appear not to be propositional in form. The latter, like the view of evidence as a ‘reliable sign’ (Section 13.2), is compatible with there being no essential connection between one’s evidence and the information one has available to one, from a ﬁrst-personal point of view. Whether or not these problems are insuperable, there is another way of thinking of evidence that preserves its essential connection with truth. Here, instead of beginning with the abstract noun (‘evidence’), we take evidentness as the root notion and treat the nominative ‘evidence’ in derivative terms, as that which makes something evident (manifest, etc.). Just as light makes manifest visible objects, evidence is the voucher for all truth (IP VI 5, W 448a). As against the argumentational view (but like the reliabilist view) there is no restricting evidence to sentence-like entities (perceptual experience, say, can vouch for the existence of some object). And as against the reliabilist view, it is not the bare fact of reliability that deﬁnes evidence. The connection with truth, again, is secured via the notion of evidentness: for something to be evident is for it to be manifestly true; that’s why, when I say, ‘It’s obvious [evident, manifest] that p’, or, ‘X makes it manifest [evident, obvious] that p’, I am thereby committing myself as to p. And, on the assumption of the general reliability of our faculties (see below), those things which we ‘comprehend . . . clearly and without prejudice’ (IP VII 3, W 482b) and judge it to be evident (hence, true) generally will be such. But while the simple ‘evident that’ is thus naturally read as a factive operator,12 evidentness is not an all-or-nothing matter; and we usually reserve ‘evident’ and related unqualiﬁed terms (‘manifest’, ‘obvious’) for things exhibiting the quality of evidentness to some rather high degree—as Conee puts it, ‘[i]t is plausible that only very strong evidence can make a proposition evidently true’ (1992: 252). And, in addition to our sometimes mistaking the degree to which something is made evident by something else, there can of course be misleading evidence—something that indicates (makes evident, to some degree) that p even though p is false. Indeed, even the most evident propositions could—logically, anyway—be false (again: see below). Even so, the important point is that the concept of evidentness—hence, of evidence itself—can’t be deﬁned independently of truth. Further, though, as the notion of something’s being evident to someone makes clear, ‘evidence’ on the present view—that which makes evident/manifest—is also a (partly) psychological notion. We already knew that, of course: insofar as one has a sound understanding and comprehends some matter clearly and without prejudice (IP VII 3, W 482b; cf. IP II 20, W 328a), evidence produces belief to whatever degree is ﬁtting. But the point here is that when something is made evident to me I see it as evident, and 12 Chisholm, by contrast, focuses on the subjectivized notion of something’s being ‘evident for S’, which is, he says, not factive (1989: 11–12).

MAKING IT EVIDENT

215

recognize it to be true. The conception of evidence, then, is doubly psychological: evidence doesn’t just cause my belief; it assures me of its truth (Lehrer 1989: 114). And whereas the reliability per se of a belief’s source need not be something to which the subject has access, at least in the case of things which are evident to some high degree, that evidentness does disclose itself.13 When something is evident to me, it is evident (to me) that it is evident to me: Perhaps evidence, as in many other respects it resembles light, so in this also—that, as light, which is the discoverer of all visible objects, discovers itself at the same time, evidence, which is the voucher for all truth, vouches for itself at the same time (IP VI 5, W 448a)

But besides these truth-linked and psychological aspects there is, further still, some normative-evaluative content to the notion of evidentness/manifestness: saying that something is evident, like saying that it’s obvious, implies that it’s something that (similarly situated and informed) others ought to recognize/accept as well. Thus, for example, if someone is standing outside in a downpour but fails to accept that it’s raining, we’d say (other things equal) that this revealed that something wrong with them, though this failing would be a case of improper functioning, rather than of epistemic irresponsibility.14 Now, it’s important to emphasize once again that thinking of evidence in terms of this ‘mixed’ notion of evidentness (/that which makes evident)—its being both psychological and evaluative—is not an alternative to taking evidence to be an inherently truthlinked notion. The difference is not that this way of thinking about evidence severs the semantic connection with truth, but that it also makes clear the connection to belief, to evidentness disclosing itself to the subject, and to some kind of normative-evaluative notion of what one ought to believe. When Reid says that the different kinds of evidence ‘are all ﬁtted by nature to produce belief in the human mind’ (IP II 20, W 328b), he clearly means the sound or ‘healthy’ human mind.15 So, just as it’s a contingent matter that evidentness tracks relations of objective evidential support, it’s contingent too that ‘evidence discerned by us forces a corresponding

13 Which is not to say that the evidence always discloses itself. For Reid, it’s not always easy to say in what one’s evidence consists (IP II 20, W 328a). On this point, there appears to be no real disagreement between Reid and Feldman and Conee, however: e.g., the latter’s ‘mentalism’ is explicitly contrasted with accessibilism (Conee and Feldman 2001: 54–8); beyond the mentalist thesis, Feldman and Conee’s discussions of that in which one’s evidence consists take the form of suggestive/illustrative examples, rather than precise characterizations; and, in defending internalism, Feldman disavows the assumption that ‘people are infallible with respect to whether they can identify a good reason’ (2005: 278, 281)—which given his (Feldman and Conee’s) usage amounts to acknowledging fallibility with respect to identifying evidence. 14 Here too, it is testament to the resiliency of the relevant notion that even those outside the epistemological mainstream acknowledge it. In the present case, even Quine—and in an explicitly ‘behavioral’ moment, no less—says that an acceptable translation must preserve more than just rudimentary logicality: ‘the native’s unreadiness to assent to a certain sentence gives us reason not to construe the sentence as saying something whose truth should be obvious to the native at the time’ (1970: 82). 15 E.g.: ‘ . . . in most cases, we measure the degrees of evidence by the effect they have upon a sound understanding, when comprehended clearly and without prejudice’ (IP VII 3, W 482b).

216

EVIDENCE

degree of assent’ (IP VI 5, W 448a16). Again, in some cases the latter depends upon our having had the appropriate experiences, or on our possessing the requisite knowledge or expertise. More fundamentally, though, it depends on our constitution—on our possessing both those speciﬁc ‘principles’ that connect various experiences, for example, with a conception and belief of what they ‘suggest’ (e.g., IHM 5 III, W 122a–b; IP II 21, W 332a–b), and common sense—that degree of judgment ‘which is common to men with whom we can converse and transact business’ (IP VI 2, W 421b), ‘which makes a man capable of acting with common prudence in the conduct of life, [and] . . . of discovering what is true and what is false in matters that are self-evident, and which he distinctly apprehends’ (ibid., W 422b). ‘Matters that are self-evident’ brings us to Reid’s ‘ﬁrst principles’,17 which speak to (inter alia) the general reliability of our various faculties (IP VI 5, W 447a). Concerning as they do contingent matters, these principles are indemonstrable; but they’re no less certain for that. Nor does our being certain about them mean that it is wrong to speak of them as having evidence in their favor18—it’s not as though evidence is something we require or fall back only when we aren’t in a position to grasp some truth ‘in an utterly direct, unmediated way’ (Kelly 2006: }3; cf. Kelly 2008: 942). All good believings are evidence-based, according to Reid. Objections to the contrary (see, e.g., Plantinga 1993: 186–93) presume an overly narrow view of the evidence available in a particular case. Thus, though the evidence’s inﬂuence is much more easily felt than described (IP II 20, W 328a), one needn’t resort, for example, to the claim that one’s evidence for memorial, a priori or de se knowledge just is that feeling, or ‘impulsional evidence’ that’s not really distinct from the belief itself (see Conee and Feldman 2001: 64–7). By the same token, however, while it’s only when a given proposition has especially good evidential backing (and so is made particularly evident) that there is something worth calling a ‘sense of obviousness’ about it (or evidentness simpliciter) (see ibid.: 66), any ‘evidence for a proposition, however weak, [provides] some indication that the proposition is true’ (Conee 1992: 253; emphasis added). In the case of self-evident propositions, such evidence is carried by the propositions themselves:

16 Again: ‘Every degree of evidence perceived by the mind, produces a proportioned degree of assent or belief’ (IP VII 3, W 482b). Notice that ‘evidence’ in this passage pretty much requires to be read as ‘evidentness’. Note too that Locke, anyway, often favors the latter term—as, e.g., in Book IV, Chapter 2 of the Essay. 17 In particular, the ﬁrst principles ‘of contingent truths’ (vs. of necessary truths, of morals, and so on); it is these that are most obviously and straightforwardly relevant to the concerns of epistemologists. 18 Reid says: ‘That there is such a city as Rome, I am as certain as of any proposition in Euclid; the evidence is not demonstrative, but of that kind which philosophers call probable. Yet in common language it would sound oddly to say, it is probable that there is such a city as Rome, because it would imply some degree of doubt or uncertainty’ (IP VII 3, W 482b). In the same way, it might seem odd to say that we have evidence that such-and-such, when we in fact have no doubt about it. Cf. Austin’s response to Wisdom’s looking in the larder and ﬁnding ‘signs’ of bread when he sees the loaf, touches it, tastes it, and so on; talk of signs, Austin thinks, makes sense only where there is some doubt about whether it is bread we’re eating, as there manifestly isn’t in the case Wisdom describes (Austin 1946/1979: 106ff.)

MAKING IT EVIDENT

217

[They] are no sooner understood than they are believed. The judgment follows the apprehension of them necessarily, and both are equally the work of nature, and the result of our original powers. There is no searching for evidence, no weighing of arguments; the proposition is not deduced or inferred from another; it has the light of truth in itself, and has no occasion to borrow it from another. (IP VI 4, W 434a)

Just as evidence isn’t merely what causes belief, self-evidence (/-evidentness) isn’t simply the inevitability of the relevant beliefs. Reid regards the ﬁrst principles as constitutive principles, in the sense that accepting them is a condition (for us, given our nature) of cognizing at all.19 A failure to accept the ﬁrst principles of common sense is, Reid thinks, just plain and literal lunacy. The ﬁrst principles are, for us, and in our view, the ‘ﬁxed point’ upon which cognizing rests (cf. IP VI 4, W 435a). So, while the distinction between self-evidence and being strongly inclined to believe a proposition is real, and is easy to draw at the level of non-basic propositions (especially where there is unclarity about whether something is evident), since the ﬁrst principles are ﬁrst principles their being self-evident and our being strongly inclined to believe them are, as one might expect, not in practice separable. So too, Reid thinks, for our accepting the ﬁrst principles and our being justiﬁed in accepting them: they typify, even deﬁne, what (self-)evidentness is for us, given our constitution. If this seems unclear or implausible, note that essentially the same point holds for our beliefs about obvious necessary or analytic truths: it is not their necessity or analyticity per se which leads me to accept these things as true. (There are many analyticities I’ve never even considered, and many (other) necessary truths I don’t recognize as such.) What gets me accepting such things is my ‘seeing’ that they can’t be false, my inability to understand how things could turn out such that they’re not true. Reid, like Descartes, allows that even simple mathematical judgments are not impervious to error. Further, Reid thinks that we can conceive of impossible things—our ability to use reductio ad absurdum arguments requires it (IP IV 3, W 379a). So while the ordinary use of ‘conceivable’ disguises the distinction (ibid., W 377a), even for analyticities, it’s not the inconceivability but the de facto unbelievability of their negations that accounts for our thinking them true (ibid., W 375a–379b, esp. 378a). Hence Reid’s saying that ‘the rules of demonstrative sciences . . . have no authority but that of human judgment’ (IP VII 4, W 486a). Thus, whether it concerns contingent or non-contingent truths, the (self-) evidentness or simple manifestness of certain things, the bruteness of certain such judgments, is in the end the ﬁnal court of epistemic appeal.20 If evidence and evidentness straddle the boundary between the psychological and the normative, self-evident 19

This idea is spelled out at greater length in Rysiew (2002). Greco has suggested that Reid may plausibly be read as a reliabilist about evidence, and that ‘once we are reliabilists about evidence, evidence is not so special any more’ (2002: 562)—that is, it is not important that we have evidence which bears some necessary relation (either logical or probabilistic) to the beliefs that it makes evident. I see no reason why Reid would object to the latter claim. But as Greco notes (2002: 562), Reid rejects this narrow, rationalistic conception of evidence; evidence, in the sense of evidentness, does play a crucial role in his epistemology. 20

218

EVIDENCE

propositions mark the limits of this boundary for us, as the epistemic subjects that we are. These ideas suggest an answer to doubts that Earl Conee has raised about the utility of appeals to obviousness in elucidating the concept of evidence, and of self-evidence in particular.21 Conee says that when we ask after the evidence of those propositions we’re inclined to regard as ‘self-evident’, in many instances we are inclined to report that the proposition is ‘just obviously true’. (2011: }1.2). ‘That does not,’ he says, ‘sound like a report of some evidence’ (ibid.). And later: [T]he obviousness in obvious truth consists i[n] an open manifestation—of either a justiﬁcation or a truth-sufﬁcing fact. The phrase ‘ground for truth’ covers both possibilities. In those terms, the obviousness of something’s truth is the open availability with which a ground for its truth is presented. Obviousness characterizes the accessibility of a ground. It is not itself the ground. The ground is what gives justiﬁcation. We add ‘just’ to ‘obviously true’ as an expression of an explanatory incapacity. We ﬁnd a proposition to be ‘just obviously true’ when we cannot further identify the manifest justiﬁer or truth-maker, except to say that it is manifest. (Ibid.: }3.2)

How might Reid respond? He would resist the attempt to drive a wedge here between the accessibility of the ground and the ground itself.22 In the case of self-evident propositions, what grounds belief is just the evidentness of the proposition in question. Remember: evidence for Reid resembles light; in terms of that comparison, it’s as though Conee wishes to pull apart light’s disclosing various things to us and its being visible itself. Calling something ‘just obviously true’ may not be so much ‘an expression of an explanatory incapacity’—an inability to ‘further identity the manifest justiﬁer’— as an acknowledgment of the fact that there is no deeper ground one might cite.

13.4 Revisiting the roles: the Reidian view as a unifying account It was noted early on that Reid is a pluralist about the types of things falling under the extension of ‘evidence’: there are different types of things ‘ﬁtted by Nature to produce belief in the human mind’ (IP II 20, W 328b) when they are carefully considered by someone with a sound understanding, without prejudice and, in some cases, with the requisite experience and/or knowledge. This they do by making evident to us various things. This is the ultimate or fundamental role of evidence: evidentness is, one might say, the face of truth; and evidence is what indicates and impresses truths, as such, upon the (healthy, informed, etc.) mind. In terms of this idea, the roles of evidence discussed at the outset can in turn be explained.

21 Which is not to say that they put to rest all of the various issues Conee raises in the course of his discussion of self-evidence (or ‘self-support’). 22 Again, we’re speaking here of self-evident propositions. With others, this distinction does get a grip.

MAKING IT EVIDENT

219

First, while Reid tends not to speak in terms of ‘justiﬁcation’, it’s pretty clear that evidence, as he understands it, is what justiﬁes beliefs. ‘Evidence’ refers to ‘whatever is the ground of belief’ (IP II 20, W 328a). But, as we’ve seen, ‘ground’ isn’t a merely causal notion: it is something like a proper cause of belief. Hence the claim that ‘the evidence of sense, when the proper circumstances concur, is good evidence, and a just ground of belief’ (ibid., W 328b), and that ‘[a]ll men of common understanding agree, that each [kind] of evidence may afford just ground of belief’ (IP II 20, W 328a). Further, Reid’s view seems to be that, in a healthy (informed, etc.) mind, evidence is the only thing that matters to the justiﬁedness of belief (role 1)—if evidence is what ‘grounds’ belief, any ground (or proper cause) of belief should be understood in evidentiary terms. On a straightforward reading of the evidentialist thesis, then, Reid’s view vindicates it. Equally straightforward is the endorsement of the idea that rational thinkers respect their evidence (2). This, Reid appears to take as not really in need of argument: All good evidence is commonly called reasonable evidence, and very justly, because it ought to govern our belief as reasonable creatures; (IP II 20, W 328a) To believe without evidence is a weakness which every man is concerned to avoid, and which every man wishes to avoid. (IP II 20, W 328a)

However, that evidence justiﬁes belief and is something to which rational thinkers must be responsive doesn’t mean that considerations of reliability are beside the point in understanding evidence; and this is just as well since, on the face of it, the reason that justiﬁcation, evidence, rationality, and related notions are of epistemic interest is because they are our (defeasible) guides to truth. So, consider again Kelly’s suggestion that ‘[p]erhaps the root notion of evidence is that of something which serves as a reliable sign, symptom, or mark of that which it is evidence of’ (2006, }3) (3). From Reid’s perspective, that’s almost but not quite right. Recall that Kelly describes a tension between thinking, on the one hand (as with (3)), that it is a connection with truth which marks something out as evidence and, on the other hand, thinking that evidence is what justiﬁes belief (1), or is something which rational thinkers must respect (2). (Or, equally, between the thought ‘that evidence should be the kind of thing which is, in general, a reliable indication of that for which it is evidence, and that it should be relatively easy to recognize’ (Kelly 2008: 943).) Whether there is such a tension, however, depends upon how exactly this truth connection is spelled out. What does give rise to the tension in question is a bare reliable indicator conception of the connection between evidence and truth: so long as there is some objective fact-indicating relation between X (/its occurrence) and some proposition or state of affairs, the former may be said to be evidence of the latter. Again, though, on Reid’s view it is not the mere fact of reliability that deﬁnes evidence; the connection with truth is secured via the notion of evidentness: to take something as evident or manifest is to take it to be (at least) true; and Reid thinks that our faculties are generally reliable—hence, that those things which are (to whatever degree) evident to

220

EVIDENCE

us are likely (to whatever degree) to be true. But when something is evident (to whatever degree) to me, it is also evident to me that that’s so; the reliability of the indication is itself indicated.23 Insofar, then, as the tension between (3) and each of (1) and (2) is derived from the possibility that one could somehow be in possession of evidence which one doesn’t, even couldn’t, recognize as such, it doesn’t arise here. Interestingly, this way of understanding the truth connection is very much of a piece with what Conee says about the latter: a proposition is justiﬁed to S when it is (very) evident to S that it is true; and the evidence for a proposition is simply an indication of its truth (1992: 252). Yet, Feldman and Conee hold a different view to Reid as to the sorts of things which can qualify as evidence. As we’ve seen, on Reid’s rather liberal view evidence includes such varied things as propositions, states of consciousness, perceptual experiences, memorial seemings, statements, others’ sayings and gestures, the rings on a tree, smoke on the horizon, the distinctive coloring of some type of bird, and so on. Whereas, as we saw above, some of the staunchest defenders of (1) and (2), including Feldman and Conee, think that evidence’s fulﬁlling these roles means that evidence must be ‘the information one has to go on’ (see above), and that this in turn favors mentalism about justifying factors (hence, evidence). Again: While we might ordinarily say that your reason for thinking that the tree is a maple is that its leaves are a particular shape, the fact that the leaves are that shape is not part of your evidence. What you are going on in judging the tree to be a maple is your belief that it has leaves of a particular shape, and perhaps ultimately you are going on how the tree looks to you (your perceptual experience). These are internal, mental states you are in. (Feldman 2005: 273)24

Granted: the shape of the leaf can and should make a difference to what you believe only insofar as you are aware of it. Similarly, suppose, absent an awareness of it, ‘[t]he sheer reliability of the cause of a belief’ is justiﬁcatorily impotent (Conee 1988: 49). But it doesn’t follow that the needed state of awareness—be it a belief, experience, or some other mental state—is the evidence. Such an awareness seems, if anything, to be a condition on the evidence’s being evidence that one ‘possesses’ in the relevant sense of that phrase; and it is the evidence one possesses in terms of which the justiﬁedness of one’s beliefs, or the rationality of one’s epistemic ‘actions’, should be assessed; but nothing immediately follows about the ontology of evidence. 25

23 Recall the analogy with light: ‘Perhaps evidence, as in many other respects it resembles light, so in this also—that, as light, which is the discoverer of all visible objects, discovers itself at the same time, evidence, which is the voucher for all truth, vouches for itself at the same time’ (IP VI 5, W 448a). 24 Conee describes essentially this form of argument—an argument from ‘reﬂective access’, as we might call it—as ‘[t]he best grounds for accepting internalism’ (1988: 48). I do not think that the argument establishes the need for reﬂective access speciﬁcally; and I do not think that the requirement of having access to justifying factors favors mentalism about justiﬁers (evidence). 25 Of course, if one is persuaded that there’s no good sense in which worldly things (facts, events, etc.) are accessible to the mind, that might favor mentalism: (a) one should make one’s beliefs ﬁt the evidence, but (b) the only thing(s) to which one ‘really’ has access are one’s own mental states; therefore (c) mentalism. But (b) here needs arguing, particularly if one’s opponent is a ‘direct realist’ such as Reid (see, e.g., Van Cleve 2004).

MAKING IT EVIDENT

221

Similar considerations apply to another main form of argument for mentalism (e.g. Conee and Feldman 2001: 58–61): we consider paired cases wherein there is some intuitive difference between the justiﬁedness of the subjects’ beliefs, but where the only salient difference between them is a mental difference. Thus, S1 and S2 both believe that it’s warm outside, but only S2 has felt the warmth; S1 and S2 both judge some bird to be a woodpecker, but only S2 has expert knowledge of woodpeckers’ distinctive markings. In both cases, we judge S2’s belief to be better justiﬁed (other things equal). Mentalism requires that ‘these differences have an entirely mental origin’ (ibid.: 61). But that’s far from clear: granted that crude reliabilism about evidence (justiﬁcation) is implausible, one could perfectly well insist that the mental differences in such cases make an epistemic difference only because they are among the relata—the most salient relata, perhaps, in the cases described—in some reliable process (perception, inductive reasoning, etc.). Again then: that the difference in mental states is intrinsically differencemaking needs to be argued. ‘But what of subjects in the demon world? Aren’t those beliefs justiﬁed, rational, and based on their evidence, even though here the connection with truth has been entirely severed?’ (cf. Feldman 2003: 94–5; Conee and Feldman 2001: 60–1). The Reidian view might seem to recommend a negative answer: various things appear evident to the demonly deceived; but (almost) nothing really is (the straight ‘evident that’ is factive, recall); so those in ‘the bad case’, as Williamson (2000a, 2000b) puts it, lack the evidence which their undeceived counterparts enjoy; so, by (1), their beliefs can’t be justiﬁed.26 To the extent that that strikes us as wrong, reliability can’t be necessary for justiﬁcation; or, in the present case, actual evidentness can’t be required for it either, nor for the presence of genuine evidence. I don’t myself ﬁnd a ‘heroic’ (Sosa, in Bonjour and Sosa 2003: 159) response to the new evil demon problem—claiming that those in the bad case do lack justiﬁcation (and/or genuine evidence) for their beliefs—obviously unacceptable. After all, it seems natural to say that it’s an essential feature of the gross and utter epistemic misfortune of those in the bad case that they really have no genuine evidence at all; all of their evidence is ‘false’ (i.e. merely apparent) (cf. Williamson 2000a: 623 n. 10; cf. Kelly 2006: }2). Still, they may be blameless; and so they—the subjects—may be justiﬁed in holding their beliefs, even if the beliefs themselves are not (Bach 1985: 251–3). Still, the Reidian isn’t forced to adopt the ‘heroic’ line. As a number of writers have observed (Plantinga 1993; Greco 2002, 2004; Bergmann 2008), and as should be clear from the discussion of Section 13.3, when Reid says that the various forms of evidence ‘are all ﬁtted by Nature to produce belief in the human mind’ (IP II 20, W 328b), he has in mind a sound human mind; and he has in mind too a certain sort of environment—viz., the sort of environment in which our faculties are supposed to function. So it’s very natural to sink Reid’s views on evidence into—or, rather, to see them as 26 This would, of course, be the evidentiary analogue of Williamson’s response to the new demon problem (2000a, 200b).

222

EVIDENCE

presuming—a broadly ‘proper functionalist’ framework. (Or, equally, they presume a theory of the etiological functions of our faculties or belief-forming mechanisms—see Graham 2011; the present point is equally well served by either view.) We can then say that the subject’s beliefs in the bad case are justiﬁed because in the sort of environment in which their belief-forming faculties arose, or for which those faculties are ‘designed’, what appeared evident to them would generally be such (hence, their beliefs would be reliably formed and based on genuine evidence).27 Thus, we can preserve the thought that these subjects’ beliefs are justiﬁed. But we can also insist that focusing on the respect in which subjects in the demon world are just like us— viz., in respect (suppose) of the purely ‘internal’ facts—belies the fact that they are epistemically quite different from us, and not just because all of their beliefs are false, or just because their beliefs are not reliably formed. Just as importantly, the normal interface between mind and world—including various truths’ being evident to them—has been hijacked.28 Does this represent a back-tracking with respect to (1)? Is the justiﬁedness of a belief really not a function merely of one’s evidence after all? Not quite: the justiﬁedness of a belief is solely a function of one’s evidence, provided that one is in possession of ‘a sound understanding’ and in roughly the kind of world we take ourselves to be in. The latter requirements are what binds what appears evident to us to the truth. But those requirements—which amount to something very like the assumption of the reliability of our faculties—are also, I would suggest, very much things that are taken for granted when those who consider the evidentialist thesis (and/or (1)) judge it to be plausible. Absent that assumption, it’s not clear whether we would place any particular epistemic value upon evidence.29 Lastly, consider (4): the idea that evidence can function as a neutral, intersubjective arbiter—hence, that it enjoys a certain degree of objectivity and publicity. Once again: the core notion in thinking about evidence as Reid conceives of it is, I claim, that of evidentness: evidence is that which makes things evident, to varying degrees. But isn’t evidentness (to a person) a psychological phenomenon? And if so, isn’t it essentially private, and so not suited to play the role of publicly available neutral arbiter? No, that doesn’t follow: First, evidentness may be psychological, but—with the possible exception of self-evident propositions—evidentness is distinct from evidence, and the latter can be as publicly available as you please. (This is one of the advantages of 27 Bergmann (2008: }I(D)) is especially emphatic in recommending the preceding as the proper Reidian response to the new evil demon problem. 28 So too, we avoid the potential costs that come with insisting that there’s no epistemic difference, and no evidentiary difference, between the good and bad cases: see, e.g., Kelly (2006, 2008); Williamson (2000a, 2000b). 29 Cf. Bonjour: ‘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 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’ (1985: 7–8).

MAKING IT EVIDENT

223

the present view over mentalism.) Not all evidence is like that, of course—the introspective evidence I have for my belief that I’ve got low blood sugar isn’t publicly available (Kelly 2006: }4). But that isn’t news: plausibly construed, what (4) requires is that evidence can be such as to play the role of neutral arbiter; and, on the present view, it can. Of course, that publicly available evidence can’t play the role of neutral arbiter unless people’s responses to it exhibit a certain robust similarity. But given certain nonskeptical assumptions, people’s (considered) judgments as to the evidentness of a given proposition can be expected to track the relevant facts (truths); and given substantial psychological commonalities across subjects, those judgments themselves can serve as intersubjective, neutral arbiters of belief. This, of course, won’t please someone who thinks that any ‘neutral arbiter’ worthy of the name must not include or make any essential reference to human judgment, even when the latter is somewhat idealized (e.g. restricted to those with sound understandings, a lack of prejudice, and perhaps the requisite education and training as well; cf. IP VII 3, W 482b; IP II 20, W 328a). But if the need for a ‘neutral arbiter’ derives from a desire to rule out, or at least combat or minimize, the role of dogma, prejudice, and so on in our intellectual lives, there is no reason Reidian evidence can’t play that role. Further, as we’ve seen, it’s one of Reid’s central points that there is no escaping the role that human judgment plays in shaping and under-girding our epistemic lives.

13.5 Conclusion As usually stated, evidentialism constitutes the outline of a theory. The present theory of evidence is in many ways programmatic too. But it is, it’s hoped, an attractive theory: beginning with a conception of evidence as that which makes evident, we’ve seen how we might begin to construct a uniﬁed theory of evidence, one that avoids the pitfalls of both an externalist theory which ignores the epistemic signiﬁcance of the subject’s perceptive and an internalist theory which denies any essential connection between evidence (/justiﬁcation) and truth. In addition, we’ve seen how psychological and epistemic notions might come together in our understanding of certain fundamental epistemic phenomena: if it’s true, as Conee says, that ‘[t]he relation of giving evidence is not obviously within the ontology of any current or prospective science’ (1992: 253),30 that of producing belief certainly is. So too, it’s not as though we lack any good pre-epistemological grasp of such things as a given mind’s not functioning as it should, an environment’s being abnormal, someone’s being inﬂuenced by one or another kind of ‘prejudice’, as opposed to by experience and relative expertise, and so

30 Conee’s point is not to raise worries about the naturalistic acceptability of evidentialism or related notions; as he goes on to point out (1992: 253–4), there’s nothing in the view that’s clearly contrary to a plausible naturalism. For discussion of Reid’s relation to current debates surrounding epistemic naturalism, see Rysiew (2002).

224

EVIDENCE

on. Insofar, then, as evidence is what indicates and impresses truths, as such, upon the (healthy, informed, suitably situated, etc.) mind, it’s perfectly scientiﬁcally investigable; and its effects are no more mysterious, and no less, than the fact that some things are just more or less obvious.31

References *

Pagination of in-text citations follows that of the reprint, where listed

Alston, W. P. 1985. ‘Thomas Reid on Epistemic Principles,’ History of Philosophy Quarterly, 2: 435–52. ——. 1989. ‘Reid on Perception and Conception’, in M. Delgarno and E. Matthews (eds), The Philosophy of Thomas Reid, pp. 35–47. London: Kluwer. Austin, J. L. 1946. ‘Other Minds,’ Proceedings of the Aristotelian Society, Supp. Vol. 20. Reprinted in Austin’s Philosophical Papers, 3rd edn. New York: Oxford University Press, 1979: 76–116. Bach, Kent. 1985. ‘A Rationale for Reliabilism,’ The Monist, 68: 246–63. Bergmann, Michael. 2008. ‘Reidian Externalism’, in Vincent Hendricks and Duncan Pritchard (eds), New Waves in Epistemology. Basingstoke: Palgrave Macmillan. Bonjour, Lawrence. 1980. ‘Externalist Theories of Empirical Knowledge’, in Peter French et al. (eds), Midwest Studies in Philosophy, Vol. V, pp. 53–73. Minneapolis: University of Minnesota Press. ——. 1985. The Structure of Empirical Knowledge. Cambridge MA: Harvard University Press. —— and Ernest Sosa 2003. Epistemic Justiﬁcation: Internalism vs. Externalism, Foundations vs. Virtues. Malden MA: Blackwell. Chisholm, Roderick. 1964. ‘The Myth of the Given,’ in Chisholm et al., Philosophy. Englewood Cliffs, NJ: Prentice-Hall, 261–86. ——. 1989. Theory of Knowledge, 3rd edn. Englewood Cliffs, NJ: Prentice-Hall. Clifford, William K. 1877/1901. The Ethics of Belief, reprinted in Lectures and Essays, London: Macmillan, 1901. Cohen, Stewart. 1984. ‘Justiﬁcation and Truth,’ Philosophical Studies, 46: 279–95. Conee, Earl. 1988. ‘The Basic Nature of Epistemic Justiﬁcation,’ The Monist. Reprinted in Conee and Feldman 2004: 37–52. ——. 1992. ‘The Truth Connection’, Philosophy and Phenomenological Research, 52: 657–69. Reprinted in Conee and Feldman 2004: 242–58. ——. 2011. ‘Self-Support’, Philosophy and Phenomenological Research. Available at http://onlinelibrary.wiley.com/journal/10.1111/(ISSN)1933-1592/earlyview. —— and Richard Feldman. 2001. ‘Internalism Defended,’ in Hilary Kornblith (ed.), Epistemology: Internalism and Externalism, pp. 231–60. Oxford: Blackwell. Reprinted in Conee and Feldman 2004: 53–82. —— and ——. (eds) 2004. Evidentialism: Essays in Epistemology, Oxford: Clarendon Press.

31 An ancestor of this paper was presented at the University of Alberta; my thanks to those who attended for helpful comments and discussion.

MAKING IT EVIDENT

225

Cuneo, Terence, and Rene´ van Woudenberg (eds) 2004. The Cambridge Companion to Reid, Cambridge: Cambridge University Press. Feldman, Richard. 2003. Epistemology. Upper Saddle River, NJ: Prentice-Hall. ——. 2005. ‘Justiﬁcation is Internal,’ in Matthias Steup and Ernest Sosa (eds), Contemporary Debates in Epistemology, pp. 270–84. Malden, MA: Blackwell. —— and Earl Conee. 1985. ‘Evidentialism,’ Philosophical Studies, 48: 15–34. Reprinted in Conee and Feldman 2004: 83–107. Goldman, Alvin 1979. ‘What is Justiﬁed Belief?’, in George Pappas (ed.), Justiﬁcation and Knowledge. Reprinted in A. Goldman, Liaisons: Philosophy Meets the Cognitive and Social Sciences. Cambridge, MA: The MIT Press, 1992: 105–26. Graham, Peter J. 2011. ‘Epistemic Entitlement,’ Nous. Available at http://onlinelibrary.wiley. com/journal/10.1111/(ISSN)1468-0068/earlyview. Greco, John. 1999. ‘Agent Reliabilism,’ in James Tomberlin (ed.), Philosophical Perspectives, 13, Epistemology, pp. 273–96. Atascadero, CA: Ridgeview Press. ——. 2002. ‘How to Reid Moore,’ The Philosophical Quarterly, 52, 209: 544–63. Reprinted in Haldane and Read (eds) 2003: 131–50. ——. 2004. ‘Reid’s Reply to the Skeptic’ in Cuneo and van Woudenberg (eds): 134–55. Haldane, John and Stephen Read (eds). 2003. The Philosophy of Thomas Reid: A Collection of Essays, Oxford: Blackwell. Hume, David. 1748/1987. Enquiries Concerning Human Understanding and Concerning Principles of Morals, edited with an analytical index by L. A. Selby-Bigge, 3rd edn, Oxford: Clarendon Press. Kelly, Thomas. 2006. ‘Evidence’, The Stanford Encyclopedia of Philosophy (Fall 2006 Edition), Edward N. Zalta (ed.), URL = Last accessed 28 January 2011. ——. 2008. ‘Evidence: Fundamental Concepts and the Phenomenal Conception,’ Philosophy Compass, 3/5: 933–55. Lehrer, Keith. 1989. Thomas Reid, Oxford: Blackwell. Locke, John. 1690/1959. An Essay Concerning Human Understanding, the A. C. Fraser edition. New York: Dover Publications. Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford University Press. Quine, W. V. O. 1970. Philosophy of Logic. Englewood Cliffs, NJ: Prentice-Hall, Inc. Reid, Thomas. 1895/1983. Thomas Reid: Philosophical Works, 8th edn, ed. William Hamilton. New York: Georg Olms Verlag. Rysiew, Patrick. 2002. ‘Reid and Epistemic Naturalism,’ The Philosophical Quarterly, 52, 209: 437–56. Reprinted in Haldane and Read (eds) 2003: 24–43. ——. 2005. ‘Reidian Evidence,’ The Journal of Scottish Philosophy, 3, 2: 107–21. ——. 2008. ‘Rationality Disputes—Psychology and Epistemology,’ Philosophy Compass, 3. Van Cleve, James. 1999. ‘Reid on the First Principles of Contingent Truths,’ Reid Studies, 3, 1: 3–30. ——. 2004. ‘Reid’s Theory of Perception,’ in Cuneo and van Woudenberg (eds): 101–33. Williamson, Timothy. 2000a. ‘Scepticism and Evidence,’ Philosophy and Phenomenological Research, 60, 3: 613–28. ——. 2000b. Knowledge and Its Limits. Oxford: Oxford University Press.

14 In Defense of Propositionalism about Evidence Trent Dougherty

14.1 Introduction I will here offer a rebuttal to Conee and Feldman’s (2008) defense of a non-propositional theory of evidence against arguments put forward in Williamson 2000. First, I will set forth Williamson’s argument for the (exclusively) propositional nature of evidence in a way I think is regimented and explained for optimum clarity. Next, I set out and critique Conee and Feldman’s reply. Then, I give a new argument which makes many matters implicit in Williamson’s explicit. Finally, I consider the consequences for evidentialism of a propositional theory of evidence. On at least one view of propositionalism about evidence—inspired by Richard Jeffrey’s thoroughgoing rejection of C. I. Lewis’s ‘given’—the evidentialist supervenience thesis as Conee and Feldman formulate it comes out false.

14.2 Williamson’s argument Timothy Williamson argues (2000: }9.5, 194ff) that evidence consists in propositions (he further argues that they must be known propositions, but that is another matter). He gives three arguments for this conclusion, and they are admirably clear. They pertain to three principle roles evidence plays in epistemology. The arguments center around three different kinds of reasoning: explanatory, probabilistic, and deductive. The arguments all have essentially the following form. 1. Evidence is X. 2. X consists in propositions. 3. Therefore evidence consists in propositions. The X-factor varies by the three types of considerations above: X1 = that which is explained by hypotheses, X2 = the relata of probabilistic relations, X3 = the relata of deductive relations. Thus the three arguments are as follows.

I N D E F E N S E O F P RO P O S I T I O N A L I S M A B O U T E V I D E N C E

227

The Explanatory Argument (2000: 195) 10 . Evidence is the kind of thing which hypotheses explain. 20 . But the kind of thing which hypotheses explain is propositional. 30 . Therefore evidence is propositional. It is clear from context that by ‘propositional’ in (20 ) he does not just mean having a propositional content, but rather consisting in propositions. Furthermore, it is clear from the context that he doesn’t mean (20 a) Some of the kinds of things which hypotheses explain are propositional, but rather (20 b) The only kinds of things which hypotheses explain are propositional. He defends this thesis by reference to the nature of explanations as answers to why-questions. The basic syntax is as follows. Q: ‘WHY is it the case that X?’ A: ‘X is the case BECAUSE Y is the case.’ The ‘BECAUSE’ is ﬂanked by two indicative sentences expressing propositions. The ﬁrst sentence, expressing the explanandum, the item in need of explanation, is the indicative grammatical variant of what appears in the interrogative mood following the ‘WHY’. The indicative following the ‘BECAUSE’ expresses the explanans, the item that does the explaining. That X might be ﬁlled in with an event-term in ordinary speech is quite beside the point, for it would be mere synecdoche to refer to the event itself as needing explanation. As Williamson points out, the imperative ‘Explain World War II’ is a request to ‘explain why it occurred, or had some distinctive feature’ (2000: 195). The application to experiences is obvious. ‘[T]he sensation in my throat is evidence for the conclusion that I am getting a cold in the sense that the hypothesis that I’m getting a cold would best explain why I have the sensation in my throat’ (2000: 195, emphasis added). But the evidence is expressed in the sentence ‘I have a sensation in my throat.’ The Probabilistic Relation Argument (2000: 195–6) 100 . Evidence has probability. 200 . Only propositions have probabilities. 300 . Therefore, evidence consists in propositions. Premise 100 follows from theorems of the probability calculus (such as Bayes’s theorem) together with the fact that evidence confers probability. For example, Bayes’s theorem states that the probability of an hypothesis H given some item of evidence E—Pr(H/E)—is equal to a quotient that has in its numerator the converse probability— Pr(E/H). So if items of evidence confer probability on hypotheses, and Bayes’s theorem holds, it follows that items of evidence have probabilities. Again, it is not relevant that statisticians and formal probability theorists sometimes talk of the probability of an event, for this is clearly a shortcut for the probability of the event occurring. Premise 200 seems safely true in the context of epistemology (see Skyrms 2000: 23 and 109 on this). What it would be for an experience to have a probability is mysterious. That it should occur is something we might well assign a probability to however.

228

EVIDENCE

The Deductive Relation Argument (2000: 196–7) 1000 . Sometimes, evidence rules out hypotheses by being inconsistent with them. 2000 . But inconsistency is a relation deﬁned over propositions. 3000 . Therefore, evidence consists in propositions. I don’t expect Premise 1000 to be disputed. The notion of inconsistency in 2000 is logical inconsistency, underwriting the ability to deduce ~H from E. There is overlap here with the explanatory argument in that good hypotheses are often taken to entail the observational evidence. Yet only propositions can be entailed. It seems indisputable that logical relations are deﬁned over propositions. Appeal to some sui generis notion of non-logical epistemic inconsistency would seem both uneconomical and mysterious. So Williamson offers three or four very simple arguments for the thesis that evidence consists of propositions. He provides a single kind of entity to play all three roles, and so if the reader ﬁnds one argument weaker than the others, that weaker argument may draw support from the strength of the other two in virtue of this unifying power. In the next section, I will examine Conee and Feldman’s recent rebuttal.

14.3 Conee and Feldman’s reply In their recent ‘Evidence’ (2008) Conee and Feldman defend the experiential account of evidence by appeal to three examples. They are each examples of where we naturally talk in such a way that we refer to experiences as evidence (2008: 87). My feeling of warmth is evidence that it is warm in here, ‘a palpable sense of frustration’ is evidence that I am frustrated, my visual experience of how the car looks is evidence that it is red. That people can and do talk this way with propriety is not in dispute. Williamson admits ‘It is natural to say that my evidence that I am getting a cold includes various sensations’ (2000: 194). The question is how we are to understand this talk in the context of epistemological theorizing, explicating ordinary language for the purpose of an adequate theory of justiﬁcation. One way to do so is to take the language at face value. This is perhaps the simplest option, but it leaves us with the serious theoretical problems of how logical, explanatory, or probabilistic relations hold between items of evidence and the propositions they support. Furthermore, Conee and Feldman say that ‘beliefs are . . . derivatively evidence’ (2008: 87), so they are stuck with explaining how one relation can bear on conclusions from two radically different types of relata. They give no indication of whether they have a recursive deﬁnition in mind, an appeal to multiple realizability, or some other notion. Furthermore, they go on in a later section to characterize evidence as ‘the information or data the person has to go on in forming beliefs’ (2008: 88). It is hard to see how this is going to ﬁt in with the experiential account of evidence, since, though experiences might convey or contain information, information consists in propositions. They almost seem to blur this distinction in the following example where they avow that they ‘need not deny that

I N D E F E N S E O F P RO P O S I T I O N A L I S M A B O U T E V I D E N C E

229

all evidence is propositional. A visual experience as of something blue against a white background might consist in an awareness of propositions to the effect that certain visual qualities are arranged in a certain conﬁguration’ (2008: 101). The problem is that it is not enough that evidence be propositional. The state of awareness mentioned might have a proposition as a content, but the state of awareness is not itself a proposition, and therefore cannot stand in the relations necessary to play the key functional roles of evidence. The content itself certainly can, and that is the thesis I’m defending. They go on to say ‘It is not just other propositions that you believe that contribute to your justiﬁcation. The experience itself contributes . . . conscious awareness is how we gain whatever evidence we have’ (2008: 87). But of course this is consistent with the denial of the thesis that experience is evidence. Experience can contribute to justiﬁcation precisely by being that by which we gain our evidence. It is our tether to the propositions which are our evidence. This was part of Williamson’s critique of the experiential account of evidence. He said, ‘Experiences provide evidence; they do not consist of propositions. . . . But to provide something is not to consist of it.’ The question, he says, is ‘whether experiences provide evidence just by conferring the status of evidence on propositions’ (2000: 197). Conee and Feldman come close to adopting this view, but in an odd way. They write, ‘Propositions to the effect that the evidence [experience] exists or the like, can serve to state the reasons the evidence gives that adjudicate among hypotheses’ (2008: 102). But this unnaturally divides evidence from reasons. The simplest view entails that a reason to believe—in an epistemic sense, an objective support relation—should just be an item of evidence. The sum of our reasons would just be equated with our total evidence. (Conee (in Conee and Feldman 2004: 50) also introduces this distinction, but it’s not at all clear how it resolves the current problem or which premises of the above arguments it leads him to deny. Further, they often write in ways that imply that ‘reasons’ and ‘evidence’ are simply terminological variants. See, for example, pp. 19 and 296 and often in between. Indeed, the equation seems implicated on the very ﬁrst page of the volume.) The part of Conee and Feldman’s reply I have the hardest time understanding is where they simply seem to capitulate to Williamson, then claim a victory. Recall above the caveat about event-talk, that it is merely a verbal shortcut: a request to explain an event is a request to explain why it happened, that it happened being the item in need of explanation. They consider this statement then say ‘precisely the same’ can be said about experiences because they are perhaps a kind of event. But isn’t this just to accept Williamson’s account? They make a similar move with probabilistic argument (they don’t address the deduction argument, but the pattern has been established), saying that the phrase ‘the probability on the evidence’ could be ‘harmlessly understood as the probability on the proposition that the evidence occurred’ (2008: 102). The problem is that this simply runs up against clear, key conceptual roles of evidence: making things probable and having a probability. As they have it here, their view denies the simple

230

EVIDENCE

platitudes upon which Williamson forms his arguments. And it is around such platitudes that analytic philosophers of most stripes build their theories, which seems the safest route. Are we really to understand Conee and Feldman as denying that evidence has and confers probabilities? As they have stated it here, that seems to be a consequence, for on the understanding they suggest it is propositions about evidence that confer probability, not evidence. That would be a very bad result. And it would look just as bad applied to the case of deductive relations.

14.4 Methodological argument In this section, I’ll try to set forth a regimented version of my argument for a propositional theory of evidence. 1. If the metaphysics implicit in ordinary language is theoretically inelegant (uneconomical) and that inelegance is easily eliminable by taking part of ordinary language to be loose talk while preserving key functional roles, then we ought to interpret part of ordinary language as loose talk in a way that preserves key functional roles. 2. The metaphysics implicit in ordinary language is theoretically inelegant (uneconomical). 3. That inelegance is easily eliminable by taking part of ordinary language to be loose talk while preserving key functional roles. 4. Therefore, we ought to interpret part of ordinary language as loose talk in a way that preserves key functional roles. 1,2,3 &I MP. I have not mentioned ‘evidence’ or evidence or functional roles of evidence in this argument, but the application is obvious: Williamson’s interpretation of the natural speech which calls experiences evidence takes sentences about experience to be sentences about propositions delivered by experiences. I’ve said this is simply a case of synecdoche, but it is also more expedient to talk about the experiences than the propositions experiences bring into our stock of evidence. So to speak of experiences as evidence is a sort of loose talk: innocent enough unless we take it to reveal bedrock truth. Conee (2005) has applied this kind of argument to contextualism. He notes that though a straightforward acceptance of people’s varying standards in ‘knowledge’ ascriptions would support contextualism, it is simpler to take the actual truth conditions of such ascriptions as invariant and realize that what varies are speakers’ standards for making the ascriptions (simpler because otherwise all manner of expressions turn out to be, surprisingly, contextual, and all sorts of talk isn’t what it appears to be). As Conee says, ‘In ordinary contexts, when nothing much turns on it, people will claim knowledge, and attribute knowledge to themselves and others, in belief and in speech, on a basis that is signiﬁcantly weaker than what is actually required. This is an efﬁcient way to communicate assurance about the proposition and to facilitate taking it for granted’ (2005: 52). Conee proposes a simple ‘really’ test. Suppose someone says they know

I N D E F E N S E O F P RO P O S I T I O N A L I S M A B O U T E V I D E N C E

231

their mom is happy in Heaven. You ask them ‘Do you really know that?’ ‘Well, no, I suppose I don’t, but I believe it and sure do hope it’ is a likely reply. This is just my experience with the experience as evidence thesis: I read Williamson as saying ‘Do you really think evidence consists in experiences rather than the propositions those experiences deliver?’ ‘No, I suppose not,’ I said to myself. My own amateur research has turned up this result in the others I’ve asked as well. I urge Conee and Feldman to put themselves to the ‘really’ test on this issue again. With regard to the details of the argument, premise 1 strikes me as the basis for most of the most sound analytic philosophy, something akin to a truism, so I won’t defend it. Premise 2 is entailed by the fact that ordinary language treats items in widely disparate ontological categories as being evidence. It is worsened by the fact that key conceptual roles of evidence simply cannot be played by experiences, something Conee and Feldman don’t seem to dispute. Premise 3 is made true by the fact that Williamson’s retranslation schemes work, as admitted by Conee and Feldman, and are built around the key functional roles in explanation, probability, and deduction. I conclude that this argument puts considerable pressure on Conee and Feldman to either concede that evidence consists of propositions or to address the arguments above—of Williamson’s as I’ve set them out and of my own generalized argument.

14.5 Results for evidentialism The truth of propositionalism has signiﬁcant results for evidentialism. Feldman and Conee ﬁrst introduced evidentialism as the following thesis EJ. EJ. Doxastic attitude D toward proposition p is epistemically justiﬁed for S at t if and only if having D toward p ﬁts the evidence S has at t. (2004 [1985]: 83) Surprised to ﬁnd top epistemologists writing about epistemic justiﬁcation in ways that encumbered it with considerations pertaining to the (dis)abilites and obligations of agents, they endeavored to show how EJ does not fall prey to any pitfalls related to agency. Key to their defense was the fact that it was an account of synchronic justiﬁcation, which rendered what people might have had as evidence or should have had as evidence irrelevant. (Feldman goes on to suggest an extension to diachronic rationality in ‘The Ethics of Belief’ (2004 [2000]). Their defense seemed cogent if not universally agreed upon, yet in the afterward added in 2004, they say ‘Our bedrock epistemic view is a supervenience thesis’ (2004: 101). They state the supervenience thesis as follows. ES. The epistemic justiﬁcation of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at the time. (2004: 101) In light of the cogency of the defense of EJ the move to ES can seem surprising, especially due to the common complaint that supervenience theses are non-explanatory. They

232

EVIDENCE

note—(2004: 102)—that there are a constellation of questions about ﬁt: How does epistemic ﬁt differ from moral ﬁt or prudential ﬁt? Is there a summary notion of ﬁt/ justiﬁcation that encompasses and organizes all particular notions of ﬁt? Do some forms of ﬁt over-rule other forms of ﬁt? They also note that there are also other hard questions glossed over by the supervenience thesis: Is believing justiﬁed when the evidence just barely favors a proposition? What about propositions we don’t understand? They say it is precisely the fact that such questions are circumvented by the supervenience thesis that recommends it. But the questions above are important and substantive questions, and it is better to hit them head on than to sweep them under the supervenience carpet. What we have here is a serious research project which we may hope will inspire a generation of scholars to explicate the notion of epistemic ﬁt and epistemic support. One way to motivate that research project is to note that if my defense of the propositional theory of evidence above—that is, the defense that evidence consists in propositions—is successful, then ES is false, though EJ still might be true. To see that ES is false if evidence consists of propositions, consider two individuals, Ted and Todd, who have similar but not type-identical experiences as of a green taxicab. They will later be asked if all taxis in town are yellow. Ted’s experience is more clear and distinct than Todd’s because Ted’s takes place at sundown while Todd’s takes place a half hour later, as he is preparing to star gaze. They both seem to see a green taxi, but the vividness of Ted’s experience is greater than that of Todd’s, so though each brings into their stock of evidence the proposition There was a green taxi on the road this evening, different attitudes ﬁt the proposition All the taxis in town are yellow, for Ted should have a stronger degree of belief than Todd. It would be a great liability to evidentialism if retaining ES was incompatible with varying strengths of belief. They note that ‘EJ is compatible with the existence of varying strengths of belief and disbelief. If there is such variation, then the greater the preponderance of evidence, the stronger the doxastic attitude that ﬁts the evidence’ (2004: 83, n1). Thus there are two individuals with the same evidence such that different doxastic attitudes ﬁt the same proposition being considered. Thus, though ES runs afoul of the propositional theory of evidence, EJ may ﬁt just ﬁne (because the experiential differences could be incorporated into the ‘ﬁts’ relation). This puts pressure on evidentialists to make clear the notion of epistemic ﬁt and address the constellation of questions above.

References Conee, Earl. 2005. ‘Contextualism Contested,’ in Matthias Steup and Ernest Sosa (eds), Contemporary Debates in Epistemology. Oxford: Blackwell. Conee, Earl and Richard Feldman. 2004. Evidentialism. Oxford: Oxford University Press. —— and ——. 2008. ‘Evidence,’ in Quentin Smith (ed.), Epistemology: New Essays. Oxford: Oxford University Press. Skyrms, Brian. 2000. Choice and Chance. Belmont, CA: Wadsworth. Williamson, Timothy. 2000. Knowledge and its Limits. Oxford: Oxford University Press.

PART VII

New Syntheses Duncan Pritchard argues that Conee and Feldman have made a mistake by tying their evidentialist views to a classical internalism which accepts the following theses: (MENT) If S and S* do not differ in their mental states then they will not differ in the degree of epistemic justiﬁcation that they have for their beliefs. (ACCESS) If S and S* do not differ in the facts that they are able to know by reﬂection alone then they will not differ in the degree of epistemic justiﬁcation that they have for their beliefs. (DISC) If the experiences had by S and S* are indiscriminable then S and S* will not differ in the degree of epistemic justiﬁcation that they have for their beliefs. The polar opposite of this classical internalist position is classical externalism, which denies MENT, ACCESS, and DISC. But, Pritchard observes, there are positions intermediate between classical internalism and classical externalism. In particular, there is the epistemic disjunctivist’s position which accepts MENT and ACCESS but denies DISC. Pritchard argues that this disjunctivist position is favorable to the classical internalist position because it provides a better response to skepticism, while overcoming all of the most threatening objections urged against it. Alvin Goldman, historically, perhaps, the biggest critic of evidentialism offers to lend the evidentialist a helping hand. He argues that it needs a helping hand by offering three supposed counter-examples to evidentialism—speciﬁcally to Conee and Feldman’s version. First, it cannot account for justiﬁcation by introspective evidence because there needs to be a distinct mental state of introspecting x to be explained by the mental experience of x, and it appears that there is none. Second, no evidencing mental state accompanies preservative memory. Third, arithmetic inference does not explain the states from which the inference was made. He argues that the value of superior explanatoriness is that it reliably indicates the truth of the explanation, suggesting that best-explanationism derives from reliabilist principles. Goldman argues that deﬁning evidence as a reliable guide to truth is more promising than the internalist move—which he thinks circular—of deﬁning evidence in terms of

234

NEW SYNTHESES

justiﬁcation. He argues that the reliable guide conception of evidence is incompatible with Conee and Feldman’s view insofar as evidential relationships are supposed to be necessary, and even in evil-demon worlds perceptual experiences count as evidence. Also, the property of truth-indicatorship is not the same in all possible worlds. Goldman contends that truth-indicatorship can be ‘rigidiﬁed’ and deﬁned as relationships that obtain in the actual world. Furthermore, reliabilism can also help evidentialism, says Goldman, in anchoring the concept of evidential ﬁt. It is Conee and Feldman’s failure to offer adequate accounts of both ‘evidence’ and ‘ﬁttingness’ which is a central criticism. And Goldman renews his charge concerning problems associated with the internalist requirement that all justiﬁcation-relevant states are ones that hold at the time of (actual or potential) belief—rather than in the history of the belief. But Goldman’s offer is not wholly altruistic, as he argues that reliabilism would beneﬁt from incorporating ﬁttingness into its account of inferential justiﬁcation. He argues that ﬁttingness is attractive because in instances where one proposition supports another, evidentialism shows that the appropriate degree of conﬁdence in the target proposition is the degree to which the other proposition supports it. He argues that ﬁttingness with non-doxastic evidence could have a role in reliabilism’s treatment of non-inferential justiﬁcation. This would address reliabilism’s lack of appreciation for the role experience plays in justiﬁcation. The ﬁrst component of experiential justiﬁedness is ﬁt with experiential evidence, and the second is a reliable process of beliefformation.

15 Evidentialism, Internalism, Disjunctivism Duncan Pritchard

15.1 Many in epistemology will want to grant the claim that justiﬁcation is a function of one’s supporting evidence, in the sense that it is only when one’s belief is adequately supported by evidence that one is justiﬁed. In general terms, anyone who endorses such a position is an evidentialist. In contrast, those who wish to hold that some forms of justiﬁcation are non-evidential—for example, Crispin Wright (2004)—would be non-evidentialists. In their important and inﬂuential book, Evidentialism, Earl Conee and Richard Feldman (2004) offer a powerful defence of a distinctive epistemically internalist variety of evidentialism. What I propose to do here is not so much confront this position directly, but rather describe an alternative view—a form of epistemological disjunctivism—a position which could also be cast as an evidentialist variety of epistemic internalism, but which is strikingly different to the view that Conee and Feldman defend and which also, crucially, has a number of theoretical advantages over Conee and Feldman’s position. Along the way, I will consider some objections to disjunctivism more generally that Conee (2007; cf., Conee 2008) has put forward, some of which are relevant to the view that I describe here.

15.2 For Conee and Feldman, what makes their view properly epistemically internalist is that it is committed to a thesis that they refer to as mentalism. According to the mentalist, two subjects with exactly the same mental states are necessarily alike epistemically. Or, as Conee and Feldman (2004: 56) put it, ‘if any two possible individuals are exactly alike mentally, then they are alike justiﬁcationally’. Given their wider evidentialism, such that justiﬁcation is a function solely of one’s evidence—where one’s evidence concerns one’s mental states, such as one’s beliefs and

236

NEW SYNTHESES

experiences—this view entails that any two possible individuals who are exactly alike mentally are also alike with respect to the evidence that they possess. There is a very natural motivation for both claims. For consider an agent who is experiencing the world in roughly the way that he thinks he is experiencing the world, and a counterpart agent who is radically deceived, but in such a way that he is unable to tell the difference between these deceived experiences and normal experiences. On one standard view of what constitutes an agent’s mental life, given that the two kinds of experiences are ex hypothesi indiscriminable it follows that the two agents in question have the same mental states (or, at least, that they are exactly ‘alike’ mentally, if this is thought to be a different notion). If this is right, then it follows on the mentalist view that such subjects are alike epistemically as well—in particular, that they are alike with respect to their evidence and thus with respect to whether they are justiﬁed in what they believe—and that seems to accord with intuition on this score. Interestingly, most standard forms of epistemic internalism do not explicitly commit themselves to this mentalist thesis, but rather to the (potentially) distinct view of accessibilism. According to this thesis (construed as a thesis speciﬁcally about justiﬁcation, which is our concern here), justiﬁcation is a function solely of what is reﬂectively accessible to the subject. Hence, an accessibilist evidentialist would hold that one is justiﬁed only if one has adequate evidence, where what constitutes one’s evidence is understood in turn along accessibilist lines. As it happens, Conee and Feldman’s mentalist evidentialism is almost certainly extensionally equivalent to accessibilism evidentialism. On their view, the kind of mental states that are relevant to determining one’s evidence—essentially, one’s beliefs and experiences—are also the kind of states that are reﬂectively accessible if any states are. Thus, their mentalist evidentialism is not meant to be inconsistent with accessibilist evidentialism. The point they are making, I take it, is rather that it is the mentalist thesis which is constitutive of epistemic internalism, even when it is extensionally equivalent to the accessibilist thesis.

15.3 I ﬁnd the idea that mentalism is more central to the epistemic internalist position than accessibilism very questionable. After all, one could imagine a type of content externalism which, while endorsing mentalism, nevertheless regarded some of the mental states in question as being inaccessible to the agent. Would such a view count as epistemically internalist? I think that this is unlikely, and this suggests that it is the accessibilist thesis which is more central to epistemic internalism. We will come back to this point in a moment. First, though, we need to put another thesis on the table that is central to Conee and Feldman’s view. Following Keith Lehrer and Steward Cohen (1983), we will call this thesis the ‘New Evil Genius’ (NEG) thesis:

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

237

NEG. The extent to which S is justiﬁed at t in believing that p is just the same as the extent to which S’s envatted duplicate is justiﬁed at t in believing that p. Epistemic externalists famously deny (NEG). Consider, for example, process reliabilism, which is a paradigm example of an epistemic externalist view.1 According to this view, the justiﬁcation for one’s belief depends on whether that belief was formed in a reliable fashion. On this proposal, then, one would expect there to be an epistemic difference between the beliefs held by S and by her envatted counterpart, since one would expect there to be a far greater degree of reliability exhibited in the former case than in the latter case. Thus, (NEG) is incompatible with process reliabilism. Moreover, given that whether or not one’s beliefs are reliable will (usually) not be constituted by one’s mental states, and given that it is (usually) not reﬂectively accessible to one that one’s beliefs are reliably formed, it follows that process reliabilism is inconsistent with accessibilism and mentalism too. In contrast, epistemic internalists typically take (NEG) as a datum that needs to be explained, and which can only satisfactorily be explained by an internalist view. Indeed, once one endorses mentalism and accessibilism, as Conee and Feldman do, then a very straightforward explanation of why (NEG) is true manifests itself. After all, one might plausibly hold that one shares the same mental states as one’s envatted counterpart.2 Moreover, one might also plausibly hold that it is only one’s mental states that are reﬂectively accessible. Accordingly, if one’s justiﬁcation is constituted by one’s mental states—the only kind of thing that is reﬂectively accessible to one—then it is no surprise that there is no epistemic difference between one’s beliefs and the beliefs of one’s recently envatted counterpart. A paradigm form of epistemic externalism is thus committed to the denial of accessibilism, mentalism, and (NEG), while a paradigm form of epistemic internalism is committed to the adoption of accessibilism, mentalism, and (NEG). If there were no viable epistemic positions lying between these two extremes, then we could take these paradigm types of epistemic externalism and epistemic internalism as roughly representative of the epistemic externalism/internalism distinction. As we will now see, however, there are a number of viable positions lying between these two poles, and this complicates the picture somewhat.

15.4 It will be useful to specify the three claims in play here a little more formally. Mentalism can be understood as the following thesis:

1

See, for example, Goldman (1986). Whether one ﬁnds this sort of line generally plausible will depend, in part, on whether one is attracted to content externalism. For while many would grant that there could be no difference between one’s mental states and the mental states of one’s recently envatted physical duplicate, most content externalists would argue that, at least eventually, a difference would emerge. For more discussion of content externalism, see Nuccetelli (2003). 2

238

NEW SYNTHESES

MENT. If S and S* do not differ in their mental states then they will not differ in the degree of epistemic justiﬁcation that they have for their beliefs. Accessibilism can be understood as the following thesis: ACCESS. If S and S* do not differ in the facts that they are able to know by reﬂection alone then they will not differ in the degree of epistemic justiﬁcation that they have for their beliefs. Finally, I take it that the thesis that underlies (NEG) is the following: DISC. If the experiences had by S and S* are indiscriminable then S and S* will not differ in the degree of epistemic justiﬁcation that they have for their beliefs.3 After all, what motivates (NEG) is the idea that the subject is unable to discriminate between her non-envatted experiences and the envatted experiences of her counterpart. Thus, what goes for the envatted counterpart applies with equal force to any counterpart whose experiences are indiscriminable from the experiences that our subject is presently undergoing. We will call the view which denies all three of these theses classical epistemic externalism, and we will call the view which endorses all three of these theses classical epistemic internalism. The question that now faces us is whether there are interesting versions of epistemic externalism and internalism which are not of the classical variety. Since there are three theses in play here, there are in principle eight views available in this regard: 1. 2. 3. 4. 5. 6. 7. 8.

(ACCESS) + (MENT) + (DISC) (ACCESS) + (MENT) + ¬ (DISC) (ACCESS) + ¬ (MENT) + (DISC) (ACCESS) + ¬ (MENT) + ¬ (DISC) ¬ (ACCESS) + (MENT) + (DISC) ¬ (ACCESS) + (MENT) + ¬ (DISC) ¬ (ACCESS) + ¬ (MENT) + (DISC) ¬ (ACCESS) + ¬ (MENT) + ¬ (DISC)

Views (1) and (8) are classical epistemic internalism and classical epistemic externalism, respectively. What about views (2)–(7)? Are any of them plausible? Some of them clearly aren’t very compelling. Views (3) and (4), for instance, both of which endorse (ACCESS) while denying (MENT), do not seem to be very attractive. After all, I take it that most will want to argue that it is only one’s mental states that are in the market for being reﬂectively accessible. Accordingly, the prospect of endorsing (ACCESS) while denying (MENT) will not be appealing, since it would commit one 3 More precisely, S is unable to discriminate between her experiences and the experiences had by S* (and, for that matter, vice versa).

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

239

to allowing that one’s justiﬁcation is constituted by factors that one can know by reﬂection alone while at the same time treating the range of the relevant factors to extend beyond what is reﬂectively accessible to one. View (7) also looks like a pretty odd mix of theses. If one denies both (ACCESS) and (MENT), then what would be the motivation for endorsing (DISC)? That is, why would one hold that the properties that one shares with one’s envatted counterpart determine one’s epistemic standing if one does not hold that one’s epistemic standing is either constituted by what one has reﬂective access to or one’s mental states? Finally, while view (5) is not as implausible as views (2), (3), and (7), it is far from being an attractive position to hold either. After all, whatever reason one has for denying (ACCESS) while accepting (MENT) will almost certainly count against (DISC). In particular, the natural explanation of why (ACCESS) might be false even though (MENT) is true would be because one endorses a form of content externalism which ensures that one’s mental states—which, on this view, constitute one’s justiﬁcation—are not generally reﬂectively accessible. If one endorses a content externalism of this variety, however, then one will inevitably be suspicious of (DISC) as well. In particular, one will naturally suppose that the agent in question does have at least some different mental states when compared with her envatted counterpart, even though the difference in question is indiscriminable. Hence, given (MENT), it follows that (DISC) has to go.4 Let’s simplify our discussion then by focusing on the remaining four views, two of which we have already identiﬁed as classical versions of epistemic externalism and internalism: 1. 2. 6. 8.

(ACCESS) + (MENT) + (DISC) Classical Epistemic Internalism (ACCESS) + (MENT) + ¬ (DISC) ¬ (ACCESS) + (MENT) + ¬ (DISC) ¬ (ACCESS) + ¬ (MENT) + ¬ (DISC) Classical Epistemic Externalism

Interestingly, both of the remaining unidentiﬁed views constitute plausible epistemological proposals. Moreover, as we will see, one of these views—view (2)—is also very close in spirit to the account of justiﬁcation offered by Conee and Feldman.

15.5 Let’s start with view (6). I noted in the last section that there was something odd about any position which accepted (ACCESS) while denying (MENT). Given that, as we will now see, there is nothing odd about accepting (MENT) while denying (ACCESS), and given

4 At least, this follows if the relevant difference in mental states concerns states that have a bearing on one’s epistemic standing. As far as the envatted case goes, I think it is a fair assumption that such a view would regard the difference in mental states between the agent and her envatted counterpart to be extensive enough to ensure that this is the case, though this will depend, in part, on how recent the envattment is (see note 2).

240

NEW SYNTHESES

that such a proposal is in addition naturally classiﬁed as a version of epistemic externalism, there seems every reason to hold, contra Conee and Feldman, that it is (ACCESS) which is the central thesis of epistemic internalism, rather than (MENT). Indeed, we can get a ﬂavor of the kind of argument in play here by considering the question of whether view (5), which we have just noted to be unattractive, is naturally taken to be an epistemically internalist or externalist position. I suggest it is hard to see why we would consider such a position an internalist view, given that it denies (ACCESS) and so makes one’s epistemic standing a function of, in part at least, factors that are opaque to one. The same goes for view (6). Like view (5), one might be tempted by such a position because one endorses a version of content externalism. As with view (5), one might then plausibly argue that (ACCESS) is false since many of one’s mental states will not be reﬂectively accessible. Nevertheless, one might still maintain that one’s justiﬁcation is constituted by one’s mental states. Furthermore, since this view denies (DISC), it doesn’t face the problem that faced view (5) of how to consistently deny (ACCESS) while endorsing both (MENT) and (DISC). On this proposal, just as it follows from the content externalism in play that many of one’s mental states are not reﬂectively accessible, so it also follows that one does not share the same mental states with one’s envatted counterpart.5 Indeed, one could regard view (6) as being motivated by some of the theses argued for in recent work by Timothy Williamson. Williamson famously holds that knowledge is a mental state and that one’s knowledge is to be identiﬁed with one’s evidence.6 Suppose one now held that one’s justiﬁcation was constituted by one’s evidence. This would mean that it is constituted by what one knew, where that in turn meant that it was constituted by one’s mental states. Such a view is therefore naturally allied to (MENT). Moreover, it is not hard to see why someone who held such a position would want to deny (ACCESS), since on this conception of one’s mental states they are not obviously generally reﬂectively accessible.7 Finally, any view of this sort would be required to deny (DISC). After all, many of the beliefs held by one’s envatted counterpart are false, and so one would naturally expect that one’s counterpart therefore has less knowledge that one has, and therefore has less evidence too. It follows, therefore, that one’s counterpart’s beliefs are justiﬁed to a lesser degree.8

5 In particular—see note 4—it follows provided the difference in mental states between the agent and her envatted counterpart is extensive enough to ensure that it makes the required epistemic difference. 6 See Williamson (1995; cf. Williamson 2000a: ch. 1) and Williamson (1997; cf. Williamson 2000a: ch. 9), respectively. 7 Williamson famously denies that one’s mental states are ‘luminous’, of course, though a mental state could still be reﬂectively accessible and yet not be luminous, so this does not decide the matter. See Williamson (1996; cf. Williamson 2000b: ch. 4). 8 Williamson is actually quite explicit that he denies (DISC). See, for example, Williamson (2000b; cf. Williamson 2000a: ch. 8). Just to be clear, however, I am not suggesting that the account of epistemic justiﬁcation just described is one that Williamson would endorse, only that it is the sort of position that one who advocates some of the key theses that Williamson holds would ﬁnd attractive. For further discussion of Williamson’s epistemology, see Greenough and Pritchard (2009).

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

241

Clearly, however, despite the commitment to (MENT) here, no one would describe this Williamsonian conception of epistemic justiﬁcation as an epistemically internalist position. Instead, this view would be regarded as a form of epistemic externalism, albeit one that diverged through its endorsement of (MENT) from classical epistemic externalism. Again, then, we ﬁnd that it is a commitment to (ACCESS) that is more central to epistemic internalism, contra Conee and Feldman.

15.6 In any case, it is not this non-classical form of epistemic externalism that is our main interest here, but rather the non-classical form of epistemic internalism represented by view (2). That is, given that we have now granted that it is a commitment to (ACCESS) that determines whether one’s view is a version of epistemic internalism or epistemic externalism, we can categorize our four positions as follows: 1. 2. 6. 8.

(ACCESS) + (MENT) + (DISC) (ACCESS) + (MENT) + ¬ (DISC) ¬ (ACCESS) + (MENT) + ¬ (DISC) ¬ (ACCESS) + ¬ (MENT) + ¬ (DISC)

Classical Epistemic Internalism Non-Classical Epistemic Internalism Non-Classical Epistemic Externalism Classical Epistemic Externalism

With this taxonomy of views in mind, let us turn our attentions to the case that can be made for non-classical epistemic internalism, a view which departs from the position advocated by Conee and Feldman only in its rejection of (DISC). This position does exist in the literature, and it is usually motivated via appeal to disjunctivism. Disjunctivism is standardly understood as a (broadly speaking) metaphysical proposal about the content of perceptual experience, though the position potentially has application beyond perception. In essence, the guiding claim is that cases of veridical perception share no content in common with corresponding cases (indiscriminable to the subject) where the perception is ‘subverted’, such as cases of hallucination.9 A characteristic feature of such a position is therefore to deny that two experiences which are indiscriminable to the subject must thereby share content. Such a position seems to directly motivate a speciﬁcally epistemological disjunctivism, which is the view that most concerns us here.10 After all, it follows on this view that the contents of one’s mental states will be different in deceived and non-deceived cases. 9 Contemporary disjunctivism of this broadly metaphysical stripe largely owes its source to the work of Hinton (e.g. 1973). For some of the key more recent discussions, see Snowdon (1980–1, 1990–1), McDowell (1986), and Martin (2003, 2004). For an excellent recent anthology on disjunctivism, see Haddock and Macpherson (2008a). Note that I am here glossing over the issue of what to say about deceived perceptual experiences which are nonetheless veridical. I will return to this complication later on. 10 For more on the contrast between metaphysical and epistemological disjunctivism, see Byrne and Logue (2008) and Haddock and Macpherson (2008b). As far as I am aware, Byrne and Logue (2008) and Pritchard (2008b)—which appear in the same edited collection (Haddock and Macpherson (2008a))—were the ﬁrst to use the phrase ‘epistemological disjunctivism’ to mark out a speciﬁc epistemological variant of the disjunctivist proposal which was not necessarily committed to metaphysical disjunctivism.

242

NEW SYNTHESES

Thus, if one endorses (MENT) from a disjunctivist perspective, then one will treat the subject and her deceived counterpart as having different mental states and hence as having beliefs with different epistemic standings. Clearly, then, (MENT) on this view will entail the denial of (DISC). But what about (ACCESS)? This is where the view gets interesting, since it is characteristic of epistemological disjunctivism to retain this thesis and argue that the content of one’s experiences is in the relevant sense reﬂectively accessible to the subject. Of course, the subject cannot discriminate between deceived and non-deceived cases, but the thought is that this fact does not sufﬁce to show that in the non-deceived case the subject lacks the relevant reﬂective access. So construed, epistemological disjunctivism is motivated by metaphysical disjunctivism. Interestingly, one can also regard the view as being motivated by independent considerations. Indeed, the foremost exponent in the contemporary literature of the view that we are calling epistemological disjunctivism, John McDowell (see especially 1995, but also McDowell 1982, 1994, 2002a, 2008), these days almost exclusively motivates the position by appeal to independent considerations. The distinctive thesis that McDowell offers is that the reﬂectively accessible rational support that one has for one’s beliefs can be both factive (i.e. entail that which it is a reason for) and yet be empirical (i.e. it can be a reason for belief in an empirical proposition). Before we get into some of the detail of McDowell’s position, I want to make a few simplifying assumptions that will keep our discussion manageable. First, we will focus on perceptual justiﬁcation. McDowell’s epistemological disjunctivism is meant to be a more general thesis (see, e.g., McDowell 1994), but it would take us too far aﬁeld to explain how this view manifests itself across different types of justiﬁcation. If one is skeptical about the application of McDowell’s view to other types of justiﬁcation, then one should regard the debate here as being speciﬁcally about Conee and Feldman-style classical epistemic internalism about perceptual justiﬁcation versus McDowellian non-classical epistemic internalism about perceptual justiﬁcation. This debate is important enough for our purposes. Second, I want to take it that we can freely move between talk of reasons and talk of evidence without too much concern. If we allow this assumption, then it enables us to make a direct comparison of epistemological disjunctivism and Conee and Feldmanstyle evidentialism. Still, one might think this move contentious.11 If so, then one should simply reinterpret my defence of epistemological disjunctivism, and thus my critique of Conee and Feldman, in non-evidentialist terms. That is, the attack is merely on the classical form of epistemic internalism that Conee and Feldman endorse, rather than speciﬁcally on the evidentialist variety of classical epistemic internalism that they espouse. Such a critique is surely substantive enough for our purposes. Finally, third, in what follows the version of epistemological disjunctivism that I describe is not meant to faithfully represent the McDowellian position, even though it is meant to embody the key features of that view. This is partly because I do not want 11 As it happens, Conee (2007: 32ff.; cf. Conee 2008: 20ff.) ﬁnds such an equivalence contentious, and I’m inclined to agree with him.

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

243

to get bogged down in the murky world of McDowellian exegesis, partly because there are some details of McDowell’s view that are not particularly salient to us here, and partly because there are some issues where it may be that the view presented diverges in detail from McDowell’s avowed position. Thus, while it is reasonable to think of the view presented as ‘McDowellian’, this should not be thought to imply that it is a view that McDowell himself would sign up to in its entirety.

15.7 With these points in mind, let us now look at McDowell’s position in a little more detail. First, let us distinguish between pairs of ‘good’ and ‘bad’ cases. A ‘good’ case, as we are using the term, is a case in which the agent’s veridical perception takes place in epistemically advantageous conditions, and consequently results in knowledge (and, thereby, justiﬁed belief). In contrast, the corresponding ‘bad’ case is a scenario which (i) is indiscriminable to the subject from the good case; (ii) is such that the subject’s perception is non-veridical; and (iii) takes place in epistemically disadvantageous conditions. Since the subject is unaware of being in the bad case, she forms the same belief that she forms in the good case. Clearly, there is a difference in terms of what the agent knows in the two cases, since knowledge of the target proposition is acquired in the good case but impossible to acquire in the bad case, given how it is described (because the agent’s belief is false). Crucially, the epistemological disjunctivist will also claim that there is a difference in the extent of perceptual justiﬁcation that the agents’ possess for their beliefs in the target propositions too. Here is an example to illustrate this distinction between good and bad cases. First the good case. Our agent sees a barn in good cognitive conditions (e.g. there are no undefeated misleading defeaters present, she’s not in barn fac¸ade county, and so on). Consequently, she thereby comes to know, and so justiﬁably believe, that there is a barn before her. In contrast, the corresponding bad case could be where the same agent merely seems to see a barn in bad cognitive conditions (e.g. she is, unbeknownst to her, in barn fac¸ade county, and looking at a fake barn). Since the subject cannot discriminate between the good and bad cases, she falsely believes that there is a barn before her in the bad case and so lacks knowledge of this proposition. Furthermore (and we will explain what motivates this claim in a moment), according to the epistemological disjunctivist, the agent in the bad case will have less justiﬁcation for her belief in the target proposition when compared with her counterpart (if, indeed, her belief in this proposition is justiﬁed at all). Note that there are cases which are neither good nor bad (i.e. the distinction is not meant to be exhaustive of the options available). For example, one’s perception could be both veridical and formed in epistemically disadvantageous conditions, as when one sees a genuine barn while in barn fac¸ade county.12 In order to offer a complete 12 Things get more complicated still once we bring in the distinction between whether the cognitive conditions are objectively or subjectively epistemically advantageous. For example, one could, in fact, be in

244

NEW SYNTHESES

description of what epistemological disjunctivism involves, we would have to deal with these cases as well, but we can side-step this task here. What is important for our purposes is just to identify a range of cases in which the epistemological disjunctivist tells a different epistemological story to the classical evidentialist epistemic internalist. If this alternative account can be motivated even in this restricted range of cases, then that would sufﬁce to motivate taking the view seriously as an alternative to classical evidentialist epistemic internalism. The key claim made by the epistemological disjunctivist is that in the good case the rational support available for the agent’s belief is that she sees that there is a barn before her, where seeing that p entails p. Moreover, the epistemological disjunctivist also claims that the rational basis for this belief is reﬂectively accessible. Accordingly, in the good case the agent has a reﬂectively accessible factive reason. In contrast, in the bad case the rational basis for the agent’s belief cannot be that she sees that p, since, of course, this is not the case. At most, it would appear that the rational basis for her belief is that it seems to her as if there is a barn before her, a rational basis which is clearly not factive. It follows that even though the agent is unable to distinguish between the good and bad case, the rational support that the agent possesses in each case is radically different. We can illustrate this point further by considering a skeptical pairing of good and bad cases, such as a good case in which the agent sees that she has two hands and a corresponding bad case in which the agent has been envatted and merely seems to see that she has two hands. According to the epistemological disjunctivist, in the good case the agent’s reﬂectively accessible epistemic support for her belief that she has two hands will be the factive ground that she sees that she has two hands, while in the bad case the agent will lack factive epistemic support of this kind for her belief. Given that the epistemological disjunctivist holds that the rational basis for one’s belief is reﬂectively accessible to one, it follows that she clearly endorses (ACCESS). Moreover, since she holds that the epistemic support that an agent’s belief enjoys is very different in good and bad cases, it follows that she also clearly rejects (DISC). Finally, there is nothing in the epistemological disjunctivist line that prevents her from endorsing (MENT). In particular, insofar as epistemological disjunctivism is motivated by appeal to metaphysical disjunctivism, then the grounds for endorsing (MENT) are straightforward indeed. On this view, the reason why the rational support is both reﬂectively accessible and so very different in good and bad cases is because the relevant mental states of the subjects in these cases—in particular, the content of their perceptual experiences—is so very different.

very good cognitive conditions and yet be in possession of an undefeated misleading defeater. In such a case, the cognitive conditions would be subjectively epistemically disadvantageous and yet objectively epistemically advantageous. With this in mind, I think the epistemological disjunctivist will need a six-fold distinction between cases of this sort in order to accommodate the various nuances. I explore this issue more fully in Pritchard (2008a).

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

245

But even if epistemological disjunctivism is not allied to metaphysical disjunctivism, then there could still be a basis for endorsing (MENT). For example, suppose one denied metaphysical disjunctivism and so held that the perceptual experiences of an agent in corresponding good and bad cases shared content. It needn’t follow that the content of the agent’s perceptual experience is the same in the two cases, and thus that her mental states are the same. Indeed, that would depend on the account of perceptual experience that one offered. If, for example, one opted for a causal theory of perceptual experience—of a sort most famously championed by H. P. Grice (1961), for example—then one would still regard the two experiences as having different contents because of the different causal settings of those experiences, even though they shared some content. What is important to the epistemological disjunctivist who wishes to retain (MENT) is thus not that they endorse metaphysical disjunctivism, but only that they resist the idea that the shared content of perceptual experiences in corresponding good and bad cases constitutes the common rational basis for both beliefs.13

15.8 McDowell tends to motivate epistemological disjunctivism in largely negative terms. That is, rather than offering arguments which speciﬁcally support this position, instead his general dialectical approach is to argue that such a view has many attractions when compared with its rivals and that, contra the prevailing wisdom, it is theoretically open to us.14 On the former point, one such attraction that McDowell emphasizes is the fact that such a view accords with common-sense. It is certainly true that outside of philosophical discussions it is natural to describe the epistemic support one has in favor of one’s belief in good cases in factive terms. It is only when impressed by philosophical arguments, such as skeptical arguments, or arguments from illusion, that one feels inclined to re-describe that epistemic support in non-factive terms. The import of this consideration is moot, of course, since perhaps we shouldn’t put all that much stock in the verdicts of common sense. Still, I take it that the thought in play here is that common sense occupies a ‘default’ position in our theoretical thinking, in the sense that we require good reasons to deny common sense, and that absent such reasons we should endorse common sense views. In addition, McDowell argues that once we endorse epistemological disjunctivism then some of the key problems of epistemology simply disappear. Consider, for 13 This is not the place to explore this possibility further. For additional discussion of this possibility, see Millar (2007, 2008, 2009). See also Haddock and Macpherson (2008b). 14 There is one exception in this regard, which is McDowell’s transcendental argument to the effect that it is only if it is sometimes the case that one’s perceptual experiences involve a ‘taking in’ of the relevant fact, such that one sees that p, that it could be intelligible that there exists perceptual experiences with representational content. This argument is subtle, as is the extent to which it presupposes metaphysical disjunctivism. In order to keep this discussion of a manageable size, I have set this motivation for the view to one side here. This transcendental argument for epistemological disjunctivism is given its clearest expressions in McDowell (1982, 2008). For a helpful discussion, see Haddock and Macpherson (2008b: }2).

246

NEW SYNTHESES

example, the problem of radical skepticism. First, there is the epistemological disjunctivist’s claim that as far as other epistemological proposals go this problem looks unsolvable. In particular, the claim is that epistemological proposals which do not take the epistemological disjunctivist route will have to swallow one of two bitter pills. On the one hand, epistemically externalist positions will only be able to evade the skeptical problem by making the epistemic support for our beliefs opaque to us, even in epistemically optimal conditions. On the other hand, classical epistemically internalist positions seem unable to say very much at all in response to the skeptical problem, since for them the reﬂectively accessible epistemic support we have for our beliefs is no better than the epistemic support possessed by our envatted counterparts. Thus, the skeptical problem seems inevitable. We can get a ﬂavor of this last point by noting how Conee and Feldman (2004: ch. 12) struggle with the skeptical problem. After all, on their view it is hard to see how the reﬂectively accessible evidence that an agent might have in support of her beliefs could ever sufﬁce to give her an epistemic basis for her everyday beliefs as opposed to their skeptical alternatives, something which Conee and Feldman in effect admit (2004: 303). Ultimately, the maneuver that Conee and Feldman are forced into attempting is to bring to bear some sort of inference to the best explanation strategy. That is, we have better evidence for our everyday beliefs than for their skeptical alternatives because we have additional abductive evidence in favor of the former. Even Conee and Feldman agree that this is a fairly contentious strategy to employ, particularly in this regard. Even they would surely admit, then, that if there is a better anti-skeptical response available to the epistemic internalist then it should be taken. The next key step is to argue that epistemological disjunctivism is in a far better position than its rivals on this score. In particular, the claim is that epistemological disjunctivism has an easier time with the skeptical problem since it denies a key prop in the skeptic’s reasoning—viz., the idea that the reﬂectively accessible epistemic support that one has for one’s beliefs is no better than the reﬂectively accessible epistemic support available to one’s counterparts who are the victims of skeptical scenarios. Thus, even though it is true that we are unable to discriminate between our normal experiences and, say, the experiences of an envatted counterpart, it does not follow that we are epistemically no better off. Moreover, since this epistemic standing is reﬂectively accessible, one does not face the problem of conceding, with the epistemic externalist, that the nature of one’s epistemic support is opaque to one. More generally, I take it that a key attraction of epistemological disjunctivism is that it appears to enable us to incorporate the best features of both epistemological internalism and epistemological externalism. In the case of the former, we retain the idea that where one’s belief is justiﬁed the epistemic support for that belief must be reﬂectively accessible to one. In the case of the latter, we retain the idea that one’s epistemic support is paradigmatically truth-conducive. Given these putative theoretical advantages, the natural question that arises is why more people aren’t epistemological disjunctivists. This is where the point that

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

247

McDowell makes about how epistemological disjunctivism is theoretically open becomes important, since he charges that we are led towards the opposing position by reasoning which is far from compelling. In particular, he argues that while the opposing line is often presented as the inevitable outcome of philosophical theorizing, it is in fact nothing of the sort. Indeed, he claims that it is only when we are in the grip of (what he regards as) a faulty ‘highest common factor’ conception of epistemic support that one is led to deny epistemological disjunctivism, and that the way this conception of epistemic support is motivated has an important lacuna, such that it is resistible. I take it that what McDowell has in mind in this regard is reasoning something like as follows: The Highest Common Factor Argument 1. In the bad case, the epistemic support for one’s belief is weaker than factive epistemic support. [Premise] 2. Agents are unable to discriminate between the good case and the corresponding bad case. [Premise] C1. So, the epistemic support that the agent has in favor of her belief in the good case can be no better than it is in the bad case. [From (2)] C2. So, in the good case, the epistemic support for one’s belief is weaker than factive epistemic support. [From (1), (C1)] I think McDowell is right to ﬁnd this style of reasoning dubious, as it stands, for why should we grant that the epistemic standing of one’s belief in the bad case dictates the epistemic standing of one’s belief in the good case? Note that the motivation for this move cannot come from epistemological internalism, since epistemological disjunctivism itself endorses the two key epistemic internalist theses of (ACCESS) and (MENT). Indeed, it ought to be clear that the motivation in fact comes from the contested (DISC), since what underlies this principle is the thought that where a subject is unable to discriminate between her actual experiences and the experiences of her deceived counterpart then there can be no difference in epistemic standing in the two cases. Since all will agree that the epistemic support available in the bad case is fairly limited, it follows that the epistemic support available in the good case will also be similarly limited. But given that (DISC) is the key suppressed premise in play in this argument, then there is no reason why an epistemological disjunctivist, who explicitly rejects (DISC), should be persuaded by it. In particular, the inference from premise (2) to (C1) is blocked. The highest common factor conception of epistemic support thus does not give one grounds to reject epistemological disjunctivism, and so this option is still live. But if epistemological disjunctivism has signiﬁcant theoretical advantages when compared to its rivals, and also accords better with common-sense, then why not endorse it?

248

NEW SYNTHESES

15.9 As it happens, I think there are some serious problems with the way McDowell presents his form of epistemological disjunctivism. In particular, I think McDowell owes us an explanation of why his view is not subject to a ‘McKinsey-style’ argument to the effect that we are able to have a priori knowledge of speciﬁc external world propositions. After all, if we have reﬂective access to factive empirical reasons then, given that one can surely come to know a priori that these factive empirical reasons entail speciﬁc empirical facts, why isn’t it possible on this basis to acquire a priori knowledge of the target speciﬁc empirical facts? Relatedly, I think there is an unresolved tension between allowing reﬂectively accessible empirical reasons and endorsing the truism that one is unable to discriminate between good and bad cases. For given the former thesis it surely follows that the rational basis for one’s belief when one is in the good case actually entails that one is not in the bad case. But given that this rational basis is reﬂectively accessible, why isn’t one in a position to make the relevant discrimination? Finally, I think there is a problem for McDowell’s view as it stands when it comes to explaining why ﬂat-out anti-skeptical assertions of a broadly ‘Moorean’ stripe seem so inappropriate. After all, in good cases at least such assertions are not only true on this view but also supported by reﬂectively accessible factive reasons, so it is hard to understand what the source of such impropriety is supposed to be. Indeed, I think there are other, less central, problems with McDowell’s presentation of the epistemological disjunctivism (for example, some of his more philosophically quietistic remarks). Crucially, however, I also think that there are some pretty compelling things that one can say in response to these problems. For example, I think that one can draw an independently plausible distinction between favoring and discriminating epistemic support that offers the basis on which one can deal with the three core problems just outlined. Moreover, as I have argued quite extensively elsewhere, once epistemological disjunctivism is given the additional support that it needs to avoid these problems then it starts to look like a very attractive thesis indeed (and so the philosophical quietism so associated with McDowell’s work in this regard becomes unnecessary).15 Rather than review the nature of this additional support here, however, I want instead to focus on the very speciﬁc series of objections that Conee (2007; cf. Conee 2008) raises to disjunctivism—interestingly, he doesn’t appear to be troubled by the concerns just raised—since these will obviously be more salient for our present purposes.

15.10 For the most part, Conee’s critique of disjunctivism focuses on metaphysical disjunctivism. As noted above, it is not obvious that the speciﬁcally epistemological disjunctivism that we

15

See in particular, Neta and Pritchard (2007) and Pritchard (2007, 2008a, 2008b, 2009, 2010).

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

249

are interested in is hostage to metaphysical disjunctivism, and Conee himself doesn’t argue that it has to be, so the import of much of his critique of disjunctivism is therefore moot.16 That said, Conee does raise three key objections against epistemological disjunctivism which would apply regardless of whether the position were wedded to metaphysical disjunctivism, so let us focus on them here. We will take them in turn. The ﬁrst is that epistemological disjunctivism is unable to explain the intuition underlying (NEG). In particular, Conee (2007: 20) charges epistemological disjunctivism with being unable to explain the ‘equal reasonableness’ of the beliefs held by the agent and her deceived counterpart. Thus far, we have represented the epistemological disjunctivist as simply rejecting this thesis, which naturally prompts the question of why so many ﬁnd this claim so intuitive if it is meant to be baldly false. In particular, why is it that so many epistemologists who, like epistemological disjunctivists, are drawn to epistemological internalism ﬁnd this thesis intuitive if it is simply false? There is, however, a very straightforward explanation available to the epistemological disjunctivist as to why these two beliefs are equally reasonable, which is that both beliefs are equally epistemically blameless. After all, given the nature of the deception involved, it is no failing of the deceived subject that she believes what she does (the epistemological disjunctivist doesn’t deny, after all, that the agent is unable to distinguish between the two cases). Unless one further holds that two agents who are equally epistemically blameless are thereby of equal justiﬁcatory standing more generally, however, then accepting this claim is entirely consistent with endorsing epistemological disjunctivism. It could be that Conee ﬁnds this distinction between epistemic blame and justiﬁcatory standing more generally dubious, or perhaps he thinks that mere epistemic blamelessness cannot capture the intuitive sense of reasonableness that is in play here. But if this is the case then we need an argument to support these claims. Indeed, on the ﬁrst point note that the epistemological disjunctivist has a pretty straightforward explanation available to them of why we shouldn’t reduce general justiﬁcatory standing to mere epistemic blame. After all, on their view the rational support that is reﬂectively available to the agent in the good case is signiﬁcantly superior to that available to the agent in the bad case. Thus, there is a difference in the respective justiﬁcatory standings of the two agents even though, ex hypothesi, they are equally blameless and thus in this sense equally reasonable in how they have formed their beliefs. Accordingly, as matters stand I don’t see any reason why the epistemological disjunctivist cannot help herself to this distinction in order to account for why (NEG) can seem so intuitive. Hence, Conee’s objection here lacks bite.17

16

As Conee himself concedes—see, e.g., Conee (2007: 35). Similar remarks apply to the parallel objection that Conee raises against Millar’s account of what the epistemological disjunctivist should say about ‘zebra’ cases. See Conee (2007: 32). 17

250

NEW SYNTHESES

15.11 The second criticism raised by Conee as regards epistemological disjunctivism concerns the problem posed by defeaters. Conee (2007: 19) argues that if epistemological disjunctivism were indeed to offer an epistemic advantage over rival views then it would have to be the case that a belief formed in response to a veridical perception is less susceptible to defeat than a belief formed in response to a non-veridical perception (by one’s envatted counterpart, say). But this is not the case, argues Conee, since if one were reliably informed that one is subject to a deception then this would give one just as much reason to doubt the target proposition regardless of whether the belief was formed in response to a veridical perception. Epistemological disjunctivism, as we are understanding the view here at least, is not susceptible to this problem. What is key to this view is only that in the good case the agent’s epistemic support is both factive and reﬂectively accessible. Crucially, however, if one is in possession of an undefeated misleading defeater then one is not in the good case as we are understanding that notion, since the cognitive conditions are not then epistemically advantageous. Thus, since the epistemological advantages of epistemological disjunctivism are meant to be manifest only in the good case, it is by-the-by to evaluate the view by comparing the epistemic standing of two beliefs formed in nongood cases. Still, Conee raises an important point here which concerns what to say about the range of cases that fall between good cases and bad cases as I have deﬁned them here. In particular, let us focus on what we should say about the epistemic differences between (i) a good case; (ii) a case exactly like the good case except that the agent is in possession of a misleading undefeated defeater; and (iii) a bad case. We argued above that the epistemic standing of the agent’s respective beliefs in (i) and (iii) are different. The question, though, is what to say about (ii). There are, it seems to me, two ways for the epistemological disjunctivist to go here. One way is to simply claim that in a type-(ii) case the agent is not in possession of the target factive reason. An example in this regard of a type-(ii) case is a situation where one is not in fact in barn fac¸ade county, and one forms one’s belief that there is a barn before one by looking at a genuine barn, but where one is told, falsely, by a reliable informant that one is in barn fac¸ade county. The line just canvassed would say that in this case one does not see that there is a barn (even though one does obviously see a barn), and hence this cannot be the rational basis for one’s belief. Although I can see the attraction of such a proposal, on balance I think that it is not the right thing for the epistemological disjunctivist to say about such cases. After all, from an ‘objective’ point of view at least, the cognitive conditions are epistemically advantageous, so why is it that the agent doesn’t ‘take in’ the relevant fact? With that in mind, I think a second possible view in this regard is to allow that the agent does see that p in this case but argue that, nonetheless, one’s seeing that p cannot form the rational basis of one’s belief in the target proposition given the presence of the

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

251

undefeated misleading defeater. Indeed, I think this line chimes better with what McDowell says at a number of junctures about ‘doxastic responsibility’ (e.g. McDowell 1993: 429)—viz., the idea that when an agent has knowledge then the target belief must be responsibly formed (as clearly it wouldn’t be in the case just described). Moreover, if the idea driving (ACCESS) is that the rational basis of one’s belief must be reﬂectively accessible to one, then there is no tension between taking this line while holding on to this thesis, since on this view one’s seeing that p cannot be the rational basis of one’s belief (even though one does see that p). Viewing the matter in this way also enables to us to deal with a third problem that Conee (2007: 21) raises, which is that merely seeing that p will not sufﬁce for knowledge in cases in which the target belief is not rationally formed on this basis. Hence, argues Conee, there is no epistemic advantage offered by epistemological disjunctivism, since even when one’s belief is based on one’s seeing that p it does not follow that knowledge is any more assured thereby than if one’s belief has a different, non-factive, basis. But the cases that we are interested, the cases which mark the difference between epistemological disjunctivism and standard epistemic internalist views, are precisely those cases in which the belief is appropriately formed on the factive basis. Accordingly, that there may be no epistemic advantage in the cases that Conee describes does not entail that there is no epistemic advantage to be had by opting for epistemological disjunctivism.18

15.12 In conclusion, then, I have argued that there is a viable epistemological proposal available which can incorporate the sort of epistemically internalist insights that drive Conee and Feldman’s classically internalist view but which also delivers some distinctive theoretical advantages over its rivals. Moreover, I have argued that Conee’s reasons for rejecting such a view do not convince.19

References Byrne, A. and H. Logue. 2008. ‘Either/Or’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge, pp. 57–94. Oxford: Oxford University Press. Conee, E. 2007. ‘Disjunctivism and Anti-Skepticism’, Philosophical Issues, 17: 16–36. ——. 2008. ‘Opposing Skepticism Disjunctively’, typescript. —— and R. Feldman. 2004. Evidentialism. Oxford: Oxford University Press. Goldman, A. (1986). Epistemology and Cognition. Cambridge, MA: Harvard University Press.

18 I discuss these issues further in Pritchard (2008a), where I distinguish between a range of different ‘good’ and ‘bad’ cases and discuss the epistemological differences that they generate. See also note 15. 19 I am grateful to Evan Butts, Adam Carter, Trent Dougherty, Adrian Haddock, Alan Millar, and Ram Neta for helpful discussion on issues covered in this paper.

252

NEW SYNTHESES

Greenough, P. and D. H. Pritchard (eds). 2009. Williamson on Knowledge. Oxford: Oxford University Press. Grice, H. P. 1961. ‘The Causal Theory of Perception’, Proceedings of the Aristotelian Society (suppl. vol.), 35: 121–52. Haddock, A. and F. Macpherson (eds). 2008a. Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. ——. 2008b. ‘Introduction: Varieties of Disjunctivism’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge, pp. 1–24. Oxford: Oxford University Press. Hinton, J. M. 1973. Experiences. Oxford: Oxford University Press. Lehrer, K. and S. Cohen. 1983. ‘Justiﬁcation, Truth, and Coherence’, Synthese, 55: 191–207. McDowell, J. 1982. ‘Criteria, Defeasibility, and Knowledge’, Proceedings of the British Academy, 68: 455–79. ——. 1986. ‘Singular Thought and the Extent of Inner Space’, in P. Petit and J. McDowell (eds), Subject, Thought, and Context. Oxford: Clarendon Press. ——. 1993. ‘Knowledge by Hearsay’, in B. K. Matilal and A. Chakrabarti (eds), Knowing from Words: Western and Indian Philosophical Analysis of Understanding and Testimony. Dordrecht: Kluwer. ——. 1995. ‘Knowledge and the Internal’, Philosophy and Phenomenological Research, 55: 877–93. ——. 2002a. ‘Knowledge and the Internal Revisited’, Philosophy and Phenomenological Research, 64: 22–30. ——. 2002b. ‘Responses’, in N. H. Smith (ed.), Reading McDowell: On Mind and World. London: Routledge. ——. 2008. ‘The Disjunctive Conception of Experience as Material for a Transcendental Argument’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge. Oxford: Oxford University Press. Martin, M. 2003. ‘The Transparency of Experience’, Mind and Language, 17: 376–425. ——. 2004. ‘The Limits of Self-Awareness’, Philosophical Studies, 120: 37–89. Millar, A. 2007. ‘What the Disjunctivist is Right About’, Philosophy and Phenomenological Research, 74: 176–98. ——. 2008. ‘Perceptual-Recognitional Abilities and Perceptual Knowledge’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge, pp. 330–47. Oxford: Oxford University Press. ——. 2009. ‘Disjunctivism and Skepticism’, in J. Greco (ed.), Oxford Handbook to Skepticism. Oxford: Oxford University Press. Neta, R. 2008. ‘In Defence of Disjunctivism’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge, pp. 311–29. Oxford: Oxford University Press. Neta, R. and D. H. Pritchard. 2007. ‘McDowell and the New Evil Genius’, Philosophy and Phenomenological Research, 74: 381–96. Nuccetelli, S. (ed.) 2003. New Essays on Semantic Externalism and Self-Knowledge. Cambridge, MA: MIT Press. Pritchard, D. H. 2005. Epistemic Luck. Oxford: Oxford University Press. ——. 2006. ‘Review of Evidentialism, by E. Conee and F. Feldman’, The Philosophical Quarterly, 56: 147–9. ——. 2007. ‘How to be a Neo-Moorean’, in S. Goldberg (ed.), Internalism and Externalism in Semantics and Epistemology. Oxford: Oxford University Press.

EVIDENTIALISM, INTERNALISM, DISJUNCTIVISM

253

——. 2008a. Epistemological Disjunctivism, book manuscript (under contract with Oxford University Press). ——. 2008b. ‘McDowellian Neo-Mooreanism’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge, pp. 283–310. Oxford: Oxford University Press. ——. 2009. ‘Wright Contra McDowell on Perceptual Knowledge and Scepticism’, Synthese, 171: 467–79. ——. 2010. ‘Relevant Alternatives, Perceptual Knowledge and Discrimination’, Noûs, 44: 245–68. Snowdon, P. 1980–1. ‘Perception, Vision, Causation’, Proceedings of the Aristotelian Society, 81: 175–92. ——. 1990–1. ‘The Objects of Perceptual Experience’, Proceedings of the Aristotelian Society (suppl. vol.), 64: 121–50. Williamson, T. 1995. ‘Is Knowledge a State of Mind?’, Mind, 104: 533–65. ——. 1996. ‘Cognitive Homelessness’, Journal of Philosophy, 93: 554–73. ——. 1997. ‘Knowledge as Evidence’, Mind, 106: 717–41. ——. 2000a. Knowledge and its Limits. Oxford: Oxford University Press. ——. 2000b. ‘Scepticism and Evidence’, Philosophy and Phenomenological Research, 60: 613–28. Wright, C. J. G. 2004. ‘Warrant for Nothing (and Foundations for Free)?’, Proceedings of the Aristotelian Society, 78 (supp. vol.): 167–212. ——. 2008. ‘The Disjunctive Conception of Experience as Material for Underestimation of the Challenge of Scepticism’, in A. Haddock and F. Macpherson (eds), Disjunctivism: Perception, Action, Knowledge, pp. 376–89. Oxford: Oxford University Press.

16 Toward a Synthesis of Reliabilism and Evidentialism? Or: Evidentialism’s Troubles, Reliabilism’s Rescue Package Alvin I. Goldman

For most of their respective existences, reliabilism and evidentialism (that is, process reliabilism and mentalist evidentialism) have been rivals. They are generally viewed as incompatible, even antithetical, theories of justiﬁcation.1 But a few people are beginning to re-think this notion. Perhaps an ideal theory would be a hybrid of the two, combining the best elements of each theory. Juan Comesana (2010) takes this point of view and constructs a position called ‘Evidentialist Reliabilism’. He tries to show how each theory can proﬁt by borrowing elements from the other. Comesana concentrates on reliabilism’s problems and how it might be improved by infusions from evidentialism. This paper follows a similar tack. My emphasis, however, is the reverse of Comesana’s. I highlight problems for evidentialism and show how it could beneﬁt by incorporating reliabilist themes. I am not sanguine that evidentialists will see it my way. They might even view my proposals as an insidious attempt to convert evidentialists to reliabilism. Well, I won’t debate the best way to formulate this paper’s recipe. At any rate, it began with the idea (which anteceded my reading of Comesana) of creating a synthesis of reliabilism and evidentialism. It retains signiﬁcant strands of that idea, although the synthesis theme does not pervade the entire chapter. What is mentalist evidentialism? Its original formulation was succinct: EJ. Doxastic attitude D toward proposition p is epistemically justiﬁed for S at t if and only if having D toward p ﬁts the evidence S has at t. (Feldman and Conee 1985/ 2004: 83)

1 See Goldman (1979, 1986), Feldman and Conee (1985), Conee and Feldman (1998), Goldman (1999c), Conee and Feldman (2001).

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

255

I begin by raising worries about the notions of evidence and ﬁttingness that are crucial to evidentialism as formulated by [EJ]. To help with many of the problems raised, I recommend supplements borrowed from reliabilism. As we proceed I shall also present reasons why reliabilism might be improved by doses of evidentialism.

16.1 The concept of evidence in evidentialism A theory of justiﬁcation that calls itself ‘evidentialism’ and makes evidence possession its focus can reasonably be asked to explain which concept of evidence it means to invoke. If the only viable concept of evidence available to it introduces elements from a ‘foreign’ approach such as reliabilism, then this should provide some motivation toward a synthesis or uniﬁcation with that approach. This is the situation facing evidentialism, as I argue in this section. In an overview of possible conceptions of evidence, Thomas Kelly (2006) offers several conceptions worthy of attention. First, evidence may be understood as ‘that which justiﬁes belief.’ As Jaegwon Kim puts it, ‘When we talk of “evidence” in an epistemological sense we are talking about justiﬁcation: one thing is “evidence” for another just in case the ﬁrst tends to enhance the reasonableness or justiﬁcation of the second’ (1988: 390–1). Can this be the sense of the term Feldman and Conee have in mind? Not if their evidentialist theory is intended to provide an analysis or explanatory account of justiﬁcation. Yet, clearly, this is what their evidentialist theory, as expressed in [EJ], is intended to provide. If ‘evidence’ is deﬁned as ‘that which justiﬁes belief’, then the deﬁnition of ‘justiﬁed’ in terms of ‘evidence’, as proposed in [EJ], is circular and unhelpful. Next consider a conception of evidence proposed by Timothy Williamson (2000), in which knowledge is equivalent to evidence. At any rate, according to Williamson, all items of evidence are pieces of knowledge and vice-versa. Could Feldman and Conee avail themselves of the sense of ‘evidence’ in which evidence is equivalent to knowledge? This poses the same problem of circularity as the ﬁrst deﬁnition. Feldman and Conee hold that ‘knowledge’ is to be analyzed in terms of justiﬁcation, which, of course, is analyzed in terms of evidence.2 In light of these commitments, it would again be circular to analyze or deﬁne ‘evidence’ in terms of ‘knowledge’. Another conception of evidence Kelly considers is evidence ‘as a guide to truth’. In other words, something is evidence for p just in case it is a reliable sign, symptom, or mark of the truth of p. This is an appealing conception of evidence, which makes sense of the term’s use in many walks of life. In criminal law, it is plausible that what a court admits as evidence should be items that (when properly interpreted) are reliable guides

2 A commitment to deﬁning ‘knowledge’ in terms of ‘justiﬁcation’ is clear at least for Feldman, who writes: ‘knowledge requires justiﬁed true belief that does not essentially depend upon a falsehood’ (Feldman 2003: 37). I know of no reason to suspect that Conee disagrees, at least about justiﬁcation being deﬁnitionally necessary for knowledge.

256

NEW SYNTHESES

to truth, or signs of that for which they are taken to be evidence. Similarly, in science the reading of a gauge or instrument is evidence for a certain object’s having property F just in case the reading is a reliable indicator of the object’s possessing F. Is this deﬁnition one that Feldman and Conee could adopt? It is certainly compatible with certain strands of their approach. There is no looming circularity of the sort that besets the ﬁrst two deﬁnitions. Furthermore, it is compatible with Feldman and Conee’s mentalism about evidence, because mental states, events, and processes often qualify as indicators of the truths for which they are (ostensible) evidence. At least this is so if we don’t inhabit an evil-demon-world and are not otherwise badly deceived about the actual world. When it visually appears to someone that there is a computer monitor before him, this visual state is usually a reliable sign or indicator that a computer monitor is before him. When a person has an ostensible memory impression that she ate oatmeal for breakfast this morning, the memory impression usually indicates that she did eat oatmeal this morning. Under the truth-indicator conception of evidence, it looks like these two experiences will constitute (prima facie) evidence for just the propositions one would expect them to be evidence for—and that Feldman and Conee take them to be evidence for. So far, then, the reliable indicator interpretation of evidence is one that Feldman and Conee should ﬁnd congenial. Feldman and Conee would probably resist this suggestion, of course, for two closely related reasons. First, they hold that even in an evil-demon world a visual experience as of a computer monitor is evidence for the presence of a computer monitor (Feldman 1985), and one is justiﬁed in believing that a computer monitor is present even though such visual experiences are not reliable indicators of the presence of a monitor in such a world. A related problem is that truth-indicatorship properties vary from world to world, so that what is evidence for what is contingent under this approach. But Feldman and Conee want evidential relationships to be necessary. It should not be assumed, however, that truth-indicatorship in the world of an example is the optimal interpretation of the reliable indicatorship approach. Instead, truth-indicatorship relationships might be rigidiﬁed so that they are ﬁxed by the correlations that obtain in the actual world. If this approach is adopted, even a cognizer in an evil-demon world will be justiﬁed in his perceptual beliefs (see Goldman 2008a). Comesana embeds roughly this approach in a two-dimensional semantics (Comesana 2002, 2010) and calls it ‘indexical reliabilism’. In particular, following Stalnaker’s (1978) version of two-dimensional semantics, there are two different propositions associated with any attribution of justiﬁcation: the ‘diagonal proposition’, which implies that the belief is produced by a method that is reliable in the world where the belief is held; and the ‘horizontal’ proposition, which implies that the belief is produced by a process that is reliable in whichever world the proposition is considered. The latter interpretation might be acceptable to evidentialists, both in terms of allowing perception-based justiﬁedness in an evil-demon world and in terms of complying with the necessity constraint on the evidential relation.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

257

In this section we have noted that there are multiple conceptions, or deﬁnitions, of ‘evidence’. Thus, evidentialism owes us an indication of what it means by ‘evidence’. Until we are told how to interpret the term ‘evidence’, for purposes of this theory, we cannot begin to assess its (extensional) adequacy. As we have seen, moreover, some deﬁnitions of ‘evidence’ are not viable for evidentialism because of circularity. The only deﬁnition I am able to pinpoint that does not suffer from this liability employs some sort of reliability notion. Could evidentialism adopt this deﬁnition? Pure evidentialism, understood as a rival or opponent of reliabilism, could not adopt it. But when we consider the prospects of a hybrid approach, this might be an attractive option. Mental states would serve as the items of evidence under the hybrid theory, but they would qualify as items of evidence (ultimately) because they—or their ilk—stand in reliable-indicator relationships to facts in the world. Although Feldman and Conee themselves are unlikely to applaud this maneuver, other epistemologists in search of the best overall theory might ﬁnd it a congenial solution to the deﬁnitional problem.

16.2 Justiﬁcation and ﬁttingness with non-doxastic evidence Two notions of justiﬁedness are commonly recognized in the literature: propositional and doxastic justiﬁedness. Feldman and Conee’s (1985) label for the latter is ‘wellfoundedness’.3 Let us concentrate on this second notion. Feldman and Conee’s account of doxastic justiﬁedness (well-foundedness) is formulated as follows: WF S’s doxastic attitude D at t toward proposition p is well-founded if and only if (i) having D toward p is justiﬁed for S at t; and (ii) S has D toward p on the basis of some body of evidence e, such that – S has e as evidence at t; – having D toward p ﬁts e; and – there is no more inclusive body of evidence e0 had by S at t such that having D toward p does not ﬁt e0 . (1985/2004: 93) Obviously, mentalist evidentialism owes us an account of ﬁttingness. Two types of cases need to be covered, ﬁttingness as applied to inferential justiﬁedness and ﬁttingness as applied to non-inferential justiﬁedness. I begin with non-inferential justiﬁedness. Feldman and Conee certainly appear to believe in non-inferential justiﬁedness. Feldman’s textbook Epistemology (2003) seems to favor (modest) foundationalism as the best approach to justiﬁcation, and foundationalism is committed to basic justiﬁedness, which is noninferential. Conee’s paper ‘The Basic Nature of Epistemic Justiﬁcation’ (1988/2004), on the other hand, appears to favor a combination of foundationalism and coherentism. 3 That Conee and Feldman equate well-foundedness with doxastic justiﬁedness was indicated in personal communication with Conee and afﬁrmed by Feldman at a 2008 conference at Rochester where Feldman served as commentator on an early version of this material.

258

NEW SYNTHESES

Conee agrees, however, that such a theory must acknowledge foundational experiences (1988/2004: 43). Thus, both theories seem to be committed to some strand of nondoxastic justiﬁedness, and, given evidentialism, this must arise from some non-doxastic species of ﬁttingness.4 Non-doxastic ﬁttingness is what I examine in the present section. Let us begin with non-inferential beliefs based on perceptual experience or introspection. Almost all foundationalists will accept introspectively-based beliefs (at least some of them) as non-inferentially justiﬁed. Modest foundationalists, such as Feldman and Conee, also accept some perceptual beliefs as non-inferentially justiﬁed. The question is how, in detail, evidentialism can handle these cases of justiﬁedness. In both cases the problem I have in mind may be called the selection problem. Some perceptually based beliefs are justiﬁed and other beliefs are unjustiﬁed; similarly for introspectively based beliefs. How can mentalist evidentialism explain which ones are which? Feldman (2003) is fully aware of the problem and provides a good example. Maurice enters a room, sees a 12-year-old table, and forms two beliefs about it: a belief that it is a table (B(T)) and a belief that it’s a 12-year-old table (B(TYOT))—although there’s nothing in its appearance that hints at its age. Presumably, B(T) is well-founded and B (TYOT) is ill-founded. Why? Feldman wants to say, of course, that the former belief but not the latter ‘ﬁts’ Maurice’s experience. But how can ﬁttingness be spelled out so as to yield this result? Feldman tries to explain this ﬁttingness by appeal to the notion of a ‘proper response to experience’. Unfortunately, ‘proper response’ seems like a mere paraphrase of ‘ﬁttingness’, so there is a clear threat of the explanation being unilluminating. Feldman does try to explain in non-circular terms what is meant by a ‘proper response’ to experience. The explanation, however, does not cover all of the relevant territory (see Goldman 2008b for details). For example, Feldman says that when the contents of a belief are ‘closer’ to the direct contents of the experience, they are more apt to be properly based on experience. This approach assumes that proper responsiveness is always a matter of content matching, but this cannot be right. If I form a belief that this mental state I am internally demonstrating is a desire, this belief may well be justiﬁed. But the belief ’s content—that this mental event is a desire—need not match the desire’s content, which could be anything, and, in particular, need not concern a desire.5 Another move Feldman makes is to appeal to the subject’s training. He uses this to explain the difference in well-foundedness between an expert and a novice birdwatcher, who both identify a certain bird as of type X. The expert, says Feldman, should be credited with a well-founded belief, the novice (who was just guessing) with an ill-founded one. But how, exactly, is their respective training relevant? A simple and

By ‘non-doxastic ﬁttingness’ I mean ﬁttingness of a belief to one or more non-doxastic mental states, for example, perceptual experiences. This has nothing to do with the distinction between propositional and doxastic justiﬁedness. 5 I don’t mean to dismiss content matching or ﬁttingness entirely in an account of non-inferential justiﬁedness. I return to this topic in Section 16.7 below. 4

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

259

direct answer comes from process reliabilism. The relevance of training, quite simply, is that it makes the bird expert’s bird-spotting judgments consistently come out true; the absence of relevant training makes the novice’s bird-spotting judgments come out only randomly, or occasionally, true. This is a thinly disguised way of saying that the expert has acquired a reliable process for spotting birds (at least birds of type X); the novice has not acquired any such reliable process, although he lucked out with his present guess. By my lights, evidentialism would be greatly improved if it availed itself of this ingredient of process reliabilism. Its capacity to illuminate cases of non-inferential well-foundedness could thereby be much enhanced. The foregoing diagnosis and recommendation also apply to the problem of introspective justiﬁedness. A well-known problem in this arena is Chisholm’s (1942) problem of the speckled hen. Exactly which types of belief about one’s own present experience are justiﬁed? Even if there is a determinate number of apparent speckles associated with a given visual experience, a subject who believes that the hen appears to have exactly 43 speckles will not necessarily be justiﬁed in so believing. That depends on how the subject deploys his introspective process and associated processes. Psychologists give the label ‘subitizing’ to a process of simply scanning a set of objects and coming up with a numerical estimate of them. For a small number of objects, subitizing is quite reliable. There is no need for deliberate counting. For larger sets of objects, subitizing is not accurate. Now suppose that a subject uses subitizing to arrive at a belief that his current visual experience features exactly 43 apparent speckles; then he isn’t, intuitively, justiﬁed—even if, by chance, 43 is the correct number. This diagnosis can be accommodated within a theory of justiﬁcation that highlights mental processes and their reliability, but—as far as I can see—only within such a theory. Evidentialism as currently formulated is not such a theory. I do not mean to reject categorically the notion of ﬁttingness for purposes of a theory of justiﬁcation. Indeed, Feldman’s appeal to content matching in the account of ﬁttingness may hold some merit, at least as part of the story. I return to these topics in a more ecumenical spirit in Sections 16.4 and 16.7 below.

16.3 Memory belief and current evidence Another domain of non-inferential justiﬁedness to which foundationalists commonly appeal is memory-based beliefs. A familiar idea is that beliefs about the past can be justiﬁed by ostensible memories, viewed as non-doxastic states. If I seem to recall eating a bagel for breakfast this morning, this ostensible memory creates justiﬁcation for my belief that I did eat a bagel this morning. Presumably, Feldman and Conee will be happy to say that memory beliefs of this sort are justiﬁed insofar as they ‘ﬁt’ the nondoxastic evidence provided by a conscious memory experience of this kind. I will not challenge this theory when it comes to beliefs about the past based on such experiences. But another kind of memory—so-called ‘preservative’ memory—has a rather different epistemic role to play in our cognitive life. Preservative memory does not create or

260

NEW SYNTHESES

generate justiﬁedness ‘from scratch’, but instead transmits a belief ’s justiﬁedness (or unjustiﬁedness) from one time to a later time. Here, it seems to me, evidentialism runs into serious trouble when asked to account for this epistemic role. Here is an example. Years ago Ichabod formed a belief in proposition q by acquiring it in an entirely justiﬁed fashion. He had excellent evidence for believing it at that time (whether it was inferential or non-inferential evidence). After ten years pass, however, Ichabod has forgotten all of this evidence and not acquired any new evidence, either favorable or unfavorable. However, he continues to believe q strongly. Whenever he thinks about q, he (mentally) afﬁrms its truth without hesitation. At noon today Ichabod’s belief in q is still present, stored in his mind, although he is not actively thinking about it. I stipulate that none of his other beliefs confers adequate evidence either for believing q or for disbelieving it. Since Ichabod remembers q’s being the case, and since he originally had excellent evidence for q, which was never subsequently undermined, Ichabod’s belief in q at noon today is justiﬁed.6 Moreover, as I argue elsewhere (Goldman 2009), if we refuse to grant justiﬁedness to beliefs of this sort, which derive from preservative memory, there will be serious skeptical ramiﬁcations: people will fail to know a great many things that common sense credits them with knowing. But does Ichabod’s noontime belief in q satisfy Feldman and Conee’s account of well-foundedness? Does it qualify as doxastically justiﬁed under their theory? No. At noontime Ichabod undergoes no evidential experience of seeming to remember q and (by hypothesis) possesses no stored beliefs on which his belief in q is based—or which it ﬁts. Years ago, as indicated, he had evidence that his belief in q ﬁtted, and perhaps his noontime q-belief might be said to be ‘based on’ that old, forgotten evidence. But can a basing relation help Feldman and Conee in this case? Their analysis of well-foundedness explicitly requires that justifying evidential states must be held at the same time as the target attitude. But the earlier justifying evidence is no longer possessed, neither consciously nor unconsciously. So evidentialism implies that Ichabod’s noontime belief in q is not justiﬁed (well-founded), an intuitively incorrect verdict. It should be emphasized that in the Ichabod case there is no memory experience that triggers his noontime belief in q, the sort of experience foundationalists typically invoke to account for justiﬁed memory-based belief. But many justiﬁed memorybased beliefs—especially stored, or non-occurrent, memory-based beliefs—are unaccompanied by memory experiences. Nonetheless, such beliefs need to be credited with justiﬁedness. Unless they are justiﬁed, they cannot be known—certainly not on a justiﬁcationist view of knowledge. Yet non-occurrent knowledge—that is, knowledge involving non-occurrent belief—constitutes the vast bulk of our knowledge at any

6 It may be contended that Ichabod’s noontime belief in q is not as justiﬁed as his belief in q was when it was originally formed and he was in possession of the original evidence. This point may be conceded. Nonetheless, given the history described, the noontime belief has a substantial measure of justiﬁedness, and that measure of justiﬁedness cannot be explained by evidentialism, as far as I can see.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

261

given moment. If, as theorists, we abandon such knowledge, the skeptic will have won a major (but unearned) victory. How can evidentialism cope with this problem? It could improve its handling of these cases by abandoning the simultaneity requirement, the requirement that justifying evidence must be possessed at the same time as the belief. But this requirement is a core part of internalism, to which mentalist evidentialism adheres. In any case, how much would it help evidentialism to follow the above advice and abandon the simultaneity requirement? It would still have to add a rule or condition to mark preservative memory as a justiﬁcation-transmitting feature. It would be extremely ad hoc to simply add such a rule or condition without offering a rationale for it. Why, epistemologists are entitled to ask, does this sort of memory process qualify as a justiﬁcation transmitter? This calls for explanation, and evidentialism has little in its toolbox to offer. What would improve evidentialism’s explanatory prospects on this topic is to borrow two basic ingredients of process reliabilism. These ingredients are (1) the epistemological importance of belief-forming or belief-retaining processes, and (2) the importance of their reliability (or conditional reliability). Preservative memory is a cognitive belief-retaining process that is able to transmit justiﬁedness from an earlier to a later time. This is why Ichabod is still (substantially) justiﬁed in believing q at noon although he has no evidence at noon that this belief ﬁts. Moreover, reliabilism has a plausible-looking explanation of why preservative memory is justiﬁcation transmitting, namely, its conditional reliability. Belief outputs of preservative memory at later times are mostly true if its inputs at earlier times were true (see Goldman 1979, for a discussion of conditional reliability).7 It is not clear how evidentialism can accommodate any of this within its existing (pure) ﬁttingness story, which is why it needs a ‘rescue’ by reliabilism. Some readers might not ﬁnd my treatment of Ichabod persuasive. I stipulated that Ichabod has no other beliefs at noon that provide adequate evidence either for or against q. But is this plausible, really? A critic might argue as follows. Won’t anybody in Ichabod’s situation have general beliefs both about the quality of his memory and about the quality of his usual belief-forming processes? Won’t he typically believe— and believe justiﬁably—that his memory is quite good and that he usually forms beliefs in a justiﬁable fashion? So, if he ﬁnds himself remembering that q and takes account of these background (justiﬁed) beliefs, won’t this justify him in believing q without any reliance on his original evidence? So there is no need to advert to his forgotten evidence to account for his justiﬁedness. The case just described may well be a typical one, but it isn’t the only possible case. So let us consider a different scenario. Suppose Ichabod is being treated by a clinical psychologist, who falsely persuades him that his once-robust memory is no longer 7 I am not assuming that preservative memory is deﬁned as memory with mostly accurate retention of believed content. Rather, I assume it’s a contingent fact that what is held in memory from previous periods is generally ‘true to’ what was there earlier. (Of course, this does not preclude a large amount of forgetting.)

262

NEW SYNTHESES

working well; so he has no reason now for supposing that a stored belief of his (like q) was accurately transmitted from the past. The epistemological critic imagined in the previous paragraph would then have to say that Ichabod is not justiﬁed in believing q. But is this verdict intuitively correct, given that Ichabod in fact reliably retained q from the past and originally acquired the belief in a perfectly sound fashion? At a minimum, Ichabod’s original acquisition and retention of the belief constitute positive forces, or vectors, in support of the justiﬁedness of his belief. The newly formed beliefs about his powers induced by the psychologist may provide a countervailing epistemic force, a vector that militates in the opposite direction of justiﬁedness. But that does not negate the fact that his past evidence for q and subsequent retention of q are factors prior to the current (noontime) belief that positively affect his current justiﬁcational status, something denied by evidentialism. We might supplement this last point by noticing another way to understand the critic’s intuitions without conceding his central claim. His central claim is that the deceived Ichabod is (deﬁnitely) unjustiﬁed in believing q. I am not prepared to concede this (intuitively). On the other hand, I may be prepared to concede a related thesis, namely, that Ichabod is unjustiﬁed in believing that he is justiﬁed in believing q. Since Ichabod believes (albeit falsely) that he no longer enjoys the same powers of memory that he once had, he is not justiﬁed in believing (upon reﬂection) that his current belief in q is justiﬁed. In other words, he lacks second-order justiﬁedness with respect to q. It does not follow from this, however, that he lacks ﬁrst-order justiﬁedness; that is, it does not follow that he is not justiﬁed in believing q. As a general matter, being justiﬁed in believing that one is justiﬁed in believing p does not entail being justiﬁed in believing p ( J Jp 6¼> Jp). Similarly, being unjustiﬁed in believing that one is justiﬁed in believing p does not entail that one is unjustiﬁed in believing p (J Jp 6¼> Jp). We might concede that (J Jp) is always a defeater for Jp, but this is not incompatible with there being defeasible factors working in favor of Jp. Indeed, my contention is that Ichabod’s sound acquisition of q in the past and his retention of q via reliable preservative memory are both factors that work (defeasibly) in favor of his being justiﬁed with respect to believing q. Even if these defeasible factors are overridden by Ichabod’s other current beliefs—a debatable matter—this would not save evidentialism from the current critique. The provisionally conceded defeat of his justiﬁedness with respect to q still allows factors prior to t to be (positively) relevant to the justiﬁcational status of his belief at t (i.e., noon), contrary to the thesis of evidentialism.

16.4 Toward a two-component theory of inferential justiﬁcation In this section I return to the ecumenical project of proposing that reliabilism should incorporate something like the evidential element emphasized by evidentialism. Earlier versions of process reliabilism tried to make do with the reliability of cognitive

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

263

processes as the linchpin. To be sure, it has never hesitated to invoke mental states, both doxastic and non-doxastic states, in its set of resources (Goldman 1979, 1986). But mental states have usually been treated as mere inputs and outputs of processes. No ﬁttingness relation between the target belief and these states has ﬁgured in the formulations of reliabilism. Now, however, I am inclined to suggest that an incorporation of a ﬁttingness relation into a reformulation of process reliabilism might be a salutary addition. This would generate a two-component theory of justiﬁedness as opposed to a single-component theory. The attraction of a two-component approach seems especially obvious in inferential justiﬁcation. Let us see how it would work by ﬁrst replacing the familiar tripartite framework of doxastic attitudes—belief, disbelief, and suspension—with the richer framework involving degrees of credence or subjective probabilities. Next let us assume that for any ordered pair of propositions (p, q), there is some relation of degree of support or conﬁrmation that takes values on the unit interval. Next, assume that p expresses S’s total doxastic evidence relevant to q, that is, the total evidence vis-a`-vis q possessed in doxastic form,8 and that S has no non-doxastic evidence relevant to q. Finally, assume that the degree of conﬁrmation p confers on q is N; in other words, C(q, p) = N. Then it is plausible to hold that doxastic attitude of degree N toward proposition q is precisely the attitude that ﬁts his evidence for q. In other words, if C (q, p) = 0.63, then adopting the degree of credence, or subjective probability, 0.63 would ﬁt the (total) evidence consisting of p. For an evidentialist about justiﬁcation, the doxastic attitude a person in the foregoing conditions is justiﬁed in having vis-a`-vis q is 0.63 (and nothing else).9 This account is squarely in the spirit of the ﬁttingness approach; it contains no trace of processism (thus far). The need for a process-based evaluation can be motivated, however, as follows. Shirley is very poor at determining conﬁrmation relations. When she reﬂects on her total set of beliefs relevant to a hypothesis, she typically draws a blank about the strength of conﬁrmation. She then hazards a wild guess about their degree of support for the hypothesis and forms that degree of belief in it. Proceeding in this fashion, Shirley assigns degree of belief 0.45 to a proposition H. Now consider Madeleine, a highly proﬁcient conﬁrmation theorist. Madeleine has exactly the same

8 In saying that he possesses evidence in ‘doxastic form,’ I mean that he possesses evidence in virtue of either believing categorically or having weaker degrees of credence in propositions that are conﬁrmationally relevant to q. Plausibly, these propositions don’t constitute evidence unless his doxastic attitudes toward them are justiﬁed (at least semi-justiﬁed). I won’t try to settle the latter issue here. 9 Actually, there are more problems lurking here. What dictates the appropriate degree of credence in q is not only the strength of the conﬁrmational relation between the evidence propositions and q, but also how strong are the (justiﬁed) credal attitudes in these various evidential propositions. (I am assuming that they don’t all have to be 1.0, or ﬂat-out belief. I do not feel that it is my job to solve this problem. It is a problem that squarely confronts evidentialism, so the main duty of solving it falls on the shoulders of evidentialists. My proposal in this section is that, assuming this problem can be solved, we will then have on our hands one kind of justiﬁcational factor that could and should be added to the reliable-process factor that reliabilism traditionally emphasizes.)

264

NEW SYNTHESES

evidence vis-a`-vis H as Shirley does. She uses her accurate, well-honed skills at determining degrees of support and arrives at the conclusion that her evidence for H is 0.45. She therefore assigns degree of belief 0.45 to H. Now, on one dimension of justiﬁedness—the ﬁttingness dimension—Shirley and Madeleine’s doxastic attitudes vis-a`-vis H deserve the same rating. Equally clearly, however, there is another dimension of justiﬁedness—call it the process dimension—on which their doxastic attitudes merit different ratings. Madeleine’s degree of belief is much more aptly, or competently, chosen than Shirley’s—despite the fact that they arrive at the same result. On this second dimension of justiﬁedness, Shirley’s degree of belief is not at all justiﬁed or well-founded, whereas Madeleine’s degree of belief is very well-founded. A twofactor theory handles this case nicely. But no single-factor theory, of either the purely evidentialist or purely reliabilist sort, can do so. That’s a good reason to promote a synthesis of the two. In their recent paper, ‘Evidence,’ Conee and Feldman (2008) anticipate part of what I have just said. They anticipate the objection that assigning a belief or degree of belief as a function of the degree of logical or probabilistic support from one’s total evidence does not sufﬁce for justiﬁedness. They do not, however, explicitly consider endorsing any move toward a process factor to accommodate this objection. Instead they write: A person may know some propositions that logically entail some proposition that the person scarcely understands and surely does not know to follow from the things she does know. The logical route from what she knows to this proposition may be complex and go beyond her understanding, or even the understanding of any person. In our view, the person is not then justiﬁed in believing the consequence, even though it is entailed by her evidence. It is noteworthy that, to become justiﬁed in believing the proposition, she has to learn something new— namely, its logical connection to her evidence. (Conee and Feldman 2008: 94)

Let us set aside the possibility that the subject scarcely understands the target proposition. That feature needs separate attention.10 The main problem, correctly identiﬁed in this passage, is that even entailment by the evidence does not sufﬁce to make the subject justiﬁed in believing the target proposition (call it ‘H’). What is Conee and Feldman’s solution? Initially it appears that they seek to solve the problem by adding a requirement that the subject knows the logical connection between the evidence propositions and H. This invites two worries. First, knowledge of a logical (or, more generally, probabilistic) connection does not seem to be necessary for being justiﬁed in believing H. They themselves concede this in the following section of their paper. People do not have to learn principles of logic or logical theory (and similarly for probability theory) to come to be justiﬁed in believing on the basis of logically or probabilistically supportive evidence. Second, how would an acquisition of principles

10 In any case, it isn’t a problem for doxastic justiﬁedness, the variety of justiﬁedness on which I am concentrating for now. Unless the subject understands the proposition, he does not believe it, so the issue of doxastic justiﬁedness does not arise.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

265

of logic or probability help solve the problem? Unless the subject applies appropriate processes, or operations, to the newly learned principles to form a belief in H, he will still fail to have a justiﬁed belief. At least this is so if we are discussing doxastic justiﬁedness (well-foundedness), as I am doing here.11 Where well-foundedness is in question, causal processes of belief-formation (and belief-retention) cannot be brushed under the rug. That is why a two-factor theory, which includes a process dimension, is needed.

16.5 Ultimate and derivative evidence In this section I consider the kinds of states that qualify as evidential states according to Conee and Feldman and the ﬁttingness relation that is supposed to hold between them and justiﬁed doxastic attitudes. Conee and Feldman write: The justiﬁcatory status of a person’s doxastic attitudes strongly supervenes on the person’s occurrent and dispositional mental states, events, and conditions. (Conee and Feldman 2001/ 2004: 56)

In another statement, which appears in a 2004 ‘Afterward’ to Evidentialism (Conee and Feldman 2004), they write: ES. The epistemic justiﬁcation of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at the time. (Conee and Feldman 2004: 101, emphasis added) Thus, mental states, events, and conditions, according to Conee and Feldman, exhaust the evidence (evidential states) that a person ever has, in the preferred sense of evidence they clarify in ‘Evidence’ (Conee and Feldman 2008). What they call ultimate evidence is experiential evidence, including perceptual and memorial experience. Beliefs are also said to be evidence, but not ultimate evidence. ‘We hold that experiences can be evidence, and beliefs are only derivatively evidence’ (2008: 87). Beliefs are ‘kosher’ items of evidence, presumably, because they are mental states and hence internal states. States of knowing, on the other hand, should not be evidential states, because knowing entails truth and truth is not internal. (Conee and Feldman do not subscribe to Williamson’s view that knowing is a mental state.) Strictly speaking, then, it was a little slip when they wrote the passage quoted earlier (from p. 94 of ‘Evidence’) that seems to speak of ‘knowing’ as evidence. If one puts together the passage quoted above from the 2004 ‘Afterward’ with the statement from the 2008 paper ‘Evidence’ about beliefs being merely derivative evidence, it appears that Conee and Feldman mean to hold that the evidential status

11 It is also relevant, however, to propositional justiﬁedness, as the selection problem discussed in Section 16.2 reveals. The reason that Maurice is not propositionally justiﬁed in believing that the table is 12 years old is that he lacks any reliable cognitive process that would take his visual appearance as input and yield a belief that the table is 12 years old as output.

266

NEW SYNTHESES

of any belief held at time t is derivative from the evidential status of experiential states undergone at t. Let us reﬂect on the implications of this view. Call the total body of evidence—that is evidential states—that S possesses at a given time his concurrent body of evidence (CBE). What exactly comprises S’s CBE at a chosen time t? Presumably, all of his experiential states at t plus all of his beliefs and other doxastic states at t—or at least the justiﬁed doxastic states. What I now wish to ask is whether the belief states in CBE are merely redundant. In other words, does the ‘evidential juice’ possessed in CBE reside fundamentally and exclusively in the ultimate evidence portion of CBE? Or do some of the CBE beliefs contribute additional evidential juice beyond what the ultimate, experiential evidence provides? If we ask what attitude S should adopt toward a given hypothesis H at t, is it necessary to consider S’s beliefs in addition to S’s experiences? Or is the ﬁttingness or non-ﬁttingness of a possible doxastic attitude toward H determined by the experiential portion alone of the CBE? I interpret Conee and Feldman to mean that the experiential portion of CBE sufﬁces to determine whether the ﬁttingness relation holds or does not hold between any hypotheses and the subject’s evidence. Call this the sufﬁciency of current experience thesis (SCE thesis). The SCE thesis appears to be what they mean in saying that experience is the ultimate evidence. Let us ask, then, whether the SCE thesis is defensible, and whether it helps constitute an adequate theory of justiﬁedness. Does the experiential portion of CBE exhaust CBE’s evidential power? No, I respond. Many beliefs in CBE can add extra, independently derived evidential weight to the total evidence beyond the portion of evidence arising from the experiential portion of CBE. Why? Because, in a great many cases, justiﬁed beliefs held at time t were originally acquired earlier, as a result of previous experiences. The source of the evidential power of these beliefs, therefore, may partly reside in those previous experiences, not in CBE experiences. Although the evidential potency of these beliefs may derive from experience, as Conee and Feldman contend, this does not entail that it all derives from CBE experience. Thus, the doxastic portion of CBE need not be a merely redundant portion of CBE. Its justiﬁcational juice may derive from experiences other than CBE experiences. Conee and Feldman might reply that the mere fact that a CBE belief (justiﬁedly) originated in past experiences does not imply that it still possesses evidential power at t. It retains such power only if, at t, S remembers either the originating experiences or intervening experiences that support the target belief. However, a memory at t of earlier experiences is itself an experience that occurs at t, and hence belongs to CBE. So the CBE belief acquires whatever evidential power it has from a CBE experience, just as the SCE thesis maintains. In the cases in question, though, the pertinent CBE experience is a memory experience. This reply is unpersuasive. People frequently fail to recall their original sources of evidence for things they know or justiﬁably believe. First, they often don’t have memory-based beliefs about their original sources. More to the present point, they rarely have memory experiences of speciﬁc perceptual episodes (what psychologists call

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

267

‘episodic’ memories) in which they were exposed to relevant observational evidence. It is extremely implausible to claim that for every moment at which a justiﬁed belief is stored in your mind, you undergo an episodic memory experience of one or more past (perceptual) experiences that constituted your evidence. Five minutes ago I had the stored, dispositional belief that my social security number is such-and-such. Did I also have, at the same time, a memory experience of earlier perceptual experiences that provided evidence for this belief? This is dubious. It is especially dubious if ‘experiences’ are restricted to conscious mental events. I simply had no such conscious memories ﬁve minutes ago. What about unconscious memory experiences of the relevant kind? This is what Conee and Feldman would need to invoke to shore up the currently hypothesized line of defense. But what evidence could they adduce for the psychological thesis that whenever a person has a stored factual belief, it is accompanied by memories of speciﬁc perceptual observations of the required sort? Even if such psychological evidence were forthcoming, are such episodic memories necessary for the original belief’s justiﬁedness? A psychologically more plausible view is that many beliefs, including my social-security-number belief, are retained via preservative memory from earlier stages of the same belief.12 An epistemologically more plausible view is that preservative memory of this kind sufﬁces (absent defeaters) to transmit justiﬁedness from the earlier period. The epistemic role of preservative memory also explains how the evidential status of a belief held at t can partly derive from earlier experiences without positing event memories, at t, of such earlier experiences. Thus, the SCE thesis is not well supported.

16.6 Internalism and the historicity of justiﬁedness The failure of the SCE thesis and the importance of preservative memory in a satisfactory account of justiﬁedness pose a serious problem for evidentialism, especially the internalist element in evidentialism. This is so for two reasons. First, the failure of the SCE thesis and the epistemic signiﬁcance of preservative memory demonstrate the general historicity of justiﬁedness (see Goldman 2009 for more detail). Although there may be some cases in which a doxastic attitude’s justiﬁcational status is wholly determined by events occurring at its own doxastic decision time, t, in general an attitude’s justiﬁcational status is partly determined by events and states before t. Hence, an attitude’s justiﬁcational status does not supervene on mental states occurring at t. This is incompatible with Feldman and Conee’s evidentialism. According to their principle EJ, it will be recalled, an attitude’s justiﬁcational status is said to depend on the evidence the subject has at t. This is too restrictive. Earlier evidence is also relevant to justiﬁedness.

12 This is the kind of ‘factual’ memory that psychologists refer to under the label ‘semantic’ memory (in contrast to ‘episodic’ memory).

268

NEW SYNTHESES

It is not just earlier evidence that is relevant. Earlier processes leading to the belief can also be relevant. Preservative memory is one such process. Suppose S originally formed a justiﬁed belief that P based on perceptual experiences of his between to and t1. At a later time, tn, S also believes that p. Is S’s belief that p at tn also justiﬁed? That depends on how S arrived at this belief state at tn. Suppose that the tn-belief in p is the product of preservative memory operating from time t1, and S has encountered no defeating evidence of P during the interim. Then his p-belief at tn is presumably justiﬁed. By contrast, suppose S totally forgets p at some point after t1, but comes to believe it once again at tn for entirely frivolous, or spurious, reasons. Then S’s p-belief at tn is unjustiﬁed. Thus, which causal process leads to the (dated) belief is crucially relevant. Evidentialism seems to ignore causal processes. In discussing the evidence that a doxastic attitude must ﬁt in order to be justiﬁed, Conee and Feldman (especially in Conee and Feldman 2008) talk only about beliefs and experiences. Moreover, the evidential status of beliefs is said to derive entirely from that of experiences. Ultimately, then, only experience provides evidence; and it is ﬁttingness with evidence, and only evidence, that confers justiﬁedness. It appears, then, that processes are left out of the picture. This is unacceptable, as the case of preservative memory illustrates. In principle, an evidentialist might try to evade this problem by declaring preservative memory itself a species of evidence. But this would be a very odd use of the term ‘evidence’. Moreover, Conee and Feldman show no signs of wanting to go down this path. Their examples of items (or states) of evidence do not incorporate any such phenomena. Conee and Feldman might reply that the role of preservative memory is already covered by the basing relation that they admit as relevant in their account of wellfoundedness (or doxastic justiﬁedness). Perhaps, but this is problematic. Clause (ii) of their analysis of well-foundedness speaks of S having doxastic attitude D toward p on the ‘basis of ’ evidence e. Does it make sense to view preservative memory as a species of evidence on the basis of which one might have doxastic attitudes? The things that serve as evidence, under Conee and Feldman’s construal, seem to be states or events with propositional contents. But preservative memory is not a state or an event, although the things it preserves or retains are states or events (or at least the contents of such states or events). There are two other problems facing Feldman and Conee’s supervenience thesis (namely, that an attitude’s justiﬁcational status at t supervenes on the subject’s mental states at t). Some properties that do not seem to be purely mental also have to ﬁgure in the supervenience base of an attitude’s justiﬁedness. First, ﬁttingness with one’s total set of beliefs (at t) is not exactly what determines justiﬁedness, even waiving the problem of prior acquisition. Only ﬁttingness with justiﬁed beliefs (and other doxastic states) should determine the justiﬁedness of a further attitude. Suppose that believing q is a ‘proper’ inferential response to the total set of beliefs B*. But suppose that no members of B* are justiﬁed. All of the beliefs are just products of wishful thinking, or other unsound methods of belief formation. Then a belief in q (on the basis of B*) does not constitute a justiﬁed belief. Justiﬁed beliefs cannot be derived from unjustiﬁed beliefs.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

269

This would not be a problem for the mentalist supervenience thesis if the justiﬁedness property of a belief held at t supervened on mental states obtaining at t. But we have already seen that this is not so. The justiﬁcational status of an attitude held at t supervenes on what happens both at t and prior to t. A second problem is the ﬁttingness relation itself. Fittingness is not a mental state of affairs. According to evidentialism, given the composition of a particular CBE at time t, believing H either ﬁts or does not ﬁt the CBE. Whether the ﬁttingness relation is instantiated by the ordered pair (B(H), CBE) does not depend on whether the subject (or anybody else) thinks it is instantiated. Thus, the justiﬁcational status of the subject’s believing H depends on an instantiation fact (as it is called in Goldman 2009), which is not a mentalistic fact. Here is yet another problem for the mentalistic supervenience thesis about justiﬁcation. Another lurking problem is one for internalism more generally. Fittingness is a seemingly attractive relation for internalism because it seems to be an internal relation, hence a welcome tool or resource for this type of approach. Consider what the internalist Chisholm wrote about the nature of justiﬁcation: ‘the 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’ (1989: 7). The ﬁttingness relation looks like it would ﬁnd a comfortable home in Chisholm’s internalist conception. Surely, one can tell by reﬂection alone (ostensibly an a priori process) whether a given attitude toward hypothesis H does or does not ﬁt with one’s current CBE. After all, can’t one tell by reﬂection what one’s current CBE is, and can’t one tell by reﬂection whether the selected attitude toward H does or does not ﬁt one’s CBE? Unfortunately, this simple picture is incorrect. One cannot tell by reﬂection (if reﬂection is supposed to be an a priori process) what one’s current mental states are. One can only determine the composition of one’s total mental state by introspection, which is not an a priori process (see Goldman 1999b). (However, perhaps Chisholm meant to include introspection under reﬂection.) Second, and more importantly, one’s CBE—one’s total body of evidential states—is not identical with one’s total body of mental states. The total body of evidential states includes only one’s justiﬁed beliefs, not the unjustiﬁed ones. Although one can tell by introspection which beliefs one has (at least conscious beliefs—stored beliefs presumably are not introspectible), one cannot tell by introspection which ones are justiﬁed. The justiﬁcational status of one’s beliefs, as we have seen, partly depends on their history of acquisition and conﬁrmation (prior to t). None of this is accessible to introspection at time t. If one cannot tell by introspection (at t) which of one’s beliefs are evidential states, then one cannot tell by reﬂection (at t) whether the ﬁttingness relation holds between one’s CBE and the selected attitude in question. So, instantiations of the ﬁttingness relation are not generally detectable (at the time of attitude selection) by reﬂection alone. It emerges from these points that the ﬁttingness relation should hold less appeal to internalists than it initially appeared. This may not matter to Feldman and Conee, who do not accept Chisholm’s accessibilism. For accessibility internalists, however, the

270

NEW SYNTHESES

above discussion is quite pertinent, insofar as it suggests that the ﬁttingness relation is a less promising piece of armor for defending their cause than they might imagine. This point is independent of the many complications that would be encountered in trying to work out a determinate ﬁttingness relation that meets Feldman and Conee’s desiderata—an enormous problem given the special difﬁculty of balancing the respective roles for current experience and current beliefs. However, since I am tentatively prepared to sign on to the notion of a ﬁttingness relation (as indicated in Section 16.4), I won’t say that it cannot be done. The question I have mainly addressed in this section is whether a well-speciﬁed ﬁttingness relation would be such a major victory for internalism. I have argued in the negative.

16.7 Experiential evidence: toward a two-component theory As indicated at the outset, one aim of this chapter is to explore the prospects for a hybrid theory of justiﬁcation that combines the strengths of reliabilism and evidentialism. Thus far, however, rather limited steps have been taken toward embracing evidentialist themes. The main such step (in Section 16.4) was to advocate a two-factor theory of inferential justiﬁcation, in which one factor is a belief’s ﬁttingness relation to the subject’s doxastic evidence. Should reliabilism go a step further and acknowledge an analogous role for ﬁttingness with non-doxastic evidence, for cases of non-inferential justiﬁcation? This is what Comesana recommends in his proposed synthesis. He motivates this move in response to the oft-cited counterexamples to reliabilism offered by BonJour (1985) and Lehrer (1990). The examples of Norman the clairvoyant (BonJour) and Mr Truetemp (Lehrer) purport to demonstrate the non-sufﬁciency of reliability for justiﬁedness. Comesana’s diagnosis of these cases is that the subject lacks any evidence—especially experiential evidence—on which his belief is based. Nonetheless, the beliefs in these cases are all reliably caused. So reliabilism must classify them as justiﬁed—a mistaken classiﬁcation according to most commentators. Comesana therefore proposes a strengthened sufﬁciency condition according to which a belief is justiﬁed if it is caused by a reliable process that includes some evidence. We earlier acknowledged a need for doxastic evidence in the case of inferential justiﬁcation. Why not join Comesana in taking the further step of requiring non-doxastic evidence, that is, experiential evidence, for non-inferential justiﬁedness? This would handle the clairvoyance and Truetemp cases and would add more heft to the proposed synthesis of reliabilism and evidentialism. An intriguing counter-argument against this concession to evidentialism is made by Jack Lyons (2009). Lyons argues that experiential grounds (or evidence) play no essential role in the structure of justiﬁcation. In particular, he holds that there can be—and there are—immediately justiﬁed (‘basic’) perceptual beliefs that do not get their justiﬁcation from perceptual experiences that serve as grounds. Instead, these

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

271

perceptual beliefs get their justiﬁedness from their being (A) basic and (B) the products of reliable processes. He defends this position by appeal to both possible and actual cases of justiﬁed perceptual belief without consciously experienced grounds. One hypothetical case he adduces is zombies,13 who have justiﬁed perceptual beliefs in the absence of any conscious perceptual experience. Another hypothetical case is Block’s (1995) super-blindsighter, who has visual beliefs without visual experiences. An actual case is one of J. J. Gibson’s (1966) examples of ‘sensationless perception’. This example involves blind people who possess an obstacle sense: they detect obstacles like walls and chairs without having any associated conscious sensations. They tend to think they are picking up information somehow through the skin of the face, when in truth the information is coming through the ears as a subtle form of echolocation (Lyons 2009: 52). Lyons regards the perceptual beliefs associated with this obstacle sense as justiﬁed.14 Lyons also offers an account of basic beliefs that requires them to be outputs of a certain type of modular system—what he calls a ‘primal system’—but makes no requirement concerning an experiential ground. His way of handling the Norman and Truetemp cases is quite different from Comesana’s. He says that their beliefs are not basic because basic beliefs must be the products of a primal system and no such system would underpin Norman’s clairvoyance or Truetemp’s temperature detection. Under Lyons’s theory, reliability sufﬁces for the justiﬁedness of basic beliefs; but Norman’s and Truetemp’s beliefs do not qualify as basic. I shall not try to decide if Lyons’s alternative strategy ultimately succeeds. However, even if we agree with Lyons that experience is inessential for basic perceptual justiﬁedness and that a different explanation of the clairvoyance and Truetemp cases is possible (without abandoning reliabilism), this does not demonstrate that all is well with reliabilism’s treatment of experience. Few epistemologists would deny that people have both perceptual and memorial experiences, and it is reasonable to expect such experiences to play a distinctive role in justiﬁcation. The one traditional epistemology that ignores or downplays the evidential role of experience is coherentism, and its standing in epistemology takes a hit precisely because of this feature. Reliabilism shows no comparable opposition to a role for experience, but it does little to highlight or acknowledge such a role. Its silence about experiential evidence is, at a minimum, a noticeable lacuna. Here is a speciﬁc argument to motivate a substantive role for experiential states in a theory of justiﬁcation, an argument from defeaters. Sidney inferred it would be sunny

Lyons deﬁnes ‘zombies’ in a weaker sense than usual in philosophy, as beings that are as psychologically similar to one of us as possible consistent with their lacking conscious experiences (2009: 51–2). 14 Lyons provides a number of other examples drawn from perceptual psychology. The reader should consult his book for detailed exposition and interpretation of these examples (2009: 52–9). In addition to perceptual cases where there is room for doubt that justiﬁedness depends on experiential evidence or grounds, I would adduce the case of introspective beliefs. I can be justiﬁed in believing that I am now in this or that mental state, but there is no additional mental state (distinct from the target state itself) that serves as ground or evidence for such a belief. 13

272

NEW SYNTHESES

this afternoon from what he read in this morning’s newspaper. On the basis of this evidence, he continues to believe an updated version of this proposition—namely, that it is sunny right now (in the middle of the afternoon)—despite the fact that he is walking in the middle of a rainstorm. Surely his current perceptual experience is a defeater for this belief; he is not justiﬁed in believing that it is sunny right now. But he has used no belief-forming or belief-revising process that takes this perceptual experience as an input; he just ignores this experience. It appears, then, that there is no process we can appeal to—at least no instantiated process—to account for the defeat of his sunniness belief. If we want to say what defeats Sidney’s current justiﬁcation for his sunniness belief, the obvious candidate is his perceptual experience. Moreover, it is natural to say that this justiﬁcation-undermining experience is a piece of evidence he possesses. 15 In Section 16.4 I offered a two-factor approach to inferential justiﬁcation in which doxastic evidence occupied a central role. A similar move is in order with respect to experiential evidence, in light of the foregoing argument from defeat (among other things). So, an analogous two-component approach to non-doxastic justiﬁcation (that is, justiﬁcation that rests on non-doxastic states) is in order, in which perceptual and memorial experience would occupy a central role. Let us concentrate on perceptual experience. To construct a two-factor theory of perceptual justiﬁcation we need an account of perceptual ‘ﬁttingness’ to provide the ﬁrst factor. A preliminary question is how to construe perceptual states. Presumably they are not doxastic states; but might they still be contentful, or representational, states? To remain in the vicinity of Feldman’s (2003) approach to perceptual justiﬁcation, we would have to interpret them as being representational, or contentful, states. Otherwise, there could be no closeness of content between a perceptual state and beliefs that it justiﬁes (or helps to justify). If perceptual states have content, of course, this might scare up the ghost of the Sellarsian dilemma. For justiﬁcation to derive from a contentful state, doesn’t that state itself have to be justiﬁed? And if it must be justiﬁed, how does it attain this status? This initial problem is not so daunting. Both philosophers of mind and cognitive scientists commonly treat perceptual states as representational states. This does not mean that they are doxastic (belief-like) states, and if they are not doxastic states they do not need to be justiﬁed themselves in order to confer justiﬁedness on other states. But how can they confer justiﬁedness if they don’t possess it themselves? As evidentialism says, they can confer justiﬁedness without being justiﬁed because they are evidence. In virtue of what are they evidence? As proposed in Section 16.1, we can provide an 15 In my earlier formulations of (pure) reliabilism (Goldman 1979, 1986), I tried to accommodate the problem of defeat or undermining by invoking reliable processes that could or should have been applied to the undermining mental states, processes which, if used, would have prevented the relevant belief from being formed. But these attempts to accommodate defeat in terms of the process framework were rather strained, I would now say. They were not straightforwardly inadequate, but the treatment can deﬁnitely be improved upon by incorporating evidential states explicitly into the theory, as I now propose.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

273

answer congenial to reliabilism: they are evidence (or evidential states) because they are (commonly) reliable indicators of the subject’s environment. Perceptual states are produced by systematic relations between properties of the environment, properties of a perceptual medium (light waves, sound waves, etc.), and perceptual systems that pick up information from these sources and transmit it to the brain. In favorable circumstances, the resulting conscious states are systematically correlated with tokenings of relevant properties of (objects in) the environment, and hence are evidence for such tokenings. Granted that these states are evidence (hence capable of conferring justiﬁcation), how can beliefs about the environment ﬁt this evidence? A feature of the ecumenical account of inferential justiﬁcation proposed in Section 16.4 was acceptance of ﬁttingess as one component of the story. Is the ﬁttingness relation applicable here as well, in the perceptual case? As remarked above, if the approach considered here is to mesh with the one sketched in Feldman (2003), it would have to involve some sort of content matching, content overlap, or content ‘implication’ between perceptual states and beliefs (or other doxastic states). How good are the prospects for such a contentful relationship? It is widely held that perceptual states have a different species of content than doxastic states do: nonconceptual rather than conceptual content. Is this correct, and would it preclude serious prospects for a suitable contentful relationship between the two kinds of states? There are many different interpretations of the distinction between conceptual and nonconceptual content. It is not clear, however, that any of them deﬁnitely implies that these are two distinct kinds of content. Jeff Speaks (2005) offers a general critique of this conclusion. Here is how he argues, for example, against just one defense of this idea. Many writers appeal to the richness, or ﬁne-grained character, of experience to show that the contents of experience are nonconceptual. A thousand words, it may be said, would hardly begin to convey a complete description of how the world appears to you in a brief perceptual state. Speaks responds that even if experience is far more detailed and full of information than could be captured in a single thought, or even a lifetime of thoughts, this does not show that the information given in perception is of a different kind than the information represented by a belief. It only shows, at most, that there is more of it. Without resolving this issue, let us turn to the question of ﬁttingness. How could a doxastic attitude toward a proposition ‘ﬁt’ an experiential state, assuming that both have content (whether the same or different kinds of content)? In the case of inferential ﬁttingness, we proposed that the relevant relation is that the truth of the contents of the premises should make probable the truth of the content of the conclusion. Perhaps the same idea can work here. Perhaps we can say that a belief in proposition p ﬁts the evidence provided by experiential state E just in case the truth of the content of E makes (highly) probable the truth of p. We can illustrate this idea with the help of a particular proposal in vision science for how beliefs about visible objects are formed based on visual representations of

274

NEW SYNTHESES

shape.16 According to this proposal, the visual system represents object shapes in terms of parts constructed out of geometrical ‘ions’ (elements), called geons. Geons are members of a family of representations of volumetric shapes that can be modeled as generalized cones, that is, volumes swept out by a cross-section moving along an axis. A typical geon is a cylinder. Groups of geons can be combined by relating several of them to one another in ways familiar from tinkertoy sets. A combination of geons plus selected relations among them (e.g. ‘side-attached’) can be recognized as familiar kinds of objects, such as a chair, a giraffe, a mushroom, or a pail. Assume now that geonic combinations are a type of (high-level) visual representation, and that, in normal circumstances, a speciﬁc geonic combination G is tokened only when a giraffe is present. Then the truth of a G-token’s content makes it highly probable that the proposition ‘A giraffe is in the vicinity’ is true. Hence, it is ﬁtting for a subject to believe the latter proposition when a G-representation is tokened. This, then, is an adaptation of the evidentialist notion of ﬁttingness to the case of visual experiential evidence, thus making ﬁttingness one suitable factor, or component, of experience-based justiﬁcation. If we stop here, of course, we have a purely evidentialist account of experiential justiﬁcation. It doesn’t resemble an evidentialist-reliabilist hybrid of the sort I have been touting. Must we add a process-reliability component to obtain a more satisfactory account? Yes. Otherwise, we won’t cover all the necessary bases. In particular, we won’t have a satisfactory account of doxastic as opposed to propositional justiﬁedness (i.e. well-foundedness) in this domain. Obviously, to have a justiﬁed belief that object X is an F, where the justiﬁcation arises from shape perception, it is not enough that a geonic F-representation is tokened in the subject’s head while he forms the indicated belief. This condition might be met even by the novice bird-watcher in Feldman’s example (see Section 16.2 above) who should not be credited with a justiﬁed belief. The novice bird-watcher might even undergo a tokening of the same geonic representation of the pink-spotted ﬂy-catcher as the expert bird-watcher undergoes. In that case, the ﬁttingness condition will be satisﬁed by both of them if it is satisﬁed by one. But, intuitively, it is still possible that one should be justiﬁed and the other unjustiﬁed (as Feldman recognizes).17 Biederman (1987, 1990), the leading proponent of the geonic approach to visual object-recognition, gives us the materials for explaining the difference between the two bird-watchers in terms of the geon theory. Biederman says that subjects construct

16

Not all vision scientists would sign on to this approach, but it has received considerable support and development, especially by Irving Biederman (1987, 1990), and it builds on the very popular ideas of David Marr (1980). However, it is used here mainly because of its vividness, not because of its empirical adequacy as presently judged by the research community. 17 The bird-watching example makes for a poor ﬁt with the geon theory of object recognition, for the obvious reason that colors of body parts constitute a major feature of bird identiﬁcation, and colors are not included in the shape-oriented geon theory. For present purposes, however, let us pretend that all of the relevant features involved in bird identiﬁcation are shape features.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

275

‘object-models’ for a large number of object names in English, where each ‘objectmodel’ is a particular geonic conﬁguration of speciﬁc geons and relations between them. Many object types require several models each, both because different models are required for different viewing orientations and because some names or concepts have more than one conﬁguration associated with them. For example, lamps come in different conﬁgurations, and there are several types of pianos (grand, baby-grand, upright, and spinet) each of which has a distinctive geonic conﬁguration. Biederman hypothesizes that people store (geonic) object-models in memory where they are linked to object names or concepts. Then, when they have a visual experience, they automatically try to match a geonic conﬁguration tokened in visual consciousness with one of the geonic conﬁgurations stored in memory. When such a ‘match’ is secured, the system recognizes (or judges) the perceived object as an F, where F is the label or concept paired with the recovered object-model.18 Presumably, this matching process would be the kind of process used by the expert bird-watcher, who has a stored geonic object-model for the pink-spotted ﬂy-catcher. If his classiﬁcation of the bird as a pink-spotted ﬂy-catcher is the product of a matching process of the indicated kind—more speciﬁcally, a matching process with a high threshold or standard for a match (see note 18)—he uses a very reliable process of classiﬁcation, or belief-formation. So this would satisfy a process-reliabilist condition of the kind I mean to incorporate into our hybrid account. The novice bird-watcher, by contrast, would not be using such a reliable process. His lack of training suggests that he has not constructed a suitable object-model for the pink-spotted ﬂy-catcher, and therefore does not secure much of a ‘match’ at all between his visual experience and a stored object-model of the ﬂy-catcher. Hence, he is just guessing when he classiﬁes the bird as a ﬂy-catcher, and guessing is not a reliable process. I propose, then, in parallel with the proposal of Section 16.4, that experiential justiﬁedness is a function of two factors, or components. One factor says that belief in a proposition is prima facie justiﬁed in experiential terms only if the belief ﬁts with the subject’s current experiential evidence.19 The second factor says that an experiencebased belief is justiﬁed only if it is the product of a reliable experience-based process. The two-factor theory says that an experiential belief is fully justiﬁed (doxastic considerations

18 As used here, the term ‘match’ does not refer to a perfect correspondence between all parts of the visually experienced geonic conﬁguration and the complete object model stored in memory. Perfect correspondence may be atypical in perceptual experience because not all parts are usually in view, either because of the object’s orientation to the viewer or because of occluding objects. ‘Match’ here refers to a correspondence with respect to enough elements of the two geonic conﬁgurations to meet the threshold, or criterion, of matching that the subject’s psychological system employs on the occasion in question. This threshold can vary from occasion to occasion. Of course, threshold variation can affect the reliability of the matching process upwards or downwards. But Biederman argues, on theoretical grounds, that even a fraction of the elements of a full geonic conﬁguration often sufﬁces to obtain a reliable identiﬁcation of an object. 19 I say ‘prima facie’ justiﬁed because one might have non-experiential evidence from background beliefs that also affects one’s ultima facie evidence for believing the proposition. The problem of how to balance conﬂicts between experiential and non-experiential evidence is not on the present agenda.

276

NEW SYNTHESES

aside) only if it is justiﬁed in terms of both factors. This two-factor approach to experience-based justiﬁedness, which marries the requirement of evidential ﬁttingness with production by a reliable process, has signiﬁcant attractions, I have argued.

16.8 The source of epistemic principles (best explanationism) In their recent paper ‘Evidence,’ Conee and Feldman (2008) rightly criticize Chisholm’s approach to epistemic principles. Chisholm denies that epistemic principles derive from any more fundamental source, and he does not seek a uniﬁed account of them. By contrast, Conee and Feldman contend that if perceptual and memorial experiences are justifying, there must be something about them that makes this the case. And if religious experience is similarly justifying, this must be so for the same underlying reason. Here we are in complete agreement. Conee and Feldman are pointing to what I call (Goldman 2009) a criterion or rationale of justiﬁcational rightness. An adequate theory of justiﬁedness should specify the content of this criterion or rationale. I endorse the externalist idea that what rationalizes various epistemic rules or principles has to do with the proﬁciency or efﬁcacy of the principles at meeting such goals as acquiring true belief and avoiding error. A generalization of this idea for degrees of belief is offered in Goldman (1999c) under the label ‘veritistic value’ (although that theory is not directed at justiﬁedness per se). Conee and Feldman, of course, seek an internalist criterion or rationale. The one they propose in ‘Evidence’ is the criterion of ‘best explanations.’ Here is how they put it: We believe that the fundamental epistemic principles are principles of best explanation. Perceptual experiences can contribute toward the justiﬁcation of propositions about the world when the propositions are part of the best explanation of those experiences that is available to the person. Similarly, the truth of the contents of a memory experience may be part of the best explanation of the experience itself. Thus, the general idea is that a person has a set of experiences, including perceptual experiences, memorial experiences, and so on. What is justiﬁed for the person includes propositions that are part of the best explanation of those experiences available to the person. Likewise, one’s inferences justify by identifying to one further propositions that either require inclusion in one’s best explanation for it to retain its quality or enhance the explanation to some extent by their inclusion. (Conee and Feldman 2008: 98)

Unfortunately, there are a number of counter-examples to this approach, plus the worry that best explanation is itself subservient to a different rationale, namely true belief acquisition. I turn ﬁrst to the counter-examples. It should be emphasized that the success of even a single counter-example among the three offered below would sufﬁce to make my critical case. All the counterexamples would have to be successfully resisted to overcome this line of criticism.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

277

1. Introspection. How would Conee and Feldman subsume a principle of justiﬁedness for ﬁrst-person-current-mental-state beliefs under the explanatory inference model? The most promising tack is to posit an introspective principle of evidence analogous to a visual principle of evidence. In the latter case, a principle might say: ‘Having a visual experience V(E) is prima facie evidence for E’s being the case.’ This can subsumed under the best explanation mode by indicating that having, say, a visual experience as of a mouse running across the ﬂoor is best explained by its being the case that a mouse is running across the ﬂoor. Similarly, it might be proposed, what gives me prima facie evidence that I am now undergoing a tickle sensation is that I am having an introspective experience as of a tickle sensation. This would be subsumed under the best explanation theme by indicating that having an introspective experience as of a tickle sensation is best explained by its being the case that I do have a tickle sensation. The viability of this approach depends, however, on there being introspective experiences of mental events or states of affairs distinct from the mental states that are allegedly introspected. Is there an introspective experience of a tickle sensation in addition to and distinct from the tickle sensation itself? This is extremely doubtful. I don’t doubt that there are tickle experiences. And I don’t doubt that there are (self-reﬂective) beliefs that one is having a tickle experience. What I doubt is that there is any ‘intermediate’ mental event that stands between the ‘objective’ state of affairs in question (here, the tickle sensation) and the belief in this state of affairs in the same way that a visual experience of a mouse is an ‘intermediate’ event that stands between the mouse scurrying across the ﬂoor and the belief that a mouse is scurrying across the ﬂoor. If there is no such intermediate introspective experience, then what is the explanandum (item to be explained) that the occurrence of the tickle sensation (the explanans) supposedly explains? I deny that there is anything that can function in the guise of a suitable explanandum, thereby casting introspective justiﬁcation in the best-explanationist mold. 2. Preservative Memory. As reviewed above, a satisfactory theory of justiﬁedness needs a principle of preservative memory: If S was justiﬁed in believing p earlier, and S retains her belief in p now via preservative memory, then S is prima facie justiﬁed in believing p now. Like introspection, however, preservative memory lacks any type of mental experience or episode that invites explanation. In particular, there is no (conscious) act of ‘recollection’ that invites explanation. An epistemic principle that covers preservative memory, then, cannot be rationalized by the best-explanation approach. 3. Arithmetic Inference. I think there are two squirrels on my deck, and I think there are two birds. So I infer that there are (at least) four animals. Presumably, this arithmetic inference is justiﬁed. Is it a case of explanatory inference? Surely not. How does there being four animals explain there being two squirrels and two birds? It doesn’t.

278

NEW SYNTHESES

Still, here is a justiﬁed belief that some epistemic principle must cover. But that principle, in turn, cannot be grounded in terms of best explanation. In each of these cases, the best-explanation account does not work. (Or, more cautiously, it does not work in some of them.) At the same time, each is very plausibly accommodated by process reliabilism. The belief that one has a tickle sensation is justiﬁed because introspection is a reliable process. Preservative memory contributes to justiﬁcation because preservative memory is a conditionally reliable process.20 Arithmetic inference (of the right sort) is justifying because it is a conditionally reliable process. Thus, process reliabilism does a better job of accounting for the justiﬁedness ‘data’ than best-explanationism. In addition to these counter-examples, we should reﬂect on why inference-to-thebest-explanation is often a good pattern of non-deductive inference. I submit that it is a good pattern (where it is) because it is conducive to true belief. So, even if the inference-to-the-best-explanation rationale were accepted, there is still a deeper rationale of an externalist variety. In defending this thesis, I don’t wish to appeal to the argument that no explanation genuinely explains anything unless it is true. That is doubtless correct in one sense of ‘explain’. However, people like Conee and Feldman who invoke inference-to-the-best-explanation must mean ‘explain’ to be understood in some non-truth-entailing sense of ‘explain’ (whatever that sense is, exactly). As dyed-in-the-wool internalists, they would not want a fundamental rationale of theirs to appeal directly to truth considerations. Suppose, then, that ‘superior explanatoriness’ is not deﬁned in terms of truth-conduciveness. Nonetheless, I submit, superior explanatoriness strikes us as a rationalizing property of a type of inference only because superior explanatoriness is, in general, an excellent indicator of truth. In other words, even if we accept superior explanatoriness as a mark (only one mark, not a universal mark) of justiﬁcation, its being such a mark derives from its correlation with truthconduciveness. The most fundamental principle of epistemic justiﬁcation, then, is truth conduciveness—as reliabilism, of course, maintains.21

References Biederman, I. 1987. ‘Recognition-by-Components: A Theory of Human Image Understanding,’ Psychological Review 94: 115–47. ——. 1990. ‘Higher-Level Vision,’ in D. Osherson, S. M. Kosslyn, and J. M. Hollerbach (eds), An Invitation to Cognitive Science, vol. 2, Visual Cognition and Action, pp. 41–72. Cambridge, MA: MIT Press. 20 As deﬁned in Goldman (1979), a belief-dependent process is conditionally reliable just in case it yields a high ratio of output truths in cases when all of its belief inputs are true. 21 I have received valuable comments from Holly Smith, from Richard Feldman, and other discussants at the 2008 Rochester graduate epistemology conference, and from audiences at other venues where earlier versions of this paper were presented, i.e., Brown University, Union College, and the Metropolitan University of Mexico City.

T OWA R D A S Y N T H E S I S O F R E L I A B I L I S M A N D E V I D E N T I A L I S M ?

279

Block, N. (1995). ‘On a Confusion about a Function of Consciousness,’ Behavioral and Brain Sciences 18: 227–47. BonJour, L. 1985. The Structure of Empirical Knowledge. Cambridge, MA: Harvard University Press. Chisholm, R. M. 1942. ‘The Problem of the Speckled Hen,’ Mind, 51: 368–73. ——. 1989. Theory of Knowledge, 3rd edn. Upper Saddle River, NJ: Prentice-Hall. Comesana, J. 2002. ‘The Diagonal and the Demon,’ Philosophical Studies, 110: 249–66. ——. 2010. ‘Evidentialist Reliabilism,’ Noûs 94: 571–601. Conee, E. 1988. ‘The Basic Nature of Epistemic Justiﬁcation,’ The Monist. Reprinted in Conee and Feldman, Evidentialism (2004). —— and Feldman, R. (1998). ‘The Generality Problem for Reliabilism,’ Philosophical Studies 89: 1–29. —— and Feldman, R. 2001. ‘Internalism Defended,’ in H. Kornblith (ed.), Internalism and Externalism, pp. 231–60. Malden, MA: Blackwell. Reprinted in Conee and Feldman, Evidentialism (2004). —— and R. Feldman. 2004. Evidentialism: Essays in Epistemology. Oxford: Oxford University Press. —— and R. Feldman. 2008. ‘Evidence,’ in Q. Smith (ed.), Epistemology: New Essays. Oxford: Oxford University Press. Feldman, R. 1985. ‘Reliability and Justiﬁcation,’ The Monist, 68: 159–74. ——. 2003. Epistemology. Upper Saddle River, NJ: Prentice-Hall. —— and E. Conee. 1985. ‘Evidentialism,’ Philosophical Studies 48: 15–34. Reprinted with an afterward in Conee and Feldman, Evidentialism (2004). Gibson, J. J. 1966. The Senses Considered as Perceptual Systems. Boston: Houghton Mifﬂin. Goldman, A. I. 1979. ‘What Is Justiﬁed Belief?’ in G. Pappas (ed.), Justiﬁcation and Knowledge. Dordrecht: Reidel. ——. 1986. Epistemology and Cognition. Cambridge, MA: Harvard University Press. ——. 1999a. ‘Internalism Exposed,’ Journal of Philosophy 96: 271–93. ——. 1999b. ‘A Priori Warrant and Naturalistic Epistemology,’ in J. Tomberlin (ed.), Philosophical Perspectives, vol. 13. Cambridge, MA: Blackwell. ——. 1999c. Knowledge in a Social Wortd. Oxford: Oxford University Press. ——. 2008a. ‘Reliabilism,’ in E. N. Zalta (ed.), Stanford Encyclopedia of Philosophy (Fall 2008 Edition), URL = http://plato.stanford.edu/archives/fall2008/entries/reliabilism/. ——. 2008b. ‘Immediate Justiﬁcation and Process Reliabilism,’ in Q. Smith (ed.), Epistemology: New Essays, pp. 63–82. Oxford: Oxford University Press. ——. 2009. ‘Internalism, Externalism, and the Architecture of Justiﬁcation,’ The Journal of Philosophy 106: 309–38. Kelly, T. 2006. ‘Evidence,’ in E. N. Zalta (ed.), Stanford Encyclopedia of Philosophy (Fall 2008 Edition), URL = http://plato.stanford.edu/archives/fall2008/entries/evidence/. Kim, J. 1988. ‘What Is Naturalized Epistemology?’ in J. Tomberlin (ed.), Philosophical Perspectives 2. Atascadero, CA: Ridgeview. Lehrer, K. 1990. Theory of Knowledge. Boulder, CO: Westview Press. Lyons, J. 2009. Perception and Basic Beliefs: Zombies, Modules, and the Problem of the External World. New York: Oxford University Press.

280

NEW SYNTHESES

Marr, David 1980. Vision: A Computational Investigation into the Human Representation and Processing of Visual Information. San Francisco: W. H. Freeman. Speaks, J. 2005. ‘Is There a Problem about Nonconceptual Content?’ Philosophical Review 114: 359–98. Stalnaker, R. 1978. ‘Assertion,’ in P. Cole (ed.), Syntax and Semantics, vol. 9, Pragmatics. New York: Academic Press. Williamson, T. 2000. Knowledge and Its Limits. Oxford: Oxford University Press.

PART VIII

Replies

This page intentionally left blank

17 Replies Earl Conee and Richard Feldman

In the pages that follow, we offer brief responses to each of the essays in this volume. We are grateful to the authors for their contributions. We are equally grateful to Trent Dougherty for the hard work that he put into bringing this project to completion. We have found it challenging and rewarding to think about the numerous themes that are discussed in the essays. There are, of course, many interesting points that we do not take up here. And, as is customary, our responses focus on claims that we wish to dispute rather than on the many points with which we agree.

Response to Bergmann Michael Bergmann considers what he calls the Great Pumpkin Objection (GPO) to arguments against skepticism. Mike concludes that the GPO ‘is more worrisome for evidentialists than it is for externalists’. The GPO makes use of the fact that the things that defenders of views such as evidentialism say in response to skepticism can be mimicked by defenders of ridiculous views, such as the Great Pumpkin view that a great gift-delivering pumpkin exists and the ‘conjecturalist’ view (C) that intentionally guessing that a proposition is true provides good reason to believe that proposition. The objection concludes that the responses of the mimicked views are no better than those of the ridiculous mimicking views. Applied to evidentialism (E) as a response to skepticism, the GPO is this. 1. The E response to skepticism is mimicked by a C response to skepticism. 2. If so, then the E response is no better than a C response. 3. Therefore, the E response to skepticism is no better than a C response. Relying on conjecturalism to object to skepticism would be silly. So, we can infer from 3 that the E response is unsuccessful. The version of evidentialism that Mike considers is ‘seeming evidentialism’, the view that the fact that something seems to a person to indicate the truth of a

284

REPLIES

proposition provides evidence for the person that the proposition is true. This account of evidence, together with facts about how things seem to people, can enter into a response to skepticism. The external world seems to be various particular ways, and perhaps these appearances give good enough evidence to know external world facts. As Earl has noted, conjecturalists can say something somewhat parallel to this, claiming that guessing yields knowledge and thus developing a response of sorts to skepticism.1 Mike goes on to argue that Earl fails to provide an adequate response to the GPO. Earl does not discusses the GPO and thus he does not offer the responses to the GPO that Mike discusses. The most similar topic that Earl does discuss is a hypothetical objection to an E response to the Wheel problem. The Wheel problem asserts that starting legitimate epistemology about knowledge requires assuming epistemological doctrines. It is alleged that we must assume either a doctrine about what we know or a doctrine about how we know. The proposed E response asserts that legitimate epistemology can start by using evidence for what we know or how we know it. The evidence gives good reasons. The evidence we actually have includes evidence for the E thesis that evidence gives good reasons. An objection to this E response that Earl takes up is the closest thing to the GPO. The objection asserts that this procedure shows only that E is self-conﬁrming. It merely shows that E gives good reasons according to itself. The objection adds that a silly thesis such as C might well also be self-conﬁrming—C is self-conﬁrming for a conjecturalist who has guessed that guesses give good reasons. This objection is not the GPO. It just about self-conﬁrmation, not the comprehensive mimicry claim that is central to the GPO. Earl’s reply to this objection is similar to an adequate reply to the GPO, though. The reply claims that it is no part of the E response to the Wheel problem to assert that E is supported by being self-conﬁrming. The response rather claims that E is supported by good reason to think that it is true. We have good reason to believe E according to our best understanding of what good reasons are. We do not have good reason to think that C is true. In fact we have good reason to think that C is untrue. This same point goes strongly against premise (2) of the GPO. Assume that some theory, T2, mimics the response to skepticism of another theory, T1, by replacing each proposition that is used in giving the T1 response with a proposition that says the same thing about T2. It seems clear that conjecturalists, as well as defenders of virtually any other theory can, in this sense, mimic any E response to skepticism. So premise (1) of the GPO is clearly true. It is just as clear that (2) is untrue. The key proposition in the E response to skepticism minimally asserts that we have evidence that supports the denial of skepticism that is stronger than the evidence for skepticism. The claim is reasonable. We have plenty of perceptual experiences as though of external things with no hint of deception or error, and memories of many more. So we have available for

1

‘First Things First’, p. 20 (in Evidentialism).

REPLIES

285

our reﬂection plenty of undefeated evidence for external world propositions. This gives us better evidence for our having knowledge of external world beliefs than for our not having it. This in turn gives us inferential justiﬁcation for the denial of skepticism. The mimicking claim on behalf of C would be that we have conjectures that deny skepticism and these are stronger justiﬁcation for the denial than are conjectures for it. That claim is unreasonable. A guess that something is true is no justiﬁcation for it and we have no good reason to think that it is, whether or not some conjecturalist guesses otherwise. Mike claims that this sort of refutation of the GPO does not ﬁt well with internalist views such as evidentialism. He says that internalists object to externalist views because the externalists focus on the actual facts about what is justiﬁcation-conferring rather than on facts about what is justiﬁcation-conferring from a believer’s perspective. Yet this E response to the GPO has that same focus. The response concerns what is actually justiﬁed by a believer’s evidence rather than what is justiﬁed from the believer’s perspective. We contend in reply that all theories of justiﬁcation are theories about what actually justiﬁes, our versions of E included. Some things that are in a person’s perspective are justiﬁcation-conferring. For example, perceptual experiences help to justify some perceptual beliefs, whether the experiences are deceptive or not. But these facts about someone’s perspective are justiﬁcation-conferring because they are actually and objectively part of the evidence that the person has for the beliefs. Having evidence that is actually supporting on-balance is what confers justiﬁcation on a belief. We object to externalist views because we think that they are mistaken about what actually justiﬁes, not because we think that they are mistaken about what is justiﬁcationconferring from a believer’s perspective. What actually justiﬁes is the sole topic of E.

Response to Greco John Greco discusses something he identiﬁes as ‘k-normative status’, which is the ‘full normative status required for knowledge.’ John characterizes evidentialism about knowledge as the following view about this status: KE. The facts about k-normative status supervene on facts about evidence. John says that we ‘come closest to explicitly endorsing’ (KE). He argues that (KE) fails to account for memory knowledge and other kinds of knowledge and that (KE) is undermined by contemporary cognitive science. We do not endorse (KE). We do not have any clear idea what normative status is. Perhaps as a result of this unclarity, we are not conﬁdent that either justiﬁcation or knowledge is normative. We do hold that justiﬁcation is necessary for knowledge and that a person’s justiﬁcation supervenes on the person’s evidence. But we do not hold that other arguably necessary conditions for knowledge, such as well-founded belief, and whatever blocks the Gettier problem, supervene on the person’s evidence. If these

286

REPLIES

other conditions enter into determining the normative status required for knowledge, then we reject (KE). If justiﬁcation is the only normative element of knowledge, then we accept (KE). John seems to think that conditions on knowledge beyond justiﬁcation are normative. He points out that, ‘In the post-Gettier world, no one thinks that knowledge can be understood in terms of merely internal justiﬁcation and truth.’ He concludes from this that internalist versions of evidentialism about k-normative status are false. That would be okay with us. We’ve defended internalism only about justiﬁcation. We concur with John that there is memory knowledge. However, we emphatically deny that people lack evidence for the things that they know by memory. John dismisses the evidence we think people have, claiming that what is conscious when recalling something is counted as evidence only by a ‘stretch’. But a vivid sense of recalling some fact seems to us clearly to be some evidence for the belief, and less vivid senses of recalling are weaker evidence. Furthermore, memories are usually accompanied by corroborating factual recollections or recollective imagery. Those things are further evidence. John also questions the existence of evidence for a priori knowledge by reﬂection. However, reﬂection seems clearly to generate some sort of reason to believe the propositions. Any such reason is evidence. The nature of the evidence for a priori knowledge is not obvious. One possibility is that the evidence is the seeming truth of the proposition. Another possibility is that some a priori known propositions are literally evidence for themselves. Alternatively, the evidence might be some immediate inclination to believe the proposition, an inclination that the person has learned to be reliable by previous conﬁrmation. Finally, it may be, as one of us has argued, that the evidence is an impression of detecting something truth-making about the proposition.2 According to John, recent developments in cognitive science undermine evidentialism. He says research into blindsight indicates that it is reliable but does not involve evidence. We reply that there is a problem for evidentialism here only if blindsight yields justiﬁcation (or knowledge) without evidence. Some people who have blindsight make perceptual judgments with some remarkable accuracy, perhaps as accurate and ordinary as the judgments that yield perceptual knowledge. But they seem at ﬁrst to experience the basis for their blindsighted judgments as a guess. When this is the basis, intuitively their resulting judgments are no more knowledge than are Mr Trutemp’s guesses. If later the blindsighted learn the accuracy of these apparent guesses, then they have as evidence that learning in addition to the guesses. They may thereby have knowledge. Finally, John says that knowledge requires a ‘modally strong relation to the world.’ He seems to think that this involves some kind of reliability. We do not agree that there is something that must be reliable in every case of knowledge. Unreliable people

2

See Earl’s ‘Self-Support’ (Philosophy and Phenomenal Research, forthcoming) for a defense of the latter.

REPLIES

287

plainly know some things. Unreliable faculties can deliver knowledge when the judgment is easy enough. For instance, a person with very blurry and unreliable vision can know by vision that something moved. People can know while not exercising any reliable intellectual virtue. Various types of the belief-forming processes involved are reliable whenever there is a case of knowledge. The generality problem stands in the way of giving a speciﬁcation of the types the reliability of which is necessary for justiﬁcation. We see no reason to be optimistic that the generality problem is solvable. These are the main types of reliability that have been associated with knowledge. None seems to us to be necessary.

Response to DeRose Keith DeRose objects to the evidentialist thesis (EO) that at any time one ought to have whatever doxastic attitudes are justiﬁed by one’s evidence at that time. Keith’s discussion focuses on an example in which Henry has excellent evidence, E1, for the proposition, B, that Jones did something terrible. In addition to having this excellent evidence, E1, for B, Henry has no reason to doubt B. Henry also has some lousy evidence, E2, for B. Henry mistakenly and for no good reason thinks that E2 is good evidence for P. He does not even consider B in light of E1. Instead, Henry considers B in light of E2 and believes B on that basis. Keith considers the claim: P. Henry ought not to believe B. EO implies that P is false. Henry’s evidence supports B, so EO implies that he ought to believe B. As we see it, this implication is correct. We maintain this in spite of Keith’s claim that there is ‘certainly’ some ‘very good sense’ in which P is true. In our view, accepting B as true is the reasonable attitude for Henry to take. It is made reasonable for Henry by his having E1 and his not having any reason to doubt B. As a result, he ought to believe B. One can be misled about the credibility of P. There are other nearby facts about things that Henry ought not to do. These ‘ought not’s’ may lead some, perhaps including Keith, mistakenly to believe P. For example, consider P2. Henry ought not to ‘believe as he does’ concerning B. P2 is true, because for Henry to ‘believe as he does’ is for him to base his belief in B on E2, and E2 is bad evidence. Believing as he does is thus not ‘well-founded’, in the sense that we have deﬁned. But denying P is compatible with the claim that Henry ought not to believe anything, including B, on an ill-founded basis. Consider also P3. Henry ought not to infer B as he does. P4. Henry ought not to base his belief on E2. P5. Henry ought not to come to believe B as he does.

288

REPLIES

P3–P5 are true. They refer to things that Henry does, but he ought not do. The reasons he ought not do these things are evidentialist and fully compatible with evidentialism’s verdict on P. He ought not make bad inferences such as the one he actually makes in accepting B, he ought not base his belief on bad evidence such as the evidence on which he actually bases his belief in B, and he ought not come to believe in the way he did. Nevertheless, B is a proposition that he ought to believe. Keith argues that P is true. His premise is the fact that if we add to Henry’s situation the provision that B is true, Henry still does not know B. We respond that this absence of knowledge does not validly imply P. In any Gettier situation the person is believing what the person ought to believe. Furthermore, in Gettier situations the person has arrived at his belief as he ought. Yet although the belief is true, the person does not know the believed proposition. Keith suggests that adding to the story that Henry neglects to get evidence about B that he could easily have gotten, and should have gotten, enhances the credibility of P. We respond that these additional stipulations do not make P credible, when P is carefully understood. P is a claim about what attitude Henry should take toward B, given how things actually are. Henry’s negligence, consisting in not having gotten other readily available evidence, does not support P. The negligence does make for another potentially distracting ‘ought not’: P6. Henry ought not to have only the evidence he has about B. P6 is about a sort of practical evaluation of Henry’s evidence gathering, This is emphatically not what P is about. P is about a doxastic attitude that Henry has toward B. P is mistaken because having that attitude is correct, given the evidence that Henry actually has. In making our claim about P, we have assessed Henry’s believing B relative to the evidence he actually has. Keith notes that we also can assess belief in a proposition relative to the evidence one actually considers (E2 in his example), we can assess the belief relative to all the evidence the person has, and we can assess the belief relative to the evidence the person should have. Keith claims that it is arbitrary to select ‘all the evidence the person has’ as the one of these alternatives that determines what the person ought to believe. We ﬁnd the selection of our alternative from this list—the total actual evidence— not to be arbitrary, when ‘ought to believe’ is understood so that it is closely connected to epistemic justiﬁcation. On this reading of ‘ought to believe’, one ought not ignore any relevant evidence that one has and one is in no position to consider any evidence that one does not even have. This leaves all the evidence one does have as a nonarbitrary determinant of what one ought to believe. Shifting now to the topic of external world skepticism, Keith suggests that people lack ‘any effective evidence one way or the other’ concerning skeptical hypotheses. Let’s consider what a typical person would express by K. I am a brain in a vat having these experiences as of hands. As we see it, that people have strong evidence against propositions like K is clear. Why the evidence that people have is evidence against K is harder to explain. Each of us has

REPLIES

289

evidence of the following sort: (E3) our sensory experiences that are to us as though we are perceiving hands and our memories ﬁtting with our having hands. E3 is strong reason to deny, K, even though E3 itself is compatible with K. Beyond this compatibility, critics might note that, given suitable assumptions, people would also have this evidence, if K were true. Critics might further note that the vat hypothesis is the core of an explanation of why the experiences exist, just as is the existence of perceived hands. But none of these facts is a test of whether E3 supports not-K. To see that these sort of facts do not decide what evidence supports, it is helpful to compare the current sort of claim about perceptual evidence with what seems clear about evidential support by enumerative induction. Clearly, for the denials of the conclusions of the strongest enumerative inductive arguments from observed true premises, counterparts exist to those same facts: compatibility with the evidence, counterfactual duplication of the same evidence, and alternative explanations of the evidence. Yet it is quite intuitive that the strongest inductive arguments do support their conclusions. Similarly, our evidence does support not-K. Having argued for the existence of this support, we grant that a complete explanation of how it works would be most welcome. Keith suggests that what evidence one has about something depends on what is at issue. In particular, he suggests that the fact that one has hands may be evidence, when the issue is some everyday question, perhaps a question about how one might perform some cleaning task. But the suggestion is that the fact that one has hands is not evidence that one has when K is at issue. We respond that in the anti-skeptical evidence that we are describing, the proposition that one has hands is not serving as evidence. The sort of evidence that we have given against K does not depend on anyone having the fact one has hands as evidence. For its ultimate evidence, the argument relies only on things that skeptics do not dispute that we have as evidence—experiences and a priori reﬂections. We think that the proposition that one has hands is often had as evidence too. But this is a derivative sort of evidence that depends on having ultimate evidence such as E3 in support of it.3

Response to Pritchard Duncan Pritchard contrasts two accounts of internalism: (Ment), according to which individuals are alike with respect to justiﬁcation if they are mentally alike, and (Access), according to which individuals are alike with respect to justiﬁcation if they are alike with respect to what is accessible to them by introspection alone. We have claimed that (Ment) best captures the generic internalist idea. Duncan contends that (Access) is preferable. We continue to think that (Access) has no advantage over (Ment) for specifying internalism. Duncan suggests that a philosopher who held that inaccessible mental states contribute to justiﬁcation would not be an internalist. However, we see

3

See also ‘Response to Steup’ below.

290

REPLIES

no problem with classifying as internalist views that count mental states, however inaccessibly suppressed, as relevant to justiﬁcation. Such views might not be plausible theses about justiﬁcation. But they contrast sharply enough with paradigmatic externalist views. They do not factor into justiﬁcation contingent properties of the environment, as do reliabilism and proper functionalism. Inaccessible mental states are deﬁnitely internal to the region that is complementary to the environment, namely, the mind. In this respect, such theses are like paradigmatic internalist views. We think (Access) has troubles of its own. We stayed away from specifying internalism as some sort of accessibilism partly because ‘access’ is more technical than ‘mental’ and no stipulation of its meaning is satisfactory. (Access) uses ‘able to know by reﬂection alone’ to deﬁne ‘access’. This phrase is multiply obscure. If someone’s mind had no reﬂective capacity, would nothing be justiﬁed for the person, according to an internalism deﬁned by Access? Would knowing what perceptual states one is in, knowing what external world propositions they support, and inferring conclusions about the external world, be knowing those conclusions by reﬂection alone? If a mental state is not clearly enough manifested to someone by reﬂection for the person to know herself to be in that state, but it is nevertheless reasonably clear to her that it is present, in the way of a hazy memory, say, is that state ineligible to participate in justiﬁcation, according to (Access)? Turning to another main theme in Duncan’s paper, he explores a perspective on justiﬁcation that he calls Epistemic Disjunctivism (ED). He argues that ED provides a basis for responding to external world skepticism that is superior to our version of evidentialism. We see no advantage to ED, and some disadvantages. As we understand ED, it claims three things. First, ED asserts the metaphysical thesis that there is no shared content between a veridical perception and an indiscernible non-veridical illusion or hallucination. Second, ED makes the epistemic claim that a veridical perception more strongly justiﬁes its content factual proposition than any indiscernible non-veridical experience justiﬁes the corresponding false proposition. And third, ED claims that the presence of a veridical perception is reﬂectively accessible to the perceiver (thus allowing the perception to meet the condition on internalist views that is imposed by (Access)). One problem for ED is this. Duncan acknowledges that it appears to many who consider the matter that the BIVs who are the closest counterparts to the likes of us are as reasonable in their external world beliefs as we are in ours. Duncan offers an explanation of this appearance that is compatible with ED. His suggestion is that the appearance is explained by the BIVs being equally epistemically blameless in their false external world beliefs. To support this, Duncan notes that the BIVs are unable to distinguish their deceptive perceptual experiences from veridical perceptions like ours. This explanation is mysterious when accompanied by an afﬁrmation of ED. According to ED, the two sorts of experiences are so different that they share no content. The initially appealing account of their indistinguishability is that they are indistinguishable because they are identical in their reﬂectively accessible character. But it is quite unclear

REPLIES

291

how that could be true if they share no content. It is equally unclear why the BIVs would be our equals in epistemic blamelessness. According to ED, the BIVs get weaker justiﬁcation for their beliefs from the deceptive experiences that are reﬂectively accessible to them. Their epistemically justiﬁed strength of belief on this evidential basis is weaker than is ours from our veridical perceptions. So they seem to be epistemically blameworthy for believing as strongly as we do when our strength of belief is properly proportioned to our evidence. Another problem for ED is that it denies that we have a particular sort of rational basis for some perceptual beliefs. Suppose that Smith sees that the lights are on and Smith believes that he sees this. It seems that Smith has evidence for his belief that he sees that the lights are on. The evidence is that it visually appears to him that the lights are on. According to ED, he does not have this evidence. An appearance state with that content does not occur in veridical perception. The fact that Smith sees that the lights are on might be claimed to be Smith’s ultimate evidence. But again, intuitively, when he sees that the lights are on, this is a factual proposition for which he has something helpful as evidence. It is quite plausible that this evidence is the visual experience that the seeing shares with the most similar visual deception. It seems clear that Smith could legitimately appeal to this evidence to support his belief, if he were questioned about it. The evident availability of the visual appearance for use in support of the belief argues that we have such evidence. Finally, there is a serious question of why the perceptual state would be stronger evidence than the counterpart illusion. One might defend this claim on the grounds that the perception entails the truth of its content. But in general, the fact that a bit of evidence entails a proposition does not show that the evidence provides reason for believing the proposition. For example, the perceptual state of seeing an elm to be present entails that an elm is present. But one who is in that perceptual state might be quite unjustiﬁed in believing that an elm is present. The person might be unfamiliar with the look of an elm and he might have no good reason to think that the thing he sees is an elm. Generally, apart from the difference in entailment, the illusory counterpart of a visual perception seems exactly as good as a reason for belief. So it is not at all clear that there could be any good explanation of why an entailing perceptual state would be qualiﬁed to justify more strongly. We do not think that replying to external world skeptical arguments needs help from the sort of epistemic view offered by ED. We do not see a need for stronger justiﬁcation from perception than from illusion. The experiences of BIVs (and of ordinary perceivers) are sufﬁciently justifying for knowledge. As we acknowledge in ‘Making Sense of Skepticism’ there is a serious difﬁculty in explaining how this justiﬁcation is provided.4 But we see no good reason to doubt that it has this strength. And if the apparent strength ﬁnally has no adequate explanation, then as we also say there, we think that the absence of an explanation would count in favor of the truth of external world skepticism. It would not be as reasonable to look elsewhere for more 4

In our Evidentialism: Essays in Epistemology, Oxford, 2004.

292

REPLIES

justiﬁcation as it would be to acknowledge that we don’t have knowledge. Our external world beliefs would still often be ‘close’ to knowledge: they would often be true, quite reasonable, and ungettiered.

Response to Rysiew Patrick Rysiew describes a Reidian account of evidence as part of a view of justiﬁcation that he describes as ‘neutral as to the truth of evidentialism’. Patrick identiﬁes four roles that evidence is frequently thought to have. To his description of the ﬁrst role—that which justiﬁes belief—we would add that at least some evidence can justify beliefs without itself being justiﬁed. Sensory experiences, memory impressions, and perhaps other states, are external world ‘inputs’.5 They are our most direct access to the extramental world. As evidence for some propositions, they can at least help to justify the propositions without being the sort of thing that can be justiﬁed. This suits such evidence to stop regresses and prevent isolation from reality—two standing issues that must be resolved satisfactorily by any full theory of justiﬁcation. We think that the evidence that justiﬁes belief also plays the second of the roles that Patrick identiﬁes—being something that is respected by rational thinkers. The third role, that of being a guide to truth, we accept as well, although this phrase is subject to multiple interpretations, not all of which we would ﬁnd congenial. We will say more about this below. The ﬁnal role, that of being an objective neutral arbiter (of truth), is played by what we called ‘public evidence’. Roughly, this consists in detectable external conditions that can be justiﬁably believed to correlate with some sorts of truth. While we welcome a Reidian version of evidentialism. We will note a few differences between the account that Patrick developed and our preferred view of evidence. One Reidian theme is that evidence is the ground of a belief when the belief has a good ground. We have no objection to that, when the goodness is understood in a speciﬁcally epistemic way. This must be done to distinguish evidential grounds from ‘good grounds’ that arise when a belief is foreseen by the believer to be of practical beneﬁt. Even a believer with a ‘healthy mind’ might be naturally constituted to have such a beneﬁcial belief. This does not imply that the belief is supported by evidence or that it is epistemically justiﬁed. Another theme in the paper is to identify evidence by appeal to a ‘root’ notion of ‘evidentness’. The evidence would be ‘that which makes evident’. Later Patrick uses the phrase ‘evident (to some degree)’, and suggests that a false proposition can have this status. When a proposition is to some degree evident to someone, it is made to have that status by its support by the person’s evidence. In this reading, anything that has the potential to contribute to making a proposition evident to some degree is evidence. Evidence need not support a proposition so strongly as to make the proposition evident 5 See our paper ‘Evidence,’ in Epistemology: New Essays, ed. by Quentin Smith, Oxford, 2008, pp. 83–104.

REPLIES

293

to any given degree. It can support a proposition while being outweighed by evidence against the proposition. It can also support a false proposition. Those are things that evidence can do, as we understand it. Our candidates for what qualiﬁes as evidence, so understood, include conscious states like sensory experiences and memory impressions that intuitively support external world propositions. Evidence on this account has no essential connection to reliability. It is possible for someone to be in the conscious states that intuitively support external world propositions while in an environment where those propositions are never true. This result conﬂicts with Patrick’s claim about a Reidian account of evidence bearing out the role of evidence as reliable indicator of truth, which is one interpretation of the third role mentioned above. This reading also conﬂicts with taking ‘being evident’ to be factive, as some passages indicate that Patrick is inclined to do. For instance, if being evident is being obviously true, as he sometimes suggests, then it is factive. Evidence as we understand it can exist for a proposition without making the proposition in this sense ‘evident’. Here is another detail from Patrick’s paper about his Reidian account of evidence: ‘evidence is what indicates and impresses truths, as such, upon the (healthy, informed, etc.) mind’. Patrick takes Reid to include having generally reliable belief-forming faculties to be part of having a sound mind. Patrick uses this idea to infer, from a proposition’s being evident to some degree to ‘us’ (presumably, we are included in those of sound mind) that the proposition is likely to that degree to be true. In light of this last feature of the account of evidence, we do not understand Patrick’s Reidian account of justiﬁed beliefs in a demon world. He seems to count the beliefs as justiﬁed by appeal to the demon victims having minds that form reliable beliefs from the same experiential evidence in the environments in which they were extracted. We think that attributing the same justiﬁcation to the demon victim beliefs that correspond to ordinary justiﬁed beliefs is strongly intuitive, and thus highly desirable in an evidentialist account of justiﬁcation. But, for one thing, the demon’s victims need not have existed in any other environment. They could have been created by the demon for the purpose of deceiving them all along. If so, then by the Reidian account of evidence, there might still be such a thing as evidence in the demon world. The evidence there would be considerations that would indicate and impress the truth on minds in that world equipped to form reliably true beliefs. The victims would not have such a mind, though the demon might. But the considerations that do this for the demon’s mind may be quite alien capacities to discern its environment that are not had by the demon’s humanoid victims. If not, then the victims would not be justiﬁed on the Reidian account. Turning to another topic, we do not understand Patrick’s Reidian account of selfevidence. He asserts that the ‘ground’ of belief in some self-evident proposition is just the ‘evidentness’ of the proposition. Patrick says that a proposition’s self-evidence is just its having this evidentness. But as we understand the term, for a proposition to have ‘evidentness’ it must be such that, upon consideration of the proposition by someone, it is made evident to the person by some suitable evidence. If so, then a proposition’s ground for belief by someone cannot be its ‘evidentness’. The person must also have evidence that makes the proposition evident to the person.

294

REPLIES

Patrick also suggests that a Reidian self-evident proposition’s evidentness is itself evident to the person. This is supposed to be analogous to our seeing light itself as well as seeing objects that are illuminated by it. We think that there is a basic disanalogy here. Light enables us to see objects by appropriately relating those objects to our visual apparatus. Evidentness does not make a proposition evidently true by appropriately relating the proposition to our cognitive apparatus. The appropriate relation consists in our considering the proposition with adequate understanding. But no amount of mere understanding is sufﬁcient for a proposition to be evidently true. Any proposition asserts something about the world. Its truth depends on the world bearing out that assertion. Its evident truth does not consist in its relation to us or to our apparatus (except when those things are the aspects of the world that the proposition’s assertion addresses). Only evidence for how things really are in the asserted aspect of the world ‘illuminates’ for us the truth of the proposition. The evidentness to someone of the proposition that some proposition is evidently true to the person requires that the person have evidence indicating that the proposition really has the status of being evident. We do not ﬁnd in Patrick’s Reidian account any provision for this evidence.

Response to Swinburne Richard Swinburne makes a friendly explanatory offer: he offers for our acceptance a conception of the evidence that a person has. We certainly welcome the offer. We think that the nature of the evidence had raises signiﬁcant standing issues for evidentialism, some of which we will discuss further in our response to Alvin Goldman. But, unfortunately, we ﬁnd trouble in the offered conception of a person’s evidence. Richard observes that some of the inclinations that we have to believe propositions seem to us to be forced upon us by the world, rather than their seeming to be derived from other beliefs or produced in us by our own choices or preferences. He holds that all of our seemingly world-imposed inclinations to believe give us the evidence that we have. The inclinations vary in strength. Richard’s view, as we understand it, is that our evidence ultimately consists in these inclinations to believe, whether they are strong inclinations or weak ones. Subjects who understand probabilistic concepts may have beliefs corresponding to these inclinations that attribute probabilities to these propositions. Richard goes on to say that the propositions about which we have these inclinations are basic propositions. And those basic propositions for which our subjective probabilities correspond to their logical probabilities are ‘rightly basic’. (We are grateful to Richard for correspondence that helped us considerably in arriving at this interpretation of his proposal. An alternative interpretation is suggested to us by several passages in his essay. It is the view that our fundamental evidence is not the inclinations to believe but rather the beliefs resulting from those inclinations. We will make use of the former interpretation here, but our comments below could be adapted to the latter interpretation in a relatively straight-forward way.) Richard calls any of the propositions that we ﬁnd ourselves with some seemingly forced inclination to believe a ‘basic proposition’. The probabilities associated with

REPLIES

295

these basic propositions need not be high. Also, as Richard notes, we may have inclinations to believe propositions that are relatively remote from immediate experience. This seems advantageous in that it counts us as having evidence for the many historical and general propositions that we are justiﬁed in believing. However, one of Richard’s examples points toward a problem. He illustrates the range of propositions toward which we might have basic beliefs with the inclination to believe that Marks and Spencer sells good quality food. We see trouble in allowing a non-observational external world proposition like that to ﬁgure in the evidence that someone has. A fundamental difﬁculty is that these inclinations might have no rational basis, yet the theory seems to have the implausible implication that such a belief is welljustiﬁed if it results from a strong inclination of this sort. Richard believes that his theory avoids this result because the evidential support provided by these inclinations will be outweighed by other evidence the subject will have. This other evidence results from beliefs in (or inclinations to believe) propositions that make these unjustiﬁed propositions improbable. He also mentions some sort of a priori probability as something that might help with the problem in some cases. As we will now argue, we doubt that these considerations provide even a partial solution. A difﬁculty arises whenever there are conﬂicts among basic propositions. Suppose, for instance, that the advertising campaign of the PlastiMart grocery store chain has had its intended effect on you. As a result of it, you are strongly inclined to believe, in a way that seems forced on you by the world, this: B1. The food that PlastiMart sells is good food. But from other sources, you are equally strongly inclined to believe the following two propositions: B2. Preservative-laden food is not good food. B3. PlastiMart sells preservative-laden food. These three beliefs cannot all be true. Suppose that in fact you are not justiﬁed in believing B1. You are not having a memory impression that you recall the truth of B1. Rather, you have an inclination to believe B1 as a purely habitual association that was instilled in you by the ad campaign to link the idea of PlastiMart food with an attribution of good food. Although we lack a full account of what evidence consists in, we hold that a person’s sense experiences and memories that bear on a proposition are evidence about it, while an inclination toward belief by brute association like this one is not evidence for the proposition. So in our view, under the described circumstances you have no good evidence for B1 (assuming too that you lack any evidence that this inclination is reliable.) But if all three of these propositions B1–B3 are evidence for you (or the inclinations to believe them are evidence), as on Richard’s proposal, then how does the resulting version of evidentialism deliver the verdict that B1 is unjustiﬁed for you? When Richard comes closest to addressing this problem, he discusses how an evidentialist theory of justiﬁcation using his inclinations (or basic beliefs) as evidence might deal properly with ‘wild’ basic beliefs, such as someone’s happening strongly to believe:

296

REPLIES

B4. The Great Pumpkin returns on Halloween. He suggests that usually the person will have theories that are rendered probable by other of the person’s basic beliefs. These theories will rule out propositions such as B4 from being overall probable. The conditional probability of B4 on the theories will be low. We do not see how this sort of fact about conditional probabilities solves the problem. The conditional lowering of probability is symmetrical. We need some reason to conditionalize on some basic beliefs rather than others. Posing the problem in terms of our example of B1–B3, we note that each one of them is improbable, given the other two. The trouble is that nothing in the account of evidence that Richard offers tells us which of the conditional probabilities of some basic beliefs on others determines the strength of justiﬁcation. (Even in the extreme case of B4, the conjunction of the sensible theories to which Richard alludes does make B4 improbable, but equally that conjunction is improbable, given B4 and other typical basic beliefs. Which conditional probability shows strength of justiﬁcation?) Unlike Richard, we think that the sort of problem raised here is resolved by appeal to experiential evidence. One reason Richard proposes the theory of evidence he does is that, as he sees it, states without doxastic content cannot serve as evidence. The reason he gives is that ‘[we] cannot use these things as evidence, except in virtue of what [we] believe (true or false) about them.’ But this is a doubtful psychological claim. Doxastic attitudes always arise in causal sequences with eventual non-doxastic antecedents. This seems to happen when conscious sensory states lead to observational beliefs. These same sensory states seem often to be positively relevant to the truth of resulting observation beliefs. This relevance strikes us as holding just because of how things appear, not because of what we believe about the sensory state. Why could not any of this causing be ‘using’ the sensations as evidence for the observational beliefs? And if conscious sensory states can serve as evidence, then our evidence can get beyond our beliefs, and there is a way to avoid the sort of problem Richard’s theory faces. Support by the sensory evidence helps to distinguish the genuinely justiﬁed beliefs from the likes of B4.

Response to Fumerton We concur with a main thesis of Richard Fumerton’s paper: No plausible theory of epistemic justiﬁcation will entail this truth connection. TC1. Necessarily, if there are justiﬁed beliefs, then most justiﬁed beliefs are true. Some possible situations that are counterexamples to TC1 have massively deceptive evil demons. Some possibilities have infelicitous situations that were not intentionally contrived, where believers who follow epistemic reason just happen to be led to more errors than correct beliefs. This might even occur because of some relatively isolated engine of justiﬁed error that overwhelms in number the tendency toward truth that

REPLIES

297

justiﬁed beliefs otherwise have. Perhaps some obsessive recreational intellectual pursuit of those in the situation makes the judgments’ form rest on some false presupposition, with the consequence that all of the overwhelming number of resulting beliefs are untrue. Many such contingencies that are independent of justiﬁcation might possibly have decisively affected the balance of true to false beliefs. We do not think the falsehood of TC1 is a discouraging fact about epistemic justiﬁcation. Related to this insouciance on our part are plausible answers that Richard does not consider to an initial question that he raises. He asks why philosophers should be interested in the concept, if it were not even likely that justiﬁed beliefs are true. As a preliminary point, we think that it is in fact ‘likely’ that justiﬁed beliefs are true. First, if this likelihood is given a statistical interpretation, we see no good reason to believe that massive error actually overwhelms the vast preponderance of justiﬁed beliefs that are true. At least, they are true according to the best available evidence in the matter. And second, if this likelihood is given an epistemic interpretation, then we think that this preponderance is on balance supported by evidence that makes for both epistemic likelihood and epistemic justiﬁcation. So the two never come apart. Richard notes in the paper that one might deﬁne ‘epistemically likely’ this way, as a stipulation. The result of that stipulation would be to give an uninformative meaning to a claim that what is epistemically justiﬁed is ‘likely’. But we propose the evidential interpretation of ‘likely’ as a substantial semantic hypothesis about an existing meaning. This is what is commonly meant by ‘likely’ and ‘probable’. In any case, in this sense what is justiﬁed must be likely to be true. But suppose that we knew the contingency that justiﬁed beliefs were not ‘likely’ to be true in the statistical sense: it was somehow known to us that most justiﬁed beliefs were untrue. Here are some reasons that having justiﬁcation could still be reasonably desired. We suppose that a basis for being reasonably desired is something Richard intended to be asking about, in asking why philosophers should be interested in the concept if justiﬁed beliefs are not likely to be true. First, having epistemic justiﬁcation is necessary for having knowledge. So if knowledge is something reasonable to seek for its own sake, as it seems to be, then justiﬁed belief would be reasonably desired in order to have knowledge. The assumed known preponderance of falsehoods among justiﬁed beliefs would make aiming for knowledge a somewhat daunting quest. But still, the justiﬁcation would be needed in order to have such knowledge as was available. Justiﬁcation of beliefs may also be necessary for the other valuable epistemic ends of understanding and wisdom. Secondly, believing with justiﬁcation is sufﬁcient for having a reasonable or a rational attitude toward a considered proposition, even while also knowing that most justiﬁed beliefs are false. Having a rational attitude might be sensibly wanted for its own sake, or because it is intrinsically good. Thirdly, justiﬁed believing might be reasonably desired as part of a life of intellectual virtue. This in turn might be wanted as part of a good life for the likes of us.6 6

We are grateful to Bill Rowley for this point.

298

REPLIES

Finally, while it was known by someone that the totality of everyone’s justiﬁed beliefs was mostly false, the person might also know that the totality of unjustiﬁed beliefs included an even smaller fraction of true beliefs. For the sake of having the largest available fraction of true beliefs, such a person might reasonably desire having justiﬁed beliefs.

Response to Huemer We think the distinction Michael Huemer draws between agent-centered and agentneutral epistemological views is well worth thinking about. As we understand it, the difference amounts to this. A Centered view holds, concerning at least some considerations that people might have that can positively affect the epistemic justiﬁcation of a belief—reasons, evidence, or the like—that a person’s having the positive factor himself or herself is stronger reason for that person than is the person’s knowing for certain that someone else has that same positive factor. For instance, one’s own intuition that X is true more strongly supports X than does one’s knowing for certain that someone else has that intuition. A Neutral view holds that any positive epistemic factor justiﬁes to the same extent, whether the factor is present in the believer or equally well-known to be present in someone else. Perhaps Mike is right that our mentalist evidentialism places us ‘under strong pressure’ to be Neutralists. Our view, simply put, is that mental states are the evidence that makes all the justifying difference, and the same evidence does the same justifying, whoever has it. That sounds Neutral. However, a thesis that is extensionally equivalent to a Centered view is compatible with the view. The idea is that the overall support yielded by one’s own mental state is greater than the support resulting from knowledge that another person is in the same state, because one always knows better that one is in any given mental state than one knows that another is in the same state. If this thesis is true, then one could not have quite the same total evidence, by knowing as well as possible that another was in some mental state, as one has by being in that state oneself. If so, one who took our mentalist evidentialist view could consistently maintain that there was greater support by one’s evidence in some cases, or even always, in virtue of this difference in having the evidence oneself versus knowing about it in another. 7 Suppose instead that we can know what another’s mental states are well enough for the knowledge of them to be had as evidence in the same way as we have evidence in virtue of being in our own mental states. If so, then Neutralism still would not follow from our mentalist evidentialism. Contrast a person’s evidential situation when she has the intuition that p is true with her situation when she knows that another person has that same intuiton. Suppose she is equally well justiﬁed in believing, in the ﬁrst case, 7 The Neutralist thesis formulated as the necessity of same support by a consideration if the consideration is well enough known, would come out true, on this hypothesis, because the antecedent would be vacuously true. Thanks to John Shoemaker for this observation.

REPLIES

299

that she herself has that intuition as she is, in the second case, in believing that the other person has it. Actually having the intuition is a different mental state from knowing that another person has the intuition. These states might differ in their evidential signiﬁcance. Perhaps having an intuition that p ‘automatically’ provides at least some evidential support for p, whereas knowing that another person has an intuition that p provides support only in conjunction with background information about the evidential value of intuitions. Similarly, it is possible that perceptual experiences directly justify propositions while knowledge that others have comparable experiences justiﬁes only in conjunction with suitable background information. If such views are correct, then the letter of a Centered view would be upheld, since the knowledge of another’s justifying mental state would not give one the same justiﬁcation as would undergoing it oneself. But the spirit of Neutralism would also be upheld, since the justifying difference would really be a matter of having importantly different evidence, rather than the same evidence doing different justifying for oneself than it does for another. Suppose that the knowledge available of what another experiences is good enough and suppose also that when one does know this, it is always obvious enough what the ascribed mental state justiﬁes. We think that on these hypotheses, Neutralism is independently quite plausible, apart from speciﬁcally evidentialist views. We cannot see why a justifying consideration would be stronger, just in virtue of undergoing it oneself rather than knowing that it occurred in some other person. Generally, the proposition justiﬁed is not about either oneself or the other. So generally, the fact that one person undergoes the state, rather than another, is irrelevant to the truth of the proposition. We concur with Mike that one’s own mental states are ultimately all one has to go on, in justifying one’s beliefs. But this does not argue for a Centered view. It does not give any good reason to believe that one’s own mental state of, say, intuiting X, is any better support for X, than is one’s own mental state that is the basis by which one knows well enough that another intuits X. (If Timothy Williamson is right that knowledge is itself a mental state, then not merely the basis but also the knowledge itself would be one of one’s own mental states.) Both of these states are one’s own and both are part of what one has to go on, to justify one’s beliefs. They might work equally well in favor of X. When it comes to intuitions, it is difﬁcult to know that another really has the sort of intuition that seems most promising as supporting evidence, instead of some other sort of uninferred belief. But this is an obstacle to knowing well enough to have it as evidence that the intuition exists. If it is had as evidence, though, and appreciation of what intuitions justify was somehow automatic, then it would do the same justifying work.

Response to Kvanvig Jonathan Kvanvig attempts to provide a rational basis for the view that ‘rational disagreement is unproblematic’. Jon gives the name ‘molliﬁcationism’ to a contrasting

300

REPLIES

view, and he suggests that we have supported some version of this contrasting idea. In general terms, molliﬁcationism is the thought that learning that an epistemic peer (or an epistemic superior) has an attitude differing from one’s own toward a proposition is some reason to adjust one’s attitude toward the proposition toward the peer’s attitude. There are a number of ways to make this general idea precise, but instead of trying to formulate here an optimal version, we’ll call whatever is the most plausible version MPM and make claims about what it implies.8 We do ﬁnd the molliﬁcationist viewpoint attractive. Jon argues that MPM is contingently self-defeating, where this means that, in light of contingent truths about rational beliefs concerning one’s epistemic peers and superiors, MPM implies that it is not rational for any current epistemologist to believe MPM. This is supposed to be true because for any current epistemologist, CE, who believes MPM, CE is justiﬁed in believing: D. CE has some epistemic peer, or superior, now or in the future, who disbelieves MPM. MPM is supposed to imply that, in light of this peer (or superior) disagreement over MPM, it is not rational for CE to believe MPM. While some versions of molliﬁcationism may have the implication Jon identiﬁes, MPM does not. For one thing, MPM, being the most plausible molliﬁcation principle, makes the numbers count for something. At the extreme, if CE knows that only one of her peers or superiors about MPM will ever disbelieve it, and thousands of them will believe it along with her, then MPM will allow CE to continue her rational belief in it. Perhaps MPM implies that a single disagreeing peer makes some epistemic difference: CE will be slightly less justiﬁed in believing MPM in light of the sole dissenter than if there were known complete unanimity on the topic. But the belief can remain very well-justiﬁed. More realistically, if CE is justiﬁed in believing that the majority of her peers and superiors on MPM will be on her side, then she will have on balance reason favoring MPM from this source of evidence about MPM. This majority belief can be justiﬁed by a combination of CE’s evidence about current trends concerning MPM and CE’s best judgment about the merits of arguments directly for MPM or against it. We estimate that this is the actually justiﬁed attitude for proponents of MPM to take toward the balance of peer and superior attitudes toward MPM. Suppose instead that MPM, together with CE’s justiﬁed belief about the overall balance of peer or superior attitudes toward MPM, implies that CE is not justiﬁed in believing MPM. This situation is a possibility. Perhaps this counts as a certain kind of ‘self-defeat’, in the sense that the theory implies that in these circumstances belief in the theory is not justiﬁed. Jon counts this as an embarrassment for defenders of molliﬁcationism. We think that it is no embarrassment. The possibility accommodates the manifest variety of reasons that can exist for and against philosophical theses like 8 For a lot of valuable work on this topic, see Jon Matheson’s dissertation, ‘Dealing with Disagreement: Uniqueness and Conciliation’.

REPLIES

301

MPM. In fact, we think that it is a liability in a view about rational belief, if it implies that belief in the view is exempt from defeat by counter-evidence. We think that it is another liability in a view about rational belief if the view does not count disbelief by epistemic peers as counter-evidence. If these thoughts are correct, then the only substantial remaining question about having one’s view of justiﬁcation defeated by this counter-evidence is that of how strong the defeating actually is. As we say, we estimate that the evidence against molliﬁcationism from peer dissent is not defeating on balance. But there is also the following sort of counter-evidence. Any well-informed philosopher’s justiﬁed beliefs about the fate of most historical philosophical theses makes a strong inductive case for suspending judgment on her currently favored philosophical theses. At a minimum, these two sources of counterevidence together call for considerable tentativeness in one’s attitudes toward one’s favored philosophical theses. As part of his case for his own preferred view about the epistemology of disagreement, Jon discusses what he calls ‘closure of inquiry’ and he contends that self-trust plays a key role in settling the issues. He holds that rational belief in a proposition, B, requires adequate evidence in support of the truth of B and also adequate evidence about something else. He puts this further condition in a few different ways. The idea seems to be that sometimes support for a belief is strong and stable enough that it would be reasonably thought by the person not to be signiﬁcantly weakened by further inquiry. Jon holds that it is rational for someone to believe B only when the person’s evidence for B justiﬁes both B and the claim about the rational durability of B in any further inquiry into B’s truth. Let’s call the second condition on rational belief The Closure Requirement. Jon holds that when CE rationally believes B and ﬁnds out that she is in epistemic peer disagreement over B, there is a conﬂict in CE’s meta-level evidence concerning B. On one side is CE’s knowledge of the peer disbelief in B. This seems to defeat the support CE has that The Closure Requirement is satisﬁed with respect to B. Given CE’s heretofore rational belief in B, though, CE’s total body of information relevant to B, prominently including whatever made CE’s evidence for B previously meet The Closure Requirement, provides meta-level evidence to the effect that the original evidence is indeed sufﬁcient. As Jon says, in light of this conﬂict, some sorting is needed to determine which meta-evidence ‘takes priority’. In Jon’s view, in such a competition at the meta-level ‘the only way to go is up’. To resolve the rational status of continued belief in the face of the peer disagreement, Jon looks to the subject’s extent of trust in herself in the matter. This seems to be the extent of relative conﬁdence that the subject has in her own assessment of the case concerning the belief at issue in comparison to her conﬁdence in her peer’s assessment. If the subject lacks discrediting information affecting the actual strength of self-trust, then it is decisive. If the self-trust is strong enough, then the rational attitude toward the belief at issue is continued belief; if the self-trust in the matter is weak enough, then rationality requires a less accepting attitude toward the proposition.

302

REPLIES

We see no epistemic role for this comparative conﬁdence. Self-trust appears to be a non-epistemic emotional factor, unless it is understood to be an epistemically justiﬁed belief that the subject is better at coming to the right view of the proposition. This justiﬁcation will not be there, if the subject continues to have justiﬁcation that the dissenter is a peer. We can assume that there is the competing meta-level evidence that Jon describes: on one side, the subject’s evidence about the belief that at least previously was strong enough to meet The Closure Requirement, and on the other side, the subject’s knowledge of the peer dissent. As always, we hold that the rational attitude toward the proposition for someone is settled by the strength of the person’s overall evidence. In this sort of peer conﬂict, there will be facts of detail about such things as how strong the subject’s belief-independent case for the belief was, and how good are the subject’s reasons for continuing to count the dissenter as an epistemic peer on the proposition. Assuming that all of this results in equi-balanced evidence concerning the proposition, suspension of judgment is the rational attitude. If the overall evidence tilts one way or the other, then so does the rational attitude. It may seem that this balance of overall evidence will be difﬁcult for the subject to determine, or perhaps even indeterminate. If the former is the case, then we hold that the actual balance of evidence the subject has is decisive, but the subject may be more or less blameworthy for failing to follow it. If the status of overall evidential support is indeterminate, then suspension of belief is justiﬁed. This is true because, just as in equibalanced evidence cases, indeterminate evidence does not favor the proposition or its negation. Also, this verdict about any indeterminate cases has some separate plausibility. If there is no fact of the matter about whether or not one’s evidence supports a proposition, then one would not be following one’s rational considerations either to believe the proposition or to disbelieve it. Alternatively, if there is indeterminacy in the evidential support, then perhaps the justiﬁcatory status is also indeterminate. Either way, the evidence makes all of the difference and sheer self-trust does not affect the outcome.

Response to Goldman Alvin Goldman seeks clariﬁcation of the concept of evidence in our evidentialist principles such as EJ and WF. In response to this inquiry we observe that the concept of evidence that we intend is perhaps the commonest ordinary one. Dictionaries for the general public offer synonyms that express what we have in mind. For instance, dictionaries that one of us happens to have ready to hand say ‘that which serves to indicate or suggest’ (New College Edition American Heritage Dictionary, 1981), ‘An appearance from which inferences may be drawn; an indication, mark, sign, token or trace’ (OED), and ‘something, esp, a fact, that gives proof or a reason for believing’ (Longman Dictionary of the English Language, 1991). More recent references say, ‘A thing or things helpful in forming a conclusion or judgment’ (American Heritage, online,

REPLIES

303

2009); ‘ground for belief or disbelief ’ (Collins English Dictionary—Complete and Unabridged, 2003). A bit more carefully, but entirely along the same lines, we can say that some evidence of the truth of X for S is an indication to S that X is true, understanding an indication to be of any strength up to, and including, proof. One of the conceptions that Alvin cites from Tom Kelly’s ‘Evidence’ article in the Stanford Encyclopedia of Philosophy is also entirely congenial to convey the idea. The quoted phrase is: ‘a guide to truth’. Alvin immediately paraphrases this as ‘a reliable sign, symptom, or mark of the truth’. The addition of ‘reliable’ is unwarranted. It is not in the Kelly quote, and it is not part of the ordinary concept. (It is notable in this regard that none of the evidence in popular dictionaries that we consulted mentions reliability as a requirement for being evidence.) For clariﬁcation we should add that, in our evidentialist principles like EJ and WF, the having of evidence is meant in a fairly technical way. It is intended as a sort of mental hosting. This is clearly much narrower than the most commonsensical notion of possession. Also, not all that might be thought to be included in someone’s mental hosting of facts will count as the person’s ‘having’ the facts as evidence, since a more selective account of what aspects of the experiences are evidence for the person is necessary in order to solve the speckled hen problem.9 But again, the concept of evidence that we are employing is the ordinary one and our notion of the having of evidence seems to us to be familiar enough, though its understanding can take some prompting. Moving to a more substantive topic, Alvin argues that evidentialism needs help. He proposes reliabilist elements to provide the help. We are grateful for the offer of assistance. But we do not see the need. One place where Alvin contends that our version of evidentialism is inadequate concerns knowledge of stored information. Alvin makes his case on the basis of his example of Ichabod. Here is the example. Years ago Ichabod formed a belief in proposition q by acquiring it in an entirely justiﬁed fashion. He had excellent evidence for believing it at that time (whether it was inferential or noninferential evidence). After ten years pass, however, Ichabod has forgotten all of this evidence and not acquired any new evidence, either favorable or unfavorable. However, he continues to believe q strongly. Whenever he thinks about q, he (mentally) afﬁrms its truth without hesitation. At noon today Ichabod’s belief in q is still present, stored in his mind, although he is not actively thinking about it. I stipulate that none of his other beliefs confers adequate evidence either for believing q or for disbelieving it. Since Ichabod remembers q’s being the case, and since he originally had excellent evidence for q, which was never subsequently undermined, Ichabod’s belief in q at noon today is justiﬁed. Moreover, as I argue elsewhere (Goldman 2009), if we refuse to grant justiﬁedness to beliefs of this sort, which derive from preservative memory, there will be serious skeptical ramiﬁcations: people will fail to know a great many things that common sense credits them with knowing. But does Ichabod’s noontime belief in q satisfy Feldman and Conee’s account of well-foundedness? Does it qualify as doxastically justiﬁed under their theory? 9 For discussion, see Richard Feldman, ‘Foundational Justiﬁcation,’ in Sosa and His Critics, edited by John Greco, Blackwell, 2004, pp. 42–58.

304

REPLIES

No. At noontime Ichabod undergoes no evidential experience of seeming to remember q and (by hypothesis) possesses no stored beliefs on which his belief in q is based—or which it ﬁts. Years ago, as indicated, he had evidence that his belief in q ﬁtted, and perhaps, his noontime q-belief might be said to be ‘based on’ that old, forgotten evidence. But can a basing relation help Feldman and Conee in this case? Their analysis of well-foundedness explicitly requires that justifying evidential states must be held at the same time as the target attitude. But the earlier justifying evidence is no longer possessed, neither consciously nor unconsciously. So evidentialism implies that Ichabod’s noontime belief in q is not justiﬁed (well-founded), an intuitively incorrect verdict. It should be emphasized that in the Ichabod case there is no memory experience that triggers his noontime belief in q, the sort of experience foundationalists typically invoke to account for justiﬁed memory-based belief. But many justiﬁed memory-based beliefs—especially stored, or non-occurrent, memory-based beliefs—are unaccompanied by memory experiences. Nonetheless, such beliefs need to be credited with justiﬁedness. Unless they are justiﬁed, they cannot be known—certainly not on a justiﬁcationist view of knowledge. Yet non-occurrent knowledge—that is, knowledge involving non-occurrent belief—constitutes the vast bulk of our knowledge at any given moment. If, as theorists, we abandon such knowledge, the skeptic will have won a major (but unearned) victory.10

In response, we contend that at noon Ichabod is in a mental state that is evidence that he has for q. The state is his disposition to recollect q. By hypothesis, at noon Ichabod is not thinking about anything relevant to q, and he has forgotten all of the initial evidence that he had when he learned q. But he has not forgotten q. Since this is so, one aspect of his noontime mental condition makes it true that he is able to recall the proposition. He has the potential to bring q to mind with the phenomenology of activating a memory, speciﬁcally, the memory that q. This is not true merely because he has the stored belief that q. Many of our stored beliefs are recalled as things we merely believe, not as things we know. But some are recalled as things we know. We will say that a ‘disposition to recollect’ a proposition is a disposition to bring to mind the proposition as known. Ichabod has as evidence at noon a disposition to recollect that q. That is evidence justifying his stored belief that q. If this disposition to recollect is sufﬁciently strong and clear, then in the absence of defeaters, it is strong enough evidence for him to know q. Thus, there is something among his mental states at the time that provides justiﬁcation for this stored belief. The justifying is not done by any active conscious occurrence at the time, evidentialists need not appeal only to conscious occurrences as justiﬁcation. Having one of these dispositions to recollect is not being in a factive state. They can justify false beliefs that we only seem to have learned when the dispositions were formed. This sort of recollective disposition is not what Alvin calls preservative memory. He says, ‘Preservative memory does not create or generate justiﬁedness 10 [footnote from Goldman] It may be contended that Ichabod’s noontime belief in q is not as justiﬁed as his belief in q was when it was originally formed and he was in possession of the original evidence. This point may be conceded. Nonetheless, given the history described, the noontime belief has a substantial measure of justiﬁedness, and that measure of justiﬁedness cannot be explained by evidentialism, as far as I can see.

REPLIES

305

from “scratch”, but instead transmits a belief ’s justiﬁedness (or unjustiﬁedness) from one time to a later time.’ In contrast, a disposition to recollect is a potentially momentary state. In ordinary cases it exists by its having been formed in the past when a proposition was learned, or at least the proposition seemed to have been learned, and then retained by a process of long-term memory consolidation. But a recollective disposition toward a proposition might come about from brain malfunction or tampering. Whatever causes a disposition to recollect does create evidence for the content proposition from scratch. Having the disposition constitutes having some defeasible evidence for its content, whatever its historical origins. When a recollective disposition resulting from malfunction or tampering leads to a justiﬁed true belief, it may nevertheless be a Gettier case rather than a case of knowledge. We turn now to another topic of Alvin’s chapter. Alvin cites work by Jack Lyons to cast doubt on the evidentialist idea that experiential evidence is necessary for the justiﬁcation of perceptual beliefs. Two sorts of examples are mentioned. The ﬁrst sort involves the ‘zombies’ of philosophy of mind fame. They are psychological duplicates of ordinary people, except that the zombies lack all conscious states. Alvin reports that Lyons credits such zombies with justiﬁed perceptual beliefs. The other sort of case is that of some blind people who sense obstacles without touching them. Reportedly, they think that they are doing this by feelings in the skin of their faces, while in fact they are making use of a subtle form of echolocation. Lyons proposes to account for the justiﬁcation of these beliefs by reliabilist considerations. Alvin is sympathetic, though he does not endorse the proposals. We see no problem in these examples for unmodiﬁed evidentialism. By feedback from the environment or by testimony, the zombies and the blind people may come to have evidence that certain of their states are correlated with certain environmental facts. In particular, these states would be spontaneous unconscious perceptually formed beliefs in the zombie case, and facial skin sensations in the blind case. When they are in one of those states and they have suitable correlation evidence, their resulting beliefs in the correlated sort of environmental fact would be justiﬁed. Without any correlation evidence, it is questionable that either group would be justiﬁed in the perceptual beliefs. Arguably, we do have correlation evidence about our own perceptual beliefs. This is contestable. But if our visual beliefs are justiﬁed without evidence about correlations, then that must be partly because the perceived environment looks certain ways to us. This looking is a conscious state, and it constitutes part of what makes our beliefs justiﬁed. The environment does not look any way at all to the zombies (though, since they duplicate our non-conscious psychology, they will falsely say that things look, sound, smell, etc. various ways to them, just when we say these things.) The zombies would lack these conscious reasons for their perceptual beliefs. In that case, lacking any evidence correlating these beliefs to states of the world, it is doubtful that they are justiﬁed. Even so, evidentialism can be developed in a way that accommodates the contention that the zombies’ perceptual beliefs are justiﬁed. If they are justiﬁed, then the zombies

306

REPLIES

are in non-conscious perceptual states that lead to these beliefs just as our conscious perceptual states do. If the zombies are justiﬁed, then their non-conscious perceptual states seem to qualify as evidence for them. What evidentialism cannot accommodate is the contention that these beliefs are justiﬁed but are not accompanied by any justifying conscious or non-conscious mental state. We think that this negative epistemic verdict on beliefs that would arise as if from nowhere is entirely plausible. In the case of the blind people, they are not getting perceptual beliefs without experiences. For one thing, they are having facial skin sensations. These do not seem to be good reasons to form beliefs about obstacles, because, in ordinary cases at least, our skin sensations do not represent to us the presence of untouched obstacles. The blind people are in fact using subtle echolocation to form the obstacle beliefs. Perhaps the sounds bring about subtle conscious aural experiences. If so, then this aural evidence may well have a justifying representational content. It would be a highly interesting fact that the blind people mis-identify the sensory modality of their perceptual evidence. But mentalistic evidentialism does not require people to have any particular prowess at identifying their justifying evidence. It just requires that the evidence be within their mental lives. This is an instance of a familiar general phenomenon. We have justifying evidence for our beliefs about the identity of people or things we encounter, but often we are unable to identify which aspects of the evidence enable us to make that identiﬁcation.

Response to Lehrer Keith Lehrer discusses evidentialism as it bears on the question of what it is reasonable to accept in a circumstance of peer disagreement that Keith calls ‘the parity condition’. Two philosophers are in the parity condition after they have exhaustively exchanged views concerning a proposition that one believes and the other denies, when each ‘recognizes that the other is as trustworthy philosophically and epistemically as the other.’ Keith argues that in these conditions one can reasonably accept (or deny) the proposition by the very act of accepting (denying) it. As Keith puts this thought, ‘Acceptance trumps parity’. To understand Keith’s argument, it is important to note that he distinguishes belief from acceptance. We take it that the difference, amounts to this. The idea of belief is the ordinary notion of a relatively stable doxastic attitude of afﬁrmation of a proposition. Belief is not generally subject to adoption or alteration at will, though voluntary conduct can variously inﬂuence what one believes. Belief need not be responsive to one’s evidence. Acceptance, in contrast, is a voluntary act of positive evaluation of a proposition on the basis of a body of evidence. Acceptance of a proposition is sufﬁciently subject to voluntary control that the question of whether or not to accept a proposition can be decided by reasonable practical deliberation. Keith also distinguishes the reasonableness of a person’s accepting a proposition from its justiﬁcation. He conducts his discussion almost entirely in terms of reasonableness.

REPLIES

307

But the principal evidential condition that ﬁgures in Keith’s discussion of the parity condition is the same for each. Both reasonable and justiﬁed acceptance of a proposition have as a necessary condition that the person’s evidence more strongly supports the proposition than its negation. Suppose that philosophers P1 and P2 ﬁnd themselves in the parity condition concerning proposition X, with P1 believing X and P2 denying X. Keith asks us to consider whether it can be reasonable for P1 then to accept X. Keith holds that this can be reasonable. He contends that P1’s acceptance of X can give P1 evidence for X and this new evidence tips the balance of P1’s evidence in favor of X, thus meeting the necessary condition for reasonable acceptance. Keith argues for the conclusion that P1’s acceptance of X gives P1 evidence of X. The argument is intricate. As we understand it, a key step is this. In order for P1 to recognize that P1 is in the parity condition with P2, P1 must have a capacity for judgment about X that is worthy of P1’s conﬁdence. So if P1’s reﬂection on the total evidence at this point in fact leads P1 to an acceptance of X, then P1 has this acceptance by one whose judgment is worthy of P1’s conﬁdence as evidence that X is true. Keith offers a further argument about how P1’s judgment could have become worthy of P1’s conﬁdence. It begins with a premise asserting that P1 is worthy of trust in pursuing the objectives of reason. While we concur that P1’s judgment could be worthy of P1’s conﬁdence, we see no need for any premise about P1’s trustworthiness. We would instead explain P1’s having worthy conﬁdence in P1’s judgment about X in terms of evidence that P1 has for the accuracy of P1’s judgment in similar matters. This would be evidence for P1 as a matter of fact, not depending on P1’s attitude toward herself or the evidence. We see no reason why P1 could not have such evidence for P1’s proﬁciency in the matter. We have no objection to there being cases in which philosophers do in this way have trustworthy conﬁdence in their philosophical judgment in the matter at hand. There is a small problem in the step from supported conﬁdence in ‘judgment’ about X to evidence for X from the ‘acceptance’ of X. P1’s support for this good judgment might concern P1’s proﬁciency at belief in matters such as X. This would not say anything about P1’s capacity for trustworthy acceptance. But we can suppose that P1’s support concerns judgments that derive from P1’s acceptance conduct in similar matters. With that assumed, we concur with Keith that P1’s acceptance of X, in response to the total evidence concerning X in the parity condition, would be new evidence for P1 that X is true. But even if P1 does acquire new evidence for X in the way just described, it does not follow that P1 is reasonable (or justiﬁed) in accepting (or believing) X. This is because it does not follow from P1’s having this new evidence that P1 meets the necessary condition for reasonable (or justiﬁed) acceptance. For the condition to be met, the balance of P1’s total evidence concerning X at that point would have to favor X. This balance depends on what impact this new evidence has on P1’s total evidence. But it also depends on whether P1 acquires any additional new evidence concerning the acceptance response to the parity condition by the opposing peer, P2.

308

REPLIES

On the ﬁrst point concerning the impact of the new evidence from acceptance: it is far from clear that this evidential fact, when added to P1’s initial evidence, really does cause P1’s evidence to shift to support for X. Sometimes, the addition of information that would otherwise make a difference does not do so because it is undermined. Suppose that there is some proposition on which your evidence is exactly neutral. You know that a particular person is typically reliable on the topic of that proposition and you learn that the person believes it. This would ordinarily make evidence at least slightly supportive. But if you also learn that the person’s evidence is itself neutral on the proposition, then this effect is undermined unless you happen to have reason to think that the person’s acceptance has some non-evidential positive correlation with the truth of its content. And this is precisely what is happening in Keith’s case: P1 is learning that P1 accepts a proposition on which P1’s evidence, apart from this acceptance, is counterbalanced. Furthermore, P1 may well acquire additional new evidence. Suppose that P2 simultaneously informs P1 that P2 accepts the negation of X upon reﬂection in the parity condition. Clearly, P1 does not have a balance of evidence in favor of X. P1’s new acceptance evidence is balanced off, pro and con. But even if P1is not told this by P2, P1 would have evidence that this acceptance by P2 of the negation of X is as likely as P1’s acceptance of X. The evidence that establishes the parity condition for P1 includes evidence of P2’s equal intellectual merits on the topic and P2’s disbelief at the close of their exhaustive exchange concerning X. This evidence supports that P2 would have the same tendency to ﬁnd against X, on reﬂection, that P1 has to ﬁnd in favor of X on reﬂection. And that fact about P2’s evident inclination in the matter defeats the new evidence from P1’s acceptance as well as would an explicit report by P2 to that effect. The strength of P1’s new evidence for X from P1’s actual acceptance of X derives entirely from P1’s evidence about P1’s prowess on the topic of X. P1 has recognized that P2 has equal prowess and P1 has no evidence that P2 would diverge from P1 in how this would affect P2’s inclinations concerning acceptance. So in light of the defeating effect of P1’s evidence concerning acceptance by philosophical peer P2 of the negation of X, P1’s balance of evidence upon acceptance would not favor X. Thus, we doubt that ‘acceptance trumps parity’.

Response to Steup Matthias Steup offers an evidentialist argument against a version of external world skepticism based on the possibility of our being brains in a vat (BIVs). The argument for skepticism includes the premise that we do not know that we are not BIVs. Matthias argues against this premise. He contends that we have evidence sufﬁcient to eliminate for us all reasonable doubt, and thus to enable us to know, that BIVs do not exist. He defends his argument against the charge of being question-begging. The defense includes an account of what it is for an anti-skeptical argument to beg the question. The positions that Matthias takes in his contribution are quite congenial to us. We differ with him over some details.

REPLIES

309

Matthias’s anti-skeptical argument has a premise asserting that we know that BIVs do not exist. On behalf of this premise, Matthias cites evidence that we have in support of the proposition that BIVs don’t exist. The evidence he cites consists in some familiar facts about the inadequacy of current technology to keep a human brain functioning in a vat of nutrients. One such fact is this: 1. Textbooks in neurophysiology don’t have a chapter entitled ‘Envatment’. Matthias notes that these sorts of facts are about the present time. If human brain envatment is actually accomplished in the future, then other evidence will be needed against our being BIVs. We concur with Matthias that propositions such as 1 are evidence that many people now have, and that this evidence leaves those who have it with no reasonable doubt that BIVs do not exist. But in order for an anti-skeptical argument to give good reason to deny the hypothesis that we are BIVs, it must cite some additional evidence. It cannot rely ultimately on the evidence of propositions like 1. Citing a proposition as evidence is at least claiming that the proposition has an epistemic status which is strong enough to rely on it as evidence. For propositions like 1, only knowledge of them clearly qualiﬁes to give them evidential status. In any event, presumably Matthias intends the anti-skeptical argument to cite an item of knowledge when it turns to 1 as evidence. But relying ultimately on our having knowledge of an external world propositions like 1, without defense of that claim, in support of a premise in an argument for denying external world skepticism, is claiming without defense that we have an instance of external world knowledge. Identifying such an instance need not be epistemically idle. Pointing out that a surprisingly obvious entity is actually an instance of a kind, in arguing against the conclusion that the kind is empty, gives evidence against that conclusion. But a claim that a proposition like 1 is known claims the existence of an example of external world knowledge that is not surprisingly obvious. Citing propositions like 1 as known gives us no new evidence that we have some external world knowledge. Further, it is not a brute epistemic fact that such propositions are evidence, if they are. Instead, we need evidence for them, in order to know them and have them as evidence. Unless the claim that we do have them as evidence is defended, an argument that relies on this claim gives us no good reason to deny the skeptical conclusion. This defense is readily available. There is evidence people often have that supports 1, namely, evidence about what people seem to see when they seem to read the relevant textbooks and apparent memories of such readings. As we argue below, ultimately, the case against skepticism must be made on the basis of this kind of evidence. Matthias offers an answer to the charge that his way of arguing against external world skepticism begs the question. We have been careful not to use that language. We think that the ﬂaw in any argument that merits that charge can be more helpfully described by explaining where the argument fails to give evidence that is needed to justify its conclusion. That is what we have tried to do here.

310

REPLIES

We do think that Matthias’s characterization of begging the question facilitates overlooking the sort of inadequacy that we ﬁnd in his anti-skeptical argument. Question Begging: An anti-skeptical argument, based on a premise P justiﬁed by evidence E, begs the question against a skeptical alternative, A, if and only if (i) A is intended to undermine P; (ii) E fails to favor P over A. This characterization does not count an argument as begging the question against a skeptical alternative, whenever the argument is based on evidence that does in fact favor the denial of the alternative. But it can happen, as in the present case, that although the propositions cited as the evidence make a good case for the denial of the skeptical alternative, the argument relying on those propositions fails to provide evidence against the alternative. The reason is that our reliance on those propositions as evidence against the skeptical alternative depends on our having them as evidence. This status cannot be taken for granted, since the propositions are assertions about the external world. A good reason to believe that we have external world knowledge must defend our having the propositions as evidence without assuming that we know them. Arguing from the propositions as evidence that we have against the skeptical alternative requires giving some other sort of evidence that we do have them as evidence. Matthias’s argument does not meet this requirement. Thus, in order to complete the sort of anti-skeptical argument that Matthias offers, evidence is needed that we have propositions like 1 as evidence. Again, it seems clear that we do have such evidence. Our sensory experiences and memories are not subject to the skeptical argument. They exist whether or not external world skepticism is true. They are available to us as evidence. Some of them—memories of textbooks and the like—are evidence for propositions like 1. This evidence supports these sorts of propositions sufﬁciently for us to know them. These further claims are not trivialities. It is not obvious how the mental states support the corresponding external world propositions, and it is not obvious that the support is sufﬁcient for knowledge. As we see it, the rest of the job of arguing against external world skepticism consists in explaining the evidential connections here and explaining why this evidence is sufﬁcient for knowledge. In ‘Making Sense of Skepticism’ we sketch our ideas about solving the former problem and we try to do some of the latter work.

Response to Baehr Jason Baehr proposes a friendly amendment to our evidentialist view about having propositional justiﬁcation for a belief. The amendment is a constraint pertaining to intellectual virtue. Jason motivates the amendment by the use of examples that are intended to show some epistemic sort of insufﬁciency in the purely evidential condition on justiﬁcation that is asserted by our view. In Jason’s view, support by evidence is always needed to justify. But when someone’s agency makes a salient contribution to the evidence that the person has concerning some proposition, support for the proposition by that evidence is not enough for belief to be the attitude that is justiﬁed

REPLIES

311

for the person. When the agency makes this contribution, it must also be that the person has functioned, with regard to that contribution, in a manner ‘consistent with [employing] intellectual virtue’. We certainly welcome the spirit of Jason’s proposal. Also, the examples that suggest to Jason the need for the modiﬁcation call for attention. They can seem to show trouble for evidentialism. Jason’s proposal is a carefully defended and sensibly developed modiﬁcation. It preserves some important assets of our unmodiﬁed evidentialist view—all of the assets that are implied by requiring evidence for justiﬁcation. Moreover, there is something intuitively better about evidence acquisition that meets the proposed constraint that the evidence is affected in a manner that is compatible with exercising intellectual virtue. To violate that condition is to act in a way that virtuous inquiry would have avoided, and that seems deﬁnitely worse in some way than does meeting the condition. We deny, however, that Jason’s examples are best accounted for by modifying our view. We think that it is better to keep intellectual virtue out of the account of epistemic justiﬁcation. In Jason’s ﬁrst two examples, George exempliﬁes the possibility of having on balance supporting evidence for proposition (h) Exposure to secondhand smoke poses no signiﬁcant health risks. George has this balance of evidence supporting (h) partly as a result of failing to make any inquiry concerning (h). In the second example, that of Gerry, he exempliﬁes the possibility of having on balance supporting evidence for (h) partly as a result of making an irresponsible inquiry concerning (h). Jason explicitly concurs with our view that George’s and Gerry’s evidentially supported beliefs in (1) are ‘in a sense’ justiﬁed. We’ll say more soon about that sense and its signiﬁcance. But Jason maintains that there is another sense of ‘justiﬁed’,’ one that he thinks evidentialism is ‘attempting to capture’, in which those beliefs are not justiﬁed. He also holds that the justiﬁcation for which supporting evidence is sufﬁcient ‘is not a very signiﬁcant epistemic good’. It will be helpful to take up the second complaint ﬁrst. We are quite unclear about the idea that Jason employs, derived from contentions by William Alston, to the effect that there are a variety of ‘epistemic desiderata’ that various approaches to justiﬁcation seek to explain. We are puzzled by the notion of ‘epistemic desiderata’. Presumably desiderata of any sort are goals or goods. But we are quite unsure how a goal or a good is supposed to qualify as epistemic. As we understand it, and as general purpose dictionaries conﬁrm, calling something ‘epistemic’ is saying that the thing somehow pertains to knowledge. Perhaps epistemic value is value pertaining to knowledge. If so, then widely assorted things have instrumental epistemic value, in that they causally contribute signiﬁcantly in one way or another toward the existence of knowledge of something by someone. Having justiﬁcation clearly has that sort of causal role, but so do having hope, having a good night’s sleep, and having children. So we very much doubt that just making some causal contribution to knowledge can qualify something as an epistemic desideratum.

312

REPLIES

Perhaps intrinsic epistemic value is restricted to knowledge itself and states of affairs that include it. But knowledge, and the other things that entail it, cannot exhaust the relevant epistemic desiderata, since having justiﬁcation is supposed to be sufﬁcient for having an epistemic desideratum. Perhaps there is a third category of epistemic value, contributory epistemic value. Perhaps this kind of value is had by whatever helps to constitute things of intrinsic epistemic value. The conditions that constitute knowledge have this kind of value. One way to identify the sort of justiﬁcation for which we propose evidentialist conditions is that it is a constituting condition for knowledge. So in this sense, evidentialism is about something that has ‘contributory epistemic value’. The constituting conditions of knowledge do not seem to us to be an entirely happy extension for any intuitive sort of value, however, since it seems unfortunate to have to say that belief has the same sort of epistemic value as justiﬁcation. But we do not see how to do better in understanding the notion of an epistemic good that our topic of justiﬁcation clearly exempliﬁes. The justiﬁcation that we propose evidentialist conditions to explain is, in this sense, a contributory epistemic good. Knowledge partly consists in it. Jason might be right that on its own it is not a very signiﬁcant epistemic good. In our view, this is suggested by the fact that when a proposition merely has this sort of justiﬁcation for a person, and no other constituents of knowledge obtain—in particular, the proposition is not true— then, for the sake of being in the deﬁnitively epistemic state of knowledge, the person does not seem to be in good shape. Perhaps nothing short of knowledge is a signiﬁcant epistemic good (or perhaps satisfying Timothy Williamson’s notion of being in a position to know, that is, having everything needed to know except belief, is also having a signiﬁcant epistemic good). But this sort of fact about a pure and narrow sort of epistemic signiﬁcance that does not argue either that evidentialism is an inadequate account of the justiﬁcation it addresses or that this sort of justiﬁcation is of only meager interest. In addition to helping to constitute knowledge, this sort of justiﬁcation helps to make up one familiar category of reasonable belief. Whenever someone believes a proposition on a basis that provides this sort of justiﬁcation, the belief is reasonable in at least one way (‘in one way’ because the person might have other sorts of reasons not to believe). Since this can happen when knowledge is unavailable on the topic, it is of separate interest. Additionally, evidentialism is a helpful view about this sort of reasonable belief partly because it readily accounts for the intuitive differences in strength—the stronger the support by the person’s evidential basis, the more reasonable the belief. In Jason’s examples of George and Gerry, their beliefs in (h) have the sort of justiﬁcation that evidentialism characterizes. They could not know (h), since it is untrue. But their beliefs in (h) have justiﬁcation and they are reasonable. Believing (h) is the doxastic attitude toward the proposition that accords with their reasons that give them some indication of whether or not (h) is true. Jason seems to concur with this, or at least, he does not offer any reason to doubt it. Though we say that belief is the justiﬁed and reasonable response to George and Gerry’s evidence about (h), we do not wish to say that they have done altogether

REPLIES

313

well on the topic of (h). George, after getting some initial support for (h), has given no attention to getting evidence for it. In describing the example Jason does not make it clear that this is intellectually vicious on George’s part. To escape intellectual vice we need not pursue responsible inquiry into every proposition that we ever consider. Some are of no interest to us at all. But we can suppose that George is interested in whether (h) is really true. His intellectual sloth and obliviousness have kept him from making easy efforts that would have turned up much better evidence about the truth value of (h). The description of Gerry explicitly tells us that Gerry inquires viciously— being ‘prone to gullibility, carelessness, and hasty generalization’. Given these assumptions about the examples, we can all agree that George and Gerry have been bad inquirers concerning (h). That sort of ﬂaw in their inquiries seems to us to account for the intuitive deﬁciency in their belief in (h). Jason thinks that George and Gerry’s belief in (h) are ‘unjustiﬁed’ in some sense of interest to evidentialists. The closest Jason comes to identifying this sort of justiﬁcation is when he says, concerning the evidence that George and Gerry have, ‘the evidence clearly is not what it should be’ and ‘the beliefs of George and Gerry are based on evidence that ought to be other than it is’. It is notable that these claims are extremely difﬁcult to take literally. The evidence that George and Gerry have consists in mental states like experiences and memories. It is doubtful that such states could have provided different evidence. It is yet more doubtful that such states ‘should be’ different in any way. What seems clearly true in this case is that George and Gerry should have gotten themselves different evidence. Evidentialists about justiﬁcation can readily agree. This claim about what George and Gerry should have done is an evaluation of their conduct. Perhaps it is an evaluation for which the term ‘unjustiﬁed’ can be used: George and Gerry were unjustiﬁed in inquiring into (h) as they actually did. But this ‘justiﬁcation’ claim about their inquiries is independent of whether their belief in (h), under their actual conditions, has the sort of justiﬁcation that is necessary for knowledge (though intuitively they do not have enough of it to know, even if (h) had been true). We think that it is unproblematic to afﬁrm that their belief in (h) does have this sort of justiﬁcation, although they ought to have inquired differently. We also think that virtuous inquiry is not necessary for having all of the justiﬁcation that knowledge requires. This is illustrated by any example in which a conclusive case for some proposition is foist upon someone so forcefully as to overwhelm the person’s actively engaged intellectual vices. Jason’s other two problem cases for evidentialism, those of Daphne and Doris, seem to us not to raise fundamentally new issues. Daphne investigates (h), gets defeating evidence for her initial evidence in support of (h), but ‘she quickly and conveniently (though genuinely) forgets about or suppresses’ the defeating evidence. Doris likewise has initial support for (h) that is defeated by evidence she subsequently discovers, but she ‘distorts or misrepresents’ the defeating evidence, and the ‘result is that from her standpoint, the case for thinking that environmental smoke is hazardous is weak and her belief remains well-supported.’

314

REPLIES

Jason’s descriptions of what happens to the defeating evidence in the two cases do not make it clear to us whether or not Daphne and Doris still have it. If they do still have it, then the ‘suppressing’ that Daphne does must be understood as something like refusing to pay attention to it in her thinking about (h). The ‘distorting’ that Doris does must be understood as something like forming some entirely unjustiﬁed beliefs that incorrectly describe what the evidence is. Under these conditions, their belief in (h) is unjustiﬁed according to evidentialism, since their evidence does not support (h) on balance. Since Jason intends these cases to involve belief that is in some sense ‘unjustiﬁed’, this verdict would have us all agreeing about the cases. On the other hand, if the ‘suppressing’ that Daphne does is immediately and completely forgetting the defeating evidence without a trace of it remaining, then Daphne’s perspective is now just as though she never had the evidence. Her current view about whether or not (h) is true has only her initial supporting evidence available as truth-indicative guidance. Belief is thus what evidentialism classes as her justiﬁed attitude toward (h) and that seems quite right to us. Similarly, if Doris’s ‘distorting’ of the defeating evidence results in her now having different evidence that does not at all defeat the initial support for (h), then for Doris too it is as though she never had the defeaters. Belief in (h) is her justiﬁed and reasonable attitude toward it. Jason says that ‘there can be little doubt’ that Daphne and Doris’s beliefs are ‘unjustiﬁed’ in some sense of interest to evidentialists. But instead we see in this version of the case justiﬁed and reasonable beliefs in (h). We also see two people whose investigations of a proposition are badly done as a result of exercises of intellectual vices. Intermediate versions of the sorts of suppression and distortion of the evidence that the examples involve, with the result that their evidence is partially defeated, will be correspondingly mixed as to what the surviving evidence supports on balance. We see no shortcoming in what evidentialism implies about the mixed cases.

Response to Axtell Guy Axtell charges that our evidentialism has ‘weak roots’ and ‘sour fruit’. Although that phrase has a nice ring to it, we really don’t see that either metaphor applies. The ‘weak roots’ would be problems with evidentialism in the epistemological work to which it is directly addressed. The ‘sour fruit’ would be problems in extending the evidentialist perspective to broader issues, such as evaluating actual disagreements. We do not see either sort of problem. We have defended the thought that a proposition is known only if its belief is propositionally and doxastically justiﬁed. To explain doxastic justiﬁcation, we have proposed that this amounts to basing the belief on propositionally justifying evidence. Guy describes this as ‘bringing the basing relation into the analysis of knowledge by a back door’ and he alleges that this will not satisfy non-internalists. He cites John Turri as one who is not satisﬁed.

REPLIES

315

No fault in our account of doxastic justiﬁcation is suggested by anything in Guy’s paper. We cannot even see anything ‘back door’ about it. We have claimed that knowledge of a proposition requires its well-founded belief, and we have theorized that well-founded belief requires believing based on justifying evidence. Why don’t these claims bring basing into an analysis of knowledge by the main entrance? In any event, we cannot see any weakness here. Guy cites Michael Williams as complaining that a sort of view of justiﬁcation that Guy thinks includes our evidentialism is ‘an overly demanding hyper-intellectualized conception of what epistemic responsibility demands’. Guy says that this ‘problematizes’ our sort of view according to which propositional justiﬁcation is more fundamental than doxastic justiﬁcation. This claim seems to us to be far off the mark. First, evidentialism is a view about propositional and doxastic justiﬁcation; it is not about epistemic responsibility in belief or inquiry. We have offered neither evidentialism nor any other view about the nature of epistemically responsible inquiry. As for epistemically responsible belief, we think that this diverges from belief based on justifying evidence. For instance, in one sort of case of the difference, a person’s actually well-founded belief is epistemically irresponsibly believed, because the person has a justiﬁed and responsibly acquired Cartesian view of where his epistemic duty lies that calls for his believing only what is certain. Also, our view of justifying evidence is not hyper-intellectual. We think sensory states, memories, and feelings often justify beliefs without any reasoning. Guy classiﬁes as ‘a classic false dichotomy’ our claim that evidentialism provides no guidance for what an agent should do, but only for what the agent should believe. Guy does not elaborate much on this problem. He makes proximate mention of virtuous inquiry. This suggests that he thinks that sometimes what people should believe depends on what they should do, or at least, it depends on whether they have been doing what they should have been doing. We think that all the difference that proper inquiries make for the justiﬁcation that evidentialism addresses is what evidence the inquiries turn up, if any. Improper inquiries matter in exactly the same way. We think that this is a correct dichotomy between belief and inquiry. In the paper Guy gives no reason to think that we are mistaken about this. At one point in the ‘weak roots’ section of his paper Guy asserts ‘Conee and Feldman restrict the epistemic to relations of beliefs to evidence’. This is incorrect. As discussed in our reply to Jason Baehr, we take ‘the epistemic’ quite naively to mean ‘things especially pertaining to knowledge’. In this sense, truth and Gettier-prooﬁng are as ‘epistemic’ as is any relation between belief and evidence. It is in the context of his discussion of epistemic value that Guy makes this claim about our restricting of the epistemic. It is worth emphasizing that we are not committed to what Guy calls a ‘singlesource epistemic value monism’. Guy cites one of us as having held that ‘the fundamental epistemic goal is just having reasonable beliefs’. This is a claim about what is fundamental among goals. It is not a claim about what has any sort of value. We discuss epistemic value more in our reply to Jason Baehr. Again along the lines of our reply to Jason, it

316

REPLIES

seems to us that justiﬁcation may have a kind of special epistemic value in virtue of being one of the constituting conditions of knowledge. But if so then true belief seems to us equally to qualify for this sort of value. And knowledge itself has epistemic value, if anything does. Guy mentions understanding as something that ‘ranks high in [the] epistemic axiology’ of virtue epistemologists. If this sort of understanding is understanding why some fact obtains, then it seems to us to be knowing propositions that state an explanation of the fact. We concur that this knowing has whatever epistemic value is possessed by knowledge. Also, knowing that provides some understanding may have an additional sort of ‘epistemic value’, that is a value pertaining to knowledge, that not all knowledge has. We ﬁnd that plausible and nothing in our evidentialism conﬂicts with it. On the other hand, the less centrally knowledge is involved in understanding, the less clear it is to us that understanding has non-instrumental epistemic value. This clearly allows that understanding is very good in various other ways.

Response to Williamson Tim Williamson seeks to establish the possibility that a proposition, X, is known by someone, while the probability on that person’s evidence that X is known by that person is low, indeed, arbitrarily low. Though this would be no threat to our evidentialism, as we’ll explain imminently, it would be an extremely interesting result. We do not contest any of the formal ﬁndings that Tim develops. We do doubt that the sort of example that Tim offers can illustrate the possibility of improbable knowledge that he defends. We develop a different sort of illustration below. It would be a catastrophe for evidentialism about the justiﬁcation needed for knowledge if the following were true: it is possible for a proposition to be known by someone while the probability of that very proposition, given the person’s evidence, is low. At least this would be a devastating result for evidentialism, if the ‘probability’ in question were supposed to be a sort of epistemic probability to which one’s justiﬁcation makes all the difference. And this is the sort of probability that Tim clearly intends. But nothing in Tim’s paper has any tendency to support the possibility of a known proposition being improbable, given the knower’s evidence. Tim himself would not wish to defend this, since he holds that a person’s evidence just is the person’s knowledge, and any proposition has maximal probability on itself. What Tim defends is the improbability for someone that some proposition is known by the person, when in fact the person does know that proposition. This result would create no trouble for evidentialism. Nothing in evidentialism about the justiﬁcation needed for knowledge suggests that whenever we know, we must have good evidence, not only for the known proposition, but also for the proposition that we do know it. In fact, for all that is suggested by evidentialism, it could have been difﬁcult and rare to acquire enough evidence for any knowledge-attributing proposition. Conceivably, this could have been difﬁcult because the existence of knowledge also obviously depended on the meeting of some obscure further requirement, say, some remote historical causal

REPLIES

317

condition. It might have been typically improbable on a person’s present evidence that this historical condition obtained, even when it did. This sort of improbable historical requirement for knowledge is not plausible in the least. Our point is only that the evidentialism itself is silent in the matter of the existence of evidence of the sort that could help to give knowledge of knowledge. The only relevant evidentialist commitment is that having knowledge of knowledge of X requires having sufﬁcient supporting evidence that one does have knowledge of X. So if a person’s evidence makes it improbable that X is known by the person, then according to our evidentialism, the person does not know that X is known by the person. But nothing in our view suggests that the fact of someone’s having knowledge is always probable on the person’s evidence. Our doubt about the possibility of the sort of improbable knowledge that Tim defends can be seen to arise by trying to ﬁll out needed details in the perceptual example that Tim gives to illustrate the possibility. The example is lightly sketched in Tim’s paper. Even the known proposition is not stated. The example includes the following central features. A perceiver, S, sees the tip of a pointer at a location on a dial’s edge. The pointer tip moves in units of n microns around the dial, where n might have been any ﬁnite positive number. S’s powers of visual discrimination allow S visually to locate the pointer tip only to a precision of somewhere within, say, 5n microns to either side. So, wherever the pointer exactly points, at best S visually discriminates only that the tip is somewhere within an arc of 10n microns that is exactly centered on its location. The proposition that the pointer points somewhere in the relevant arc is the proposition that S knows. The true proposition that S has this knowledge of the pointer location proposition is intended to be the knowledge that S has that is improbable on S’s evidence. Before stating more precisely the pointer location proposition that S knows, we’ll describe more about how we take it that the example is supposed to work. It is S’s limited power of visual discrimination that is supposed to make the pointer location knowledge improbable given S’s evidence. So it must be that the evidence that S has makes this pointer location knowledge improbable. Tim identiﬁes a person’s evidence with the person’s knowledge. We take it that S is to have knowledge of some large number of possibilities that are left open by the limits of S’s power of visual discrimination. To get a speciﬁc improbability from this, S will have to know the relevant quantities. Thus, in the example as we have further speciﬁed it S will know that the pointer points at places separated by n microns around the edge of the dial and S will know that S visually locates where the pointer tip points only to somewhere up to 5n microns away from its actual location either way. So S will know there are 11 possible pointer tip locations that would appear visually to S in the same way as the pointer actually appears to S. The most speciﬁc visual pointer location knowledge that S has is supposed to leave these as epistemic possibilities that are open to S. These possibilities are supposed to provide for epistemic possibilities for S that make it improbable on S’s evidence that S has the actual pointer location knowledge. Assuming that this is how the example is supposed to work, one way to formulate in English S’s pointer location knowledge is this:

318

REPLIES

(L1) ‘It is pointing somewhere in there.’ We should understand S’s use of the demonstrative ‘there’ in L1 to be meant by S to identify S’s exact arc spanning visual discriminatory limits as S knows them to be. So S’s use of L1 expresses a proposition that is true exactly if the pointer tip is at any one of the 11 places on an arc on the dial edge that S’s visual acuity leaves open as its location. In light of the calibration of the arc on the dial edge that is demonstrated by S’s use of ‘there’ to S’s precise visual discriminatory limits, it is plausible that S could use visual perception of the pointer to know L1. It is also plausible that S could not know visually any proposition that is more locationally speciﬁc than is L1. The truth of any such proposition would exceed S’s stipulated power of visual discrimination. This maximum speciﬁcity of L1 is important to Tim’s purpose for the example. It precludes S from having other perceptual knowledge from which L1 would follow, potentially making S’s knowledge of L1 more probable, given that other knowledge. Call ‘KL1’ a proposition that attributes knowledge to S of what L1 expresses as S uses it. KL1 is improbable, on the evidence S has that consists of S’s knowledge of S’s visual experience and S’s visual discrimination limits. If the pointer had been at any one of the other 10 places that S’s visual acuity leaves epistemically possible, S’s visual evidence would have been the same. Yet the evidence S has does seem to sufﬁce for S to know L1, since L1 would still have been true on any of these epistemic possibilities. But on any of these alternative epistemic possibilities L1 would not have been known by S, because S would not have made the actual visual discrimination that supports the thought that S actually expresses by L1. S would have formulated the fruits of S’s visual observation of the pointer location with the same sentence type, but S’s use of ‘there’ would have had the sentence state a different proposition. The other proposition would have designated a span of 11 places that differs from the actually designated arc by at least n microns. So although KL1 is true, it is also a fact of the example that, given the visual evidence S has and S’s knowledge of S’s limited visual acuity, it is improbable that KL1 in particular is true. This evidence leaves it open that S has any one of 10 cases of slightly different knowledge concerning a slightly different region of the dial. If this is how the example is supposed to illustrate the possibility of improbable knowledge, then it fails to do so. The crucial point is this. In order for KL1 to be the improbable knowledge that the example is intended to illustrate, KL1 has to be improbable on all of S’s evidence. We concur that KL1 is improbable on S’s visual experience evidence and the evidence S has about S’s discriminatory powers. But S has other relevant evidence. S has the evidence of S’s demonstrative intention by which S’s use of ‘there’ in L1 gets its reference. This intention is evidence for S about the content of L1. By hypothesis, the intention has the effect of making L1 assert that the pointer tip is in whatever arc on the dial edge that S’s visual discriminatory capacity enables S to isolate as the region in which the pointer tip is located, though S need not conceive of it those terms. Given this additional evidence S has, evidence that tells S that the proposition that L1 actually expresses refers to the actually discriminated arc,

REPLIES

319

and given that it is S’s use of the sentence by which L1 is formulated that expresses S’s thought that KL1 asserts S to know, KL1 must be true as long as the demonstrative does not miss its mark and the perceptual experience is not deceptive. S has no grounds for thinking that either of these remote contingencies obtains. So, with this further evidence, KL1 is considerably more probable for S, given S’s total evidence, than is any of the propositions asserting knowledge by S of other pointer locations that S might have expressed instead with L1. We do not think that having additional evidence concerning the knowledge proposition is an accident of the example as we have ﬁlled it in. Tim does not discuss in the paper the need for evidence in support of X in order to know X. He may deny this need. But of course that need is part of our evidentialist view. And presumably Tim intends his case for improbable knowledge at least to be compatible with the core evidentialist requirement that when someone knows a proposition, the totality of the person’s relevant evidence supports it. Given this requirement, knowing a proposition can be improbable, on the person’s total evidence, only if the total evidence does two things at once. It must make the knowledge of the proposition an unlikely epistemic alternative for the person and it must make the proposition itself likely enough to be justiﬁed. Yet these two might be incompatible (though we will argue soon that they turn out not to be incompatible). It might be that whatever could make it unlikely to S that S knows X, would be evidence that would also defeat S’s evidence for X itself. This clearly happens whenever the proposition attributing to S knowledge of X is made unlikely to S by evidence making X itself unlikely to S. S then does not meet the evidential requirement for knowing X. So there cannot be any such case of improbable knowledge. It is nearly as clear that the same goes for the justiﬁcation condition. Evidence supporting to someone that it is unlikely that the justiﬁcation condition is met for knowledge of X seems to defeat whatever justiﬁcation for X the person might have. It is not clear how some evidence E could make X likely for someone, while some other evidence that the person had made it unlikely for the person that anything, including E, justiﬁed X to the person. Could S’s knowledge of X be made unlikely to S via evidence that S is unlikely to be believing X in particular, from among some rivals for what S believes, although by hypothesis S does in fact believe X in knowing it? We shall defend this possibility. The question raises complex issues involving the nature of belief and its content. We note that a large obstacle exists to establishing the possibility of improbable knowledge in this way. The readiest apparent cases of believing something without knowing what proposition is believed involve the use of names or demonstratives. For instance, a ﬂuent English speaker with closed eyes points somewhere after spinning around in the middle of a dense forest and afﬁrms, ‘There are trees in that direction.’ The sentence has many potential content propositions that are about different directions. It seems that a ﬂuent speaker knowing herself to be in a dense forest would know the truth of what is thereby said. It also seems that the person would not know which potential content

320

REPLIES

proposition was actually expressed from among many candidate propositions about the many equally probable different directions. If things are as they seem in both respects, then the example shows initial promise as an instance of improbable knowledge. The large obstacle to using this sort of example is this. What most clearly makes it seem that the person knows the truth of what is said, even though the content of the sentence has those epistemic alternatives, is a metalinguistic reading of what is known. The content of the knowledge would be something along the lines of the proposition that the person would express by saying ‘My use just now of this sentence about the trees states a truth.’ But that meta-linguistic proposition is true regardless of which of the potential contents the demonstrative in the sentence actually has. So it is not made improbable by the existence of those epistemic alternatives. What would make those potential content propositions about different directions rivals for being the content of the belief would be its being a belief in the proposition that is actually expressed by the sentence. So let’s suppose that the person does indeed believe the content under such conditions. If so, then it seems also to be true that the person can know by reﬂection what proposition the person believes. The identity of the belief would be available to the person by reﬂectively noting its content by use of the same sentence that we are now supposing to be understood well enough for its content to be believed. This reﬂectively available evidence concerning what is believed undercuts the idea that the ‘rival’ content propositions for the sentence make it improbable to the person that it is the actual content that is believed. Rather, it seems that the person would also have reﬂectively available evidence that is enough to know what is believed. It remains true that the person is unable to discern the belief from others, via some other descriptions of the direction that is demonstrated. But that does not reduce the probability for the person that the believed proposition is the one that the evidence available by reﬂection supports that it is. So this too would not be a case of knowledge where the person’s total evidence makes it improbable that the known proposition was believed. That improbability could be otherwise supported, however. Someone might be seriously misinformed about whether the propositional attitude that the person has toward a proposition is believing, while knowing that belief is required for knowledge (or knowing that either belief or acceptance is required, or that it is a strong enough conﬁdence, or whatever exactly is the actual doxastic necessary condition on knowledge—we’ll continue to call the condition ‘belief ’). Suppose that, while knowing that trees exist, S has received recognized expert testimony to the effect that what S is doing with regard to the proposition that trees exist is not believing it. Rather, the expert assures S that S is ‘under a sort of illusion in which mere consideration of a proposition reﬂectively simulates belief ’. S would also have evidence for S’s believing that trees exist from S’s reﬂective access to a state that is in fact belief. But that reﬂective evidence might be overridden by the convincing expert assurance. S’s totality of evidence might support denying that the attitude is belief. By inference from this support and S’s knowledge of the

REPLIES

321

requirement of belief for knowledge, S might have a totality of evidence making it improbable to S that S knows that trees exist. We can suppose that S has a typical abundance of perceptual and testimonial evidence for the existence of trees. Nothing in S’s misleading evidential grounds for the improbability of S’s knowing that there are trees defeats this evidence that trees exist. All that is misleading in that evidence pertains to whether it is belief that is S’s attitude toward the proposition. There need not be anything else that keeps that proposition from being known by S. So this sort of case does illustrate the possibility of improbable knowledge.

Response to Dougherty Trent Dougherty defends the thesis that all evidence consists in propositions. We have argued that evidence includes experiences. This seems most intuitive to us, and it appears to have no signiﬁcant cost. But we note that the general doctrines of evidentialism do not depend on any evidence being non-propositional. One argument Trent offers for the conclusion that evidence consists entirely in propositions is this: evidence is the sort of thing that hypotheses explain, but what hypotheses explain is propositional, hence, evidence is propositional. Note, ﬁrst, that there is a difference between the intended conclusion—that evidence consists in propositions— and the conclusion of this argument—that evidence is propositional. We take it that beliefs are propositional, in that they take propositions as objects, but they are not themselves propositions. Trent makes clear, however, that he intends to defend the view that evidence consists in propositions, not just that evidence is something that has propositional content. We believe that propositions formulate in a helpful way arguments that give evidence. But propositions as standardly understood are not identical to at least some of the evidence that people get. For instance (an example that Trent cites from Timothy Williamson), a scratchy sensation in one’s throat is evidence that one is getting a cold. Tim says that this is so ‘in the sense that the hypothesis that I am getting a cold would best explain why I have that sensation in my throat.’ We respond that although asserting the proposition that I have that sensation in my throat is a good way to state what some of one’s sensory evidence is, it cannot be identical to the sensory evidence. The sensory evidence exists only while the sensation is felt, while the proposition is eternal, or at least available at other times when it is not evidence. The proposition could have existed while not true and it would not have been any evidence at all of a cold. The proposition reports the evidence or represents it. But it is not identical to the evidence. Trent says that experiences are not evidence but rather are ‘that by which we gain our evidence’. No doubt he’d respond to our previous point by saying that when one is not having the appropriate experience, one does not have as evidence the proposition that one has a certain sensation. Trent acknowledges that the ordinary way of speaking of these matters implies that the sensations themselves are evidence, and thus seems to conﬂict with the sort of response that we just suggested on his behalf. However, Trent thinks that the arguments he reviews show that our ordinary talk is ‘inelegant’ and

322

REPLIES

‘uneconomical’ and should be reinterpreted. We see no reason not to take the locations at face value. Also, getting the evidence by having the sensation is clearly different than getting the evidence by being reminded of having had the sensation. The experience of the sensation on the one hand, and the reminder on the other, seem to be alternative sorts of evidence for the proposition that one has the sensation at the time. If so, then the sensation is itself evidence. To see why we think there is no reason to abandon the ordinary conception, consider the argument Trent discusses that it is propositions that are explained. We will use another example from Tim Williamson to examine this, that of explaining World War I. As we see it, explaining the occurrence of World War I is not explaining any proposition. It is explaining a gigantic complex spatiotemporal occurrence that included thousands of battles and millions of people. The occurrence of the one huge event, WWI, that is the combination of all of these things, can be explained by giving an argument the conclusion of which read ‘and so WWI occurred’. The last two words do express the proposition that WWI occurred. That proposition is the conclusion of the explanatory reasoning. What is explained, though, is not the proposition, but the occurrence of the war that the proposition asserts to have occurred. Concluding that sort of explanatory argument is not identical to explaining the proposition in it that is the conclusion. It is explaining the worldly phenomenon that the proposition reports or represents. But this difference between the proposition that represents or reports what it is evidence for, and the explained phenomenon that is thereby represented, is often harmlessly ignored. Trent, following Tim, insists that the proper form of an explanatory statement is ‘X is the case because Y is the case’. ‘X is the case’ and ‘Y is the case’ refer to a proposition being explained and a proposition that does the explaining. We do not see the basis for this insistence. We think that those propositions respectively assert the explained worldly phenomenon, and the explaining worldly conditions. The explaining consists in using those propositions to make that sort of causal claim about those aspects of the world. That all seems to us to be unproblematic common sense. We see no good reason to identify either the explained things, or the explaining things, with the propositions that assert them. At the end of his paper Trent assumes that in his example of Ted and Todd, what they get as evidence from their differing visual experiences of the cabs is the same proposition. It is part of the example that Ted is better justiﬁed than Todd in the cab color judgment that they both make, because Ted’s experience of the cab’s color is more vivid. Trent thinks that this difference in justiﬁcation can be accommodated by EJ (Doxastic attitude D toward proposition P is epistemically justiﬁed for S at t if and only if having D toward P ﬁts the evidence S has at t), but not by ES (The epistemic justiﬁcation of anyone’s doxastic attitude toward any proposition at any time strongly supervenes on the evidence that the person has at the time), by claiming that a stronger belief in the cab color judgment ‘ﬁts’ Ted’s evidence better than Todd’s. In response we contend that if the evidence that Ted and Todd both got were just the one proposition that Trent thinks it is, then there could be no difference in what

REPLIES

323

attitude ﬁts that evidence. The only apparent way to save the thought that there is this difference in ﬁt, along with Trent’s other judgments about the example, is to say that stronger belief ‘ﬁts’ the proposition ‘relative to Ted at the time’, and to explain this by saying that ﬁt for a person at a time depends on something that Trent’s propositional view counts as non-evidence—the person’s experiences. In virtue of relying on what is supposed to be a non-evidential difference, that reading of EJ is non-evidentialist in spirit, if not in letter. In order to have the evidence be all propositional in this case, it would be better to say that different propositions are the evidence given by the different experiences: Ted’s proposition ascribes a vivid color experience to him and Todd’s ascribes a dimmer one. Of course we think that it would be better yet to say that their evidence differs because it includes their differing color experiences.

Appendix A. Chart Summary of the Arguments This chart is a rough and ready summary of the arguments in this volume. It is meant as a guide to pre-readers as an overview and tantilizer, and to post-readers as an aid to memory. It is not meant to represent the depth and richness of the entries it necessarily oversimpliﬁes in draconian fashion. Note that the distinction between an objection and a suggestion is blurry and somewhat arbitrary.

AUTHOR

OBJECTION/SUGGESTION

REPLY

Axtell, Guy

Objection 1: Evidentialism overintellectualizes justiﬁcation.

Reply to obj. 1: Evidentialism grounds justiﬁcation in experiences and memory without any need for secondorder thought. Reply to obj. 2: They are two genuinely different activities. Evidentialism provides a role for action to play in justiﬁcation: inquiry can turn up evidence. Reply to obj. 3: The epistemic in an ordinary sense encompasses any relation to knowledge. Reply to obj. 4: Anything pertaining to knowledge might have epistemic value. Justiﬁcation is a fundamental value but not the only value.

Objection 2: Evidentialism creates a false dichotomy between what to do and what to believe.

Objection 3: Conee and Feldman restrict the epistemic to relations of beliefs to evidence. Objection 4: Conee and Feldman are committed to single-source epistemic value monism. Baehr, Jason

Suggestion: Add a virtue component to evidentialism to handle the following objection.

Objection: The Cases of George and Gerry: Inquiry Gone Wrong. Someone might believe in accordance with the evidence they have, but they may have inquired with vice.

Reply 1. If someone really only has E as evidence, and what E supports is p, then what they ought to believe is p, though they may be blamed for only having E as evidence. Reply 2. It can’t be that making some causal contribution to knowledge can qualify something as an epistemic desideratum, for having hope or having children might do that. But the success of inquiry seems just as contingent. Reply 3. Evidential justiﬁcation has ‘contributory epistemic value’. I.e. it is

S U M M A RY O F A R G U M E N T S

AUTHOR

325

OBJECTION/SUGGESTION

REPLY

Warning: If evidentialists don’t bring in virtue, it’s hard to see the value of the property.

part of what constitutes knowledge. It also comprises, at least in part, reasonable belief. Reply 4. Though evidential justiﬁcation is required for knowledge, it seems virtue is not, for someone might have knowledge foisted upon them unwillingly.

Bergmann, Michael

Objection: Crazy views can mimic the seeming evidentialist’s reply to the Great Pumpkin Objection (bad news).

Crazy views can’t mimic the claim that we have evidence against skepticism that provides stronger justiﬁcation against it than for it. That claim is reasonable whereas the following is not: We have conjectures against skepticism that provide stronger justiﬁcation against it than for it.

DeRose, Keith

Reply to obj. 1: There are several other Objection 1: The Case of Henry and Bad Basing. Henry has great evidence ‘ought nots’ which explain what’s for p, but bases his belief in p on lousy wrong with Henry’s doxastic practices here, but he’s believing the proposition evidence q. So Henry epistemically ought not believe p, even though that’s he ought to believe, just for the wrong reason. what ﬁts his evidence, and so evidentialism says Henry epistemically ought to believe p. Objection 2: Irresponsible Henry. Reply to obj. 2: There is another normative judgment which correctly describes Henry: He ought not to have only the evidence he has. But given that that’s all the evidence he has, he’s believing what he epistemically ought. Reply to obj. 3: Intuitively, sensory Objection 3: We lack any evidence experience is evidence for external favoring ordinary beliefs over world beliefs. There are puzzles skepticism. concerning the nature of this support relation, but that is a separate matter. Analogy: Intuitively, instances tend to conﬁrm corresponding generalizations, yet there is no agreed upon solution to the problem of induction. The puzzle shouldn’t defeat the intuition.

Dougherty, Trent

Suggestion: Go with a propositional theory of evidence, i.e. that evidence consists in propositions.

Fumerton, Richard

Concern 1. It is possible that most justiﬁed beliefs are false, so why care about the property.

Reply: Propositions just describe what evidence one has. All the functional roles of evidence can be played by experiences. Comfort 1. Though it is possible, we have no reason to think it’s true.

326

APPENDIX A

AUTHOR

OBJECTION/SUGGESTION

Objection 1. If ‘epistemically likely’ just means ‘epistemically justiﬁed’, then this seems uninformative. Goldman, Alvin

Suggestion: Add reliabilist elements to evidentialism to help with The Case of Ichabod and with cases of stored belief. Objection: There are plausible cases of (at least possible) knowledge without mental states in zombies and in blind sight (not to be confused with blindsight).

Greco, John

Objection: Contra ‘knowledge evidentialism’: A. There are three kinds of knowledge without evidence: 1. Memory knowledge, 2. A priori knowledge, and 3. Knowledge via blindsight.

REPLY Comfort 2. Justiﬁcation is valuable as a necessary condition on knowledge. Comfort 3. Reasonable/rational responses might be intrinsically valuable, apart from truth. Comfort 4. Justiﬁcation might be a part of other epistemic virtues like understanding or wisdom. Comfort 5. Even if most justiﬁed beliefs were false, it might be that more justiﬁed beliefs were true than unjustiﬁed beliefs. Reply to obj. 1: The evidentialist understanding of the relationship between ‘justiﬁed’ and ‘likely’ is a substantive semantic thesis. Reply 1. For dispositional beliefs, the evidence consists in dispositional mental states, so there is no threat to the mentalist supervenience thesis. Reply 2a. In the case of blind ‘seeing’ in all interpretations there seems to be some kind of sensory environmental feedback in which their evidence consists. Reply 2b. Either there can be unconscious perceptual states or there cannot. If there can be, then these can serve as evidence for the zombies’ beliefs. If there cannot be, then the zombies’ beliefs are unjustiﬁed (if zombies can have beliefs). Reply to obj. A1: A sense of recollection is evidence present for many memory beliefs, the strength of which varies by vividness. Reply to obj. A2: A proposition’s seeming true (in some sense) is evidence for it. Reply to obj. A3: If blindsighters experience their guesses as guesses, then they thereby have a knowledge defeater. If they learn of their reliability, then this information is evidence.

S U M M A RY O F A R G U M E N T S

AUTHOR

Huemer, Michael

327

OBJECTION/SUGGESTION

REPLY

B. Knowledge requires a ‘modally strong’ connection to the world (e.g. reliability).

Reply to obj. B: All kinds of unreliable processes can lead to success, and the generality problem for reliabilism works against this objection as well.

Observation: Mentalists are under pressure to be neutralists about how evidence counts interpersonally. Neutralism is, roughly, the thesis that it doesn’t matter evidentially for S—with respect to some p—whether S herself has evidence for/against p or whether S merely knows that S0 has evidence for/ against p.

Counter-observation 1. It could be that it’s impossible to know another’s mental states as well as one’s own, so that Neutralism would be vacuously true and Centrism would be extensionally correct.

Counter-observation 2. The states Having M and Knowing S has M are two different mental states, and so according to mentalism, two different items of evidence. Kvanvig, Jonathan

Concern: A position on peer disagreement advocated by Feldman seems close to molliﬁcationism, which is contingently self-refuting.

Suggestion: An epistemic role ought to be give to relative conﬁdence in oneself as a cognitive being. Molliﬁcationism is, roughly, the thesis that learning of a differing attitude by an epistemic peer toward a proposition is some reason to back off one’s existing attitude toward that proposition and move in the direction of your peer. Lehrer, Keith

Suggestion: Evidentialists should see themselves as being able to solve peer disagreement in one’s favor, for one’s acceptance of p is evidence for p.

Pritchard, Duncan

Suggestion 1: Deﬁne evidentialism in terms of Access rather than Mentalism.

Reply 1. The most plausible version of molliﬁcationism is not self-defeating because it will take into account the distribution of agreement and disagreement across a relevant population, and plausibly, molliﬁcationism is entailed by the majority of positions of epistemologists. Reply 2. It may well be that some true positions are such that one can’t be rationally justiﬁed in believing them. That is no evidence against their truth. Reply 3. Self-trust appears to be a nonepistemic emotional factor. Self-trust can be rational or irrational depending on whether it is based on good evidence.

Reply: Second-order considerations might tip the scales in a ﬁrst-order stalemate, but that works both ways. Successful reﬂection can also restore parity. Ad 1: Beings capable of object level beliefs but no reﬂective capabilities should count as being able to know.

328

APPENDIX A

AUTHOR

OBJECTION/SUGGESTION

REPLY

Suggestion 2: Combine evidentialism with disjunctivism so that veridical evidence supports ordinary propositions, establishing an asymmetry between Cartesian demon worlds and veridical worlds, counting occupants of the former as blameless due to indistinguishability, even though they have inadequate support for their beliefs.

Ad 2a: How can the situations be indistinguishable if the states share no content?! Ad 2b: How can they be held blameless if they believe for inadequate reasons? Ad 2c: It is independently very natural to appeal to the way things look to justify beliefs. Ad 2d: All disjunctivism offers is evidence that entails the truth of the target belief, but—for reasons familiar in math and logic—that does not by itself put one in possession of epistemic support. Ad 2e: Having a veridical entailing perception doesn’t entail having epistemic support for another reason: One can perceive an elm, yet not be justiﬁed in believing that there’s an elm, because the person might lack the concept of elm or be unreliable in identifying them.

Rysiew, Patrick

Suggestion: Adopt a more Reidian theory of evidence.

Objection 1: The account seems to include truth in being evident. This conﬂicts with the existence of justiﬁed false beliefs. Objection 2: The account of selfevident truths as being evident but not having supporting evidence conﬂicts with evidentness arising from that which makes evident: evidence.

Steup, Mattias

Suggestion: Argue against BIV skepticism that the evidence makes it beyond a reasonable doubt that there are no brains in vats.

Swinburne, Richard

Reply: The argument is dialectically ineffective as it stands: The premise that we have such-and-such evidence stands in need of evidence. The deﬁnition used for begging the question facilitates overlooking this inadequacy. Suggestion: Adopt a doxastic theory of Objection 1. Inclinations can fail to evidence. But be liberal enough in this have any rational basis. to count inclinations to believe as evidence.

S U M M A RY O F A R G U M E N T S

AUTHOR

OBJECTION/SUGGESTION

329

REPLY Objection 2. Inclinations to believe can be inconsistent in ways that background evidence and conditional probabilities fail to correct.

Williamson, Timothy

Concern: It is possible to have knowledge that p when the epistemic probability that one knows that p is extremely low. This doesn’t seem to set easily with evidentialism.

Reply 1: Such cases pose no threat whatsoever to evidentialism. Evidentialism has no entailments concerning knowledge of knowledge. Reply 2: Williamson’s case is not successful, due to the natures of the proposition known and demonstratives. Suggestion: There is in fact such a case, however, but it involves having evidence that one is not believing what one is actually believing.

Appendix B. Bibliography of Conee and Feldman since the Publication of Evidentialism: Essays in Epistemology Evidentialism: Essays in Epistemology, published by Oxford University Press in 2004, collected together a number of important essays in epistemology—some hard to ﬁnd in their original locations—and bookends them by two new essays. A substantial amount has been written by the duo which appeared concurrently with or after that volume, again, some of it hard to ﬁnd in their original sources. I list here some of that relevant work. In many cases, I make brief annotations. Conee, Earl. 2011. ‘Self-Support.’ Philosophy and Phenomenological Research. Identiﬁes the evidence in apparent cases of self-evidence and argues that it does not justify. ——.2010. ‘Rational Disagreement Defended,’ in Richard Feldman and Ted Warﬁeld (eds), Disagreement. Oxford: Oxford University Press. Argues for suspending judgment in all mature expert peer disputes. ——.2009. ‘Peerage.’ Episteme, 6 (3): 313–23. Argues against the possibility of reasonable disagreement among epistemic peers involved in ‘mature’ disagreements. ——.2009. ‘Criterial Problems.’ Philosophical Studies, 143 (3): 417–26. A discussion of Ernest Sosa’s discussion of the problem of the criterion and epistemic ascent in his A Virtue Epistemology. ——.2008. ‘Critical Notices: Epistemology and the Psychology of Human Judgment.’ Philosophy and Phenomenological Research, 77 (3): 837–40. ——.2007. ‘Disjunctivism and Anti-Skepticism.’ Philosophical Issues, 17 (1): 16–36. Argues that disjunctivism does not solve the skeptical puzzle. Disjunctivism has many problems of its own. ——.2007. ‘Externally Enhanced Internalism’ in Sanford Goldberg (ed.), Internalism and Externalism in Semantics and Epistemology. Oxford: Oxford University Press. ——.2007. ‘Review of Jonathan Sutton, Without Justiﬁcation.’ Notre Dame Philosophical Reviews (12). Contrary to Sutton, there are false justiﬁed beliefs. ——.2005. ‘Contextualism Contested,’ in Contemporary Debates in Epistemology. Oxford: Blackwell. Contra contextualism in debate with Stewart Cohen. ——.2005. Riddles of Existence (with Ted Sider). Oxford: Oxford University Press. ——.2004. ‘Externalism, Internalism, and Skepticism.’ Philosophical Issues, 14 (1): 78–90.

BIBLIOGRAPHY OF CONEE AND FELDMAN

331

Externalist attempts to exclude skeptical possibilities leads to a form of skepticism. Skeptical possibilities do not need to be eliminated, though, for internal conditions are sufﬁcient for meeting any relevant requirement. ——.(2005). ‘The Comforts of Home.’ Philosophy and Phenomenological Research, 70 (2): 444–51. Reply to Williamson on ‘cognitive homelessness.’ Conee, Earl and Richard Feldman. 2008. ‘Evidence,’ in Quentin Smith (ed.), Epistemology: New Essays. Oxford: Oxford University Press. Addresses some of the main issues in ﬁlling in the details of the evidentialist thesis including the nature of evidence—with replies to Williams and Willamson—the evidential support relation, and what evidence there is in cases of memory and the a priori. Feldman, Richard. 2009. ‘Evidentialism, Higher-Order Evidence, and Disagreement.’ Episteme, 6 (3): 294–312. Puzzles about disagreement provide no reason to doubt evidentialism. ——.2008. ‘Modest Deontologism in Epistemology.’ Synthese, 161 (3). Gives truth conditions for deontological sentences about belief which do not assume strong doxastic voluntarism. ——.(2007). ‘Knowledge and Lotteries.’ Philosophy and Phenomenological Research, 75 (1): 211–26. Critical review of John Hawthorne’s Knowledge and Lotteries. ——.2006. ‘BonJour and Sosa on Internalism, Externalism, and Basic Beliefs.’ Philosophical Studies, 131 (3). Discussion of Epistemic Justiﬁcation: Internalism vs. Externalism, Foundations vs. Virtues, Laurence BonJour and Ernest Sosa. ——.2006. ‘Clifford’s Principle and James’s Options.’ Social Epistemology, 20 (1): 19–33. A discussion of the Clifford–James debate. ——.(2006). ‘Epistemological Puzzles About Disagreement.’ In Stephen Hetherington (ed.), Epistemology Futures. Oxford: Oxford University Press. ——.2005. ‘Review of William P. Alston, Beyond “Justiﬁcation”: Dimensions of Epistemic Evaluation’. Notre Dame Philosophical Reviews, (8). ——.2005. ‘Respecting the Evidence.’ Philosophical Perspectives, 19 (1): 95–119. Discussion of the effects of gaining higher-order evidence. ——.‘Justiﬁcation is Internal.’ In Steup and Sosa (eds), Contemporary Debates in Epistemology. Oxford: Blackwell. Contra John Greco. ——.(2004). ‘Comments on DeRose’s “Single Scoreboard Semantics”.’ Philosophical Studies, 119 (1–2). ——.2004. ‘Foundational Beliefs and Empirical Possibilities.’ Philosophical Issues, 14 (1): 132–48.

332

APPENDIX B

Argues against traditional foundationalism on behalf of the thesis that justiﬁed basic beliefs can be simple external world beliefs. ——.2004 ‘In Search of Internalism and Externalism,’ in Richard Schantz (ed.), The Externalist Challenge, pp. 143–56. Berlin: Walter de Gruyter. ***This is a much neglected paper due to being in a volume not widely circulated. It provides a historical study of the origins of the debate, discusses what kind of epistemic status internalists wish to analyze, and, most interestingly, a discussion of ‘linking principles’ requiring justiﬁed believers to be justiﬁed in believing their grounds support their beliefs. Feldman rejects such requirements. Feldman, Richard and Earl Conee. 2005. ‘Some Virtues of Evidentialism.’ Veritas, 50 (4): 95–108. Along with ‘Evidence’ in the Smith volume, this is an essential piece for ﬁlling in some of the details of evidentialism. It discusses the truth connection, a priori knowledge, skepticism, and the nature of defeat.

Index Alston, William 49, 50, 89, 102, 167, 172, 195, 205, 213, 224, 311, 331 Aumann, Robert 30, 32 Austin, J. L. 163, 216, 224 Axtell, Guy vii, ix, x, 2, 3, 8, 10, 16, 69, 70, 71, 74–7, 85, 102, 135, 314, 324

Dougherty, Trent iii, vii, viii, ix, x, 1, 3, 4, 8, 9, 12, 13, 32, 92, 101, 132, 172, 193, 226, 251, 283, 321, 325 Douven, Igor 46 Dretske, Fred 106–9, 111, 112, 114, 122 Elga, Adam 25, 32, 41, 43, 84, 86 Elgin, Catherine 21, 32 Engel, Mylan 121, 177 Erasmus 84

Bach, Kent 221, 224 Baehr, Jason v, vii, viii, x, 2, 3, 8, 10, 70, 73, 88, 99, 101, 102, 310, 315, 324 Ballantyne, Nathan 67 Basinger, David 79, 80, 86 Bender, John 178 Bergmann, Michael vii, x, 2, 114, 123, 131–3, 177, 180, 185, 221, 222, 224, 283, 325 Berkeley, George 200 Bernstein, Richard 79, 86 Biederman, Irving 274, 275, 278 Bishop, John 79, 86 Block, Ned 271, 279 BonJour, Laurence 9, 12, 13, 21, 32, 88, 95, 96, 101, 102, 200, 201, 205, 221, 222, 224, 270, 279, 331 Broad, Charlie 200 Bunny, Easter 121 ff Byrne, Alex 241, 251

Fairweather, Abrol 178 Fantl, Jeremy 19, 32 Feldman, Richard viii, ix, x, 1, 4–16, 21–3, 25, 32, 41–3, 51, 55, 67, 69, 71, 72, 74–88, 91, 95, 96, 102, 103, 106, 122–5, 133, 135, 137–44, 146, 165, 166, 168, 169, 172, 173, 176, 177, 179, 181, 182, 191, 193, 195, 196, 199, 203–6, 208, 210, 211, 215, 216, 220, 221, 224–6, 228–37, 239–42, 246, 251, 252, 254–60, 264–70, 272, 273, 274, 276–9, 283, 303, 304, 315, 324, 327, 330–2 Firth, Roderick 5, 14, 59, 67 Foley, Richard 28, 32, 69, 199, 206 Fumerton, Richard viii, x, 4, 121, 165, 179, 187, 191, 296, 325

Chappell, Timothy 86 Chisholm, Roderick 3, 4, 9, 10, 12, 13, 55, 67, 75–7, 88, 102, 186, 214, 224, 259, 269, 276, 279 Christensen, David 25, 32, 43, 47, 48, 51 Clifford, William 78, 79, 86, 209, 214, 331 Cohen, Steward 106, 119, 122, 211, 224, 236, 252, 331 Comesana, Juan 254, 256, 270, 271, 279 Conee, Earl viii, ix, x, 1, 4–15, 21, 22, 23, 32, 55, 67, 69, 71, 72, 74, 75, 76, 79, 86, 88, 91, 95, 96, 102, 103, 104, 106, 122–30, 132, 133, 135, 137, 140–4, 165, 166, 168, 169, 172, 173, 176, 177, 179, 181, 182, 186, 191, 193, 195, 196, 199, 203, 205–8, 210, 211, 214–16, 218, 220, 221, 223–6, 228–37, 239–42, 246, 248–52, 254–60, 264–70, 276–9, 283, 284, 303, 304, 314, 324, 330–2 Cowen, Tyler 30, 32

Garber, Daniel 199 Geach, Peter 93, 102 Gibson, J. J. 271, 279 Goldbach, Christian 198 Goldman, Alvin viii, x, 2, 4, 8, 20, 32, 64, 67, 102, 184, 185, 189, 191, 209, 225, 233, 234, 237, 251, 254, 256, 258, 260, 261, 263, 267, 269, 272, 276, 278, 279, 294, 302–4, 326 Graham, George 222 Graham, Peter J. 225 Greco, John viii, xi, 2, 8, 14, 73, 74, 80, 86, 90, 92, 93, 102, 133, 135, 164, 165, 167, 168, 170, 175, 208, 213, 217, 221, 225, 252, 285, 303, 326, 332 Greenough, Patrick 48, 164, 240, 252 Grice, H. P. 245, 252 Gutmann, Amy 83, 86

Delgarno, M. 224 DeRose, Keith v, vii, viii, x, 2, 3, 8, 70, 107, 122, 135, 137, 287, 325, 332

Haddock, Adrian xi, 33, 87, 241, 245, 251–3 Hanson, Robin 25, 30, 32 Harris, Sam 81, 82, 84, 86

334

INDEX

Hawthorne, John 33, 43, 105, 158, 160, 163, 164, 331 Hazlett, Alan 80, 86 Heil, John 177 Henderson, David 170, 178, 185, 186, 191 Himma, Kenneth 79, 86 Hinton, J. M. 241, 252 Horgan, Terence 170, 178, 185, 186, 191 Howson, Colin 198, 206 Huemer, Michael vii, viii, xi, 8, 10, 14–17, 32, 186, 191, 298, 327 Hume, David 3, 56, 67, 78, 175, 209, 225 Hursthouse, Rosalind 93, 102 James, William 67, 68, 69, 78, 85, 86, 168, 225, 331 Jeffrey, Richard 30, 31, 33, 199 Kelly, Thomas 25, 33, 43, 207–13, 216, 219, 221–3, 225, 255, 279, 303 Kim, Jaegwon 255, 269 Kornblith, Hilary 8, 14, 102, 135, 177, 191, 224, 279 Kraft, James 80, 86 Kvanvig, Jon vii, viii, xi, 8, 11, 14, 16, 34, 177, 299, 325 Lehrer, Keith vii, vii, xi, 8, 16, 21, 28, 33, 51, 55, 56, 58, 59, 60, 63, 66, 67, 176, 177, 178, 213, 215, 225, 236, 252, 270, 279, 306, 327 Leite, Adam 121 Lewis, C. I. 59, 68, 183, 191, 226 Lewis, David 198, 206 Locke, John 3, 65, 79, 80, 83–6, 200, 209, 216, 225 Logue, Heather 241, 251 Lyons, Jack 270, 271, 279, 305 Macpherson, Fiona 33, 241, 245, 251, 252, 253 Marr, David 274, 280 Martin, Michael 241, 252 Matthews, E. 224 McDowell, John 241, 242, 243, 245, 247, 248, 251, 252, 253 McGrath, Matthew 19, 32 McKim, Robert 79, 83, 86 Mill, John Stuart 65, 68 Millar, Alan xi, 87, 245, 249, 251, 252 Moffett, Marc 43 Montaigne 84 Morris, Stephen 32, 33, 121 Neta, Ram 33, 248, 251, 252 Norcross, Alastair 121 Nozick, Robert 106 Nuccetelli, Susana 237, 252

Olson, Philip 74, 76, 86 Page, Scott 65, 68 Pappas, George 32, 86, 225, 279 Plantinga, Alvin 4, 14, 15, 38, 64, 68, 74, 88, 102, 123, 134, 170, 178, 197, 202, 204, 205, 216, 221, 225 Pollock, John 43 Price, Richard 200 Pritchard,D.H. viii,xi,4,8,48,87,164,193,224, 233,235,240,241,244,248,251,252,289,327 Rawls, John 80, 83, 86 Reid, Thomas 10, 32, 56, 63, 64, 67, 68, 145, 193, 207, 208, 212, 213, 215–25, 279, 292, 293, 294, 228 Rorty, Richard 78, 86 Rousseau, Jean Jacques 80, 87 Russell, Bertrand 187, 191 Russell, Bruce 3, 121, 177, 187 Rysiew, Patrick viii, xi, 4, 8, 10, 193, 207, 209, 212, 213, 217, 223, 225, 292, 328 Schefﬂer, Samuel 17, 33 Sellars, Wilfrid 176, 177, 200, 206, 272 Smith, Holly 232, 278 Smith, Quintin 5, 13, 32, 164, 193, 232, 279, 291, 292, 331, 332 Snowdon, Paul 241, 253 Sosa, Ernest 9, 12, 13, 14, 32, 74, 77, 102, 104, 109, 122, 123, 133, 178, 205, 221, 224, 225, 232, 303, 330, 332 Speaks, Jeff 273, 280 Stalnaker, Robert 256, 280 Steup, Matthias vii, viii, xi, 4, 32, 69, 86, 92, 102–5, 109, 122, 177, 225, 232, 289, 308, 328, 331 Stine, Gail 106, 122 Swain, Marshall 67 Swinburne, Richard v, viii, xi, 4, 8, 10, 14, 82, 77, 88, 91, 95, 102, 193, 195, 206, 294, 328 Thompson, Dennis 83, 86 Tienson, John 170, 178 Tomberlin, James 69, 178, 225, 279 Toulmin, Stephen 71, 72, 84, 87 Turri, John 73, 74, 87, 314 Urbach, Peter 198 Van Cleve, James 213, 221, 225 Van Inwagen, Peter 38, 56, 68, 81, 87 Vogel, Jonathon 145, 146 Weatherson, Brian 25, 33 Wedgwood, Ralph 26, 33 Williams, Michael 74, 87, 331

INDEX

Williamson, Timothy vii, ix, xii, 45, 48, 135, 147, 148, 150, 152, 153, 155, 158–61, 163, 164, 193, 204, 206, 208, 209, 211, 221, 222, 225–32, 240, 241, 252, 253, 255, 265, 280, 299, 312, 316, 321, 322, 329, 331

Wood, Jay 87 Wright, Crispin 235, 253 Zagzebski, Linda 3, 64, 87, 99, 101, 102, 178

335