Seemings and Justification: New Essays on Dogmatism and Phenomenal Conservatism 0199899495, 9780199899494

Table of contents :
Dedication
Contents
Preface
Contributors
1. Seemings and Justification: An Introduction • Chris Tucker
Part I: Seemings and Seeming Reports
2. Seemings and Semantics • Andrew Cullison
3. Seeming Evidence • Earl Conee
Part II: Foundations of Dogmatism
4. Immediate Justification, Perception, and Intuition • Jessica Brown
5. Problems for Credulism • James Pryor
Part III: Seemings and Epistemic Internalism
6. Does Phenomenal Conservatism Solve Internalism’s Dilemma? • Matthias Steup
7. Phenomenal Conservatism and the Dilemma for Internalism • Michael Bergmann
Part IV: The Significance of Seemings within Specific Domains
8. Doxastic Innocence: Phenomenal Conservatism and Grounds of Justification • Robert Audi
9. Agent Centeredness, Agent Neutrality, Disagreement, and Truth Conduciveness • Michael DePaul
Part V: Dealing with Cognitive Penetration
10. Phenomenal Conservatism and Cognitive Penetration: The “Bad Basis” Counterexamples • Matthew McGrath
11. Searching for True Dogmatism • Peter J. Markie
12. Phenomenal Seemings and Sensible Dogmatism • Berit Brogaard
Part VI: Phenomenal Conservatism
13. Phenomenal Conservatism and the Principle of Credulity • William G. Lycan
14. Michael Huemer and the Principle of Phenomenal Conservatism • Michael Tooley
15. Phenomenal Conservatism Über Alles • Michael Huemer
Name Index
Subject Index

Citation preview

Seemings and Justification

This page intentionally left blank

Seemings and Justification new essays on dogmatism and phenomenal conservatism Edited by Chris Tucker

1

3 Oxford University Press is a department of the University of Oxford. It furthers the University’s objective of excellence in research, scholarship, and education by publishing worldwide. Oxford 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 trademark of Oxford University Press in the UK and certain other countries. Published in the United States of America by Oxford University Press 198 Madison Avenue, New York, NY 10016

© Oxford University Press 2013 All rights reserved. No part of this publication may be reproduced, stored in a retrieval system, or transmitted, in any form or by any means, without the prior permission in writing of Oxford University Press, or as expressly permitted by law, by license, or under terms agreed with the appropriate reproduction rights organization. Inquiries concerning reproduction outside the scope of the above should be sent to the Rights Department, Oxford University Press, at the address above. You must not circulate this work in any other form and you must impose this same condition on any acquirer. Library of Congress Cataloging-in-Publication Data Seemings and justification : new essays on dogmatism and phenomenal conservatism / edited by Chris Tucker. pages cm Includes index. ISBN 978–0–19–989949–4 (hardback : alk. paper) — ISBN 978–0–19–989950–0 (updf) 1. Knowledge, Theory of. I. Tucker, Chris, 1981– BD161.S387 2013 121—dc23 2012051577

9 7 5 3 1 2 4 6 8 Printed in the United States of America on acid-free paper

For Joseph Ezra Tucker May we always be as close as we are now.

This page intentionally left blank

{ contents } Preface ix Contributors

xi

1. Seemings and Justification: An Introduction 1 chris tucker

part i Seemings and Seeming Reports 2. Seemings and Semantics 33 andrew cullison 3. Seeming Evidence earl conee

52

part ii Foundations of Dogmatism 4. Immediate Justification, Perception, and Intuition jessica brown 5. Problems for Credulism james pryor

71

89

part iii Seemings and Epistemic Internalism 6. Does Phenomenal Conservatism Solve Internalism’s Dilemma? 135 matthias steup 7. Phenomenal Conservatism and the Dilemma for Internalism 154 michael bergmann

part iv The Significance of Seemings within Specific Domains 8. Doxastic Innocence: Phenomenal Conservatism and Grounds of Justification 181 robert audi 9. Agent Centeredness, Agent Neutrality, Disagreement, and Truth Conduciveness 202 michael depaul

Contents

viii

part v Dealing with Cognitive Penetration 10. Phenomenal Conservatism and Cognitive Penetration: The “Bad Basis” Counterexamples 225 matthew mcgrath 11. Searching for True Dogmatism 248 peter j. markie 12. Phenomenal Seemings and Sensible Dogmatism 270 berit brogaard

part vi Phenomenal Conservatism 13. Phenomenal Conservatism and the Principle of Credulity 293 william g. lycan 14. Michael Huemer and the Principle of Phenomenal Conservatism 306 michael tooley 15. Phenomenal Conservatism Über Alles 328 michael huemer Name Index 351 Subject Index 355

{ preface } The connection between seemings and justification is related to indefinitely many issues and debates, and I wanted this volume to convey that broad relevance. On the other hand, I didn’t want the contributions to be a scattered sprawl. The result is a book broad enough to address perception, moral epistemology, formal epistemology, the epistemology of disagreement, and the philosophy of mind, while also having coherent areas of emphasis, such as the nature of seemings, the connection between seemings and internalism, attempts to restrict dogmatism to avoid cognitive penetration troubles, and the viability of phenomenal conservatism. There’s plenty more to say about each of these, but there’s also more to say about issues that go beyond the book. Here are but a few examples. The relevant empirical sciences related to perception discuss various representational states, but which, if any, answer to the term seeming? Proponents of dogmatism and phenomenal conservatism have been quick to assume that seemings have propositional content, but would these views be better off by rejecting this assumption? Disjunctivism and Wright-style entitlement are two salient rivals to dogmatism/phenomenal conservatism, and further work needs to be done exploring the relative merits of these views. There are many people who should be acknowledged in some way or another. Joe Tucker, to whom the book is dedicated, deserves recognition for being such a joyful diversion. He’ll eventually realize that I’m not that interesting to hang out with, but for now he loves playing with his daddy. Despite being dragged all over the globe by a professional nerd, my wife, Jenny Tucker, has loved and cared for me in ways that I’ll never deserve. Buy the book for all your friends, so that I can use the royalties to buy her something nice. Come to think of it, buy the book for your enemies and complete strangers too. Michael Bergmann was an ideal dissertation supervisor and has graciously provided support and guidance in the years post defense. He provided instructive council in the early stages of the project, as did John Bengson and Dan Korman. Robert Audi and Michael Huemer provided support and encouragement at various stages of the project. Marinus Ferreira deserves thanks for his behind-the-scenes work as my research assistant, and so does Logan Gage for his work on the index. As a whole, the contributors were a pleasure to work with, and so were Peter Ohlin, Lucy Randall, and Emily Sacharin at Oxford University Press.

x

Preface

I’m grateful to the University of Auckland Faculty Development Fund for its financial support of this project. The Royal Society of New Zealand graciously awarded my work on dogmatism and phenomenal conservatism with a Marsden Fast Start Grant, which enabled speedy completion of the introduction’s first four sections.

{ contributors } Robert Audi is John A. O’Brien Professor of Philosophy at the University of Notre Dame. Michael Bergmann is Professor of Philosophy at Purdue University. Berit Brogaard is Professor of Philosophy at the University of Missouri-St. Louis. Jessica Brown is Arché Professor of Philosophy at the University of St. Andrews. Earl Conee is Professor of Philosophy at the University of Rochester. Andrew Cullison is Associate Professor of Philosophy at SUNY-Fredonia. Michael DePaul is Professor of Philosophy at the University of Notre Dame. Michael Huemer is Professor of Philosophy at the University of Colorado-Boulder. William G. Lycan is William Rand Kenan, Jr. Professor of Philosophy at UNC-Chapel Hill. Peter J. Markie is Curators’ Teaching Professor at the University of Missouri-Columbia. Matthew McGrath is Professor of Philosophy at the University of Missouri-Columbia. James Pryor is Professor of Philosophy at New York University. Matthias Steup is Professor of Philosophy at Purdue University. Michael Tooley is College Professor of Distinction at the University of Colorado-Boulder. Chris Tucker is Assistant Professor of Philosophy at the College of William and Mary.

This page intentionally left blank

Seemings and Justification

This page intentionally left blank

{1}

Seemings and Justification: An Introduction Chris Tucker It is natural to think that many of our beliefs are rational because they are based on seemings, or on the way things seem. This is especially clear in the case of perception. You believe that there is a paper document (or a computer screen) in front of you because it seems visually that way. Many of our mathematical, moral, and memory beliefs also appear to be based on seemings. I believe that I ate cereal for breakfast because I  seem to remember eating it for breakfast. And we believe that torturing for fun is morally wrong and that 2 + 2 = 4 because those claims seem intuitively obvious. In each of these cases, it is natural to think that our beliefs are not only based on a seeming, but also that they are rationally based on these seemings—at least assuming there is no relevant counterevidence. These initial reflections, however natural, raise three questions: 1. Is there really some state—a seeming—that is present in each case or is it just convenient to talk that way? 2. If these seemings really are genuine states or entities, what are they? Beliefs, experiences, or something else entirely? 3. What is the connection between seemings and justified belief: under what conditions, if any, can a seeming justify its content? This volume focuses on the third question—the connection between seemings and justification—but the first two questions are obviously relevant to the third. If there is no such thing as a seeming, then there’s likely no point in asking about the connection between seemings and justification. And whether it is plausible to hold that seemings can make their corresponding beliefs justified depends heavily on what seemings are. The contributors to Seemings and Justification (or S&J) generally agree that seemings exist, except for Conee (S&J, sec 2.1) who is officially neutral on the topic. On the other hand, they disagree widely over what seemings are and the circumstances, if any, under which seemings can justify their contents.

2

Seemings and Justification

Phenomenal conservatism and dogmatism hold that there is a very tight connection between seemings and justification. Phenomenal conservatism (PC) holds that if it seems to S that P, then, in the absence of defeaters, S thereby has justification for believing P.1 As I will use the term, dogmatism is essentially phenomenal conservatism restricted to some domain(s). (I define “dogmatism” this way because it captures the usage that, I think, is most prevalent in the minds of epistemologists; however, Pryor, who is largely responsible for the popularity of the term, uses it differently.)2 Dogmatism about perception is the most popular version, and it holds that, if it perceptually seems to S that P, then, in the absence of defeaters, S thereby has justification to believe P.3 This volume explores the connection between seemings and justification by exploring the prospects of PC and various versions of dogmatism. There are few topics in epistemology that have more wide-ranging implications than the one considered in this volume. The first paragraph is enough to suggest that the connection between seemings and justification is a crucial issue regarding perceptual, memorial, and a priori justification.4 The chapters in this volume illustrate some of the broader importance of this connection by concerning (in no particular order): • the content of perceptual experiences (Brogaard, Cullison); • the epistemology of cognitive penetration (Brogaard, Markie, McGrath, this introduction); • the epistemology of philosophical method (Brown); • the epistemology of disagreement (DePaul); • doxastic conservatism5 (Lycan, McGrath); • testimonial justification (Audi); • moral epistemology (Audi, Brogaard) • the debate between internalism and externalism (Bergmann, Steup); and • the correct formulation of Bayesian epistemology (Pryor, this introduction).

1 Huemer (2001, 2006, 2007) is most commonly associated with the view, but Cullison (2010, 272– 74), Lycan (S&J), Skene (2013), and I (2010a, 2011) also endorse it. 2 For Pryor, dogmatism about X holds that it’s possible to have immediate but underminable justification for X. On his usage of the term there is no presumption that the justification consists in the seeming, rather than say, the reliability of the associated process. That he does not intend the stronger, internalist reading is clearer in his later work (S&J, sec 3; ms) than in his 2000. 3 In addition to Pryor (2000), those who endorse perceptual dogmatism include Audi (1993, 366), Chisholm (1989, 65), Chudnoff (2011), Pollock and Cruz (1999, 201), and Pollock and Ovid (2005). 4 Most entries in the volume are relevant to perceptual justification. For discussion of memorial justification, see the entries of Audi and Brogaard. For discussion of a priori justification, see the entries by Brogaard, Brown, Lycan, and Markie. 5 Doxastic conservatism, aka epistemic conservatism, is the claim that believing P provides prima facie justification for P.

An Introduction

3

I preview the other entries to the volume in section 5. In the meantime, I provide a general introduction to dogmatism and phenomenal conservatism. In section 1, I consider the ontology of seemings and their relation to other mental states. In section 2, I survey the main motivations for dogmatism and phenomenal conservatism. In sections 3 and 4, I consider two main objections to dogmatism and phenomenal conservatism. The first concerns whether cognitive penetration poses a problem for those views. The second concerns the relationship between those views and classical Bayesian epistemology.

1. Seemings: What Are They? As we’ve already mentioned, it is natural to think that perceptual experiences, memorial experiences, and a priori intuitions are kinds of seemings or at least closely related to seemings. But what, exactly, are seemings?

1.1. accounts of seemings Here are three views. A seeming that P is: Belief View Inclination View Experience View

A belief that P. An inclination, disposition, or attraction to believe that P. An experience with the content P or a sui generis propositional attitude that P.

As these views are to be understood, one might accept the Belief View for some seemings and the Experience View for others. Or one might hold that there really are seemings but allow more than one kind of state to count as a seeming: perhaps both a belief in and an inclination to believe Goldbach’s conjecture count as a seeming that the conjecture is true. The first of these three views is probably the least popular these days. Lycan endorses it in early work, provided that the beliefs in question are “spontaneous,” or non-inferential (1988, 165–66).6 Swinburne (2001, 141–42) also endorses it, but he does so in an ecumenical spirit. He also allows that a seeming that P can be an inclination to believe that P (so he also endorses the Inclination View), and a belief that P has “quite a probability (but less than ½)” (142).7 The Belief View receives support from the idea that the verb “seems” is often

6

As I mention below, he now seems to hold the Experience View. If Williamson (2007, 3) holds that intuitions are intellectual seemings, then he likewise holds that a seeming that P can be a belief or an inclination to believe that P. He could be construed as saying either that intellectual seemings are beliefs/inclinations or that there is no such thing as intellectual seemings (217). I don’t know what his official view is. 7

4

Seemings and Justification

used to report beliefs, especially beliefs that are not firmly held. When I say “It seems to me that the best economic policy is so and so,” I might simply mean that I believe, perhaps tentatively, that the best economic policy is so and so. Nonetheless, the Belief View is not widely endorsed these days, because it faces the Problem of Known Illusions. According to Huemer, we can tell that seemings, or appearances, “are different from beliefs from the fact that it may appear to one that p while one does not believe that p” (2007, 31). It might seem to me that the half-submerged stick is bent, even though I don’t believe that it is. Indeed, it might seem to me that the stick is bent even though I know it isn’t.8 Once it is appreciated that it can seem to one that P even though one doesn’t believe P, a natural move is to suggest that seemings are not beliefs, but inclinations to believe. When I disbelieve the stick is bent, it is at least plausible that many of us retain the inclination to believe that the stick is bent even when we resist that inclination. Consequently, the Inclination View has held a wider following than the Belief View and some version of it is endorsed by Ernest Sosa (1998, 258–59; 2007, ch. 3), Rogers and Matheson (2011), and as previously mentioned, Richard Swinburne (2001, 141–42). Another advantage of the Inclination View is that it makes seemings non-mysterious by reducing them to something that we apparently understand, namely, inclinations to believe. Despite the advantages of the Inclination View, Huemer provides three arguments that seemings should not “be identified with dispositions or inclinations to form beliefs” (2007, 31, emphasis removed; cf. Cullison 2010). Argument 1: I can be so convinced that an appearance is illusory that I’m not even inclined to believe it. It’s at least possible that I’m so used to the bent-stick illusion that I’m not even inclined to believe that the stick is bent. Argument 2: seemings provide non-trivial explanations of what I’m inclined to believe, and they couldn’t provide such explanations if they were identical to inclinations to believe. I’m inclined to believe that a computer is in front of me because it seems that way. Argument 3: an individual can be inclined to believe things even though they don’t seem true. I  might be inclined to believe P because I really want it to be true, even though it doesn’t seem true and perhaps seems false.9 The Experience View has at least two different motivations. The first is that it avoids the problems faced by the Belief and Inclination Views. In other words, it can accommodate the following data: (i) it can seem to S that P even if S doesn’t believe that P; (ii) it can seem to S that P even if S isn’t inclined to believe that P; (iii) a seeming to S that P can explain why S is inclined to believe

8

Lyons (2009, 71–72) rejects this argument by relying on a popular, functionalist account of belief. As Pryor has pointed out to me, many of Huemer’s objections to the Belief and Inclination views are reminiscent of points made by Jackson (1977, sec 2) in his discussion of “looks.” 9

An Introduction

5

P; and (iv) S can be inclined to believe that P even if it doesn’t (and never did) seem to S that P. The second motivation applies only to those who think at least some seemings can provide non-inferential justification for their contents. Suppose you hold that seemings can justify. You may wonder what seemings must be like in order to play that justifying role. If you reject doxastic conservatism (the claim that merely believing P can provide prima facie justification for its content10), you aren’t going to think that beliefs can justify their own contents. So you will reject the Belief View (at least, you will reject the Belief View for those seemings that you think justify their contents). And if beliefs don’t justify their contents, it’s hard to see why inclinations to believe justify their contents. So you will likely reject the Inclination View of seemings. Since many philosophers allow experiences to provide non-inferential justification for their contents, it is natural to think of seemings as experiences. This second motivation may partly explain why the Experience View has been so popular among those sympathetic to dogmatism or phenomenal conservatism. Its proponents include George Bealer (2000), Eli Chudnoff (2011), Andrew Cullison (2010), Michael Huemer (2001, 2005, 2007), William Lycan (S&J), Jim Pryor (2000), Matthew Skene (2013), and yours truly (2010a, 2011). Proponents of the Experience View generally agree about three things. First, seemings have propositional content.11 Second, they are distinct from any sort of belief or inclination. Third, seemings have their own distinctive phenomenal character (e.g., Pryor 2000, 547, n37; Huemer 2001, 77; and Tucker 2010a, 530). Whether proponents of this view say that seemings are sui generis propositional attitudes or experiences seems to depend more on terminological preference than substantive disagreement. When seemings are characterized as experiences or sui generis propositional attitudes, many find seemings obscure or non-existent. A critic might think, “I understand what a belief is. I understand what an inclination to believe is. When I introspect, I can find beliefs and inclinations to believe. But I don’t have a grip of some sui generis propositional attitude thing you call a ‘seeming,’ and I can’t find it when I introspect” (cf. Williamson 2007, 217). Characterizing seemings as experiences may help insofar as people generally have a grip of what experiences are. But the progress is limited. On this view, seemings aren’t just any experience: they are a special kind of experience. Okay, but what kind?

10

Prima facie justification = justification in the absence of defeaters (e.g., relevant counterevidence). There are, of course, many accounts of perceptual experience which hold that perceptual experiences lack propositional content. In the terminology I discuss in 1.2, I think those accounts are more naturally construed as accounts about sensations rather than seemings. In any event, those who most clearly endorse dogmatism or phenomenal conservatism tend to hold that seemings have propositional content, and so for space reasons, I’m ignoring accounts of dogmatism and phenomenal conservatism that might wish to get by without the claim that seemings have propositional content. 11

6

Seemings and Justification

Well, the kind that has this really neat and distinctive phenomenal character. Okay, but what phenomenal character? It’s at this point that proponents of the Experience View resort to ostension/examples (e.g., when you are looking at a tree, it seems to you that a tree is in front of you) and metaphor (a seeming that P “recommends” P as true or “assures” the subject of P’s truth).12 If that’s the best the proponent of the Experience View can do, one may well doubt that there is any such distinctive phenomenal character and may wonder whether there is any such thing as a seeming at all (cf. Tooley S&J, especially secs 2–4). Seeming realists (those who think seemings exist) who are dissatisfied with the apparent obscurity of the Experience View have reason to look for an alternative account of seemings. At this point, it is worthwhile to reflect on the general argumentative strategy that Huemer launched against the Belief and Inclination Views. Huemer argued that seemings aren’t beliefs by arguing that a seeming that P is not a belief that P. Likewise, Huemer argued that seemings aren’t inclinations by arguing that a seeming that P is not an inclination to believe P. This raises the question of whether the Belief or Inclination Views can be resuscitated by reducing a seeming that P to a belief or inclination to believe a proposition other than P. A seeming that P might be: Evidence-Taking View A belief or an inclination to believe of some mental state M that it counts in favor of P.

According to this view, if it seems to me that a police officer is in front of me, I’m inclined to take some mental state—maybe the sensation of the blue uniform—as evidence that there is a police officer in front of me. Conee (S&J) and Tooley (S&J) can be construed as proposing some version of the EvidenceTaking View, but such a construal oversimplifies Conee’s views somewhat. For he wants to leave it open that there is no semantic or ontological unity behind our seeming talk. The Evidence-Taking View has two potential advantages. First, it avoids the potential obscurity that afflicts the Experience View. Second, it arguably avoids all the problems that afflict the Belief and Inclination Views. For example, one problem with the Inclination View was that I might be so used to the illusion that I’m no longer even inclined to believe that the half-submerged stick is bent. The proponents of the Evidence-Taking View can accommodate this sort of case if you retain the inclination to treat the visual image of the stick as evidence that the stick is bent. (See Huemer S&J, sec 1.6 for criticisms of the Evidence-Taking View.) Despite its advantages, proponents of dogmatism and phenomenal conservatism have reason to resist this account. It’s hard to see how taking something to be evidence for P, by itself, can justify P. Perhaps justifiably taking oneself

12

See my 2010a, 530.

An Introduction

7

to possess evidence for P can justify P, but dogmatism and phenomenal conservatism do not require the seemings to be justified and typically deny that seemings admit of justification. It’s no surprise, then, that the Evidence-Taking View was put forward by critics of dogmatism and phenomenal conservatism.

1.2. seemings and the ontology of the mind Recently, philosophers have been re-examining the relationship between seemings and other mental states. The most popular subject of re-examination is the relationship between seemings and sensations, or sensory experiences. Huemer (2001, 58–79) and Tolhurst (1998, 300)  hold that sensations are a kind of seeming. On their view, we have seemings of many kinds—perceptual, memorial, and intellectual, perhaps among others—and the perceptual seemings are identical to sensations. Its seeming to you that the book is before you is identical to the visual image you have of the book’s being before you. Brogaard (S&J), Bergmann (S&J, sec 1.3), Conee (S&J), Sosa (2007, 48), and I (2010a, sec 1; 2011, sec 2.2) all reject this view and hold that sensations are a distinct state from perceptual seemings.13 Since Conee and Sosa deny that seemings are experiences, perhaps it is no surprise that they think there is a distinction between seemings and sensations. It is more surprising to see a proponent of the Experience View of seemings, like me, defend this assumption. To grasp the distinction between seemings and sensations, consider the following example of facial recognition. Suppose you and I are looking at the face of some person who, unbeknownst to you, happens to be my wife. We would have a mental “picture” of her in our minds. This “picture” might look and feel exactly the same to us, i.e., our mental images of my wife would be phenomenally identical. These phenomenally identical images are (visual) sensations. Although there is no phenomenal difference with respect to our sensations, there is a phenomenal difference in the way things seem. It would seem utterly obvious to me that she is my wife. On the other hand, it would not seem to you that she is my wife, and if anything, it would seem utterly obvious to you that you have no idea who you are looking at. Despite having phenomenally identical sensations, we have different seemings. A plausible explanation is that seemings are not identical to sensations. A proponent of the seeming/sensation distinction will likely hold that typical cases of perception involve both a seeming and a sensation. Visual

13 Plantinga (1993, 91–92) may be taken as giving, at the very least, a precursor to the seeming/ sensation distinction. Lyons’ (2009, ch 3) distinction between sensations and percepts also bears some connection with the seeming/sensation distinction. Chudnoff (2011) does think there are some important distinctions to be made between those experiences that have what he calls “presentational phenomenology,” such as seemings, and those that don’t. But his distinction seems very different from the one being considered here.

8

Seemings and Justification

perception will involve a seeming and a “visual image.” Auditory perception will involve a seeming and a “mental sound,” and so on for at least gustation and olfaction. A proponent of this distinction is also likely to find a parallel distinction in other domains. I seem to remember eating cereal for breakfast. This memorial seeming accompanies but is distinct from something like a degraded visual image of my cereal bowl (cf. Plantinga 1993, 58). My seeming that some act of torture is wrong may be accompanied by my revulsion of the act in question. In each of these cases, the dogmatist may say that while only the seeming has justificatory power, sensations, memorial imagery, and moral emotions play a crucial role in belief formation, perhaps by causing it to seem a certain way.14 The distinction between seemings and sensations has important implications for both epistemology and the philosophy of mind. In epistemology, it is alleged that this distinction plays an important role in resolving the speckled hen problem (Tucker 2010a, sec 3; Brogaard S&J, sec 5). In philosophy of mind, the distinction is claimed to undercut important arguments for the claim that sensations represent not only shape and color, but also more complex properties, such as being a pine tree, being a policeman, and being an instance of causation (cf. Brogaard S&J).15

2. Motivations for Dogmatism and Phenomenal Conservatism If you listen to the proponents of dogmatism and phenomenal conservatism, the advantages of those views are legion. First, dogmatism is thought to be the natural view for perception, intuition, memory, or some combination thereof.16 Second, dogmatism avoids both skepticism and regresses of justification. Third, dogmatism has been claimed to provide straightforward resolutions to certain epistemological issues, such as the speckled hen problem, expert recognition, and certain obstacles for coherentist explanationism.17 A fourth potential advantage is that a view like phenomenal conservatism provides a single principle that unifies all non-inferential justification, including perceptual, introspective, memorial, and a priori, and perhaps also

14

cf. Brogaard (S&J, sec 3). For other examples of this distinction’s importance within epistemology, see Tucker (2010a; 2011, secs 2.2–3) and Brogaard (S&J, sec 5). For other examples in the philosophy of mind, see Cullison (S&J). 16 Pryor (2000, 538) holds that it is the natural view of perception. Chudnoff (2011, 315) agrees and adds that it is also the natural view with regard to intuition (322). Brown (S&J) argues that it is natural to hold that intuition provides immediate justification, but she does not commit herself to dogmatism or the claim that intuitions are seemings. Pollock and Cruz (1999, 48) hold that it is the natural view of memory. 17 For resolutions to the first two issues, see Brogaard (S&J, sec 5) and my 2010a (secs 3, 5). For a resolution to the third issue, see Lycan (S&J). 15

An Introduction

9

inferential justification.18 Sober reflection, however, does raise some worries about this alleged advantage. It is natural to think that, at any given time, the vast majority of our beliefs are stored in long-term memory and are not presently in our consciousness. Yet, if seemings are the sorts of things that are necessarily occurrent, presumably a belief ’s content will seem true only if it is in consciousness. How, then, can seemings explain the justification of beliefs that are not presently in consciousness?19 I think it is largely acknowledged that dogmatism and phenomenal conservatism have at least some of the aforementioned virtues, especially the first two; but detractors tend to think that there are disadvantages which more than outweigh these advantages.20 There are two additional arguments for phenomenal conservatism which are more controversial. The first is Michael Huemer’s (2007) self-defeat argument, which tries to show that you cannot justifiably believe that phenomenal conservatism is false. Skene (2013, sec 3) defends the argument, but otherwise the argument has been widely criticized for a variety of reasons.21 Even some proponents of phenomenal conservatism wish to distance themselves from the self-defeat argument.22 Regardless of whether the self-defeat argument for phenomenal conservatism fails, a more limited selfdefeat argument for dogmatism about intuitions may nonetheless be viable.23 Another motivation for phenomenal conservatism is its ability to capture certain internalist intuitions. It’s no secret that internalists tend to think that there is some very tight connection between justification and one’s first-person perspective. The driving thought behind many internalisms is something like this: given my current situation, what should I believe now? And if something like that approach appeals to you, phenomenal conservatism may be appealing also. McGrath puts the point this way: Suppose it seems to you that P and you have no defeaters (i.e., no good evidence for not-P and no good evidence that this seeming is unreliable as to whether P). Which doxastic attitude would it be reasonable for you to have toward P? Disbelieve P, without good evidence for not-P? [That doesn’t seem reasonable.] Withhold judgment on P? It does seem to you that P, and you lack evidence for not-P and for the unreliability of the seeming with

18 Huemer (S&J) holds that even inferential justification can be reduced to the way things seem. I disagree and hold only that seemings have a role to play in inferential justification (see my 2012). 19 Audi (S&J, sec 2) also has doubts that seemings can account for all testimonial justification. 20 For example, although White (2006, 527–28) rejects dogmatism, he apparently grants it has at least the first two advantages. Markie (S&J) apparently grants that perceptual dogmatism has the first three advantages (see especially sec 2). 21 These criticisms include Conee S&J; DePaul 2009; DePoe 2011; Hasan 2011; Markie S&J; and Tooley S&J. For some replies, see Huemer’s 2009, 2011a, and his chapter in S&J. 22 I hereby distance myself from the argument, and Lycan (S&J, nt 8) does too. 23 Bealer (1992) defends the more limited self-defeat argument. Audi (S&J, sec 5.1) and especially Lycan (S&J, sec 4) note sympathy with the more restricted self-defeat argument.

10

Seemings and Justification

respect to P. [So it doesn’t seem reasonable to withhold judgment either.] The only reasonable attitude to take is belief.24 (McGrath S&J, sec 1; cf. Huemer 2001, 104–5)

The plausibility of McGrath’s reasoning suggests that epistemologists need to account for McGrath’s Datum

when it seems to you that P in the absence of defeaters, you bear a relation to P which makes it irrational to disbelieve or withhold judgment about P.

Huemer (2006) gives the most detailed and sustained defense of this motivation. His basic idea is this. If you fix the way things seem, you fix what is epistemically relevant from the subject’s perspective. Suppose a theory allows something besides a seeming to contribute to what one ought to believe. Huemer holds that such a theory would be committed to this odd result: there could be some propositions P and Q such that a subject ought to affirm P and deny or withhold Q while at the same time acknowledging that, insofar as he can tell, P and Q are exactly alike in all epistemically relevant respects (cf. 2006, 151). The plausibility of Huemer’s reasoning suggests that epistemologists need to account for Huemer’s Datum when P and Q seem alike in all epistemically relevant respects, it is irrational to treat the two propositions differently (e.g., believe one but disbelieve the other).

The phenomenal conservative can provide straightforward explanations of these two data. Since an undefeated seeming that P would justify P, it would be irrational to disbelieve or withhold judgment about P. Thus, McGrath’s Datum is explained. Since there is no relevant difference in the way that P and Q seem, the two propositions shouldn’t be treated differently. Thus, Huemer’s Datum is also explained. Do these data give us reason, by themselves, to endorse phenomenal conservatism? Not obviously, for there is another explanation that Huemer and McGrath fail to rule out.25 To understand this alternative, we need to introduce rational commitment. S is rationally committed to taking some attitude A  toward P just in case, if one takes an attitude toward P, it is irrational not to take A toward P (cf. Pryor 2004, 363–64, 2012, sec V). Suppose I believe P. I am rationally committed to believing P or Q. Perhaps it is perfectly sensible to ignore P or Q, to take no attitude toward that disjunction at all; however, if I do take an attitude toward

24 Although McGrath clearly has some sympathy with this line of reasoning, it is unclear to what extent he ultimately endorses it. 25 I owe the following objection to Jackson 2011, but the presentation of the objection is my own. McGrath (S&J) provides an alternative way to press the objection.

An Introduction

11

that disjunction, as long as I  continue to believe P, I  would be irrational if I disbelieved or withheld judgment about P or Q. The problem is that such a combination of attitudes would be incoherent. In the above example, I am rationally committed to believing P or Q. Yet it does not follow that I am justified in believing P or Q. If I believe P or Q solely on the basis of my belief in P, my belief in P or Q is justified only if my belief in P is justified. If my beliefs do not violate my rational commitments, they will avoid a certain kind of incoherence. Yet justification requires more than simply avoiding this incoherence. The alternative explanation of the two data is provided by (phenomenal) semi-conservatism: necessarily, if it seems to S that P, then, in the absence of defeaters, S is thereby rationally committed to believing P.26 If semi-conservatism is true but phenomenal conservatism is false, then a seeming might make it irrational to not believe P (assuming I take an attitude toward P at all) without making me rational, or justified, in believing P. In such a case, I’m at an epistemic dead end, a situation in which there is no rational, or justified, attitude that one can take toward P. Consider an analogy. Suppose I take it as my ultimate end to eat every rock I find. I find a rock. Eating that rock is a necessary means to my ultimate end, so I’m rationally committed to eating the rock; it would be irrational for me to ignore the rock and carry on with my day. Yet my end of eating rocks is downright stupid and stupid ends can’t make it rational to take the necessary means to those ends. So I’m at a practical dead end: (as long as my ultimate aim remains the same) I’m irrational whether I eat the rock or not. Back to epistemology. Suppose that a seeming can’t even prima facie justify a proposition when it has a bad causal history, even if the subject is now completely unaware of the bad causal history (and so even if the causal history doesn’t provide a defeater). Perhaps the seeming is a memorial seeming that P which is the product of forgotten irrationality. Given semi-conservatism, this seeming that P rationally commits one to believing P, and so makes it irrational to disbelieve and withhold judgment about P. Yet, on the assumption that bad causal histories nullify (prima facie) justificatory power, one would also be irrational for believing P. So one is at an epistemic dead end: (as long as the causal history of one’s seeming remains the same) one is irrational no matter what attitude one takes toward P. The semi-conservative explanation of the two data should now be relatively clear. Since a seeming that P rationally commits one to P, it would be irrational to disbelieve or withhold judgment about P. McGrath’s Datum is thereby

26 They don’t address this question, but Markie, McGrath, and Brogaard may each endorse phenomenal semi-conservatism (or semi-dogmatism, if you prefer) despite their rejection of phenomenal conservatism.

12

Seemings and Justification

explained. If epistemic dead ends are possible, it may nonetheless be irrational to believe P and phenomenal conservatism would be false. By adding a few details, we get a plausible explanation of Huemer’s Datum. The subject may compare the memorial seeming that P, which has a bad causal history, with a memorial seeming that Q, which has a good causal history. It might seem to the subject that P and Q are alike in all epistemically relevant respects (again, the subject isn’t aware of the bad causal history of her seeming that P). In such a case, semi-conservatism can say it would be irrational for the subject to treat P and Q differently (e.g., believe one and withhold judgment about the other). Huemer’s Datum is thereby explained. If epistemic dead ends are possible, S nonetheless might be justified in believing only Q and so PC would be false. Until the phenomenal conservative shows that epistemic dead ends are impossible or finds some other reason to prefer the phenomenal conservative explanation over the semi-conservative one, internalist intuitions don’t, by themselves, support phenomenal conservatism.27

3. Cognitive Penetration Objections Dogmatism and phenomenal conservatism allow seemings to justify their contents no matter how they are caused. A natural way of attacking these views is to consider seemings whose causal histories appear incompatible with the seemings’ providing even prima facie justification. The most common ways of pushing this objection appeal to cognitively penetrated seemings. For our purposes, we can say that S’s seeming that P is cognitively penetrated by S’s mental state M just in case M (partly) causes that seeming that P. If my desires, beliefs, experiences, or other seemings (partly) cause me to have a seeming that P, then that seeming is cognitively penetrated. It’s worth stressing that cognitive penetration, by itself, is not a bad thing. Consider the lizard-like tuatara (figure 1.1). This creature is so awesome that it hung out with dinosaurs and has a New Zealand brewing company named after it (try their pilsner). As should be obvious to anyone, if alcoholic beverages aren’t named after you, you’re not that important. But I digress. More to the point: I look at the tuatara and it seems to be some kind of lizard; but it isn’t. The background beliefs of an expert might make it seem to the expert that 27 McGrath (S&J, sec 2) suggests that the best explanation of a seeming’s ability to rationally commit may be its ability to justify. If so, it would not be viable to endorse semi-conservatism while rejecting phenomenal conservatism. McGrath only mentions this suggestion and doesn’t defend it, and I don’t know to what extent he himself endorses this suggestion. He clearly rejects phenomenal conservatism, but what he says in the chapter seems compatible with semi-conservatism. In any event, the suggestion isn’t very promising if unjustified beliefs can rationally commit one to their obvious entailments. When we say things like “Your belief in P commits you to Q, which is just flat out crazy” we don’t assume that the original belief in P has to be justified in order to rationally commit the person to Q.

An Introduction

13

figure 1.1

the creature is a tuatara, not a lizard. The expert’s cognitively penetrated seeming doesn’t seem problematic, and, if anything, seems to enhance the expert’s cognition. When a seeming is cognitively penetrated by relevant and justified beliefs, there often doesn’t seem to be any problem with basing your belief on the relevant seeming.28

3.1. cognitive penetration by justifiers Although not all cognitive penetration is bad, some might be. One way of pushing cognitive penetration worries can be called the Illegitimate Boost Objection. The basic idea is that when your seeming that P is causally dependent on, so cognitively penetrated by, a legitimate justification for P, it is implausible to suppose that the seeming does any extra justificatory work.29 Suppose you have some bit of evidence that doesn’t depend on its seeming to you that P. For example, suppose you accept P on the basis of a good argument. While you are thinking about how clever you are for coming up with that argument, P begins to seem true to you. Now you have two justifications

28

See Lyons (2011, sec 2) for further examples of benign cognitive penetration. In such a case, the objector might say, you would double count the justificatory force of the original legitimate justification: you would count its force once directly and once indirectly via the seeming. Hence, in S&J, Tooley and Huemer refer to this objection as the Double Counting Objection. 29

14

Seemings and Justification

for P, the argument for P and its seeming that P. So far, so good. But now suppose that your acceptance of the argument is what makes P seem true to you and that you are oblivious to this fact. Objectors claim that according to phenomenal conservatism, the seeming that P would provide your justification a boost; however, this boost is claimed to be illegitimate because the seeming is completely dependent on the argument for P. I’ve heard this objection in conversation on multiple occasions. In print, I’ve seen the objection from Tooley (S&J, sec 5.2.1) and, oddly enough, Huemer (1999, 348).30 To see the problem with this objection, consider an analogy with testimony. Suppose that Bill testifies (P) that there is free pizza on the quad. You aren’t convinced, but his testimony does give you some reason to believe that P. Then Jill comes along and also testifies that P. You have no special reason to believe that her testimony is dependent on Bill’s. It seems reasonable to raise your confidence in P; her testimony gives your justification in P a boost. And this is so, even if, unbeknownst to you, Jill believes that P only because Bill told her. Once you learn that her testimony is dependent on Bill’s, her testimony no longer provides you with a boost. Likewise, when you don’t know that your seeming causally depends on another justification, it can provide a boost. When you learn that the seeming causally depends on another justification, it can’t provide a boost. We might quibble over the details of the example, but the key point is this. A second justification, whether dependent or not, whether in the form of seemings or testimony, can provide a legitimate boost to your justification when you reasonably believe the two justifications to be completely independent. They can’t add boosts when you reasonably believe them to be completely dependent (cf. Huemer S&J, sec 3.3). What’s less clear is what to say about the intermediate cases, such as when you have no evidence one way or another concerning the independence of justifications. These cases are challenging, but they are no more challenging for the dogmatist than they are for anyone else.

3.2. cognitive penetration by non-justifiers The Illegitimate Boost Objection appealed to seemings that are cognitively penetrated by mental states that constitute legitimate justification for the target proposition. The Tainted Source Objection,31 perhaps the most common objection to dogmatism, appeals to seemings that are penetrated by mental states that are not legitimate justifications for the target proposition, such as desires and unjustified beliefs.

30 31

Audi (S&J, sec 2.2) discusses a related worry. I get this name from Huemer (S&J).

An Introduction

15

Consider a case of cognitive penetration by desire. Suppose it seems to Wishful Willy that the yellow object is gold, because he wants it to be gold. Most epistemologists find it counterintuitive to allow his wishfully produced seeming to provide him with prima facie justification that the nugget is gold.32 Or suppose that Jill irrationally believes that Jack is angry. Jack walks into the room, and her irrational belief makes it perceptually seem to her that Jack is angry. Most epistemologists find it counterintuitive to allow Jill’s seeming to provide her with justification that Jack is angry.33 Those who press this sort of objection against dogmatism often attempt to diagnose precisely what’s wrong with the problematic sort of cognitive penetration. Siegel (2011) speculates that the problem is a kind of circularity. Lyons (2011) suggests the problem is a lack of reliability. McGrath (S&J, secs 4, 5) suggests that the problem boils down to free enrichment. Roughly, a seeming that Q is freely enriched when it is cognitively penetrated by a seeming that P and P does not support Q. I leave examination of these diagnoses for another occasion. Although I’m a proponent of phenomenal conservatism, I have the intuitions that the objectors want me to have in some of the cases at least some of the time. But not everyone shares the intuitions of the objectors. Lycan (S&J, sec 7) and Huemer (S&J, sec 5) apparently don’t have the intuition that, say, wishfully produced seemings can’t provide prima facie justification for their contents. If the subjects in such cases fail to have ultima facie justification, it is because they have a defeater (e.g., they have good reason to believe they are unreliable on the topic). Still, the intuitions are widely shared, and it would be a significant coup for dogmatists if they could account for these intuitions without giving up their dogmatism. In what follows, I will focus on the simple Wishful Willy case, because that’s the case that I  find most counterintuitive and the case that, to me, incorporates the fewest distracting details.34 A popular line of response is to concede that our intuitions are picking up on a legitimate defect but then deny that the defect is that the cognitively penetrated seeming can’t provide prima facie justification. In other words, this approach distinguishes between prima facie justification and some other status and then says that our negative reaction to the wishfully produced seeming is due to the failure to attain this other status. So what is this other status (or lack thereof) that our intuitions allegedly track?

32 See, e.g., Goldman (2009, 330), Lyons (2011), Markie (2005, 356–57; S&J), Siegel (2011), and Steup (S&J, sec 6). 33 This example is Siegel’s (2011). Forgotten irrationality objections to dogmatism about memorial justification have a similar structure. See, e.g., Goldman (2009, 323). 34 Lyons (2011) also seems to think that cognitive penetration by desire provides the most forceful variant of the cognitive penetration objections.

16

Seemings and Justification

Depending on the case in question, I’ve made two different suggestions (2010a, sec 6; cf. 2011, sec 5). I argue that a wishfully produced seeming can provide prima facie justification even though it can’t provide warrant for its content, where warrant is the property that makes true belief knowledge. This result would be parallel to a common view of new evil demon cases: the experiences of demon victims provide prima facie justification for their contents, but demon-caused seemings can’t provide warrant for their contents. I  also distinguish between justification and epistemic blameworthiness and argue that a subject would be epistemically blameworthy for having the wishfully produced seeming, even though the seeming is still capable of providing prima facie justification for its content (cf. Huemer S&J, sec 5). Markie (S&J, sec 3) and McGrath (S&J, sec 4) argue that these two suggestions cannot cover all the problematic cases. At the very least, these criticisms show that more needs to be said about blameworthiness for it to be clear that it is present in all the troublesome cognitive penetration cases. Another suggestion is put forward by Skene (2013, sec 5.1). He distinguishes between the evaluative properties of the agent and the evaluative properties of a belief based on the wishfully produced seeming. His suggestion is that our intuitions are explained by our rightly criticizing the agent for having wishfully produced seemings, but wishfully produced seemings nonetheless provide prima facie justification. In other words, our intuitions correctly identify a defect in the subject’s cognitive character, but this defect in character needn’t entail a defect in the resultant cognitive “action,” namely, believing the content of the wishfully produced seeming. This response has initial appeal, but it faces a problem parallel to that faced by my blameworthiness approach: without a more worked out theory concerning the appropriate conditions for criticizing agents and actions, it’s not clear that this move will cover all problematic cases.35

4. A Bayesian Objection to Dogmatism We will focus on one Bayesian objection to dogmatism, though there are others. This objection consists of two stages: a formal proof and some interpretation of that proof. Since Pryor (S&J, sec 6) discusses the proof at length, in 4.1, I will merely explicate the result of the proof and then present the standard

35 An alternative way of defending dogmatism against this objection is to show that if dogmatism has a problem with cognitive penetration, then so do most other views in epistemology (see my manuscript). While this response doesn’t show that dogmatism doesn’t have a problem, if correct, it would prevent others from using dogmatism’s alleged cognitive penetration problems as a reason to prefer their own view.

An Introduction

17

interpretation of that result. In the following sub-sections, I  consider three broad strategies for responding to this objection.

4.1. summary of the objection The acquisition of new evidence E raises the probability of some propositions and lowers the probability of others. In such a case, we say that there has been an update on E. Let “Old” represent the probability of a proposition before the update on E and “New” be the probability of a proposition after. The relevant Bayesian objection focuses on a case in which E1, the proposition we update on, is “it seems to me that I have a hand.” H1 is assumed to be “I have a hand” and BIV the conjunction that “I’m a handless brain-in-a-vat and I seem to have a hand.” For the sake of the objection, E1 is assumed to be evidence for H1. The proof establishes this: Formal Result New(H1) < Old(~BIV).

In other words, the proof establishes that the probability that I have a hand will always be less than the antecedent probability of ~BIV. The antecedent probability of ~BIV sets an upper limit on how high E1 can raise the probability of H1. The standard interpretation assumes that a proposition’s having a certain degree of probability is its having a certain degree of justification. E1’s raising H1’s probability corresponds to E1’s justifying H1. Since Old is the probability of a proposition before the update on E1, it’s natural to interpret Old as one’s justification for a proposition antecedent to E1’s justifying H1. S’s acquiring E1 as evidence is understood as S’s coming to know E1. With this interpretation in place, the formal result becomes the Interpreted Result

S’s knowledge that E1 justifies H1 to some degree only if S has some higher degree of antecedent justification for ~BIV.

In other words, the antecedent justification of ~BIV places an upper limit on the justification that knowing E1 provides for H1. Note that the formal result constrains, most directly, the degree of justification provided by knowing that it seems to me that I have a hand; however, dogmatism holds only that the having of the seeming confers prima facie justification. For this objection to even be relevant to dogmatism, the Bayesian objector relies on the proxy assumption: by focusing our attention on reflective subjects who know how things seem to them, we can understand the epistemic effects of seemings for all subjects. For the sake of the discussion, we shall simply grant this assumption.36 Once we do, the dogmatist is in an odd 36 For discussion of this assumption, see White (2006: 534-5) and Pryor (S&J, sec 6). In note 41, I also point out that this assumption is problematic given agent-centered conceptions of justification.

18

Seemings and Justification

position. He tells us that seemings provide foundational, immediate justification, but Formal Result says that the immediate justification produced by seemings requires antecedent justification for rejecting skeptical hypotheses. But, while this is an odd position to be in, is it a problematic one? As we’ll see, that’s unclear.

4.2. strategy 1: distinguish necessary and constitutive conditions At this point, we need a sharper definition of dogmatism. I said that dogmatism about some domain is the claim that within that domain, if it seems to S that P, then, in the absence of defeaters, S thereby has justification for P. I didn’t explain what was intended by “thereby.” The basic idea is that in the absence of defeaters, the seeming justifies P all by itself. This means, in part, that one’s seeming-based justification for P does not consist in justification for the denial of skeptical hypotheses. If the dogmatist can sensibly maintain that this antecedent justification for ~BIV is a necessary condition on one’s seeming-based justification for H1 (which is what Interpreted Result claims) while resisting the further claim that one’s seeming-based justification partially consists in this antecedent justification, it’s hard to see how the Formal or Interpreted Result poses any problem for dogmatism. Silins (2008, especially sec 4.1) was perhaps the first to make this sort of point, but he doesn’t try to explain how dogmatism could be true if seeming-based justification requires antecedent, anti-skeptical justification. Without some explanation of this surprising result, Silin’s position may seem like flat-footed stubbornness. That’s where Wedgwood (forthcoming) comes in.37 The Bayesian objector typically assumes that one’s antecedent justification must partially explain why it’s rational for one to believe P when it seems that P. Wedgwood claims the typical objector gets things backward: it’s the fact that seemings make their corresponding beliefs rational that explains why we have antecedent justification for the denial of skeptical hypotheses (cf. Pryor S&J, secs 6-7). If dogmatism is true, then any thinker with the relevant concepts and abilities has at least one course of reasoning available to her that relies on dogmatism’s truth and concludes that we have antecedent justification for the denial of skeptical hypotheses. One example involves reunderstanding the nature and purpose of the much-maligned bootstrapping reasoning.

37 Wedgwood (forthcoming, sec 4b) rejects what he calls “dogmatism,” but he’s using the term differently. He’s a dogmatist, as I use the term.

An Introduction

19

4.3. strategy 2: distinguish risk and uncertainty Suppose you are considering P and have no evidence bearing on P at all. It’s not that you have evidence for P and evidence for ~P that balance each other out. It’s not that you have evidence that the objective probability of P is .5. It’s that your evidence is completely irrelevant as to whether P. In such a case, what credence, or degree of belief, are you justified in having? This is a notoriously difficult question. Whatever the answer is, let’s say that such a credence would be based on ignorance. To the extent that ignorance justifies a certain credence in P, the credence is said to be uncertain. To the extent that your evidence justifies a credence in P that is less than 1 (say, because of less than conclusive evidence or knowledge of the objective chances), the credence is said to be risky. Those sympathetic to dogmatism tend to think that there is an important difference between the epistemic effects of uncertainty and risk. For example, a dogmatist might argue that justification to have a low credence in the proposition I’m not being deceived has far more devastating consequences when it is risky than uncertain, when it is based on evidence than ignorance. Kung’s (2010) attempt to exploit the distinction between risk and uncertainty is especially modest, because, like the approach in the previous sub-section, it can assume the classical Bayesian formalism for the sake of argument. His goal is to explain how, from within the classical Bayesian framework, one might try to model the difference between risk and uncertainty in a way that a dogmatist will find appealing. Pryor (ms), Weatherson (2007), and Jehle and Weatherson (2012), also think the Bayesian objection ignores important differences between risk and uncertainty, but they apparently think that to model these differences, we must reject the classical Bayesian’s formalism.

4.4. replace the formalism The third approach is to replace the classical Bayesian apparatus with a formal structure more friendly to dogmatism. The target is typically the classical Bayesian’s update rule, which says roughly this: when you get evidence E for H, E becomes certain and New(H) should always equal Old(H|E). There is reason to be suspicious of this rule in this context. Weisberg (2009; and especially ms, secs. 1–2) has argued that this rule can’t model pure undermining defeat.38 Now, the Bayesian objector doesn’t rely on a pure undermining defeater. If you have evidence that you are a handless brain-in-a-vat that is being deceived into thinking you have a hand, you have evidence that rebuts and undermines:  it both provides evidence against your having a hand and

38 Weisberg (ms) extends this complaint to a variety of other rules, including Jeffrey conditionalization. See Pryor (S&J, secs 7-8) for further discussion.

20

Seemings and Justification

it provides evidence that your evidence that you have a hand is misleading or untrustworthy. Although Weisberg doesn’t show that the classical Bayesian can’t model a mixed defeater of this sort, his argument should make us wonder whether the fundamental problem is with, not dogmatism, but the classical approach to modeling undermining defeat. Putting the previous sub-section together with this one, we see that the dogmatist has reason to find a formal apparatus that will be able to model two phenomena: pure undermining defeat and the epistemic differences between risk and uncertainty. Pryor (ms) and Weatherson (2007) take steps in that direction. Given how widely embraced these two phenomena are independently of dogmatism, constructing such a formal model is well motivated, even if it would overturn the classical update rule. Jehle and Weatherson (2012) challenge a different component of Bayesian orthodoxy. Classical Bayesian holds, as a matter of logic, that Pr(P) + Pr(~P) = 1. Their intuitionistic Bayesianism, on the other hand, denies that this claim is true as a matter of logic. If their intuitionistic alternative is assumed, they argue that the Formal Result does not hold. The alternative is a bit exotic, but let me ask you: are you certain that it is false? I’m not, and Weatherson and Jehle argue that the slightest lack of certainty is enough to save dogmatism. At this point, it is not clear how the dogmatist should respond to objections that rely on the classical Bayesian formalism. But nor is it clear that the Bayesian objections reveal that dogmatists really have a problem. Indeed, if anybody has a problem here, it may be the classical Bayesians.

5. What’s in the Volume I now turn to introducing each chapter in the volume. This section is divided into five sub-sections, one for each part of this book.

5.1. seemings and seeming reports The two chapters in Part I aim primarily at clarifying seemings. Other chapters contain important discussions of seemings, but I’ve put them in other parts of the book because I take their primary aim to be something else. These other important discussions include Bergmann’s section 1, Brogaard’s section  3, Tooley’s sections 2–4, Huemer’s section 1, and, of course, the first section of this introduction. In chapter  2, Andrew Cullison focuses on the propositional content of seemings. He argues that at least some seemings have Russellian propositions as contents. Roughly, a proposition is Russellian if it contains an object or property as a constituent. Whereas a Fregean might hold that the meaning of

An Introduction

21

a name (e.g., “Superman”) is some associated description (e.g., the guy who flies around in blue spandex), Russellianism would hold that the object itself (Superman) is the meaning of the name (“Superman”). Cullison rejects a number of recent arguments in favor of Fregeanism about the content of perceptual seemings on the grounds that those arguments fail to distinguish between seemings and sensations. One upshot is that philosophers of mind have something to gain by paying attention to recent developments in the epistemological literature on seemings. In chapter 3, Earl Conee’s primary objective is to defend the Evidence-Taking account of seemings (assuming that seemings do exist) and its epistemological implications. Conee’s motivation for the account pays careful attention to when it is correct to say that “it seems to me that so-and-so.” Once we think of seemings as inclinations to believe of something that it counts in favor of P, Conee maintains that we can be disabused of two extreme views: seemings never provide justification and seemings always provide justification. Suppose we are inclined to take M as evidence for P. If it turns out that M really is evidence for P, then one component of the seeming, M, justifies P. For example, Conee holds that when M is an ordinary visual sensation, M is evidence for P. In such a case, seemings provide justification; but they don’t always do so. When M is evidentially irrelevant to P, as it would be if M is my desire for P, the seeming doesn’t provide justification for P. One thing that emerges from Conee’s discussion is that he endorses the distinction between seemings and sensations, and he holds that sensations, not seemings, are what do the primary justificatory work.

5.2. foundations of dogmatism The chapters in Part II defend elements of dogmatism that are widely endorsed even among non-dogmatists. In chapter 4, Jessica Brown argues for Immediacy about Intuition: we have immediate, or non-inferential, justification to believe the contents of at least some of our intuitions. The standard way of arguing for this claim assumes that both perceptual experiences and intuitions are a form of seeming (see, e.g., Chudnoff 2011). Brown’s arguments are less controversial because she does not assume that intuitions have any particular ontology. This neutrality allows her arguments to have much broader appeal than those directed at defending a very specific dogmatist thesis about intuition. Before discussing the motivation she finds successful, she considers whether regress arguments and the avoidance of skepticism can provide suitable motivations. She argues that they do not, and, contra popular opinion, she argues that, without supplementation, they fail to motivate even Immediacy about Perception, the claim that we at least sometimes have non-inferential justification for the contents of our perceptual experiences. When she turns to her positive argument, she contends that Immediacy about Intuition provides the

22

Seemings and Justification

best explanation of our intuitions in Gettier Cases. Brown’s examination of Gettier Cases is an important contribution not only to the epistemology of intuition, but also the epistemology of philosophical method. In chapter 5, Jim Pryor defends dogmatism against the Bayesian objections, and he provides a comprehensive examination of the assumptions involved in such arguments. As he thinks of them, they target not merely dogmatism but a far more modest and widely endorsed thesis he calls “credulism.” Say that a potential underminer for P is a proposition U such that justification for U would undermine a subject’s justification for P. Credulism holds that there is at least one U, such that one’s justification for P does not (even partially) consist in antecedent justification for U. Dogmatism is a very strong version of credulism. It holds that one’s seeming-based justification does not consist in antecedent justification for any proposition, yet skeptical hypotheses are potential underminers for one’s seeming-based justification that, say, one has hands. Credulism is compatible with the claim that justification (whether inferential or non-inferential) consists in antecedent justification for lots of propositions. It holds only that there is at least one U for P such that one’s justification for P does not consist in antecedent justification for U. Pryor argues that if we reject credulism, we must take one of two alternatives. The first continues to count all intuitive cases of undermining as genuine cases of undermining. The problem with this position, when combined with anti-credulism, is that it is committed to a suspicious regress of justification. The second denies that every intuitive case of undermining is a genuine case of undermining. In other words, to reject credulism—to hold that justification for P at least partially consists in antecedent justification for every potential underminer for P—one must hold either (i) that justification requires never ending gobs of antecedent justification or (ii) that one limit the number of potential underminers in a counterintuitive way. Given that neither option is particularly appealing, we have reason to think that the Bayesian objection to dogmatism goes wrong somewhere. Much of Pryor’s essay is then devoted to identifying precisely where it goes wrong.

5.3. seemings and internalism Part III concerns whether dogmatism or phenomenal conservatism, if true, provides a way of salvaging what we can call access internalism (AI). In its most generic form AI holds: S is justified in believing P only if S is aware of a justification contributor for believing P, where a justification contributor is something that contributes to the justification of believing P. This sort of internalism, as such, imposes no constraints on which kinds of things can count as justification contributors. As far as AI is concerned, seemings, beliefs, and reliable processes could all be among the things that count as justification

An Introduction

23

contributors. AI simply demands that the subject be aware of at least some of the things—whatever they are—that contribute to the justification of believing P. Proponents of AI include BonJour (1985) and Fumerton (1995). Perhaps the most powerful objection to AI is Michael Bergmann’s (2006, chs 1, 2)  so-called dilemma for internalism. In chapter  6, Matthias Steup rejects phenomenal conservatism but argues that seemings can nevertheless be used to resolve the dilemma. In chapter 7, Bergmann argues that seemings do not provide the resources to resolve his dilemma, and he specifically criticizes the proposals proffered by Steup (S&J) and Rogers and Matheson (2011). Their disagreement over the alleged resolution of the dilemma reflects an even deeper disagreement over what is required to avoid the Subject’s Perspective Objection (SPO). The canonical presentation of the SPO is BonJour’s (1985) Norman examples. Suppose a belief that P just pops into Norman’s head. This belief isn’t based on any other mental state, P doesn’t seem true to Norman, and Norman has no relevant evidence concerning P’s truth or whether the belief was formed in a reliable or otherwise appropriate way. In short, although Norman has no defeaters, there’s nothing that supports P from Norman’s perspective. There is wide agreement that Norman’s belief that P is not justified, and this intuition remains even if we add that the belief was caused by some reliable, clairvoyant ability that Norman doesn’t realize he has. Let us describe the problem with Norman’s belief by saying that it is an accident from Norman’s perspective that P is true. To avoid the SPO, a theory of justification must prevent a subject from being justified in believing P whenever it would be an accident from the subject’s perspective that P is true. In the above Norman case, there is wide agreement that Norman’s belief that P isn’t justified; however, there is considerably less agreement concerning what is required to prevent it from being an accident from Norman’s perspective that P is true. Comesaña (2010) holds that as long as Norman has evidence that P is true (perhaps P seems true), then, from Norman’s perspective, it is not an accident that P is true. As I understand them, Bergmann, Steup, and Huemer (S&J, sec 2) all have the intuition that merely having evidence that P is not sufficient to prevent P from being an accident from the subject’s perspective. I have the same intuition, albeit one that is weaker than the intuition I have regarding the Norman case from the previous paragraph. What else might be required to prevent P from being an accident from the subject’s perspective? Steup provides one way of trying to make good on the following basic suggestion: Necessarily, (i) if S has evidence E for P and (ii) evidence E1 that E reliably indicates P, then P is not an accident from S’s perspective. On the other hand, both Bergmann (S&J, sec 3.3) and Huemer (S&J, sec 2) deny that having E and E1 is, by itself, enough to prevent P from being

24

Seemings and Justification

an accident from the subject’s perspective.39 They ultimately conclude that the intuitions behind the SPO demand too much and should be rejected altogether.40

5.4. the significance of seemings within specific domains Part IV of the volume examines the significance that seemings have for specific domains. In chapter 8, Robert Audi’s main goal is to identify what, if any, psychological and normative roles seemings play. Particular attention is paid to the domains of perception, intuition, memory, and testimony. He argues that, with the possible exception of intuitive ones (secs 5.1–2), seemings fail to play any fundamental normative role (sec 3). On the other hand, seemings have such significant bearing on which beliefs we actually form, they may be ineliminable features of a “full-scale theory of rationality” (sec 3). Even here, though, Audi warns against overestimating the importance of seemings. Consider, for example, a rather ordinary case of testimony in which a person relates to us a rather long string of events. We very well may end up believing each part of the story. Yet it’s doubtful that each part of the story seems true. A more natural description is that we get “so wrapped up in [the] story . . . that the information provided simply flows into our belief system” (sec 2). If Audi is correct that many such testimonial beliefs are not based on seemings, then he may expose an important limitation of phenomenal conservatism: it cannot account for the justification of many of our testimonial beliefs. In chapter 9, Michael DePaul considers the significance of seemings for the epistemology of disagreement, and he essentially objects to Feldman’s (2007) dictum that evidence of evidence is evidence. Consider two subjects S1 and S2; it is not assumed that S1 and S2 are distinct. Also consider some arbitrary proposition P and some condition C. The following thesis can be considered a formal presentation of Feldman’s dictum: Agent Neutrality (AN) Necessarily, for any S1, S2, P, and C, if S1’s satisfying condition C would confer prima facie justification for S1 to believe P, then S2’s knowing that S1 satisfies C would confer equal prima facie justification for S2 to believe P. 39 Huemer (S&J, sec 6.3) also claims that the details of Steup’s position commit him to a vicious regress:  having E and E1 is not enough, because one would also need, E2, evidence that E1 is reliable . . . and so on. Although Steup could be clearer on this point, he is not subject to any vicious regress. He’s essentially a coherentist who thinks that seemings provide justification when supplemented by a certain kind of coherent belief structure that affirms the reliability of the seemings upon which the subject relies. Steup’s section 8 is relevant here, especially the connection he draws between his view and Sosa’s (nt 24). 40 My own view is that the SPO can motivate a version of AI, but one importantly different from Steup’s. My 2012 essentially uses the SPO to defend an AI about inferential justification. One day I hope to generalize the argument of that paper to defend a completely general version of AI.

An Introduction

25

The denial of AN is referred to as Agent Centeredness (AC). Suppose your seeming that P prima facie justifies you in believing T to degree 9. If I know about your seeming, given AN, this knowledge of your seeming prima facie justifies me in believing T to .9. If AN is false, i.e., if AC is true, then my knowledge of your seeming may provide me with some lower degree of prima facie justification for believing P or no prima facie justification at all. In such a case, S1 and S2 might reasonably disagree even though they recognize each other as peers and they have fully disclosed their evidence. To see this, suppose S1’s total evidence is E + SEEMING, where SEEMING is S’s seeming that T is true. S2’s total evidence, on the other hand, is E + KNOWLEDGE, where KNOWLEDGE is S2’s knowledge of S1’s seeming. Given AC, E + SEEMING might provide ultima facie justification for P when E + KNOWLEDGE provides ultima justification for believing ~P. This result would obtain because we are assuming that AC and that SEEMING provides better justification for P than does KNOWLEDGE. Building on Huemer’s 2011b, DePaul shows that those sympathetic with the conjunction of dogmatism and AC have reason to be optimistic concerning the possibility of reasonable disagreements, even when the subjects recognize each other as epistemic peers and have fully disclosed their evidence.41

5.5 dealing with cognitive penetration The three chapters in Part V try to find principled restrictions which allow dogmatism to avoid worries arising from cognitive penetration, especially the Tainted Source Objection (see sec 3.2 this chapter). In chapter 10, Matthew McGrath divides seemings into two categories: those that are quasi-inferred and those that aren’t. Quasi-inferred seemings have “inference-like” dependence on either another seeming or a belief. Receptive seemings, those that aren’t quasi-inferred, always provide foundational justification for their contents. Quasi-inferred seemings, on the other hand, at best provide derivative justification for their contents. In effect, a quasi-inferred seeming that P derivatively justifies its content P just in case the “basis” of the quasi-inferred seeming also justifies P. Consider the Wishful Willy case. McGrath holds that Willy’s seeming that (G) the object is gold is quasi-inferred from his seeming that (Y)  the object is yellow. Since Y does not evidentially support G by itself or in conjunction with Willy’s desire, Willy’s seeming that G fails to justify, even derivatively, Willy’s belief in G. In chapter 11, Peter Markie explores the Knowledge How Proposal, i.e., that seemings justify only if they are the products of knowledge how. According to

41 The truth of AC might also enervate the Bayesian objections insofar as they rely on the proxy assumption.

26

Seemings and Justification

this proposal, Wishful Willy isn’t justified in believing that the object is gold because Willy doesn’t know how to use visual phenomenology to identify gold nuggets. With respect to perception, knowledge how boils down to a special kind of disposition. With respect to the a priori, knowledge how boils down to sufficient concept possession. Although some of the details are left for another occasion, Markie argues that the early returns on the proposal merit further investment. In chapter 12, Berit Brogaard contends that the key to restricting dogmatism is a principle she calls “Content Grounding.” This principle has two conditions which are individually necessary and jointly sufficient for a seeming to be content-grounded. The first requires seemings to be reliably correlated with the content of a perceptual, introspective, or memory-related experience. The second is reliabilist in character: S’s seeming that P is content-grounded only if S’s seeming that P reliably indicates P. Brogaard then defends what she calls “Sensible Dogmatism,” that a seeming that P provides prima facie justification for P iff the seeming is content-grounded. At first glance, this proposal doesn’t cover intellectual seemings because those seemings aren’t reliably correlated with perceptual, introspective, or memorial experiences. On Brogaard’s view, however, intellectual seemings spring from semantic memory; what we ordinarily call “a priori justification” is merely a special case of memorial justification.

5.6. phenomenal conservatism The three entries in Part VI focus on endorsing or rejecting phenomenal conservatism. In chapter  13, William Lycan provides a novel way of motivating and defending the view. In the early stages of the chapter, Lycan compares his earlier principle with Huemer’s phenomenal conservatism and modifies his Principle of Credulity so that it is simply a version of phenomenal conservatism. Lycan’s position is, however, no mere echo of Huemer’s. Lycan provides entirely different motivations than does Huemer: phenomenal conservatism resolves some problems in his explanationist coherentism, and Mother Nature would design us to rely on seemings. Furthermore, Lycan’s version is considerably more modest than Huemer’s (or mine for that matter) for two reasons. First, Lycan stresses that “the justification furnished by the Principle of Credulity is minute, the faintest edge, infinitesimal if you like” (sec 7), whereas Huemer and I both allow that very strong seemings may provide rather high degrees of justification. Second, and perhaps relatedly, Lycan holds that the justification provided by seemings is defeated more easily than Huemer or I  intend. Suppose you are comparing two theories, T1 and T2, and that T1 strongly seems true to you and T2 strongly seems false to you. As it turns out, T2 is marginally simpler than T1. Huemer and I would want to leave it open that you could still have greater justification for T1 than T2. Lycan, on the

An Introduction

27

other hand, says that any theoretical advantage of T2, no matter how small, will defeat the seeming-based justification you have for T1. In chapter 14, Michael Tooley launches a wide-ranging attack on Huemer’s version of phenomenal conservatism. The three most important lines of attack, perhaps, are as follows. First, Tooley argues that Huemer’s Experience View makes seemings unacceptably opaque or mysterious, and so he replaces it with the Evidence-Taking View. Second, he argues that PC is problematic because it allows seemings to justify even if (i) they are caused by other seemings, and even if (ii) they occur in the absence of any qualia or experience. Third, he argues that his own direct acquaintance approach is superior to phenomenal conservatism. In chapter 15, Michael Huemer replies to several of the entries in this volume and focuses his attention on those by Brogaard, Conee, Markie, McGrath, and Steup. The three issues that receive the most attention are the nature of seemings, cognitive penetration, and the alleged superiority of PC over rival views proposed in the volume. The discussion of seemings is, I  think, particularly helpful, and his replies confirm that phenomenal conservatism is a resilient view that deserves attention in spite of the problems it faces.

Acknowledgments Thanks to Robert Audi, Marinus Ferreira, Matthew Flanagan, Paul Silva Jr, and the audiences at the University of Otago and University of Oxford for helpful discussion on earlier drafts of this chapter. Thanks especially to Michael Huemer for helpfully commenting on multiple drafts and to James Pryor for detailed discussion when the deadline was nigh. My wife, Jenny, graciously provided the photo of the tuatara. Finally, thanks to the University of Auckland’s Faculty Research Development Fund and the Marsden Fast Start Grant for financially supporting the development of this chapter.

References Bealer, George. 2000. “A Theory of the A Priori.” Pacific Philosophical Quarterly 81: 1–30. Bealer, George and P. F. Strawson. 1992. “The Incoherence of Empiricism.” Proceedings of the Aristotelian Society 66: 99–143. Bergmann, Michael. 2006. Justification without Awareness. Oxford: Oxford University Press. BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge:  Harvard University Press. Chudnoff, Eli. 2011. “The Nature of Intuitive Justification.” Philosophical Studies 153: 313–33. Comesaña, Juan. 2010. “Reliabilist Evidentialism.” Noûs 44: 571–600. Cullison, Andrew. 2010. “What Are Seemings?” Ratio 23(3): 260–74.

28

Seemings and Justification

DePaul, Michael. 2009. “Phenomenal Conservatism and Self-Defeat.” Philosophy and Phenomenological Research 78: 205–12. DePoe, J. M. 2011. “Defeating the Self-Defeat Argument for Phenomenal Conservatism.” Philosophical Studies 152: 347–59. Feldman, Richard. 2007. “Reasonable Religious Disagreements.” In Louise Antony, ed., Philosophers without Gods, 194–214. Oxford: Oxford University Press. Fumerton, Richard. 1995. Metaepistemology and Skepticism. Lanham, MD:  Rowman & Littlefield. Goldman, Alvin. 2009. “Internalism, Externalism, and the Architecture of Justification.” Journal of Philosophy 106: 309–38. Hasan, Ali. 2011. “Phenomenal Conservatism, Classical Foundationalism, and Internalist Justification.” Philosophical Studies, doi: 10.1007/s11098-011-9751-0. Huemer, Michael 2011a. “Phenomenal Conservatism and Self-Defeat: A Reply to DePoe.” Philosophical Studies 156: 1–13. _____. 2011b. “Epistemological Egoism and Agent-Centered Norms.” In Trent Dougherty, ed., Evidentialism and Its Discontents, 17–33. Oxford: Oxford University Press, _____. 2009. “Apology of a Modest Intuitionist.” Philosophy and Phenomenological Research 78: 222–36. _____. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. _____. 2006. “Phenomenal Conservatism and the Internalist Intuition.” American Philosophical Quarterly 43: 147–58. Huemer, Michael. 2005. Ethical Intuitionism. New York: Palgrave MacMillan. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman & Littlefield. _____. 1999. “The Problem of Memory Knowledge.” Pacific Philosophical Quarterly, 80: 346–57. Jackson, Alexander. 2011. “Appearances, Rationality and Justified Belief.” Philosophy and Phenomenological Research 82: 564–93. Jackson, Frank. 1977. Perception:  A  Representative Theory. Cambridge:  Cambridge University Press. Jehle, David and Brian Weatherson. 2012. “Dogmatism, Probability, and Logical Uncertainty.” In Greg Restall and Gillian Russell, eds., New Waves in Philosophical Logic, 95–111. New York: Palgrave Macmillan. Kung, Peter. 2010. “On Having No Reason:  Dogmatism and Bayesian Confirmation.” Synthese 177: 1–17. Lycan, William G. 1988. Judgement and Justification. Cambridge: Cambridge University Press. Lyons, Jack. 2011. “Circularity, Reliability and the Cognitive Penetrability of Perception.” Philosophical Issues 21: 289–311. _____. 2009. Perception and Basic Beliefs. New York: Oxford University Press. Markie, Peter. 2005. “The Mystery of Perceptual Justification.” Philosophical Studies 126: 347–73. Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford University Press. Pollock, John L. and Joseph Cruz. 1999. Contemporary Theories of Knowledge. 2nd ed. Lanham, MD: Rowman & Littlefield. Pryor, James. Manuscript. “Uncertainty and Undermining.” _____. 2012. “When Warrant Transmits.” In Annelisa Coliva, ed., Wittgenstein, Epistemology

An Introduction

29

and Mind:  Themes from the Philosophy of Crispin Wright, 269–303. Oxford:  Oxford University Press. 2004. “What’s Wrong with Moore’s Argument?” Philosophical Issues 14: 349–77. _____. 2000. “The Skeptic and the Dogmatist.” Noûs 34(4): 517–49. Rogers, Jason and Jonathan Matheson. 2011. “Bergmann’s Dilemma:  Exit Strategies for Internalists.” Philosophical Studies, 152(1): 55–80. Schiffer, Stephen. 2004. “Skepticism and the Vagaries of Justified Belief.” Philosophical Studies 119: 161–84. Siegel, Susanna. 2013. “The Epistemic Impact of the Etiology of Experience.” Philosophical Studies 162: 697–722. _____. 2012. “Cognitive Penetrability and Perceptual Justification.” Noûs 46: 201–22. _____. 2005. “Which Properties Are Represented in Perception?” In Tamar Szabó Gendler and John Hawthorne, eds., Perceptual Experience, 481–503. Oxford:  Oxford University Press. Skene, Matthew. 2013. “Seemings and the Possibility of Epistemic Justification.” Philosophical Studies 163: 539–59. Sosa, Ernest. 2007. A Virtue Epistemology, Vol. 1. Oxford: Oxford University Press. _____. 1998. “Minimal Intuition.” In Michael DePaul and William Ramsey, eds., Rethinking Intuition, 257–70. Lanham, MD: Rowman & Littlefield. Swinburne, Richard. 2001. Epistemic Justification. Oxford: Oxford University Press. Tolhurst, William. 1998. “Seemings.” American Philosophical Quarterly 35: 293–302. Tucker, Chris. Manuscript. “If Dogmatists Have a Problem with Cognitive Penetration, You Do Too.” 2012. “Movin’ On Up: Higher-Level Requirements and Inferential Justification.” Philosophical Studies 157(4): 323–40. _____. 2011. “Phenomenal Conservatism and Evidentialism in Religious Epistemology.” In Kelly James Clark and Raymond VanArragon, eds., Evidence and Religious Belief, 52–73. Oxford: Oxford University Press. _____. 2010a. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45. _____. 2010b. “When Transmission Fails.” Philosophical Review 119: 497–529. Weatherson, Brian. 2007. “The Bayesian and the Dogmatist.” Proceedings of the Aristotelian Society 107: 169–85. Wedgwood, Ralph. Forthcoming. “A Priori Bootstrapping.” In Albert Casullo and Joshua Thurow, eds., The A Priori in Philosophy. Oxford: Oxford University Press. Weisberg, Jonathan. 2009. “Commutativity or Holism? A Dilemma for Conditionalizers.” British Journal for Philosophy of Science 60: 793–812. _____. Manuscript. “Updating, Undermining, and Independence.” White, Roger. 2006. “Problems for Dogmatism.” Philosophical Studies 31: 525–57. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Blackwell. Wright, Crispin. 2007. “Perils of Dogmatism.” In Susana Nuccetelli and Gary Seay, eds., Themes from G. E. Moore: New Essays in Epistemology and Ethics, 25–48. Oxford: Oxford University Press.

This page intentionally left blank

{ part i }

Seemings and Seeming Reports

This page intentionally left blank

 { 2 }

Seemings and Semantics Andrew Cullison 1. The Nature of Seemings: Some Preliminary Remarks I think we are familiar enough with the notion of a seeming, but a few preliminary remarks are in order. First, there are various kinds of seemings. There are perceptual seemings. It seems that there is a cup on the table. There are memorial seemings. It seems that I had eggs for breakfast. There are rational (intuition-type) seemings. It seems that 2 + 2 = 4. There are moral seemings. It seems that torture for fun is wrong. Second, I  think seemings are sui generis propositional attitudes and are not analyzable in terms of the other mental phenomena. For example, I don’t think seemings are beliefs, partial beliefs, inclinations to believe, attractions to believe, or dispositions to believe.1 Third, while it is appropriate to talk about perceptual seemings, I think it is very important to make a distinction between seemings and sensations. Perceptual seemings should not be identified with perceptual experiences, raw visual feels, or sense data. Someone could have a visual experience that is qualitatively identical to my raw visual experience of a chair, but it not seem to them that there is a chair in front of them. I  suspect that there is likely a causal link between our raw visual experience of a chair and a seeming that there is a chair, but I think there are possible cases where the two might come apart.2 This distinction between sensations and seemings will be important to keep in mind as we proceed. To further keep this distinction

1 I defend this view about seemings in Cullison (2010) so I won’t defend the view here, but I will note that treating seemings as unanalyzable should not trouble us. Unanalyzable concepts are less worrisome if there are clear instances, and seeming talk is so ubiquitous in our language and we are all familiar with the phenomenon of a proposition seeming true. 2 Chris Tucker (2010) has several compelling reasons for believing there is a distinction between perceptual sensations and seemings. First, there are actual cases of blind sight, where people lack the phenomenology of a perceptual experience but have seemings of objects in front of them. Second, the reverse also seems to happen. Some people have certain visual phenomenology but lack any seemings of the things in their visual field.

34

Seemings and Seeming Reports

clear, when I use the term perceptual experience or visual experience, I will be talking about raw feels or visual images, not the seemings that those experiences may cause.3 The role of seemings in epistemology has received a lot of attention; however, there are interesting questions about the nature of the content of seemings that have received considerably less attention. That will be the focus of this chapter. To be clear, issues related to the content of perceptual experiences have received considerable attention, and part of the purpose of this chapter is to illuminate some of the substantive differences between various content issues concerning perceptual experiences on the one hand and those same kinds of content issues concerning the content of seemings on the other hand. Let’s turn our attention to the first content issue.

2. First Content Issue: Seemings and Conceptual Content There are, at least, two distinct issues regarding the content of seemings. The first issue is whether seemings have non-conceptual or conceptual content. I’ll follow Peacocke’s (2001, 243) characterization of the distinction between conceptual and non-conceptual content. He characterizes conceptual content as the sort of content that can be the content of a belief or judgment and that conceptual content is the sort of thing that is truth evaluable. I leave open the question as to what sorts of things might play the role of non-conceptual content. It will not matter for the purposes of this chapter. I think the content of a seeming is a proposition and, thus, conceptual. In this section, I  briefly discuss a few motivations for this view. I  also discuss some reasons that have been offered for thinking that perceptual experiences lack conceptual content and argue that these reasons do not translate well into reasons for thinking that seemings lack conceptual content. While there is a lively debate concerning whether perceptual experiences have conceptual content,4 an attempt to apply these arguments to seemings would ignore important differences between sensations and seemings. One simple motivation for thinking that seemings have propositional content is that we refer to them using that-clauses. It seems to me that there is a table in the room. It seems to me that I had eggs for breakfast. It seems to me 3 This may be a controversial choice for terminology. Susanna Siegel (2006), for example, distinguishes between perceptual experiences and mere sensations. However, there is also some precedent for using the term perceptual experience interchangeably with “mere sensation.” See Siegel’s (2011) discussion of Thomas Reid for an example. In that discussion, it is noted that a part of a non-conceptualist argument against conceptualism about perceptual experiences involves identifying perceptual experiences with sensations. 4 Defenders of the non-conceptual view include Evans (1982), Crane (1992), Dretske (2003), and Peacocke (2001). Defenders of the conceptual view include McDowell (1994), Brewer (1999), and Siegel (2010).

Seemings and Semantics

35

that 2 + 2 = 4. It would be odd if propositional attitude verbs like said that, believes that, hopes that, and desires that expressed a relation between a person and a proposition, but locutions of seems that did not.5 Another prima facie reason to think that seemings have propositions as content, mirrors a simple reason offered for thinking that sentences have propositions as content. Propositions explain how sentences from different languages could mean the same thing. When Juan says “Nieve es blanco” and John says “Snow is white” they have uttered different sentences, but they have also said the same thing. If their utterances encode the exact same proposition, then we can explain how it is they said the same thing. Just as people can express the same proposition with different sentences, people with different phenomenology might have the same thing seem true. Imagine an alien with some kind of sonar-vision. Suppose Bob introduced this alien to tigers. The alien would have radically different internal sensory data from Bob when it encounters tigers, but when they both enter a room it would be very natural to say that it seemed (to both of them) that there was a tiger in the room. I grant that there is quite a bit that seemed different to both of them and discuss that more below. But for now, it’s simply important to note that in an important sense the same thing seemed true to both of them. If seemings have propositional content, we could explain how it’s possible for the same thing to seem true for different persons via radically different raw visual phenomenology. A third motivation for thinking that seemings have propositional content parallels a reason for thinking that perceptual experiences have propositional content. Seemings, like perceptual experiences, are the sorts of things that can be accurate. Suppose it seems to me that there is an orange on the table, but it turns out that I was hallucinating. My seeming was inaccurate. But for a seeming to be inaccurate it would have to encode something that is capable of being true or false. On the assumption that propositions are the primary bearers of truth and falsity, seemings have propositional content. This accuracy argument is discussed at length in the literature on perceptual content.6 However, some have argued that the mere fact that perceptual experiences have accuracy conditions is not sufficient to show that they have conceptual or propositional content. Tim Crane has recently argued as follows: But [the claim that perceptual experiences have conceptual content] does not follow from the fact that experiences can be accurate or inaccurate. Accuracy is not truth, since accuracy admits of degrees and truth does not. (The same can be said of correctness.) (Crane 2009, 458)

5 This doesn’t entail that seemings have propositions as content, but it is good prima facie motivation for thinking that they do. 6 Susanna Siegel (2010) has very recently endorsed a version of this argument.

36

Seemings and Seeming Reports

Crane argues that because accuracy admits of degrees and truth does not, there can’t be a link between something being accurate and it expressing something with a truth value. But it’s unclear how this observation undermines the main thrust of the accuracy argument. We should ask ourselves why something can be more or less accurate. The most natural answer is that degrees of accuracy depend on how close the proposition (or sets of propositions) represented are to the actual truth. Beliefs, for example, are more or less accurate in virtue of how close the proposition believed is to the truth. If I drink Pepsi and judge that it is Coca-Cola, but my friend judges that it is milk, my belief is more accurate because the proposition I believe is closer to the truth. Closeness to truth is a similarity relation between propositions. So, if accuracy just is closeness to the truth, then there is a clear route from the fact that something is accurate to the fact that it has propositional content even if accuracy is not to be identified with the truth. Being identified with closeness to the truth is sufficient.7 Before I move on to a second issue about the content of seemings, I would like to briefly discuss two reasons non-conceptualists have offered for thinking that perceptual experiences lack conceptual content and explain why these reasons would not be good reasons for thinking that seemings lack conceptual content. Some argue that perceptual experiences are so rich that they far outstrip the conceptual apparatus we have to describe our experiences, so there must be some content contained in perceptual experiences that is non-conceptual.8 However, when people think about the richness of perceptual experience they seem to have in mind the sensations and raw visual images that are present in cases of perception. Seemings, understood as something distinct from the sensations, are not rich in the sense that sensations are, so it’s unclear how this popular reason for non-conceptualism about experiences could be modified to apply to seemings. A second strategy for resisting the thesis that perceptual experiences have accuracy conditions and conceptual content argues that we confuse

7 Crane has an auxiliary argument to bolster his position concerning accuracy. Pictures can be more or less accurate, but Crane argues that we should not think pictures have propositional content. Crane argues that this is an example of something capable of accuracy but not having propositional content. However, it seems clear that pictures do have propositional content. And pace Crane, you can make assertions with drawings and pictures. Crane argues that you cannot make an assertion with a picture without also saying something. This seems false. Political cartoonists are in the business of saying things with their drawings without also saying something using non-pictorial language. If someone draws a thick but narrow moustache on a political leader, he has said something about that leader. 8 See Byrne (2005) for a clear presentation of this argument. He also has a good candidate response to this argument. I do not go into that response in detail, because my primary concern here is to note that this argument does not apply to seemings even if it is successful against the thesis that perceptual experiences have conceptual content.

Seemings and Semantics

37

experiences and the judgments or beliefs that they often cause. Some maintain that perceptual experiences just are “raw feels, or sensory affectations of the subject that do not purport to represent the world in anyway at all.” If we construe experiences this narrowly, then it might not make sense to assess them for accuracy conditions. But proponents of this strategy argue that we can easily see why they might think experiences have content if they often occur to us in conjunction with something that does have accuracy conditions, like a judgment.9 Note that this second argument is a less compelling argument if we think of seemings as a distinct thing from a raw feel or sensory affectation of the subject. Something can seem true to someone without that person having a sensory experience and without having made a belief or judgment, and yet in those cases it still seems like we’re dealing with something that has accuracy conditions and conceptual content. If that is the case, then we cannot explain away temptations to think that seemings have content as a confusion between seemings and the things with content they cause. Furthermore, seemings might be precisely the sort of thing that non-conceptualists about experiences could appeal to in order to explain why people are tempted to think that perceptual experiences (understood as raw visual feels) have conceptual content. Seemings should be part of the non-conceptualists’ strategy for arguing that we confuse perceptual experiences with other contentful mental states that are intimately connected to perceptual experiences. Seemings would simply be one of those contentful states. As we shall see, the distinction between seemings and sensations may help to illuminate other aspects of the debate about the nature of perceptual content. Let’s turn our attention to the second content issue.

3. Second Content Issue: Fregeanism versus Russellianism about Seemings A second question about the content of seemings is what sort of propositional content do they have? In the literature on the nature of perceptual content, there are two dominant views. Fregeanism

the content of a perceptual experience is a descriptive proposition or mode of presentation. Russellianism the content of a perceptual experience is a structured proposition constituted by objects and properties.

These views correspond to the analogue views in philosophy of language about the content of sentences. According to Russellianism, the content of a proper 9

See Siegel (2011) for a good summary of this strategy.

38

Seemings and Seeming Reports

name is the object it refers to, the content of a predicate is the property it refers to, and the content of a sentence is a structured proposition that is constituted by objects and properties. The Fregean view about the content of perceptual experiences closely parallels Fregeanism about names, kind terms, and sentences. The content of a name is a sense or mode of presentation. One classic Fregean view is that the content of a name is a description (or cluster of properties) that the speaker associates with a name. The same goes for kind terms. In this section, I  defend the thesis that seemings often encode singular, Russellian, structured propositions by arguing against a candidate Fregean alternative. I say often because Russellianism about seemings need not maintain that seemings never encode non-singular propositions. It can seem to me that Tony is over there, but it can also seem to me that there are some tigers over there. When we would describe a seeming using a proposition that Russellians would maintain encodes a singular proposition, I  think the seeming has a Russellian content. The natural Fregean alternative to Russellianism about the content of seemings that I have in mind is very much like classical descriptivism in philosophy of language about proper names and sentences. According to classical descriptivism, the meaning of a proper name is some description that a speaker associates with that name. To articulate descriptivism about seeming states, one would presumably need an analogue to proper names in natural languages that are tokened in the head. Whatever this thing is that plays the role of a name in the person’s head, let’s call that a mental proper name. A Russellian about mental proper names would maintain that the content of that mental proper name just is the individual picked out by the name. Descriptivism about mental proper names would maintain that the meaning of that mental proper name is always some description the person associates with the name. Now we can ask ourselves, is the meaning of a mental proper name always some definite description? I think there are interesting Kripke-esque reasons to think that classical descriptivism about certain seemings is false. Patrick Somerville is the author of This Bright River. Suppose that is the only thing that Sarah knows about Patrick Somerville. Now, consider the following sentences that describe possible seeming states for Sarah. (A) It seems that if the author of This Bright River is the author of This Bright River, then Patrick Somerville is the author of This Bright River (B) It seems that if the author of This Bright River is the author of This Bright River, then the author of This Bright River is the author of This Bright River.

Suppose we ask Sarah if these sentences accurately describe her. Sarah might say that (B) definitely does. If Descriptivism about Mental Proper Names is true, then we would be committed to the thesis that (A)  also describes her seeming state, the very same seeming state as (B). This may be a consequence

Seemings and Semantics

39

that descriptivists can embrace. If Sarah really does associate “the author of This Bright River” with “Patrick Somerville,” then perhaps it’s not so bad to think that the seemings described in (A) and (B) have the same content. But we can push the objection further in a way that should be more troubling. Seemings come in degrees. Some things seem more strongly true than others. It seems to me that I have two hands. And that seeming is much stronger than my seeming that there is a car parked outside of my window. So, consider (A) and (B) but add strongly in the appropriate place and you get (C) and (D) below. It strongly seems that if the author of This Bright River is the author of This Bright River, then Patrick Somerville is the author of This Bright River. (D) It strongly seems that if the author of This Bright River is the author of This Bright River, then the author of This Bright River is the author of This Bright River.

(C)

While Sarah (and anyone else) who associated “the author of This Bright River” with her mental tokenings of “Patrick Somerville” might readily assent to both (A)  and (B), even someone like Sarah could accept (C)  but reject (D). This gives us some prima facie reason to reject the view that seemings involving individuals always encode purely descriptive propositions. In particular, it seems that the mental analogue of proper names ought not be regarded as having purely descriptive content that the person associates with the name. What’s interesting about this objection to descriptivism about mental proper names is that it seems to resist some common attempts to respond to Kripke’s original modal and epistemic objections to descriptivism. Many descriptivists maintain that Kripkean objections fail, because they fail to accurately characterize what the description is that people associate with a name. Rigidified descriptivism is a popular example. Rigidified descriptivists maintain that actual is always part of the description we associate with a name.10 No one ever really associates the author of This Bright River with Patrick Somerville according to Rigidified Descriptivism, what someone who seems to do this really associates with Patrick Somerville is something like—the actual author of This Bright River. So, for any occurrence of “Patrick Somerville” in (C)  we should substitute “the actual author of This Bright River.” This makes (C) and (D) have different content, and so it might pose a problem for Kripke’s original modal and epistemic objections, but it’s less clear that it avoids the worry mentioned earlier with respect to seemings. However, there are other pairs of sentences involving strongly seems where even Rigidified Descriptivism gets the wrong result.

10

See Everett (2005) for a good survey of some recent defenses of descriptivism.

40

Seemings and Seeming Reports

It strongly seems that Patrick Somerville is the actual author of This Bright River. (F) It strongly seems that the actual author of This Bright River is the actual author of This Bright River. (E)

Rigidified Descriptivists must say that (E)  and (F)  have the same content. However, someone who is fully convinced that Patrick Somerville was the actual author of This Bright River might still consider those two sentences and have one strongly seem true, but not the other. Despite being fully convinced that Patrick Somerville is the actual author of This Bright River, the person might still consider it enough of a live option that he is not, and this consideration might weaken the strength of the seeming characterized by (E) but not weaken the seeming characterized by (F). A second reason to be worried over descriptivism about mental proper names is related to one of the motivations mentioned earlier for thinking that seemings have propositional content. I noted that the same thing can seem true to people with radically different internal phenomenology. Imagine that Bob and the alien meet Tony the Tiger. Suppose Bob and the alien observe Tony eating Frosted Flakes, and they both even utter “Tony is eating frosted flakes.” I can accurately report their seeming states. It seems to both of them that Tony is eating frosted flakes, but part of the descriptive content that the alien might associate with his internal mental proper name likely involves descriptions of weird phenomenal qualia that he has that Bob doesn’t. So if the contents of their seemings states are always descriptions they each associate with mental proper names, we can’t say the same thing seemed true to both of them. I don’t think my objections to classical descriptivism are necessarily decisive, but in conjunction with some basic motivations for the Russellian view about the content of seemings, they yield prima facie reason to accept a Russellian view for the contents of many seemings. I think we should only reject it if we have compelling objections against it. In the next two sections, I’ll consider some possible objections to Russellianism about seemings.

4. Some Objections: Russellianism about Sensations versus Russellianism about Seemings In this section, I  present two objections to Russellianism about the content of perceptual experiences and argue that they are not plausible objections to Russellianism about the content of seemings. The first, I will call The Argument from Color and Location. Imagine two individuals, Jack and Jill, sitting in separate rooms. Each individual is looking at a ball on the table in their respective rooms. Suppose the balls are objectively different colors: they reflect different frequencies of light. However, you can imagine that Jack and Jill are having phenomenally identical experiences.

Seemings and Semantics

41

What it is like in both cases for each of them is the same. Fregeans about the content of experiences argue that there is an important sense in which they are having the same experience, and so the content must be the same. They argue that a Fregean view is better equipped to capture these facts. If the content of their two experiences are restricted to Russellian contents, then the contents must not be the same because the content of Jack’s experience will encode a different color property and a different location property than Jill’s experience.11 There are two reasons to think this is not a successful argument against Russellianism about seemings. First, there are plausible responses to this argument for those who wish to defend Russellianism about perceptual content. Second, the differences between experiences (understood as visual feels or sensations) and seemings make it difficult to construct a parallel argument against Russellianism about seemings. Let’s consider each reason in turn. A Russellian about perceptual experiences might note that we should say something similar in this case that they would say about Twin Earth Cases. Consider what a Russellian might say about Twin Earth cases. Imagine a pair of individuals, Oscar and Twoscar. Oscar and Twoscar are, in some very important sense, perfect internal duplicates. They appear to speak the same language and, for our purposes, let’s assume they have the same internal phenomenal qualia. However, Twoscar lives on Twin Earth. On Twin Earth, the clear, potable liquid that Twoscar drinks and that fills the rivers and oceans of Twin Earth is not H20; it has a completely different chemical composition— XYZ. This thought experiment is used to motivate content externalism in philosophy of language and philosophy of mind because, intuitively, Oscar and Twoscar have different thoughts. One is thinking about water and the other is thinking about XYZ. However, when Oscar says “Water is wet” and Twoscar says “Water is wet” there is still some temptation to think that they believe the same thing.12 Russellians will point to something other than the content of the belief to capture the sameness intuitions. Oscar and Twoscar might have similar modes of presentation when they think about water and XYZ, respectively, but the contents of their thoughts are different. Perhaps they grasp this content via these modes, but the modes are not themselves the content. 11 Chalmers (2004) discusses a version of this argument. A related objection to Russellianism about the contents of perception deals with thinking about brain-in-a-vat twins. Imagine someone who is an internal duplicate of you, but is a brain-in-a-vat being deceived into thinking he or she is in a world like yours. He or shee has all of the same internal phenomenal properties as you do. We might go so far as to say that everything visually appears exactly as it would if your twin were in the real world. It’s plausible to suppose that your perceptions and your twin’s perceptions are representing the world to each of you in the same way, so the content of those perceptions must be (largely) the same (cf. Chalmers 2004). This objection is very similar to the objection from color and location; it has just been expanded to include a wider range of objects and properties. The responses I’ll offer to the argument from color and location will apply equally well to this objection. 12 Putnam (1973), (1996) presents these cases to argue against semantic internalism, the view that semantic content is entirely internal to the mind.

42

Seemings and Seeming Reports

Russellians about the contents of perceptual experiences could say something similar in the case of phenomenologically identical perceptions of distinct properties. There are ways in which colors and locations can be presented to us, and persons could have different contents in their experiences represented to them via similar internal ways, or modes of presentation. The ways or modes in which properties are represented to us need not be the contents of the perceptual experiences themselves. Let’s turn our attention to the second reason that the argument from color and location should not trouble the Russellian about seemings. Even if the Argument from Color and Location is successful against Russellianism about perceptual experiences, it won’t necessarily translate well into an argument against Russellianism about seemings. Again, this is where it is important to keep the distinction between sensations and seemings in mind. As noted earlier, the same proposition could seem true to people with radically different phenomenology. Just as there could be persons with different phenomenology such that it seems to both of them that there is a tiger in the room, persons with same phenomenology could have, in a very important sense, different seemings. If it’s possible for people with the same phenomenology to have different things seem true, then it’s unclear how a case where Jack and Jill have the same phenomenology would put pressure on us to reject Russellianism about seemings. Consider the following case. Imagine Loki discovers Twin Earth before Thor does and tricks the people of Twin Earth into thinking that he (Loki) is their God. Wanting to be as successful as Thor, Loki imitates everything Thor did on Earth and disguises himself so that he looks exactly like Thor does to the people on Earth; he also makes sure to replicate Thor’s interactions with Earthlings. Now imagine Loki reveals this to Thor. He might say things like the following: It seems to Oscar on Earth that you (Thor) came down from the sky. But on Twin Earth it seems to Twoscar that I, Loki, came down from the sky. Sure it seems to each of them that their God has features that look at lot like your actual features, but it’s me that Twoscar is having those seemings about. To them I, Loki, seem to be that way.

Here, again, the difference between seemings and sensations is relevant. It may be that Oscar and Twoscar’s sensations descriptively represent the world the same way, but those descriptive representations are not the entire story. It may also be that those raw sensations lead to de re seemings with distinct contents. So, again, we should be careful to move from considerations about the nature of the content of sensations or visual experiences to the nature of the content of seemings. A Russellian view about the content of seemings would enable us to preserve the idea that in some sense, to Oscar it seems that Thor is swinging a hammer, but to Twoscar it seems that Loki is swinging a hammer. This is

Seemings and Semantics

43

important, because this is why the Argument from Color and Location won’t translate well into an Argument about Russellianism about seemings. We don’t have to think that internal duplicates have to have all of the same content with respect to their seemings (even if we thought they had to have the same content with respect to their raw visual experiences), which is something the parallel arguments would need to presuppose. A second objection to Russellianism about the phenomenal content of perceptual experiences comes from Bradley Thompson. The core idea behind Thompson’s objection is as follows: Consider the way that a ripe tomato looks at noon in full sun versus at five o’clock in the afternoon. The tomato, both at noon and in the late afternoon, will appear to be the same colour. This is so even though the lighting conditions might change between noon and five o’clock. One might view the tomato continuously during that time. Clouds will occasionally pass overhead, changing the amount and proportions of the spectrum of light that reaches the tomato and your eyes. As the sun sets, the atmosphere filters the light in different ways, changing the percentages of various wavelengths of light that strike the tomato and are then reflected from the tomato to the retina. Light from the sun as it sets is heavier on the red end of the spectrum than the more bluish light at midday. These changes in illumination are adjusted for by the visual system in such a way that the tomato will appear to be the same colour despite differences in the spectral distribution of the light reaching the eye. This ability of the visual system to extract information about relatively stable surface colour across variations in illumination is called “colour constancy.” But despite the fact that the tomato will look to be the same colour, the tomato nonetheless looks different at noon versus five o’clock. (Thompson 2006, 79)

Again, the distinction between seemings and sensations may give the Russellian about the content of visual experiences (not necessarily seemings) the resources to reply to Thompson. The Russellian could say that we should still be Russellians about the contents of the visual experience and explain the sameness intuitions by appeal to a particular seeming that is generated by these visually distinct sensations. The sensations and their contents really do change. It doesn’t look the same. But it does seem that the tomato has not changed color. We don’t need Fregean contents of the sensations to capture the sameness intuitions here. Russellian contents of both the sensations and seemings would capture the intuitions here. It is tempting to think that the issues concerning the content of seemings in cases of perception just are issues concerning the content of perceptual experiences, but I hope this section has helped drive a wedge between these two issues. Many of the arguments against Russellianism about perceptual experiences aren’t clearly applicable to Russellianism about seemings precisely

44

Seemings and Seeming Reports

because of the differences between sensations and seemings and the fact that the arguments about perceptual content tend to focus on aspects of perceptual experience that are peculiar to sensations and not seemings.

5. Frege’s Puzzle and Seemings In the previous section, I focused on objections to Russellianism about perceptual experiences and explained how the difference between seemings and sensations renders those arguments less effective against Russellianism about seemings. In this section, I focus on a more general problem for Russellianism in philosophy of language called Frege’s Puzzle. I defend a particular solution called Ways-Millianism. However, this solution is more difficult to apply to seemings because of the role that seemings are supposed to play in certain theories of epistemic justification. Proponents of views that maintain that seemings play a role in justification may need to modify their epistemic principles to embrace the Ways-Millian solution. The Ways-Millian solution may also be committed to a particular conception about the nature of evidence. I will defend both consequences. As noted earlier, according to Russellianism, the content of a proper name is the object it refers to, the content of a predicate is the property it refers to, and the content of a sentence is a structured proposition that is constituted by the objects and properties that its parts refer to. A classic problem for Russellianism is Frege’s Puzzle. Consider the following two sentences. (G) Mark Twain is the author of Huckleberry Finn. (H) Samuel Clemens is the author of Huckleberry Finn.

If Russellianism is correct, then (G)  and (H)  express the same proposition. That seems problematic. It seems obvious that (G) and (H) do not express the same proposition. The problem seems worse when we consider sentences like the following with propositional attitude verbs. (I) Sarah believes that Mark Twain is the author of Huckleberry Finn. (J) Sarah believes that Samuel Clemens is the author of Huckleberry Finn.

While one might simply insist that (G) and (H) express the same proposition, it seems that (I) and (J) clearly express different propositions. Surely, (I) and (J) could differ in truth value. If (I) and (J) could differ in truth value, then they do not express the same proposition. If (I) and (J) do not express the same proposition, then neither do (G) and (H). Naive Russellians have a problem. We can solve this problem by appeal to modes of presentation or ways of believing. The solution takes belief to be a mediated relation. The belief relation is still construed as a relation between a person and a proposition; however, persons believe propositions via some way of believing. A person believes

Seemings and Semantics

45

a proposition in virtue of standing in some psychological relation to some third thing. One theory of what ways are holds that ways are sentence-like mental representations. According to this theory, people believe propositions in virtue of standing in some relation to these representations. Let’s call the relation that a person bears to the sentence like mental representations “accepting.” So a person believes a proposition P by accepting a sentence-like mental representation that semantically expresses P.13 Call the combination of this metaphysics of belief with Naive Russellianism “Ways-Millianism.”14 Ways-Millianism can explain why we have the intuition that (I) and (J) could differ in truth value. Consider Sarah again. She could believe Mark Twain is the author of Huckleberry Finn by accepting some mental sentence, say, “Mark Twain is the author of Huckleberry Finn.” However, she could fail to accept the mental sentence “Samuel Clemens is the author of Huckleberry Finn.” We could say that Sarah believes Mark Twain is the author of Huckleberry Finn (in the Twain way), but fails to believe Mark Twain is the author of Huckleberry Finn (in the Clemens way). Since Sarah could believe the same proposition in one way and fail to believe it in another way, it is easy to see how we might have the intuition that (I) and (J) could differ in truth value. We imagine Sarah accepting a sentence like “Mark Twain is the author of Huck Finn,” so we think (I) is true. However, we think that she could accept that sentence while also accepting the denial of the sentence “Samuel Clemens is the author of Huck Finn,” so we think that (J) is false. However, we are simply mistaken in our judgment about (J). (I) and (J) do not differ in truth value, but it is easy to see why we might think so.15 Once we realize that we can generate a Frege’s puzzle for Russellianism using that-clauses such as believes that, it should be clear that the puzzle could generalize to other that-clauses to produce similar counterintuitive consequences. The Ways-Millian must relativize any propositional attitude verb to ways. You can fear that Darth Vader is flying the fighter plane in one way (when you think about him in the “Darth Vader” way), but you can fail to fear that Anakin Skywalker is flying the fighter plane (when you think about him in the “Anakin Skywalker” way). Similar issues arise with respect to seemings if they are propositional attitudes with Russellian contents. Suppose Sarah was friends with Samuel Clemens when he was a boy, and doesn’t yet realize that “Mark Twain” and “Sam Clemens” refer to the same individual. Suppose she

13 Braun (2001) suggests this theory of ways, but notes that his defenses of Millianism/Russellianism do not rely on this particular theory. 14 It’s called Ways-Millianism because the modification to Naive Russellianism focuses on the Millian direct reference thesis that Naive Russellians endorse. 15 For further discussion of Frege’s Puzzle, see Salmon (1986). For detailed discussion of how WaysMillianism can respond to various problems for Naïve Russellianism, see Braun (2000; 2001; 2002).

46

Seemings and Seeming Reports

sees an older man at the bar who looks like classic pictures of Mark Twain. Now consider the following two propositions. (K1) It seems that Mark Twain is drinking a pint of beer. (K2) It seems that Samuel Clemens is drinking a pint of beer.

Someone might argue that (K1) is true of Sarah, but (K2) is false. Sarah herself would assent to (K1) but not (K2). This suggests that (K1) and (K2) do not express the same proposition, but Russelianism about seemings is committed to denying this. We can accommodate our intuitions that they differ in truth value by relativizing seemings to ways, just like the Russellian can relativize belief to ways. It seems to Sarah (in the Mark Twain Way) that Mark Twain is drinking a pint of beer. It does not seem to Sarah (in the Samuel Clemens Way) that Mark Twain is drinking a pint of beer. However, the extension of the Ways-Millian strategy is not as simple as this because of the role that seeming states in epistemic justification. WaysMillians who want to embrace something like phenomenal conservatism need to state the theory to take this into account.16 Phenomenal Conservatism is the following thesis: PC

If it seems to S that P, then (absent defeaters) S is justified in believing that P.

Consider the following instance of phenomenal conservatism: (1)

If it seems to Sarah that Mark Twain is drinking a pint of beer, then (absent defeaters) Sarah is justified in believing that Mark Twain is drinking a pint of beer.

(1) identifies a connection between a kind of evidence Sarah has and what Sarah is justified in believing. When Sarah goes into the bar and sees Mark Twain drinking a beer, she sees him in a way that corresponds best with her “Mark-Twain-way” of thinking about Mark Twain. Now consider the following two mentalese sentences, both of which express the same proposition. (MT) “Mark Twain is drinking a pint of beer.” (SC) “Samuel Clemens is drinking a pint of beer.”

It seems plausible from this example that Sarah is rational in accepting (MT), but not rational in accepting (SC). If this is true, then phenomenal conservatism needs to be clarified in order to accommodate this. The most natural proposal would be to relativize phenomenal conservatism to ways of believing. PC-W

16

If it seems to S that P in way W, then S is prima facie justified in believing P in way W.

This has already been recognized by Heimir Geirsson (2002).

Seemings and Semantics

47

This is a simple modification to PC that I think we should accept. A full defense of this modification is beyond the scope of this chapter, but this certainly seems like the most natural way for a Ways-Millian to embrace something like PC in light of the previously mentioned Frege’s Puzzles.17

6. Two Virtues of the Naive Russellian View about Seemings In this last section, I highlight two virtues of the position about seemings that I have defended.

6.1. epistemic internalism and semantic externalism It has been argued that epistemic internalism about justification is incompatible with semantic externalism. Recall Oscar and Twoscar from Twin Earth. They are internal duplicates. But if semantic externalism is true, then they are justified in believing different propositions. One might think that Oscar’s total internal evidence justifies him in believing that he is drinking water and Twoscar’s total evidence justifies him in believing that he is drinking XYZ. If they are internal duplicates and have the exact same internal states, then we have a reason to reject internalism and some versions of evidentialism.18 A Ways-Millian (about belief) will say that Oscar and Twoscar have their internal mentalese sentences (or way of believing) linked to different external world things. And so while they might accept the same way of believing or mentalese sentence, they believe different things because of the accidental facts about which ways of believing represent which external things. A proponent of phenomenal conservatism who extends Ways-Millianism to relativizing seemings to ways can say something similar and reconcile internalism 17 To my knowledge the most worrisome potential consequence of this modification is that it might entail that the ways in which propositions seem true are evidence rather than the propositions themselves. However, this seems not problematic to me. First, there are very good reasons to think that mental states are evidence not propositions. John Turri (2009) calls this view statism. Conee and Feldman (2008, 2011) and Turri (2009) both have good defenses of the view that evidence is a mental state not a proposition. Dougherty (2011) defended the view that evidence is propositional, but he has recently converted to a mental state view (cf. Dougherty and Rysiew, forthcoming). Second, it’s not clear this modification would be committed to statism anymore than ordinary phenomenal conservatism. If you thought seemings were evidence and not the propositions themselves, then you’d be saddled with a similar consequence. Third, people on both sides of the statist/non-statist debate have been attracted to some version of phenomenal conservatism; apparently there is room in that debate to be either a statist or non-statist and still embrace phenomenal conservatism. If that is true, then it’s unclear to me why someone who embraced a more precise formulation of phenomenal conservatism that relativized to ways of seemings would be committed to statism. 18 The more general issue here is that there is a prima facie conflict between semantic externalism and epistemic internalism. Goldberg (2007) contains several essays that discuss this problem, including ones by Brown (2007), Conee (2007), and Fumerton (2007).

48

Seemings and Seeming Reports

about justification with externalism about semantic content. Oscar might have a way in which it seems to him that there is water on the floor and that makes it rational for him to accept a mentalese sentence that expresses the proposition there is water on the floor. When Twoscar encounters XYZ on the floor he experiences the same way of seeming that Oscar experiences and accepts the same mentalese sentence that Oscar accepts. While they believe different propositions and are justified in believing different propositions, there is an important sense in which their internal states, the ways of seemings, are the primary determiners of what they are justified in believing. A view that relativizes seemings to ways preserves this core internalist idea while maintaining the core of externalism about content.

6.2. clarify issues in the debate about the content of perceptual experiences Susana Siegel argues for the Rich Content View of Perceptual Experiences. According to the Rich Content View, perceptual experiences encode a wide range of properties that are more complex than the simple, observable properties that are more obviously part of the content of perceptual experiences. An expert on trees could, for example, see that the tree in front of her is a sugar maple. The property being-a-sugar-maple can be part of the content of a perceptual experience on Siegel’s view. Non-conceptualists argue that there could be other entities in the neighborhood that are always present in the case of an expert’s experience, such as a judgment. They defend a more austere thesis about the contents of the perceptual experiences themselves. More austere conceptualists do this too. One could, however, by making a distinction between seemings and sensations, defend a rich content view of seemings and be neutral about the content of the sensations themselves. So, we could be non-conceptualists (or austere conceptualists) about the content of perceptual experiences (understood as sensations or raw visual images) and rich-content conceptualists about seemings. We then have a way to preserve several non-conceptualist intuitions and several conceptualist intuitions including conceptualists who endorse a Rich Content View. The upshot is that making a distinction between seemings and sensations, and endorsing a rich content view of seemings that permits seemings to sometimes have Russellian contents could go great lengths to resolving some disputes in the philosophy of mind and the nature of perceptual content. Related to this, there is an additional, epistemic virtue of endorsing a Naive Russelian view of seemings. One central concern for non-conceptualism about the content of perceptual experiences comes from McDowell. He argues that conceptualism about perceptual experiences is the only way to explain how perceptual experiences could justify beliefs. As McDowell notes, “The point of

Seemings and Semantics

49

the claim that experience involves conceptual capacities is that it enables us to credit experiences with a rational bearing on empirical thinking” (1994: 52). If perceptual experiences are distinct from seemings, but stand in some important causal connection, then one could be a non-conceptualist about perceptual experiences (e.g., sensations), but a conceptualist about seemings. If seemings are ultimately what do the justifying, then McDowell’s view and the non-conceptualist views are both to some extent vindicated. If we go on to further embrace the thesis that the contents of seemings are sometimes Russellian, then we have a nice rich content view of seemings that is well suited to justify a wide range of propositions that phenomenal conservatives tend to think that seemings can justify.

7. Conclusion I think a Russellian view about the content of seemings is very plausible. It is resilient to arguments against conceptualism about perceptual experiences and Russellianism about perceptual experiences, in part, because of the differences between seemings and perceptual experiences. However, even if my arguments are not persuasive, I hope it is clear that issues concerning the content of seemings on the one hand and the content of perceptual experiences or sensations on the other hand are distinct issues, and keeping the distinction in mind may aid in our thinking about important issues in philosophy of mind, philosophy of language, and epistemology.

Acknowledgments Thanks to Neil Feit, Dale Tuggy, Stephen Kershnar, and Chris Tucker for helpful comments on this version of the chapter. Earlier versions benefited from very helpful discussion from Richard Feldman, Earl Conee, Ed Wierenga, Chris Tillman, David Braun, Joshua Spencer, Andrei Buckareff, and Dan Mittag.

References Braun, David. 2002. “Cognitive Significance, Attitude Ascriptions and Ways of Believing Propositions.” Philosophical Studies 108: 65–81. _____. 2001. “Russellianism and Prediction.” Philosophical Studies 105: 59–105. _____. 2000. “Russellianism and Psychological Generalizations.” Noûs 34: 203–36. Brewer, Bill. 1999. Perception and Reason. Oxford: Oxford University Press. Byrne, Alex. 2005. “Perception and Conceptual Content.” In Ernest Sosa and Matthias

50

Seemings and Seeming Reports

Steup, eds., Contemporary Debates in Epistemology, 231–50. Oxford: Blackwell. Brown, Jessica. 2007. “Externalism in Mind and Epistemology.” In Sanford C. Goldberg, ed., Internalism and Externalism in Semantics and Epistemology, 13–43. Oxford: Oxford University Press. Chalmers, David. 2006. 2004. “The Representational Character of Experience.” In Brian Leiter, ed., The Future for Philosophy, 153–181. Oxford: Oxford University Press. Conee, Earl. 2007. “Externally Enhanced Internalism.” In Internalism and Externalism in Semantics and Epistemology, 51–67. Conee, Earl. and Richard Feldman. 2008. “Evidence.” In Quentin Smith, ed., Epistemology: New Essays, 83–104. New York: Oxford University Press. Crane, Tim. 2009. “Is Perception a Propositional Attitude?” Philosophical Quarterly 59(236): 452–469. _____. 1992. The Contents of Experience:  Essays on Perception. Cambridge:  Cambridge University Press. Cullison, Andrew. 2010. “What Are Seemings?” Ratio 23(3): 260–74. Dougherty, Trent. 2011. Evidentialism and Its Discontents New York: Oxford University Press. Dougherty, Trent and Patrick Rysiew. Forthcoming. “Experience First.” In Ernest Sosa, Matthias Steup, and John Turri, eds., Contemporary Debates in Epistemology. Oxford: Blackwell. Dretske, Fred. 2003. “Sensation and Perception.” In York H. Gunther, ed., Essays on Nonconceptual Content, 25–41. Cambridge, MA: MIT Press. Evans, Gareth. 1982. Varieties of Reference. Oxford: Oxford University Press. Feldman, Richard. 2003. Epistemology. Upper Saddle River, NJ: Prentice Hall. _____. 1993. “Proper Functionalism.” Noûs. 27,1 34–30. Feldman, Richard and Earl Conee. 1985. “Evidentialism.” Philosophical Studies 48: 15–34. Fumerton, Richard. 2007. “What and About What Is Internalism?” In Internalism and Externalism in Semantics and Epistemology, 35–51. Geirsson, Heimir. 2002. “Justification and Ways of Believing.” Disputatio 12: 43–53. McDowell, John. 1994. Mind and World. Cambridge: Harvard University Press. Peacocke, Christopher. 2001. “Does Perception Have a Nonconceptual Content?” Journal of Philosophy 98(5): 239–264. Putnam, Hilary. 1996. “The Meaning of Meaning.” In Heimir Geirsson and Michael Losonsky, eds., Readings in Language and Mind, 157–197. _____. 1973. “Meaning and Reference.” Journal of Philosophy, 70(19): 699–711. Salmon, Nathan. 1989. 1986. Frege’s Puzzle. Cambridge, MA: MIT Press. Siegel, Susanna. 2011. “The Contents of Perception.” In Edward N. Zalta, ed., The Stanford Encyclopedia of Philosophy (Winter 2011), http://plato.stanford.edu/archives/win2011/ entries/perception-contents/. _____. 2010. The Contents of Visual Experience. Oxford: Oxford University Press. _____. 2006.  “Subject and Object in the Contents of Visual Experience.” Philosophical Review 115: 355–88. Speaks, Jeff. 2009. “Transparency, Intentionalism, and the Nature of Perceptual Content.” Philosophy and Phenomenological Research 79(3): 539–573.

Seemings and Semantics

51

Thompson, Brad. 2006. “Colour Constancy and Russellian Representationalism.” Australasian Journal of Philosophy 84(1): 75–94. Thompson, Brad. 2009. “Senses for Senses.” Australasian Journal of Philosophy 87(1): 99–117. Tucker, Chris. 2010. “Why Open-minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–545. Turri, John. 2009. “The Ontology of Epistemic Reasons.” Nous 43: 490–512.

{3}

Seeming Evidence Earl Conee 1. Introduction Sometimes we attempt to make a thorough airing of our reasons for thinking that something is true. When we do this, sooner or later we have occasion to cite how things seem to us, as in the following exchange. Smith reports to Jones, with some consternation, “A snake is slithering toward us.” Jones asks Smith, “What reason do you have to think so?” Smith replies, “I see the thing.” Jones follows up, “And what reason do you have to think that you see an approaching snake?” Smith replies, carefully but with rising alarm, “I am visually experiencing what looks to me exactly like a snake slithering right at us!” Jones continues the inquiry, “And what reason do you have to think that you are experiencing that?” Smith replies, again carefully but on the verge of flight, “Well, it seems to me that I am!” Jones inquires again, “And what reason do you have to think that it seems to you are having that experience?” Smith replies, after reflecting intensely but quickly, “It just does seem that way to me!” Smith flees.

As Smith’s final response illustrates, when we consider the reason that some claim of seeming truth serves to report, at least initially nothing presents itself as a candidate to be support that we have for that reason. Thus, when we delve as deeply as we can into our ultimate rational resources, attributions of seeming truth can come to play a central role. It is easy to be led by this to believe that seeming truth is our ultimate evidence for anything and everything.

Seeming Evidence

53

It will be argued here that a proposition seeming to us to be true is never our ultimate evidence for it. When we correctly claim that it seems true, this is a way to communicate that we have a certain sort of consideration in support of the content proposition. Usually the consideration actually supports the proposition. But the state of affairs of a proposition seeming to us to be true, if there is such a state of affairs, is not evidence on its own for the proposition. That is, there is no relation of seeming truth, relating a person to a proposition, such that being in that relation on its own gives the person evidence for the truth of the proposition. Evidential support is usually involved. It is of varied sorts. When a claim of seeming truth is correct, there is a combination of two psychological factors that might have a positive evidential bearing on the proposition. First, some occasioning mental event occurs, such as visually experiencing or recalling. Second, the person is inclined to regard the event as a certain type of rational support for the content proposition. The person is inclined to regard the event as a type of support that makes the truth of the proposition apparent; its being true is manifested or displayed in some way. Various influences determine the existence and strength of the support from the putatively supporting mental event. The result of this variety is that correct claims of seeming truth do not having any uniform justificatory role. This view of seeming truth will be elaborated here. Also, two arguments that threaten the view will be criticized. One argument is by Chris Tucker. It contends that seemings do the justifying that is attributed here to sensory events, memory impressions, and other evidence that occasions correct claims of seeming truth. The second argument is by Michael Huemer. It contends that any denial of an exclusive role for seemings in justification is in a certain way self-defeating.

2. Correct Seeming Reports 2.1. a declaration of semantic and ontological minimalism The topic of interpretation here is some correct uses of the terms seems and appears and their variants. This “correct” means nothing technical. It means that the terms are employed with semantic and conversational propriety to communicate something. The interpretation offers a description of the conditions that occasion this proper use. No more specific sort of interpretive claim is intended. The interpretation is not being alleged to identify a Fregean sense of the “seeming” or “appearing” sentences, or to describe their implicatures, their presuppositions, or their pragmatically conveyed propositions. These things might be true of correct uses. The interpretive proposal takes no stand

54

Seemings and Seeming Reports

on that. Its purpose is to clarify and assess the epistemic contributions of what makes the uses correct. The interpretive aim is to say the least about meaning that is required to get this epistemic job done. No ontology is proposed for correct seeming truth attributions. The interpretation is neutral about whether a correct attribution is made true by a state of affairs of a proposition’s seeming true to a person. Some such state may be constituted by the occasioning mental event and the inclination that make correct some central uses of the “seeming true” terminology. Again, the aim here is epistemic assessment. This does not require determining the ontology.

2.2. the inclination One psychological factor that is present in any correct report by S of a seeming truth is an inclination by S to regard some mental event of S’s as presenting the truth of the content proposition. Philosophers have used various terms to try to describe how the content of a seeming truth is presented. Michael Huemer (2001, 77–79) has called this manner of presentation “forcefulness.” William Tolhurst (1998, 298–99) has said that these seemings “have the feel of truth, the feel of a state whose content reveals how things really are.” Tucker (2010, 530)  prefers “assertiveness” for this aspect of seemings; he also says that it is as though the seemings “recommend” their content or “assure” us of its truth. None of these descriptions is quite right. The event strikes us as somehow showing, displaying, or making manifest the truth of the proposition. Tolhurst’s description comes closest. It goes awry by ascribing to the event what it calls a sort of feeling, “the feel of truth.” The inclination is not an affective state, nor is it any metaphorical sort of “feeling” of a proposition’s truth. That is too direct. Rather, it is an impression of being presented with something pertaining to that truth. The presenting may be something that strikes us as displaying some truthmaking fact. This happens in typical cases involving perceptual experience. For instance, suppose that Smith sees that there is an asp in the apse and Smith asserts that the asp is there. Suppose further that Smith is told that no asp is really there and Smith is asked whether Smith has any good reason to think that an asp is there. It would be natural for Smith to reply, “Well, it does seem to me that an asp is in the apse.” This claim is correct in virtue of Smith’s having a visual experience that strikes Smith as being a visual display of the fact of an asp being in the apse. Non-sensory sorts of mental episodes can prompt correct claims of seeming truth. These include some instances of a priori reflection. In such cases it is not clear what we are discerning that strikes us an appearance of truth. What gives us this impression is something that we metaphorically “see” by having our conceptual perspective on the proposition.

Seeming Evidence

55

Thus, this psychological aspect of our grounds for correctly asserting that a proposition seems true is not a response to some unified mental state type. The relevant sort of basis for the response may be better understood by observing a contrast with otherwise similar cases in which the basis is absent. For instance, suppose that Smith sees no flaw in the premises or the logical structure of a certain argument. But its conclusion remains implausible to Smith. The conclusion seems to Smith to be untrue, in spite of the argument. Smith does not find the truth of the conclusion to be revealed or displayed by the argument. In such cases, the argument gives someone supporting evidence for a proposition without the person having the impression that the truth of the proposition is presented. As a result, these are not cases in which the proposition is rightly said to “seem true” even though, as when a proposition does seem true, someone is aware of the positive bearing of some evidence that the person has for the proposition. Smith could correctly say that the argument “seems to be sound,” and because of this appearance, “it seems that the conclusion must be true.” Nevertheless, while the conclusion remains implausible, doubtful, and not manifest to Smith, it would not be correct for Smith to say that the conclusion seems to him to be true. Whenever a proposition seems true, something must be taken as a sort of presentation or display of the truth to the person. The mental state or event that is the specific occasion for the seeming claim need not incline us to regard it as a display of a truth-making fact. Another sort of occasioning takes place when a proposition occurs to us as though it is remembered. A sense of recalling the content can be an occasion for correctly making a claim of seeming truth. We are not inclined to take this as a display of a fact that makes the content proposition true. Rather, we are inclined to take the sense of recall as something like a presentation to ourselves that this truth has already been somehow settled for us. The truth must be regarded as appearing or being presented in some way. If a factual state of affairs is taken as simply beheld, not presented in some way, then it is not correctly reported as seeming true. This is why, in the last exchange in our opening dialogue, Smith replies to Jones’s request for a reason to think that something seems a certain way by saying “it just does,” rather than that it “seems to seem” that way. This is characteristic of our response to a request for a reason when we are inclined to find a direct awareness of the fact. The same conscious episode can be the basis of a correct seeming report, by becoming regarded as a presentation. Suppose that the dialogue about a snake in the Introduction is supplemented by one more exchange. Jones continues by asking Smith about the seeming visual appearance of an approaching snake: Jones inquires of the fleeing Smith, “Did it really seem that way to you or could you possibly have made a mistake about its seeming like that?” Smith stops at a safe distance, reflects, and replies, “Well, it surely seemed to me that it seemed that way.”

56

Seemings and Seeming Reports

The further reflection by Smith made it correct for Smith to report the seeming as seemingly there. Smith reflected on what Smith previously took as unmediated awareness of a way things seemed to him. As a result of this reflection, the awareness became correctly reportable as a way that this way of seeming presented itself to Smith.

2.3. the occasion for the inclination The mental occurrence that is the basis for the inclination to find the truth displayed is typically one that has a positive evidential bearing on the proposition. It may be a sensory experience as though of observing a truth-making fact, a remembering of such a fact, or an a priori reflection that somehow displays such a fact. In addition, correct claims of the seeming truth of a proposition that involve evidence for it extend beyond cases that include such relatively direct displays of truth-making facts. For instance, it is correct for me to say, “You seem to me to be angry,” when your behavior or demeanor gives me sensory evidence that you are manifesting anger. I need not be inclined to think that I am aware of the fact of your anger itself. Sometimes we lack good reason to believe the content proposition, though. One of the dangers of any view according to which each correct claim of seeming truth reports a prima facie justifying state or event is that the claims can be correctly made on the basis of an erroneous evidential association. That is, the prompting mental event can be mistakenly thought to be good evidence for the truth of the content. It can be correct for me to say, “You seem to me to be depressed,” when I have sensory experience of your manifesting behavior that I associate with depression. Perhaps I see you moving in a way that I take to be characteristically slowed by depression. I need not be justified in taking this appearance to be connected to depression. Suppose that I unreasonably flatter myself by thinking that I apprehend a connection between your pace and your state of mind. Nevertheless, in spite of my lacking good reason for making the connection, it is correct for me to claim that you seem depressed to me, just because your behavior strikes me as “depressed-appearing.” There is a complication. The basis for this correct claim can still be part of some actual evidence that I  have for your being depressed. Suppose that I  recall a history that inductively supports the conclusion that my correctly made seeming truth claims tend to be associated with true contents. If so, then I  have that inductive evidence for your being depressed in this case. I  have this inductive evidence, in spite of this being a case in which I am wrong and unreasonable about a sign of depression. I need not have evidence that I have any such good track record. Even if I do not, in this sort of case I have an occasioning occurrence and an inclination to regard it as displaying the truth. So my claim of seeming truth is correct. Yet what makes it correct does not give me any good evidence for the content proposition.

Seeming Evidence

57

3. Phenomenal Conservatism and Seeming Truth A primary source of the epistemological interest in correct claims of seeming truth is their role in Phenomenal Conservatism (PC). In a few words, PC is the view that seeming truth is an epistemic reason to believe. A principal proponent of PC, Michael Huemer, favors this formulation: PC

If it seems to S that p, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p. (2007, 30)

Again, in the present view there is no relation of seeming truth that supports its content proposition. This view might be thought to be directly contrary to PC. But it is not. PC does not require what is definitely denied here—that a whole state of affairs of p’s seeming to S to be true is always evidence for S of p’s truth. PC might have been contingently true. It might have been that something always included in the seeming was always evidence. It might have been that, whenever PC’s antecedent held, the mental state that any S was inclined to regard as a manifestation of p’s truth was actually some supporting evidence that S had for p. When undefeated, this evidence would justify p for S. If this possibility had obtained, then seeming would be an episode whereby the person would have prima facie justifying evidence. The evidence would have been included in what made the proposition correctly said to seem true. In the present view of what can occasion a correct claim of seeming truth, this is not the way things actually are. It is enough for the correctness of S’s claim that p seems true to S that S takes S to be aware of some manifestation of p’s truth, whether or not the taking is for any good reason. Evidence that supports p is not required and sometimes it is lacking. Nothing else about the correctness of the seeming claim entails that any evidential support is possessed. So PC is not true. It is conceivable that there is an especially strict use of “seeming” locutions. Suppose that for this use to be correct, one must have actual evidence for p as what one is inclined to regard as the manifestation to oneself of p’s truth. On that strict reading of PC’s antecedent condition, PC would be acceptable, though potentially misleading. These “seemings” would always entail possessing prima facie justifying evidence for the proposition. The evidence would not be any whole state of affairs of p’s seeming to S to be true. It would be the mental event that occasions the inclination. We shall return to PC. Huemer defends it with an argument that concludes that opposition to PC is self-defeating. The support Huemer offers for his argument includes support for the proposition that only what he calls “appearances” justify belief. It will be argued that this is not so. Any sort of evidence can justify belief and not all evidence consists in “appearance” in any pre-theoretical sense of the term. Huemer’s self-defeat argument will be discussed in section 7 with special attention to opposing the claim that only appearances justify belief.

58

Seemings and Seeming Reports

4. Opining Some claims of seeming truth are arguably correct on the basis of conditions that are quite lax. It can be correct to say, “It seems to me that thus-and-so,” when all that makes this utterance correct is having the opinion that thus-andso. For instance, “It seems to me that all politicians are crooks,” or “It seems to me that Paris is the best city in the world.” The basis for the correctness in such cases is often just a settled conviction. No inclination to cite anything as support is required. This is a casual and broad use of “seems.” It is distinguishable from a more careful use. At least, its being thus distinguishable is important for the defensibility of PC. If this sort of correct use is one way that the antecedent of PC actually is satisfied, then PC is thereby refuted. No prima facie justification is required for the correctness of these uses. In fact, they are sometimes offered instead of justification. They are used as an expression of undefended belief. For example: “It seems to me that we are going to have a problem with Robinson one of these days. I can’t give you a reason. But I can’t shake the thought that he’s going to make trouble. It seems that way to me.” Two tests for this use: First, another phrase often employed for the same report is “It just seems to me that so-and-so.” When that formulation is equally apt, this broadest use is a very likely interpretation. In contrast, this use is not happily paraphrased as “It appears to me that so-and-so,” and it is still less felicitously paraphrased as “Apparently, so-and-so.” Unlike the most natural understanding of these expressions, it is no part of the opining use to communicate an appearance of truth. The opining use is not the interpretive target here, because it is not a promising interpretation for “seems” in PC. Much more promising is the reading on which “It seems to me that so-and-so” is well paraphrased by “It appears to me that so-and-so” and, more verbosely, “From my perspective it is an apparent truth that so-and-so.” Uses that are well paraphrased in these ways correlate pretty well with having some justification for their propositional contents.

5. The Evidence in the Seeming Views about the justifying capacities of seeming truth differ quite widely. Chris Tucker (2010, 531–34) argues that sensory perceptual experiences do nothing to justify the perceptual beliefs that they bring about. Rather, what does all of the justifying of those beliefs is their seeming truth. This usually accompanies such experiences. At the other extreme, Richard Fumerton (2008, 79) sees no justification at all to be provided for a proposition by its seeming truth. Fumerton thinks that the attitude of p seeming true to S is as justificationally

Seeming Evidence

59

neutral for S, concerning the truth of p, as is any other intentional attitude that S might have toward p, such as S’s hoping for p or S’s fearing that p. Both of these extremes are understandable, given the present view of the circumstances that ordinarily occasion a claim of seeming truth. On the one hand, having an impression of being shown the truth of a proposition can be reasonably thought to support the proposition, whether or not whatever is taken to be the showing really does so. Tucker’s view is thus credible. On the other hand, just having an attitude toward a proposition, with nothing that makes the attitude appropriate, does nothing to justify the proposition no matter what the attitude is. This makes Fumerton’s view credible. If the impression of a showing is occasioned by actual support, then of course there is that support. It is a reasonable empirical hypothesis that such support exists in the typical cases of correct claims of seeming truth, at least among reasonable people. This makes Tucker’s view an understandable generalization. But again, the impression of being shown the truth need not be occasioned by anything that actually supports the proposition. It can be a misimpression. Even when it is a misimpression, it will usually be part of a good inductive argument. Usually people have some awareness of a tendency for these inclinations to have panned out. When this is so, the inclination is part of that sort of inductive support. Still, it will be the inductive case for the content proposition that is evidence for it, not the event that occasions the correct the claim of seeming truth. And the inductive support need not exist. Thus, if p’s seeming to S to be true is a propositional attitude by S toward p, then, as Fumerton holds, having that attitude does not entail that S has even prima facie justification for p. But again, the mental occurrence that occasions the attitude, the event that the person is inclined to regard as a presenting of p’s truth, is usually evidence for that truth. And again, even when it is not evidence, the inclination to take a mental occurrence this way usually has a good track record, and cognizance of that association makes for inductive evidence of p.

6. The Imagination Argument For simplicity, let’s restrict ourselves to visual experiences and visually formed perceptual beliefs. As Chris Tucker uses the term, the sensations that we get by, say, seeing the asp in the apse, are the phenomenal qualities of our visual image of the asp in the apse. To help to illuminate how this notion of a sensation differs from a visual seeming, Tucker (2010) suggests that certain optically impaired people who do remarkably well at some visual tasks—the blind-sighted—may have visual seemings, but not visual sensations; while certain people who are unable to categorize familiar objects, but have no trouble reporting their visual qualities—visual agnosiacs—may have visual sensations without visual seemings (530–31).

60

Seemings and Seeming Reports

Tucker argues for the following views. Sensations do not justify perceptual beliefs. Sensations are no part of the evidence we have for perceptual beliefs. Their contribution to justifying these beliefs is to affect how things seem to us. It is these perceptually caused seemings that do the justifying of perceptual beliefs. This underwrites the correctness of PC (531). Tucker’s argument should be opposed here, since in the present view evidence does all of the epistemic justifying and a correct claim of seeming truth entails nothing that is essentially evidence for its content. The first step in Tucker’s reasoning argues that sensations do not prima facie justify their contents. This first step is what will be disputed here. Tucker argues by generalizing on the basis of a thought experiment. We are to suppose that someone, call her Imogene, has such a powerful imagination that she is able to imagine a scene that is phenomenally just like a full-blown ordinary perception. Let’s suppose that the perception is Imogene’s seeing that the asp is in the apse. Some abbreviations:  let’s call the proposition that the asp is in the apse “ASP,” Imogene’s visual perception of the asp in the apse “Perceptual-ASP,” the phenomenal array of Imogene’s sensations that she receives by seeing that the asp in the apse “Sensory-ASP,” and Imogene’s perfect imaginary rendition of the scene “Imaginary-ASP.” Tucker argues as follows: (P1)

Sensory-ASP and Imaginary-ASP have the same capacity to justify propositions. (P2) Imaginary-ASP cannot prima facie justify believing ASP.

Therefore, (C1) Sensory-ASP cannot prima facie justify believing ASP. (P3) If Sensory-ASP does not prima facie justify believing ASP, then no sensation prima facie justifies its content.

Therefore, (C2) No sensation prima facie justifies its content. (523–24)

It is plausible that P1 is true. Sensory-ASP and Imaginary-ASP do not differ in what are reasonably regarded as their justifying capacities—their qualitative or representational contents. They differ primarily in the mental relations that the subject has to them, sensing versus imagining. We shall see that P2 is objectionable. Tucker thinks it intuitive that P2 is true. And P2 is initially plausible. After all, we should concur that Imogene’s merely imagining the scene does not provide Imogene with any justification for ASP. Conceding that point is easily thought to amount to acknowledging the truth of P2.

Seeming Evidence

61

Tucker offers some help in overcoming resistance to C1. He suggests that the temptation to think Sensory-ASP does some justifying on its own arises by confusion. We are convinced that Imogene gets some justification for ASP from Perceptual-ASP. We think that the Sensory-ASP component of PerceptualASP must do this justifying. But in Tucker’s view the Sensory-ASP component on its own lacks the assertive force of a seeming. In this way it is like merely entertaining ASP, or wishing for ASP. The assertive force of Perceptual-ASP is a phenomenal quality that ordinarily occurs during perceptual experience, but it is not a sensory quality. It is not a constituent of the visual image that is Sensory-Asp. In Tucker’s view, when Imogene is imagining Imaginary-ASP, it does not seem to her that ASP is true, though ordinarily when Imogene undergoes Perceptual-ASP, it does seem to her that ASP is true (532). In contrast, as Tucker understands it, Perceptual-ASP includes a state with the assertive force. That state is a seeming to Imogene that the asp is in the apse. Let’s call this seeming state “Assertive-ASP.” Tucker holds that when Imogene perceptually experiences the asp in the apse, it is Assertive-ASP that does the prima facie justifying of ASP for Imogene. Tucker suggests that once we disentangle Assertive-ASP from Sensory-ASP in Perceptual-ASP, it becomes intuitively unproblematic to accept C1 (532–33). It can be agreed that perfectly imagining a visual scene does not prima facie justify believing that the scene is present and that perceptually experiencing the same scene prima facie justifies. We shall see that there is a better candidate than Assertive-ASP for what does the justifying. Imogene’s stipulated power to imagine is drastically different from ours. It may well be that even our best and most vivid visual imagining does not present to us the sensory qualities of any part of any perceptual experience that we have had. What is visually imagined is not nearly as rich and immediate. That difference does not make Imogene’s imaging ability inconceivable. But it requires us to proceed with caution in thinking about how things consciously are for Imogene as she exercises her perfect imagination. When we imagine, we are conscious of our pallid imitations of sensory images and we are conscious of producing and altering them. Though the latter awareness has an obscure non-sensory phenomenology, we are conscious in some way of our devising and guiding the imagining. In order to be sure that we are thinking about Imogene imagining when we consider Imaginary-ASP, we must include that she has some such awareness of her imagining. Without that, the perception-like image that is presented to Imogene is not consciously different from her hallucinating a perfect sensory replication of the scene. If instead we think of her experience as being consciously just like hallucinating the scene with perfect sensory replication, though, then it would display to her exactly the look of perceiving that the asp is in the apse, while she has no awareness of the experience having any

62

Seemings and Seeming Reports

internal origin. Considering this sort of case, we lose any sense that Imogene is not receiving prima facie justification for ASP. This is justification that would usually be at least somewhat defeated by its incongruity with whatever else she would usually be justified in thinking about her environment. But it is quite plausible that the experience would give her reason to believe ASP. So the conscious character of imagining must be part of the experience she has in producing Imaginary-ASP, if we are to get the impression that this experience does not justify believing ASP. Thus, to be as clear as possible about what is doing the justifying, we should consider the imagining that has as its content Imaginary-ASP to be a complex conscious state, composed of both the output rendition of the scene, Imaginary-ASP, and some awareness of her imagining. This opens the way to taking the latter awareness to defeat the prima facie justification from the former sensory state. The net epistemic effect of the imagining is a defeated prima facie justification for ASP. Thus, it is an open alternative view that, without the defeat, Imaginary-ASP does prima facie justify ASP. Given that this alternative is open, it is attractive. After all, the nearest counterpart experience is a perfect hallucinatory one without awareness of imagining. It is quite plausible that this experience would prima facie justify believing ASP. Thus, the intuitive difference between Imogene’s lacking prima facie justification for ASP as she imagines Imaginary-ASP, and having prima facie justification for ASP as she hallucinates it, goes with the presence and the absence of her awareness of imagining. In order to find a prima facie justifying experience, we have only to subtract awareness of the producing and guiding of imagining. We have no need to add any forcefulness, assurance, or assertoric character to it. Assertive-ASP is not required. The phenomenology here is difficult to apprehend clearly and distinctly. In the end though, it is most reasonable to think there is no assertive phenomenal quality in ordinary perceptual experience. If we seek such a quality by introspection, we do not find it. We are familiar with the relative forcefulness of some perceptions. They include perceiving bright lights, loud sounds, and strong pressures. This is vigorousness in the degree of particular sensory qualities. It is not any sort of assertiveness or an assurance that pervades the whole of every ordinary perceptual experience, however dim and feeble the qualities are. The experiences do not generally press themselves forward; they just do consciously occur, sometimes quite subtly. Setting aside the contingent and variable forcefulness of some perceptual experiences, no conscious assertive quality is presented to our introspective investigation. If this is right, then an ordinary sensory display can prima facie justify the corresponding external world proposition without being accompanied by any special conscious force or assertion. It is most reasonable to think that the justifying aspect, missing from ordinary imaginings, is provided by some

Seeming Evidence

63

combination of the richness and immediacy of its sensory character and its characteristically involuntary presentation and evolution over time. The characteristic fullness of the sensory character in perception is, by hypothesis, present in Imogene’s Imaginary-ASP. What is still consciously different about Imogene’s imagining Imaginary-ASP is that its contents are voluntarily created and configured. That difference might well be what grounds any mistaken inclination to judge that Imaginary-ASP does not prima facie justify ASP. This basis for the justificatory contrast suggests that Imogene does get prima facie justification for ASP from her experience of Imaginary-ASP when she imagines it. That prima facie justification is defeated by her awareness of its imaginary origin. Tucker takes up this hypothesis in an endnote. He reports that when he considers just the imagining of Imaginary-ASP and he is careful about the difference between sensations and seemings, his intuition is that the imagined Imaginary-ASP does not even prima facie justify (2010, 544). This apparent intuition is a mistake. Perhaps the mistake is based on projecting into Imogene’s Imaginary-ASP some vestige of ordinary imagining, some slight absence of the sensory detail of perceptual experience. Prima facie justification can be delicate and readily defeated. It is helpful to consider the following question. When Imogene imagines Imaginary-ASP, does she have any evidence at all for ASP, defeated or otherwise, that she lacks when she does not have any visual experience? The answer is clear. She has the following evidence when she imagines Imaginary-ASP: it looks to her exactly as though the asp is in the apse. She has no such evidence when she has no visual experience. Further support for this view emerges when we consider what would happen if Imogene were quite convincingly told the following falsehoods: She is not imagining Imaginary-ASP. Her effort to imagine is actually causally ineffective. Instead, she perceives the asp in the apse. She has been fooled by an elaborate hoax, involving the actual visual display to her of things that she was instructed to imagine in the specific details that were then perceptually shown to her. This hoax has her reasonably but mistakenly believing that she has the power of perfect visual imagination. It is clear that in this version of the case, despite the deception, Imogene would have some visual evidence for ASP. Her awareness of her intentional imagining would be defeated by the credible testimony about the hoax. Under the circumstances, what remains to serve as the visual evidence that she has for ASP is the conscious presentation to her of Imaginary-ASP. This evidential status of Imaginary-ASP goes against P2.

7. The Self-Defeat Argument Justified belief in a proposition–“doxastic justification”–requires two things. The person must have justification for believing the proposition–“propositional

64

Seemings and Seeming Reports

justification”–and the person must employ a proper basis for believing the proposition. Propositional justification enters into a proper basis for doxastic justification, and epistemically insignificant factors such as the consolation of the belief do not. Proper basing is accomplished when someone uses in a rational way the propositionally justifying evidence that the person has for the truth of a proposition to believe the conclusion that the proposition is true. (See John Turri [2010] for several citations presenting standard accounts of doxastic justification and also for examples arguing for problems with those accounts.) Huemer’s self-defeat argument goes as follows. The first premise is an empirical claim about our actual method of belief formation, whenever the believing stands some chance of being doxastically justified rather than being clearly irrational, wishfully thought, or the like. Let’s call the beliefs that stand some chance of being justified “potentially doxastically justified.” Here is the first premise. (P1)

Any potentially doxastically justified belief, including belief in the negation of PC, is formed on the basis of the way things seem to the one who believes it—the basis is, in a word, “appearances.”

The first step taken in Huemer’s reasoning is an inference from P1 and PC. P1 says that whenever some actual belief is potentially doxastically justified, the belief is based on appearances. Doxastic justification requires some proper basis and propositional justification is the only clear candidate. Roughly speaking, PC says that appearances are what propositionally justify. From P1 and PC, it is inferred that unless PC is true, the actual basis of our belief in the negation of PC is not a justifying basis. (C1)

If PC is untrue, then appearances are not a justifying basis for believing the negation of PC.

The other premise in Huemer’s self-defeat argument is a safe assumption about doxastically justified belief. (P2) A belief is unjustified if it is not formed on a justifying basis.

The ultimate conclusion, from C1, P1, and P2, gives the intended sense in which “opposition to PC is self-defeating.” (C2) If PC is not true, then belief in its negation is unjustified. (2007, 39–41)

One fault in this reasoning is that C1 does not follow from P1 and PC. Appearances do not have to justify in the particular way that PC asserts in order for appearances to have the role in the basing of beliefs that P1 asserts for them. For example, suppose that I  have no idea where Jones was born. I  ask Smith, who I know to be Jones’s friend and my friend too. Smith tells me

Seeming Evidence

65

that Jones was born in Cleveland. Smith does not give any sign of untrustworthiness. Hearing that sort of assertion from my friend justifies my belief in its content. Still, when I  have just been told this, the proposition that Jones was born in Cleveland does not seem true to me. The proposition does not seem untrue either. At most, it seems to be just one possibility among many. I have a justified belief because my believing is based on the proposition being asserted by my friend Smith, who is giving plainly trustworthy testimony. It may be that appearances are involved in this justification. For instance, it may be that the justification that I have for believing that Jones was born in Cleveland includes evidence that I get from its appearing to me that Smith is sober and sincere. Any such role for appearances is enough for P1 to have no trouble with this sort of example. My basis for belief at least includes appearances, and that is close enough to what P1 says about a justifying basis. But appearances in this example are not playing the more specific role that PC asserts for them. The sufficient condition that PC states for having prima facie justification for a proposition requires the proposition itself to appear to be true. That does not happen in this case. Some cases of justified belief by inference also include being aware of evidence for a proposition without the proposition itself seeming true. For instance, suppose that you learn from a clearly trustworthy source that either Smith or Jones stole a copy of a certain book from the bookstore. Your prior knowledge of the two of them gives you some reason to deny that either would do this. You also know that Smith in particular could not have stolen the book, since you saw Smith pay for Smith’s copy. You infer that Jones did the stealing. At that point, Jones does not seem to you to have done it. That thought seems dubious, unlikely, rather than true. But you have a strong and undefeated argument as justifying evidence for the belief. In such cases, your evidence supports the conclusion. Belief in the conclusion on the basis of the argument is doxastically justified. Thus, even if P1 and PC are true, appearances in the role that PC assigns for them do not have to be included in the basis for a justified belief. C1 follows from P1 and PC only if appearances must play the PC role. At least, C1 follows only if appearances play this role concerning the proposition that is the negation of PC. Nothing in P1 and PC gives any reason to think that this proposition in particular must be justified by appearances in their PC role. So C1 does not follow. It might be thought that the spirit of the self-defeat argument remains unimpeached. The problem cases do not include the apparent truth of the justified propositions themselves. But they do include the appearance of the reliability of a testifier and of the cogency of an argument. Perhaps appearances always figure essentially somewhere in the justification. If this view were right, then we would always be doxastically justified ultimately on the basis of appearances, though not quite as PC asserts.

66

Seemings and Seeming Reports

Appearances need not play any justifying role. Even when the evidence is communicated by correct appearance claims, the correct claims are merely occasioned by what really justifies. Here is a rival evidentialist sufficient condition for doxastic justification: EJB

If S’s belief in X is based on propositionally justifying evidence S has supporting X, then, in the absence of defeaters, S thereby has at least some degree of justification for believing X.

What follows is an example that argues against any need for appearances in justification. EJB accounts for the justification in the example. Seemless has an ordinary range of perceptual beliefs. But Seemless has been extraordinarily influenced by some bad grounds for external world skepticism. The grounds consist in some empty rhetoric. Seemless has been persuaded by the rhetoric of the truth of a sweeping skepticism. This persuasion has effected an exceptional psychological transformation. Seemless no longer makes one type of normal response to perception. When Seemless sees the asp in the apse, he experiences the same visual sensory presentation as an ordinary human viewer of the scene, one who thereby knows that the asp is in the apse. Also quite ordinary are Seemless’s understanding of the external world concepts that are involved and his appreciation of how different perceptual experiences are usually taken to bear on different external world propositions. But Seemless is thoroughly sold on the idea that external world experiences are justificationally irrelevant to external world propositions. The extraordinary result is that Seemless is not at all inclined to regard his visual experience of the asp in the apse as anything like a display of the asp being in the apse. The crucial consequence of that missing inclination is this: it not correct to say that it seems to Seemless that he sees the asp in the apse. Again, Seemless retains the ordinary tendency to form perceptual beliefs on the basis of ordinary visual experiences, such as those that Seemless has in seeing the asp in the apse. This belief formation frustrates Seemless as a failure to live up to his skepticism. But he does it. If it is difficult to take seriously the possibility that Seemless’s skepticism could have this peculiar psychological effect, then we should suppose instead that the missing element has some purely physiological malfunction as its cause. The force of the example does not depend on the present view of correct claims of seeming truth. Substantially the same example refutes any need for a seeming truth in perceptual justification, whether or not claims of seeming truth depend on having the specified sort of inclination with regard to the experience. The claims might depend instead on the experience including an assertoric phenomenal character, or its including a feel of truth, or something else. The example does its work as long as a correct seeming claim in a case like this requires anything in addition to the visual experience. Whatever it is, we should suppose that this requirement for correct claims

Seeming Evidence

67

of seeming truth is what Seemless’s skeptical view has removed from his perceptual psychology. Nevertheless, Seemless’s visual experiences give him justifying evidence for his visual beliefs. Since he embraces his skepticism on bad grounds that do not defeat the experiential evidence, the justifying effect of the evidence is undefeated. Seemless’s perceptual evidence justifies his perceptual beliefs. Yet seeming truth does none of the justifying, since it does not occur. If this is right, then the spirit of the self-defeat argument is refuted. Seeming truth need not make any contribution at all in order for us to have doxastic justification. Reflection on the Seemless example can also give reason to think that, even when seeming truth is present, it is not the seeming that does the justifying. Absent defeaters, it is sufficient for the person to believe based on the evidence that prompts the inclination. EJB is in the clear here. Seemless does have undefeated perceptual evidence. By EJB, that is an adequate basis for having some doxastic justification.

8. Conclusion Correct claims reporting that something seems true are efficient ways to convey the possession of otherwise unspecified evidence, including evidence from perception, memory, and introspection. The claims are correct in virtue of one’s having an inclination that these sorts of evidence can prompt, and one’s undergoing a conscious episode that one has the inclination about. It is roughly an inclination to regard that episode as a presentation of the truth of the content that is said to seem true. This efficient versatility of claims of apparent truth makes them extremely handy, and indeed almost irreplaceable. (The present work makes no assertion to the effect that things seemingly are some way or other and no assertion to the effect that something is apparently true. Omitting all such claims has been difficult!) Though the great utility of such assertions at conveying evidence possession must be granted, the inclinations involved and its occasioning episode need not give evidence. They can exist without having any good evidence for the content of the corresponding claim of seeming truth. This happens when we are inclined to take something to present a truth that does not actually indicate that truth, although we are inclined to think that it does because that pleases us, consoles us, or the like. Also, seeming truth is not the best account of the evidence that we do have. We can have at least equally good evidence without its being reportable as a correct claim of seeming truth, as when testimony or a powerful argument gives us good reason to believe a proposition that does not seem to us to be true. In fact, we can have at least equally good evidence when nothing relevant seems true, as the Seemless example illustrates.

68

Seemings and Seeming Reports

Acknowledgments I appreciate helpful comments on previous drafts from Kevin McCain and Chris Tucker.

References Fumerton, Richard. 2008. “Epistemic Conservatism:  Theft or Honest Toil?” In Tamar Szabó Gendler and John Hawthorne, eds., Oxford Studies in Epistemology, 63–86. Oxford: Oxford University Press. Huemer, Michael. 2001. Skepticism and the Veil of Perception. Lanham, MD:  Rowman and Littlefield. Huemer, Michael. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. Tolhurst, William. 1998. “Seemings.” American Philosophical Quarterly 35: 293–302. Tucker, Chris. 2010. “Why Open-minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45. Turri, John. 2010. “On the Relationship between Propositional and Doxastic Justification.” Philosophy and Phenomenological Research 80(1): 312–26.

{ part ii }

Foundations of Dogmatism

This page intentionally left blank

{4}

Immediate Justification, Perception, and Intuition Jessica Brown 1. Introduction A number of philosophers have recently defended the claim that one has immediate justification to believe the contents of intuitions. Many of these philosophers take intuitions to be seemings or experiences of a certain kind. Further, they suggest that the kinds of consideration which motivate the idea that we have immediate justification to believe the contents of perceptual experiences also motivate the view that we have immediate justification to believe the contents of intuitions (see Huemer 2007; Chudnoff 2011). By contrast, I  defend the claim that one has immediate justification to believe the contents of intuitions while separating this thesis from the claim that intuitions are seemings. Further, I argue that certain standard motivations offered for the claim of immediacy in the case of perception do not work in the case of intuition. I start in the next section by showing how the notion of immediate justification can be applied to intuitions even on the view that they are not experiences, but rather beliefs or inclinations to believe. Thus the thesis that one has immediate justification to believe the contents of intuitions is not hostage to the controversial view that intuitions are seemings or experiences. In sections 3 and 4, I  reject several potential common motivations for the immediacy thesis in the case of intuition and perception. In section 3, I set aside the regress argument. I then turn in section 4 to consider a motivation which Pryor offers to defend the view that we have immediate justification to believe the contents of perceptual experiences, namely, the aim of avoiding skepticism about knowledge of the external world. However, I argue that no analogous argument motivates the claim that one has immediate justification to believe the contents of intuitions. Instead, in section 5, I argue that the best motivation for the view that one has immediate justification to believe the contents of intuitions is that it is a natural and plausible view about certain

72

Foundations of Dogmatism

cases, such as the Gettier thought experiment, and can overcome standard objections.

2. Immediate Justification The question of whether there is immediate justification can be formulated either doxastically or propositionally. I will start with a propositional formulation although we will see later that, on certain views about the nature of intuition, the thesis of immediacy is more naturally formulated in doxastic terms. Formulated propositionally, the thesis of immediacy is the claim that some justification to believe is immediate (e.g., Pryor 2000).1 One can have a propositional justification, or justification to believe a proposition, even if one lacks a justified belief in it. This could either be because one doesn’t believe it, or one does but one fails to base that belief on the good reasons one has to believe it. By contrast, one has doxastic justification for a proposition when one has a justified belief in that proposition. We can explain the relevant notion of immediate justification by contrasting it with the notion of mediate justification. Suppose that the only kind of justification a detective has to believe that Burglar Bill committed a burglary rests on her justification to believe various propositions about the manner and timing of the burglary and the preferred manner of operation of local burglars. In such a case, were the detective to lose justification to believe the latter propositions concerning the time and manner of the burglary and preferred manner of operation, then she would lose justification to believe that Burglar Bill committed the burglary. Thus, her justification to believe that Burglar Bill committed the burglary is mediate since this justification rests on, or depends on, her justification to believe some other proposition. By contrast, one’s justification to believe a proposition, p, is immediate if it does not depend on, or rest on, one’s justification to believe any other proposition (Pryor 2000). Immediacy is a thesis about the architecture of justification, not its strength. Contemporary defenders of immediacy deny that immediate justification entails infallibility or incorrigibility (e.g., Pryor 2000). In particular, they hold that immediate justification is defeasible. This is how I will understand immediate justification throughout the chapter. Further, the immediacy thesis merely claims that some justification is immediate without specifying further the circumstances in which one has immediate justification. One popular suggestion is that perceptual experience can provide immediate justification.

1 Compare Goldman (2008) who offers a doxastic formulation of the thesis of immediate justification as the claim that “some beliefs or propositions are made (prima facie) justified in virtue of states of affairs, or processes, etc that confer justification without themselves being justified” (p. 64).

Immediate Justification

73

For instance, Pryor (2000) defends the view that when one has a perceptual experience with the content that p, one has immediate justification to believe that p.2 The claim that some perceptual justification is immediate is compatible with a range of different views about (1) whether one has immediate justification to believe the content of all or only some perceptual experiences; and (2) which features of perceptual experience make it the case that one has immediate justification (compare, for example, Pollock 1986, Brewer 1999, Dretske 2000, Pryor 2000, Huemer 2001, Peacocke 2004, Goldman 2008, and Chudnoff 2011). Many of the key defenders of immediacy in the case of intuition take it that intuitions are seemings and use the perceptual case to model the claim that we have immediate justification for the contents of intuitions. For instance, Huemer (2007) and Chudnoff (2011) defend an immediacy thesis for intuition understood to be a certain kind of seeming: Immediacy (intuition = seeming): In some cases, having an intellectual seeming with the content that p provides immediate justification to believe that p.

Although the idea that intuitions are seemings or experiences is popular among those who defend the idea that intuition provides immediate justification, it is a controversial view about the nature of intuitions. So, it is useful to see how the thesis of immediacy could be applied to a wide range of other views about the nature of intuitions. According to one minimal account, intuitions are beliefs, or perhaps tendencies to believe, without any further restriction (e.g., Williamson 2007). If intuition is understood as belief, then the thesis of immediacy for intuition is most easily formulated doxastically as follows: Immediacy (intuition = belief): Some beliefs, including some paradigm philosophical intuitions, are immediately justified.

An alternative propositional formulation would claim that some justification to believe is immediate, including justification to believe the contents of some paradigm philosophical intuitions. Whereas Williamson argues that the notion of intuition is no more restricted than the notion of belief, a number of other philosophers have argued that

2 Compare Pryor’s definition of dogmatism on which perception provides immediate justification to believe: “[t]he Dogmatist about perceptual justification says that when it perceptually seems to you as if p is the case, you have a kind of justification for believing p which does not presuppose or rest on your justification for anything else, which could be cited in an argument (even an ampliative argument) for p” (Pryor 2000, 519; see also 520, 532). Although Pryor defines “dogmatism” as the view that perceptual experience provides immediate justification to believe its content, I think the view is sometimes associated with the more particular view that it is in virtue of the phenomenal character of perceptual experiences that they provide immediate justification. So, I will avoid the expression “dogmatism.”

74

Foundations of Dogmatism

intuitions are best regarded as a sub-species of belief. It is common to impose restrictions of content (e.g., the relevant proposition is abstract or modal) and aetiology (e.g., the relevant belief is not based on perception, inference, introspection, or mere memory), as is done by Sosa (1998, 2006) and Ludwig (2007). One might worry whether the definition of intuitions as beliefs not based on inference begs the question at hand of whether those beliefs are immediately justified. However, any such worry is ill founded. The psychological claim that intuitions are not based on inference does not entail the claim that they are immediately justified. Even if one acquires a belief non-inferentially, say, as a result of wishful thinking, it does not follow that it is immediately justified. Indeed, it would surely be unjustified. Having noted how defining intuitions partly in terms of the aetiological condition that they not be based on inference does not beg the question of immediacy, let us examine how the thesis of immediacy could be applied to a definition according to which intuitions are a restricted kind of belief. Most of those who suggest that intuitions are a restricted kind of belief define intuitions not only negatively in terms of the absence of certain kinds of aetiology, but also positively in terms of possessing a certain kind of aetiology. For instance, some suggest that an intuition is a certain kind of belief, or attraction to believe, formed by a certain kind of competence (e.g., Ludwig 2007; Sosa 2007). Once intuition is understood as a restricted kind of belief, one could state the immediacy thesis as follows: Immediacy (intuition = restricted belief): (Sometimes) when S has the intuition that p, the belief which is S’s intuition is immediately justified.

We have now seen how one could apply the thesis of immediacy to intuitions on all of the main approaches to intuitions on which they are variously treated as experiential states, beliefs (or dispositions to believe) without restriction, or beliefs (or dispositions to believe) with some restriction. Thus, we have seen that the idea that we have immediate justification for the contents of intuitions is not hostage to the controversial view that intuitions are seemings or experiences. Just as in the case of perception, the immediacy thesis in the case of intuition is compatible with different views about what feature of intuitions it is in virtue of which one has immediate justification,3 and the range of intuitions in which one has immediate justification. In my defense of the thesis of immediacy as applied to intuition I remain neutral on these issues.

3 For instance, suggested views include that one has immediate justification to believe the contents of intuitions in virtue of their phenomenal character, that they are reliable indicators of the truth, or that they are the product of a competence (compare Sosa 1998, 2006; Huemer 2007, Ludwig 2007, Chudnoff 2011).

Immediate Justification

75

Some contemporary defenders of immediacy in the case of intuition don’t provide any detailed defense of the view but seem to think that the case is already established. For instance, Chudnoff (2011) assumes that Pryor and Huemer have provided good reason to suppose that we have immediate justification for the contents of intuitions. His main contribution is to try to provide an explanation of how this is true, assuming that it is. Sosa (2007) similarly concentrates on providing an account of how we have immediate justification in the case of intuition. However, I want to focus on what might motivate the thesis of immediacy in the case of intuition. Some have suggested that the same considerations motivate the immediacy thesis in the case of perception and intuition. For instance, Huemer’s argument for phenomenal conservatism would, if successful, motivate immediacy in the case of both perception and intuition. However, like others, I find his defense of phenomenal conservatism unpersuasive (see DePaul 2009). Instead, I consider two other standard ways to motivate immediacy in the case of perception and examine whether they apply to the case of intuition. After rejecting these potential common motivations, I argue that immediacy in the case of intuition is best motivated by the fact that it is a natural view of certain cases, such as Gettier’s thought experiments, and can be defended against standard objections (section 5).

3. The Regress Argument Suppose that you have justification to believe one proposition, p, in part in virtue of your justification to believe a second proposition, q. We can then ask the question in virtue of what you have justification to believe the second proposition, q. Perhaps, this justification consists in part in your having justification to believe some third proposition, r. If so, then we can simply raise the question again: in virtue of what do you have justification to believe the proposition, r? There are three main options in response to the repeated pressing of this question: (i) an infinite regress; (ii) a circle; or (iii) immediate justification. A traditional way of arguing that some justification is immediate is by arguing that the other possible options are untenable. The strength of such an argument for immediate justification depends on the strength of the considerations mounted against the rival options of an infinite regress or a circle. Importantly, the strength of the rival options may differ between perception and intuition. For instance, consider a coherence response to the regress argument. Although coherence is not popular as an account of justification for claims about the external world, it is more popular in certain philosophical domains, such as ethics.4 In addition, some of the key arguments against

4 See Daniels (2003), which surveys and defends the idea that reflective equilibrium provides a coherence account of justification in inductive and deductive logic, and ethics and political philosophy.

76

Foundations of Dogmatism

coherence theories of justification exploit the idea that justification is a function not only of coherent relations among beliefs, but also the relationship between experience and belief,5 e.g., Sosa (1980); see also Feldman’s (2003) strange case of magic Feldman, and Plantinga’s (1993) case of the epistemically inflexible climber. But such arguments are inapplicable to the case of intuition if intuitions are not experiences. (Notice that this may leave in play other arguments against coherence theories which do not depend on the relation between a subject’s belief system and her experiences but rather the relation of her belief system to the facts.) A more fundamental concern about the regress argument is that, even if successful, it is simply an existence proof for immediate justification. It doesn’t by itself show that either perceptual experience or intuition provides immediate justification. It is useful to illustrate this in the case of perceptual experiences. The regress argument by itself does not adjudicate in favor of the popular view that one has immediate justification for propositions about the external world, such as here is a hand, and against a more traditional form of foundationalism according to which one has immediate justification for propositions concerning one’s own mental states, but lacks immediate justification for propositions about the external world. Of course, there are the numerous objections to the traditional foundationalism outlined above. But, these objections are in addition to the regress argument which, in itself, is simply an existence proof for immediate justification. Given that the regress argument by itself fails to motivate the claim that one has immediate justification to believe the contents of intuition, I turn in the next section to consider another possible motivation for immediacy given by Pryor.

4. Pryor and Skepticism Pryor (2000) argues that the idea that experience provides immediate justification has the advantage of providing a reply to skepticism about the external world. Similarly, one might wonder if the immediacy thesis in the case of intuition could be analogously motivated. However, I will argue that considerations of skepticism in fact fail to motivate the immediacy thesis in either case. For, as we will see, immediacy is a much stronger claim than is needed to block the skeptical argument.

5 Sosa (1980, section 9) asks us to take a belief at the periphery of the total belief system, say, the belief that this table is rectangular. Such a belief will have rather few connections to other beliefs in the total system. Leaving experience unchanged, substitute that belief for another (say, that the table is square), making the few adjustments in belief necessary to maintain the same level of coherence with other beliefs. He argues that the substituted belief is not as justified as the original, even if it coheres just as well with the overall belief set, for it is at odds with experience.

Immediate Justification

77

Pryor argues that the notion of immediate justification enables us to answer what he takes to be the most powerful way of formulating the skeptic’s argument.6 On Pryor’s formulation, the skeptic’s argument starts from the premise that the only way to have justification to believe that one is not being deceived by an evil demon is for that justification to rest in part on perceptual justification one has for believing things about the external world. It is then argued that if one is to have justification to believe anything about the external world on the basis of one’s perceptual experiences one must have antecedent justification to believe one’s not being deceived by an evil demon. Combining these claims, it follows that one cannot have justification to believe that one is not being deceived by an evil demon. The argument concludes that one cannot have justification to believe anything about the external world on the basis of one’s perceptual experiences. Pryor recommends replying to this argument by rejecting the second premise, that if one is to have justification to believe anything about the external world on the basis of one’s perceptual experiences one must have antecedent justification to believe one’s not being deceived by an evil demon.7 He correctly observes that one can reject the second premise by endorsing the claim that some perceptual experiences provide immediate justification to believe their content. However, the latter claim is stronger than is required to reject the second premise. One could claim that one can have justification to believe a proposition about the external world on the basis of perceptual experience without having antecedent justification to believe one’s not being deceived by an evil demon, even without supposing that perceptual experience provides immediate justification. For instance, one line of resistance to skepticism treats perceptual beliefs about the external world as inferentially justified while denying that one has justification to believe propositions about the external world only if one has antecedent justification for the negation of the relevant skeptical hypothesis. For example, Vogel (2008) suggests that one’s beliefs that here is a hand and that one is not a BIV acquire justification simultaneously by inference to the best explanation: they are part of a wider set of propositions which is the best explanation of one’s perceptual experiences. Of course, there are well-known objections to any explanationist approach to our perceptual knowledge of the external world. My point here is not to defend the explanationist approach, but rather to point out that consideration of skepticism 6 His formulation is designed to capture skepticism about both knowledge of, and justification to believe, propositions about the external world. Further, it is designed to capture the skeptical power of both the evil demon hypothesis and the dreaming hypothesis, where these are traditionally understood so that the latter but not the former is compatible with the truth of one’s beliefs. 7 He formulates the second premise as follows: “If you’re to have justification for believing that p on the basis of certain experiences or grounds E, then for every q which is “bad” relative to E and p, you have to have antecedent justification for believing that not-q—justification which does not rest on or presuppose any E-based justification you may have for believing that p.”

78

Foundations of Dogmatism

most obviously supports the minimal claim that one has perceptual justification to believe propositions about the external world without having antecedent justification to believe the negation of the relevant skeptical hypothesis. To move to the stronger thesis that one has immediate perceptual justification to believe propositions about the external world, one needs to argue, as Pryor does, that the thesis of immediacy in the case of perception is the best implementation of the more minimal claim. This lesson is particularly important when we consider the possibility of immediacy as a thesis about intuitive justification. A parallel skeptical argument in the case of intuition would show at most that we have intuitive justification to believe the contents of intuitions without needing antecedent justification against relevant skeptical hypotheses, such as that whenever it intuitively seems to me that p, an evil demon makes it seem to me as if p although in fact, not p. The crucial question is whether the best way to implement this insight is one on which one has immediate justification to believe the contents of intuitions, or rather mediate justification to believe their contents. In the next section, I try to answer this question by defending the view that we do have immediate justification to believe the contents of certain intuitions.

5. Appeal to Cases Defenders of immediacy claim that it is a natural view about certain kinds of examples. Pryor (2005) defends the immediacy thesis in the case of perception in this way. Chudnoff (2011) claims that it is natural to suppose that we have immediate justification for certain mathematical propositions. Analogously, I will argue that it is natural to suppose that we have immediate justification for judgments about certain philosophically important thought experiments, such as the Gettier case. In order to set up the positive argument, we need to look in a little more detail at Gettier’s thought experiment and how it refutes the JTB theory of knowledge. This is especially important to the project of defending immediate justification for intuition since some leading philosophers take thought experiments to support their conclusions inferentially, which may seem to undermine the thesis of immediacy (e.g., Williamson 2007). Gettier (1963) is standardly taken to have refuted the JTB theory of knowledge according to which, necessarily, if a subject has a justified true belief that p, then her belief that p constitutes knowledge that p.  He did so by asking his readers to consider certain imaginary cases that have been widely taken to establish a certain possibility claim, namely, that it is possible for a subject to have a justified true belief that is not knowledge. Initially, one might take this possibility claim to be a candidate for immediate justification. However, several authors have recently suggested that our justification for this possibility claim is inferential (e.g., Williamson 2007, Ichikawa and Jarvis 2009). On their view, when Gettier presents certain fictional examples, such as the Ten Coins Case,

Immediate Justification

79

we realize that it is possible that someone could stand in the relation described in this case. (In the Ten Coins Case, Smith infers the truth that the person who will get the job has ten coins in his pocket from the justified true belief that Jones has ten coins in his pocket and the justified false belief that Jones will get the job.) Furthermore, we realize that a certain generalization holds of anyone who stands in that relation, namely, that they have a justified true belief that is not knowledge. Combining these pieces of knowledge, we realize that it is possible that someone has a justified true belief that is not knowledge. More fully, we can represent our justification to believe the latter possibility claim as arising from the following inference concerning the Ten Coins Case: (1)

Possibly, someone stands in the relation described in the Ten Coins Case. (2) Necessarily, anyone standing in the relation8 described in the Ten Coins Case has a justified true belief that is not knowledge.

Therefore, (C)

Possibly, someone has a justified true belief that is not knowledge. (From 1 and 2.)

Of course, we could go through a parallel inference using other examples of Gettier cases, such as Gettier’s own Brown in Barcelona case.9 Some have queried the suggestion that we have only inferential justification for (C) and have instead suggested that we have immediate justification for this claim (e.g., Martin 2009; Malmgren 2011). Rather than settle this issue here, I will argue that we have immediate justification for the generalization which is the second premise ((2): anyone who stands in the relation described in the Ten Coins Case has a justified true belief that is not knowledge). This leaves it open that we have immediate justification for other judgments. Before presenting the argument that we have immediate justification for (2), it is worth further clarifying just what (2) amounts to. In particular, notice that if the generalization at premise 2 is not to be completely trivial, it cannot be written into the details of the Ten Coins Case that it involves someone who

8 There is some controversy about the best formulation of the generalisation (2) (Williamson 2007). In my view, the best formalization suggested so far understands the expression “the relation described in the Ten Coins Case” to refer to the relation described in the story expressed by the text of the Ten Coins Case, not merely what is explicitly stated in the text itself (Ichikawa and Jarvis 2009). In general, a proposition may be true in a fiction even if it is not explicitly contained in the text. For instance, it is part of the fiction of Shakespeare’s Hamlet that he is not telepathic, even though this is nowhere stated in the text. I understand (2) in this way in the rest of the chapter. 9 In this case, Smith has a justified belief that Jones owns a Ford and thereby infers the disjunction (Jones owns a Ford or Brown is in Barcelona). He has no evidence whatsoever for the second disjunct. However, the disjunction is true since, unbeknownst to him, Jones does not own a Ford but Brown is in Barcelona.

80

Foundations of Dogmatism

has a justified true belief that is not knowledge. Rather, the Ten Coins Case must be understood more minimally so that it contains enough information to enable us to realize that the subject as described has a justified true belief that is not knowledge, without this being explicitly part of the case. For instance, following Gettier, we could understand the Ten Coins Case as follows: Suppose that Smith and Jones have applied for a certain job, and that Smith has strong evidence for the following conjunctive proposition: (a) Jones is the man who will get the job, and Jones has ten coins in his pocket. Smith’s evidence for this proposition might be that the president of the company assured him that Jones would in the end be selected, and that he, Smith, had counted the coins in Jones’s pocket ten minutes ago.

Proposition (a) entails: (b)

The man who will get the job has ten coins in his pocket. Suppose that Smith sees the entailment from (a) to (b) and accepts (b) on the grounds of (a) for which he has strong evidence. Imagine, further, that unknown to Smith, he himself, not Jones, will get the job, and that he himself has ten coins in his pocket. Proposition (b) is then true, although proposition (a), from which Smith inferred it, is false. (1963, 122)

If we take this description to be the ten coins story, then we can understand how the generalization (2) is not trivial. It claims that anyone who stands in the relation described in the story has a justified true belief that is not knowledge. I argue that we have immediate justification for (2) by arguing against the alternative view that one’s justification for (2) is mediate, i.e., that one’s justification to believe (2) rests in part on one’s justification to believe some other proposition.10 I  start by arguing against one possible version of the mediate view, which I call the “template view,” and later argue against the other possible implementation of the mediate view, which I call the “coherence account.” Suppose that, prior to reading Gettier’s article, one already had justification to believe that true beliefs fitting a certain template do not constitute knowledge, where by “template,” I mean a certain specified set of features. Thus, one had prior justification to believe what I will call “the template claim”: Template claim: If a true belief has features F1–n, then that belief does not constitute knowledge.11 10 Compare Nagel (2012) who claims that our judgments of knowledge are intuitive and not made on the basis of explicit theory, and may even contradict explicit theory. 11 This seems to me the most plausible version of the template account. By contrast, suppose that the relevant template instead stated that if a belief has features F1–n, then it is a case of justified true belief but not knowledge. This template explicitly raises the possibility of a justified true belief that is not

Immediate Justification

81

Realizing the Ten Coins Case has features F1–n, one has mediate justification to believe that the Ten Coins Case is not a case of knowledge. Combining this with one’s knowledge that the Ten Coins Case is so set up that the relevant belief is justified and true, one has mediate justification to believe (2). In a similar way, someone might gain from a bird book justification that any bird with a certain set of features is a magpie, and then on perceiving that the bird before her has those features, have mediate justification that the bird is a magpie. According to this template version of the inferential story, one has justification to believe (2) only if one has prior and independent justification to believe the template claim. However, this first inferential account of one’s justification to believe (2) does not seem to fit the history of philosophical discussion of Gettier cases, and in particular the way in which philosophers responded to counterexamples to their favorite templates. When one’s justification to believe a certain proposition depends on one’s prior and independent justification to believe some other proposition, then to the extent that one’s justification for the latter proposition is undermined, one loses one’s justification to believe the former proposition. For instance, suppose that a policeman’s justification to believe that Burglar Bill committed the burglary depends on his independent justification to believe that certain fingerprints were left at the scene. If his justification for the latter proposition is undermined, so is his justification for the former. Further, if he realizes that the latter justification is undermined, then he should no longer believe that Burglar Bill is the culprit but should be agnostic about the matter. Similarly, if one’s justification to believe the Gettier verdict (that Gettier cases are not cases of knowledge) depends on one’s prior justification to believe a certain template claim, then when the latter justification is undermined so is the former. But since the initially suggested templates were almost immediately counterexampled, the template view runs the risk that the Gettier verdict was unjustified. Further, on the template view, if one realizes that one’s justification to believe the relevant template claim has been undermined, one should withdraw one’s verdict that Gettier cases are not cases of knowledge. But philosophers continued to make the Gettier verdict even after they recognized that their suggested templates had been counterexampled. Instead of withdrawing their verdict, they attempted to revise their templates to get around the suggested counterexamples. But the same dialectic repeated itself. Many of the more complicated principles turned out to be subject to counterexamples. Still, the philosophers did not withdraw their verdict

knowledge. But it seems to me implausible that philosophers had a justified belief in such a template before Gettier produced counterexamples to the JTB theory of knowledge. Why would the JTB theory have been defended if philosophers already justifiably believed that beliefs with certain features are cases of justified true belief but not knowledge?

82

Foundations of Dogmatism

about Gettier cases but instead complicated their templates yet further in the light of the counterexamples. Given their increasing complexity, the revised templates hardly seem to be the kinds of principles for which one would have justification prior to the Gettier debates. But, on the template view, one has justification to believe the Gettier verdict in virtue of one’s prior and independent justification to believe a template claim. Thus, the template view risks the conclusion that for long periods during the Gettier debate, no one had justification to believe that Gettier cases are not cases of knowledge. I will illustrate the way in which suggested templates were quickly counterexampled leading to ever more complicated templates by looking at two main lines of approach to Gettier cases, no false lemmas, and no defeaters.12 According to a line of approach suggested almost as soon as Gettier had offered his cases, Gettier cases involve the presence of a false proposition in the relevant inference. Clarke (1963) suggested adding a fourth condition to the original JTB analysis, requiring that S’s belief be fully grounded, where by this he meant that the belief be based on an inference, no element of which is false. So, we could understand Clark as proposing the following template: Template (no false lemmas): If the justified true belief that p is not fully grounded, then it is not knowledge.

However, it was immediately recognized that the suggested necessary condition for knowledge is too strong. Non-inferential beliefs can constitute knowledge. Further, a subject may know that p by inference from a large set of evidence, even though one element of that set, q, is false (Lehrer 1965). To the extent that one’s justification to believe the Gettier verdict depends on one’s justification to believe the no false lemmas template, the counterexampling of the latter template undermined one’s justification to believe the Gettier verdict. As a result, the template view has the consequence that those relying on the no false lemmas template should have withdrawn their verdict about Gettier cases once they learned that it had been counterexampled. But they didn’t. Instead, proponents of the no false lemmas view offered further refinements and complications of the view to deal with the counterexamples. But even the refined diagnoses faced further counterexamples that led to further refinements in turn. As a result, the no false lemmas approach became ever more complex, which made it even less plausible that one had justification to believe it prior to one’s justification to believe the Gettier verdict. To get

12 A separate feature of the debate less important for our purposes is that many of the necessary conditions turned out to be insufficiently general in the sense that Gettier cases could be constructed which did not flout them. So, the suggested necessary conditions for knowledge were not effective as diagnoses of Gettier cases in general.

Immediate Justification

83

some sense of the complexity of the resulting views, let us consider just one portion of the subsequent dialectic. Lehrer (1965) considered various fourth conditions to be added to the original analysis on which S knows h iff (i) h is true, (ii) S believes h, (iii) S is completely justified in believing h. Lehrer first formulated the fourth condition as follows: (iv a) If S is completely justified in believing a false statement p which entails (but is not entailed by) h, then S has evidence adequate to completely justify his believing h in addition to the evidence he has for p.

However, this formulation of the fourth condition implausibly rules out knowledge in the following case. Suppose that I infer the proposition that someone in my office owns a Ford by inference from the following propositions, for both of which I have strong evidence: n h

Nogot, who is in my office, owns a Ford. Havit, who is in my office, owns a Ford.

I may know that someone in my office owns a Ford on this basis even in the circumstance in which n is false but h is true. In this case, I am completely justified in believing the following falsehood (n and h), that both Havit and Nogot, who are in my office, own Fords. Further, this falsehood entails that someone in my office owns a Ford, and I have no evidence for the latter proposition in addition to the evidence I have for (n and h). So, by this first formulation of the fourth condition, I do not know that someone in my office owns a Ford. In the light of this difficulty, Lehrer (1965) further developed this line of approach suggesting the following analysis: (iv c) If S is completely justified in believing a false statement p, which entails (but is not entailed by) h, then S would be completely justified in believing h even if S were to suppose that p is false.

However, this alternative also faced problems, e.g., with Skyrm’s pyromaniac case. More important for the plausibility of the template view, the complexity of Lehrer’s modified fourth condition undermines the plausibility of the claim that we had justification to believe it prior to reflection on Gettier cases. But the template view holds that one has justification to believe the Gettier verdict in virtue of one’s prior justification to believe some template. The same process of counterexamples leading to ever more complicated diagnoses of what’s going wrong in Gettier cases is also exhibited by another suggested template, the no defeaters template. According to this approach, Gettier cases arise because there are some truths that would have destroyed the believer’s justification had he believed them. In an early version of the no defeaters approach, Lehrer and Paxson (1969) suggest that non-basic knowledge that h requires true belief that h plus the condition that there be some statement p that completely justifies S in believing that h and no other statement which

84

Foundations of Dogmatism

defeats this justification. However, it turned out to be very difficult to specify a no defeaters condition that avoided counterexamples. According to a first attempt, when p completely justifies S in believing that h this justification is defeated by q iff q if true and the conjunction of p and q does not completely justify S in believing that h (33). However, the suggested condition on knowledge is not in fact necessary, as illustrated by the Neurotic Grabbit case: Sam sees his acquaintance, Tom Grabbit, steal a book from the library right in front of his eyes. But unsuspected by Sam, Tom’s mother asserts that Tom was miles away at the time of the theft and has a twin brother, John, who the parent tends to visually mistake for Tom who was in the library at the time. Yet the parent’s statement is only a neurotic lie.

In this case, Sam does know that Tom has stolen the book, but Sam fails the initial no defeater requirement. The Tom Grabbit case was just one of a series of cases that led to evermore complex refinements of the original no defeaters approach in a series of subsequent articles by Lehrer and others. To give some flavor of the complexity of subsequent definitions of the no defeaters condition, consider the following suggestion made by Lehrer and Paxson (1969, 36): If p completely justifies S in believing that h, then this justification is defeated by q if and only if 1) q is true, 2) the conjunction of p and q does not completely justify S in believing that h, 3)  S is completely justified in believing q to be false and 4) if c is a logical consequence of q such that the conjunction of c and p does not completely justify S in believing that h, then S is completely justified in believing c to be false.

This more sophisticated view has much less immediate intuitive plausibility than the initial statement of the approach. Indeed, it seems implausible that we had justification to believe this more sophisticated variant in advance of the Gettier debate that led to its development. But since the template view holds that we have justification to believe the Gettier verdict in virtue of a prior and independent justification to believe some template, the template view runs the risk of having the consequence that we lacked justification to believe the Gettier verdict. In summary, it seems that the actual history of philosophical discussion concerning Gettier cases does not fit the template account according to which one’s justification to believe that the Gettier cases are not cases of knowledge rests on one’s prior and independent justification to believe that cases fitting a certain template are not cases of knowledge, and realizing that the Gettier cases fit the template. The history of the Gettier debate suggests that, for much of the debate, we lacked a plausible template. So, the template model has the implausible consequence that we lacked justification to believe that Gettier cases are not cases of knowledge. Further, on the template model, philosophers should have retracted the claim that Gettier cases are not cases of knowledge when their favorite template was undermined. However, they did not do so but

Immediate Justification

85

rather continued to insist that Gettier cases are not cases of knowledge while they hunted for a new template. I have illustrated these general points by reference to the no false lemmas and no defeater approach. However, the same is true for other leading approaches, such as the causal approach.13 The philosophical dialectic concerning Gettier cases not only undermines the template picture of how we have justification to believe (2), but also undermines a possible coherence account of how we have such justification. On the relevant coherence account, one’s justification to believe (2)  is not prior to one’s justification to believe some template, and one’s justification to believe some template is not prior to one’s justification to believe (2). Rather, on the coherence picture, the fit between the template and (2) provides justification to believe both simultaneously. Like the template picture, the coherence picture holds that one’s justification to believe (2) is not prior to one’s justification to believe some template. As a result, it is undermined by the actual dialectic concerning Gettier cases, which suggests that we had justification to believe (2) prior to identifying a plausible template. In conclusion, immediacy in the case of intuition is motivated by the fact that it is the best view of certain central cases of intuition, such as the Gettier case. It remains to be shown that the view does not have problematic consequences for understanding philosophical dialectic.

6. Immediate Justification and Philosophical Dialectic Someone might wonder whether there is any tension between the view that intuitions provide immediate justification to believe their contents and the fact that philosophy as practiced involves a great deal of argument.14 In fact, we will

13

Goldman (1967) suggested the following template: Template (causation): If a true belief that p is not caused by the fact that p, then that belief does not constitute knowledge. However, Goldman himself realized that there are clear cases of knowledge in which the known fact does not cause the relevant belief, such as knowledge of the future, and knowledge of many generalizations. As a result, Goldman produced a much more complicated causal approach that is much less plausibly the kind of thing that we had justification to believe prior to considering our verdicts about a range of cases of knowledge and non-knowledge, including Gettier cases. On his more complex view, “S knows that p if and only if the fact that p is causally connected in an “appropriate” way with S’s believing p.” He added that appropriate causal processes include (i) perception, (ii) memory, (iii) a causal chain exemplifying one of two general patterns which is “correctly reconstructed by inferences, each of which is warranted,” and iv) combinations of (i), (ii), and (iii). The two general patterns include cases in which the fact is connected to the belief by a process of inference (pattern 1), and cases in which there is a common cause of both the subject’s belief that p and the fact that p (pattern 2). 14 This idea surfaces in the criticism that moral intuitionism cannot account for moral disagreement, e.g., Mackie (1977), and in Cappelen’s (2011) attack on the role of intuitions in philosophy more generally. Huemer (2005) responds to the criticism of moral intuitionism (chapter 6).

86

Foundations of Dogmatism

see that it is easy for a defender of immediate justification to account for the extent of philosophical argument. Suppose that one has immediate justification to believe the content of some intuition, say, the content that necessarily, Gettier cases are not cases of knowledge. As we previously noted, much of the discussion in the Gettier literature is about the correct diagnosis of the fact that Gettier cases are not cases of knowledge. The controversy about the correct diagnosis of Gettier cases and the arguments provided for and against each diagnosis are hardly surprising on the view that one has immediate justification to believe that Gettier cases are not cases of knowledge. Indeed, I exploited the controversy about the correct diagnosis of the cases to motivate the view that we have immediate justification to believe that Gettier cases are not cases of knowledge. Almost all contemporary defenders of immediate justification deny that immediate justification is infallible or incorrigible and allow that immediate justification may be defeated. This enables them to explain much of the philosophical argument concerning judgments about cases as concerning whether there are defeaters for the relevant immediate justification. Interestingly, in the Gettier case, very few defeaters have been offered for the belief that Gettier cases are not cases of knowledge.15 However, for many other key philosophical intuitions, a wide range of potential defeaters has been offered. For example, consider non-contextualist explanations of contextualist intuitions about bank cases (e.g., see Rysiew 2001; Hawthorne 2004; Bach 2005; Stanley 2005; Williamson 2005; Nagel 2008). So far, we’ve seen how a defender of immediate justification can easily explain extensive philosophical argument both concerning the diagnoses of accepted intuitions and concerning whether there are defeaters for the relevant intuitions. It may be useful to delineate one further way in which a defender of immediacy can explain the prevalence of philosophical argument. Even in a case in which the relevant immediate justification has not been defeated, its content might not be widely accepted. In general, philosophers and other enquirers are not merely concerned with the epistemic standing of their own beliefs, but also concerned that the truth should be accepted by the community of enquirers. They are concerned to try to ensure that the community of enquirers is undertaking the relevant enquiry in the right way and is not being deflected by false assumptions and a failure to see the evidence. As a result, philosophers have reasons to provide arguments even for a claim that they take themselves and other philosophers to already have justification to believe.

15 Weatherson (2003) suggested that the intuition that Gettier cases are not cases of knowledge might not reflect the nature of knowledge on the grounds that the referent of “knowledge” is determined by two factors, use and naturalness, and naturalness might cut against use. The idea, subsequently retracted, was that JTB theory might be more natural than any other alternative.

Immediate Justification

87

7. Conclusion I have been developing and defending the idea that intuition provides immediate justification. I started out by showing how one could apply the thesis of immediacy to a wide range of views about the nature of intuition and not only the view that intuitions are experiences. I then argued that the best motivation for immediacy in the case of intuition is via an appeal to cases. By considering in detail the history of philosophical discussion concerning Gettier cases, I argued that it is plausible to suppose that we had immediate justification to believe the relevant verdict about these cases, namely, that these are cases of justified true belief that is not knowledge. Last, I showed how the claim that some intuitions provide immediate justification is easily compatible with the nature and extent of argument in philosophy.

Acknowledgments Thanks to Justin McBrayer, Daniele Sgaravati, Ernie Sosa, Margot Strohminger, and Chris Tucker for comments, and to audiences at St. Andrews and Rutgers who provided useful feedback on earlier versions of the chapter. I would also like to gratefully acknowledge grants from the AHRC and Leverhulme Trust for work on intuitions and philosophical methodology, which supported this research.

References Bach, Kent. 2005. “The Emperor’s New ‘Knows.’ ” In Gerhard Preyer and Georg Peter, eds., Contextualism in Philosophy:  Knowledge, Meaning and Truth, 51–90. Oxford:  Oxford University Press. Brewer, Bill. 1999. Perception and Reason. Oxford: Oxford University Press. Cappelen, Herman. 2011. Philosophy without Intuitions. Oxford: Oxford University Press. Chudnoff, Elijah. 2011. “The Nature of Intuitive Justification.” Philosophical Studies 153: 313–33. _____. Forthcoming. “What Intuitions Are Like.” Philosophy and Phenomenological Research. Clark, Michael. 1963. “Knowledge and Grounds:  A  Comment on Mr. Gettier’s Paper.” Analysis 24(2): 46–48. Daniels, Norman. 2003. “Reflective Equilibrium.” In Edward N. Zalta, ed., Stanford Encyclopaedia of Philosophy, http://plato.stanford.edu/archives/spr2011/entries/ reflective-equilibrium/. DePaul, Michael. 2009. “Phenomenal Conservatism and Self-Defeat.” Philosophy and Phenomenological Research 78: 205–12. Dretske, Fred. 2000. “Entitlement: Epistemic Rights without Epistemic Duties?” Philosophy and Phenomenological Research 60: 591–606. Feldman, Richard. 2002. Epistemology. Upper Saddle River, NJ: Prentice-Hall.

88

Foundations of Dogmatism

Gettier, Edmund. 1963. “Is Justified True Belief Knowledge?” Analysis 23(6): 121–23. Goldman, Alvin. 1967. “A Causal Theory of Knowing.” Journal of Philosophy 64: 357–72. Goldman, Alvin. 2008. “Immediate Justification and Process Reliabilism.” In Quentin Smith, ed., Epistemology: New Essays, 63–82. Oxford: Oxford University Press. Hawthorne, John. 2004. Knowledge and Lotteries Oxford: Oxford University Press. Huemer, Michael. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74(1): 30–55. _____. 2005. Ethical Intuitionism. Basingstoke: Palgrave McMillan. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Ichikawa, Jonathan and Benjamin Jarvis. 2009. “Thought Experiment Intuitions and Truth in Fiction.” Philosophical Studies 142(2): 221–46. Lehrer, Keith. 1965. “Knowledge, Truth, and Evidence.” Analysis 25(5): 168–75. Lehrer, Keith and Thomas Paxson Jr. 1969. “Knowledge: Undefeated Justified True Belief.” Journal of Philosophy 66: 225–37. Ludwig, Kirk. 2007. “The Epistemology of Thought Experiments,” Midwest Studies in Philosophy 31(1): 128–59. Mackie, John. 1977. Ethics: Inventing Right and Wrong. New York: Penguin. Malmgren, Anna-Sara. 2011. “Rationalism and the Content of Intuitive Judgements.” Mind 120(478): 263–327. Martin, Michael. 2009. “Re-upholstering a Discipline.” Philosophical Studies 145(3): 445–53. Nagel, Jennifer. 2012. “Intuitions and Experiments:  A  Defence of the Case Method in Epistemology.” Philosophy and Phenomenological Research 85(3): 495–527. Nagel, Thomas. 2008. “Knowledge Ascriptions and the Psychological Consequences of Changing Stakes.” Australasian Journal of Philosophy 86(2): 279–94. Peacocke, Christopher. 2004. The Realm of Reason. New York: Oxford University Press. Plantinga, Alvin. 1993. Warrant: The Current Debate. Oxford: Oxford University Press. Pollock, John L. 1986. Contemporary Theories of Knowledge. Totowa, NJ: Rowman and Littlefield. Pryor, James. 2005. “There Is Immediate Justification.” In Matthias Steup and Ernest Sosa, eds., Contemporary Debates in Epistemology, 181–201. Oxford: Blackwell. _____. 2000. “The Sceptic and the Dogmatist.” Nous 34(4): 517–49. Rysiew, Patrick. 2001. “The Context-Sensitivity of Knowledge Attributions.” Nous 35(4): 477–514. Sosa, Ernest. 2007. A Virtue Epistemology, Vol. I. Oxford: Oxford University Press. _____. 2006. “Intuitions and Truth.” In Patrick Greenhough and Michael P. Lynch, eds., Truth and Realism, 208–26. Oxford: Oxford University Press. _____. 1998. “Minimal Intuition.” In Michael De Paul and William Ramsay, eds., Rethinking Intuition, 257–70. Lanham, MD: Rowman and Littlefield. _____. 1980. “The Raft and the Pyramid.” Midwest Studies in Philosophy, 5(1): 3–25. Stanley, Jason. 2005. Knowledge and Practical Interests. Oxford: Oxford University Press. Vogel, Jonathan. 2008. “Internalist Responses to Scepticism.” In John Greco, ed., The Oxford Handbook of Scepticism, 533–56. Oxford: Oxford University Press. Weatherson, Brian. 2003. “What Are the Good Counterexamples?” Philosophical Studies 115: 1–31. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Blackwell. _____. 2005. “Contextualism, Subject-Sensitive Invariantism and Knowledge of Knowledge.” Philosophical Quarterly 55(219): 213–35. _____. 2000. Knowledge and Its Limits. Oxford: Oxford University Press.

{5}

Problems for Credulism James Pryor 1. Introduction I’ve defended views about the epistemology of perception that I call “dogmatist.” They are close to views defended by Michael Huemer and John Pollock, and somewhat more broadly to a range of views defended by others. These views have been criticized for being incompatible with Bayesianism. I will take Roger White’s “Problems for Dogmatism” (White 2006) as representing these criticisms.1 This chapter reviews and assesses some central pieces of the criticisms. Doing so will reveal some limits of Bayesianism when it comes to representing undermining evidence. I won’t argue directly that the criticisms fail, nor concede directly that they are correct. As you’ll see, the dialectic is not that straightforward. Additionally, my aim here isn’t to speak on behalf of my own idiosyncratic views of perception. I hope instead to show that whatever problems lie here aren’t just problems for me. We don’t just have “Problems for Dogmatism,” but, more generally—if there are problems—then “Problems for Credulism.” That’s a term I’ll introduce to capture a broad range of epistemologies. It includes my own view of perception, but also many other contemporary epistemologies. All of them want to say things about undermining that are awkward to represent in Bayesian terms. We’ll begin by trying to get sharper about what “undermining” consists in. That will turn out to be not so easy. We will be able, though, to get sharper about what “dogmatism” and “credulism” are. Then the bulk of our discussion will critically

1 There is not a single criticism, but a cluster, only some of which I will directly engage with. The provenance of these criticisms is messy. I was aware of some of the difficulties myself and began exploring formal alternatives to Bayesianism in 2002. Cohen and Wright were at that time also pressing these complaints against me in correspondence. Shortly after, various of the criticisms were spelled out in print in: Cohen 2005; John Hawthorne 2004, 73–76; Schiffer 2004; White 2006; and Williamson 2005. Since then they’ve been more widely discussed. See http://www.jimpryor.net/research/bayesian.html for links to the work I’m aware of that deals with these issues.

90

Foundations of Dogmatism

exposit some of the supposed Bayesian problems for dogmatism. We’ll see that the “problems” require some substantial assumptions about the philosophical import of different elements of the Bayesian formalism. These assumptions are widely held, but they go beyond anything that’s part of Bayesianism proper, or that familiar arguments for Bayesianism establish.

2. Undermining The way I’ll use the term defeat, it’s only things you learn or get justification to believe that count as defeaters. There’s an alternate use of the term, where mere facts in the world can “defeat” some epistemic status you would otherwise have, even if you’re ignorant of those facts. That’s also a legitimate and interesting notion, but let’s not call it “defeating.”2 I want to reserve “defeating” for a kind of justification you can acquire. What are our different paradigms of defeating evidence? Let E be the fact that Ernie tells me that his aunt’s pet Precious is a bird. This supports the conclusion H, that Precious has the ability to fly. However, Orna gives me opposing evidence. She says that Precious is a dog rather than a bird.3 Defeating evidence need not oppose Ernie’s testimony in this direct way. There are other ways to weaken the support I  have for believing H, where the new evidence doesn’t itself intuitively speak one way or the other about Precious’s flight ability. An example:  Ursula tells me that Ernie has no idea what Precious’s species is; he’s just guessing. She doesn’t herself weigh in about Precious’s real species or flight ability. I  call defeating evidence of this sort undermining evidence.4 Yet other kinds of evidence are also possible:  perhaps I  get refining evidence: I learn that Precious is indeed a bird, but a predominantly flightless one, such as a penguin. Or I might learn that Precious is no bird but is capable of flight all the same. (A flying squirrel?) We will focus on undermining defeaters.5 Most of us discern an intuitive kind there, but if it is a real epistemic kind, it’s difficult to say in a rigorous way what makes it distinctive. 2

We might call it “disabling.” Sometimes this is called “overriding” or “rebutting” evidence. Since opposing is a matter of degree, the terminology I’m suggesting is better. Depending on how the example is filled out, you may end up trusting Ernie on balance more than you do Orna; but so long as her testimony has even some impact on your credence in H, it will have opposed Ernie’s testimony. 4 Another term sometimes used is undercutting. 5 Some epistemic effects have an undermining feel, but don’t obviously involve the acquisition of new evidence, so they may not be “defeaters” as I understand this term. Hartry Field reminded me of the idea that the mere articulation of a new scientific theory could “undermine” an old one, even in advance of our acquiring evidence for the new theory. I won’t try to settle here how such phenomena should be categorized. My disposition, though, is to count new insights and recognitions as a species 3

Problems for Credulism

91

The natural first thought is that evidence that opposes the support E gives me for H will do so by speaking for the negation of H. Orna’s evidence is an example:  her testimony that Precious is a dog supports his actually being a dog, and so incapable of flight. Whereas undermining evidence, the thought goes, speaks not for the negation of H, but rather for some claim such as that E is unreliable with respect to H, that E should not be trusted on this matter, and so on. This sounds like a natural way to distinguish undermining evidence from (at least some) other species of defeating evidence. Yet it takes only a moment to notice that each of the claims we just made can be applied with some merit to the other kind of defeating evidence, too. The opposers may speak in the first place for not-H, but then by doing so, don’t they also constitute some evidence against the reliability, in this context, of any evidence like E that speaks for H? The underminers may speak in the first place for the unreliability of E, at least concerning H, but then by doing so, don’t they contribute to its being less reasonable for me to believe H—and so more reasonable to believe not-H—than it was before? Perhaps the thought is significant that the opposing evidence speaks “directly” or “in the first place” for one kind of upshot, and then only by way of that for other evidence’s unreliability. Similarly the thought that the undermining evidence speaks “directly” or “in the first place” for some other evidence’s being in this context unreliable, and then only by way of that against the hypotheses that evidence formerly supported. But to turn those thoughts into a fleshed-out, rigorous articulation of what undermining evidence is and why it’s distinctive is no easy matter. I welcome attempts to do so, but I won’t pursue it further here. A different thought is that underminers work by “screening off ” the epistemic contribution E makes to H. What does this mean? Let’s at this point introduce a notion I’ll call your “epistemic probability function.” This is a probability function that is intended to represent what credences you have ex ante or “propositional” justification to have.6 Of course, it

of evidence, too; so these cases could after all involve our acquiring defeating evidence. It’d just be evidence of a reflective, intellectual kind, rather than the sort scientists usually talk about. 6 Some authors call this your “rational” as opposed to your actual credence function. I resist that usage because this probability function is intended to track just the character of your evidence, and the notions that informal epistemologists understand by “rational” in my view depend on more than just that. Also, authors sometimes understand “rational credences functions” to track credences that are both held and that one has justification for holding. But I mean to discuss only the latter. For expository convenience, I will allow myself to use expressions like “more justification to believe” to mean “justification to be more confident.” But it is a substantive question, that I do not intend to take a stand on, what epistemic probabilities in the sense we’re considering really have to do with acquiring more justification to have the attitude of categorical belief. Nor do I make any effort here to address what they have to do with epistemic probability locutions in natural language.

92

Foundations of Dogmatism

may turn out that such facts aren’t representable by a probability function— this is a prospect we’ll return to. Let E and H be the propositions about Ernie and Precious from before, and let U be the putative undermining proposition. Let Old(.) be your epistemic probability function before acquiring evidence E, and New(.) be your probability function after acquiring evidence E. I assume that it makes no difference whether we consider E coming first, and then U, or U coming first and then E. (We’ll talk more about this assumption later.) I assume also that we can usefully talk about your probability function conditioned on the hypothesis that U, though generally you’ll never get justification to be certain that U, but only more confident that it’s so. With that background in place, let’s say that U screens off a contribution E makes to H when: (i) E supports H, that is: Old(H) < New(H). (ii) But against the assumption of U, E does not support H: New(H|U) ≤ Old(H|U).7

Now this is not enough to give us a useful notion of undermining, for it may be that (ii) holds because U already includes all the information E would provide. Alternatively what if (ii) holds because U opposes the evidence provided by E, and does so more effectively against an evidential background that includes E: for instance, Orna might tell me not that Precious is a dog, but that if Ernie says Precious is a bird then Precious is a dog. Let’s refine the conditions as follows. We keep (i) and add: (iii) U doesn’t speak for or against H on its own.

Let’s also modify (ii) to include not just the case where U wholly defeats the effect of E, but also cases where U merely reduces it: (ii*) Against the assumption of U, E supports H less: New(H|U) < New(H).

This will count as evidence that Ernie has a ≥70 percent chance to be making things up as undermining, as well as evidence that he is definitely doing so. Putting these all together, and relying on a natural interpretation of (iii), we get: Old (H | U) = Old (H) ≤ New (H | U) < New (H)

7 Standardly the “screening off ” language is reserved for cases where New(H|U) = Old(H|U). Also, sometimes it is further required that E doesn’t return the favor to U. The primary home of this notion has been in discussions of causation; see, for example, Reichenbach (1956) and Salmon (1984).

Problems for Credulism

93

These refinements help, but they don’t fully address the concerns voiced a moment ago. Like positive evidence, undermining also comes in degrees. And it can also itself be defeated, and can come mixed with other opposing or supporting information. It may be no easy matter to disentangle these different evidential components, when they do come mixed together. Here’s a series of examples that illustrate that possibility, and obstacles it poses to analyzing undermining in terms of screening-off. In each of the examples, I’ll possess some auditory evidence, which will be undermined by olfactory evidence of smoke. I’ll assume background knowledge that smoke tends to make me have random auditory hallucinations more frequently. (Perhaps it’s opium smoke?) In the first case, the evidence that the smoky smell threatens to undermine is my auditory experience of a radio. On its own, that experience would give me some justification to believe a radio is nearby. In the presence of the smoky smell, though, this experience cannot reasonably be relied on to the same degree. I now have more justification than before for thinking the auditory experience is hallucinatory. It doesn’t matter how much less justification to believe there’s a radio nearby this results in; it’s enough for our purposes that it results in some. Of course, I can’t be certain that there is smoke, much less that I’ve auditorily hallucinated. What I’m hallucinating might instead be the smoky smell. But I needn’t be certain for the evidence of smoke to have some undermining effect on my auditory justification that there is a radio nearby. In the second case, the evidence that the smoky smell threatens to undermine is instead my auditory experience of singing birds. We will assume background knowledge that when there really is smoke present, birds flee. They don’t stick around and sing. Now, in this case too the smoky smell contributes to making it reasonable for me to rely less on my auditory experience. However, what’s different in this case is that the smoky smell also on its own speaks against the same proposition my auditory experience seemed to support. That is, the smoky smell on its own testifies against there being any birds present. So what we have here is that the smoky smell both undermines and opposes my auditory evidence for birds.8 In the third case, the evidence that the smoky smell threatens to undermine is instead my auditory experience of an approaching fire engine. On its own, that auditory experience seems to support the hypothesis that firemen are nearby. However, in this case too, the smoky smell calls for me to be more guarded toward my auditory experiences. Perhaps this too is just another smoke-generated auditory hallucination. On the other hand, my olfactory experiences themselves seem to improve the likelihood that firemen are nearby. Where there’s smoke, there are usually firemen. There may be a

8

For another example like this, see the case of the butler and the maid in Weisberg ms.

94

Foundations of Dogmatism Evidence that on balance is somewhat defeating Undermining evidence Opposing evidence Quotidian Non-quotidian

figure 5.1

complex interaction between the undermining effect my smoky experiences have on my auditory evidence and the independent olfactory evidence for firemen they put in its place—for there actually being firemen there to be heard presumably makes it again somewhat less likely that I’m hallucinating. We haven’t said enough to know what equilibrium this leads to. What’s important is that here the smoky smell to some degree undermines one body of evidence for conclusion H, while at the same time itself providing other evidence for H.9 I call these last two cases mixtures: in the one case, a mixture of undermining and opposing, and in the other case, a mixture of undermining and supporting. They are the shaded regions in figure 5.1. I’ll explain the contrast between quotidian and non-quotidian undermining later. Some of the time, as perhaps in these examples, we may be able to separate out different considerations in a mixed body of evidence, which point in the different directions. However, I don’t think we’re in any position to assume it will always be so. Mixed bodies of evidence may not always be easily decomposed; they may, for all I know, sometimes be in principle un-decomposable. Mixed evidence of these sorts makes difficulties for the project of explicating undermining in terms of screening-off. A screening-off test seems as though it would, at best, capture what’s going on in cases of pure undermining, with no mixture of additional opposing or supporting effects. It won’t capture cases where the combination of U and E on balance increases the likelihood of H (because U adds more support of its own for H than it diminishes E’s). Because of condition (iii), which aimed to exclude mere opposing evidence, our screening-off test can fail to capture cases where U both undermines and

9

Silins (forthcoming) gives some examples of this sort.

Problems for Credulism

95

opposes E’s support of H.  We should hope for an account of the distinctively undermining evidential effect as it shows up in all of these cases.10 Some have tried to do better at spelling out undermining in probabilistic terms.11 We won’t pursue those attempts here, either. It’s enough that we’ve seen some of the initial difficulties. In the rest of this essay we’ll see other sorts of awkwardness for reconciling Bayesianism with some intuitive views about undermining. For these purposes, I propose we just proceed with our intuitive, as-yet-unanalyzed understanding of undermining.12 10 David Barnett offered the following example, where your belief is intuitively undermined but no change of credence in the relevant proposition is called for. You’re at a crossroads, inquiring the way to Camelot. Your informant tells you it lies to the right. You know your informant is either a knight, who is perfectly reliable, or a knave, who is perfectly anti-reliable, or a fool, who answers at random. You initially estimate he’s 45 percent likely to be knight, 45 percent knave, 10 percent fool, and so you suspend judgment whether Camelot really does lie to the right. Still, you do have some evidence to think so; it’s just balanced by equally weighty evidence to think not. Next you learn that your informant is the fool. Intuitively, some kind of undermining has now taken place. Earlier the informant’s testimony seemed 50 percent likely to be accurate, and 45 percent likely to be accurate because reliable. Now your evidence still seems 50 percent likely to be accurate, but not at all likely to be reliable. Because of the way the case is set up, no change in your credence about Camelot’s direction is called for; but your grounds for believing Camelot lies to the right seem intuitively weakened from what they were before. This case invites the idea that although undermining your justification for H may not be a matter of how your credence in H changes, it may be a matter of how your credence in some other proposition changes—perhaps a proposition about the reliability of your evidence for H. But read on, and see the variety of undermining considerations we display in section 4. Then try to say specifically what other proposition we can analyze the undermining of your justification for H in terms of. Also, would you insist that no subject’s justification for H can be undermined who lacks defined credences for propositions about reliability, their own epistemic status, and so on? I acknowledge that the notion of “defined credence” we’re working with is an epistemic one, not a psychological one. But I would still be reluctant to think every subject who is vulnerable to undermining must have defined credences for these things. 11 See, for example, Kotzen ms. 12 I’ll mention briefly a last strategy for explicating undermining evidence. John Pollock was the first I know to discuss the phenomenon of undermining in a sustained way. (For a survey of his views, see Sturgeon, forthcoming.) Pollock called this phenomenon “undercutting,” and proposed that U undercut E’s support of H just in case U was evidence against the conditional “E wouldn’t be true unless H were.” Interpreting this is not easy, because it involves an “unless” construction, the subjunctive mode, and the notion of evidence against a conditional—each of which is tricky in its own right. (Also, it’s not clear how seriously we should interpret its similarity to Dretske’s (1971) defintion of a “conclusive reason.”) I think the proposal is most plausible when the relevant conditionals are understood epistemically. If we understand them truth-functionally, then the proposal seems to be that something undercuts E’s support of H iff it’s evidence against E⊃H. Both directions of this are odd: why couldn’t there be underminers that make E and not-H less likely (though more likely than they make E and H)? And if Orna tells me that Precious can’t fly (not-H) but that Ernie will tell me it can (E), must that be counted an underminer of E? Understanding these conditionals counterfactually brings other difficulties. So as I said, the proposal seems most plausible if the conditionals are understood to express something epistemic, along the lines of “If E, then likely also H,” or “If E, then it might be that not-H.” (Pollock himself glosses his conditionals as “E does/doesn’t guarantee H”; but these are open to more-or-less the same range of interpretation.) And yet, what does it mean for U to be justification for (or against) an epistemic conditional? On some views, such conditionals don’t have truth-conditions and so perhaps aren’t legitimate targets of justification. Even if that thought is mistaken, and it really does make sense for there to be justification for or against epistemic conditionals, I submit it’s not a sense we yet have any good understanding of. So even if Pollock’s criterion ultimately does prove to be correct, I don’t think it’s a suitable place from which to begin our inquiries.

96

Foundations of Dogmatism

3. Dogmatism and Credulism When I began several years ago to use the term dogmatism, I meant it to be the view that immediate justification exists even where it might be undermined by skeptical (or mundane) defeaters that one has no epistemically antecedent grounds for ruling out. That is, justification is sometimes both immediate and underminable. More specifically, what I mean by immediate justification is some amount (it matters not how much) of prima facie justification to believe something (that is, prospective or ex ante justification) that does not even partly come from—or, using other language, is not even partly constituted by—your having justification to believe something else. Justification is on the other hand mediate when it does in part come from, or is in part constituted by, your having justification to believe something else. So understood, your justification for some belief might be mediate even if the belief was arrived at spontaneously, not via any chain of explicit reasoning. And even in cases where your belief was explicitly (and competently) inferred, it might not be based on all the upstream considerations whose justification entitled you to have it.13 I’ve advocated a view about perception that emphasizes the possibility of immediate but underminable justification, and many have come to associate the term dogmatism with that particular view. I’ve also said things favorable about “Moorean” arguments against skeptical hypotheses (though also some things unfavorable). And again, some have come to associate the term dogmatism with sympathy for such Moorean arguments. However, it is a substantive claim that these go together and should not be assumed as a matter of definition.14 It’s not up to me how others will use the term dogmatism. But I will use it to name the general thesis that justification is sometimes both immediate and underminable. It doesn’t include by definition any commitments about Moorean arguments. Neither does it include by definition any commitments about why some body of justification is immediate. Some of us favor an internalist view of perceptual justification that ascribes the justificatory power to the quality of our perceptual phenomenology.15 But that’s just one species of dogmatism. Other stories are also possible. And those stories need not always

13 It’s a substantial question whether the grounds of a competently based belief need to include all the considerations justification for which make up the justification you have for that belief. If they do, and we subscribe to commonly held views of which considerations those are, and commonly held understandings of the basing relation, it’s doubtful that very many beliefs are competently based. 14 Silins (2008) argues for the first without the second. See also Wedgwood (forthcoming), Neta (2010), and Kotzen (2012). 15 Besides myself, see also Huemer (2001, 2006, 2007), Tucker (2010), and Chudnoff (2011).

Problems for Credulism

97

be internalist. A reliabilist or a disjunctivist about perceptual justification can also claim that justification to be immediate yet underminable. Finally, “dogmatism” in the general sense used here isn’t a thesis specifically about perception. Perhaps you deny that perceptual justification is immediate; but you think your knowledge of what you intend to do is often immediate— yet also underminable. Or you think mathematical justification is sometimes immediate—yet also underminable. My own inclination is to expect all justification to be underminable, and I’ll try to persuade you to think so too. So if you think we ever have immediate justification about anything, you should be a dogmatist, too. Many philosophers are already dogmatists in this sense, though they may never have applied that label to themselves. However, much of what we’re going to discuss bears also on an even more inclusive group, which has not before been named. I  will call this group the credulists.16 We will look at what defines the group more carefully in the next section. For the moment, here is a quick gloss. Credulists think you can be justified in believing H, in a way that would be undermined by evidence for U, without antecedent justification to believe not-U needing to be a constitutive part of your original justification for H. We can sum this up in the slogan: Your justification for believing H is vulnerable to being undermined in ways you didn’t need to antecedently rule out. A  dogmatist adds the additional commitment that this is because no antecedent justification to believe anything was part of your original justification to believe H. Credulists need not be dogmatists, though. They are allowed to think that antecedent justification to believe some things was part of your justification to believe H—maybe even antecedent justification to rule out some underminers. But there are some propositions like U that would undermine and didn’t need to be antecedently ruled out.17

16 This is just an arbitrary label; there’s not supposed to be any tight connection to the folk meaning of “credulity.” I wanted a term that sounded reminiscent of “dogmatism” but more moderate and so more inclusive. 17 A comment on locutions like “In order to be justified in believing H, you need/require antecedent justification to believe not-U.” When I say this, I always have in mind the claim that the antecedent justification is included in your justification to believe H. There’s another (perhaps more literal) reading of the locution, where it only says that having some antecedent justification to believe not-U is a necessary condition to have justification for H. I do acknowledge the difference between these claims (see note 33 of Pryor, 2000); and Silins (2008) argues for the importance of separating them—partly on the basis of the Bayesian issues we’ll be examining later. (See also Silins (forthcoming) and McGrath (forthcoming), section 3.) However, I will neglect the difference in this discussion, because the dialectic is already very complex. “Neglect” not “ignore”: I won’t assume there is no difference. But it may be that some of what I say needs to be revised when we attend to the difference more carefully. Neta (2010) argues for this distinction and for others, some of which I  also acknowledge but describe differently. My dogmatism about perception is what he calls “mediate liberalism.”

98

Foundations of Dogmatism

In the next section, we will call underminers of this sort non-quotidian. So credulism is commitment to the possibility of non-quotidian undermining. In later sections, we will see difficulties for representing non-quotidian undermining in either Classical or Jeffrey Bayesian terms. These difficulties won’t come from the formalisms on their own, but from the combination of the formalisms with a popular set of interpretive assumptions, which I  will identify. It shouldn’t be surprising that epistemic effects belief in which make one a credulist should be difficult to represent in Bayesian terms. For consider the following: Hypotheses H1 and H2 are logically equivalent, and as a matter of fact you do justifiably give them equal credence. However, your own mental states are not transparent to you. Suppose you now acquire evidence that your credence in H2 is in fact lower than your credence in H1. It is debatable what effect such evidence should have, but there is a presumptive case that it would put some kind of pressure on you to raise your credence in H2 and/or lower your credence in H1. Refusing to alter your current credences would seem unreasonably unresponsive. True, the evidence you’ve just acquired is misleading, but it’s part of the case we’re imagining that you’re in no position to know this.

Or: Hypotheses H1 and H2 are logically equivalent, and as a matter of fact you do justifiably give them equal credence. However, their logical relationship is not transparent to you. Suppose you now acquire evidence that H1 is in fact logically stronger than H2, and that the possibility of H2-but-not-H1 should have a positive credence. Here too, there is a presumptive case that the evidence would put some kind of pressure on you to raise your credence in H2 and/or lower your credence in H1. Here too, the evidence is misleading, but you’re in no position to know this.

It’s already commonly granted that Bayesian formalisms aren’t straightforwardly able to represent epistemic effects of these sorts. But many cases of non-quotidian undermining seem, at least to this author, to be cut from the same cloths. If it’s in fact true that non-quotidian undermining can’t be represented in Bayesian terms, what should be our response? Should we conclude that nonquotidian undermining is impossible? Should we rein in our explanatory ambitions, and say the formalisms only apply where we can idealize in such a way that the possibility of non-quotidian undermining can be ignored? Or should we give up some of the interpretive assumptions that make non-quotidian undermining unrepresentable? We will return to these choices.

Problems for Credulism

99

4. Credulism More Carefully To get a better handle on what credulists believe, let’s consider the following example. You have the evidence E, that a certain barometer is falling. E together with other things you are justified in believing—for example, about reliable connections between that barometer and the upcoming weather—justify you in believing H, that there will soon be rain.

Though our focus is prospective or ex ante justification, for some of the variations I want to consider next, we should suppose that this transition from E to H is an inference you’ve explicitly drawn. So, how can the justification you have for H be defeated, while leaving your original evidence in place—and intuitively without directly speaking for or against H? The most quotidian way for this to happen is for the defeating evidence to oppose some auxiliary hypothesis, where it’s also plausible that you needed antecedent justification for that hypothesis, to have the justification you did for H. In our example, evidence that your barometer is not reliable about the weather would be of this sort.18 There are things one can fuss about here. Perhaps we shouldn’t say it was ever E itself that supported H, but only the combination of E and the hypotheses that connected it to H. If so, does this really deserve to be called a case of undermining? Undermining requires that what was evidence for H is still available, but supports H less. This instead is a case where you start with evidence E-and-connecting-hypotheses to believe H, and new evidence takes that old evidence away. You’re then no longer (or less) justified in believing E-and-connecting-hypotheses. It doesn’t matter to me whether these quotidian examples do deserve to be called examples of undermining. What interests me is that many of us think there are possibilities of undermining other than what’s just been described. That is, non-quotidian forms of undermining. For example, a dogmatist can think your justification for E itself is underminable, but immediate, so that there are no hypotheses you needed antecedent justification for to have the justification you did for E. Credulists more generally take the same attitude toward some underminer that the dogmatist does:  it’s a way in which your justification is vulnerable, that you didn’t need antecedent justification to believe wasn’t realized. But a

18 As mentioned before, your needing antecedent justification for an auxiliary hypothesis needn’t mean it had to be among the grounds on which you based your inference that H, for that inference to be competent.

100

Foundations of Dogmatism

credulist needn’t insist that no antecedent justification was involved. In our example, your evidence for H did require antecedent justification to believe the barometer was reliable. But arguably it didn’t require antecedent justification to believe the higher-order claim that E plus evidence of reliability are enough to justify belief in H.  Yet the credulist may think that evidence that these aren’t enough would justify you in being less confident, or in some way more guarded, toward H.19 Such evidence may take the form of a compelling philosophical argument that there is no justified belief about the future. Or that, though there may be such, it is not available to subjects with the cognitive shortcomings you know yourself to have. Shouldn’t evidence for those claims undermine your confidence in H at least somewhat?20 Suppose we concede that your original justification for H does have to include antecedent justification to reject the conclusions of those arguments.21 Then consider instead evidence that what was in fact your antecedent justification for doing so, wasn’t. Shouldn’t evidence for that undermine your confidence in H at least to some degree? And did you need antecedent justification to reject it? In all of these cases, I am supposing that your original evidence really did support H; that enables us to sidestep questions about how to deal with illusions of support. Hence, the undermining evidence is evidence for a false claim. But that surely is no obstacle to your acquiring it. Some philosophers I’ve discussed these issues with have resisted the idea that philosophical arguments for false conclusions can provide justification for those conclusions. They seem to me to have unrealistic conceptions of philosophical justification. But we needn’t fight about that. The undermining evidence needn’t be provided by any mistaken philosophical argument, itself. Rather, let it come in the form of empirical evidence about how your philosophical mentors assess arguments you don’t yourself understand, or haven’t yourself seen. Other philosophers I’ve discussed these issues with have complained about this defeating evidence being directed in the first place at questions like

19 I count such cases as examples of undermining, though there may be interesting differences between them and cases of undermining that aren’t higher order. Some philosophers will urge that degrees of confidence don’t by themselves provide the right structure to model the relevant guardedness. I have some sympathy for that, but won’t pursue it here. 20 Cases like this, and the ones that follow, have been much discussed in the disagreement literature, especially by David Christensen. See Christensen (2007, 2008, 2010a, 2010b, 2011); Elga ms; Feldman (2005, 2006, 2009); and Kelly (2010, esp. section 4). See also Huemer (2011) and Schechter (2013). 21 This is in the vicinity of what I call “Inferential Internalism” in Pryor (2001). I take that term from one of its proponents: see Fumerton (1995, chapter 3). Alston (1989) opposes the view in several essays. For more recent work on this issue, see Tucker (2012) and the work he references. Boghossian (2003) argues that for deductive inference, simple forms of Inferential Internalism and Inferential Externalism are both wrong.

Problems for Credulism

101

whether you’re justified in believing H, rather than H itself. Of course I agree those are separate questions. It may even be that we can not expect a straightforward correlation between your epistemic position with respect to the one and your epistemic position with respect to the other.22 But that is compatible with evidence for the higher-order claim having some defeating effect on your justification for the lower.23 Another kind of underminer would be (misleading) evidence that no claims about the future are true. Shouldn’t that put some pressure on you to withhold belief about the future? Must your justification to believe H have included antecedent justification to disbelieve any such metaphysical evidence? One can have credulist views, not only about vulnerabilities in your justification to believe some hypotheses, but also about vulnerabilities in transitions or inferences you make. You might be entitled to less confidence in H because of evidential challenges to your move to H from premises that do in fact support it. The challenges we considered testify that such moves are inappropriate even in the ideal. There can also be challenges to your actual performance of the move: that is, (misleading) evidence that you have not reasoned competently in inferring H from E.24 It is debatable what effect such evidence should have, but there is a presumptive case that it should put some kind of pressure on you to be less confident in H on that basis. This is compatible with there also being some pressure on you to believe H on that basis, since the basis does in fact support H. But so long as the first pressure also exists, and is not wholly trumped by the second, we would have a case where your confidence in H was undermined to some degree. Now, in order for you to have the justification you originally did for H, did you need, as a constituent, antecedent justification to believe you were inferring H competently? 25 Many epistemologists will be reluctant to think so. That would make considerations about your epistemic biography essential parts of bodies of evidence where we would not have expected to find them. There would then be no wholly mathematical or a priori justification, since evidence about your competent performance would have to be part of the story too, before anything got to be justified.26 Nor could there be any wholly historical justification. But for the sake of argument, suppose we concede that. Let it be agreed that any time you believe H on the basis of E, part of what justifies the belief in H

22 Williamson (2011) argues that knowing P is compatible with the epistemic probability that you know P being arbitrarily low. See also Williamson ms. 23 See here also the end of Kotzen (2012). 24 Willenken (2011) focuses on defeaters of this sort. 25 This seems to me different from, and much less plausible than, Inferential Internalism. 26 Silins (forthcoming) also mentions this threat.

102

Foundations of Dogmatism

must be antecedent justification to believe you’re performing that inference competently. Then consider instead what should be the effect of (misleading) evidence that you are not properly taking account of the evidence of your competent performance in believing H on the basis you do. Again, it is debatable what effect such evidence should have, but there is a presumptive case that it should put some kind of pressure on you to be less confident in H on that basis. Now was that something you needed antecedent justification to disbelieve, as part of your original justification for H? If so, we can keep going. The credulist thinks that at some point, we will find some way in which your original justification is vulnerable to being undermined, without your original justification needing to include antecedent justification that that vulnerability isn’t realized. I can see only two non-credulist alternatives. One says: No, you really did need a tower of antecedent justification as part of your original support. And every underminable part of that tower needs to be supported by its own antecedently justified tower, and so on ad infinitum.27 The other alternative is a specific kind of “externalism.” At some point— perhaps at the very first step—it would say that if you are in fact doing everything properly, then evidence that you’re not should have no defeating effect. Moreover, it wouldn’t say this just once, but everywhere the prospect of a notantecedently-ruled-out vulnerability arose. Don’t worry, this view would say. You don’t need an infinite tower of antecedent justification to be justified in believing H. But that’s not because the credulist is right, and some vulnerabilities don’t need to be antecedently ruled out. Rather, it’s because these are not really vulnerabilities of the operative sort. If in fact you reasoned properly in such-and-such a respect, your confidence shouldn’t be threatened by evidence that you didn’t.28 Both of these alternatives are intelligible. But they seem to be minority views among epistemologists I interact with. I expect that most epistemologists will be reluctant to hold them, and will instead turn out to be credulists.29 To summarize:  Dogmatists think some vulnerabilities don’t need to be antecedently ruled out, because the justification that’s vulnerable is immediate, and doesn’t include any antecedent justification. That’s one way to be a

27 This couldn’t even be given a coherentist spin, because the view we’re envisaging says that every underminable part of a tower is supported by antecedent justification to believe other things. This is a notion of epistemic not temporal priority, but presumably any notion of priority will exclude cycles. 28 This is reminiscent of the view of disagreement defended in Kelly (2005); though Kelly himself explicitly refrained from claiming that higher-order evidence has no defeating effect (see his section 6). And he argues against that claim in Kelly (2010, section 4). 29 The view that Willenken (2011) calls “liberalism” is roughly equivalent to credulism (modulo the issue mentioned in note 17). The “defeasibility argument” that Silins (forthcoming) considers is, in one of its initial variations, an argument that all underminers are quotidian (and so against credulism). Silins criticizes both that argument and some more restricted variations.

Problems for Credulism

103

credulist. Other ways to be credulist say that, even where the justification in question requires some other antecedent justification, it doesn’t require antecedent justification against all of the ways—perhaps arbitrarily higher-order ways—in which it may be undermined. I have no master argument that it’s impossible to be a credulist and also a Bayesian. However, as we’ll see, there are widespread assumptions about how to interpret aspects of the Bayesian formalism—what their philosophical “cash-value” is—that seem to get in the way of doing so.

5. Some Assumptions and Ambitions The first assumption to consider is: Assumption-1 Getting more justification for H coincides with raising H’s probability.30

This has been challenged. Geoff Pynn imagines that you start out with higher credence in H than you should have. Say you should have a credence of 0.5, but in fact you have a credence of 0.9. Then you acquire evidence in the light of which you should have a credence of 0.6. At this point, should your credence go up? Well, not up from 0.9. It should go down, because it was too high to start with.31 There are interesting issues here about the relation between the attitudes it would be normatively correct for you to have, and the attitudes you actually do have. As I discuss elsewhere, I do not think those facts are insulated from each other.32 However, in the present context, we are understanding talk of your “credences” and your “probability function” to be confined to the attitudes it is reasonable for you to have, prospectively or ex ante, regardless of what attitudes you do in fact have. So in the given example, your epistemic probability for H—the credence your epistemic position calls for—does go up, from 0.5 to 0.6, regardless of your actual doxastic mistake. A different concern with Assumption-1 is that our informal notion of justification is in the first place rooted in thoughts about prima facie support, whereas the Bayesian works with the resultant sum of many different justificatory pressures. For example, you may initially estimate the probability that the

30 White (2006, note 10) points out that his argument against dogmatism needs only a more cautious version of this, which says that getting more justification for H is incompatible with lowering H’s probability. That is true, but I have never heard any challenges to the bolder assumption that would not also bear against the more cautious one, so I’ll just discuss the bolder, simpler assumption. 31 Pynn (forthcoming, in section 3). See also Enoch (2010, section 5) and the example Christensen (2011) calls “Wrong and Wronger.” (The name is Christensen’s, but the case comes from Kelly (2010).) 32 See Pryor (2004, 2012, and ms.).

104

Foundations of Dogmatism

notorious pet Precious is a bird at 0.6, and that it’s a penguin at a minuscule 0.01. Peter then testifies to you that Precious is in fact a penguin. You may trust Peter enough that the probability Precious is a penguin climbs substantially, say to 0.4. But the prospect of a pet penguin strikes you as so odd that you also suspect Peter may have seen some other, non-bird animal that he confused with a penguin. So you’re somewhat less sure now that Precious is a bird— perhaps only 0.5. You acquired justification to believe Precious is a penguin. Did you fail to acquire justification to believe it’s a bird? I’m reluctant to say so. I’d rather say, Yes you did acquire some justification to believe that, it’s just that all-things-considered, you’re left in a position where it would be, on balance, reasonable for you to be less confident that Precious is a bird than it was before. Our simple informal epistemic notions more closely track the prima facie notions. It takes more words to describe the all-things-considered facts. I explained dogmatist views as in the first place concerned with what’s required to have prima facie justification. Still, it would be a real disappointment to hear that, despite our experiences giving us prima facie immediate justification to believe our environments are as they seem, we’re never allthings-considered justified in being more confident of that than we are that we’re in some skeptical scenario. So dogmatists will want to say that their prima facie justification can in some cases amount to ultima facie justification, too. They won’t just turn their back on the Bayesian and say, we’re talking about different epistemic notions. But it is worth keeping in mind that the theories aren’t directly working with the same epistemic quantities. This point will come up again later. I will not directly contest Assumption-1 any further, though some of what we’ll say below should inform our later assessment of it.33 We’ll work our way to the next assumption by starting with a concern a fan of immediate justification may have about how to represent her view in the Bayesian framework. The Bayesian says that epistemic changes are (when reasonable) always a matter of updating on propositions. But why should we think that acquiring immediate justification in the way our informal theorist believes possible corresponds to that? For example, a core claim of my dogmatist view of perception is that it’s merely having certain experiences that justifies you in believing you have hands—not learning that or getting reason to believe that you do have those experiences. And we can expect dogmatist views generally to have this character. They’ll propose some kind of epistemic situation S, and they’ll say it’s merely being in S, not having justification to believe you are, that justifies you in believing other things. Wouldn’t modeling such epistemic changes in terms of updating on a proposition distort what the dogmatist thinks is going on?

33

For further discussion, see Achinstein (2001, chapter 4), and Kung (2010).

Problems for Credulism

105

This is not a simple issue. I’ll sketch a first response now, but shortly we’ll see this worry re-arise in a different form. The first response says: That’s OK, this much of what the dogmatist thinks is going on can be reconciled with the Bayesian model. For on these dogmatist views, there will be some first propositions that getting into S makes it reasonable for you to believe. So we can say it’s those propositions that you reasonably update on. That is compatible with our having an extra-formal, dogmatist story about why you’re in a position to update on it. All right. But let’s think more carefully about which propositions it is that you update on. The response we just heard suggested: Assumption-2 What you update on coincides with what you’re immediately justified in believing.

This has two directions: first, that your probability function updates on E only when you’ve acquired immediate justification for E.34 The other direction is that changes in your credence in H that aren’t themselves updates on H but are rather a function of your updating on something else cannot represent acquirings of immediate justification to believe H, but only the acquirings of mediate justification. If a theorist doubts the existence of any immediate justification at all, she might uphold the second of these without the first. Assumption-2 looks natural, but it should not be assumed uncritically. Recall that immediate justification is in the first place defined for prima facie justificatory contributions, whereas what you update on will be the ultima facie, net sum of different such pressures. So what you update on might not be any proposition you then acquired prima facie justification for, much less immediate justification. Moreover, alternatives to Assumption-2 are intelligible. On some views, our probability functions just simply evolve, and facts about what propositions we’ve “updated on” are reconstructions from that evolution. We are not given as independent facts that now you should update on this proposition, and now on that. On such a view, I see no reason to expect we could read off facts about immediate justification from what propositions you’ve updated on. (On the other hand, I’m not sure such views are hospitable to the possibility of immediate justification, in the first place.) Another view says that at a given time you become entitled to certainty that you’re having hand-like experiences, and reasonable confidence (but not certainty) that you have hands, and the second of these does not epistemically depend on your justification to believe the first. You become immediately

34 Remember, these “updates” concern how the credences you’re justified in having change. Of course, unreasonable subjects may as a matter of descriptive psychology change their doxastic attitudes in ways that don’t correspond to any justification they’ve acquired.

106

Foundations of Dogmatism

justified in believing each. (Since we’re credulists, we may allow that evidence against the first would tend to undermine your justification for the second.) This view seems coherent. Assumption-2, however, implies that it’s only the claim about your experiences that you could have acquired immediate justification for. If we’re going to engage with dogmatists, then we should be able to at least try to represent justification that’s both immediate and underminable. But if it’s immediate, then Assumption-2 tells us you update on it. And in the Classical Bayesian framework, when you update on something it becomes maximally justified and you can never take it back. No future evidence can defeat it. When we turn to Jeffrey Bayesianism, later, the formalism will permit you to update on hypotheses without becoming certain of them. Perhaps that will prove more hospitable to the dogmatist. But is there any way to make Classical Bayesianism (at least initially) more accommodating of what the dogmatist thinks happens? Perhaps we should just give up Assumption-2. We could suppose that updating on the claim that you have hand-like experiences is compatible with your having acquired immediate justification to believe bolder claims. We will see White take this approach. An advantage of such views is that it will be possible for the subject to recognize she’s in a situation that gives her immediate justification to believe H, without thereby committing herself to H, because it has been undermined. Alternatively, we might try appealing to “Popperian” conditional probabilities, which are defined even for conditions that have an unconditional probability of 0. So it could be that your New(H) is 1, yet you still have well-defined, non-trivial values for New(.|U), where U is incompatible with H. This is not the domain where these tools are usually used, but it may be a viable application of them. On the face of it, this would permit us to update on the immediately justified propositions, but allow them later to be defeated. But how should we think about cases where you acquired the defeating evidence first, before your immediate justification for H? Intuitively, that shouldn’t make a difference to what your final epistemic position is. We will return to this intuition later, and discuss it more carefully. For the time being, assume it’s right. But if, when you get the defeating evidence second, the credence you’re justified in having in H goes lower than 1, then in the case where you get the defeating evidence first and the immediate justification for H second, you shouldn’t there be Classically updating on H.  Your credence in H should in that case not go to 1. Yet H is what you acquired immediate justification to believe—it was just preemptively defeated. So even if we want to use “Popperian” conditional probabilities to let you update on H in some cases, we’ll still end up denying that Assumption-2 always holds. Recall that the Bayesian formalism itself is completely silent about when subjects should update on which propositions. Plantinga tells a useful story

Problems for Credulism

107

of a mountain climber whose beliefs “freeze” at a certain moment, though his sensory experiences continue to evolve like everyone else’s. His friends carry him home, but he continues to believe he’s sitting on a ledge watching a hawk glide below him. Yet he’s not hallucinating: he has exactly the perceptual phenomenology any of us would have, at home in his bed. It’s only his beliefs—including his beliefs about his experiences—that have frozen; not the experiences themselves.35 Plantinga’s point is that if the subject’s beliefs were coherent just before the freeze, then so far as their internal relations go, they’ll continue so—as long as no new beliefs are added or removed. But surely this subject manifests some kind of epistemic defect. The defect is in the correlation between what’s happening in the world and how his beliefs are updating. This is a matter that can go epistemically better or worse. The Bayesian formalism doesn’t itself say anything about it. Instead, it just starts from the point where it’s given that the subject now does (or should) update on such-and-such. By itself, that’s no weakness in the formalism. It’s just one of the things the formalism doesn’t try to explain. And against that background, we can understand the view sketched a moment ago like this: when the subject has hand-like experiences and reasonably low credence in undermining hypotheses, then it’s epistemically appropriate for her to update on bolder claims, like the claim that she has a hand (or perhaps the claim that she sees herself to have a hand). When she has the same experiences, but reasonably higher credences in undermining hypotheses, she should instead update on more cautious claims about hand-like experiences. So far, nothing here conflicts with Bayesianism. It just goes further than Bayesianism does and says something about when the world makes one update appropriate rather than another—a matter about which Bayesianism itself is silent. Jeffrey Bayesianism promises to be an even more hospitable setting for this strategy, too, because it can model partly updating in the bold way and partly updating in the cautious way. But despite the crudeness of the present version, it should still be intelligible. As I said, this view gives up on Assumption-2, for even in cases where the subject has high credence in the underminers, she still acquires immediate justification to believe the bold claim. It’s just that that justification is preemptively defeated. Yet on this view, in such cases the bold claim is not what the subject updates on. This view also gives up the idea that there is a single proposition P such that acquiring a given piece of immediate justification always coincides with updating on P. So the worry we voiced earlier would be vindicated: acquiring that immediate justification isn’t correctly modeled by any one update.

35

Plantinga (1993, 82). See also Feldman (2003, 68) and Christensen (1992, note 1).

108

Foundations of Dogmatism

Third, this view abandons the following idea: Assumption-3 The negative effect of undermining evidence, even when it purports to be immediate justification that’s undermined, should be represented inside our formal models, rather than “off-stage,” in terms of when the world supplies the model one input rather than another.

Though I call this an “Assumption,” it is really more of an explanatory ambition. And it’s quite a strong ambition. Recall that the examples of introspective and logical opacity we discussed before are already generally thought to be places where it fails. Those involved epistemic effects that these formal models are already acknowledged not to represent. I think that abandoning Assumption-3 for non-quotidian underminers is in fact a theoretically fruitful strategy. But for this discussion, we will try to see how far we can get without abandoning it.36 I will identify three other assumptions when they become relevant in the following discussion .

6. Threats from Classical Bayesianism Now we’re ready to assess the alleged Bayesian problems for my dogmatist view of perception. As we’ll see, if there are problems here they aren’t specific to perception—and they may not be specific to dogmatism either, but may be issues that all credulists need to sort out. Additionally, any problems to be found here depend on one’s making certain choices about the interpretive assumptions. We’ve already seen a range of options for what a dogmatist might say you update on, when you acquire immediate justification to believe H. White and other Bayesian complainers against dogmatism rely on a specific one of those options. We’ll soon encounter yet another interpretive assumption that is essential to their complaints. White’s discussion contains several threads. One of these concerns “bootstrapping”; I will not attempt to sort those issues out here. A second concerns justification or suppositions about what justification you will have in the future; we will take these issues up in the next section. A third thread concerns “Moorean” arguments like the following: E Hand

36

I am having experiences as of hands. So I have a hand. If I have a hand, I am not a handless brain in a vat being fed illusory experiences as of hands.

Section 4 of Christensen (1992) emphasizes the costs of abandoning this Assumption.

Problems for Credulism

109

Thus, Good

I am not a handless brain in a vat being fed illusory experiences as of hands.

It will be convenient to have a short label for the negation of Good, which I’ll dub: Bad

I am a handless brain in a vat being fed illusory experiences as of hands.

White claims that by Bayesian lights, having experiences as of hands should make it more likely, not less, that you are a handless brain in a vat having such experiences. For the brain in a vat hypothesis we’re considering, Bad, entails that you would have those experiences. Hence, when you update on E, the probability of Bad should go up, and the probability of its negation Good should go down. It would be strange to say that such reasoning increases or contributes to the justification you have to believe Good, when the justification you acquire for the premises makes the probability of Good go down.37 White is here criticizing those of “Moorean” sympathies, who think this kind of reasoning can contribute to one’s justification to believe Good. I am such a philosopher. However, my project today is to speak on behalf of all dogmatists and credulists, not just those who share some specific and controversial views with me. So I won’t speak to this complaint directly. Note that White is assuming here that acquiring the immediate justification my view concerns itself with—which happens when you have hand-like experiences—should be modeled by your updating not on Hand but on E. Hence, White is rejecting what we called Assumption-2. He’s supposing that a charitable representation of what the dogmatist thinks happens can involve you updating on something other than what you (allegedly) acquire immediate justification for. If he’s not to have already begged the question against the dogmatist in modeling things this way, then the mere fact that your credence in Hand goes up as a function of your updating on something else shouldn’t be understood to already mean that the justification you acquired for Hand was epistemically posterior to justification you acquired to believe something else, and hence non-immediate. Attend to this point well, for it will bear on what happens later. If we thought that modeling the situation in this way did mean that the justification you acquired for Hand was epistemically posterior to something else—as we would if we held Assumption-2—then White’s “argument” against the dogmatist would already here be concluded. It would consist in his assertion that when you have hand-like experiences, you don’t get immediate justification to believe Hand. All the subsequent details about probability would be irrelevant.

37

See note 30.

110

Foundations of Dogmatism

So in charity to White, we assume he is charitable to his target and does not take this aspect of the model to have that significance. Later, though, we’ll see he seems to be of two minds about this. Here is what White himself says to justify modeling the case this way. His idea is that he’s focusing on subjects who not only have hand-like experiences but are also reflectively aware of having them. These subjects are better informed about their own epistemic situation. Surely that should not make them worse off with respect to whether they have hands, right? True, philosophers have sometimes told stories where knowing more can make you epistemically worse off overall; but there’s no apparent reason to think these cases are like that. So if we find constraints on how justified even these betterinformed subjects can be, we should expect those constraints to apply with at least as much force to less-informed subjects, too.38 White’s strategy will be to identify such a constraint. In fact, given the specific arguments he wants to present, White’s hands were sort of tied to proceed in this way. His arguments posit a skeptical hypothesis U that both entails the proposition you update on, but is incompatible with the proposition it undermines. That posit could never succeed in a case where what gets undermined is the same proposition you update on. So White’s presentation seems to require modeling the dogmatist’s idea that your immediate justification is underminable in a way that doesn’t involve you updating on the proposition you’re acquiring immediate justification for.39 We’ve mentioned and set aside three of the threads in White’s discussion. A fourth thread is what we will focus on. This thread can also be presented using the argument from E to Good; but it targets all dogmatists about perception, regardless of their attitude toward such “Moorean” reasoning.

38 See White (2006, 534–35); see also section 3 of Wedgwood (forthcoming) and Silins (2008, note 22). Silins raises some worries in his forthcoming work (note 30 and preceding text). I’ve been disposed to go along with White here; his strategy seemed to me a reasonable one. But recently David Barnett has persuaded me that the issues here are not so straightforward. White’s proposal encourages the idea that the subjects are just coming to discern facts about the epistemic situation that were already in place. It’s as though we asked them “How many US States have names beginning with the letter M?” The relevant information is already there in their mind; reflection just has to locate it. Contrast a case where we ask the subject “How many US States remind you of your grandmother?” Here too, they may be able to answer the question without leaving the armchair. But most subjects would perform mental experiments to answer the question, or engage in other mental activity that intuitively changes their epistemic situation as they proceed. Are the better-informed subjects White focuses on more like the first group of subjects? Or more like the second? That’s not entirely clear. And if they’re more like the second group, then the differences between them and less-informed subjects might not be so harmless. 39 As Maria Lasonen-Aarnio pointed out to me, though, White only uses U’s entailing the proposition you update on in order to secure the result that Old(U|E) > Old (U). He might substitute a different U that retains that property but doesn’t entail the proposition you update on.

Problems for Credulism

111

This criticism turns on the claim that your new probability for Hand—the credence you’re justified in having after updating on E—cannot be higher than your prior probability for Good. The proof of this is not hard. Old(.) is your epistemic probability function before learning E or Hand. New(.) is your epistemic probability function after E becomes true. White has defended his choice of letting New(.) be the result of conditionalizing on E; that is, New(.)  =  Old(.|E). We note that observations that are entailed by a hypothesis, as E is entailed by Bad, contribute positively to the probability of that hypothesis. That is, Old(Bad|E) will be greater than Old(Bad) when E is not yet itself epistemically certain. This is equivalent to: Old(not-Bad|E) < Old(not-Bad).

Next we observe that, since you have updated on E: New(not-Bad) = Old(not-Bad|E).

And further, we observe that Bad is incompatible with Hand. In other words, Hand entails not-Bad and its probability can be no higher than not-Bad’s: New(Hand) ≤ New(not-Bad).

Putting these three equations together, we have: New(Hand) ≤ New(not-Bad) = Old(not-Bad|E) < Old(not-Bad).

Ignoring the middle terms, we have: New(Hand) < Old(not-Bad).

So Bayesianism tells us that your new probability for Hand cannot be higher than your prior probability for not-Bad, that is, your prior probability for Good. Though White doesn’t argue it, an even stronger formal result is possible. We can show, not merely that New(Hand) < Old(not-Bad), but also that New(Hand) < Old(E⊃Hand), where the latter quantity may be even lower than Old(not-Bad). So this is a constraint that will bind at least as tightly as the original, and sometimes more tightly.40 One can see from the way these results were derived that nothing here is specific to perception. Any case where one thinks there is underminable immediate 40 Proof:  Old(E⊃Hand)  =  Old(not-E) + Old(EшHand)  =  Old(not-E) + Old(Hand|E) Old(E)  =  Old(not-E) + Old(Hand|E)(1-Old(not-E))  =  Old(not-E)(1-Old(Hand|E)) + Old(Hand|E) = Old(not-E)Old(not-Hand|E) + Old(Hand|E). When Old(not-E) and Old(not-Hand|E) are each > 0, this will be > Old(Hand|E), which is New(Hand). So in those circumstances, New(Hand) < Old(E⊃Hand). Finally, since Bad entails that E and not-Hand, Old(E⊃Hand) ≤ Old(not-Bad). It is more difficult to prove in the Jeffrey framework that New(Hand) < Old(E⊃Hand) ≤ Old(notBad); however, the same result does also hold there. Bad in these examples is generally assumed to entail that I have the specific experiences I do (though see note 39). Presumably the prior probability of my having specifically experiences E will be extremely low; so not-Bad’s prior probability will be extremely high. As David Christensen reminded me, though, it also follows that the prior probability of E⊃anything will be extremely high. So Old(E⊃Hand) shouldn’t be expected to usually be much lower than Old(not-Bad).

112

Foundations of Dogmatism

justification should display the same structure. For example, perhaps I have introspective justification to believe I intend to confront my father about something. Then I read a psychological study that says subjects are unreliable about whether they even intend, as opposed to merely fantasize, such things. I  presume that would undermine my introspective justification. And reasoning just like White’s would seem to establish the same inequalities. Arguably even my justification to believe E would also have this structure. What is the significance of the inequalities we’ve established? Here is how White interprets them:41 So its appearing to me that this is a hand can render me justifiably confident that it is a hand, only if I am already [justifiably] confident that it is not a fake-hand. (p. 534)

In taking this to tell against dogmatism, White commits himself to what I’ll call: Assumption-4 If you already need to be justified to a certain degree in believing A, in order to acquire some new quantity of justification to believe B, then it’s false to say that your new justification for believing B is immediate. It did after all need to include at least that degree of antecedent justification to believe A.

To sloganize it, we might think of this as the idea that prior probabilities have epistemic antecedence.42 This idea promises to be as problematic for credulism more generally as it is for dogmatism in particular. Any attempt to model non-quotidian 41 White clearly intends the inserted “justifiably.” If we genuinely do consider the effects of mere, possibly unjustified, confidence that Good, this introduces new issues. See the papers cited in note 32 for discussion. 42 Assumption-4 makes two steps, one signaled by the word “include” and the other by the word “antecedent.” The first of these steps was mentioned in note 17. Silins (2008, section 4) argues that this step is not mandatory, and I agree. The second step takes the justification represented by prior probabilities to be “antecedent” in the sense we’re working with—regardless of whether it’s an included part or merely a necessary condition of the justification represented by your posterior probabilities. This is what I will focus on. Willenken (2011) calls Assumption-4 “Auxiliary Thesis about Independent Justification” and argues against one variation of it. The notion of justification I’m working with is sensitive to his “reasoningdirected defeaters,” and so is what he engages with under the heading of “robust” views. However, because of the issues discussed in the papers cited in note 32, I officially have what he defines as an “anemic” view of justification. Some care is needed in attributing Assumption-4 to White. What’s clear is that (i) his paper nowhere attends to, and is sometimes insensitive to, the difference between its being a necessary condition that you have some prior probability, and your justification including antecedent justification to believe something else. It’s also clear that (ii) the theories he’s taking his formal result to arbitrate between are defined by their proponents in terms of the second notion, so something like Assumption-4 is needed to make the formal result relevant. After that, things are less clear: (iii) the language in which White himself defines the theories he’s considering can in places be read as invoking the second notions; but it can also be read just in terms of the first notions, and White’s paper strongly suggests that this is his intent. (See esp. his note 14; and he has confirmed this in discussion.) So the best way to understand him may be as

Problems for Credulism

113

undermining in the Bayesian framework is going to face structural choices like the ones we’ve been discussing. It’s going to be natural to (try to) model cases with non-quotidian underminers in ways that don’t look any formally different from how we model quotidian undermining. For example, suppose you believe H on the basis of E, and “E seems not to support H” is a non-quotidian underminer. One expects a Bayesian model of this to be such that Prob(H) ≤ Prob(H|E шE seems not to support H) < Prob(H|E)

In just the same way that Prob(It will rain) ≤ Prob(It will rain|This barometer is falling but is unreliable) < Prob(It will rain|This barometer is falling)

If we then subscribe to Assumption-4, we’ll interpret these structural facts as implying that, in fact, you need justification for “E seems to support H” epistemically antecedent to any justification for H. So this case turns out to be not one of non-quotidian undermining after all. Surely Assumption-4 is philosophically optional. If it’s going to be possible to model non-quotidian undermining in Bayesian terms, we must do so without that Assumption in place. So why should a dogmatist or credulist agree to it? And we can do better than complain that Assumption-4 is optional. We can argue that White has himself already presumptively relied on its falsity; and we can show that in the general form we’ve stated it, it has intolerable consequences. In what sense has White “presumptively relied” on Assumption-4 being false? I invite you to compare it to Assumption-2, which said that the immediately justified propositions are what you update on. These two ideas seem to naturally go together. Yet White rejects Assumption-2—and as we saw, he needs to do so in order for his model of the situation to be charitable to the dogmatist. Remember: if all the business about probabilities is to do any work—if there’s to be more to his argument than the bare assertion that when you have hand-like experiences, you don’t get immediate justification to believe Hand—then the mere fact that Hand goes up as a function of your updating on something else shouldn’t be understood to mean that the justification you acquired for Hand was epistemically posterior to anything else. So what White has relied on is that the functional dependencies in

arguing for a position such as we’ll entertain in the next section. He just incorrectly takes that position to be incompatible with dogmatism, and to suffice for the kind of view espoused in Wright 2002. For two reasons, I won’t keep re-acknowledging this. Instead I will proceed as though White’s criticisms are intended to oppose dogmatism as I understand it, and so he really does commit himself to Assumption-4. The first reason is that many of his readers have understood him that way and sometimes endorsed what they so understood. The second reason is to keep our dialectic more manageable. It is instructive to figure out how the criticism I’ll treat White as advancing fares. The biographical facts about what he or I ever intended are of less interest.

114

Foundations of Dogmatism

the formalism can’t in general be assumed to mirror the facts about epistemic dependence and priority. Assumption-2 and Assumption-4 turn on different functional dependencies. But if the dependencies of the latter justify an interpretation that those of the first do not, this needs special motivation. And when you think about, it doesn’t seem like Assumption-4 can be true, not in the general form we’ve stated it. The implications are much too strong. Let H be any hypothesis whose probability goes up when you learn that E. Let H* be any old logical implication of H. Now it’s straightforward that New (H) ≤ New (H*) = Old (H* | E)

Moreover, whenever Old(E) and Old(H*|E) are both less than 1, then Old(E⊃H*) will be strictly greater than Old(H*|E).43 So we can add New (H) ≤ New (H*) = Old (H* | E) < Old (E ⊃ H*)

That is, you can’t acquire justification to believe H above a given threshold, in response to some evidence E, unless you already had at least that high a prior probability that E⊃H*—for any implication H* of H.  Do we really want to conclude that the justification you acquired for H was partly constituted by or came from antecedent justification to believe each such conditional? 44 Rejecting that interpretation of the inequality New(H) < Old(E⊃H*) means rejecting Assumption-4. Perhaps some more restricted version can fare better, but we’ll have to wait and assess it when we see it. Without some such assumption, the formal results we derived don’t directly threaten anything a dogmatist or credulist wants to say. In the next section, I’ll spell out some ways of thinking that develop the idea of prior probabilities not having epistemic antecedence in the way that Assumption-4 envisages.

7. Life without Assumption-4 What might explain an inequality like New(H) < Old(E⊃ H*), or New(Hand) < Old(not-Bad), if not the fact that the justification you acquire for the former proposition is partly constituted by the antecedent justification you have for the latter? Consider what Boghossian 2000 says about our epistemic relation to modus ponens. We may well be justified a priori in believing Valid

43 44

If P, and P⊃Q, are true, then Q must be true.

See note 40. Thanks to Philip Ebert for stressing the awkwardness of this to me.

Problems for Credulism

115

However our justification for applying the rule of modus ponens in a deduction doesn’t come from our justification to believe Valid. Rather our justification to believe Valid relies instead on the reasonability of our applying the rule. There’s an important question of epistemic priority here that isn’t settled by the fact that our justification for Valid is a priori. (See also Boghossian (2001, 2003).) We don’t need to assume that Boghossian’s story about that is right; nor that, even if it is right, the same story extends to our present concerns. I merely ask you to attend to the epistemic structure he posits between our justification to believe Valid and our justification to reason by modus ponens. A similar structure could obtain in the case of perception, or anywhere else we have immediate justification. That is, even if belief in E ⊃H is a priori, our justification for believing H in response to E needn’t come from our justification to believe that conditional. The epistemic priority may in fact be reverse. Several philosophers have thought that wherever a transition from some evidence E to H is epistemically legitimate, the conditional E⊃H must be a priori justifiable.45 A dogmatist needn’t oppose this. Whether it conflicts with anything he says depends on whether it’s the justification of the conditional that underwrites the reasonableness of the transition. It may be the other way around. (Or maybe neither underwrites the other.) Similarly, a credulist needn’t oppose its being a priori that (E шE supports H)⊃H. That only forces him to say his E-based justification for H includes antecedent justification to believe “E supports H” if his justification for H is epistemically posterior to his justification for this conditional. It need not be. Just because one claim is a priori and another isn’t, it doesn’t follow that the second is epistemically posterior to the first—that your justification for it needs to include antecedent justification for the first. Not even when the

45 See BonJour (1998, section 7.7) and Hawthorne (2002). This is an idea with a longer history, though I’m not sure how much longer. Chisholm held that something of the form “If . . . then E⊃I am justified in believing H” was synthetic a priori (see Chisholm,1989, 72–73). See also van Fraassen (1989, section 6.3). I heard the epistemic descent from “I am justified in believing H” to H made several times in the 1990s. The formalisms we’re working with require a proposition to have some defined initial probability if it’s ever to have any posterior probability. Philosophers who interpret the probabilities as representing justification, as we are, often interpret the initial probabilities as representing what you have a priori justification to believe. I won’t resist this; though I will point out a different interpretation of “a priori” later. But in fact I have serious doubts about this practice. The formalism requires initial probabilities for E, H, and E⊃H. Am I sure that I had a priori justification to have any particular confidence in these claims?—for example, the claim E, that I’d have just these experiences now? This is an extraordinary claim. Perhaps there may be ways to understand the probabilities as representing facts about justification, without interpreting the initial probabilities that way. If Assumption-4 is optional, this may well be optional too. (Kung (2010, 12) also suggests not interpreting a high prior in Good to mean you have any justification to believe Good.) If it’s not optional, then we should at least want to move to versions of these formalisms that can represent probabilities as initially very imprecise; even better would be ones that permit probabilities to be initially undefined. Claims like Assumption-4 would then be even less straightforward.

116

Foundations of Dogmatism

claims’ contents are relevant to each other. There may be routes to the second that proceed by way of the first; but there may also be routes that don’t. The dogmatists and credulists think that the reasonableness of your move to H doesn’t derive from antecedent justification to believe the conditionals we mentioned. When you learn E, you don’t need to rely, not even implicitly, on an application of modus ponens. Doing so would be a mistake akin to Achilles’ mistake in agreeing that Valid is another premise he needs to argue that Q. Cohen and Wedgwood have recently argued for the a priority of E⊃H as a result of using non-deductively good reasoning inside a conditional proof.46 Their idea is that if it’s prima facie reasonable to conclude you have hands when you really do have experiences as of hands, then it’s also prima facie reasonable to conclude you have hands under the supposition that you have experiences as of hands. And then you can discharge the supposition to infer that You have experiences as of hands ⊃ you do have hands.

It’s not obvious that those epistemic transitions are legitimate. The epistemic effects of having some experiences needn’t be the same as, or even inherited by, the proposition that you have them. Or there may be general reasons that we can’t expect arbitrary good reasoning moves to also be licensed in suppositional contexts. Certainly some reasoning moves that have been claimed to be good are not so licensed. Alex Byrne claims that the transition from P to I believe P is justificatory (Byrne 2005, 2008). Perhaps he’s wrong, or perhaps not. We should agree, though, that under the supposition P I should not be entitled to conclude that I believe P, and then use a form of conditional proof to infer that P⊃I believe P.  As David Barnett points out, if I  could do that I should also be entitled to argue in the same way that not-P⊃I believe not-P. For any P at all. And the conjunction of all those conditionals looks like the claim that I am omniscient.47 Moreover, Weatherson 2012 gives a compelling case where a very weak inductive rule looks to be illegitimate in a suppositional context. So it’s not clear to me that the Cohen/Wedgwood strategy is correct. But some other story about why we’re a priori justified in believing E⊃H may be. Such stories need not make the reasonableness of moving from E to H epistemically dependent on our justification for the conditional. The story might make that justification parasitic, in some other way, on the reasonableness of moving directly from E to H. (Or maybe neither is epistemically posterior to the other.)

46 See Cohen (2010) and Wedgwood (forthcoming). White (2006, section 6) anticipates their proposals in some ways. 47 Wedgwood acknowledges this specific limitation; see his note 13.

Problems for Credulism

117

Juan Comesaña discusses the possibility of combining such stories with a commitment to immediate justification.48 He says: [T]his position is not very stable. . . . [W]henever you are justified in believing H on the basis of E you will have available to you a different justification— one which depends not only on E but also on the conditional if E then H.

Comesaña complains: Of course, the friend of immediate justification is free to hold that even though you have justification for believing this conditional, your justification for believing the consequent “comes from” just the antecedent. . . . What determines whether your [propositional] justification for believing H comes from (in part) the conditional or not? Perhaps one could try to argue for the claim that it doesn’t by saying that you would still be justified in believing H even if you were not justified in believing in the conditional. But remember that, according to the neo-rationalist, justification for the conditional will be available whenever the antecedent justifies the consequent.

But it seems readily imaginable for you to have the immediate justification but lack justification for the conditional: they may have different defeasibility profiles. Let your philosophical mentor give you a mistaken but compelling argument against the conditional, which allegedly doesn’t threaten the immediate justification. As we said in section 4, what effect such evidence should have on your other justification is controversial. But a natural view is that it would drag the justification you have for the conditional in a defeated direction, and need not defeat the immediate justification you have from E to the same degree.49 So the combination of belief in the a priority of E⊃H, and continued commitment to immediate justification, looks entirely feasible to me. The a priority of some such conditional doesn’t establish its epistemic antecedence over anything. At best, it only establishes its lack of epistemic posteriority on certain kinds of evidence. Our notion of “a priority” may have an indeterminacy that blinds us here. Since the 1980s, it has come to be widely appreciated that some experiences may be necessary to have a belief, without threatening that belief ’s claim to be a priori.50 For example, maybe no one can have the concept of umami who’s never had taste experiences. But the beliefs that umami things are umami, or that integers are not umami, are presumably a priori. Experiences are necessary to have these beliefs, but don’t contribute to their justification. (Other times, experiences that are necessary to have beliefs can contribute to their

48

Comesaña (forthcoming); I’ve changed the labeling in the quotations. See also my 2000, note 6. 50 Kant also emphasized this; but the point was long unappreciated in the twentieth century. 49

118

Foundations of Dogmatism

justification.) What the preceding reflections suggest is a converse possibility, that experiences may play a role in the justification of some beliefs even without ever being had. You may be justified in believing E⊃H in part because of the justificatory power of experiences E—even though you don’t now have E. Perhaps you can be so justified without ever having had E but just by reflecting on what it would be epistemically like were you to have E. Does that make your justification for the conditional E⊃H a priori? I feel pulled in two directions here, and that suggests that my concept of a priority can be extended in either direction. We might say justification is “a priori” if it can be had in advance of having any experiences.51 So understood, our justification to believe E⊃ H would be a priori, because it can be had in advance of ever having experiences E. Despite that, though, it may still be the epistemic powers of E that justify the conditional, rather than your justification for the conditional justifying you in concluding H when you really do come to have E. Alternatively, we could use “a priority” to track whether the epistemic powers of any experiences or experiential capacities contribute to the justification of your belief. Understood in that way, your justification for these conditionals is not a priori, even if it could be in place before you’ve ever had any experiences, and so even if it could be reflected in your initial probability function. It’s not a priori because it’s parasitic on what the experiences can justify you in believing, when you do have them.52 In sum, I  think it’s open to discussion whether prior probabilities in the Bayesian sense should be understood either as a priori or as having any epistemic antecedence. White and others have been assuming that the debate between dogmatists and their opponents boils down to what our prior probabilities have to be. Whether it does so or not depends on how we settle some of these open questions. It may for all I know be the best course for a dogmatist to agree with White about the prior probabilities, but to wrestle about what their philosophical significance is. This may also be an open avenue for credulists who aren’t dogmatists. They may agree to attribute high prior probability to claims like “E supports H,” and “I am a Competent Reasoner,” without your justification for those claims yet playing a constitutive role in the justification of the inferences you make.

51 Though we shouldn’t also make that a necessary condition, because we still want “Integers are not umami” to be justified a priori, and maybe I can’t have that justification until I have the concept of umami, which may require having experiences. 52 Yablo (2002, sections 12 and 15) argues that imaginative judgments about ovals fail to be a priori even though they don’t evidentially rely on any sense-experiences. (Neither does it seem right to say they evidentially rely on the deliverances of introspection.) Williamson (2007, 165 ff.) argues similarly about imaginative comparisons of inches to centimeters. I understand them to be working with something in the vicinity of this second notion of “a priority”—though we may also want to make finer-grained distinctions.

Problems for Credulism

119

8. Jeffrey Bayesianism We’ve looked at some philosophical assumptions behind the Bayesian-based objections to dogmatism. We’ve seen that these assumptions are not mandatory. However, neither is it really straightforward how to model the possibility of immediate yet underminable justification—especially if we retain Assumption-2, and so think that that needs to involve an underminable update. We only gestured at some ways this might be understood. As we said before, some of the ideas we were gesturing at look like they would be more promisingly developed in a Jeffrey Bayesian framework. So let’s turn our attention there. This section will say a bit to explain how this framework works. The next section will summarize some arguments due to David Christensen and Jonathan Weisberg that suggest that Jeffrey Bayesianism isn’t after all very hospitable to the possibility of underminable updates. These arguments also involve some substantial philosophical assumptions, which I will identify. OK, so what is Jeffrey Bayesianism? Suppose you start off with a probability distribution that looks like figure 5. 2.

E

not E

E&H

not E & H

figure 5.2

Now in the Classical framework, if you update on E, then your probability distribution will change as in figure 5.3. That is, we just erase the not-E part of your old distribution and retain the E part. All the relative relationships within the E part remain the same. That is, when you update on E, then for any proposition H, New (H | E) = Old (H | E)

120

Foundations of Dogmatism

Old

E

New

E&H

not E

E

E&H

not E & H

figure 5.3

This fact is called your update being rigid with respect to E.53 Now Jeffrey Bayesianism shares much of what we just described and has the Classical update as a limiting case. Jeffrey’s transitions look as in figure 5.4. Old

E

not E

New

E&H

not E & H

E

not E

E&H

E

E&H

not E & H

figure 5.4

That is, we don’t just get a copy of the E-part of our old distribution. We also get a copy of the entire old distribution too. Of course, there will be a question of the relative sizes of these two components. That might not always be the same fixed ratio. When specifying an update, it somehow needs to be settled which proposition you update on (in this case, E), and also what scaling factor should be applied to the right-hand component of your new distribution. Let’s suppose the two components are scaled as in figure  5.5, where B is some real value, and d is a normalizing factor so that Old(E)/d + Old(not-E)/d + B*Old(E)/d = 1. In figure 5.5, B looks to be roughly 1. But we can suppose it to be any non-negative real. As B approaches 0, New’s right-hand component becomes relatively much smaller, and it’s as though you didn’t become much more confident of E at all. As B approaches infinity, the right-hand size dwarfs the left-hand, and we approach the situation we had in the Classical case, where you moved just to a copy of the E-part of your original distribution. 53 In the Classical framework, the left-hand term of the equation is equivalent to New(H) simpliciter, but that won’t also be true in the Jeffrey framework.

Problems for Credulism

121

Old

E

New

E&H

E

E&H

Size = Old(E)

not E

not E & H

Size = Old(not-E)

Size = Old(E)/d

not E

E

E&H

Size = B*Old(E)/d

not E & H

Size = Old(not-E)/d

figure 5.5

Perhaps there is some magic single value for B that all updates should use. Or perhaps, as most theorists assume, different learning episodes would make different Bs appropriate. A good look at an object in noon sun may call for updating on the proposition that it’s red, with a high B; a quick look in twilight may call for doing so with a low B. It’s part of the Jeffrey machinery that these transitions are also rigid with respect to E. The two left-hand components of your new distribution are internally just the same as they were in your old distribution; and the right-hand component retains all the internal relationships of Old(.|E), just as we saw in the Classical update. So in the Jeffrey framework, it’s also true that an update on E is rigid with respect to E. This story is simplified in some ways; but it is the essence of what many now understand by “Jeffrey updating.”54 What stands out is that when you update on a proposition, as we did here on E, you needn’t become certain of that proposition, but merely more confident of it. That leaves open the possibility that you might later acquire other evidence that opposes that proposition. Perhaps it might also leave open the possibility that your update on E could be defeated by being undermined. Let’s wait and see. An interesting question is what it would be for someone who started with a different probability distribution to update in the same way as we just illustrated. It’s natural to expect this will involve also updating on the same proposition E. But how big should the scaling factor be, for it to be the same update? In figure 5.6, what does “?” need to be, for this to constitute updating in the same way as before, albeit from the starting point of Old'(.) rather than Old'(.). It turns out that if we want updates to commute, it is necessary and sufficient for us to have the same update that it is an update on the same proposition and 54 The model presented here purports to be exhaustive in a way that Jeffrey would not himself endorse; see note 60.

122

Foundations of Dogmatism

Old

E

New

E&H

Size = Old′(E)

not E

not E & H

Size = Old′(not-E)

E

E&H

Size = Old′(E)/d′

not E

E

E&H

Size = ? * Old′(E)/d′

not E & H

Size = Old′(not-E)/d′

figure 5.6

using the same scaling factor B. This may look natural in the diagrams we have, but that’s because I deliberately explained Jeffrey updating in a way that makes it look so. The result wasn’t at all initially obvious.55 But should we want evidence to commute? Isn’t it sometimes significant which of two pieces of evidence we acquired first? For example, what if we knew in advance that our measuring devices deteriorate rapidly? Then the first use of them is more likely to be reliable. Yes, of course. But that worry is not envisaging two scenarios where we have all the same evidence, acquired in different order. It’s imagining that in one scenario, you have the evidence that you got instrument reading A first, and in the other scenario, evidence that you got instrument reading B first. So those are cases where your total evidence is different. Who would deny that such a difference may sometimes be epistemically relevant? What would be needed for a case that demonstrated an intuitively compelling violation of commutativity would be a case where in each scenario you had all the same evidence, including the same evidence about which evidence you acquired first, but the facts about which evidence you acquired first were different, and that by itself made an intuitive difference to what it was reasonable for you to believe. It’s not clear there could be any such intuitively persuasive case.56

55 For a time, it was widely accepted that Jeffrey updating must be non-commutative. Field (1978) demonstrated that making the update partition and B invariant suffices for commutativity. Wagner (2002) extended Field’s result, and showed that on minimal assumptions this invariance is also necessary for commutativity. See also James Hawthorne (2004). 56 For externalists about evidence, it may prove difficult to even describe such a case, since they allow that such facts could make a difference to what evidence you have. Lange (2000) criticizes putative demonstrations of (Jeffrey’s commitment to) non-commutativity in roughly the same way I do in the text.

Problems for Credulism

123

I think the best way to think about commutativity is to think about its application in cases where there is no empirically significant difference in the order in which you acquire two pieces of evidence. Instead, the choice of which to update on first is just a formal artifact. For example, you may have your left hand and your right hand in each of two holes in the ground, and you simultaneously feel something with each hand. Shouldn’t we get the same results by doing any of the following: • first update on your right-hand evidence, then on your left-hand evidence, then on the claim that they were acquired simultaneously; • first update on your left-hand evidence, then on your right-hand evidence, then on the claim that they were acquired simultaneously; • update in the first place only on some composite of your left- and right-hand evidence. At root, I think commutativity just amounts to a commitment that our formalism will not attach any significance to which of these routes we take. Whether empirically real differences in the order we acquire different pieces of evidence make an epistemic difference, on the other hand, depends upon the details of our situation. That issue shouldn’t be confused with commutativity. So now we’ve gotten an intuitive grip on what Jeffrey Bayesianism involves; we’ve observed that Jeffrey updates, like Classical updates, are rigid with respect to the proposition being updated on; and we’ve heard that commutativity entails that updating “in the same way” means updating with the same scaling factor B. That’s all the machinery we need to see that Jeffrey Bayesianism ultimately isn’t that hospitable to the key ideas of dogmatism, either.

9. Threats from Jeffrey Bayesianism There are three arguments for this conclusion, using somewhat different, plausible but still controversial assumptions. The first of these arguments is from Weisberg (2009), and turns on facts about rigidity and independence.57 Recall, rigidity is the property that when you update on E, all your conditional probabilities given E, and given notE, remain the same. This is a core feature of Classical Bayesianism, of Jeffrey

57 Weisberg’s paper is titled “Commutativity or Holism? . . .” “Holism” here is roughly the thesis that all acquisitions of (at least empirical) justification are underminable. Weisberg (forthcoming) extends his arguments to some other formal models of belief update.

124

Foundations of Dogmatism

Bayesianism, and of some other popular formal models of rational credence as well. It’s sometimes motivated by saying: Assumption-5 When what you learn is E, you don’t learn anything about the epistemic relations between E and other propositions.

However, whether that intuitive assumption really is correct, or whether it really is equivalent to the formal property we’ve identified, seem substantial questions. I will not pursue them. Weisberg presents the problem this poses as follows. We suppose you will undergo some learning episode that will involve your Jeffrey updating to some degree on E, but this hasn’t yet happened. We suppose also there could be a pure underminer U for the support you thereby acquire to believe E. This is a notion we discussed back in section 2: an underminer not mixed with any opposing or supporting elements.58 It’s at least initially natural to think this should be possible. It’s also natural to think that what this amounts to is that, where Old(.) is your probability function before the learning episode, Old (E | U) = Old (E) That is, before the learning episode takes place, we expect U to be probabilistically independent of E. Getting more evidence for U should not yet either increase or decrease your credence in E. Let’s label this Assumption-6

If the support you acquire for updating on E is underminable at all, then it could be undermined by some U such that Old(E|U) = Old(E).

Now let the learning episode take place. It is the very proposition E you’re updating on whose support U undermines. If we keep Assumption-2’s identification of what you should update on with what you acquire immediate justification for, then this is just the kind of possibility the dogmatist posits. Let New(.) be your probability function after the learning episode. Observe that we would not think that U should be probabilistically independent of E in New(.). Instead, the higher your credence was in U, the less support we should think you end up with for E in New(.). If U thoroughly undermines the support you newly acquired for E, then New(E|U) should leave you back at Old(E). But it is not essential for U’s undermining effect to be so thorough. All that is important is that we think, now New (E | U) = New (E)

58 Hence, U in Weisberg’s argument is very different from U in White’s argument. White’s U was incompatible with the proposition Hand whose justification we were considering.

Problems for Credulism

125

However, what this means is that U and E started out independent, and then ended up no longer independent, as a result of a Jeffrey update on E, which by definition will be rigid with respect to E. And what Weisberg observes is that that is impossible. If an update is rigid with respect to E, then whatever started out independent of E must end up independent of E. So if Assumption-2 and Assumption-6 are right, then the possibility the dogmatist posits is not representable in Jeffrey Bayesianism, after all. You can represent the support you acquire for E, when updating on E, as being defeasible—for example, you could go on to acquire opposing evidence for not-E. But you can’t represent this support as being purely underminable in the way envisaged by Assumption-6. There are ways out of this. We might decide that pure underminers aren’t possible after all, or that Assumption-6 doesn’t correctly formally capture what they involve. Or we might try one of the strategies from section 5, and give up Assumption-2 and/or -3. What you gain immediate justification for might not be the same as what you update on. Immediate justification might not correspond to any specific update. The negative effect of undermining evidence might not be wholly represented inside the formal model: it might help determine what you update on, or to what degree. Maybe the appropriate scaling factor should be a function of how much justification you independently have for U.59 But something needs to give.60 I said there were three arguments. The second and third arguments turn not on the requirement of rigidity but rather on the requirement of commutativity.61

59

I have developed an extension of the Jeffrey formalism which works that way. Wagner (2013) criticizes Weisberg’s argument, making the following points: (i) Jeffrey did not intend the updating rule we described in section 8 to be the “be-all-and-end-all of probability revision.” Moreover, (ii) you can reasonably update with that rule on a proposition E only when it is antecedently reasonable for you to believe that the conditional probability for any U on E should be unaffected by what you’ve learned. Finally, (iii) in cases like the ones we’re discussing, what it’s reasonable for you to update on is not E but E щU. I’m prepared to believe that (ii) was Jeffrey’s view. Certainly it is a natural way to read the end of Jeffrey (1987). But if point (ii) is correct, it severely limits the applicability of the model described in section 8. I suspect that only theorists who embraced the “towers of antecedent justification” we described in section 4, or theorists who thought we only acquired non-underminable justification—that is, only non-credulists—would then be able to regard the model as an exhaustive theory of reasonable beliefupdate. (As Wagner points out, Jeffrey did not himself so regard it.) Wagner’s point (iii) tacitly concedes the main point I’m taking Weisberg to be arguing: this model is not suited to represent updates where it’s your justification for the very proposition E you’re updating on that’s underminable. (Wagner instead recommends updating on a proposition whose acquired justification isn’t underminable by any of the other hypotheses in your algebra.) So long as our Assumption-2 is in place, this means that the model is not suited to represent updates of the sort dogmatists believe possible. 61 I’ve been helped in thinking about the second argument by discussions with Matt Kotzen. The third argument comes from Christensen (1992, section 6) and is also presented in Weisberg (2009). Christensen’s paper is titled “Confirmational Holism and Bayesian Epistemology.” “Holism” here can be understood in the way explained in note 57. 60

126

Foundations of Dogmatism

Theorists argue about what is the best way to measure the amount of evidential support some update contributes to a given proposition E. One natural such measure is called the E:not-E Bayes factor of the update. This is not the only measure that has advocates, but it does have some broad support. We needn’t dwell on the details of how this measure is defined.62 The bottom line is that the E:not-E Bayes factor of an update on E will turn out equivalent to 1 + B, where B is the scaling factor displayed in our earlier diagrams. That’s the same scaling factor we said Jeffrey formalism forces to be invariant, for subjects who update “in the same way,” on pain of violating commutativity. So in other words, if you and I have a good look at an object in noon sun, and it looks the same to us—and that means that our credal models should be supplied the same input—then we’ll need to update on some single E and do so with the same E:not-E Bayes factor; though our prior probability functions will differ. But then if Bayes factors are also the right way to measure how much evidential support our updates contribute to E, it will follow that each of us will have acquired the same amount of such support. But we’ve said nothing about what other ways we might differ! My Oldʹ(U) might be much higher for some underminer U than your Old(U):  that is, I  may have much more evidence than you do for thinking vision is unreliable in noon sun. Shouldn’t that have undermined, and so weakened, the support I get for E, when my credal system gets the same input as yours? We have a conflict between the following ideas: i. As the dogmatist claims, what you acquire immediate justification for is itself underminable. ii. As Assumption-2 claims, a proposition you acquire immediate justification for is what you update on. iii. If the support I acquired for E is undermined to a greater degree than the support you acquired, then the measure of how much support I got for E should be less. iv. The right way to measure that support is in terms of the E:not-E Bayes factor of the update. v. But commutativity requires that when we receive “the same input,” we update on the same proposition E, and the E:not-E Bayes factor of our updates is the same. In other words, once it’s settled what the input is, commutativity leaves no further room for sensitivity to one’s prior credence in underminers. Now there are ways out of this too. Some of them are the same ways we mentioned before. We could give up Assumption-2. We could give up the idea 62 For reference, it is the quantity (New(E)/New(not-E))/(Old(E)/Old(not-E)). Sometimes the log of this quantity is used instead. In a Classical setting, this is equivalent to the “likelihood-ratio” measure of evidential support.

Problems for Credulism

127

that the world settles which input we receive without consulting our current credence in underminers. So even if you and I have the same experiential phenomenology, it might be appropriate for our credal models to be supplied different inputs. But the most controversial element in the preceding is (iv). The third argument dispenses with (iv), and also with (iii). In their place it puts the following fact: vi.

Given an E:not-E Bayes factor, and an initial probability function Old(.), your ending credence in E will then be a function solely of Old(E). It will not be sensitive to the credence Old(.) assigns to anything else—in particular, not to Old(U)—except insofar as that affects the factors already specified.

Claims (i), (ii), (v), and (vi) tell us that if you and I have the same initial probability in E, and acquire the same further immediate justification to believe it, then we should end with the same probability in E, too—even if I have more justification to believe some underminer of the immediate justification we acquired than you do.63 Again, this is not tolerable. Something here still needs to give. Getting rid of (iii) and (iv) didn’t really help.

10. Conclusion I will close by summing up some of the stable packages that do still look viable for a dogmatist or credulist. They’re all built out of pieces we’ve already mentioned. Option 1.  We could limit the explanatory ambitions of our probabilistic frameworks even further than is commonly acknowledged and say they just don’t apply in cases where there is the prospect of non-quotidian undermining. Or at least, they don’t apply fully. As we saw, we can try to model non-quotidian undermining, imperfectly, as if it were a case of quotidian undermining. We might explore other formal systems to see if they can better model the behavior of non-quotidian underminers. Option 2. We might stay with Bayesian systems and say that our prior credence in non-quotidian underminers plays a role in determining what the 63 As Christensen points out, this is arguably the same difficulty that Garber (1980) presses against Field. Garber phrases his complaint in terms of intuitive redundancies, rather than undermining effects, that Field’s proposal is not respecting. But formally, it’s natural to expect those effects to stand or fall together. Wagner’s (2002) response to Garber is different from the recommendation (iii) described in note 60. I would describe the move he recommends here as: don’t assume that what update you should perform is determined just by what experiential phenomenology you have, or what prima facie justification you newly acquire. It is also sensitive to facts about what you’ve previously learned. This amounts to giving up our Assumption-2 and Assumption-3. James Hawthorne (2004) responds similarly to Garber.

128

Foundations of Dogmatism

inputs to these formal systems are. The fact that a range of subjects acquires the same prima facie justification, informally understood, doesn’t guarantee that our formal models of their epistemic state should all be updated in the same way. This gives up Assumption-2 and Assumption -3. Option 3.  We mentioned the possibility of your sometimes updating on bold claims, like Hand. In Classical Bayesianism, we would have to assume “Popperian” conditional probabilities, so that the credence might later be defeated. But reflecting on what happens if you acquired the undermining evidence first pushed us to include at least some elements of Option 2. In Jeffrey Bayesianism, we can immediately make sense of your updating on claims like Hand, but we would have to reconcile ourselves with the impossibility of pure underminers, in the sense envisaged by Assumption-6. We would also have to include elements of Option 2 here, too, because of difficulties raised in the second and third arguments from section 9.  We would need to have some response to the worries raised in the second argument about how to measure the amount of support we acquired for what we update on. Option 4.  We might say we always update on claims that are more cautious than can possibly be undermined—if there are any such claims, which I myself doubt. This strategy can still make room for dogmatism or credulism if Assumption-2, and possibly some other assumptions, are rejected. Section 7 explored some ways this might be developed.

Acknowledgments Thanks to several forums in New York; and to audiences at Brown, Fribourg, York, Geneva, Bologna, Svolvaer, Toronto, and the Institute for Advanced Studies at Hebrew University; and to the Institute for financial support. In different ways, Maria Lasonen-Aarnio, David Barnett, David Christensen, Annalisa Coliva, Philip Ebert, Adam Elga, Hartry Field, Matt Kotzen, Ram Neta, Eugenio Orlandelli, Christian Piller, Sherri Roush, Stephen Schiffer, Nico Silins, Levi Spectre, Scott Sturgeon, Chris Tucker, Ralph Wedgwood, Jonathan Weisberg, Roger White, Tim Willenken, and Crispin Wright made especially helpful contributions to the end result. Thanks also to Marinus Ferreira for making the diagrams.

References Achinstein, Peter. 2001.The Book of Evidence. Oxford: Oxford University Press. Alston, William. 1989. Epistemic Justification. Ithaca: Cornell University Press. Boghossian, Paul. 2003. “Blind Reasoning.” Aristotelian Society Supplementary Vol. 77: 225–48.

Problems for Credulism

129

_____. 2001. “How Are Objective Epistemic Reasons Possible?” Philosophical Studies 106: 1–40. _____. 2000. “Knowledge of Logic.” In Christopher Peacocke and Paul Boghossian, eds., New Essays on the A Priori, 229–54. Oxford: Oxford University Press. BonJour, Laurence. 1998. In Defense of Pure Reason. Cambridge: Cambridge University Press. Byrne, Alex. 2008. “Knowing That I  Am Thinking.” In Anthony Hatzimoysis, ed., Self-Knowledge, 105–24. Oxford: Oxford University Press. _____. 2005. “Introspection.” Philosophical Topics 33: 79–104. Chisholm, Roderick. 1989. Theory of Knowledge. Englewood Cliffs, NJ: Prentice-Hall. Christensen, David. 2011. “Disagreement, Question-Begging and Epistemic Self-Criticism.” Philosophers’ Imprint 11. xxxx _____. 2010a. “Higher-Order Evidence.” Philosophy and Phenomenological Research 81: 185–215. _____. 2010b. “Rational Reflection.” Philosophical Perspectives 24: 121–40. _____. 2008. “Does Murphy’s Law Apply in Epistemology? Self-Doubt and Rational Ideals.” In Tamar Szabo Gendler and John Hawthorne, eds., Oxford Studies in Epistemology, Vol. 2, 3–31. Oxford: Oxford University Press. _____. 2007. “Epistemology of Disagreement:  The Good News.” Philosophical Review 116: 187–217. _____. 1992. “Confirmational Holism and Bayesian Epistemology.” Philosophy of Science 59: 540–57. Chudnoff, Eli. 2011. “The Nature of Intuitive Justification.” Philosophical Studies 153: 313–33. Cohen, Stewart. 2010. “Bootstrapping, Defeasible Reasoning, and A  Priori Justification.” Philosophical Perspectives 24: 141–59. _____. 2005. “Why Basic Knowledge Is Easy Knowledge.” Philosophy and Phenomenological Research 70: 417–30. Comesaña, Juan. Forthcoming. “On an Argument against Immediate Justification.” In Matthias Steup and John Turri, eds., Contemporary Debates in Epistemology, 2nd ed. Oxford: Blackwell. Dretske, Fred. 1971. “Conclusive Reasons.” Australasian Journal of Philosophy 49: 1–22. Elga, Adam. Manuscript. “Lucky to Be Rational.” http://www.princeton.edu/ ~adame/ papers/bellingham-lucky.pdf. Enoch, David. 2010. “Not Just a Truthometer:  Taking Oneself Seriously (but not Too Seriously) in Cases of Peer Disagreement.” Mind 119: 953–97. Feldman, Richard. 2009. “Evidentialism, Higher-Order Evidence, and Disagreement.” Episteme 6: 294–312. _____. 2006. “Epistemological Puzzles about Disagreement.” In Stephen Hetherington, ed., Epistemology Futures, 216–36. Oxford: Oxford University Press. _____. 2005. “Respecting the Evidence.” Philosophical Perspectives 19: 95–119. _____. 2003. Epistemology. Upper Saddle River, NJ: Prentice-Hall. Field, Hartry. 1978. “A Note on Jeffrey Conditionalization.” Philosophy of Science 45: 361–67. van Fraassen, Bas. 1989. Laws and Symmetry. Oxford: Oxford University Press. Fumerton, Richard. 1995. Metaepistemology and Skepticism. Lanham, MD:  Rowman and Littlefield. Garber, Daniel. 1980. “Field and Jeffrey Conditionalization.” Philosophy of Science 47: 142–45.

130

Foundations of Dogmatism

Hawthorne, James. 2004. “Three Models of Sequential Belief Updating on Uncertain Evidence.” Journal of Philosophical Logic 33: 89–123. Hawthorne, John. 2004. Knowledge and Lotteries. Oxford: Oxford University Press. _____. 2002. “Deeply Contingent A Priori Knowledge.” Philosophy and Phenomenological Research 65: 247–69. Huemer, Michael. 2011. “The Puzzle of Metacoherence.” Philosophy and Phenomenological Research 82: 1–21. _____. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. _____. 2006. “Phenomenal Conservatism and the Internalist Intuition.” American Philosophical Quarterly 43: 147–58. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Jeffrey, Richard. 1987. “Alias Smith and Jones:  The Testimony of the Senses.” Erkenntnis 26: 391–99. Kelly, Thomas. 2013. “Disagreement and the Burdens of Judgment.” In David Christensen and Jennifer Lackey, eds., The Epistemology of Disagreement: New Essays. Oxford: Oxford University Press. _____. 2010. “Peer Disagreement and Higher Order Evidence.” In Richard Feldman and Ted A. Warfield, eds., Disagreement, 111–74. Oxford: Oxford University Press. _____. 2005. “The Epistemic Significance of Disagreement.” In Tamar Szabó Gendler and John Hawthorne, eds., Oxford Studies in Epistemology, Vol. 1, 167–96. Oxford:  Oxford University Press. Kotzen, Matt. 2012. “Silins’s Liberalism.” Philosophical Studies 159: 61–68. _____. Manuscript.“A Formal Account of Epistemic Defeat.” http://matthewkotzen.net/ matthewkotzen.net/Research_files/defeatersweb.pdf. Kung, Peter. 2010. “On Having No Reason.” Synthese 177: 1–17. Lange, Marc. 2000. “Is Jeffrey Conditionalization Defective by Virtue of Being NonCommutative? Remarks on the Sameness of Sensory Experience” Synthese 123: 393–403. McGrath, Matthew. Forthcoming. “Dogmatism, Underminers and Skepticism,” Philosophy and Phenomenological Research. Neta, Ram. 2010. “Liberalism and Conservatism in the Epistemology of Perceptual Belief.” Australasian Journal of Philosophy 88: 685–705. Plantinga, Alvin. 1993. Warrant: The Current Debate. Oxford: Oxford University Press. Pryor, James. 2012. “When Warrant Transmits.” In Annelisa Coliva, ed., Mind, Meaning, and Knowledge:  Themes from the Philosophy of Crispin Wright. Oxford:  Oxford University Press. _____. 2004. “What’s Wrong with Moore’s Argument.” Philosophical Issues 14: 349–78. _____. 2001. “Highlights of Recent Epistemology.” British Journal for the Philosophy of Science 52: 95–124. _____. 2000. “The Skeptic and the Dogmatist.” Noûs 34(4): 517–49. _____. Manuscript. “Hypothetical Oughts.” http://www.jimpryor.net/research/papers/ Hypothetical.pdf. Pynn, Geoff. Forthcoming. “The Bayesian Explanation of Transmission Failure.” Synthese. Reichenbach, Hans. 1956. The Direction of Time. Berkeley : University of California Press. Salmon, Wesley. 1984. Scientific Explanation and the Causal Structure of the World. Princeton, NJ: Princeton University Press.

Problems for Credulism

131

Schechter, Joshua. 2013. “Rational Self-Doubt and the Failure of Closure.” Philosophical Studies 163: 429–52. Schiffer, Stephen. 2004. “Skepticism and the Vagaries of Justified Belief.” Philosophical Studies 119: 161–84. Silins, Nicholas. 2008. “Basic Justification and the Moorean Response to the Skeptic.” In Tamar Szabó Gendler and John Hawthorne, eds., Oxford Studies in Epistemology, Vol. 2, 108–41. Oxford: Oxford University Press. _____. Forthcoming. “Experience and Defeat.” Philosophy and Phenomenological Research. Sturgeon, Scott. Forthcoming. “Undercutting Defeat and Edgington’s Burglar.” In Lee Walters, ed., Conditionals, Probability, and Paradoxi:  Themes from the Philosophy of Dorothy Edgington. Oxford: Oxford University Press. Tucker, Chris. 2012. “Movin’ on Up:  Higher-Level Requirements and Inferential Justification.” Philosophical Studies 157: 323–40. _____. 2010. “Why Open-minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45. Wagner, Carl. 2013. “Is Conditioning Really Incompatible with Holism?” Journal of Philosophical Logic 42: 409–14. _____. 2002. “Probability Kinematics and Commutativity.” Philosophy of Science 69: 266–78. Weatherson, Brian. 2012. “Induction and Supposition.” The Reasoner 6: 78–80. http://www. kent.ac.uk/secl/philosophy/jw/TheReasoner/vol6/TheReasoner-6(5).pdf. Wedgwood, Ralph. Forthcoming. “A Priori Bootstrapping.” In Albert Casullo and Joshua Thurow, eds., The A Priori in Philosophy. Oxford: Oxford University Press. Weisberg, Jonathan. 2009. “Commutativity or Holism? A Dilemma for Conditionalizers.” British Journal for the Philosophy of Science 60: 793–812. _____. Manuscript. “Updating, Undermining, and Independence.” http://www.utm. utoronto.ca/~weisber3/new/Research_files/Updating%20Undermining%20and%20 Independence.pdf. White, Roger. 2006. “Problems for Dogmatism.” Philosophical Studies 131: 525–57. Willenken, Tim. 2011. “Moorean Responses to Skepticism: A Defense.” Philosophical Studies 154: 1–25. Williamson, Timothy. Manuscript. “Very Improbable Knowing.” http://www.philosophy. ox.ac.uk/__data/assets/pdf_file/0015/19302/veryimprobable.pdf _____. 2011. “Improbable Knowing.” In Trent Dougherty, ed., Evidentialism and Its Discontents, 147–64. Oxford: Oxford University Press. _____. 2007. The Philosophy of Philosophy. Oxford: Blackwell. _____. 2005. “Knowledge and Scepticism.” In Frank Jackson and Michael Smith, eds., The Oxford Handbook of Contemporary Philosophy, 681–700. Oxford:  Oxford University Press. Wright, Crispin. 2002. “(Anti-) Sceptics Simple and Subtle:  G.  E. Moore and John McDowell.” Philosophy and Phenomenological Research 65: 330–48. Yablo, Stephen. 2002. “Coulda, Woulda, Shoulda.” In Tamar Szabó Gendler and John Hawthorne, eds., Conceivability and Possibility, 441–92. Oxford: Oxford University Press.

This page intentionally left blank

{ part iii }

Seemings and Epistemic Internalism

This page intentionally left blank

{6}

Does Phenomenal Conservatism Solve Internalism’s Dilemma? Matthias Steup In his recent book, Justification without Awareness, Michael Bergmann (2006, 3–24) presents a serious objection to internalism. According to this objection, advocates of internalism are confronted with a dilemma. As he defines it, internalism is the view that if S has a justified belief, S must have awareness of at least some of the justification contributors: the factors upon which the belief ’s justification depends. Since such awareness is either strong or weak, internalism itself is either strong or weak. Bergmann argues that no matter whether weak or strong, internalism succumbs to a problem the view can’t overcome. Strong internalism generates a regress Bergmann considers vicious. Weak internalism yields a notion of justification that he thinks falls victim to one of the very objections its advocates level against externalism. Since in addition to its weak or strong construal, there is no middle way, internalism seems doomed. In this chapter, I will present a version of weak internalism that I take to be immune to Bergmann’s objection.

1. Internalism Internalism is best understood as a constraint on the kinds of items that qualify as sources of justification.1 Different internalists have their preferred ways of articulating what that constraint is. Mine goes as follows: there is one and only one source of justification for our beliefs, namely our reasons. According to this view, believing p while having an undefeated reason for p is both necessary and sufficient for being justified in believing p.2

1 By “sources of justification,” I mean the items or factors that determine whether a belief is justified. Bergmann refers to them as “justification contributors.” 2 Some epistemologists hold that justified belief must be causally based on the reasons that justify it. Like Feldman (2003, 46), I  take basing to be a necessary condition of not justification but wellfoundedness. What I am going to say in response to Bergmann’s dilemma is independent of what position is taken on the basing requirement. Readers who hold that justification does require basing could

136

Seemings and Epistemic Internalism

Two questions present themselves immediately: What are reasons, and what is it for a subject to have, or be in possession of, a reason? Among the bona fide candidates for being reasons are the following: beliefs, rational intuitions, memory impressions, introspective states, and perceptual experiences. What qualifies such items to play the role of reasons is that, provided certain conditions (to be discussed later) are met, they make it likely that the beliefs they support are true. To have a reason for p is simply to have a belief, intuition, or experience with a suitable content. Making a subject’s justification for her belief in this way dependent on her reasons yields paradigmatic internalism, for reasons—items like the ones on the aforementioned list—should count as internal to the subject’s mind on just about any construal of internalism.3 Since internalism restricts reasons to a subset of our mental states and experiences, facts as such do not count as reasons. For example, a fingerprint is not the sort of thing that can justify a jury member to find a defendant guilty. But the jury member’s memory of what the prosecutor said about the fingerprint can. Likewise, the mere presence of smoke can’t justify one in believing that a fire is nearby. But a perceptual experience of smoke, brought about through either smell or sight, can. In general terms, internalism is the view that facts located outside of our cognitive grasp never justify our beliefs. Rather, if something is to give us justification for a belief, it must have been brought within our cognitive perspective through a faculty like reason, memory, introspection, or perception. Internalism has two characteristic implications. First, since basing beliefs on undefeated reasons is sufficient for justification, subjects in the evil demon world have justified beliefs.4 Second, since having an undefeated reason is necessary for justification, a belief ’s having its origin in a reliable process or a belief ’s meeting some other external condition—a condition that might not supply the subject with a reason—is not sufficient for justification.5 Let’s consider an example that illustrates how internalist justification works. Sitting on his deck, Ed observes a little yellow bird. He believes: “It’s a goldfinch.” Let’s stipulate that Ed’s belief is justified on the basis of Ed’s visual experience simply take internalism to be the view that basing one’s belief that p on an undefeated reason for p is both necessary and sufficient for being justified in believing p. 3 What the internality internalists demand exactly amounts to is difficult to pin down. Typical attempts proceed by requiring that justification contributors are mental states, or that they are properly part of the subject’s perspective, or that the subject have proper cognitive access to them. According to Bergmann’s construal, the defining mark of internalism is an awareness requirement. For reasons of space, I will assume an intuitive understanding of internality, leaving aside the issue of how best to define the notion. 4 The claim that subjects in the evil demon world have justified beliefs enjoys a rather high degree of intuitive appeal. Hence the claim is used as a premise in a prominent argument against externalism (see Ginet 1985; Cohen 1984; Sosa 1991, 132ff.). 5 For this point, see BonJour’s (1985, 45–51) influential counterexample involving Norman, the clairvoyant.

Solving Internalism’s Dilemma

137

of the bird he is observing. The specific experience that gives Ed a reason for, and thus justifies, his belief that a goldfinch is present can be described as “Ed’s being appeared to goldfinch-like,” or by saying that “it seems to Ed as though a goldfinch is present.” Since Ed knows what a goldfinch looks like and since nothing defeats his goldfinch-like perceptual seeming, Ed’s belief in the presence of a goldfinch is justified. As mentioned earlier, Bergmann’s dilemma is based on the premise that internalism is either weak or strong. What characterizes weak internalism is the absence of any requirement to the effect that the subject have an epistemic meta-grasp of her reasons as reasons for the beliefs they support. In contrast, strong internalism includes such a requirement. Consider a theory according to which Ed’s goldfinch-like experience by itself qualifies as a reason that, if undefeated, is sufficient for justifying Ed’s belief in the presence of a goldfinch. Such a theory would be an example of weak internalism. Strong internalists would say that Ed’s goldfinch-like experience by itself does not so qualify. What’s required in addition is that Ed have some kind of understanding of the fact that his goldfinch-like experience is a reason for believing that the observed bird is a goldfinch.6 According to Bergmann’s dilemma argument, each form of internalism comes with a fatal liability. Whereas strong internalism generates a vicious regress, weak internalism is too weak to satisfy the very standards that internalists themselves impose on justification. Next, I’ll discuss each of the dilemma’s two horns.

2. Bergmann’s Objection to Strong Internalism According to Bergmann, strong internalism generates a vicious regress. The mechanism responsible for generating the regress comes clearly into view when we focus on the question of what is involved in a meta-grasp of one’s reasons as reasons for the beliefs they support. Clearly, just having an epistemic belief of the form r is a reason for p is not enough if the expectation is

6

Bergmann (2006, 14–15) articulates the additional requirement as follows: S’s belief B is justified only if (i) there is something, X, that contributes to the justification of B and (ii) S is actually aware of X in such a way that S justifiedly believes that X is in some way relevant to the appropriateness of holding B

Applying this to the case at hand, the additional demand of strong internalism would be that Ed believes, with justification, that his goldfinch-seeming contributes to the justification of his goldfinch belief. The passage just quoted expresses what Bergmann (2006, 14–19) calls an actual awareness requirement. He also considers and argues against a potential awareness requirement, and he distinguishes between doxastic and non-doxastic awareness requirements. Since I do not aim at defending strong internalism, there is no need to discuss these distinctions here.

138

Seemings and Epistemic Internalism

that the required meta-awareness contribute to the target belief ’s justification.7 After all, the belief that r is a reason for p might be lacking justification or even be utterly irrational. Thus, if the epistemic meta-grasp that strong internalism requires is of any use toward contributing to the target belief ’s justification, the epistemic belief that r is a reason for p must itself be justified. Hence one way of defining strong internalism goes as follows: S is justified in believing p if and only if (i) S has an undefeated reason r for believing p; (ii) S is justified in believing that r is a reason for believing p.

Unlike condition (i), condition (ii) requires level-ascent. That is why strong internalism generates a regress. Assume S is justified in believing p, and let S’s belief that p = B1. Since B1 is justified, conditions (i) and (ii) are both met: S has a reason r1 for B1, and S has a second justified belief, B2, namely, the belief that r1 is a reason for B1. Since B2 is justified, it too satisfies conditions (i) and (ii). By virtue of satisfying condition (i), S has a reason r2 for B2. By virtue of satisfying condition (ii), we get a third justified belief, B3. The content of B3 is that r2 is a reason for B2. Since B3 is justified, it meets conditions (i) and (ii). So we get a fourth belief, B4, with the content that r3 is a reason for B3. Due to condition (ii), this series continues ad infinitum. So if strong internalism is true, every justified belief comes with an infinite number of the kinds of beliefs condition (ii) requires. Not every infinite regress is vicious. But there is good reason to believe that this one is, since the regress at hand is one of meta-beliefs of increasing complexity. B1 is simply a belief with p as its content. B2 is about B1 and the reason it supports; its content is r2 is a reason for believing that r1 is a reason for believing p.

B3 ascends one further level: r3 is a reason for believing that r2 is a reason for believing that r1 is a reason for believing p.

Each further belief in the series adds another level of complexity. Pretty soon it’s no longer possible to keep straight what the content of these beliefs is. When we arrive at, say, B16, we lose comprehension of exactly which proposition the belief ’s content is supposed to be. It’s safe to conclude that having an infinite number of such beliefs is humanly impossible. The regress strong internalism generates is, therefore, vicious. For strong internalism, epistemic ascent internalism would be a fitting label. I  agree with Bergmann that making epistemic ascent a necessary condition

7 The expression “epistemic beliefs” is meant to refer to beliefs about what justifies what, what is a reason for what, or the justificational status of one’s beliefs.

Solving Internalism’s Dilemma

139

of justification is fatal. Epistemic ascent internalism is not a good version of internalism.8 Let’s move on and consider weak internalism.

3. Bergmann’s Objection to Weak Internalism According to weak internalism, having an undefeated reason for p is both necessary and sufficient for being justified in believing p.9 The only kind of awareness weak internalism requires is awareness of the reason itself. Awareness of any epistemic facts is not needed. Bergmann claims that weak internalism is too weak by internalism’s own lights. The view is too weak, Bergmann argues, because it succumbs to a well-known objection to externalism, due to archinternalist Laurence BonJour (1985, 41–45): the well-known case of Norman the reliable clairvoyant. Norman, as you might recall, holds a belief about the president’s present location, and this belief has its origin in a reliable cognitive process. However, Norman is not aware of anything supplying him with a reason in support of his belief. Therefore, BonJour concludes, Norman’s belief is not justified even though it is reliably produced. Norman, then, is lacking a reason for his belief about the president’s present location. According to BonJour, that’s an epistemic defect robbing Norman’s belief of any justification it might otherwise have. But why, Bergmann asks, think of the lack of a reason as an epistemic defect? The answer to this question, he suggests, is stated in the following passage by Bonjour: since Norman’s belief is reliably produced, the truth of his belief would not be an accident from the standpoint of [a] hypothetical external observer who knows all the relevant facts and laws. But how is this supposed to justify Norman’s belief? From his subjective perspective it is an accident that the belief is true. And the suggestion here is that the rationality or justifiability of Norman’s belief should be judged from Norman’s own perspective rather than from one which is unavailable to him. (Bonjour 1985, 43–44)

Bergmann refers to BonJour’s line of reasoning as the Subject’s Perspective Objection (SPO). According to Bergmann, the condition that weak internalism places on a belief ’s justification—the subject’s having a reason for the belief—is not enough to protect against BonJour’s SPO, that is, to ensure that the subject’s true belief count as non-accidental when considered from within the subject’s perspective. Even if, Bergmann argues, a subject has a reason for her true belief that p, it is possible, from the subject’s own perspective, that the

8 Some internalists think that strong internalism withstands the infinite regress objection (see Rogers and Matheson, 2011, and Tucker, 2012). 9 See note 2.

140

Seemings and Epistemic Internalism

belief ’s truth is accidental. Therefore, having a reason for p, even if undefeated, is not sufficient for being justified in believing p. Since weak internalism does not require that the subject have an epistemic meta-grasp of her reasons as reasons, weak internalism allows, according to Bergmann, for counterexamples in which a subject holds a true belief that the view counts as justified even though the belief ’s justification is effectively undermined by BonJour’s SPO. Call such a counterexample a Bergmann Case. The details of a Bergmann Case are as follows: (i) S has a true belief that p. (ii) S has a reason r for p. (iii) S does not have any form of justified meta-belief with the content that r is a source of justification for his belief that p. (iv) Because of (iii), the truth of S’s belief must, from within S’s perspective, be considered accidental. (v) Because of (iv), S’s belief fails to be justified by the very standards internalists themselves endorse. When conditions (i)–(v) are met, a subject holds a true belief that meets the condition that according to weak internalism is both necessary and sufficient for justification. Yet the subject’s belief is not justified. If Bergmann Cases are possible, they are effective counterexamples to weak internalism. In what follows, I argue that a well-known version of weak internalism does indeed fall victim to Bergmann Cases. However, I also argue that if weak internalism is supplied with suitable bells and whistles, it is immune to Bergmann Cases.

4. Phenomenal Conservatism The version of weak internalism that I claim to be refuted by Bergmann Cases is phenomenal conservatism, which can be seen as a theory telling us what a subject’s reasons are.10 According to phenomenal conservatism, whatever support a subject has for her beliefs arises from her seemings. So what phenomenal conservatism tells us is this: a subject’s reasons = a subject’s seemings. The main thought motivating phenomenal conservatism is that for the justification of our beliefs about the world, all we have to go on is how the world seems or appears to us through the full array of our cognitive faculties. Each of our faculties—perception, introspection, memory, and reason—presents the world as being this way or being that way. Since we have nothing else that could serve as a rational ground for our beliefs, whatever justification our beliefs might enjoy must come from the seemings with which our faculties provide us. 10 The term “phenomenal conservatism” is due to Huemer (2000), a main advocate of that view. Pryor (2000) defends a similar view, and arguably Chisholm’s (1977) view could be counted as a version of phenomenal conservatism as well.

Solving Internalism’s Dilemma

141

On my initial list of bona fide candidates for being reasons, I included beliefs. According to phenomenal conservatism, beliefs function as reasons only derivatively, namely, by virtue of the seemings that support them. So phenomenal conservatism presents us with a rather simple theory of what reasons are: a subject’s reasons consist of all and only her seemings. However, the simplicity of the starting point is compensated for by the fine-grained and near infinite variety of seemings we can have, ranging over the full spectrum of seeming states our faculties can induce in us. According to phenomenal conservatism, the having of a seeming that p is sufficient for having prima facie justification for believing p; put differently, having an undefeated seeming that p is sufficient for being justified in believing p.11 For a subject’s undefeated seeming that p to justify her belief that p, it’s not necessary that the subject form a justified belief of the form my seeming that p is a reason for believing p. Let’s consider again the example of Ed, and let’s list the salient aspects of the case: (i) (ii) (iii) (iv)

it seems to Ed that the bird he is observing is a goldfinch; it is true that the bird is a goldfinch;12 nothing defeats Ed’s seeming; Ed does not perform epistemic level-ascent; he does not form any belief with the content that his being appeared to goldfinch-like is a reason for him to believe in the presence of a goldfinch.

Bergmann would argue that because of (iv), the truth of Ed’s belief must, from within his perspective, be considered accidental. Hence, Bergmann would say, the case is a counterexample to phenomenal conservatism, for phenomenal conservatism implies that Ed’s goldfinch belief is justified. So, weak internalism construed as phenomenal conservatism falls victim to BonJour’s SPO. Although I agree that BonJour’s SPO does refute phenomenal conservatism, I will argue next that Ed’s belief is justified after all, notwithstanding Ed’s failure to perform epistemic level-ascent.

5. Protection against Bonjour’s SPO In describing the goldfinch example by listing features (i) through (iv), something has been left out that I consider essential to internalist justification: the contribution of Ed’s memory to his epistemic perspective. If Ed were to ask

11 There is no reason here to be worried about the absence of a basing requirement, since the issues I am concerned with in this chapter do not hinge on the presence or absence of this requirement. See note 2. 12 As far as the justification of Ed’s belief is concerned, this condition is irrelevant. However, since a false belief can hardly be accidentally true, the kind of case of interest in the present context requires that the belief be true.

142

Seemings and Epistemic Internalism

himself whether the kind of seeming on which his goldfinch belief is based is reliable, he would (let us stipulate) remember that seemings of the kind in question have rarely led him astray.13 What he remembers about the track record of his visual discriminations is that there haven’t been any cases in which what seemed to him to be a bird was in fact a squirrel, or what seemed to him to be a dog was in fact a cat, or what seemed to him to be a frog was in fact a salamander. Rather, what his memory would tell him is that in the past, by far most of the beliefs based on his visual seeming have been true. So, if Ed were to consult his memory to check on the credentials of his visual seemings, his memory would tell him that they have an excellent track record. Ed’s memory, then, would upon reflection supply him with an important layer of additional seemings, namely higher-order seemings that justify the attribution of reliability to Ed’s lower-order seemings. As condition (iv) makes explicit, Ed does not in fact engage in any such reflections. Thus the memorial seemings just mentioned are not included in his current epistemic perspective. Let us distinguish, however, between memorial seemings and the data that are embedded within one’s memory. When I want to recall my telephone number, a certain series of numbers seems to be the right one. In normal cases, such a seeming retrieves content that is stored in one’s memory. Let’s refer to such content as memory data. Ed does not actively consult his memory on the reliability of his goldfinch seeming. Thus he doesn’t have a memorial seeming with the content that his goldfinch seeming is an instance of a reliable kind of visual seeming. Nevertheless, his memory contains a rich body of data such that, if Ed were to reflect on the reliability of his goldfinch seeming, he would—if he succeeds in reflecting competently—retrieve these data and have memorial seemings telling him that its track record is excellent. Let’s capture this aspect of Ed’s situation by adding another feature to our description of the case: (v) Ed has memory data that justify the attribution of reliability to the type of seeming of which his goldfinch seeming is an instance.14

Feature (v) protects Ed’s justification for his goldfinch belief against BonJour’s SPO. That’s why, despite the absence of level-ascent, the example of Ed’s 13 There are also possible cases in which upon reflection Ed would notice that his belief in the presence of a goldfinch is not based on a reliable kind of seeming. But that’s not the kind of case I wish to consider here. 14 Bergmann distinguishes between an actual and a potential awareness requirement. It might be objected that making the possession of memory data of this kind a necessary part of justification amounts to imposing a potential awareness requirement—which, according to Bergmann (2006, §3.2), leads to an infinite regress. Since a full discussion of this objection would go beyond the scope of this chapter, I will only give a brief response. According to the view I am defending here, perceptual justification requires memory data the nature of which I describe by saying that they would support the attribution of reliability to S’s perceptual faculties if S were to reflect on them, provided S succeeds in reflecting competently. So S’s actual or potential abilities at epistemic reflection are irrelevant to

Solving Internalism’s Dilemma

143

goldfinch belief is not a Bergmann Case. Embedded within Ed’s perspective is a body of memory data that make the truth of Ed’s goldfinch belief highly likely. This, I claim, is sufficient for eliminating any accidentality that might otherwise adhere to the truth of Ed’s goldfinch belief. Thus, if Bergmann were to argue that, due to feature (v), the truth of Ed’s goldfinch belief must be considered accidental from within Ed’s perspective, I would have to disagree with him. On behalf of Bergmann, it could be objected that, notwithstanding the enrichment of Ed’s perspective by including in it memory data that ground the attribution of reliability, the dilemma remains in full force. Either I require that Ed conceive of his memory data as justification contributors, or I do not. If I do, an infinite regress takes off. If I do not, the truth of Ed’s goldfinch belief remains accidental from within Ed’s perspective. My response is that I reject the dilemma’s second horn. The question at hand is whether the elimination of accidentality requires of Ed a mental operation consisting of his retrieving the memory data in question and conceiving of them as contributing to the justification of his goldfinch belief. I deny that accidentality elimination requires any such operation. The mere possession of the relevant memory data is enough. Since Ed is in possession of such memory data, there are, embedded within his subjective perspective, reasons that ground the ascription of reliability to his goldfinch-like experience. Given the presence of such reasons within his perspective—his intellectual outlook, as it were—it would be rather odd to claim that, from within Ed’s subjective point of view, the truth of his goldfinch belief must be considered accidental.15 The example of Ed’s goldfinch belief is representative of typical cases of perceptual belief. By far most of our perceptual seemings are such that if we were to reflect on their epistemic credentials, our memory would tells us that they are instances of seeming types that have an excellent track record. It’s obvious

this point. What matters is this: if, as a matter of objective fact, S’s memory data support the attribution of reliability, then they do so irrespective of what the subject’s potential for epistemic reflection happens to be. So what the view requires is merely a certain kind of memorial content. In describing this content, a story must be told about the support structure in which such memorial content serves as a kind of foundation. What the view does not require is that the subject have the ability, actually or potentially, to form—via reflection on the required memory data—epistemic meta-beliefs about her perceptual experiences as reasons for her perceptual beliefs. In short, the view does not require any mental operations resulting in level ascent. Therefore, the requirement the view imposes does not generate an infinite regress. 15 A question that’s relevant in the present context is whether evidential justification requires basing one’s belief on one’s evidence. Epistemologists who think it does might disagree with me. They might argue that if justification does indeed require memory data, then the beliefs such data support are justified only if they are based on these data, which in turn requires conceiving of them as justification contributors. However, I reject the basing requirement, and thus consider this line of reasoning to be headed in the wrong direction. There is much to be said on this issue, but for reasons of space I can’t pursue this issue here.

144

Seemings and Epistemic Internalism

why this is so; as perceivers, we typically enjoy a high degree of reliability, and we can know that we do via memory. Alas, there are also atypical cases, cases in which seemings are, as it were, bad: unaccompanied by any memory data that would justify the attribution of reliability to them. Next, I’ll consider such a case.

6. Against Phenomenal Conservatism The case of Ed’s goldfinch belief is not, as we have seen, a Bergmann Case and thus is not a counterexample to phenomenal conservatism. That doesn’t mean there aren’t other cases that are. To build such a case, we need to describe a case with the following features: (i) the subject has a seeming that p resulting in a true belief that p; (ii) the subject’s epistemic perspective fails to include the very ingredients that protect against BonJour’s SPO, namely, memory data that justify the attribution of reliability to the subject’s seeming; (iii) because of (ii), the truth of the subject’s belief that p must be considered accidental from within the subject’s perspective; (iv) because of (iii), the subject’s belief fails to be justified by internalist lights. Let’s consider such a case. Suppose it seems to Carl that it’s going to be a sunny afternoon, and it’s true that it’s going to be a sunny afternoon. Carl wants to go hiking, which is why he hopes it’s going to be sunny, which in turn is why he believes it’s going to be sunny. So his belief is the result of wishful thinking. Assuming Carl doesn’t happen to be the rare sort of being for whom wishful thinking is a reliable belief-forming process, what’s going to be missing from his epistemic perspective are memory data justifying the attribution of reliability to his sunny afternoon seeming. So, from within Carl’s subjective perspective, the truth of his belief must be considered accidental. Hence Carl’s belief is not justified. But according to phenomenal conservatism, it is.16 The case, therefore, is a counterexample to phenomenal conservatism.17 Similar cases can be construed at will. The recipe goes as follows. Let a subject have a true belief based on a bad seeming that p—one that comes from a

16 Advocates of phenomenal conservatism might challenge this claim by arguing that Carl’s belief is defeated. However, it’s not plausible to assume that there must be defeaters in Carl’s epistemic perspective. Let’s make two further assumptions: first, he does not have any evidence against believing that it’s going to be sunny. So he doesn’t have a rebutting defeater. Second, it isn’t evident to him that his belief is grounded in wishful thinking. So he doesn’t have an undermining defeater either. 17 Markie (2005, 356–57; 2006, 119–20) has a further objection along the same lines. Tucker (2010, 523–42) offers a defense of phenomenal conservatism against this kind of objection.

Solving Internalism’s Dilemma

145

bad origin like wishful thinking, drug-induced hallucinations, brain lesions, and the like. Choose details so as to ensure that the seeming is (i) not defeated and (ii) not backed up by any memory data that justify the attribution of reliability to it. The result will be a Bergmann Case: a belief that is not justified because its truth must be considered accidental from within the subject’s perspective.18 The example we just considered undermines phenomenal conservatism. It shows that if phenomenal conservatism is to withstand BonJour’s SPO, it must be amended with an evidence-of-reliability requirement. According to this amendment, seemings as such are not a source of justification; only seemingly reliable seemings are.19 In the next section, I propose such an amended version of phenomenal conservatism. In the remainder of this section, I wish to supplement the counterexample-based rejection of phenomenal conservatism with a second argument. According to phenomenal conservatism, having an undefeated seeming that p is sufficient for being justified in believing p. We might wonder whether there is an explanation of the justificatory relation between a seeming that p and the belief that p. Advocates of phenomenal conservatism cannot offer an explanation of this relation. They must treat it as a brute fact.20 In and of itself, this need not be considered a liability. After all, an attempt at explanation will have to stop somewhere. However, the counterexample we considered previously indicates that having an undefeated seeming that p is not sufficient for being justified in believing p.  So what phenomenal conservatism considers a brute fact turns out to be not a fact at all. Moreover, the theoretical move needed to protect weak internalism against BonJour’s SPO provides us with the desired explanation. If we endorse that seemings, if they are to succeed in doing justificatory work, must be accompanied by evidence of reliability in the form of suitable memory data, we can say the following: an undefeated seeming that p is a source of justification if and only if it is a seemingly reliable seeming—one that would upon reflection seem reliable given the subject’s memory data. If a seeming that p is a seemingly reliable seeming, then its being seemingly reliable is precisely what explains why there obtains a justificatory relation between the seeming that p and the belief that p.

18

For a different kind of counterexample, see my echolocation case (Steup 2004). By “seemingly reliable seemings” I  have in mind seemings that upon reflection seem reliable. Again, normally our seemings are not accompanied by memorial seemings of their reliability. However, we do normally have memory data such that were we to reflect on the trustworthiness of our perceptual or memorial seemings, they would seem reliable. So when I say that justification requires seemingly reliable seemings, I  do not require actual memorial seemings but merely memory data that, upon reflection, will ground memorial seemings of reliability. 20 I refer to this aspect of the view as ipso factism (Steup 2004). 19

146

Seemings and Epistemic Internalism

The explanation offered in the previous paragraph is based on the following internalist picture of justification: whatever justification we might have for our beliefs can only arise from the input of our faculties: perception, introspection, memory, and reason. According to phenomenal conservatism, this input consists of a certain subset of our mental states: seemings. But, as we will recognize when we treat Bergmann’s objection to weak internalism seriously, the mere fact that p seems true to S just isn’t enough to generate any justification for S to take p to be true. For such justification to arise, the seeming that p must, from within S’s perspective, make it likely that p is true. And within S’s perspective, there is one and only one thing that can succeed in making S’s seeming that p a reason for thinking it likely that p is true: evidence for S to think that her seeming that p is an instance of a reliable type of seeming, a type of seeming whose content is true more often than not. To conclude this section, I put my rejection of phenomenal conservatism in a more general perspective. Conservatism in epistemology is the view that beliefs are innocent (justified) until proven guilty (until defeated by elements of the subject’s total evidence). The standard criticism of this view is that a belief is not justified unless the subject’s evidence fully supports it. Unlike conservatism about beliefs, phenomenal conservatism is a view about the justificatory role of experiences. The main thought is that a seeming with a certain content is innocent (justifies a belief with that content) unless the seeming is proven guilty (unless it is undermined by parts of the subject’s total evidence). My main objection to phenomenal conservatism parallels the standard criticism of belief conservatism. Beliefs are justified only by virtue of evidential support. Likewise, it is only by virtue of evidential support that seemings qualify as a source of justification. To thus qualify, the subject must have memory data that provide the subject with reasons for attributing reliability to them. This evidence-of-reliability requirement can be supported by way of counterexample, and by arguing that phenomenal conservatism fails to supply us with an explanation of the justificatory relation between seemings and beliefs.

7. Internalist Reliabilism Consider again what phenomenal conservatism asserts: Necessarily, S is justified in believing p if and only if (i) S believes p; (ii) it seems to S that p; (iii) S’s seeming that p is undefeated.21 21 As Ali Hasan has pointed out to me, some phenomenal conservatives might want to restrict this account to non-inferential beliefs. They might hold that if S competently infers p from propositions that seem true to S, S is justified in believing p even if p itself does not seem true to S. Be this as it may, it doesn’t affect my objection to phenomenal conservatism. According to the version of phenomenal conservatism Hasan proposes as an option, condition (ii) is not necessary for non-inferential justification. My objection, on the other hand, is that the conjunction of conditions (i)–(iii) is not sufficient for justification, irrespective of whether it is inferential or non-inferential.

Solving Internalism’s Dilemma

147

What’s missing in this account is a condition that requires evidence of reliability, ensuring that only seemingly reliable seemings count as sources of justification. If we amend phenomenal conservatism with such a condition, we get a view that could appropriately be labeled internalist reliabilism.22 Here is what this view says: Necessarily, S is justified in believing p if and only if (i) S believes that p; (ii) S has a seeming that p of kind K; (iii) S’s seeming that p is undefeated; (iv) S’s memory data support the attribution of reliability to kind K seemings.

According to internalist reliabilism, seemings are not innocent until proven guilty. To do justificatory work, they need to satisfy an evidence-of-reliability condition. Hence internalist reliabilism is not a version of phenomenal conservatism. The latter succumbs to Bergmann Cases because it does not come with an evidence-of-reliability condition. In contrast, internalist reliabilism does not succumb to Bergmann Cases because it includes such a condition, namely, condition (iv). If this condition is met, then S has justification for believing that her seeming that p is an instance of a reliable kind of seeming. From within S’s perspective, the truth of her belief that p is then not at all accidental. So where does internalist reliabilism stand with regard to Bergmann’s dilemma? As we have seen, if weak internalism comes in the form of phenomenal conservatism, the view allows for Bergmann Cases and must therefore be judged too weak. If, on the other hand, we move to strong internalism, we get the requirement of level ascent: S must form a meta-belief with the content that her seeming that p is a reason for believing p, or some such belief. Bergmann is right when he claims that such a requirement leads to a vicious regress. Internalist reliabilism is comfortably situated between the two extremes. It avoids the excessive permissiveness of weak internalism and steers clear of the pernicious move of requiring the subject to form epistemic meta-beliefs. Unlike phenomenal conservatism, internalist reliabilism requires evidence of reliability. So, it withstands BonJour’s SPO and is therefore immune to Bergmann Cases. Unlike strong internalism, it does not require that the subject make use of her reliability evidence by way of forming justified meta-beliefs to the effect that her reasons make the beliefs they support likely true. Internalist reliabilism, then, is a version of internalism that is neither too weak nor too strong. Let me sum up. As Bergmann’s objection to weak internalism makes clear, advocates of the view must rise to the challenge of explaining what, without requiring level ascent, protects internalist justification against BonJour’s SPO. I have argued that what fills the bill is a requirement to the effect that seemings

22

Steup (2004) defends this view.

148

Seemings and Epistemic Internalism

be seemingly reliable seemings. Given this evidence-of-reliability requirement, it is difficult to see how weak internalism might still be vulnerable to counterexamples involving beliefs that, though true, must count as merely accidentally true from within the subject’s perspective. Hence, I suspect that externalist opponents of weak internalism will object to internalist reliabilism on different grounds. They might claim that even though the view succeeds in providing protection against BonJour’s SPO, it nevertheless succumbs to other objections. In the remainder of this chapter, I discuss what I consider to be the two most serious.

8. The Circularity Objection It might be objected that internalist reliabilism is infected with a vicious kind of circularity. Consider internalist reliabilism’s fourth condition: S has memory data with the content that kind K seemings are reliable. These memory data, I claim, give subjects justification for attributing reliability to their seemings. But if these memory data are to justify an attribution of reliability to one’s seemings, they themselves must be trustworthy too. And for this to be the case, S must have further memory data that justify the attribution of reliability to S’s memorial seemings. Thus, if justification depends, as I claim it does, on the possession of evidence of reliability, then it must be possible to derive justification for trusting one’s memorial seemings from one’s memorial seemings themselves. According to the objection, such source circularity is as bad as using perception to justify reliance on perception, or using induction to justify reliance on induction.23 This is a serious objection deserving full-length treatment in a separate paper. Here I briefly outline the bare bones of what I take to be an effective reply to it. From the internalist point of view, source circularity is indeed inescapable. The question is whether it is bad. The standard argument for the badness of source circularity as vicious goes as follows. If it’s legitimate to rely on memory to justify reliance on memory, then it’s also legitimate to justify trusting one’s crystal ball by consulting the crystal ball itself regarding its own reliability (see Fumerton 1995, 177). I don’t find this argument convincing. Relying on one’s crystal ball for answering the question “Is my crystal ball reliable?” is undoubtedly bad. Hence, if memory circularity and crystal ball circularity are indeed analogous, then the former would be bad as well. Alas, it is far from clear that the two are really analogous. Surely we ought to think that there is a satisfying answer to the question: why is

23 Fumerton (1995, 177)  has a nicely articulated rejection of source circularity. See also Alston’s (1993) excellent discussion, and Bergmann’s (2006, ch 6) defense of source circularity.

Solving Internalism’s Dilemma

149

it bad to rely on a crystal ball for an assessment of its own reliability? If the only answer available is that doing so is bad because it’s an instance of source circularity, then the analogy would indicate that memory circularity is bad as well. However, there is an alternative explanation. Crystal ball circularity is obviously a bad maneuver because we know to begin with that crystal balls are unreliable. Using a source known to be unreliable is bad, whatever the intended purpose of its use might be. According to this alternative explanation, crystal ball circularity is bad, not because it is an instance of source circularity but rather because we know that basing one’s beliefs on what a crystal ball says is a highly unreliable method of belief acquisition. Since the same cannot be said for memory, the analogy falls apart. It is one thing to have a response to the crystal ball argument against source circularity; it is another to have a satisfying account of when source circularity is irrational and when it is not. Such an account, I believe, can be supplied by explaining the difference between a rational and an irrational use of source circularity in terms of the coherence, depth, and richness of one’s total body of evidence.24 So my response to the circularity objection is twofold: (i) arguments for the viciousness of source circularity can be defused, and (ii) an account of the rationality of source circularity is available.

9. The Generality Problem According to the second objection, internalist reliabilism is subject to a generality problem comparable to the one for process reliabilism.25 The target of the objection is again condition (iv): S has memory data with the content that kind K seemings are reliable. The question at hand is this: since each seeming token is an instance of many seeming-types, how are we to arrive at the right specification of kind K seemings? Consider again Ed and his goldfinch seeming. Ed’s belief is justified, according to internalist reliabilism, because he has memory data that justify the attribution of reliability to the kind of seeming of which his goldfinch seeming is an instance. There are many different kinds of seemings of which his goldfinch seeming is an instance. For example, it is a perceptual seeming, a visual seeming, a seeming involving a small feathered creature, and a seeming had (let’s assume) in the month of August. The problem is that of identifying which of the many seeming-types is the relevant one. The generality problem for process reliabilism can be exploited for the construction of painful counterexamples, which is one reason the problem is 24 Ernest Sosa (1997, 2007, 2009) has a view that emphasizes the importance of a coherent metaperspective on one’s faculties and attempts to justify source circularity along these lines. I defend such circularity in my 2013 paper. 25 Goldman (1979) gives the locus classicus on process reliabilism. Feldman (1985) and Conee and Feldman (1998) have excellent discussions of the generality problem.

150

Seemings and Epistemic Internalism

formidable. There are several recipes for constructing counterexamples. Here is one. Describe a case involving a belief that is clearly unjustified but happens to be true. Let the belief-producing process be wishful thinking. By choosing a maximally specific description of the process type, there will be one and only one process token of it, namely, the specific belief in the case under consideration. Since that belief is true, the chosen process type is perfectly reliable. Hence, unless amended with bells and whistles that solve its generality problem, process reliabilism yields the bad implication that the clearly unjustified belief is justified.26 Is internalist reliabilism subject to similarly constructed counterexamples? It seems to me it is not. The construction of generality-based counterexamples against internalist reliabilism would require types of seemings that go too far in the direction of either specificity or generality. Call such types deviant. For condition (iv) to be satisfied, the subject would have to have memory data with the content that the deviant type of seeming invoked in the intended counterexample is reliable. If the subject doesn’t have the requisite memory data, this condition will not be met. Consider again Carl and his wishful-thinking-based belief that it’s going to be sunny in the afternoon. Carl’s desire for good weather causes him to have a seeming with the content that it will be a sunny afternoon. If a critic of internalist reliabilism wishes to turn this case into a generalitybased counterexample, she must rise to the following challenge: find a deviant seeming-type for Carl’s weather-seeming such that Carl has memory data that would justify the attribution of reliability to this seeming-type. The strategy of devising a one-token-only type won’t work in this case, for it would ensure that Carl won’t have the requisite memory data. Hence the critic would have to go in the opposite direction and choose a seeming-type of high generality. A potentially promising move would be to let Carl’s seeming be an instance of the following type: a seeming about the weather. But does Carl have any memory data that would justify the attribution of reliability to his weather seemings? There are two possibilities to consider. First, Carl is a normal human being. In that case, he is not going to have memory data that would justify the attribution of reliability to his weather seemings. Suppose Carl bases lots of beliefs about the weather on weather forecasts from reliable sources. In that case, he is going to have memory data indicating that his forecast-based weather-seemings have a good track record. What these memory data justify is the attribution of reliability, not to Carl’s generic weather-seemings, but rather to Carl’s forecast-based weather seemings.27 So the memory data required for the objection 26

Feldman (1985) refers to this as the “single case problem.” Here’s an analogy. Suppose Carl has a Brand A model X car. His memory data tell him that his car has a good track record of functioning as intended. Carl has memory data that justify the attribution of reliability to the vehicle type Brand A model X. He does not have memory data justifying the attribution of reliability to the, say, vehicle-type four-door sedan. The lesson here is that evidence for attributing reliability to a certain type of item might not be evidence for attributing reliability to a more general type of the same item. 27

Solving Internalism’s Dilemma

151

will not be available. Second, we might suppose that Carl is a non-normal but possible human being who does indeed have memorial seemings grounding the attribution of reliability to his generic weather seemings. In that case, we would be looking at a belief that is (i) caused by wishful thinking and (ii) supported by Carl’s overall body of evidence. Because of (ii), the implication that internalist reliabilism yields—namely, that Carl’s belief is justified—would no longer be implausible.28 The example we just considered shows the following: when it comes to the generality problem, the main difference between process reliabilism and internalist reliabilism is crucial. Whereas the former makes de facto reliability a necessary condition of justification, the latter requires merely evidence of reliability. Finding deviant process types that are either excessively broad or excessively narrow is easy. But it isn’t easy to find evidence that justifies ascribing reliability to deviant seeming-types. The best strategy for giving the objection traction would be to say the following: for some unjustified beliefs based on some seeming-token SN, we can find a suitably general seeming type ST such that some possible subject has memory data that justify the attribution of reliability to ST. According to the objection, a case like that would be a counterexample to internalist reliabilism, for the view would then count an unjustified belief as justified. However, it is doubtful that such cases are coherent. For if the subject does indeed have memorial evidence for ascribing reliability to ST, then it is no longer plausible to claim that the belief based on SN is really unjustified.29 To conclude, I briefly sum up the main points made in this chapter. First, Bergmann’s objection to strong internalism sticks. Second, Bergmann’s objection to weak internalism meets with success when it comes to phenomenal conservatism, but fails when phenomenal conservatism is amended with an evidence-of-reliability requirement, resulting in a view I  call “internalist reliabilism.” When a subject has a true belief that p while having a seeming that p accompanied by memory data justifying the attribution of reliability to her seeming, there is, from within her perspective, nothing accidental about the truth of her belief. However, although internalist reliabilism is similar to

28 Note that beliefs resulting from wishful thinking are not necessarily unjustified. It’s possible for a subject to live in a benevolent demon world in which beliefs originating in wishful thinking are known to be nearly always true. 29 The evidentialist solution to the generality problem I have proposed might be of use to those who think that justification—or alternatively, knowledge—requires reliable belief-producing processes. Either way, the problem demands a solution. Externalists who think that justification requires de facto reliable belief production can solve the problem by pairing this view with an evidence-of-reliability requirement, thus making justification an evaluative status with both an internalist and an externalist dimension. The same strategy is open to internalists who think that the external dimension of knowledge can be captured in terms of reliability. (Few internalists claim that knowledge is lacking any externalist dimension.) In each case, the proposed solution is that the relevant process types are the ones that connect with the subject’s reliability-confirming memory data.

152

Seemings and Epistemic Internalism

phenomenal conservatism, it is a different view. Hence I answer as follows the question that is the title of this chapter: Internalist reliabilism solves internalism’s dilemma, but phenomenal conservatism does not.

Acknowledgments I’m indebted to Mike Bergmann, Chris Tucker and Ali Hasan for excellent comments on earlier versions. An even earlier draft was presented at the third Midwest Epistemology Conference (October 2009) at St. Louis University. For helpful discussion on that occasion, I’m indebted as well to Al Casullo, Marian David, Trent Dougherty, Mylan Engel, Richard Fumerton, John Greco, Bruce Russell, Joe Salerno, and Casey Swank.

References Alston, William. 1993. The Reliability of Sense-Perception. Ithaca: Cornell University Press. Bergmann, Michael. 2006. Justification without Awareness. Oxford: Oxford University Press. Chisholm, Roderick. 1977. Theory of Knowledge, 2nd ed. Englewood Cliffs: Prentice-Hall. BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, MA:  Harvard University Press. Cohen, Stewart 1984. “Justification and Truth.” Philosophical Studies 46: 279–95. Conee, Earl and Richard Feldman. 1998. “The Generality Problem for Reliabilism.” Philosophical Studies 89: 1–29. Feldman, Richard. 2003. Epistemology. Upper Saddle River: Prentice Hall. _____. 1985. “Reliability and Justification.” The Monist 68: 159–74. Fumerton, Richard. 1995. Metaepistemology and Skepticism. Lanham, MD:  Rowman and Littlefield. Ginet, Carl. 1985. “Contra Reliabilism.” The Monist 68: 159–74. Goldman, Alvin I. 1979. “What Is Justified Belief?” In George Pappas, ed., Justification and Knowledge, 1–23. Dordrecht: Reidel. Huemer, Michael. 2000. Skepticism and the Veil of Perception. New  York:  Rowman and Littlefield. Markie, Peter. 2006. “Epistemically Appropriate Perceptual Belief.” Noûs 40: 118–42. _____. 2005. “The Mystery of Perceptual Justification.” Philosophical Studies 126: 347–73. Rogers, Jason and Jonathan Matheson. 2011. “Bergmann’s Dilemma:  Exit Strategies for Internalists.” Philosophical Studies 152: 55–80. Pryor, James. 2000. “The Skeptic and the Dogmatist.” Noûs 34: 517–49. Sosa, Ernest. 2009. Reflective Knowledge. Oxford: Oxford University Press. _____. 2007. A Virtue Epistemology. Oxford: Oxford University Press. _____. 1997. “Reflective Knowledge in the Best Circles.” Journal of Philosophy 94: 410–30. _____. 1991. “Reliabilism and Epistemic Virtue.” In Ernest Sosa, Knowledge in Perspective, 131–47. Cambridge: Cambridge University Press.

Solving Internalism’s Dilemma

153

Steup, Matthias. 2013. “Is Epistemic Circularity Bad?” Res Philosophica 90: 215–236. _____. 2004. “Internalist Reliabilism.” Philosophical Issues 14: 403–25. Tucker, Chris. 2012. “Movin’ on Up:  Higher-Level Requirements and Inferential Justification.” Philosophical Studies 157(4): 323–40. _____. 2010. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45.

{7}

Phenomenal Conservatism and the Dilemma for Internalism Michael Bergmann In previous work I  have argued against internalism by means of a dilemma intended to force all internalists to accept one of two undesirable options: either their internalism is unmotivated or it is saddled with vicious regress problems (Bergmann 2006, esp. chs. 1 and 2). Recently it has been argued that Phenomenal Conservatism is a kind of internalism that can avoid being impaled on either horn of this dilemma.1 Phenomenal Conservatism is a theory of justification according to which justification depends on seemings.2 One common version of Phenomenal Conservatism says that a person’s belief that p is prima facie justified if it is based on its seeming to that person that p. A stronger version says that a person’s belief that p is prima facie justified if and only if it is based on its seeming to that person that p. Although Phenomenal Conservatism is commonly thought of as an internalist position in epistemology, it is only the stronger version of Phenomenal Conservatism that has a chance of being so. Internalist positions are those that require (for justification) awareness of justification-contributors (see Bergmann 2006, 55–57). Only the stronger version of Phenomenal Conservatism mentions any requirements at all for justification.3 For this reason, I will focus on the stronger “if and only if ” version and I’ll refer to it as “PC.”4 1 See Rogers and Matheson (2011). Steup (2004; this volume) suggests something similar (see my discussion at the end of this chapter). Other arguments that internalists can avoid my dilemma have been offered by Rogers and Matheson (2011), Markie (2009), and Crisp (2009) and, in section 2 of this chapter, I consider ways that Phenomenal Conservatives might draw upon them for help. 2 Some Phenomenal Conservatives will say that this applies only to non-inferential justification. For accounts and defenses of Phenomenal Conservatism, see Huemer (2001, 99–100; 2006, 148; 2007, 30), Swinburne (1998, 20–21; 2001, 141–49), Tucker (2011; 2010), and Conee (2004, 15–16). Conee stops short of endorsing the version of Phenomenal Conservatism that he discusses. 3 It’s compatible with the weaker version of PC (which says only that being based on a seeming that p is sufficient for the prima facie justification of the belief that p) that being reliably formed is also sufficient for a belief ’s prima facie justification. A view that endorsed both of those sufficiency claims would not be an internalist position in epistemology, any more than reliabilism is. 4 There’s reason to think that Huemer (2001, 109)  endorses the stronger version. He says that a principle like

The Dilemma for Internalism

155

I argue that PC cannot escape my dilemma for internalism. To do this, it is helpful to clarify what PC is saying, and for that reason I begin, in section 1, by examining in some detail what seemings are. I then argue, in the remainder of the chapter, that, whichever horn of the dilemma for internalism they take, Phenomenal Conservatives are in trouble.5

1. What Are Seemings? 1.1. three uses of “it seems to me that p” Chisholm (1989, 21) draws our attention to at least three different uses of “It seems to me that p.” Sometimes you say things like “It seems to me that the room is getting warmer” and in doing so you “play it safe, indicating that what [you] say carries no guarantee at all, and that anyone choosing to believe what [you] say does so at his or her own risk.” In such a case, “seems” has a kind of hedging function, providing the speaker with a way out should the claim turn out to be false. But there are also purely descriptive uses of “It seems to me that p” in which hedging is not the purpose. Among these there are comparative and non-comparative uses of the phrase. The comparative descriptive use of “It seems to me that p” is shorthand for something like “It seems to me the way it does when p”—for example, “When I wear those eye blinds, it seems to me that (or as if) it’s completely dark” can be shorthand for “When I wear those eye blinds, it seems to me the way it does when it’s completely dark.” PC1

a person’s belief that p is prima facie justified if it is based on its seeming to that person that p

is the only principle of non-inferential justification. And from this he concludes that the only thing that can defeat prima facie justification that results from a seeming is another seeming that provides rebutting or undermining evidence (110). That suggests that he also endorses PC2 a person’s belief that p is prima facie justified only if it is based on its seeming to that person that p. After all, if only seemings can be evidence and justification requires evidence, then having prima facie justification implies that one has the relevant seemings.In addition, in a 2006 paper, Huemer (2006, 148) makes it clear that he endorses internalism understood in accord with what he calls the “appearance account” of internalism, which he explains as follows: Appearance Account: Internalism holds that all of the conditions that confer justification supervene on how things seem to the subject. Given his endorsement of internalism so understood, we can conclude that, according to Huemer, if a belief is prima facie justified, its justification is conferred by conditions that supervene on seemings of the subject. And that too suggests that he endorse PC2. And of course the stronger version of Phenomenal Conservatism just is the conjunction of PC1 and PC2. 5 Elsewhere, I  present a different kind of objection to Phenomenal Conservatism and argue that there are ways for externalists about justification to accommodate the main insights of PC (see Bergmann forthcoming).

156

Seemings and Epistemic Internalism

By contrast, the non-comparative use of “It seems to me that p” is intended as a description of how things seem without comparison to any other way things seem. In this chapter, I am concerned not with the first two uses of “It seems to me that p” but instead with the purely descriptive, non-comparative use of it. When someone makes such a descriptive, non-comparative claim, that person is reporting an occurrence or state of its seeming to him or her that p. That occurrence or state is a mental event or a mental state that we can think of as a seeming, and that is what I have in mind here when I speak of seemings.

1. 2. beliefs, inclinations to believe, and seemings I believe that 2 + 2 couldn’t equal 5. On what basis? One natural answer is to point out, first, that it seems very strongly to me that 2 + 2 couldn’t equal 5 and, second, that I believe that 2 + 2 couldn’t equal 5 on the basis of that seeming. This example of a seeming is the sort of thing that Phenomenal Conservatives have in mind when proposing their views on justification. They are most definitely not thinking of a seeming that p as what it feels like to believe p. After all, they require for a belief ’s justification that the belief is based on a seeming— that it is held because of that seeming. It wouldn’t make sense to require for justification that a belief be based on what it feels like to have that very belief. But what exactly are these seemings on which our beliefs are based? It’s common to point out that seemings can be of different types:  there are, to name a few, intellectual (or a priori) seemings such as the simple mathematical one mentioned in the previous paragraph; in addition, there are perceptual seemings, memory seemings, introspective seemings, and moral seemings. It’s also common to point out that seemings are distinct from beliefs (see Bealer 1999, 31; Huemer 2001, 99–100; 2007, 30–31). Thus you can have the perceptual seeming that the previously straight ruler that you just inserted into the water at an angle is now bent. But you might not believe that it’s bent. Or you can have the perceptual seeming that your friend is angry but not believe that she’s angry (given that she has just assured you that she is not angry). Cases like these suggest that a seeming that p is distinct from a belief that p. It’s not uncommon to say that a seeming that p is just an inclination to believe that p (Swinburne 2001, 141–42; Rogers and Matheson 2011, 60; Conee 2004, 15). But Michael Huemer (2007, 31) offers some good reasons for thinking that this is a mistake. One is that inclinations to believe are explained by seemings: the reason I’m inclined to believe that 2 + 2 couldn’t equal 5 is that it seems to me that 2 + 2 couldn’t equal 5. Another reason for distinguishing seemings from inclinations to believe is that inclinations to believe are in some cases explained by other things. William Tolhurst (1998, 297–98, cases III and V) gives some examples of such cases. Suppose a man’s son has been charged with a horrible crime and the evidence presented in court all points to the

The Dilemma for Internalism

157

conclusion that he is guilty. The father might be strongly inclined to believe that his son is not guilty but this inclination may be caused by his desire that his son not be guilty, not by any seeming that his son is not guilty (since he lacks any such seeming). Or the father might feel that it is his moral duty as a father to believe that his son is not guilty and this might cause him to be inclined to believe his son is not guilty, even though it doesn’t seem to him that his son is not guilty.6 So seemings are not inclinations to believe, even though seemings almost invariably produce inclinations to believe (which explains why people are tempted to mistakenly identify the two). These same examples from Tolhurst also show how it’s possible to have a belief that p even when it doesn’t seem to you that p. The father might believe that his son is not guilty because he wishes it to be true even though it doesn’t seem to him that his son is not guilty. Another way in which seemings differ from beliefs is that while it is typically not irrational for a person S to have two seemings that S explicitly recognizes as contradictory, it is typically irrational for S to have two beliefs that S explicitly recognizes as contradictory. It might seem to you that your friend is angry (due to your observation of her behavior) and yet at the same time it might seem to you that your friend is not angry (due to her apparently sincere testimony to that effect). It isn’t irrational to have these conflicting seemings because, in having them, you aren’t committed to their contents in the way you are to belief contents when you have conflicting beliefs. To sum up what we have so far: seemings are mental states that typically cause or explain inclinations to believe (so they’re distinct both from such inclinations and from beliefs). Other states (such as desires or felt duties) can also cause inclinations to believe. Seemings, then, are a particular kind of belief-inclination-causing state.7 Exactly what kind is something I discuss in section 1.4.

1.3. sensory experiences and seemings Perceptual seemings are distinct from sensory experiences.8 Consider a perceptual seeming where it seems to you (on the basis of using your hands to feel around in a completely dark room) that there is a small ball on a table in front of you. It’s important, in this case, to distinguish the sensory experience—i.e.,

6 Cases like these should not be confused with cases in the literature where a seeming (rather than an inclination to believe) is caused by a desire. See Markie (2005, 357) and Steup (2004, 416). 7 When I call seemings “belief-inclination-causing” states, I don’t mean that of necessity they cause us to be inclined to believe things. Rather, I mean that they typically do, especially when they’re strong enough and we have no contrary seemings. 8 Up to this point, I’ve been mostly following Huemer’s lead on understanding seemings. However, I don’t think Huemer would follow me on this point (see his 2001, 66–77).

158

Seemings and Epistemic Internalism

the phenomenal feel which is the tactile experience—from the seeming that it produces. The tactile feel is one thing; the seeming that there’s a ball on the table is another. Moreover, the seeming that there’s a ball on the table is caused and explained by the tactile feel. One can imagine these tactile sensations being experienced without any seemings at all, at least with no seemings (produced by these sensations) that are about external objects in one’s environment.9 In fact, this seems to be the way things go with perception more generally: there’s the sensory experience, which produces the seeming that p, which in turn produces the belief that p (at least a belief is produced in most cases where the seeming is strong enough and there is no contrary seeming). Given that the sensory experience and the seeming that produces it are constantly conjoined, it might be tempting to conflate them, but that would be a mistake.10 Because some philosophers conflate sensory experiences with the seemings that produce them, the term perceptual seeming can be misleading. For this reason, it might be helpful to distinguish thin and thick perceptual seemings as follows: Thin Perceptual Seeming a seeming (i.e., one kind of belief-inclinationcausing mental state) that is produced by (but distinct from) a sensory experience. Thick Perceptual Seeming a combination of a thin perceptual seeming and the sensory experience that produces it.11

I think of perceptual seemings as thin, but others think of them as thick (see Pryor 2000; Steup 2004). However, even those who think of them in the latter way (in my view, mistakenly) might be willing to admit that thick perceptual seemings have both a sensory phenomenal aspect (the feel that is tactile or visual or auditory, etc.) as well as some distinct aspect that corresponds to the seeming. My view is that these two aspects are distinct kinds of mental states,

9 Consider alien cognizers—ones who have never had and do not naturally have the sorts of tactile sensations that humans have—being miraculously given such sensations while sitting still in contemplation. They might simply be puzzled by them and not form any of the seemings we would on the basis of them. See the extended thought experiment Thomas Reid describes in chapter 5, section VI of his ([1764] 1997). In that thought experiment a person has tactile sensations without having any seemings based on them about the external world. 10 I think the view that sensory experience has propositional content stems from the temptation to conflate the experience and the seeming. It’s the seeming that has the propositional content, not the sensory experience. But the sensory experience naturally produces that seeming and so we are (mistakenly) tempted to assign the seeming’s content to the experience. (In my example, in the previous note, of the alien cognizer with artificially induced tactile sensations, the alien has the sensations without the seeming and without the content of that seeming.) I can’t defend this view here, but see Bergmann (2006, 121–32) and Alston (2005) for some relevant discussion. Pryor (2000) and Steup (2004) seem to take for granted that sensory experiences are perceptual seemings and do have propositional content. 11 Some would say that this so-called combination just is a sensory experience and that the thin perceptual seeming is only an aspect of it. That’s not how I think of sensory experiences.

The Dilemma for Internalism

159

with the sensory feel being the state that causes the seeming (which, in turn, typically causes one to be inclined to believe the seeming’s content). One reason people might be tempted to conflate seemings and sensory experiences into one thing they call a “perceptual seeming” is that they may be thinking of seemings comparatively instead of non-comparatively. For example, they might be thinking of a visual perceptual seeming that (or as if) there’s a large red ball nearby as an experience that is just like (in all ways, including visually) the experience one typically has when confronted by a large red ball that is nearby in good lighting. If we are careful to focus on non-comparative seemings (perceptual and otherwise), we won’t be led astray in this way. I’ve spoken of perceptual seemings (and, as noted above, I think of them as thin perceptual seemings, and that’s what I’ll have in mind when speaking of seemings for the remainder of this chapter). But, as I’ve already mentioned, there are other types of seeming as well, including intellectual, memory, introspective, and moral seemings. And what I’ve said about the importance of distinguishing perceptual seemings from what causes them applies as well to these other types of seeming. Each of these other types of seeming is, of course, a seeming—a particular kind of belief-inclination-causing mental state (again, I’ll be saying what kind in section 1.4). What makes these types of seeming different from each other is what causes them to occur and what they are about: perceptual seemings are caused by one’s current sensory experience and are about external objects in one’s environment; introspective seemings are caused by one’s current conscious mental states (sensory experience included) and are about those mental states. It’s not as clear what intellectual, moral, and memory seemings are caused by. But we can say something about their subject matter: intellectual seemings are about logical or mathematical or other allegedly necessary truths; moral and memory seemings are about alleged facts of morality or alleged facts about the past, respectively. In addition, we notice a difference in the feel of each of these types of seeming, presumably at least partly because we notice a difference in what these seemings are caused by and in what they’re about.

1.4. seemings and felt veridicality So far, I’ve said that seemings are distinct from what causes them (e.g., sensory experiences in the case of perceptual seemings) and also distinct from what they cause—in particular, the inclinations to believe that they cause. They are, I’ve said, a particular kind of belief-inclination-causing mental state. But what kind exactly? In what way are seemings that p different from other mental states that cause inclinations to believe p, states such as a desire that p or a feeling of moral obligation to believe p? William Tolhurst (1998, 298–99) proposes that what differentiates seemings from desires and feelings of moral obligation is that seemings have felt veridicality—i.e., they have “the feel of a state whose

160

Seemings and Epistemic Internalism

content reveals how things really are.” It is because of this feeling associated with seemings that our believing the content of the seeming (in those cases in which we do believe it in response to the seeming) is experienced as “an objectively [epistemically] fitting or proper response to the seeming” (Tolhurst 1998, 299). Tolhurst’s description does a nice job of capturing what is unique about the belief-inclination-causing mental states that are seemings. A desire that p doesn’t feel like a mental state whose content reveals how things really are.12 Likewise, a feeling of moral obligation to believe p doesn’t have the feel of a mental state whose content reveals how things are; nor is believing p in response to a feeling of moral obligation to believe p experienced as an objectively epistemically fitting or proper response (instead it’s experienced, at most, as a morally fitting response). Both a desire that p and a feeling of moral obligation to believe p create some felt pressure to believe p, but in neither case is this due to their having the feel of a state whose content reveals how things really are. I will follow Tolhurst in thinking of seemings as mental states with felt veridicality—the feel of a state whose content reveals how things really are. But it’s important to remember that it can seem to one that p while at the same time also seeming to one that ~p. It might sound odd to say that both a seeming that p and a seeming that ~p feel like their contents reveal how things really are. But this does not imply that any state with the content p & ~p feels like it reveals how things really are. On the contrary, the seeming that p and the seeming that ~p can be had by someone who also has an even stronger seeming that ~(p & ~p). Although a seeming that p is similar to a belief that p in that neither guarantees that p is true, they differ in that a seeming that p doesn’t rationally commit you to p in the way that believing that p does. Likewise, your acknowledgment that you have a seeming, which has the feel of a state whose content reveals how things really are, does not rationally commit you to believing that things really are as its content says. As noted earlier, this difference helps to explain why it can be rational to have recognized inconsistencies in one’s seemings but not to have recognized inconsistencies in one’s beliefs.

1.5. are seemings conscious? It is controversial whether any mental state with a feel to it can be unconscious. If an unconscious feel is impossible, then (given the understanding of

12 It’s possible that a desire that p can cause a seeming that p, which in turn causes a belief that p. In this case, the desire that p causes a state that does feel like a state whose content reveals how things really are. But the desire that p doesn’t itself feel like a state whose content reveals how things really are.

The Dilemma for Internalism

161

seemings with which I’m working—where they are states with a certain feel) an unconscious seeming is impossible. But let’s suppose that an unconscious feeling is possible. Then perhaps it’s possible for a state to have the unconscious feel of a state whose content reveals how things really are. Nevertheless, a further question is whether it can seem to one that p even if one isn’t aware of its seeming to one that p. For it may be that a seeming that p is, of necessity and by definition, a state with the conscious feel of a state whose content reveals how things really are. (I’m assuming that if a mental state is a conscious mental state, then one is aware of it; this is so even if one isn’t aware that one is in that state, which would require conceiving of oneself as being in that state—something not required for awareness of.) Instead of trying to settle this question (of whether there can be unconscious seemings), let’s focus instead on whether PC is speaking of conscious seemings only or whether it says that even unconscious seemings can contribute to prima facie justification. Huemer (2001, 178), one of the main proponents of PC, says that seemings contribute to prima facie justification. But he also says it is false, on his view, “that what renders a belief justified is always something one is aware of (that is an object of awareness).” This suggests that he thinks seemings can contribute to a belief ’s prima facie justification even if the seemings are unconscious. However, this doesn’t fit well with his earlier characterization in the same book of his view as an internalist view: When I  talk about what beliefs are “justified,” . . . I  mean what things it is reasonable for a person to believe, given their situation at the time. This is what I  mean by “justification from the first-person perspective.” . . . Justification from the first-person perspective is what matters to us. The reason is that we have only our own perspective from which to decide what to believe. We never have the omniscient point of view, and we never have anyone else’s point of view. If I am to decide what to believe, I have to do it on the basis of the information I have; I can’t do it on the basis of information you have and I don’t. Similarly, when you want to decide what to believe, you have to do it on the basis of your information, not mine. (22)

Huemer endorses internalism and explains here that according to the internalist view of justification, justification depends on the information you have; clearly he is speaking here about what you are aware of (what’s in your perspective). Later he explains that justification depends on seemings. If we weren’t aware of seemings—if they were unconscious—then they wouldn’t count as information we have (in the relevant sense). Suppose that one could experience a seeming that p without being aware of it; then one could also experience a seeming that p without being able to become aware of it by reflection alone (such a limitation could be imposed by damage resulting in malfunction or by the interference of an evil demon). If justification depended on such unconscious seemings (of which one might not be able to become aware), it would depend on information

162

Seemings and Epistemic Internalism

we don’t have and, perhaps, couldn’t get. It would be like reliabilist justification, which depends on reliability we typically aren’t aware of and often aren’t able to be become aware of, at least not by reflection alone. Any justification arising from such unconscious seemings would not be justification from the first-person perspective. It would not be internalist justification. So if Phenomenal Conservatives want to be internalists, then they should claim that S’s belief that p is prima facie justified only if it consciously seems to S that p.13 This is so whether or not seemings can be unconscious. But suppose there are Phenomenal Conservatives who don’t care to be internalists and are happy, instead, to say that justification depends on seemings of which one isn’t and, perhaps, couldn’t become conscious (even on reflection). What do I have to say to them? Their position is not one at which my dilemma for internalism is directed. Much of what I have to say about such a view I’ve said in my discussions of mentalism, of which this non-internalist version of PC is an instance (see Bergmann 2006, ch 3). I won’t be saying anything more about such a view here. Instead, for the remainder of this chapter, I focus on the sort of PC which insists that justification requires conscious seemings.14 To review: because our interest in this chapter is in whether PC can avoid being impaled on the horns of my dilemma for internalism (which imposes an awareness requirement on justification), our focus is on conscious seemings. A conscious seeming that p has the conscious feel of a state whose content reveals how things really are. Such conscious seemings are distinct from and often caused by sensory experiences. A conscious seeming that p is also distinct from and typically causes an inclination to believe p; likewise, a conscious seeming that p is distinct from and typically results in (via the causal mediation of an inclination to believe p) a belief that p.

13 This explains the justification of occurrent beliefs. Non-occurrent beliefs don’t seem to be accompanied by seemings, certainly not by occurrent seemings. Are there such things as non-occurrent seemings? If so, how are they typically related to non-occurrent beliefs? I don’t have the space to explore these questions here. However, it’s worth noting that if it turns out that there are no non-occurrent seemings, then PC could be altered to say that a person’s belief that p is prima facie justified only if either (a) the belief is occurrent and it seems to that person that p or (b) the belief is non-occurrent but it was occurrent and when it was occurrent it seemed (at that time) to that person that p. More would need to be said, but this suggests a possible strategy for dealing with this worry. 14 Because of limitations of space, I will be ignoring the view that justification requires potentially conscious seemings—which might include unconscious seemings of which one could become conscious at will on reflection alone. Nevertheless, what I say against the view that conscious seemings are required for justification can be modified to apply as well to the view that it is potentially conscious seemings that are required. See, for example, Bergmann (2006, 14–19) for an illustration of how the argument in section 2.2 can be extended to apply to the view that justification requires potentially conscious seemings.

The Dilemma for Internalism

163

2. Strong Awareness The main question before us is whether PC can be used to escape my dilemma for internalism. Now that we have a clearer picture of what seemings are, we have a fairly good grasp of what PC is saying when it says that a person’s belief that p is prima facie justified only if it seems to that person that p. In this section, after briefly explaining my dilemma for internalism, I consider how PC fares if it takes one horn of that dilemma. In section 3 I consider how it fares if it takes the other horn.

2.1. the dilemma for internalism The first premise of my dilemma for internalism is that an essential feature of internalism is that it makes a subject’s awareness of some justification-contributor a necessary condition for the justification of any belief held by that subject. (One could object to that premise but I will simply be taking it for granted here because I don’t have the space to address this objection and, in any case, the premise is considered plausible by many.)15 The second premise of my dilemma says that the awareness required is either strong awareness or weak awareness. Strong awareness of a justification-contributor involves conceiving of the object of awareness as being in some way relevant to the justification or truth of the belief. Weak awareness of a justification-contributor doesn’t involve such conceiving (though it might involve some other sort of conceiving). The third premise says that requiring strong awareness leads to one sort of problem; the fourth premise says that requiring weak awareness leads to another sort of problem. The fifth premise says that if internalism leads to either problem, we shouldn’t endorse it. The conclusion is that we shouldn’t endorse internalism (see Bergmann 2006, 13–14). In section 1.5, I argued that PC of the internalist variety (henceforth “IPC”) says that justification requires awareness of a justification-contributor: it says that seemings are justification-contributors and that in order to do their justifying work, they must be conscious, which is just to say that the subject must be aware of them. But the question is whether IPC requires strong awareness or merely weak awareness of seemings. To put it another way, the question is whether IPC should be interpreted as weak awareness IPC or strong awareness IPC: IPCW

S’s belief that p is prima facie justified only if (1) it seems to S that p and (2) S is aware of this seeming (i.e., it is a conscious seeming); it is not necessary that (3) S conceives of this seeming that p as being in some way relevant to the truth or justification of the belief that p.

15 But see section 1.5 above and, especially, Bergmann (2006, 55–57) for reasons to think internalism requires awareness.

164

Seemings and Epistemic Internalism

IPCS

S’s belief that p is prima facie justified only if (1) it seems to S that p, (2) S is aware of this seeming (i.e., it is a conscious seeming), and (3) S conceives of this seeming that p as being in some way relevant to the truth or justification of the belief that p.

I will argue that both interpretations fall prey to my dilemma for internalism. The focus in this section is IPCS and the focus in section 3 is IPCw. The problem with IPCS has to do with clause (3). Clause (3) can be interpreted in two ways: (a) as saying that the conceiving it requires must be justified and (b) as saying that the conceiving it requires needn’t be justified. Either way, clause (3) is implausible. If we go with option (a) and say that the conceiving it requires must be justified, then we have the implausible result that an infinite regress of ever more complicated concept applications is required, even for the justification of simple beliefs such as the first-person belief that I’m in pain. But if we go with option (b) and say that the required conceiving needn’t be justified, then it is implausible to claim that such conceiving is required for justification because such unjustified conceiving contributes nothing to a belief ’s justification. In the remainder of this section, I explain the problems with each of these options in a little more detail.

2.2. option (a): the conceiving must be justified As I mentioned earlier, if we go with option (a) for interpreting clause (3) of IPCS and say that the required conceiving must be justified, then we have the implausible result that an infinite regress of ever more complicated concept applications is required, even for the justification of simple beliefs such as the first-person belief that I’m in pain. To see why this is so, suppose that my belief that I’m in pain is based on S1

the seeming that I’m in pain.

Clause (3) requires (for the justification of the belief that I’m in pain) the following concept application: A1

the application to S1 of the concept being relevant to the justification of my belief that I’m in pain.

And when understood in accord with option (a), clause (3) says that this concept application must itself be justified. Now presumably what’s required for the justification of concept applications is like what is required for the justification of beliefs. So if we assume that a suitably modified version of clause (1) is required for the justification of a concept application such as A1, then (in order to be justified) A1 must be based on S2

the seeming that S1 is relevant to the justification of my belief that I’m in pain.

The Dilemma for Internalism

165

And if we assume that such justification also requires a suitably modified version of clause (3), then (in order for A1 to be justified) I must also conceive of S2 as being relevant to the justification of A1, which is to require the following concept application: A2

the application to S2 of the concept being relevant to the justification of my concept application A1.

But, if we go with option (a), then A2 must be justified too. And of course this will continue on and on ad infinitum. The problem is that even concept application A2 is a very complicated concept application. Here it is, in full without using shorthand notation such as “S1” or “A1”: A2

the application to ⎡it’s seeming to me that my seeming that I’m in pain is relevant to the justification of my belief that I’m in pain⎤ of the concept ⎡ being relevant to the justification of my application to my seeming that I’m in pain of the concept being relevant to the justification of my belief that I’m in pain⎤.

It is difficult to get one’s mind around A2. Maybe some people can do so; but most ordinary people who believe they’re in pain will typically be unable to manage it. And A2 is, of course, only the tip of the iceberg. It gets more difficult at each higher level. The bottom line is that if we take option (a), the justification of even a simple belief such as I’m in pain requires an infinite number of ever more complicated concept applications, ones that ordinary mortals simply aren’t capable of making. As a result, option (a) has the implausible consequence that none of our beliefs is justified. Before turning to option (b), it’s worth noting that not all philosophers think option (a)  is as bad as I’ve just said it is. There are some who think that although requiring strong awareness leads to a regress of beliefs or concept applications, it needn’t be as vicious as suggested in the previous paragraph (Rogers and Matheson 2011, section 3; Crisp 2009; Fales ms). I don’t have the space here for a full response to these philosophers. I’ll simply say this:  I’d be happy to retreat temporarily, for the sake of argument, to the conclusion that if this regress is vicious, then internalism should be rejected. If internalists were to concede that conditional, then the battle would be over whether the regress is vicious. I am confident that internalism would be widely (and rightly) viewed as a highly unattractive position if internalists made such a concession; this is because I think there are few internalists who would want to require such a regress (because most think it is vicious, at least for ordinary believers). If internalists won’t concede such a conditional, it is presumably because they think there are other problems with my dilemma for internalism besides its claim that the regress is vicious—in particular, they will think either that option (b) for interpreting clause (3) of the strong awareness requirement isn’t implausible or that weak awareness requirements aren’t problematic in the way I say they are. I address the first

166

Seemings and Epistemic Internalism

of these two alleged problems with my dilemma in the next subsection (2.3) and I deal with the second in section 3.

2.3. option (b): the conceiving needn’t be justified Let’s turn to option (b) for interpreting clause (3) of IPCS. According to option (b), the conceiving that clause (3) requires needn’t be justified. I noted earlier that if this were so, then it would be implausible to claim that such conceiving is required for justification because such unjustified conceiving contributes nothing to a belief ’s justification. The rough idea here is that if an internalist Phenomenal Conservative says that satisfying clauses (1) and (2) of IPCS is insufficient for a belief ’s justification, she won’t be appeased if we add that the believer has a completely irrational and insane concept application, conceiving (in this irrational and insane way) of the seeming as being relevant to the belief ’s justification. If the belief was unjustified beforehand, adding this irrational and insane concept application won’t help. This shows that if the conceiving needn’t be justified, it is implausible to require it, in which case it is implausible to require clause (3) interpreted in accord with option (b). Peter Markie (2009, n. 8) has pointed out that the requirement of a higherlevel belief or concept application that needn’t be justified isn’t viewed by internalists as sufficient for a belief ’s justification. His suggestion is that although I  may be right in saying that an irrational and insane higher-level belief or concept application won’t be enough for the target belief ’s justification, that isn’t a problem for internalists who allow that insane and irrational higherlevel beliefs or concept applications satisfy the higher-level requirement; for they never said that satisfying that higher-level requirement was sufficient for justification, only that it was necessary. But, by way of response, suppose a belief satisfied all the requirements for justification except the requirement of higher-level belief or concept application of the sort mentioned in clause (3). According to the proponent of IPCS, that belief won’t be justified. Now suppose we make just one change to the case in question: we say that the believer has an irrational and insane higher-level concept application (e.g., she conceives—in an irrational and insane way—of the seeming that p as being relevant to the justification of the belief that p). Will that change the belief from being unjustified to being justified? Intuitively, the answer is “no,” even though the proponent of IPCS is forced to say otherwise. This shows that an insane and irrational higher-level concept application or belief doesn’t contribute to a belief ’s justification, in which case it isn’t plausible to say it is required. Hence, it isn’t plausible to require clause (3) of IPCS if it is interpreted in accord with option (b). Given that it is also implausible (as I argued in section 2.2) to require clause (3) of IPCS if it is interpreted in accord with option (a), I conclude that IPCS is implausible. It’s time now to consider IPCW.

The Dilemma for Internalism

167

3. Weak Awareness I’ve argued that internalist Phenomenal Conservatives should endorse one of the following two views: IPCW

S’s belief that p is prima facie justified only if (1) it seems to S that p and (2) S is aware of this seeming (i.e., it is a conscious seeming); it is not necessary that (3) S conceive of this seeming that p as being in some way relevant to the truth or justification of the belief that p.

IPCS

S’s belief that p is prima facie justified only if (1) it seems to S that p, (2) S is aware of this seeming (i.e., it is a conscious seeming), and (3) S conceives of this seeming that p as being in some way relevant to the truth or justification of the belief that p.

It’s central to the view of Phenomenal Conservatives to endorse clause (1) and, insofar as they are internalists, they will also endorse clause (2). In section 2, I argued that clause (3) is implausible and so IPCS should be rejected. But that conclusion probably won’t create much dismay among Phenomenal Conservatives because they have never been particularly enamored with higher-level requirements such as clause (3). In mentioning what’s required for the prima facie justification of the belief that p, they tend to emphasize only that you have a seeming that p (and, insofar as they’re internalists, they’ll say it must be a conscious seeming, in which case you’re aware of it). There is no need to conceive of this seeming in any way at all. In short, it’s weak awareness of justification-contributors (in particular, of seemings) that they require for justification, not strong awareness. My dilemma for internalism argues that if internalists require only weak awareness and not strong awareness, their position falls prey to the Subject’s Perspective Objection (the SPO); and since avoiding the SPO is the main motivation for endorsing an awareness requirement, the weak awareness internalist view is unmotivated. As a result, we should not endorse it. Some internalists think that PC provides a satisfying response to this sort of objection to weak awareness internalism. In this section I  argue that weak awareness PC (i.e., IPCw) succumbs to this horn of the dilemma and that attempts to show otherwise fail. But first, I need to say a little more about what the SPO is.

3.1. the subject’s perspective objection The SPO is an objection to externalist accounts of justification, which say that so long as the belief in fact does have a certain thing going for it (e.g., it’s reliably formed or it’s produced by properly functioning faculties), then the belief is justified, even if the person isn’t aware of this thing the belief has going for it. The basic idea of the SPO is this:

168

Seemings and Epistemic Internalism

SPO If the believing subject isn’t aware of what her belief has going for it, then from her perspective, it is an accident that the belief is true, in which case the belief isn’t justified (see BonJour 1985, 42–44; Bergmann 2006, 11–12).

In previous work, I haven’t examined in much detail how best to understand the SPO. In particular, I haven’t considered how to understand the notion of a belief ’s being an accident from the subject’s perspective. To remedy this, let’s consider the following five ways of understanding the claim that “It’s false that it’s an accident from S’s perspective that her belief B is true”: (I) It’s false that it’s an accident from S’s perspective that her belief B is true iff: S is aware of X and X in fact indicates B’s truth.16 (II) It’s false that it’s an accident from S’s perspective that her belief B is true iff: S is aware of X and S epistemically should believe that X indicates B’s truth.17 (III) It’s false that it’s an accident from S’s perspective that her belief B is true iff: S doesn’t believe that B is formed in an unreliable way (so that it’s unlikely to be true).18 (IV) It’s false that it’s an accident from S’s perspective that her belief B is true iff: it’s false that S epistemically should believe that B is formed in an unreliable way (so that it’s unlikely to be true). (V) It’s false that it’s an accident from S’s perspective that her belief B is true iff: S is aware of X and S believes that X indicates B’s truth (or at least conceives of X as being relevant to B’s truth or justification). The problem with proposals (I), (II), and (IV) is that they are compatible with both (a)  S’s not realizing that the specified condition is satisfied and (b)  S’s 16

Rogers and Matheson (2011, 64) mention, as a potentially plausible interpretation of the SPO, a view like this when they say: And even the very weakest version of internalism could avoid this sort of accidentality (and thus this “revised” SPO) simply by requiring some mere awareness of a relevant justificationcontributor, and nothing else. Non-accidentality would be achieved, according to such views, when one simply has some awareness of a relevant justification-contributor(s), as one does not in cases like Norman’s. 17

Rogers and Matheson (2011, 64) suggest that this sort of position is their own when they say: to be “an accident, from the subject’s perspective, that his belief is true” is for it to be the case that the subject lacks the kind of awareness of some justification-contributor that makes it reasonable for him to suspect that his belief is true.

This says that B is not accidentally true from S’s perspective when S is aware of some justificationcontributor and this awareness makes it reasonable for S to suspect that B is true, which is pretty much what proposal (II) says. 18 Rogers and Matheson (2011, 62) say: “Given that the proposition seems true to him, it is not surprising or accidental, from his perspective, that it is true.” This might be interpreted as saying that a belief ’s truth is accidental from a subject’s perspective just in case it’s surprising from that subject’s perspective (i.e., just in case the subject thinks it is unlikely to be true). This is very much like what proposal (III) says.

The Dilemma for Internalism

169

believing that B is formed in an unreliable way (so that it’s unlikely to be true). Consider (I):  a thing, X, of which S is aware could indicate B’s truth without S realizing it does. A similar point applies to (II): it might be that S epistemically should believe that X indicates B’s truth even though (due to some cognitive malfunction) S doesn’t realize that she epistemically should believe this. Likewise with (IV): it might be that it’s false that S epistemically should believe B is unreliably formed, even though (due to some cognitive malfunction) she has no idea this is false. And in all three cases, S might believe that B is formed in an unreliable way and is, therefore, unlikely to be true. If S did think B was unlikely to be true (and she also held belief B) and, in addition, S had no idea that the conditions mentioned in (I), (II), and (IV) were satisfied, then of course it would be an accident from her perspective that B is true, and this despite the fact that the conditions mentioned in (I), (II), and (IV) are satisfied. Hence, proposals (I), (II), and (IV) are implausible ways of understanding the condition of it’s being false that it’s an accident from S’s perspective that her belief B is true. Proposal (III) is more plausible: it says that it’s an accident from your perspective that your belief B is true if you believe that B is formed in an unreliable way and, therefore, that it is unlikely to be true. In effect, proposal (III) says that so long as you have no believed defeaters for your belief, then it is not an accident from your perspective that your belief is true. The problem here is that many externalists agree that a belief isn’t justified if it is in this sense an accident from the subject’s perspective that it’s true; this is because many externalists impose a no-believed-defeater condition on justification or warrant (see Bergmann 2006, 175–77). Hence, proposal (III) makes the SPO ineffective against all such externalist views. Moreover, BonJour, who introduced the SPO, clearly didn’t intend this understanding of being an accident from the subject’s perspective. BonJour (1985, 41) was arguing that it isn’t enough for the justification of your belief B that you don’t have any believed defeaters for B. Something more is required. It was in the context of explaining what this “something more” is that he introduced the SPO. So proposal (III) should also be rejected. Proposal (V) is more plausible than the others. With proposal (V), unlike with proposals (I), (II), and (IV), if you satisfy the stated condition, you clearly avoid its being an accident from your perspective that your belief is true. In addition, proposal (V) clearly goes beyond the externalism-compatible requirement mentioned in proposal (III). It is for these reasons that the SPO should be interpreted in accord with proposal (V)  or something more like it than like proposals (I)–(IV).19 Any attempt to avoid my dilemma for

19 If we interpret the SPO in accord with proposal (V)—i.e., so that to avoid it, a view must require (for justification) awareness of a truth-indicator or justifier conceived of as such—then it will be easy to show that weak awareness requirements cannot avoid the SPO. Is the fact that it will be easy to show this a problem? I don’t see why it has to be. The fact is that we have good reason to reject interpretations (I)–(IV) of SPO and to prefer an interpretation such as (V).

170

Seemings and Epistemic Internalism

internalism that relies on an interpretation of the SPO that is in accord with proposal (I), (II), (III), or (IV) should be considered a failure.

3.2. the weak awareness horn and ipc w Having said a little about how best to think of the SPO, I  can now explain why IPCw falls prey to my dilemma. Recall that, according to my dilemma, strong awareness versions of IPC are subject to a vicious regress whereas weak awareness versions of IPC are unable to avoid the SPO and, as a result, are unmotivated (since the main reason for imposing awareness requirements on justification is to avoid the SPO). Let’s consider why it is that IPCw is unable to avoid the SPO. According to IPCw, for a belief that p to be justified, the subject must have a seeming that p and the subject must be aware of that seeming that p; however, the subject needn’t conceive of that seeming as being in any way relevant to the truth or justification of the belief that p. The question before us is whether S could have a conscious seeming that p (without conceiving of it as being appropriately relevant to the belief that p) even though it is an accident from S’s perspective that her belief that p is true. I will argue that this is possible. If I’m right about that, then it’s false that requiring a conscious seeming (which is all IPCw requires) is enough to avoid the SPO. To make my case, it is helpful to consider first the feeling of jaw pain and arm pain that are indicative of a heart attack. Let us assume that these pains (when they are of the right phenomenal type) are in fact good indicators that one is having a heart attack. Pains that are good indicators of your having a heart attack might nevertheless not be recognized by you as being good indicators of that occurrence. Now suppose one believes, for very silly reasons, that one is having a heart attack (maybe one has just heard a song with the words “heart attack” and this leads to the belief). Such a person can also be aware of pains that are in fact good indicators of her having a heart attack without recognizing them as such. Consistent with the description of this case so far, we can add that it is an accident from this person’s perspective that her heart attack belief is true. This gives us a scenario with the following form: Scenario A S is aware of mental state X and S holds belief B (but not because of X). X is in fact a good indicator of B’s truth but S doesn’t recognize X as a good indicator of B’s truth. From S’s perspective, it is an accident that B is true.

In this scenario, X is the pain that is indicative of a heart attack and B is the belief that one is having a heart attack. Now imagine another example. This time, we can let X be the belief that it’s false that Jack and Jill are in the room while Jane isn’t. And we can let B be the belief that if Jack is in the room, then if Jill is in the room, then if Jack is in the

The Dilemma for Internalism

171

room then so is Jane. Could a person hold the latter belief for a silly reason (e.g., it sounded nice as a lyric in a song the person was writing) and also be aware of the former belief without recognizing that the former belief is indicative of the truth of the latter belief? Obviously this is possible, even though the two beliefs are logically equivalent.20 Thus, this example is also an instance of scenario A: here too we can say that it is an accident from the subject’s perspective that B is true. In both this somewhat complicated logical equivalence case and the heart attack case, a person is aware of something indicative of a belief ’s truth and yet it’s an accident from that person’s perspective that the belief is true. Intuitively, the belief ’s truth is an accident from that person’s perspective because the person has no idea at all that the truth-indicating thing she’s aware of is indicative of that belief ’s truth. In the cases just mentioned, we can see how sensible people might fail to recognize that the pains are indicative of a heart attack or that the complicated logical equivalence holds. But there are also instances of Scenario A in which an adult human would have to be seriously malfunctioning not to see that the mental state X was indicative of belief B’s truth. Here are two: (i) where X is the tactile sensation one has when grabbing a billiard ball in the dark and B is the belief that one is holding a hard smooth spherical object and (ii) where X is the belief that Jack is at the party and if Jack is at the party, then Jill is too and B is the belief that Jill is at the party. These beliefs (that one is holding a hard smooth spherical object and that Jill is at the party) can be held for silly reasons and one can be aware of things—such as the corresponding substitutes for X just mentioned in describing examples (i) and (ii)—that obviously (to us at least) indicate the truth of these beliefs even though, as a result of serious cognitive malfunction, one fails to recognize this. If the malfunction is serious enough, the modus ponens connection in example (ii) might be as opaque to the believer in question as the complicated logical equivalence is to some normal people; likewise, the tactile sensation connection in example (i) might be as opaque to a malfunctioning believer as the heart attack pain connection is to many normal people. Once we see that serious malfunction can cause connections to be opaque that would otherwise be obvious, I think we can see that instances of Scenario A can include more cases than one might initially think. In particular, I think it is possible, if the cognitive malfunction is serious enough, for there to be a case that is an instance of Scenario A in which X is a seeming that p and B is a belief that p—i.e., it is possible to hold the belief that p for a silly reason and at the same time to be aware of the seeming that p, all the while (because of severe malfunction) not recognizing any connection between the seeming that p and the truth of the

20 In this example, the belief that is X can be formalized as ~(p & q & ~r) and the belief B can be formalized as p → (q → (p → r)). Truth tables show that these two beliefs are logically equivalent.

172

Seemings and Epistemic Internalism

belief that p.21 In such a case, the believer will have a conscious seeming that p and yet it will be an accident from that person’s perspective that her belief that p is true. We acknowledge that Scenario A is possible in the case of the heart attack belief and in the case of the complicated logical equivalence. And we can see how there can be instances of Scenario A with unappreciated modus ponens evidence as well, if there is cognitive malfunction. The same thing is true with cases of severe cognitive malfunction where the connection between a seeming that p and the truth of the belief that p is unrecognized. Hence, it is possible for a person to have a conscious seeming that p even though it is an accident from her perspective that her belief that p is true, in which case it’s false that requiring a conscious seeming (which is all IPCw does) is enough to avoid the SPO. These considerations show that IPCw doesn’t enable one to avoid the SPO. I conclude, therefore, that my dilemma for internalism holds against all internalist versions of PC: IPCs leads to vicious regress problems and IPCw falls prey to the SPO (and, therefore, is unmotivated).

3.3. objections Some philosophers have argued explicitly that PC is a way of taking the weak awareness horn of my dilemma without being impaled on it. In this section, I briefly explain why I think they fail to make their case. I begin with a paper by Jason Rogers and Jon Matheson (2011) and then turn to a paper by Matthias Steup (this volume). One problem with the discussion by Rogers and Matheson, in their attempt to show that PC can avoid the problems that I say arise when internalists take the weak awareness horn of my dilemma, is that they say of seemings that (a) they are mere inclinations to believe and that (b) they needn’t be conscious (i.e., we needn’t be aware of them) for them to have their justifying effects (see Rogers and Matheson 2011, 60–61).22 In section 1.2, I explained why it is a mistake to equate seemings with inclinations to believe (it’s true that seemings typically cause such inclinations, but so do other things such as desires). And in section 1.5, I  explained why only a version of PC that requires conscious

21 Having a seeming that p can be compared to your being told by someone that p. We normally believe what we are told, but it’s possible for a person to be damaged in such a way that she has no tendency at all to believe what she is told—no tendency to take testimony that p as a reason to think p is true. Likewise, it’s possible for a person to be damaged in such a way that she has no tendency to take a seeming that p (with its characteristic feel—i.e., the feel of a state whose content reveals how things really are) as a reason to think p is true. Just as cognitive damage could remove any tendency to believe what other people tell you, so also it could remove any tendency to believe what your seeming states “tell” you (without making it the case that they don’t “tell” you things). 22 In saying that unconscious seemings can confer justification, Rogers and Matheson (2011, 62, n. 23) acknowledge that they are following Huemer (see my discussion of Huemer on this point in section 1.5).

The Dilemma for Internalism

173

seemings is relevant to my dilemma for internalism.23 However, Rogers and Matheson eventually consider a version of PC which requires that the seemings are conscious (63), so here I focus on that view when I consider what they say in support of the claim that PC avoids the weak awareness horn of my dilemma (though I take for granted my own understanding of seemings rather than their “inclination-to-believe” understanding, which I think is mistaken). One point Rogers and Matheson insist on is that the seeming makes a difference from the subject’s perspective, especially if the seeming is conscious (61 and 63). Now in one sense this is trivially true: if the seeming is conscious, the subject is aware of it and her perspective is affected (since one’s perspective includes what one is aware of). But to establish that a seeming makes a difference, in this trivial sense, to the subject’s perspective is a far cry from showing that a seeming that p makes it the case that the belief that p is not an accident from the subject’s perspective. And yet this is exactly what they claim: Given that the proposition seems true to him, it is not surprising or accidental, from his perspective, that it is true. If this is right, then (and this is the crucial point) one’s merely being in a state like the one described [i.e., having a seeming with the same content as the belief] is sufficient to make it the case that it is not an accident, from one’s own perspective, that one’s belief in the relevant proposition is true. (62)

They go on to say that the “believed proposition still seems true to the subject, and so is not accidentally true from his perspective”—as if its seeming true guarantees that it isn’t accidentally true from the subject’s perspective (63). They appear to take this implication to be obvious because they don’t defend it.24 But, for the reasons laid out earlier in section 3.2, it is a mistake to think that a conscious seeming that p guarantees that it is not an accident from the subject’s perspective that the belief that p is true. Being aware of something that is, in fact, a good reason for B doesn’t guarantee that it isn’t an accident from the subject’s perspective that B is true; for the subject might not recognize that the thing she’s aware of is a good reason (or any reason or indication at all) in support of B. This guarantee is still missing even when only severe malfunction would prevent a person from

23 Rogers and Matheson (2011, 60–62) draw attention to a view which says that the justification of S’s belief that p requires (i) that S is aware of some justification-contributor X and (ii) that X non-deviantly causes a seeming that p (which they say is an inclination to believe p) of which S needn’t be aware. And then they note that because of clause (i), this counts as an internalist view according to my definition of internalism. I won’t take time to discuss this view because I want to focus here on PC and this view is quite different from PC (in part because it relies on a mistaken account of seemings). 24 We need to remember here that, as I explained in section 3.1, there are different ways of understanding the notion of its not being an accident from the subject’s perspective that her belief is true. It may be that one of the problems with these claims by Rogers and Matheson is that (as I suggested in notes 16–18) they are working with proposal (I), (II), or (III) for understanding this notion of subjective accidentality, without recognizing that, as I noted in section 3.1, it is problematic to do so.

174

Seemings and Epistemic Internalism

recognizing that the reason is a good one. To obtain such a guarantee, it is also necessary that this sort of severe malfunction is absent.25 Let’s turn next to Steup (this volume, sec 2). He agrees that strong awareness internalism (including, presumably, IPCs) is hopeless insofar as it leads to vicious regress problems. His focus is on weak awareness internalism, where what is required is that one is aware of some justification-contributor without needing to conceive of it as such. But instead of endorsing IPCw, Steup endorses the following internalist reliabilist view: IR

S’s belief that p is prima facie justified only if (1a) (1b)

it seems to S that p, S’s memory data support the attribution of reliability to seemings of the sort mentioned in (1a), (1c) S’s memory data support the attribution of reliability to memory data of the sort mentioned in (1b),26 and (2) the seeming in (1a) and the memory data in (1b) and (1c) are internal mental states of S; it is not necessary that (3) S conceives of this seeming that p—or the memory data mentioned in (1b) and (1c)—as being in some way relevant to the truth or justification of the belief that p.27

It is tempting to interpret IR as a version of PC. For one thing, Steup (this volume) uses the term memory data in places where “memory impressions” is used in previous work (Steup 2004), suggesting he has in mind the same thing in each case. And he says that “what should count as evidence of reliability will be the mere memory impression, putative memory, or memorial seeming, of a positive track record,” which suggests that memory impressions just are memory seemings (Steup 2004, 408). In addition, he often speaks of the 25 Remarks just like the ones in this paragraph apply also to accounts of justification according to which it depends not on conscious seemings (as it does according to PC), but on direct acquaintance with justification-contributors, as it does according to Richard Fumerton (1995, 74). Rogers and Matheson (2011, 64) speak not only of seemings but also of direct acquaintance states and say that the latter will, of necessity, confer prima facie justification. My response to this (which is parallel to the response just given in the text in connection with PC) is given in Bergmann (2006, 29–32). In the same passage, Rogers and Matheson also cite approvingly Fumerton’s remark that direct acquaintance states bring intellectual satisfaction not had in connection with beliefs satisfying only externalist conditions on justification (they cite his 2007 but he makes a similar remark in his 2006, 189). For reasons I give in Bergmann (2006, 221–23, esp. n. 12), this remark of Fumerton’s is unconvincing. 26 This confirmation of memory data by memory data itself raises circularity worries, which Steup, this volume, sec 8, addresses. 27 This is my formulation. Clauses (1a) and (1b) correspond to (ii) and (iv) in Steup’s statement of internalist reliabilism (see Steup, this volume, sec 7). Clause (1c) is inspired by IR3 on p. 408 of Steup (2004). Clause (2) comes from Steup’s definition of internalism, where he says that he counts as an internalist in virtue of the fact that he thinks justification depends on one’s reasons, which are mental states (Steup, this volume, sec 1).

The Dilemma for Internalism

175

seemings confirmed by memory data as “seemingly reliable seemings” (Steup, this volume, secs 6–7), thereby suggesting that the memory data supporting the reliability of the first-order seemings are themselves memorial seemings. Moreover, given that he thinks seemings and memory data are internal mental states, it’s natural to think of them as states of which the subject is aware. Putting all that together, one might interpret IR as equivalent to the following view: IPCW*

S’s belief that p is prima facie justified only if

(1a) it seems to S that p, (1b) it seems (memorially) to S that she has had a track record of successfully forming true beliefs in relying on seemings of the sort mentioned in clause (1a), (1c) it seems (memorially) to S that she has had a track record of successfully forming true beliefs in relying on memorial seemings, (2) S is aware of the seemings mentioned in clauses (1a)– (1c)—i.e., they are conscious seemings; it is not necessary that (3) S conceives of this seeming that p as being in some way relevant to the truth or justification of the belief that p.

IPCw* is very much like IPCw. The main difference is that memorial seemings are required by IPCw*—in particular, memorial seemings of successful track records in relying on the first sort of seeming and in relying on memorial seemings. The main problem with IPCw* is that we can imagine another possible instance of Scenario A where B is the belief that p and X is the combination of the three seemings mentioned in clauses (1a)–(1c) of IPCw*. It is possible to hold the belief that p for a silly reason and at the same time be aware of the three seemings mentioned in clauses (1a)–(1c) of IPCw*, all the while failing (due to severe malfunction) to recognize any connection between those three seemings and the truth of the belief that p.28 Since that is possible, clauses (1a), (1b), (1c), and (2) of IPCw* can be jointly satisfied with its being an accident from S’s perspective that her belief that p is true. This is enough to show that IPCw* cannot avoid the SPO. However, in the end, these criticisms of IPCw* don’t apply to Steup’s work. Although it is tempting to interpret IR as being equivalent to IPCw*, we should resist this temptation. It’s true that Steup says that seemings whose reliability is confirmed by memory data are “seemingly reliable seemings.” But he makes it clear that this wording shouldn’t keep us from seeing that he thinks memory

28

See note 21 where this sort of possibility is briefly defended using a comparison with testimony.

176

Seemings and Epistemic Internalism

data are not themselves memory seemings (Steup, this volume, n. 19).29 What matters for justification, according to IR, is that we have memory data supporting the reliability of our seemings. We don’t need to reflect on these data or to have any seemings about them or their content; we don’t even need the potential to reflect on them in this way (though normally we will).30 Thus, the memory data mentioned in clauses (1b) and (1c) of IR needn’t involve seemings, actual or potential. As for whether IR requires the memory data mentioned in those clauses to be conscious, that is doubtful. We typically don’t think of memory data as essentially conscious; instead, we think of them as accessible (although they often aren’t as accessible as we wish they were— sometimes we are unable to retrieve them when we want to). Thus, Steup’s IR is not a version of PC. IPCw* is a view in the neighborhood of Steup’s IR that is a version of PC. But, as noted, it does no better than IPCw at avoiding the SPO. As for Steup’s IR, it too falls prey to the SPO. He claims it doesn’t: Embedded within [S’s] perspective is a body of memory data that make the truth of [S’s] belief highly likely. This, I claim, is sufficient for eliminating any accidentality that might otherwise adhere to the truth of [S’s] belief. (Steup, this volume, sec 5)

Steup seems to be assuming here something like (VI) It’s false that it’s an accident from S’s perspective that her belief B is true iff:  X is embedded in S’s perspective (perhaps in the way unretrieved memory data are so embedded) and X in fact makes B likely.

But the same points made in section 3.1 about proposals (I), (II), and (IV) apply to (VI) as well. Something in S’s unretrieved (and perhaps unretrievable) memory data could make B likely without S realizing or having any idea that it does, just as the jaw and arm pain could indicate a heart attack without the person feeling that pain having any idea that it does. The proponent of the SPO would say that the heart attack sufferer was not justified in believing she was having a heart attack if she had no idea that the jaw and arm pain indicated a heart attack. Likewise, the proponent of the SPO would say that S’s belief that p would not be justified by the mere fact that her memory data made p likely if S had no idea that she had memory data that made p likely (and she might have 29 Though, as he makes clear in the same footnote, he thinks memory data supporting reliability would normally ground memorial seemings, if we were to reflect on the matter. 30 Steup (this volume, sec 5) says that the “mere possession of the relevant memory data is enough” and that

S’s actual or potential abilities at epistemic reflection are irrelevant to this point. What matters is this: if, as a matter of objective fact, S’s memory data support the attribution of reliability, then they do so irrespective of what the subject’s potential for epistemic reflection happens to be. (Steup this volume, n.14, emphasis original).

The Dilemma for Internalism

177

no idea that this is so, even if she has such memory data). So although Steup’s view isn’t a version of PC, it is like PC in that it fails to avoid the SPO.

Conclusion My conclusion is that PC, insofar as it is an internalist position, falls prey to my dilemma for internalism. The Phenomenal Conservative must either require strong awareness of the seemings on which justification depends or settle for only weak awareness of them. If she insists on strong awareness, she will face vicious regress problems. If, as is more likely, she requires only weak awareness of seemings, then her view falls prey to the Subject’s Perspective Objection. As a result, attempts to use PC to avoid my dilemma for internalism are doomed to failure.

Acknowledgments Thanks to Trenton Merricks, Michael Rea, Jason Rogers, Matthias Steup, audience members at the Sixth Biennial Rochester Graduate Epistemology Conference where this chapter was presented in October 2010, Earl Conee (my commentator at that conference), and especially Chris Tucker for comments on earlier drafts.

References Alston, William. 2005. “Perception and Representation.” Philosophy and Phenomenological Research 70: 253–89. Bealer, George. 1999. “A Theory of the A  Priori.” Philosophical Perspectives 13: Epistemology, 29–55. Bergmann, Michael. Forthcoming. “Externalist Justification and the Role of Seemings.” Philosophical Studies. ———. 2006. Justification without Awareness. New York: Oxford University Press. BonJour, Laurence. 1985. The Structure of Empirical Justification. Cambridge, MA: Harvard University Press. Chisholm, Roderick. 1989. Theory of Knowledge, 3rd ed. Englewood Cliffs, NJ: Prentice-Hall. Conee, Earl. 2004. “First Things First.” In Earl Conee and Richard Feldman, Evidentialism: Essays in Epistemology, 11–36. New York: Oxford University Press. Crisp, Thomas. 2009. “A Dilemma for Internalism?” Synthese 174: 355–66. Fales, Evan. Manuscript. “Turtle Epistemology.” Presented at the Midwest Epistemology Workship VI at Indiana University, Bloomington, September 2012. Fumerton, Richard. 2007. “Review of Michael Bergmann’s Justification without Awareness.” Notre Dame Philosophical Reviews. http://ndpr.nd.edu/review.cfm?id=9104.

178

Seemings and Epistemic Internalism

———. 2006. “Epistemic Internalism, Philosophical Assurance and the Skeptical Predicament.” In Thomas Crisp, Matthew Davidson, and David Vanderlaan, eds., Knowledge and Reality:  Essays in Honor of Alvin Plantinga, 179–91 Dordrecht, The Netherlands: Springer. ———. 1995. Metaepistemology and Skepticism. Lanham, MD: Rowman and Littlefield. Huemer, Michael. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. ———. 2006. “Phenomenal Conservatism and the Internalist Intuition.” American Philosophical Quarterly 43: 147–58. ———. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Markie, Peter. 2009. “Justification and Awareness.” Philosophical Studies 146: 361–77. ———. 2005. “The Mystery of Direct Perceptual Justification.” Philosophical Studies 126: 347–53. Pryor, James. 2000. “The Skeptic and the Dogmatist.” Noûs 34: 517–49. Reid, Thomas. [1764] 1997. An Inquiry into the Human Mind on the Principles of Common Sense, ed. Derek Brookes. Edinburgh: Edinburgh University Press. Rogers, Jason and Jonathan Matheson. 2011. “Bergmann’s Dilemma:  Exit Strategies for Internalists.” Philosophical Studies 152: 55–80. Steup, Matthias. 2004. “Internalist Reliabilism.” Philosophical Issues 14: 403–25. Richard Swinburne. 2001. Epistemic Justification. Oxford: Oxford University Press. ———. Providence and the Problem of Evil. Oxford: Oxford University Press. Tolhurst, William. 1998. “Seemings.” American Philosophical Quarterly 35: 293–302. Tucker, Chris. 2011. “Phenomenal Conservatism and Evidentialism in Religious Epistemology.” In Raymond VanArragon and Kelly James Clark, eds., Evidence and Religious Belief, 52–75. Oxford: Oxford University Press. ———. 2010. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45.

{ part iv }

The Significance of Seemings within Specific Domains

This page intentionally left blank

{8}

Doxastic Innocence: Phenomenal Conservatism and Grounds of Justification Robert Audi In thinking about our beliefs, it is natural to focus on examples of what we believe from memory, or perceptually, or introspectively, or in an abstract realm such as mathematics. The beliefs I think of most readily seem true to me, and surely most people trust at least most of their own beliefs. Many of us tend to take our own beliefs and indeed those of friends and acquaintances to be rational. Are beliefs, then, in some sense innocent unless “proven” guilty?1 If we take doxastic innocence to be roughly equivalent to rationality, this idea—the doxastic innocence view—seems to reflect epistemological common sense, at least as applied to rational persons. For most of us, when we believe or presuppose that another person is rational and we know the person believes something, then if we have neither evidence regarding it nor any inclination to believe or disbelieve it, we tend to assume or presuppose that the belief is rational. Why might one hold the doxastic innocence view, at least for rational persons? One possible answer is that, in at least a large proportion of cases in which we consider a proposition we believe, it seems true to us: it presents an appearance of truth, in a phenomenal sense I will clarify. This chapter explores the connection between such propositional seemings and the rationality of our believing the propositions that, in the phenomenal way in question, seem true to us. The chapter will also explore a related common-sense view: that rational beliefs possess their rationality on the basis of the believer’s grounds for them. I will begin, then, with a sketch of grounds for believing. I will argue that although many kinds of grounds yield propositional seemings or may even be considered a kind of propositional seeming, not all grounds do so, and not all propositional seemings are grounds or otherwise suffice for justification.

1 Cf. the principle R. M. Chisholm (1982, 14) formulates as a plausible standard applying to this matter: “anything we find ourselves believing may be said to have some presumption in its favor—provided that it is not explicitly contradicted by the set of other things that we believe.”

182

The Significance of Seemings

1. Sources and Grounds of Belief The kind of ground of main interest in epistemology is normative: a basis on which believing the proposition for which it is a ground is (at least prima facie) rational, a status I take to entail the belief ’s having some degree of (prima facie) justification. There are, however, many kinds of grounds. In part because normative grounds are commonly also causal bases of beliefs they support, the term “ground” can be used to designate a causal basis, as where a belief is said to be grounded on prejudice. The term may also be used to designate sources of belief, such as perception. Historically, reason and experience have been considered the basic sources of justification and knowledge, and these have been taken to comprise perception (and here I, at least, include proprioception), consciousness, memory, and intuition (including intuitive reflection as we see it in such areas as logic and pure mathematics—more will be said about intuition in Section 5) (see Chisholm 1966; Audi 2001, 2010a).

1.1 basic grounds Perception, consciousness, memory, and intuition are basic sources of belief, of its rationality, and of its justification. Roughly, to call them basic in this way is to say that, first, the sources can yield (in beings with sufficient conceptual sophistication) (a) beliefs, which they commonly produce or sustain causally, and (b)  rational support or justification for beliefs (which does not entail a causal relation to them), and, second, that they can yield (a) and (b) without dependence on any other source of the beliefs in question or their rationality or justification. Seeing typeface, for instance, can produce and justify believing there is typeface, without dependence on my remembering anything: I need the concepts of, e.g., typeface, but it seems in principle possible to have that without even having a past. Normally, conceptual capacity is memorially stored, but I consider this a background condition for, not a partial producer of, beliefs involving the stored concepts. Other sources of belief include inference from one or more premises one believes. I regard inference as a secondary source given its dependence on premise-belief(s).2 Although I take these basic sources to yield knowledge in a great proportion of the cases in which they yield justified true belief, I do not take that to hold a priori and am not presupposing reliabilism about justification. In my

2 To be sure, inferences need only have cognitive premises, in the sense that implies having a truthvalued object, even if there is no premise that is believed by the person in question; and inference from such non-doxastic premises might produce a belief of the inferred conclusion. But this is not a typical case of inference or, more important, a case of a basic source of rationality or justification. The premise set would have to be rational (or have positive normative status) for the person in order to provide a ground for believing the conclusion.

Doxastic Innocence

183

view, it is a priori true that these sources yield justification but not that they yield knowledge (at least empirical knowledge). Most (but not all) of what I  say is consistent with reliabilism about justification, but for our purposes here there is no need to discuss the internalism-externalism controversy.3 Consider perception as a source of beliefs. I may at a given time perceive a table visually, tactually, and so forth for all five senses, in which case I have multiple grounds for believing there is one before me. My visual impression as of a table comes from a perceptual source and is a ground for believing that proposition. One could say that visual perception is my ground, but I prefer to consider the tabular visual impression the ground, in part because, even if it is an element in consciousness not produced by perception but rather by some hallucinogen, it may still justify believing there is a table before me. To be sure, we need not take perception to be a priori limited to the five senses or even to those together with inner sense, but we should not take hallucinating a table to entail perceiving one. I have granted that (as some writers on religious experience have claimed)4 perception is not thus limited. Moreover, it is at least not obvious that there cannot be another basic source of justification. The concept of justification surely does not foreclose the possibility. Still, if we set out to explicate that concept in detail, the classical basic sources are apparently ineliminable, and this chapter will illustrate their important role in understanding justification and its grounds. There are of course non-basic sources both of belief and of rationality and justification. We acquire beliefs from testimony (in the widest sense encompassing what people affirmatively say to us), but receiving testimony requires dependence on another source: perception. With inferential sources, we typically acquire beliefs on the basis of one or more other beliefs.5 When inference produces beliefs, then, it typically does not do so in a basic way, as with perception. Inference also apparently does not yield a basic ground for believing: it normatively supports its conclusion only if there is non-inferential support for its essential premises. It is possible, however, for elements in consciousness, which is a basic source of beliefs, to produce beliefs without also producing a basic ground for believing. Even dreamy musings about something might produce a belief, but they need not yield a ground for it; and of course brain manipulation might also be a direct source of beliefs. Sources of belief, then,

3 I have defended an internalist theory of justification (in, e.g., Audi 1988). This is consistent with an externalist account of knowledge, which, there and in later works, I have defended in a reliabilist version. 4 W. P. Alston (1991) does not put it this way, but he is not thinking of perception of God as a matter of the operation of one or more of the five senses. 5 Not all inferences are belief-forming, nor is every belief that arises from another necessarily based on any inferential process. I argue this (Audi 2010a, ch. 8) and also discuss it alongside the question of what constitutes inference (Audi 2006, 167–68).

184

The Significance of Seemings

may be basic or not; may or may not yield normative grounds; and may be experiential, like daydreaming, or not, as with brain manipulation. Perception is both a basic and an experiential source of beliefs, and it normally yields (normative) grounds for believing. It surely entails experiencing something, however faintly or fleetingly; and every perceptual experience is both a basis for belief formation and—at least normally—yields a ground for justified beliefs that are based on it and correspond to what might naturally be called its content.6 The same holds, I assume, for sensory experiences qualitatively identical to perceptual experiences.7 Each of the four basic sources has a distinctive phenomenology. For perception, there are qualia for each mode: there is something it is “like” to have a visual experience of print, an auditory experience of a Chopin nocturne, an olfactory experience of roses, etc. Here “like” is not comparative but qualitative. We also know what it is like to have the sense of remembering that something is so, as opposed to the sense of seeing its truth on considering it or of realizing it upon inferring it from propositions we already believe. As to introspection, if I have a pain in my knee or am silently humming a melody from Chopin, the ground of my believing that these things are occurring within me—the unpleasant quality of the pain and the mellifluous movement of the melody—is utterly plain. Regarding intuition, one paradigm for a priori intuition is the sense of intuitive plausibility that goes with comprehendingly considering the proposition that if spruces grow taller than do Japanese maples, then the latter grow shorter than the former.

1.2 non-perceptual experiential sources of belief As important as perceptual experience is in yielding beliefs and in producing propositional seemings of a kind that appear to constitute normative grounds, it is by no means the only kind of experience that does this. The point holds even if, as is essential, we include self-perception, such as awareness of the position of a limb. Memory is a different source of propositional seemings. There is of course recalling, say by imaging something one has seen. Something can seem true to me (and I can also believe it) on the basis of recalling it; for instance, asked where I saw a person, I may picture our meeting and it may only then seem

6 I assume, then, that “blindsight”—conceived as visual knowledge without any visual experience— is not seeing in the ordinary sense but rather a matter of a kind of cognition that arises from the relevant object while bypassing the conscious elements that go with seeing as ordinarily understood. (Indeed, subjects in blindsight experiments generally report that they see nothing.) 7 This is an internalist assumption I have defended elsewhere, as have many others. I do not pursue the matter since my main points about experience will be significant even if applicable only to perception and other kinds of experience conceived externalistically. Not every perception or sensory experience need yield belief, however, as is shown later.

Doxastic Innocence

185

to me that it was in Chicago. Another source of seemings is reflection on abstract matters or indeed, any subject. We also believe many kinds of propositions on the basis of reflection that provides a ground for them. Thinking about belief formation brings many options to my mind; as I ponder the topic, it is plain that one way I form beliefs is to pursue a question reflectively. Take this case. I have a twenty-foot pole saw which I would like to make five feet longer. I wonder if the pole will bear the weight of a five-foot extension. As I think of the flexibility and strength of the cedar to which I have affixed the saw, I come to believe that it will. Beliefs can also be formed from daydreaming. This category may overlap that of wishful thinking. Suppose I daydream about cutting high limbs with a twenty-fivefoot pole saw at once strong enough not to break and light enough to wield, and I come away believing I can successfully make one that long. If I have not thought about flexibility or unwieldy weight, and my desire produces a belief from the sheer attractiveness of the prospect, my belief may be at once experientially produced and not rational. In addition to wishful thinking as a source of belief, there are posthypnotic suggestion, subliminal advertising, and—perhaps frighteningly close in technological development—brain manipulation. These points show that even beliefs produced directly (in the sense of “non-inferentially”) by a kind of basic source of beliefs, including some beliefs produced by an experience, are not intrinsically even prima facie rational. There need be no rational support for such beliefs.

1.3 conferral versus preservation conditions Our concern so far has been with sources and (normative) grounds of belief. Sources of belief may yield grounds that confer rationality on it, but they need not yield such normative grounds. The clear examples of conferral are cases in which some event or episode, such as viewing something, produces a belief in the way perception normally does. Much more of what we rationally believe, however, is stored in memory. There is, then, a question of how to understand preservation of a belief ’s rationality or its justification. I will focus mainly on rationality as the more permissive normative status; but on the assumption that a rational belief, even if not justified overall, has some degree of justification, we can understand justification in relation to the degree or quality of grounding of the kind that confers rationality. I should also point out that I am not now concerned with knowledge, even if beliefs constituting it must be justified. Rationality and justification are important in their own right, and many epistemologists do not take knowledge to entail them.8 8 This relation has received less attention than, in my view, it should, but I believe my own effort to show that knowledge does not entail justification (see Audi 2010a, ch. 11), and the work of Dretske, Alston, and others at least makes it reasonable to assume that what grounds justification for believing a proposition need not even begin to ground knowledge of it.

186

The Significance of Seemings

Given the distinction between conferral conditions and retention conditions, how might a basic source view of rationality account for the apparent rationality of a belief at a time when it is either not in consciousness at all or is presupposed, as where one simply affirms the proposition in giving testimony or takes it as a premise for inference? In either of these cases, the belief may enter consciousness only at a time it is expressed and even then not be a focus of the believer’s attention. Here the epistemology of memory is crucial. It is plausible to hold that if we acquire a rational (or justified) belief at time t, and we memorially retain both it and the rationality-conferring grounds in the normal way until later, then at that later time it is still rational (or justified). This seems obvious if we can, through introspection or recall, remember the conferring experience; but what if we cannot? We may still rationally hold the belief if, at the time in question, we have an appropriate sense of remembering its propositional object, a kind of phenomenal memorial support for it. Having that memorial sense does not entail actually remembering it, since remembering is factive, and rationality and justification are not; but memory impressions can surely confer rationality on a belief. The memorial sense that p (where p is some proposition) provides a ground—to be sure, a defeasible basis—for believing p even if there never was a previous time at which rationality was conferred on the belief.9 I take such a memorial sense to imply a (propositional) seeming, a kind of phenomenal sense that p. When it does, that memorial seeming, in turn, implies a degree of justification. But not just any memorial or sensory experiences entail such seemings; seemings are propositional attitudes and presuppose conceptual capacities not entailed by the experiences alone. We should leave open that an infant or even a dog can have a sensory experience qualitatively just like mine but lack the conceptual capacity to understand the proposition that p. Perhaps a sensory experience as of (say) a tree is possible for a being with no concept of a tree, but it cannot seem to such a being that there is a tree before it.10 So far, I have been characterizing sources for beliefs and grounds for holding them, and have distinguished basic from non-basic sources of both. Seeing constitutes a basic source (one of the perceptual sources) of both, whereas inference is commonly a non-basic source of both, yielding belief from belief

9

I’ve argued this (Audi 1995) and it has been defended by other non-skeptical epistemologists. Cf. Chris Tucker’s case (2010, 530–31) that a perceptual sensation does not entail a seeming. In discussing Ernest Sosa’s treatment of the speckled hen case, Tucker rightly notes that a visual experience is not a seeming. I would add that if, e.g., it immediately seems to me that there are 25 speckles, my justification is defeated by my justification for believing I’d need to count them (though counting might still not suffice given phenomenal instability). Unlike Tucker, however, I do not find it clear that the idiot savant who seems instantly to “count” them has justification simply from a seeming that there are 48 speckles. But must the belief in question even be accompanied by a seeming? Can it not automatically arise as the savant is asked the number in question? 10

Doxastic Innocence

187

and (some) justification for its conclusion only given some for its premise(s). Compare seeing. My seeing typeface before me constitutes the causal basis of my visual belief that there is typeface before me, and it provides a basic normative ground for believing this proposition: my sensory experience of a certain visual kind. Call it an experience as of typeface. I take this locution to presuppose that normally we do see typeface in reading, but the locution also applies in the case in which we hallucinate. On my view, there are basic sources of normative grounds, and these grounds include certain phenomenal seemings. The next section will bring the position so far developed to bear on our main question, which concerns the general character of phenomenal conferring conditions, and especially propositional seemings, for the rationality of belief.

2. Propositional Seemings as Candidates for Normative Grounds To gain perspective on the normative role of propositional seemings, we should also consider normal (and so undefeated) testimony, much of which gives us the sense of credibility regarding what is said. Indeed, many propositions attested to seem true to us at the time, and, perhaps in part on this basis, testimony normally yields non-inferential, prima facie rational beliefs.11 If we think of all these cases—the four classical basic sources and testimony—as sources of grounds, and so of rationality, for beliefs, we are likely to be sympathetic with the common-sense view that if a proposition seems true to us in one of the phenomenal ways in question, then we have at least some basis for rationally believing it and, correspondingly, at least some degree of justification for it. Two qualifications of this pluralistic view of seemings are needed. First, not all belief formation is based on or even accompanied by such a phenomenal sense of credibility. Wishful thinking and associational belief formation (as with adoption of one’s social set’s presuppositions about manners) each seem to provide some examples of this point. Second, even with testimony, once we take a person to be credible—as with many in our lives whom we depend on for information—there is commonly at most a sense of the person’s credibility, either in general, as with many friends, or on certain occasions when they to speak, as with, say, people giving us driving directions. It is not the case that for every proposition contained in a routine narrative, we have any phenomenal sense of its being true. We can get so wrapped up in a story we are told that the information provided simply flows into our belief system.

11 In (Audi 2011) I’ve characterized undefeated testimony in some detail, and also argued that testimony-based beliefs are non-inferential and explained how, as such, they can constitute “basic knowledge” in one sense—knowledge not inferentially based on other knowledge—without being a basic source of knowledge (since testimony yields knowledge only through perception of the testimony).

188

The Significance of Seemings

Granted, something we are told that is inconsistent with a standing belief of ours may be immediately rejected or withheld. But this monitoring capacity we normally have—which exists dispositionally when not elicited—should not be parleyed into a phenomenal sense of the truth of what is being said. The fact that we would reject a proposition if it conflicted with standing beliefs—or things we presuppose or are suitably disposed to believe—does not imply that it seems true to us. We may conclude, then, that even if a phenomenal sense of a proposition’s being true often yields believing it, not all beliefs arise from such a sense. Some of our examples also suggest that even if that sense implies some degree of justification for one’s believing it, there are propositions we are justified in believing (such as many we remember) without our having that sense at every time we are in some way conscious of them. If, however—as appears well supported by the examples we have considered and others they suggest—a belief that p can be rational at a time without, at that time (or any other) seeming true to the person, it certainly does not follow that propositional seemings are not sufficient for the prima facie rationality of believing p.12 Call the view that, if a proposition seems true to one, then one has prima facie justification for believing it unrestricted phenomenal conservatism.13 It is a wide-ranging foundationalist view in implying that there is a great deal of non-inferential justification (though unlike a full-blooded foundationalism it makes no claims about inferential justification). The next section will consider just how close the position sketched in this chapter comes to this attractively common-sensical conservative view, which, on the assumption (also to be examined below) that what we believe typically seems true to us, is naturally taken to support the doxastic innocence view. The very name “phenomenal conservatism” indeed suggests the doxastic innocence view for the huge proportion of beliefs whose propositional objects seem true to the believer. It appears to be these beliefs, rather than seemings, that are conserved by the view. In that respect, the view might be better conceived as a kind of phenomenal liberalism. It will be obvious that I take it to be very plausible to maintain that, from certain kinds of impressions that p—sensory or other impressions that are both phenomenal and characteristic of a basic source—we have rational support

12 Note that Tucker’s phenomenal conservatism (2010, 529), which he calls dogmatism, provides a sufficient, not a necessary condition, for justification: “Necessarily, if it seems to S that P, then S thereby has prima facie (non-inferential) justification for P.” Tucker does not intend “dogmatism” to carry its usual implication of excessive confidence. Note too that in the text I do not take the view to imply that the seeming is a ground for belief as well as sufficient, but that broad grounding view will also be considered. 13 Tucker (2010), Michael Huemer (2001, 99), and others also hold this. Among the valuable critical discussions are Steup (2004, 402–25) and Markie (2005, 347–63), who criticizes phenomenal conservatism. (The “mystery” that Markie refers to in the title of that piece may be in part dispelled later in this chapter.)

Doxastic Innocence

189

(and some degree of justification) for believing p. A sensory experience as of p, for instance where one has a visual impression that leaves are falling, is one kind of impression that p. A memorial sense that p is a memorial impression; and so forth for introspective and intuitive basic sources. Is there any other kind of phenomenal seeming that p? There could be, as is affirmed by some regarding certain religious experiences (though often these are taken to be perceptual in some way). But is there a source-independent kind of seeming and, if so, must it provide a (normative) ground for believing? That there in fact are source-independent seemings that provide it is far from clear. There must be something qualitative about such a seeming—call it a sense of cognitive attractiveness. This may imply an inclination to believe the proposition, but it is not entailed by that.

2.1 the normative role of phenomenal seemings However many kinds of (phenomenal) propositional seemings there may be, we know of enough kinds to pursue the question what kind of justification a phenomenal seeming might provide. The notion used by proponents of phenomenal conservatism here is prima facie justification. Let me clarify a quite similar normative notion broad enough to be preferable for use in much of the discussion of phenomenal conservatism:  rational support. Call it prima facie support if you like; it is such that, if one has no ground or set of grounds of at least equal weight either in favor of not-p or some obvious contrary, or in favor of the proposition that one’s grounds for p do not support p—then, if one believes p on that basis (on the ground constituting the basis), one is rational in so believing, at least in the minimal sense that one’s so believing is not irrational. We might now fruitfully ask whether its seeming to S that p must provide rational support for S’s believing p. Here it is crucial to distinguish between what is necessarily true and what is broadly speaking a priori. The former does not entail the latter, and I am mainly concerned with the latter, on the ground that the issue here is philosophical and of the kind that we should be able to resolve by reflection. At the very least, what we can resolve by reflection on such a topic is of great philosophical interest. Now I have said that the notion of perception is a priori open-ended in allowing instances beyond the five senses and self-perception. I have also said that the list of classical basic sources of justification does not, a priori, exhaust the list of such sources of it. Still, I cannot see that there is a good a priori argument to show that every case of its seeming to one that p provides rational support for believing it. Let us recur to two cases. Can one’s wanting it to be true that p cause its seeming to one that p? It would appear so. We must ask, then, whether the seeming that derives from desire must confer rationality on the belief in question, especially when it has none of the properties that go with the phenomenal character that reflects the

190

The Significance of Seemings

basic sources. The clear cases of normatively supportive seemings go with the basic sources—or such non-basic sources as inference. Regarding inference, it is plausible to say that there is rational support for an inferential belief provided that, as is common, there is some rational support for the premise(s) and the inference is not patently weak, roughly in the sense that accepting it—as one implicitly does in inferring the conclusion from the premise(s)—is not unreasonable. Now suppose that wanting p to be true produces in me a sense of remembering p or an intuitive sense of its credibility. This would be like its yielding a good premise. It yields a rationality-conferring ground. Is it plausible, however, to think that a seeming produced only by the desire for p to be true need have any such normative power? I doubt that, but before proposing an alternative view I want to show why one apparent source of support for phenomenal conservatism may mislead. Recall the case of testimony. If I think of various things I believe on that basis—and their number and diversity is enormous—for the most part they seem to be true and believing them seems rational. Here propositional seemings appear to indicate rational support and one might find apparent support for both unrestricted phenomenal conservatism and the doxastic innocence view. But suppose we think of testimony-based beliefs not retrospectively but as they are formed. With a friend’s narrative that I believe in every detail, I am consciously occupied with the story being told, and my phenomenology need contain nothing in the way of a sense of truth for each proposition. This illustrates the important point that even non-inferential rational belief need not be based on a seeming. This, however, is a point the proponent of unrestricted phenomenal conservatism can grant. The more important question raised by testimony and other routes to belief-formation is whether they suffice for rational believing.

2.2 seemings as grounds versus seemings as products To answer this question we should note that, even if there is no need for each point in my friend’s narrative to be supported by a phenomenal seeming as it is made, it may still be true that when I  recall a point individually, I may then have the sense that it is true. This possibility, however, indicates something else, something also applicable to non-testimonial cases: once we believe something, it is common for considering it to yield a sense that it is true—we view it assentingly, as it were. It isn’t that one must say it to oneself in a positive tone or in some other way say “yes” to it. But one commonly feels a positive attraction, such that one’s natural inclination, if asked by a friend whether it seems true, is to answer affirmatively. Granted, considering a proposition we believe need not yield a phenomenal seeming and can even raise doubts about it; but with a great deal we believe, when we consider it (without any skeptical influence operating it) we tend to have a phenomenal impression of its truth.

Doxastic Innocence

191

We can now see that phenomenal conservatism should be appraised in the light of a distinction between support by seemings for propositions already believed and support for those not at the time believed (though of course products of a belief can be evidence of its truth, as where my belief that I am nervous causes my nervous symptoms). The thesis seems more plausible for a wide range (and I suspect a good majority) of propositions already believed. What we believe we tend to “reaffirm” upon considering it—though, on reconsidering it, we may of course find reason to reject or withhold it. Moreover, in part because of the role they play in guiding thought and action, beliefs supply premises and stand ready to be expressed as relevant to one’s conversation or other activity; so, if we did not by and large tend to reaffirm them, we would be disoriented. This role beliefs play goes well with the reaffirmational sense I refer to. It is important, then, that phenomenal conservatism not be considered more plausible than it is because it is generally well confirmed in the special cases in which we call to mind things we believe and take to be rational for us and then note our phenomenal state. The fact that seeming to be true is a property we find in most of our beliefs when we consider their propositional objects—or at least consider them in a certain way—is not a good reason to think that seeming true provides rational support to propositions we consider, which, at the time, we do not believe. When we already believe p, its seeming true might be only a product of our belief rather than a basis for holding it. Since the most common and most natural source of cases in which seemings appear to imply justification are instances in which the proposition in question is believed, this point must be kept in mind in assessing the data.

2.3 seemings as implying support for non-doxastic cognitions In addition to distinguishing kinds of causal grounds for belief and kinds of cases in which we experience them—for instance, on receiving testimony, on considering a proposition we do not believe, and on recalling one we do—we should also distinguish different kinds of cognition that may receive support from grounds of any kind. Suppose, for instance, that we distinguish belief from acceptance, where acceptance is (largely) a matter of a kind of disposition to avoid rejecting p, to use p in certain kinds of reasoning (at least to see where it leads), and to guide some other kinds of behavior by p, say to act on it where one must act on either it or not-p (we need not here give an account). Might acceptance of p be rendered rational by any phenomenal seeming that p, even if belief that p is not? It is true that other things equal, less is required to render accepting p rational than to render believing it rational. But it does not follow that just any phenomenal seeming that p rationally supports accepting p. Roughly, conferring rational support implies giving a certain kind of merit

192

The Significance of Seemings

to a proposition; where phenomenal seemings apparently do not confer it on believing they apparently do not confer it on acceptance either. Hoping that p is, in the convictional spectrum, still weaker than accepting p. Might unrestricted phenomenal conservatism apply to hoping? I doubt it, for the same reasons. (Another reason for doubt is that it is not rational for one to hope that p unless it is rational to have a positive attitude toward p’s being the case, but there is no need to pursue this here.) Still, the main point here is that, whatever the case for unrestricted phenomenal conservatism, it is stronger as applied to weaker attitudes than belief, such as acceptance, inkling, and conjecture. One further point is essential in this section. Since a supporting consideration can be outweighed or undermined by a contrary one, we must keep in mind something implicit in the notion of prima facie justification: the difference between a seeming’s not providing any support from its doing so but failing, because of a defeater, to provide support on balance. To illustrate, suppose I realize that I have been hallucinating squirrels in a spruce tree. Then, although a clear visual impression of one perched on a high branch supports the proposition that there is one there, I  should not either believe or even accept it. Perhaps, then, in the cases mentioned earlier in which seemings apparently do not to support their propositional object—say, with no ground provided by a basic source and mere wishful thinking as a causal basis, or where a proposition seems true only because already believed—there really is support but its normal effect is absent owing to defeat. This is an abstract possibility, but to appeal to it to defend unrestricted phenomenal conservatism one must indicate what kinds of defeaters are in question. The most plausible account of the relevant kind of defeat might well be this. When one rationally believes, or rationally should believe (say because one has strong evidence for it), something to the effect that there is no support from a basic source of justification, then (other things equal) a seeming is justificationally defeated.14 This might occur in all the cases we have considered that appear to be counterexamples to unrestricted phenomenal conservatism. Defeat may also occur where p is “isolated” from both those sources and cognitions that are supported, directly or indirectly, by them. I might, for instance, simply have a sense that it will rain tomorrow (a propositional seeming with this content), where I have neither the testimonial evidence of anyone familiar with weather patterns nor any evidence of my own, including memory of similar conditions preceding rain. The defender of unrestricted phenomenal conservatism may hold that defeat is what makes the seeming appear to imply no support; opponents may

14 The qualifier is needed to allow for the possibility that a seeming could be so epistemically powerful (in terms, e.g., of steadfastness, vividity, representational detail) that, if the rationality of the defeating element is minimal, there might still be minimal justification.

Doxastic Innocence

193

reply that there is no support in the first place. The common ground between these views is apparently the recognition of the basic sources as crucial both for understanding rational support in the first instance and for understanding its defeat. The former view suggests that beliefs appropriately related to seemings are innocent unless (as it were) credibly impugned; the latter view suggests that only certain seemings imply such a positive status. Both views are defensible, and one reason why neither easily emerges as preferable to the other is that they strongly tend to converge on just what sorts of beliefs have the non-inferential rational support both countenance.

3. The Place of Seemings in a More Restricted Foundationalist View There is at least one datum that is better explained by the restricted foundationalist view associated with the primacy of the basic sources than by unrestricted phenomenal conservatism, which I take to be a stronger but equally foundationalist view. On this more restricted view, justification for belief is conferred directly or indirectly, by a basic ground of justification, and basic grounds include the four wide-ranging kinds sketched earlier under the heading of “experience and reason” as basic sources. When we are asked to justify or support the rationality of something we believe, then if we cite a noninferential ground, we commonly indicate one involving a basic source in this range. This does not require naming the source, though that is usually easy for perceptual or memorial sources; but even people who do not use “intuition” and its cognates can explain their basis of belief, when it is an intuitive seeming, in ways that identify this kind of source. To illustrate some implications of the basic source view, suppose I reflect on how to justify a proposition I believe which I want others to believe. If I can say to myself only that it seems true to me and can find no story associating it with a basic source, I will very likely wonder if I have failed and would be widely taken to have failed. The normal response—and a kind whose relevance is arguably a priori—is to appeal to one of the five senses, to introspective consciousness, to memory, or to intuition. Thus, “I see that p,” “I feel the pain in my left knee,” “I seem to remember that p,” and “It’s obvious that p—just think about it” are common ways these basic sources are invoked in justification. By contrast, suppose that, when realizing someone may wonder why I believe p, I rationalize p by saying to myself, “It seems true to me,” where a proposition’s seeming true to me is constituted simply by a seeming that p, embodying the relevant kind of affirmative sense toward it. If there is nothing more to say that connects the seeming to a source like perception, I would surely have failed. To be sure, if p seems true to me and I cannot on reflection find any element of, say, memorial or sensory or introspective or intuitive support for it,

194

The Significance of Seemings

then at least if the matter is important to me, I have practical reason to explore whether p. But a reason to explore a proposition—though it does imply something to be said for it and so is normative—does not imply rational support for it. It is a practical reason—say, that the proposition is worth exploring—not an epistemic reason. The question raised by these and other points (and so far only implicit in our discussion) is whether seemings are a basic ground of rationality (or justification) or rather an important sufficient condition for it. Even unrestricted phenomenal conservativism is not usually formulated in a way that entails the former.15 The question is difficult to answer in the absence of a very detailed phenomenological account of seemings (something I  hope is forthcoming), but given the sketch I have offered, together with what I can discern in much discussion by defenders of phenomenal conservatism, I believe that seemings as such are either not grounds or not basic grounds, i.e., elements that confer normative status (here rationality and justification for believing) other than by appropriately representing some distinct ground. If this is so, it does not imply that seemings are unimportant or even eliminable from a full-scale theory of rationality. Let me sketch an account of how it may be so. Consider something highly plausible: that in virtue of having a clear and steadfast visual impression of a window before me, I have prima facie justification for believing there is one before me. This impression normally implies its seeming to me that there is a window before me. Why only normally? The visual impression, just as a sensory experience, need not imply the seeming, at least not a seeming as a kind of propositional impression. It is possible that a dog (or other perceptually discriminative being) might have the same visual impression and lack the conceptual sophistication needed for the propositional impression, which requires understanding the proposition. For us conceptually sophisticated perceivers, however, it is not only normal to have the propositional impression, but extremely likely that if we perceptually believe there is a window before us, it is because we are “appeared to windowly” and it thereby visually seems to us that there is a window before us. These two points make it natural to say, in such a case, that (1) we believe this because it seems to us that there is a window there and perhaps (2) we are justified in so believing by the visual impression and the associated propositional seeming. Given that visual perception is a source of belief, it is reasonable to assume (1). The “because” in (1) does not, however, indicate a ground, as does (2). The “because” may indicate a sufficient condition with some explanatory power, but it does not follow that the propositional seeming, as such and not integrated with a visual

15 This accords with the dogmatism formulation cited from Tucker, certainly with Huemer’s (2001, 99), and with other formulations in the literature, though in these and other writings on the topic a grounding role is often what seems to be attributed to the relevant seemings.

Doxastic Innocence

195

impression, is a conferring ground. Moreover, given the phenomenology of seeming—if we do not assimilate seemings to simply having appropriate sensory or other experiences characteristic of basic sources—it is possible to have the seeming without any such impression or other deliverance of a basic source. There, I have expressed doubt that the seeming is even sufficient for rationality, and I do not believe that by itself it is a ground. This is not the place for a detailed theory of how to determine what are the (basic) grounds of rationality for beliefs. But two points should indicate part of what any theory should take into account. First, one should reflect on how we can explicate the concepts of rationality and justification for belief without appeal to the basic sources I have cited. I doubt that one can proceed very far in this task without relying on an appeal to those four. Second, for rational perceptual beliefs (an important kind for the theory of normative grounds), one should reflect on the representationality of perception. This is most easily described for vision. My visual impression of windows includes what it is natural to call a phenomenal reflection of rectangular shape, of light outside, and of objects on the far side of the frame: I am “appeared to windowly.” This does not entail acquaintance with a sense-datum object; the point is that there is a kind of correspondence: qualities figuring in my phenomenal state correspond to properties I am justified in attributing to something external. A seeming, a kind of propositional impression, abstractly represents part of what my experience visually represents. This gives the seeming an important role but does not entail the power to confer rationality.16 Indeed, it may be that where we have a visual impression as of something’s having a perceptible property, F, and we believe that it is F but not even in part on the basis of its seeming that it is F, then the belief is less likely than otherwise to be based on the visual ground, hence less likely to be actually justified. Whether this is so is an empirical question, but our sense that the causal route from experience to beliefs goes through propositional impressions—or at least something it is natural to call a seeming—is sufficiently strong and pervasive to make seemings an element in any full-blooded conception of doxastic rationality. We can learn more about phenomenal conservatism by considering intuitions as elements both in yielding seemings and in grounding rationality. Before we do that, however, we can discover something more general about both seemings and their normative power from considering how they may confer or at least imply rationality in a non-doxastic domain.

16 I have provided supporting discussion of perceptual justification (Audi 2010a, chs. 1, 2) and also (Audi, 2010b), especially as concerns rationality and justification in for counterpart practical attitudes.

196

The Significance of Seemings

4. The Practical Analogy Desires are often compared with beliefs and are analogous in many ways. They are, for instance, intentional, can be rational or irrational, can be based on other desires in an inferential kind of way, and are crucial for explaining action. My hypothesis here is that where a very general normative thesis holds for beliefs, we should either find a close counterpart plausible for desires or find an explanation why not. Are desires innocent unless proven guilty? And is there a conative analogue of the sense of the truth of a proposition?17 Consider the idea that if something is attractive to one for a desirability characteristic—say as enjoyable, delicious, or lovely—then wanting it is prima facie rational. The idea is that wanting the desirable is conatively “veridical”: broadly, it is “correct,” in a way that is analogous to a belief ’s being true. To be sure, something’s being attractive to one is very close to wanting it; but perhaps no closer than a proposition’s seeming true is to believing it. In any case, even if this is not so, there does seem to be a parallel between the weak conation entailed and the inclination to believe that is apparently entailed by a proposition’s seeming true to one. Some might speak of a low degree of belief here, but I see no need to do that. There is a significant analogy in any case. Consider, then, the plausibility of conative conservatism—the view that if something is attractive for one, in the phenomenal sense designating a felt attractiveness—a conative attractiveness—then one’s wanting it is prima facie rational. Normally, just as one is cognitively attracted to a proposition on the basis of a ground such as a visual impression, one is conatively attracted to something for a (real or imagined) desirability characteristic of it, such as being delicious. If there are basic sources of rational desire analogous to basic sources of rational belief, it should not be surprising if this is plausible. And is it not prima facie rational to want to do something one takes to be enjoyable, to want to eat something one sees as delicious, to want to enter a conversation one views as nicely engaging, and—more clearly still, to want to avoid something one sees as painful? Here it is important that the sense of desirability, as we might call it, rest on a sense of a real or imagined desirability characteristic. If one has it simply because one already wants the thing in question, this would be like a proposition’s seeming true not, say, perceptually or memorially, but merely because one already believes it. Such a case may be rare, but in those instances it is not clear that seeming true by itself entails prima facie rationality of believing the proposition in question.

17 I have explored the analogy between desire (roughly, wanting) and belief at length (Audi 2001, Pt. II), and criticize the view of some Humeans, who, taking reason only “to judge of truth and falsity” (which do not apply to desires), do not take desires to admit of rationality at all (as opposed to grounding reasons to satisfy them). I here simply assume that desires do admit of rationality and irrationality.

Doxastic Innocence

197

What the practical case appears to show, then, is this:  phenomenal elements can confer rationality and indeed justification on desires and other conative attitudes, but conative attractiveness does not do so when detached from appropriate grounding, the kind analogous to grounds that go with basic sources, such as visual and memory impressions. It seems to me that normally both cognitive and conative seemings are so based, but they apparently need not be. Granted, if I  find myself attracted to a prospect, or if a proposition seems true to me, I may thereby have reason not to try to eliminate the desire or belief toward which these phenomenal elements incline me. That may give the impression that having the desire or belief would be rational on the basis of the element(s) in question. But this does not follow: that it would be rational not to try to eliminate a desire or belief does not entail that having it is rational. Indeed, where a desire is ineliminable, one is excusable for having it; and, where nothing significant is at stake, one can be excusable for having a desire that is eliminable, yet only with great effort. But, with desires as with beliefs, excusability, though a normative status, does not imply rationality.

5. The Epistemic Status of Intuition A major source of (cognitive) seemings is intuition, which I take to be a central element in our rational capacity (“human reason”). Intuition is crucial for understanding phenomenal conservatism and also of great general interest in epistemology because it is commonly considered, in one form, a kind of seeming, a basis of rationality (and justification), and a kind of non-inferential cognition—often a kind not in need of an independent ground. Its status is especially important for philosophy, which, like other fields, treats intuitions as data that, barring abnormalities, should be accounted for by definitions, general claims, and theories. Let us consider how intuitive seemings might figure in providing grounds.

5.1 intuitive seemings There surely are intuitive seemings. Some propositions seem true to us in a non-inferential but also non-sensory way—owing to our understanding of their content, or at least of their content seen in a certain context—and not owing to any premise(s) for them. Consider the proposition that if p is equivalent to q, then q is equivalent to p. This is luminously self-evident. Now turn to a normative proposition of a kind that helps to motivate phenomenal conservatism. If I have a clear and steadfast visual experience of printing before me, then I have some degree of justification for believing there is printing before me. With both propositions, there is what I call a sense of non-inferential credibility. This of course does not imply any epistemic belief about the propositions

198

The Significance of Seemings

or even having the concept of the non-inferential; the point is that considering them (perhaps merely entertaining them) evokes, on the basis of understanding their content, the sense that they are true. I take this sense to a normative ground for the proposition in question. The ground has its source in understanding, much as a sense impression has its source in perception. These examples may make it appear that intuitions have only a priori objects. But the objects of many intuitions are clearly empirical. Still, the visual experience principle I noted is a substitution instance of a more general one that is presumably a priori if true: If a person has a clear, steadfast visual experience as of x’s being F, then (assuming the person can understand the proposition in question) the person has prima facie justification for believing that x is F. Does the same hold for all intuitions? What of a linguistic intuition that locutions of the form of “Just because p doesn’t mean that q” are not good English. If good and bad English are matters of de facto practice at a time, this would seem an empirical truth. It appears that not all intuitions are connected with the a priori in the way a great many philosophical intuitions are. Suppose that there are intuitions of both empirical and intuitive kinds. Should they be regarded as providing rational support? If we deny this, we surely need an argument to do so. But how are we to judge the validity of argument forms apart from appeals to intuitions?18 Isn’t it our basic rational capacity (embodying the “faculty” of intuition) in virtue of which we see the validity of anything having the form of “If q then p, and q is true, then p is also true”? And what of “ultimate premises” in any argument against intuitions’ providing rational support? I can think of no plausible way to attack the conferring power of intuitions that does not itself rely on them. Here I must emphasize that even grounds deriving from basic sources do not confer indefeasible justification. Perhaps it is in part the association of intuitions with clearly self-evident propositions—their most prominent objects— together with the mistaken idea that the self-evident is unprovable, that fuels the impression that intuitions alone do not provide rational support for their propositional objects. Some people also decline, or are unable, to support their intuitions by arguing (even indirectly) for propositions they intuitively believe; this may also give the impression that intuitions (as seemings) are more like prejudices than like rational responses to propositions, scenarios, and cases.

5.2 the scope of intuitive seemings and the normative force of intuitive reflection As some writers use “intuition,” intuitions simply are intuitive seemings. But certain beliefs equally deserve the name “intuition.” That there is reason to 18 Timothy Williamson (2007) and Walter Sinnot-Armstrong (2011) take views quite different from mine on the status of intuitions in philosophy.

Doxastic Innocence

199

keep promises is, for many, both an intuition and a belief. To be sure, doxastic intuitions are non-inferential and commonly have some degree of rational support from intuitive seemings. But they are not necessarily rational or justified (nor are phenomenal seemings).19 Doxastic intuitions may but need not arise from intuitive seemings; but they are apparently supportable by them where, on comprehendingly considering p, it seems true, in the non-inferential way instantiated when, apart from any sense of premises for it, a proposition seems true to one. Given the kinds of considerations just noted, I  suggest that phenomenal conservatism does hold for intuitive seemings. But of course these seemings manifest the operation of a basic source: reason. It should not be thought that all its deliverances are a priori propositions, as we have seen. Many instances are, but many are not, including many in ethics. There is no precise answer to the question what proportion of such deliverances is a priori. In philosophy, however, the proportion is apparently high. Even if it is not high, if intuitions confer some degree of normative support, that is significant. They would be data that theoreticians must consider. Since many claims in philosophy depend on intuitive seemings, if phenomenal conservatism holds for those seemings, it has important epistemological implications. One might, for instance, think that skepticism can be readily answered. Phenomenal conservation and the basic source view converge in implying that, insofar as we may rationally believe we have intuitive seemings, we are entitled to believe we have some degree of justification for some propositions. But may we believe the former or that we ever have justification on balance? Moreover, whatever we say about our possession of rational and even justified belief, knowledge is factive, and the case for our having any significant kind of knowledge is not obviously advanced by these points. Still, in a world with little or no knowledge and a rational basis for enough important propositions, we could be intellectually in the right, even if objectively in the wrong.20 The idea that phenomenal conservatism holds for intuitive seemings conceived as a kind of deliverance of a basic source, suggests another plausible view in the range of positions on the normative power of seemings. Suppose one restricts phenomenal conservatism to propositions one considers, in the sense that one focuses on them and entertains their content clearly. If a proposition then seems true to one, and seems so other than on the basis of one or 19 Seemings can be abnormal or natural or deviant, but not, I think, rational or irrational. This is one reason why, like sensory impressions, they are candidates to confer rationality and justification. If, as a kind of occurrent phenomenal impression, they do not admit of these, they cannot stand in need of them. Cf. Ernest Sosa’s (2007, 49) view that “an intellectual seeming” is “evaluable” in terms of “justification and unjustification.” 20 Might this be such a world, where “might” expresses epistemic possibility? I think not. Could it be, in terms of sheer logical possibility? I am aware of no sound argument showing that it could not be, and have treated skepticism in the light of some of the points made here (Audi 2010a).

200

The Significance of Seemings

more premises, does this not imply some degree of justification for it? I find this “reflectional” phenomenal conservatism more plausible than the unrestricted version. But it still appears too strong in not ruling out (perhaps among others) cases in which one already believes p wholly on the basis of a desire that it be true and lacks an intuition (or other basic ground) that it is true. Still, considering a proposition clearly provides an opportunity to think of counterevidence or of an origin or basis that undermines justification. Thus, p’s seeming true upon such consideration represents passing a kind of test. Reflectional phenomenal conservatism may well not be too strong if restricted to certain classes of propositions, above all certain a priori propositions. Exploring this and other restricted versions of phenomenal conservatism would, however, require more space than I have here. The point needed now is that, if this and other restricted versions of the view hold, that may yet be because of the pervasive operation in cognitive life of one or more of the basic sources. It may also show that their operations range more widely, and are manifested more subtly, than has generally been noticed.

6. Conclusion The view that beliefs are innocent unless “proven” guilty—that beliefs by their very nature have a measure of rationality unless there is some mark against them—seems too strong. It might, however, hold (empirically) for all the beliefs rational persons in fact have. Indeed, there is some plausibility in the view that, much as what seems right to a morally sound person has a plausible presumption in its favor, so what seems true to a rational person—one who, we might say, has the virtue of rationality—has at least some such presumption. Even if that presumption is true, beliefs as such do not intrinsically have support. If, however, they are grounded, as many beliefs are, on a propositional seeming that manifests a basic source of justification, as where they are based on a memory impression, they thereby have a measure of rationality. They are at least prima facie rational. As to whether seemings by themselves imply some degree of rational support for their propositional objects—the central claim of unrestricted phenomenal conservatism—I cannot see that the weight of argument sustains the view. Certain restricted versions, such as intuitional conservatism, may well be true and may yield results virtually equivalent extensionally to those implied by the moderate foundationalist view just described, which posits a plurality of basic normative grounds provided by experience and reason. Even if there is such an extensional equivalence, it does not appear that seemings alone are, as on unrestricted phenomenal conservatism, basic grounds of rationality or justification for beliefs. They may themselves be based on such grounds—as I think they typically are—and will then imply some degree of justification. Those grounds, moreover, may normally

Doxastic Innocence

201

produce seemings whose phenomenal qualities reflect them. If, as it appears from our discussion, seemings alone do not suffice for justification, they may still be essential elements in a full-scale epistemology. They are a natural and common response to the basic grounds of justification.

Acknowledgments This chapter has benefited from discussion in seminars at the Jagiellonian University in Krakow and the University of Notre Dame, from comments by an anonymous reader, and from comments by Scott Hagaman and, especially, Chris Tucker.

References Alston, WP. 1991. Perceiving God. Ithaca: Cornell University Press. Audi, Robert. 2011. “Testimony as a Social Foundation of Knowledge,” Philosophy and Phenomenological Research. DOI: 10.1111/j.1933-1592.2011.00525.x. _____. 2010a. Epistemology: A Contemporary Introduction to the Theory of Knowledge, 3rd ed. New York: Routledge. _____. 2010b. “Practical Reason and the Status of Moral Obligation.” Canadian Journal of Philosophy, Supplementary Volume 33: 197–229. _____. 2006. Practical Reasoning and Ethical Decision. New York: Routledge. _____. 2001. The Architecture of Reason:  The Structure and Substance of Rationality. Oxford: Oxford University Press. _____. 1995. “Memorial Justification.” Philosophical Topics 23(1): 31–45. _____. 1988. “Justification, Truth, and Reliability.” Philosophy and Phenomenological Research 49(1): 1–29. _____. Chisholm, Roderick. 1982. The Foundations of Knowing. Minneapolis:  University of Minnesota Press. _____. 1966. The Theory of Knowledge. Englewood Cliffs, NJ: Prentice-Hall. Huemer, Michael. 2001. Skepticism and the Veil of Perception. Lanham, MD:  Rowman and Littlefield. Markie, Peter. 2005. “The Mystery of Direct Perceptual Justification.” Philosophical Studies 126(3): 347–43. Sinnot-Armstrong, Walter. 2011. “An Empirical Challenge to Moral Intuitionism.” In Jill Hernandez, ed., The New Intuitionism, 11–28. New York: Continuum. Sosa, Ernest. 2007. A Virtue Epistemology, Vol. 1. Oxford: Oxford University Press. Steup, Matthias. 2004. “Internalist Reliabilism.” Philosophical Perspectives 14(1): 403–25. Tucker, Chris. 2010. “Why Open-minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Blackwell.

{9}

Agent Centeredness, Agent Neutrality, Disagreement, and Truth Conduciveness Michael DePaul 1. Introduction Since first encountering discussions of the epistemology of disagreement, I’ve thought it is possible that one rationally continue to believe in the face of disagreement. I  have believed this because I  hold that the way things seem or appear plays an important role in determining the epistemic status of beliefs. Why would recognizing the epistemic significance of seeming lead one to think rational disagreements are possible? No one denies that when there is some epistemically relevant difference between S and T it’s possible for S to rationally believe that p and T to rationally believe that not p, e.g., when they have different evidence concerning the topic in question. Hence, discussions of the epistemology of disagreement attempt to control for epistemically relevant differences by focusing on cases where S and T have the same evidence, are epistemic peers, etc. If how things seem is epistemically significant, attempts to control for epistemically relevant differences face an obstacle. If S believes that p and T believes that not p, it is reasonable to suppose that, all things considered, it seems to S that p (in some way) while it seems to T that not p (in some way).1 Can we control for this difference? We might imagine that each reports to the other their respective seemings, but this does not obviously put S and T in the same epistemic situation. Compare a case where S and T discover they disagree about p, and as a first step they explain to each other the evidence they have bearing on p. Assume that both are trustworthy and mutually recognized as such. Plausibly, S thereby acquires the evidence T had but she lacked and T acquires the evidence S had but he lacked. They will then both have to take the new evidence 1 The parenthetical “in some way” is meant to acknowledge that p might seem true in various ways, e.g., via perception, memory, inference, or rational intuition.

Disagreement and Agent Centeredness

203

into account, either by finding some sort of defeaters for the new evidence or making suitable adjustments to the conclusions they previously drew. It is at least tempting to think that if S and T have exactly the same evidence, and there are no other epistemically relevant differences between them, it would be rational for them to believe the same things. So it seems easy enough to control for different evidence—you just have the parties lay out all their evidence. It is not so easy with seemings. Explaining to someone else how things seem to you typically does not make things seem that way to the other person as explaining one’s evidence might plausibly be thought to provide the other party with that evidence. After S and T tell each other how things seem to them, it is entirely possible that it still seems to S that p, and it still seems to T that not p.2 T will of course now know that it seems to S that p, and S will know that it seems to T that not p. But if how things seem is epistemically significant, reporting seemings cannot control for this epistemic variable—it is possible that a difference remains, one ripe to explain why it might be rational for the parties to the disagreement to persist in their disparate beliefs. That’s obviously not the end of the story. Few hold that its seeming to one that p is sufficient, all by itself, for rationally believing p. So there are various ways one might argue that, in the kinds of cases on which the disagreement literature focuses, it would not be rational for S and T to hold different beliefs regarding p in spite of there being epistemically relevant differences between them. But my aim here was only to explain why the starting position for one who takes seemings seriously would be that rational disagreement is possible, not to argue that such a one could not be forced out of this position. Michael Huemer also holds that seemings are epistemically significant. Indeed he has vigorously defended phenomenal conservativism: PC if it seems to one that p, then in the absence of defeaters, one thereby has some justification for believing that p. (22–23)3

He has recently provided an interesting framework for approaching questions about disagreement. He distinguishes two positions: 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. 2 Of course when T learns that it seems to S that p, this might cause it to seem to T that p as well. S might be a reliable eyewitness to some event that T missed, for example. But even in this case, there will be a difference in how things seem to S and T, since S will have a memorial seeming along with a wealth of associated memorial experiences, while T will lack these. 3 Parenthetical page references in the text all refer to Huemer (2011).

204

The Significance of Seemings

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. (18)

I’ll refer to these as AN and AC, respectively, and use corresponding abbreviations for agent neutral and agent centered theories, norms, views, etc.4 Huemer defends AC. Noting the disjunction in the final clause of the definition, let’s distinguish two versions. Soft AC takes the first disjunct. It allows that S’s knowing that T satisfies C confers some prima facie justification for S to believe that p, just not as much as T gets for p by actually satisfying C. According to hard AC, it is possible for S to know that T satisfies C and yet have no prima facie justification at all for p. In the end, Huemer defends hard AC. Here is some terminology that will be convenient. C is a justification conferring condition (JC-condition) for p just in case, necessarily, for any person S, if S satisfies C, then S has some prima facie justification for p in virtue of satisfying C. We can then say that a JC-condition is fundamental according to an epistemological theory if it is the condition (or one of the conditions) identified by the fundamental principle (or principles) of the theory. Thus, having been produced by a reliable cognitive mechanism is the fundamental JC-condition for process reliabilism, being a member of a coherent system of beliefs is the fundamental JC-condition for coherentism, and so on. Huemer argues that PC is most naturally understood as AC. PC is an austere doctrine, identifying only one fundamental JC-condition:  seeming. As already noted, frequently S knows that it seems to T that p, without its seeming to S that p. So unless some supplemental principle is added to PC, there is no reason for phenomenal conservatives to think that S’s knowing that T satisfies even their fundamental JC-condition for p automatically provides S with some prima facie justification for p. Huemer also surveys some leading contemporary epistemological theories and concludes most are naturally construed as AN with the remainder open to being understood as AN or AC. It therefore behooves Huemer, as an advocate of PC, to defend AC. He also presents a disagreement scenario and argues that AC allows that the parties could disagree in the scenario while AN does not allow disagreement. I have several aims in this chapter. First, although I  agree overall with Huemer about AC and AN, and the role of seeming, I believe there are easy

4 It may go without saying, but I want to make it clear at the outset that I understand “has some justification for believing that p” in PC, and related phrases in AN and AC literally, as what is often called propositional justification. I do not take this to mean that S justifiably believes that p. S might have a justification for believing that p and not believe p, not base her belief in p on the justification she has for p, and so on.

Disagreement and Agent Centeredness

205

arguments for AC over AN. Huemer cannot use the easy arguments because he thinks the issue must be decided by focusing on an idealized situation. I maintain this is a mistake, and that there are also problems with the idealized disagreement scenario he focuses on as well as the idealized scenarios the broader disagreement literature uses. I explain all this in section III. In section IV, I seek to bolster Huemer’s case for AC and against AN by describing and then undermining a significant motivation for AN. Section II provides an overview of Huemer’s argument in preparation for sections III and IV. But before closing this introduction, I want to relate Huemer’s discussion of AC, AN, and PC with the ongoing discussions of the epistemology of disagreement. The lively and growing literature on disagreement focuses on what I’ll call the typical idealized disagreement scenario:5 TIDS

S and T are epistemic peers and know that they are. S believes p and T believes not p. S and T discover their disagreement. S and T share all the evidence they have that is relevant to p.

The question is whether S and T could rationally continue to believe what they did before discovering their disagreement. It is not immediately apparent AN or AC would be relevant to the established discussion of disagreement; they concern the level of justification one gets for p when one learns another person satisfies a JC-condition for p. The established discussion makes no mention of JC-conditions and is concerned with rationality, not justification. Viewing disagreement from the perspective of PC, however, brings the relevance of AC and AN into focus. For one thing, PC claims that a proposition that seems true to a person will thereby have only some prima facie justification. The distance between this level of justification and the rationality at issue in discussions of disagreement is not large, as it might be if justification according to PC were, e.g., akin to Plantinga’s “warrant”—whatever it is that distinguishes knowledge from mere true belief. More significantly, PC asserts that one has some prima facie justification for p simply in virtue of its seeming to one that p. Hence, assuming (plausibly) that it is common knowledge that when someone believes a proposition it seems to them (in some way) that the proposition is true, when the parties in TIDS become aware of their disagreement, according to PC each typically learns the other satisfies a JC-condition since T learns it seems to S that p and S learns it seems to T that not p. Hence, for phenomenal conservatives, whether AN or AC is true has an important bearing on discussions of TIDS. For if AN is true, then necessarily when the parties in TIDS learn of their disagreement, both will acquire prima

5

Richard Feldman and Ted A. Warfield (2010) provide an excellent introduction.

206

The Significance of Seemings

facie justification for believing the proposition the other party believes, and in fact prima facie justification that is equal to that had by the party who originally believed the proposition. But if AC is true, the parties in TIDS could each learn the other believes something different without acquiring as much, or any, prima facie justification for believing what the other party believes. Whether AN or AC is true will not automatically settle what the parties in TIDS would have justification for believing, all things considered, since they only concern prima facie justification. But the choice will clearly have an impact. Moreover, it is plausible that what one has or lacks justification for, as understood by phenomenological conservativism, would be intimately related to what it would be rational for one to believe. For example, if it seems to S that p, and nothing that S believes or that seems true to S defeats the prima facie justification p has as a result, it apparently would also be rational for S to believe p. More work is obviously required to show exactly how justification as understood by phenomenological conservativism is related to rationality as used in the disagreement literature. But they are at least near neighbors, and I must confess I find it hard not to think of them as interchangable. I therefore expect any results attained by considering disagreement in terms of AC, AN, and PC could be applied, perhaps with minor adjustments, to the more familiar discussion of what it would be rational to believe in TIDS.

2. Huemer’s Case for AC and the Possibility of Justified Disagreements One theme Huemer’s discussion of AC and AN develops is an analogy between ethics and epistemology. AC ethical norms evaluate actions of the same type differently depending upon whether the agent performs the action or some other person does. A standard illustration involves an agent forced to choose between killing one innocent person himself and the perpetrator killing five innocent people. AN ethical theories assign equal negative values to each killing, regardless of who kills, and so conclude that the agent should kill the single innocent person. On AC ethical theories, in contrast, it matters whether the agent kills or someone else does, so they can allow that it would be wrong for the agent to kill one innocent person himself even though this would result in fewer killings. AC theories might agree with common sense morality about this case. But notice that one could not support AC moral norms over AN norms by arguing that it would be wrong to kill a single innocent person oneself to save five innocent persons from being killed by someone else because one couldn’t be certain the five would be murdered if one doesn’t kill that one person. To zero in on the choice between AC and AN norms, Huemer asserts that we must ask what’s right in the case where one is certain that five will be killed if one doesn’t kill the one.

Disagreement and Agent Centeredness

207

By parallel reasoning he claims, 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. (19)

Huemer presents an idealized disagreement scenario designed to highlight the difference between AN and AC while satisfying this condition on a good test of AC. HIDS 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. (19)

According to Huemer, AN theories rule out disagreements in HIDS while AC theories allow disagreements. I think he is reasoning as follows: Let CS1, . . ., CSn be the JC-conditions satisfied by S and CT1, . . ., CTm the JC-conditions satisfied by T. Let pS1, . . ., pSn be the propositions for which S has some prima facie justification as a result of satisfying CS1, . . ., CSn, respectively; similarly for T and pT1, . . ., pTm. Because S and T each knows with certainty the other’s epistemically relevant states, S will know that T satisfies CT1, . . ., CTm. And T will know that S satisfies CS1, . . ., CSn. Because of this knowledge, according to AN, S will have prima facie justification for pT1, . . ., pTn and T will have prima facie justification for pS1, . . ., pSn. HIDS does not explicitly stipulate that S and T believe what they are justified in believing; it stipulates that they do not believe what they are not justified in believing. So let’s add that S and T believe what they are justified in believing. We still fall short of the conclusion that they could not disagree, since we have only shown they would have prima facie justification for the same things. A prima facie justified proposition is all things considered justified just in case it is not defeated. So S and T might be justified in believing different things, even though they are prima facie justified in believing exactly the same things, if they could have different defeaters. Perhaps it goes without saying, but let’s also stipulate that defeaters, defeaters of defeaters, etc., must all come from what they are prima facie justified in believing. This brings us closer to the conclusion that the same propositions will be all things considered justified for S and T—and to the very same degrees—and hence that they will believe the very same things.

208

The Significance of Seemings

But for all that is now packed into HIDS, we still fall short of the conclusion that S and T could not disagree. Recall Huemer’s stipulation: “neither party makes any procedural error in forming beliefs.” This stipulation allows us to conclude that S and T could not disagree only if correct procedure is not permissive in the sense that someone beginning with a given set of prima facie justified propositions and defeaters who followed correct belief-forming procedures might end up being all things considered justified in either believing, disbelieving, or witholding some proposition. If correct procedure is permissive, then even though S and T begin with exactly the same propositions being prima facie justified and they each work correctly from this starting point, they might end up with different beliefs. Since the question of whether rationality is permissive is a significant point of contention in the debate regarding the epistemology of disagreement, we should not take it for granted that correct belief-forming procedures are not permissive. I do not have space to examine the question further, however, so I will just explicitly stipulate that correct procedure is not permissive. I believe this suffices to conclude that if AN is true, S and T would believe the same things in HIDS. However, even in HIDS so augmented disagreements between S and T are possible on AC theories. According to such theories, if S knows that T satisfies CT1, . . ., CTm, S will not automatically have the same level of prima facie justification for pT1, . . ., pTm as T has in virtue of satisfying CT1, . . ., CTm. And similarly for T and pS1, . . ., pSn. Hence, S and T might not begin with the same propositions prima facie justified to the same degrees, so even if they proceed correctly from their starting points they could end up with different propositions being justified. Hence, they could disagree. Huemer does not directly argue against AN or for AC. He instead makes a proposal about what motivates AN, and resistance to AC, and explains how an AC position might accommodate the motivation. As he explains, Agent-centeredness 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.” (24)

Huemer responds that AC positions do not always privilege the believer’s own experiences. This happens only in special circumstances, where the believer has no information about the other person’s cognitive capacities. If S has no reason of her own for believing p and knows nothing at all about T except that T satisfies some JC-condition for p, e.g., that it seems to T that p or that T is having a visual experience as of p, then hard AC theories will deny that S has any prima facie justification for p. But we are rarely or never in such a situation. We know a great deal about other people, even those whom we

Disagreement and Agent Centeredness

209

do not know personally. We usually know that they are cognitively normal, that they have reasonably good memories, sensory faculties, and so on. So, in virtue of this background knowledge, when we come to know that such a person satisfies some JC-condition for p, this does provide us with a prima facie justification for p. The point of contention between AC and AN, then, is not so much whether knowing that others satisfy JC-conditions provides one with justification, but how this happens. AN theories treat such knowledge as a kind of basic source of prima facie justification, on a par with satisfying a JC-condition oneself. One needs nothing beyond the knowledge that another satisfies a JC-condition for p to have a prima facie justification for p of one’s very own. AC theories instead see the knowledge that someone else satisfies a JC-condition for p as supplying one with a reason that, in conjunction with one’s knowledge about that person, or normal people in general, can provide one with a prima facie justification for p. This is an important, fundamental difference between the two views, but it is a rather “theoretical” difference. In normal circumstances, where we have lots of background knowledge supporting the reliability of others, both kinds of theories hold that learning what JC-conditions another satisfies provides one with prima facie justification for believing the relevant propositions. So contrary to initial appearances, it will not be possible to construct a scenario that is close to real life where AN theories entail that learning another person satisfies a JC-condition provides one with a prima facie justification and AC theories disagree. Huemer writes, 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. (28)

While this line of argument may show that AC will not have counterintuitive implications in all kinds of normal situations, it highlights an important question the combination of PC and AC faces:  why does one need background information about the reliability of another person’s seemings to gain prima facie justification for p from the knowledge that p seems true to that person but one does not need background information about one’s own seemings for one to have prima facie justification for p when it seems to one that p? Why can we simply trust our own seemings? Huemer’s answer is familiar: “a fundamental self-trust is necessary to avoid a debilitating skepticism” (28). Ultimately, our beliefs are based on our own appearances, on how things seem to us, so if we had to defend the reliability of these appearances before we could rely upon them, we would be stuck. We could never be justified in believing anything.

210

The Significance of Seemings

I do not want to dispute any of what Huemer has to say about what he takes the main motivation for AN to be. Indeed, I thought he was exactly right to turn his attention to more normal circumstances and explain the various ways in which learning how things seem to another person, in conjunction with our background knowledge, can provide us with prima facie justification. I  even agree with him about the necessity of a fundamental self-trust. But I believe there is an easier argument for AC that Huemer could have used if he had not focused on what I think are excessively idealized cases. And I also think that Huemer may have missed another significant motivation for AN that should be undermined. I devote the next two sections to these issues respectively.

3. The Easy Argument for AC and Idealizations Upon first reading the definition of AC I thought, “It’s dead easy to defend.” Let C be “is considering a proof of p that is self-evidently valid with self-evident premises.” Suppose p is nothing extraordinary, e.g., the solution to some mathematical problem for which a prize has been offered. It’s something simple, along the lines of one of De Morgan’s laws. Case 1 S satisfies condition C, that is, S is currently thinking about what she recognizes to be a proof of p. S is an ordinary person and there is no funny business. Case 2 T, another perfectly ordinary fellow, remarks to S that he is contemplating a proof of p. Nothing fishy this time either; in particular, S has no reason to doubt T’s testimony.

Setting aside extreme skeptical doubts, C is clearly a JC-condition for p. In Case 1, S satisfies C. Given that the circumstances are perfectly ordinary—S has no reasons for doubting that p is true, or that it is the sort of thing that could be proven, or for thinking that she is prone to hallucinate constructing simple proofs, etc.—S would have some justification for p. Indeed, S would have a very high level of justification, and perhaps even maximal justification, for p. What about Case 2? There’s nothing funny going on here either. So there’s no reason to deny that S would know that T satisfies C. It seems quite clear that S would thereby have some prima facie justification for p as well. S would probably even have knowledge level justification for p. But surely S would not have as much justification for p as in Case 1. So much for AN. Huemer cannot accept this argument. As we saw earlier, he holds that we must decide between AC and AN by focusing on idealized cases where S knows for certain that T satisfies a JC-condition for p. He argues for restricting debate to such idealized cases by analogy with the debate between AC and AN ethical norms. It is not obvious that this restriction should be carried over into epistemology.

Disagreement and Agent Centeredness

211

One thing that idealization can do is exclude extraneous factors. But here it is not at all clear that what’s excluded is extraneous. Particularly those with internalist sympathies might be inclined to think that knowing one satisfies a JC-condition oneself can provide more justification than knowing someone else satisfies a JC-condition precisely because one has special access to JC-conditions in one’s own case—that one can be certain one satisfies such a condition in one’s own case. I must confess, I think this line of thought leads to the heart of the matter. If that’s right, Huemer’s idealization does not throw out something like a confounding variable; it excludes consideration of an essential factor. The modal status of the idealization is also problematic. How could one possibly be certain that someone else satisfies a JC-condition? I cannot come up with a single fundamental JC-condition anyone has actually proposed that one could be certain another person satisfies. One cannot even be certain that one satisfies many fundamental JC-conditions oneself. The fundamental JC-conditions defended by externalists provide the clearest examples. One could arguably be certain that one satisfies some of the fundamental JC-conditions of internalist epistemological theories, but one clearly could not be certain that others satisfy such conditions. So as I see things, Huemer’s idealization forces us to decide between AN and AC while assuming the truth of something that’s clearly impossible. According to the standard semantics for counterfactuals, counterfactuals with impossible antecedents (counterpossibles) are all trivially true. If these semantics are correct, it will be true that (1) if S were certain that T satisfies C, S would not thereby be justified in believing p to the same degree as T

and also that (2) if S were certain that T satisfies C, S would therby be justified in believing p to the same degree as T.

This result clearly doesn’t decide between AC and AN. Things are obviously more complicated. We often idealize, and often to good effect. Idealizations may well involve considering situations that are impossible in some sense, so such impossibility does not always prevent us from drawing conclusions about the situation. Here are two easy examples: (3) If a human being were to count all the natural numbers, taking one second for each number, then it would take an infinite amount of time to finish. (4) If a human being took one second to count to one and then counted each other number in one half the time it took to count the preceding number, she would finish counting the natural numbers in two seconds.

A human being could not possibly, in various senses, satisify the antecedents of (3) and (4), yet we can see perfectly well that the consequents follow. And we do

212

The Significance of Seemings

not see this by seeing that (3) and (4) are trivially true because their antecedents are necessarily false. For we can just as clearly see it is false that (5) if a human being took one second to count each natural number, she would finish counting them in under a year,

even though it has an antecedent that is necessarily false. So where does that leave us? Sometimes we can consider an impossible state of affairs and say perfectly well what would non-trivially follow. But in other cases we cannot say what would follow from an impossible situation. Consider: (6) If one were to draw a round square, then it would look.. . (7) If a kilo of pure spirit were confined in a liter canister, then the pressure would be . . .

We simply have no idea how to fill in the ellipses in (6) or (7). I can’t say anything very insightful about distinguishing the cases where we can determine what follows from an impossible scenario from those where we cannot. I might venture that it has to do with how thoroughly we can conceive of the impossible situation before whatever it is that accounts for the impossibility comes up and starts to unravel our conception and whether we are able to get far enough along in the process to be able to say something about the particular consequent in question before the unraveling begins. In the case of (6), we can begin thinking about sitting down with a compass and straight edge. Maybe we draw a circle, and then a square that contains it, and then another circle that just contains the square. Or perhaps we start out with a square instead. But then what? We just can’t get far enough along to have any idea what the final product might look like. It’s even worse with the kilo of pure spirit—we’ve no idea whatsoever how to begin. But in (3), (4), and (5) we can ignore the fact that a human being is doing something no human being could possibly do; we just think about the necessary sub-tasks, each being done in a certain amount of time, and then add up to see how long the entire job would take.6 My examples are deliberately extreme. What about Huemer’s idealization? It is easy to say, “Well, just suppose S is somehow absolutely certain that T satisfies some JC-condition for p.” But for us to make a reasonable judgment about whether this would provide S with any justification for p, we need more details. We need enough to form a fairly specific view of how satisfying the JC-condition justifies p for T and then an equally detailed view of what it would be like for S to be absolutely certain that T satisfied the JC-condition so 6 An anonymous referee suggested another possible explanation: there are counterfactuals in the near neighborhood that do not have impossible antecedents that we can clearly see to be true or false. For example, we can consider God or some powerful counting machine not limited by the physical laws of our universe undertaking the tasks in the antecedents of (3), (4), and (5). We then just ignore human limitations and judge (3), (4), and (5) to have the same truth values. But when it comes to (6) and (7), we cannot even imagine God performing the required task.

Disagreement and Agent Centeredness

213

we could consider whether and how this certain knowledge might justify p for S. I doubt we can form a sufficiently clear and detailed picture of having the required but impossible certain knowledge. Just try thinking about the specific case of most interest to phenomenal conservatives: how does one become absolutely certain that things seem a certain way to another person?7 One might be tempted to use the analogy with ethics: “The cases are similar, and we have no problem conceiving of the agent in the moral case being certain that five will be killed unless he kills one.” But are the cases similar? We need not suppose the agent in the moral case is absolutely certain; it’s good enough if the agent is just very sure, or certain enough “for all intents and purposes.” So in the moral case we are not required to consider a situation that’s obviously impossible. We might instead bracket epistemic issues in the moral case. We can stipulate a description of the facts of the case and then consider what the agent objectively ought to do. In the epistemological case we can’t bracket questions about how the agent is certain the other person satisfies some JC-condition, since this will very likely be relevant to whether the agent thereby has some justification for p. Let’s now turn our attention to TIDS. How, I wonder, could S and T possibly share all their evidence in the requisite sense? On my view, how things seem to them is a crucial part of their evidence. As I explained in the introduction, S and T might report to each other how things seem to them. S might tell T such things as “It seems to me that p” or, if p is the right sort of proposition, “It looks to me as if p” or “It sure smells like p,” etc. But this hardly puts in T’s possession the full richness of how things seem to S. And elements of how things seem that cannot be conveyed may well be epistemically significant. Perhaps seemings are not exactly “evidence,” but they play a similar role; and in any case they are epistemically relevant and hence something that S and T should have to share. From my point of view, then, the entire discussion of whether S and T could rationally disagree in TIDS is under a cloud because it is so difficult to form an adequate conception of some crucial elements of this idealized scenario. Such concerns are amplified when it comes to HIDS, since it stipulates that each party has complete and certain knowledge of all the epistemically relevant

7 An anonymous referee responds: “Why can’t we just imagine that an angel tells you that things seem some way to someone, and you have the highest possible degree of trust that the angel tells the truth?” But this would only provide psychological certainty, not certainty in the epistemic sense. We might become very sure that a highly trustworthy angel is feeding us information, but no matter the strength of the evidence for this proposition our justification would fall short of what we have for paradigm examples of propositions that are certain, such as that I exist or that 1 + 1 = 2. I might also note that I do not think an appeal to counterfactuals in the near neighborhood will help. It might be the case that God could know with certainty that T satisfies a JC-condition for p, but God would already be certain that p along with everything else, so God’s certain knowledge that T satisfies a JC-condition for p does not impact God’s justification for p.

214

The Significance of Seemings

states of the other. If the only thing that was epistemically relevant were evidence, and evidence were limited to propositions, e.g., those a person believes, justifiably believes, or knows, then we might be able to get a serviceable handle on two people reporting all their evidence to each other. But such reports clearly would not provide certain knowledge of the other’s evidence. And if more is epistemically relevant than such evidence, as I think likely, it’s obvious on its face that two people could not possibly have perfect knowledge of one another’s epistemically relevant states. More significantly—since we can sometimes form a good enough conception of something impossible for us to see what would follow—I don’t think we can even begin to conceive of what it would be like to have perfect knowledge of another’s epistemically relevant states. Consider what Huemer explicitly mentions: sensory experiences, memories, intuitions. Sensory experiences alone seem to be so rich and fine-grained that it’s doubtful we can have complete knowledge even of our own experiences. What would it be like to have such complete knowledge of ourselves while at the same time having complete knowledge of the other person’s experiences? Even supposing we could have certain knowledge in our own case—which is plausible but not beyond question—exactly how can we attain certainty regarding the other person’s sensory experiences? What cognitive mechanism, justificatory process, or form of argument could we use to attain such certainty? I just don’t know where to begin. Including memories and intuitions only makes matters worse. Maybe intuitions are sufficiently few and phenomenologically thin that they don’t make much trouble. But we have many memories and these memories can be very rich in content, so including them will be more problematic. And sensory experiences, memories, and intuitions may not exhaust the epistemically relevant states. In the end, I fear, we’d need to include a person’s entire mental life. But this is just mind-boggling. Personally, I  have such a tenuous grasp of what having perfect knowledge of another person’s epistemically relevant states would be like that I  have no confidence in any epistemological judgments I might make about HIDS—supposing I could manage to make some such judgments. Even though I  think these concerns are significant, I’ll set them aside in the next section. But let me first reiterate that my worry is not simply that we are being asked to make judgments about scenarios that are, strictly speaking, impossible. I do not balk, e.g., at being asked to suppose that both parties in HIDS always reason correctly without overlooking anything, even though this is probably impossible. Even though impossible, I can form a sufficient conception of this supposition to make various reasonable judgments about what follows. My concern is more specific: we are being asked to consider scenarios that are not only impossible, but that we cannot attain a sufficiently clear partial grasp of to be able to make reasonable judgments about them.

Disagreement and Agent Centeredness

215

There is something else that concerns me about the high degree of idealization involved in TIDS and HIDS, a worry that applies even to elements of the scenarios that are not impossible. Basically, I’m worried that these idealizations enable a kind of bait and switch. We frequently disagree with others, often about the most important things: morality, politics, religion, etc. We frequently don’t have much regard for those with whom we disagree:  we consider them ignorant or mistaken about the facts, biased, illogical, and the like. But there are also many cases where we don’t think this. We think those with whom we disagree are well informed, unbiased, and reasonable. Hence, the disagreement literature strikes us as hugely significant. It is addressing something we encounter all the time. And much of it pushes a disturbing conclusion: we cannot rationally disagree with one another.8 I suspect we are lured in by worries about disagreements we actually face in real life, and then our attention is diverted to idealized scenarios that we never encounter. I worry that the idealized scenarios are so far removed from any situation we might actually face that the conclusions we draw about these scenarios reveal next to nothing about what it’s reasonable to believe when one disagrees with someone in real life.9 But I’m just complaining; giving voice to suspicions I’ve not substantiated and don’t have the space to substantiate. Let me close this section by reiterating the main results:  (i)  there is an easy, perfectly good argument against AN; (ii) Huemer excludes this argument by limiting our attention to idealized cases which are objectionable; and (iii) more generally, Huemer and much of the disagreement literature has focused on idealized cases which face similar objections and likely don’t have much to do with real-life disagreements.

4. Another Motivation for AN Undermined In this section I will present and undermine what I believe is a powerful motivation for AN, thereby bolstering the case Huemer has already presented. Let’s begin with a quick story about epistemic justification that is familiar to any student of epistemology. It may not be entirely accurate, but it is repeated often enough that it likely reveals something important. According to the classic modern conception of justification, held by such figures as Descartes and Hume, a justification must guarantee the truth of the justified proposition. (I’ll henceforth use SM-justification, SM-justified, etc., to identify this strong modern conception of justification.) Unfortunately,

8 9

Given how much they disagree, this result may be especially troubling for philosophers. Nathan King (2011), Bryan Frances (2010), and others have raised similar concerns.

216

The Significance of Seemings

SM-justification is so strong it leads to a sweeping skepticism; our reasons or grounds for belief at most ensure the truth of simple mathematical, logical, and conceptual truths and some propositions about our own minds. Descartes thought he could secure SM-justification for much of what we ordinarily believe, but his efforts collapsed. Hume more willingly embraced the skepticism that SM-justification entails. The contemporary conception of justification is weaker:  a guarantee of truth is not required for propositions to be justified in this sense. (I’ll use WC-justification, WC-justified, etc., for this weak contemporary conception of justification.) It therefore promises to avoid the broad skepticism SM-justification yields. But if it is to be a conception of epistemic justification, WC-justification must still be truth-connected. As Laurence BonJour explains in a very influential and widely cited passage, The distinguishing characteristic of epistemic justification is thus its essential or internal relation to the cognitive goal of truth. . . . It is this essential relation to truth which distinguishes epistemic justification from other species of justification aimed at different goals. Of course, at times the degree of epistemic justification may fall short of that required for knowledge and thus fail to make it likely to any very high degree that the belief in question is true. But any degree of epistemic justification, however small, must increase to a commensurate degree the chances that the belief in question is true. (Bonjour 1985, 8)

So, to put things in somewhat old-fashioned terms, it is a conceptual truth that SM-justified beliefs are true, and a conceptual truth that WM-justified beliefs are likely to be true. SM-justification comes close to entailing AN. If S knows that T satisfies some condition C, which is a SM-justification conferring condition for p (henceforth, SMJC-condition), then C guarantees the truth of p and S knows that T satisfies C. Knowing that T satisfies C, how could S fail to have some prima facie SM-justification for p? Only two possibilities are apparent: either (i) S does not realize that C is a SMJC-condition for p or (ii) S is not aware that SM-justification guarantees truth. It is not obvious that S would fail to have some justification for p even if (i) or (ii) obtained; what seems clear is that if these obtained, S would not recognize the justification she has for p as such. But set that aside. What happens in HIDS? Consider the following element of the specification: 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. (19)

Disagreement and Agent Centeredness

217

If S failed to recognize that C is a SMJC-condition or that SM-justification guarantees truth, S would be guilty of a rather large oversight: S would have failed to grasp or mobilize some crucial elements of the concept of SM-justification. As a result, S would have in his or her possession a ground that guarantees the truth of p, but failed to make use of it. So in HIDS, where S and T make no such mistakes, when each comes to know what SMJC-conditions the other satisfies, each gains extremely strong reasons for believing what the other is SM-justified in believing, and believes accordingly. Notice that this argument for thinking that the parties in HIDS could not disagree does not turn on anything having to do with AN or AC. The premises concern only the nature of SM-justification and the specification of HIDS. So if one is working with SM-justification, whether one accepts AN or AC, one would have to say that there will be no disagreements in HIDS. But as noted previously, few epistemologists now accept SM-justification. Does anything special follow from WC-justification regarding the choice between AN and AC? One might be tempted to run arguments similar to those described, making suitable adjustments from “true” to “likely to be true.”10 Suppose T satisfies some condition C, which is a WCJC-condition for p; then p is likely to be true. If S knows that T satisfies condition C, how could S fail to have at least some prima facie WC-justification for p? Only two ways are apparent: either S does not realize that (i) C is a WCJC-condition for p or that (ii) WC-justification makes truth likely. But while S’s failure to be aware of (i) or (ii) might be significant with respect to whether S WC-justifiably believes that p, or whether S recognizes her justification as such, it is not relevant to whether S has some WC-justification for p, and the latter is our concern. If we focus steadfastly on the question of whether S’s learning that T satisfies a WCJC-condition for p provides S with some prima facie WC-justification for p, it is quite tempting to answer affirmatively. For whether S is in a position to make use of it in belief formation, S does know something, viz., that T satisfies C, which does make p likely to be true. And S would not need to acquire additional empirical evidence to recognize this as a reason for believing p is likely to be true. The case is not like one where Sherlock knows the perpetrator left size 12 footprints while his suspect wears size 8 shoes, but does not know that on the morning of the murder his suspect bought several pairs of thick wool socks and size 12 boots. Everything S needs to know follows from the concept of WC-justification. So some might think there is a pretty good case for thinking that if S knows that T satisfies a WCJC-condition for p, S will have some prima facie

10 To save space I shall ignore the question of whether WC-justification excludes or allows disagreements in HIDS.

218

The Significance of Seemings

WC-justification for p. The only question is how much. We needn’t pause to answer that question, since this is already enough to rule out hard AC.11 Let’s take stock of the motivation for accepting AN I’ve suggested. I  began by distinguishing SM-justification, which guarantees truth, from WC-justification, which guarantees only likely truth. I then offered an argument, which I take to be sound, for thinking that SM-justification leads to AN. Since few contemporary epistemologists employ SM-justification, I turned my attention to WC-justification and presented a suitably adjusted version of the argument from SM-justification to AN. The adjusted argument does not conclude with AN but with the denial of hard AC. So it leaves us with either AN or soft AC. But perhaps we can safely ignore soft AC, as it seems an unstable position balanced between AN and hard AC. I claim that this argument—perhaps in the form of the vague thought that justified beliefs must at least be likely to be true, so when one knows that someone else has a prima facie justification for believing p, this is a prima facie reason for thinking p is likely to be true, and hence one has some prima facie justification of one’s own for p—provides significant motivation for AN. Another contributing factor might be the sense that what held true for SM-justification should still hold true, in suitably weakened form, for WC-justification. If I’m right that this argument provides a significant motivation for AN, there’s a problem. We all learned that if E implies H then E conjoined with anything else still implies H, but if E only makes H probable, it is not necessarily the case that H is probable on E conjoined with something else. In spite of being frequently used, the following natural way of expressing the truth connection inherent in WC-justification is potentially misleading: TC- If S satisfies a WCJC-condition C for p, then p is likely to be true.

The problem is that taken literally the consequent expresses an unconditional probability; it asserts that p is likely to be true full stop, no matter what. If the consequent makes sense understood in this way and TC- is the correct way of expressing the truth connection implicit in WC-justification, then the argument I’ve presenting for thinking that WC-justification excludes hard AC will be hard to resist. But the truth connection implied by WC-justification is more accurately expressed in this way: TC+

If S satisfies a WCJC-condition C for p, then p is likely to be true given C.

When the truth connection involved in WC-justification is correctly understood, the argument against hard AC collapses because of the simple truth

11 Huemer (2011, 21) provides a similar argument to explain why coherentism might be committed to AN. I have merely generalized Huemer’s argument to any WC account of justification.

Disagreement and Agent Centeredness

219

cited previously. While it may be true that p is likely to be true given that T satisfies some WCJC-condition C, it could well be highly improbable that p is true given the conjunction of T satisfies C with some other true proposition. Cases are easy to construct: suppose it appears to T that he is looking at a red ripe tomato. Having such a visual experience is plausibly a WCJC-condition for the proposition that the tomato is red ripe. Moreover, given that T is having such an experience, it’s likely that the tomato is red ripe. But suppose T is in the grocery store under those funny lights they use in produce departments. It is not likely that the tomato is red ripe given that it looks red ripe to T and T is looking at it under such lights. Now suppose S is the grocer admiring the lights she just installed to make her pale, tasteless winter tomatoes look as red as vine ripe tomatoes from an August garden. S observes T admiring a tomato. S knows how that tomato looks to T. Does S thereby gain some prima facie reason for thinking that the tomato is likely to be red ripe, and hence some prima facie WC-justification for believing it is red ripe? Certainly not! Given what S knows about the lights she just installed and the quality of her produce, it is not at all likely the tomato is red ripe no matter how deep red it looks to T, and it is hard to see why we would think S would have any prima facie WC-justification at all for believing the tomato is red in virtue knowing it looks that way to T. One might be tempted to try saving the argument against hard AC by appealing to defeaters:  “Whenever S knows that T satisfies some WCJCcondition C for p, S knows something on which it is likely that p is true. This is all that is required for S to have some prima facie reason, and hence prima facie WC-justification, for believing p. The fact that S knows other facts such that p is not likely to be true given the conjunction of those and T’s satisfying C only shows that S knows things that defeat the prima facie WC-justification provided by her knowledge that T satisfies C. This poses no difficulty, however, since the whole point of attributing only prima facie WC-justification was to recognize the possibility of defeat.” This effort to save the argument against hard AC just does not seem right. Think again about the grocer. She’s gone to some lengths to ensure that her tomatoes look a certain way to her customers so they will believe the tomatoes are deliciously ripe when they are not. Given this, it seems bizarre to claim that when she knows one of her tomatoes looks just that way to one of her customers, she thereby has any WC-justification at all for believing the tomato is deliciously red ripe, even some merely prima facie WC-justification that is immediately defeated. If the grocer example isn’t convincing, think of a variation on the Cartesian demon; this demon knows a bit of epistemology and only occasionally works his mischief. The demon knows that C is a WCJC-condition for p and knows that given C, p is likely to be true. He occasionally induces his victims to satisfy C when p is false, because he gets a kick out of WC-justified falsehoods. This

220

The Significance of Seemings

is such an occasion. Surely the demon knows that his victim satisfies C—he’s the one who intentionally caused his victim to satisfy C. But the demon clearly would not be WC-justified in believing p, and it would be extremely odd to claim that this is because he has some prima facie WC-justification for p in virtue of knowing that his victim satisfies C, but this prima facie WC-justification is defeated by his knowledge that he caused the victim to satisfy C when he knew that p was false. It makes much more sense simply to deny that such a demon would have even some prima facie WC-justification for believing p in virtue of knowing his victim satisfies C. If this is right, we’ve not only got a couple of direct counterexamples to AN, in the form of the grocer and the demon, but we’ve undermined what I take to be a very significant motivation for AN that’s grounded in the widely accepted idea that WC-justified beliefs must be likely to be true. I don’t accept that idea myself, if some sort of objective likelihood of truth is involved. Those who agree with me and are sympathetic to hard AC will have another reason for rejecting the motivation for AN I’ve described. But even I realize the vast majority of epistemologists accept the idea that justified beliefs must at least be likely to be true. It is therefore important to see that TC+ provides the accurate way of formulating this connection, and that when correctly understood one cannot use the truth connection to argue against hard AC in the way I described. Insofar as something like this argument motivates AN, a serious blunder lies at the core of the motivation.

5. Conclusion I would like to be able to return to the reasons I presented at the beginning of the paper for thinking that if seemings are epistemically significant then rational disagreements are possible and claim that I had vindicated this line of thought. But I realize I have not done so much. I think that I have provided a good case against AN, and this removes one obstacle in the way of thinking that rational disagreements are possible. I have not provided a decisive argument that rational disagreements are possible, but given that I have supported AC, I  think I  have justified somewhat more optimism about the possibility of rational disagreements. I have also tried to raise concerns about the kinds of idealized scenarios on which the disagreement literature tends to focus. Whether the results of focusing on more realistic cases favor the possibility of rational disagreement or not, I think the results will be more significant. They could not so easily be dismissed on the grounds that they apply only to situations we never find ourselves in, and that may even be impossible.

Disagreement and Agent Centeredness

221

Acknowledgments I am indebted to Chris Tucker and an anonymous referee as well as the audience at the 2011 Midwest Epistemology Workshop for comments on earlier versions of this paper. Among members of the audience at the MEW, I  am most especially indebted to Richard Fumerton, who helped me better understand the issues I address in section 4 and who also suggested the demon argument I use. If I am still somewhat confused, which is entirely possible, neither he nor anyone else who tried to straighten me out is to blame for that.

References BonJour, Laurence. 1985. The Structure of Empirical Knowledge. Cambridge, MA:  Harvard University Press. Feldman, Richard and Ted A. Warfield, eds. 2010. Disagreement. Oxford:  Oxford University Press. Bryan Frances. 2010. “The Reflective Epistemic Renegade.” Philosophy and Phenomenological Research 81(2): 419–63. Huemer, Michael. 2011. “Epistemological Egoism and Agent-Centered Norms.” In Trent Dougherty, ed., Evidentialism and Its Discontents, 17–33. Oxford: Oxford University Press. Nathan L. King. 2011. “Disagreement: What’s the Problem? or A Good Peer is Hard to Find.” Philosophy and Phenomenological Research. doi:10.1111/j.1933-1592.2010.00441.x.

This page intentionally left blank

{ part v }

Dealing with Cognitive Penetration

This page intentionally left blank

{ 10 }

Phenomenal Conservatism and Cognitive Penetration: The “Bad Basis” Counterexamples Matthew McGrath 1. Introduction Seemings and beliefs are related in important ways. Like beliefs, seemings are content-bearing mental states of a subject that have truth as their correctness condition: a seeming that P is correct, accurate, only if P is true. Not only this, but when it seems that P, this normally inclines one to believe that P. It is not a mystery why they should so incline one, because seemings “have the feel of truth, the feel of a state whose content reveals how things really are” (Tolhurst 1988, 298). Because of these relationships, a subject can be said to “conserve” a seeming by believing its content. A key question in recent epistemology is whether one is always prima facie justified in conserving one’s seemings in this way. According to what Michael Huemer (2001) calls “phenomenal conservatism,” the answer is yes: PC if it seems to you that P, then you are prima facie justified in believing that P.

Huemer and others see significant epistemological implications in PC. For Huemer, it provides a single simple overarching foundationalist principle. For Chris Tucker, it provides a way to solve the “speckled hen” problem for foundationalism. For James Pryor, the principle of dogmatism, which restricts PC to perceptual seemings,1 provides the basis for a diagnosis of where the best skeptical arguments go wrong. Philosophers divide sharply over the plausibility of PC. On the one hand, Huemer (2001, 103–4) claims the principle is self-evident when understood 1 Pryor (2000, 519) writes: “The dogmatist about perceptual justification says that when it perceptually seems to you if p is the case, you have a kind of justification for believing p that does not presuppose or rest on your justification for anything else, which could be cited in an argument—even an ampliative argument!—for p. To have this justification for believing p, you need only have an experience that represents p as being the case.” Pryor assumes that experiences have propositional content. I will follow him in this assumption here.

226

Dealing with Cognitive Penetration

correctly. To understand the principle correctly is to appreciate that the sort of justification involved is epistemic rather than moral or prudential justification and that the sort of justification at issue is justification from the subject’s point of view. A good ordinary language gloss on the sort of justification Huemer has in mind is justification in the sense of what is “reasonable” for one to think, or in a more philosophical parlance, justification in the sense of the doxastic attitude it is reasonable for a subject to have.2 We might add that understanding the principle correctly also requires understanding that “seeming” is being used for states that meet Tolhurst’s description: they feel as if they reveal how things are, not merely how things might be or which are somewhat likely. With all this in place, let us think through just what PC says. Suppose it seems to you that P and you have no defeaters (i.e., no good evidence for not-P and no good evidence that this seeming is unreliable as to whether P). Which doxastic attitude would it be reasonable for you to have toward P? Disbelieve P, without good evidence for not-P? Withhold judgment on P? It does seem to you that P, and you lack evidence for not-P and for the unreliability of the seeming with respect to P? The only reasonable attitude to take is belief. Even if we do not find this answer clearly correct, let alone self-evident, it does have intuitive plausibility. Its seeming to one that P seems to be at least some evidence weighing in favor of one’s believing P. There are several objections that arise immediately. One concerns the liberality of PC. Any seeming? What if the seeming just overtakes one out of the blue (unbeknownst to one)? Is it then evidence? Or if it is evidence, is it good enough evidence for belief? A second worry concerns the need for higher-order reasons, reasons to think seemings are reliable, or at least the relevant particular seeming is reliable on the occasion in question. The first objection might make one back off from PC to restricted versions of it, such as dogmatism about perceptual justification: if it perceptually seems to you that P then you are prima facie justified in believing P. (Other possibilities privilege memorial seemings, introspective seemings, intuitional seemings.) Surely, if the thing looks plainly like a bird, this is evidence for you that it is a bird. If we do not require too much confidence for belief, isn’t this good enough evidence for belief, in the absence of defeaters?3,4 The second objection is familiar from debates over moderate foundationalism, and it applies both to PC and its restricted versions. However, the

2

Other glosses: what one ought to think, what it is rational to think, what it makes sense to think. This answer is arguably less plausible if a great deal hinges on whether P. Perhaps the prima facie rider is unsatisfied in high stakes situations. I ignore the relevance of stakes in this paper. 4 Alternatively, the first objection might be met by retreating from “justified in believing” to “has at least some degree of justification to believe.” The safest reply might be to combine the two ways of responding, scaling back PC as follows: if it perceptually (and perhaps memorially, intuitionally) seems to you that P, then prima facie you have at least some degree of justification to believe P. This is a substantial retreat. If Huemer is right in claiming that a great many of our ordinary beliefs are based on seemings, this retreat provides an opening for skepticism. The skeptic may argue that even if PC, 3

The “Bad Basis” Counterexamples

227

objection is hardly decisive. Even if such higher-level requirements could be defended without falling into skepticism (a big “if ”), there are proposals for combining moderate foundationalism with higher-level requirements. Suppose a source S could not justify one in believing P unless one was also justified in believing S is reliable. This supposition does not rule out S’s providing foundational justification to believe P. For instance, it may well be a necessary condition on being justified in believing anything that one is justified in believing that one exists.5 This doesn’t show that justification to believe things other than “I exist” cannot be foundational. Similarly, it may be a necessary condition on being justified through source S that one is justified in believing S is reliable. This would not show that justification through S cannot be foundational. More threatening to PC than these two objections are apparent counterexamples involving cognitive penetration of perceptual seemings. Peter Markie gives an example: Mental processes that are incapable of producing prima facie justified beliefs can nonetheless determine how things seem to us. Suppose that we are prospecting for gold. You have learned to identify a gold nugget on sight but I have no such knowledge. As the water washes out of my pan, we both look at a pebble, which is in fact a gold nugget. My desire to discover gold makes it seem to me as if the pebble is gold; your learned identification skills make it seem that way to you. According to (PC), the belief that it is gold has prima facie justification for both of us. Yet, certainly, my wishful thinking should not gain my perceptual belief the same positive epistemic status of defeasible justification as your learned identification skills. (Markie 2006, 356–57)

Susanna Siegel (2012) cites examples involving cognitive penetration of visual experience. If Jill unjustifiably thinks Jack is angry, she might expect him to look angry, and this might lead her to “see” him as angry, to his looking angry to her. If it does, surely her belief does not become justified if she re-bases it on his looking angry. Or think of the preformationist looking through the microscope who “sees” tiny human embryos in an ordinary cell, due to his desire that his theory be confirmed. Intuitively, if he forms the belief that there is a tiny embryo in the cell on the basis of such an experience, his belief is not justified. If we understand Siegel’s cases to involve visual seemings—its visually seeming that Jack is angry, that there is an embryo in the cell—then her cases are potential counterexamples to PC as well as its restriction to perceptual seemings. so modified, is true, these ordinary beliefs are not justified because the degree of justification seemings provide is not enough for outright belief. See Huemer (2001, 39–41). 5

Cf. Silins 2008, McGrath forthcoming.

228

Dealing with Cognitive Penetration

We are thus faced with a puzzle. PC is plausible, especially if it is restricted to perceptual seemings, but so are the apparent counterexamples involving cognitively penetrated perceptual seemings—the “bad basis” counterexamples, as I’ll call them. Anyone, whether a proponent of PC or not, ought to want some account of how and why both the principle and the apparent counterexamples to it should be plausible. This chapter explores the options for proponents of PC (or its restriction).6 I will consider three options: (1) the defeater approach, which takes the purported counterexamples to be cases in which the prima facie proviso is unsatisfied because the subject has a defeater; (2) the “distinguish the epistemic statuses” approach, which takes the examples to refute principles resembling PC but which concern some epistemic status other than the “favored” status of reasonableness to believe; and (3) the receptivity approach, which modifies PC by restricting it to “receptive” seemings, where a seeming’s receptivity is not understood in terms of the kind of seeming it is (e.g., perceptual, memorial) but rather in terms it is lacking a certain sort of basis, what I will call a “quasi-inferential” one. I will argue that the fi rst two approaches, despite enabling conservatives to show that PC withstands many “bad basis” cases, do not save them from all. I agree with Markie and Siegel that PC stands refuted, even when restricted to perceptual seemings. However, I hope to show that its prospects brighten when it is restricted to receptive seemings. Our discussion bears on a considerable number of views defended in epistemology. As Siegel (2012) notes, many epistemological theories seem vulnerable to “bad basis” counterexamples. Many philosophers have argued that how things are presented in perceptual experience prima facie justifies the subject in believing things are that way. Among these views are leading forms of moderate foundationalism, including not only Pryor’s (2000) dogmatism but Pollock’s (1975) direct realism and Robert Audi’s (1993) fallibilistic foundationalism. On other views (e.g., White 2006, Wright 2004), experiences by themselves do not provide immediate justification but can justify only together with independently justified beliefs or assumptions about their general reliability (these background beliefs or assumptions might be taken to be a priori justified or justified because of an “unearned” warrant). The counterexamples discussed here threaten these views as well. We may imagine that Markie’s wishful gold-digger has not only a gold-experience but also whatever White or Wright require in the way of justified background beliefs or assumptions. If so, and if this doesn’t affect the intuition that his belief isn’t justified (or isn’t as justified as it would be if the experience resulted from the exercise of perceptual identification skills), then this is a problem for White and Wright. I claim that these views, too, can take advantage of the receptivity approach and improve their

6

I leave this qualification tacit in what follows.

The “Bad Basis” Counterexamples

229

prospects in the face of the bad basis counterexamples. Finally, in an appendix, I  briefly discuss how proponents of doxastic conservatism might attempt to appeal to the idea of receptivity.

2. A Preliminary Worry Before considering the three approaches, I want to address a worry about the phenomenal conservative’s favored status of what it is reasonable for the subject to think. One might doubt whether the intuitive correctness of such claims— claims made in my motivation for PC—tracks any normative epistemic status as opposed to the demands of coherence.7 So, even if the motivations I gave for PC motivate something, the worry is that they do not motivate a principle with any normative “oomph.” Moreover, if throughout this paper I am going to appeal to intuitive claims about what it is reasonable to think in various situations, these appeals will be questionable as well. The general worry is about trafficking in the language of “reasonableness” at all in doing epistemology. Here is a way to explain one source of the worry. A coherent set of beliefs can be out of keeping with one’s evidence. Suppose my evidence strongly points against P and against Q, but that nevertheless I believe P. Then, if I continue to believe P, coherence requires that I believe the disjunctive proposition P or Q if I entertain the question. Suppose I entertain the question of whether P or Q is true. We want to say that it is reasonable for me to believe P or Q, insofar as I already have a belief that P. How could it be reasonable not to believe P or Q given that I already believe P? On the one hand, surely it isn’t reasonable for me to believe P or Q, because after all, I have good evidence against P and evidence against Q! So, we want to say that it isn’t reasonable for me to believe P or Q, and yet also feel it is reasonable insofar as I already have a belief that P. John Broome (1999, 2002) explains such apparently conflicting intuitions as follows. If I  believe P and consider the question of whether the disjunction P or Q is true, and then I fail to believe P or Q (by disbelieving it or by suspending judgment), then I am being incoherent; my cognitive states will clash. To keep the terminology clear, I will use the label “coherence requirement,” in place of Broome’s “rational requirement.” So, believing P requires me, on pain of incoherence, to believe P or Q at least if I consider the question of whether it is true that P or Q. However, although my belief in P imposes this coherence requirement to believe P or Q, this does not make believing P or Q epistemically appropriate for me in a normative sense: my evidence against P and against Q together give me sufficient epistemic reason not to believe P or Q. There is no contradiction here, according to Broome: it is possible, if one

7

Such a worry is developed in Jackson (2011).

230

Dealing with Cognitive Penetration

takes a normatively inappropriate attitude, to be such that coherence requires taking another normatively inappropriate attitude. The intuitive judgment, “I am not reasonable to believe P or Q” picks up on the epistemically normative status. The intuitive judgment, “I am reasonable to believe P or Q insofar as I believe P” picks up on the demands of Broomestyle coherence requirements. Whether such requirements have normative standing is a matter of dispute: are there epistemic reasons to be coherent in this sense; is there an epistemic obligation to be coherent?8 Even if both intuitive judgments of reasonableness track something of normative significance, though, the sources of the normativity are different. In the one case, the source is the particular concrete evidence against P and against Q. In the other case, the source is some more abstract reason or obligation to be coherent.9 Let us return to the original worry about my use of “reasonable.” Broome’s distinction is clear, but in ordinary life we can use the same terms—“reasonable,” “ought,” “should”—for both what is required for coherence and for epistemic normative appropriateness (deriving from concrete garden-variety reasons rather than some abstract reason to be coherent). What reason is there to think that when I motivated PC, I wasn’t picking up on coherence requirements only rather than normative appropriateness? What reason is there to think that when I go on to consider cognitive penetration cases I won’t be simply gathering intuitions about coherence requirements? I make two points in response. First, the fact that we ordinarily use “reasonable to believe” to cover cases of epistemic normative appropriateness and cases of what’s required for coherence is by itself no good reason to abandon all such talk in doing epistemology. We need only to take caution not to confuse the two. When “reasonable” is used merely for what is required for coherence, there is an implicit relativization to other mental states which themselves can be assessed for their normative epistemic status. When we find it natural to say that believing P “is reasonable,” we need to check to see whether we want to say the likes of “reasonable insofar as one believes P” or “reasonable given that one doubts Q.” The relativizations are to the mere holding of these

8 See Kolodny (2005) on the question of the normative force of Broome-style coherence requirements. 9 Similarly, if I continue to intend to achieve some end, against all the many powerful reasons not to, then if I know M is a necessary means to that end, practical coherence demands me to intend to take means M (Broome 1999, 2002; Dancy 2000). (This is a bit rough. We may need to add the belief that unless I intend to take M I will not do M.) Here again, a judgment of reasonableness picks up on the following fact: I hold other states which are such that, to maintain consistency, if I add to them a “practical” attitude toward M, it must be intention. Such facts can hold even if it is normatively inappropriate to intend M, because I have good reasons not to intend the end in the first place. Which attitude it is normatively appropriate in a practical sense to have is one thing. Which attitude I need to have to maintain coherence with another attitude I already have is quite another thing. Intuitive judgments of practical reasonableness can pick up on either.

The “Bad Basis” Counterexamples

231

psychological states, not to the epistemically appropriate/inappropriate holding of them. Coherence requirements are blind to the normative statuses of the states involved. Consider, again, the above motivation for PC. The objector agrees that we do find it natural to say that one is “reasonable” to believe P when it seems to her that P and she lack defeaters. So, we have to ask if it feels necessary to add a relativization to other mental states assessable for their normative epistemic statuses. Assuming seemings are not assessable in this way (can there be an epistemic reason for it to seem to one that P?), there is no such relativization.10 One might say, “She is reasonable to believe P given that it seems to her that P,” but here the “given that” only functions to indicate the various features of the person’s situation which might help to make it reasonable for her to believe P in the normatively loaded sense. Second, coherence requirements are presumably not brute facts. They need explanation: why is it that combining such and such attitudes is incoherent, if it is? In some cases, the explanation proceeds solely in terms of the nature of the attitudes and the logical relations between their contents. Perhaps we can explain the coherence requirement not to believe ~P if one believes P by pointing to the fact that P and ~P are contradictories and that belief aims at truth. However, other coherence requirements cannot be explained in this fashion. Doubting whether a testifier is sincere rationally requires me not to believe what the person says merely on his word (so that if I do believe the testifier solely on his word while believing he is insincere my mental states clash). It is far from clear how this requirement is explainable solely in terms of logical relations between the contents involved and the natures of the states of doubt and belief. A paradigm way to explain coherence requirements is by referring to a corresponding normative principle. (We use “reasonable” hereafter only for the normatively loaded status.) If I am reasonable to doubt that a testifier is being insincere in testifying that P to me, then the testifier’s mere word that P is not enough to make me reasonable in believing P. That’s why there is something incoherent about doubting his sincerity while accepting P based on solely on his word. In so doing, I  would be guaranteeing that either my doubt or my belief was unreasonable. In fact, wherever there is a true normative epistemic principle about reasonableness, it grounds a corresponding coherence requirement. The fact that being reasonable in believing P suffices for being reasonable in believing P or Q (if you entertain the question) grounds the fact that believing P requires believing P or Q (if you entertain the question). If one flouts the coherence requirement, one would be guaranteeing that one falls short with respect to reasonableness either in believing P or in not believing P or Q.

10

The question whether some seemings can be normatively assessable is revisited in note 22.

232

Dealing with Cognitive Penetration

Thus, where there is a coherence requirement that appears inexplicable merely in terms of logical relations between contents and the nature of belief, a plausible potential explanation for that requirement is the corresponding normative principle. Suppose our objector agrees that there is a coherence requirement to believe P when it seems to one that P and one lacks defeaters. What is its explanation or ground? It is difficult to explain it in terms of logical relations among contents together with the nature of belief and seeming. However, there is a ready explanation that appeals to the normative principle PC. The critic who concedes that the motivations I have provided for PC motivate accepting a coherence requirement concerning seeming and believing thus may very well have to appeal to PC to explain that requirement, where PC is understood as a normative epistemic principle. With this preliminary worry addressed, I turn to the three approaches for coping with the “bad basis” counterexamples. In what follows, I often write of “justification,” because of its familiarity and its convenient grammatical features.11 But the reader should see this terminology merely as a convenient way of talking about what a subject is reasonable to think, where “reasonable” is understood in the epistemically normative sense.

3. The Defeat Approach Appealing to defeaters handles many familiar “bad basis” cases. Consider wishful thinking. Within psychology, wishful thinking is understood as judgment influenced by a directional goal (a goal of reaching a particular conclusion rather than an accurate conclusion on a given question), and there is some consensus that one primary way that directional goals work is by biasing the evidence the person brings to mind at the time of judgment (Kunda 1990, 1999). In particular, when we have defeasible evidence favoring a desired belief that P—say, it seems to us that P—we tend not to search memory as hard, if at all, for any counterevidence we might have. This need not always have the desired effect, of course, because sometimes clearly disconfirming evidence springs to mind anyway, despite the biased memory search.12 Still, it can work, and when it does, the belief formed is intuitively not justified. There

11 Thus, we can speak of something justifying a belief and of one having justification to believe something. All these could be re-expressed using unlovely neologisms derived from “reasonable.” 12 Research into judgment influenced by directional goals shows there are limits to its power. Awareness of decisive counterevidence, and even the anticipation of decisive evidence one way or the other in the future, drastically reduces its effects. It is also limited in a number of other ways (e.g., it operates less effectively when the question at hand is a hard factual question rather than an open-ended one). See Kunda (1990, 1999).

The “Bad Basis” Counterexamples

233

is no mystery about why: the subject has a defeater, one that isn’t brought to mind, but a defeater all the same.13 However, there could well be “bad basis” cases in which the subject lacks defeaters, even unconscious ones. The examples from Markie and Siegel appear to fall into this category.

4. The “Distinguish the Epistemic Statuses” Approach To have a counterexample to PC, we need a case in which it seems to the subject that P, the subject lacks defeaters, but it isn’t reasonable for her to believe P. The hope behind the “distinguish the epistemic statuses” approach is to grant that in those “bad basis” cases in which the subject lacks defeaters the subject may well lack many important positive epistemic statuses, but not the favored one of being reasonable to believe. When we take the “bad basis” examples to refute PC—the claim is—we conflate distinct epistemic statuses. The counterexamples may well refute principles like PC about those other statuses, but not PC itself. We will discuss a number of ways to pursue this approach. One strategy is to press into service the distinction between propositional and doxastic justification. PC concerns propositional justification, i.e., justification to believe, but the counterexamples show, at least in the first instance, only that the subjects lack doxastic justification, i.e., their beliefs are not justified. Markie claims that the belief of the gold-digger who relies on the wishful seeming surely does not have the “same positive epistemic status” as that of the gold-digger who believes on the basis of an appearance grounded in the exercise of perceptual learning. Siegel, too, makes claims about the epistemic

13 Taking such a line would require the phenomenal conservative to agree that one can have defeaters that one is not consciously entertaining. This is a plausible line in any case (though see Feldman, 1988, for an opposing view). In light of the relevance of stored evidence to justification, it is important to distinguish a number of related assessments of unreasonableness, which vary in the degree of associated intuitive irrationality. If a person believes something that isn’t reasonable to believe in light of her conscious information at the time, this is clearly more irrational than if a person believes something that it is reasonable to believe in light of conscious evidence but not in light of her total possessed evidence. To take an example from Goldman (1988, 202–3), Melanie knows the library normally opens at 7:00 am, but she has also noticed it opens at 1:00 pm on Sundays, although she rarely goes there Sundays. This Sunday she has a special reason to go to the library. She is heading off to the library shortly after 7:00 am. She finds it closed and thinks, “How silly of me! I knew the library doesn’t open on Sundays till 1:00 pm.” Her action seems far less irrational than it would if she had recalled the Sunday hours and still went to the library. Yet even in Goldman’s case, Melanie’s action is unreasonable: she is doing something she has evidence will not turn out well. Compare this to a case in which Melanie had never known about the Sunday hours, or to a case in which years ago Melanie knew of the Sunday hours but had “completely forgotten,” so that if you asked her beforehand whether the Sunday hours were the same as the regular hours she would say yes. There are many interesting intermediary cases in which one seems to have “forgotten” and yet is tempted to say, after being reminded, “I knew that!”

234

Dealing with Cognitive Penetration

status of beliefs in her examples. The examples, then, might be taken to refute only a principle like PC about doxastic justification, but not PC itself. However, as Siegel notes in her second paper (2013), doxastic and propositional justification are standardly thought to be related. If you are propositionally justified in believing P—if P is something you are justified to believe—then it seems you should have a route available to a doxastically justified belief (or at least a belief that isn’t doxastically unjustified because of its basis in what propositionally justified you).14 If the unhappy look on your face when I serve you lentils and cabbage for dinner provides me propositional justification to believe you don’t like lentils with cabbage, then if I  form this belief, based on my perception of your unhappy look, my belief should be justified. The conservative’s favored notion of reasonableness to believe itself has a doxastic counterpart, viz., reasonable belief. So, if propositional and doxastic justification are related in the way described, the conservative cannot say both that a seeming which is the result of wishful thinking provides you propositional justification and that the belief based on this seeming is doxastically unjustified.15 Conservatives would do better to argue that the “bad basis” cases are cases of justification in the favored sense (both propositional and doxastic) and that intuitions of epistemic inappropriateness pick up on distinct negative epistemic statuses found in the cases. Chris Tucker suggests that wishful seemings do provide propositional justification, and so provide routes to doxastic justification, but they fail to put one in a position to have knowledge. Tucker (2010, 538–42) reports a response from Markie in which Markie contrasts two brains in vats (BIVs), one whose gold-seemings rely on something like perceptual training (imagine that if the brain were embodied, the embodied subject was placed in a normal situation involving gold on Earth and he would be able to recognize gold nuggets) and one whose gold-seemings are wishful. If both rely on their seemings, neither would be in a position to know, or even to believe

14 I thank Siegel for making me see the need for the parenthetical. See her (2013) for an explanation of why further adjustments are needed to the linking thesis, insofar as propositional justification is localized in a way the doxastic justification is not. See also the footnote that follows. 15 John Turri (2010) proposes counterexamples to the standard assumption about the link between propositional and doxastic justification. In one example, a juror bases his belief in the guilt of the defendant on a set of reasons not because those reasons make guilt likely but only because the tea leaves say that they do. In this case, the juror has propositional justification owing to a certain factor but although he bases belief on that factor his belief is not doxastically justified. Turri recommends weakening the principle so that what having propositional justification to believe P ensures is that the subject “currently possesses at least one means of coming to believe P such that, were [she] to believe P in one of those ways, [her] belief would thereby be doxastically justified” (2010, 320). This is plausible—the basing must be epistemically kosher. However, this minor fix to the standard assumption is not enough for the conservative. In Markie’s gold-digger case, the subject already has the seeming, based on wishful thinking. He would have to re-base it on something else, but there seems no adequate base available. Turri’s weaker link between propositional and doxastic justification is enough to get the difficulty up and running.

The “Bad Basis” Counterexamples

235

reliably, or to have any externalist positive epistemic status, and yet the belief of the BIV who relies on wishful seemings is still epistemically worse. In reply, Tucker appeals to a difference in blameworthiness. The wishful BIV is blameworthy for his false beliefs in a way that the “expert” isn’t. Being blameworthy for getting it wrong, Tucker claims, is consistent with being justified (reasonable) in one’s belief. Instead of blameworthiness, the conservative might appeal to epistemic statuses less clearly tied to voluntariness, e.g., epistemically vicious belief or bad cognitive functioning. My belief might have these statuses even if I did my best and so I am excused and thus not blameworthy. (Perhaps the wishful BIV gold-digger was trying his very best to be objective.) One could then recast Tucker’s suggestion as follows: in cases in which one’s evidence-assembling processes are biased by the wish to arrive at a belief that P one’s belief is viciously formed but reasonable all the same. The guiding assumption behind approaches like Tucker’s is the plausible one that the following two questions are independent: (1) did one believe reasonably in light of the evidence one had at the time? (2) Was one’s belief based on evidence that was obtained in an epistemically acceptable (non-vicious) way? In the “bad basis” cases in which the subject lacks defeaters, the hope would be that the answer to (1)  is always affirmative and that claims to the contrary wrongly take the negative answer to (2) to decide the answer to (1).16 So far, so good. The problem is that there are, or could be, cognitively penetrated seemings in which (a) the subject lacks defeaters and (b) the cognitive penetration doesn’t consist in the biasing of evidence-assembling processes but is more direct than this, in a way that seems to matter to reasonableness of belief in light of the evidence one had at the time.17 Consider cases of what I will call “free enrichment” of one seeming by another, due to some cognitively penetrating state. It is natural to understand Markie’s gold-digger case as follows. The pebble looks somewhat yellowish to the gold-digger and the effect of his wish to believe it is a gold nugget is that it looks to him to be a gold nugget. The sizable gap between the yellowish-pebble-seeming and the goldnugget-seeming is closed by the wish, not by any knowledge of the observable features of gold nuggets. The gold-seeming freely enriches the yellowish-pebble seeming (if you pardon the pun).18 Similarly, a natural way to understand Siegel’s anger case is to think of Jack looking at Jill in a certain way which is

16

The independence of the two questions is well explained by Feldman (2003). Thus, Siegel (2012) defines the interesting sort of cognitive penetration as follows: there is such penetration if and only if there are cases in which two subjects, with the same distal stimuli and attending to the same things, have experiences with different contents, owing to the effects of cognitive states. However, in her second paper, Siegel (2013) seems to want to classify some etiologies involving biasing influences on attention as irrational. See note 20 for further discussion. 18 I use “free enrichment” only when the content of the resulting seeming “goes beyond” the content of the seeming on which it is based, and goes beyond it in an epistemically problematic way, so that if the transition was between beliefs this would count as “jumping to conclusions.” 17

236

Dealing with Cognitive Penetration

not decisively that of someone who isn’t angry, but that Jill’s belief that he is angry leads her to expect that he will look angry, which in turn gives rise to his looking clearly angry to her. Again, we have one visual seeming being freely enriched by another, due to the presence of a cognitive state, an expectation. In free enrichment, the wish makes one “jump to conclusions.” The jumping is not jumping from one belief to another but from one seeming to another.19 The problem for the “distinguish the epistemic statuses” approach is that when there is such jumping to conclusions within seemings, it starts to seem that the “concluding” seemings aren’t really evidence; they are non-evidence because they are arrived at in epistemically sub-par ways based on the real evidence, the upstream seemings. I think the conservative must admit that free enrichment cases are counterexamples. She must revise her view.20

5. The “Receptivity” Approach When thinking about what it is reasonable for a subject to believe, a certain intuitive picture comes to mind. The subject has certain basic evidence or grounds, which are “handed” to her. Based on what is handed to her, the 19 We need not assume that there is always a temporal gap whenever a person jumps to conclusions in beliefs, and similarly for jumping to conclusions in seemings. What matters in the belief case is that there is the sort of asymmetric dependence associated with inference. Something similar holds for the seemings case (but with a notion of “quasi-inference” developed in what follows). 20 One might worry that I have assumed, too quickly, that the conservative has no trouble with biased attention cases. Siegel (2013) considers psychological experiments in which white subjects perceive anger more clearly in black faces than in white faces which are equally angry and which the subjects would see to be angry if they paid the same attention to them as they do the black faces. In such a case, Siegel suggests it might look to a white subject as though only the black faces are angry. If the subject believes this, though, he wouldn’t be justified in believing only the black faces are angry. So, doesn’t attention here affect justification and not merely evidence? Consider possible etiologies of the subject’s experience, his perceptual seeming. Two suggest themselves. The first is that it seems to the subject that only the black faces are angry because it seems to him that the black faces are angry and it seems to him that the white faces are not angry. The latter seeming could well amount to a “free enrichment” due to positive bias. The brief unfocused scanning of the white faces doesn’t produce any visual seeming that would be enough to support the “enriched” seeming. Here attention isn’t the problem, free enrichment is. A  second possible etiology is that it seems to the subject that only the black faces are angry because it seems to him that the black faces are angry and the subject is also aware that it doesn’t seem to him that the white ones are. Here, again, the problem isn’t attention. It’s the insufficiency of the combination of the angry-black-faces-seeming and the awareness of lacking an angry-white-faces-seeming to justify the subject in believing that only the black faces are angry. We would need to add to this combination further knowledge or justified belief in the subject about his being suitably attentive to the white faces, but it is hard to see how that belief would be justified in the experiments in question. I would hazard that if they were justified well enough, then the resulting belief would be justified as well. I should add that one way cognitive penetration via free enrichment might be implemented is by the subject’s focal attention. Perhaps Markie’s gold-digger focuses, because of biased attention, on those aspects of his experience that could provide some support for thinking the stone is gold. Still, the problem is that the support is not good enough.

The “Bad Basis” Counterexamples

237

subject “makes” something of her situation—she represents things as being a certain way. Given this picture, when we ask whether something provides foundational justification, we are asking about whether the basic evidence justifies it—not whether something the subject makes of that evidence would or does justify it. The standard way of filling out this picture assumes that all experiences are part of one’s basic evidence and that what one “makes” of one’s basic evidence consists only of one’s beliefs and other doxastic attitudes. As Sosa writes: Experiences are able to provide justification that is foundational because they lie beyond justification and unjustification. Since they are passively received, they cannot manifest obedience to anything, including rational norms, whether epistemic or otherwise. Since unmotivated by reasons, they can serve as foundational sources, as regress-stoppers. When they help explain the rational standing of some other state or action, they do not thereby problematize their own rational standing. Being so passive, they have no such standing. (2007, 46)

The possibility of freely enriched perceptual seemings challenges this standard way of filling out the basic picture. Such seemings are experiences, but they are also part of what we make of the basic evidence, rather than parts of that basic evidence itself. Thus, they cannot provide foundational justification. Let us say that a transition from a seeming that P to a seeming that Q is “quasi-inferential” just in case the transition that would result from replacing these seemings with corresponding beliefs that P and Q would count as genuine inference by the person.21 (The beliefs here might need to be thought of as utilizing demonstrative or phenomenal concepts.) I claim that quasi-inferential transitions between seemings function epistemically in the way inference by the person does: they can at best transmit the relevant epistemic property of the inputs to the outputs; they cannot generate this property for the outputs when it isn’t possessed by the inputs. In the case of inferential transitions between beliefs, the epistemically relevant property that can be transmitted is doxastic justification. In the case of transitions between seemings, the property is justifying the subject in believing its content. Suppose I transition from a belief that P to a belief that Q through inference, solely basing the belief that Q on the belief that P.  If the input belief (P)  isn’t justified, the output (Q)  isn’t either. Moreover, if the inference isn’t a good one—if P doesn’t support Q sufficiently for the subject—the output belief is not justified. When all goes well, the output belief is justified, but only

21 The strategy of characterizing epistemically significant features of seemings in terms of the corresponding features of beliefs I borrow from Siegel. Again, the terminology of “transition” might suggest temporal succession: first there’s one seeming and then it produces another. What matters, though, is an inference-like dependence, rather than any temporal relation.

238

Dealing with Cognitive Penetration

derivatively, not foundationally. Similarly, suppose I transition from a seeming that P to a seeming that Q, through quasi-inference, solely basing the seeming that Q on the seeming that P. If the input seeming doesn’t justify the subject in its content (P), then the output seeming cannot justify the subject in believing its content (Q). If the quasi-inference isn’t a good one—if the content of the input— P—doesn’t sufficiently support the content of the output—Q—for the subject, then the output seeming cannot justify. When all goes well, the output seeming can justify, but as with inference, only derivatively, not foundationally.22 My way of picking out “quasi-inferential” seemings-transitions is modeled on inferential belief-transitions. To have a more intrinsic characterization, we would have to ask what makes a transition between beliefs an inference by the person and then apply this to seemings-transitions. The following are plausibly necessary: (i) the input and output states (beliefs in this case) must be mental states of the person, not merely of a sub-personal system; and (ii) there must be an explanation in terms of the person’s own mental states that “rationalizes” the transition, i.e., that allows us to see the transition as the person’s treating the content of the input state as supporting the content of the output state. One sort of explanation might appeal to the person’s grasp (perhaps good, perhaps faulty) of the support the one proposition provides for the other. Another might appeal to the person’s background “information” (true or false), and where the possession of this information could amount to know-how. A third might appeal to cognitive states that make us “jump to conclusions,” including expectations, desires, moods. No doubt there are cases in which several of these factors jointly explain the transition. So far, I  have limited quasi-inferences to seemings-transitions. However, some seemings might be based on beliefs in a way that resembles inference. This presumably happens regularly for non-perceptual seemings:  because I believe one thing, another thing can seem to be true. So, let us allow that belief-to-seeming transitions can be quasi-inferences, too, so long as, were the final seeming replaced with a belief, this would be genuine inference by the person. Here, too, the output seeming would provide, at best, derivative justification, not foundational justification.23

22 Seemings that are quasi-inferred from other seemings have some sort of normative epistemic standing in that they either count or fail to count as providing justification depending on whether the quasi-inference was a good one. Thus, one might be required in Broome’s sense to believe P when it seems to one that P and one lacks defeaters even though one “ought not” have that seeming in a normative epistemic sense. This observation holds out hope for explaining why it might seem “reasonable” in some sense for the wishful-seeming gold-digger to believe he sees a gold nugget—it seems reasonable given his gold-seeming, but that gold-seeming itself is normatively inappropriate. Thus, although his belief isn’t reasonable in the normatively oomphy sense, it is rationally required by the seeming. Here I anticipate the receptivity proposal, developed below. 23 Thus the idea of quasi-inference is of some help in resisting the charge, lodged against PC, that it treats even inferentially justified beliefs as foundationally justified.

The “Bad Basis” Counterexamples

239

Finally, the Kantian terminology:  call a seeming that is not an output of a quasi-inferential transition receptive and call one that is nonreceptive. The receptivity proposal is to take receptive seemings to be sources of justification—to confer foundational justification—and to take nonreceptive seemings as justifying derivatively if at all.24 If this is what the receptivity proposal amounts to, what reason is there to accept it? Later, I consider a number of different sorts of cases to test the proposal’s predictions. The proposal makes predictions insofar as it has implications for which beliefs are justified in various cases. Depending on whether the beliefs in the various cases are justified or not, the prediction is either correct or not. I argue that the predictions across a range of test cases are correct, or at least plausible. In each case, as I hope will be clear, the proposal gives us not only a plausible verdict but a plausible accompanying explanation of why the case is or is not a case of justified (i.e., reasonable) belief. However, there are certain hard cases that suggest the possibility of broadening the notion of quasi-inference further. I discuss these at the end of the section. Let’s start with cases of free enrichment due to cognitive penetration. I take it that the beliefs in these cases are not justified. The receptivity proposal of course predicts this (it was designed to do so). Take the gold-digger case. The pebble’s seeming somewhat yellowish would not make the novice justified in believing it is gold (even given background knowledge). So, if, due to the wish to find gold, this seeming is freely enriched to give rise to the pebble’s seeming to be gold, the gold-seeming is nonreceptive and so incapable of providing foundational justification. It can at best justify derivatively. But it doesn’t even justify derivatively, because the quasi-inference is not a good one: this stone is somewhat yellowish doesn’t sufficiently support this stone is a gold nugget for the subject. As the subject bases his belief on this seeming, and this seeming doesn’t justify, the subject’s belief isn’t justified. Siegel’s preformationism and anger cases are handled analogously.25 24 My receptivity proposal draws from Siegel’s (2013) work on rationally assessable etiologies of experiences. If we think of an irrational etiology as one such that the resulting seeming cannot provide justification, I claim that an etiology is rational or irrational just in case it involves quasi-inference, rational if the quasi-inference is good and irrational if bad. 25 Widely discussed psychological experiments by Delk and Fillenbaum (1965) suggest that subjects in certain situations see clearly gray bananas as yellow due to their knowledge that bananas are or normally look yellow (or have looked yellow in the past). There are various possible accounts of why this might be. One treats this simply as a case of making false judgments about one’s experiences: it doesn’t really look yellow to us but we judge it does because of our expectation that it will because it’s a banana and they look yellow. There is no quasi-inference involved on this account. A second account understands such cases to involve free enrichment of one color experience by another, due to the knowledge that it is a banana and beliefs about bananas being or looking yellow. This knowledge would not close the gap between its looking rather grayish and its looking yellow At best the knowledge would be part of another justification for thinking the banana is and looks yellow. A third sort of account, closer to that suggested by Hansen et al. (2006), locates the influence of the knowledge earlier in visual processing, prior to any visual seemings-transitions.

240

Dealing with Cognitive Penetration

By allowing that quasi-inferred seemings justify derivatively, the receptivity proposal gets cases of good quasi-inference right.26 Consider a version of Thomas Senor’s (2005) sunset example. In the evening, when I view the beautiful sunset over the Missouri River, I believe (and know) it is a sunset based on its looking like a sunset (let’s suppose). It looks like a sunset to me in part due to my knowledge that it is evening together with more upstream seemings. Here the content of my upstream seemings does sufficiently support that it is a beautiful sunset for me. It does so, because of my background knowledge that it is evening. For someone else who didn’t know if it was evening or morning, the contents of the upstream seemings would not provide good enough evidence (assuming that sunsets and sunrises look alike). The receptivity proposal relies heavily on the personal/sub-personal distinction. I  included as the “intrinsic” properties of quasi-inferential seemings-transitions the following properties of inferential belief-transitions:  the seemings must be seemings of the person and the transition between them must be explainable by mental states of the person. I of course admit that there are vague cases of the personal/sub-personal distinction.27 However, there are clear cases, and in these cases the receptivity proposal’s predictions seem plausible. I consider some examples. Suppose my visual system produces a perceptual state in which it seems to me that one tree is farther away than another on the basis of “cues” such as these: facts about difference in the angle of convergence between the two eyes when fixed on the center of each of the trees, facts about the relation of the trees’ retinal images to the horizon line, facts about occlusion, etc. Presumably, I do not go through this sort of calculation when I see one of the trees as farther away; rather, some part of my sub-personal visual system does. Similarly, when I hear a sound as coming from a certain direction, I am not computing information about wave interference created by the fact that the head is an obstacle; but in ongoing theories of auditory perception, our auditory perceptual systems do make such computations. The explanation for the output seeming in these cases is not given in terms of personal mental states. These are not quasi-inferential transitions. The receptivity approach therefore predicts that the resulting seemings, because receptive, provide foundational justification. This seems correct. Suppose a sub-personal visual or auditory system, for some reason, “jumps to conclusions” in its calculations. This by itself does not affect whether the person’s belief is reasonable. It would affect

26

Thanks to Chris Tucker for discussion here. He saved me from several mistakes. If a heuristic is at work but is not easily accessible to consciousness and is not available for central processing, is the person relying on the heuristic or rather the sub-personal system? When I anchor on a prime (say, my SSN), and use it in estimations of populations of cities, say, am I relying on a belief about my SSN as a reason? Or consider practical reasons: when I prefer items on the right, am I relying on it’s on the right as my reason? 27

The “Bad Basis” Counterexamples

241

reasonableness, though, if the seeming produced by the system depended for its justificatory power for the person on the quality of the sub-personal inference. Consider, also, our reactions to perceptual illusions. We do not think that someone unaware of the potential for these illusions is unreasonable to trust appearances; and our assessment would not change if we learned that what is responsible for the illusion is some overgeneralization on the part of the visual system, an instance of “jumping to conclusions.” In other cases, a perceptual system might employ information that the person does possess but the employment of which by a perceptual system does not count as the person’s employing it. For instance, we all know that lighting usually comes from above. However, if you are in a room with strange lighting, in which it is unclear from which direction the light is shining, you might still see an object as convex, as bulging outward, because of shading in its lower parts.28 Even assuming there is a seeming-transition in this case, from one personal seeming state (e.g., about shading) to another (about convexity), what explains this transition is not your possession of information about the direction from which the light is shining or usually shines, but rather your visual system’s possession of this information.29 Supposing you had not read up on this feature of the perceptual system, wouldn’t it be reasonable for you to think that the object was convex, despite having reason to doubt the lighting is from above? One might worry about how the receptivity proposal fares when we consider seemings resulting from the exercise of learned perceptual skills. Suppose through Audubon society training I learn to recognize pine warblers by sight. When I see one and it seems to me to be a pine warbler, I believe it is one on this basis. Suppose I don’t do any conscious reasoning. My eyes scan the bird, no doubt focusing on certain features of the bird, and as a result it seems to me that this is a pine warbler. Is this seeming receptive or not? First, suppose there is a seeming-transition here:  the bird seems to be a pine warbler as a result of its seeming to have certain other features which it is hard to articulate, its seeming to have a pine warbler gestalt. Is this a quasiinferential transition? If it is explained by appeal to the person’s knowledge of the observable features of pine warblers, the answer is yes. Suppose someone, by virtue of this knowledge, based a belief that the bird is a pine warbler on the belief that the bird had such and such gestalt. This would count as inference, I think. When these beliefs are exchanged for seemings, the result is a quasiinferential seemings-transition and so a nonreceptive output seeming. Thus, on the receptivity proposal, the output seeming can at best justify derivatively.

28

See Braisby and Gellatly (2005) for discussion. It’s surprising to learn about how perception of convexity and concavity works. Compare this to “learning” that one perceives someone as a police officer because of the person’s uniform. 29

242

Dealing with Cognitive Penetration

It plainly does justify, because the quasi-inference is a good one: the input seeming justifies one, and one’s know-how makes it the case that the content of that input sufficiently supports the content of the output seeming for one. This verdict seems plausible: the birder who believes the bird is a pine warbler based on quasiinference is justified, and justified in part because of her perceptual know-how.30 Suppose, instead, that there is no seemings-transition at all. What would account for the justification of the resultant belief? I am not sure how to answer this question. Perhaps we can make sense of seemings as being based on experiences with non-conceptual content, in such a way that the transition from these experiences to the seeming counts as close enough to be “inferential.” If so, we might hope to explain the output seeming as non-receptive. But in lieu of that, if the seeming comes from nothing like inference by the person, we would have to say that the seeming is receptive.31 On the receptivity proposal, this seeming would then be classified as providing foundational justification. Either way, there is no danger of ruling the birder’s belief unjustified.32 30 If we follow Turri, we might think that more is required for quasi-inference to result in a justifying seeming. Not only must the input seeming have this property, and not only must the quasiinference be a good one, the actual transition must be based appropriately on the quasi-inference. So, in a case in which there is a good quasi-inferential relationship between a seeming that P and a seeming that Q, and in which the seeming that P can justify the subject in believing that P, the seeming that Q might not justify the subject in believing that P because the person’s transition from one seeming to the other was not appropriately based on the quasi-inferential relationship. So, suppose a gold-digger has a visual seeming in which the stone does look quite yellowish, shiny, etc., and in which the quasiinferential relationship is a good one for the person, but still the gold-digger makes the transition because of wishful thinking. If we take a Turri-style view, we will say that the resulting gold-seeming does not justify the gold-digger. 31 Top-down influences can have the result of increasing our capacity for having rich receptive seemings. Goldstone and Hendrickson (2010) notes experimental evidence for a number of top-down influences on relatively early perceptual processing. The categories in our conceptual repertoire, he claims, affect perceptual unitization (the development of single functional units that are triggered by complex configurations in the stimulus) and dimensionalization (the development of distinct functional units triggered by different elements of a stimulus). The former makes possible quicker and more efficient processing of a scene. The latter makes possible selective attention in experts, where no such attention is possible for most of us (e.g., color scientists who can attend to saturation and brightness separately). See Goldstone (2003) and Goldstone and Hendrickson (2010). 32 What if the pine-warbler-seeming is the output of a quasi-inferential transition, but the transition is made not because of the person’s learned identificatory skills but because of a mental state masquerading as learned identificatory “know-how”? Would the output seeming justify? To determine this, we would have to examine the different possibilities for these masquerading imposters. Imposters might include unreliable dispositions resulting from poor training (e.g., the disposition to see something as a cardinal bird if one sees it as yellow and small) and reliable but unlearned dispositions (e.g., those that result, say, from “last night’s neurosurgery” as Jack Lyons (2011) imagines). In either case, these are not “know-how” but at best “belief how,” if you will. Whether the resulting seeming justifies will therefore depend on whether such “belief how” makes it the case that the content of the input seeming sufficiently supports the content of the output seeming for the subject. It presumably does not if the “belief how” counts as unreasonable but may if it counts as reasonable. A key question is whether doxastic conservatism can be defended. If it can, these “beliefs how” may well count as justified. If not, they will not. I leave this matter open, for different phenomenal conservatives to decide differently. See the appendix of this chapter for a brief discussion of doxastic conservatism.

The “Bad Basis” Counterexamples

243

We have considered cases in which there is quasi-inference. The receptivity proposal handles these cases well. We have also considered cases in which there is neither quasi-inference nor top-down influences in which cognitive states affect the goings-on at the sub-personal level which lead to the output seeming. Here, too, the receptivity proposal seems to do well. What of the remaining class of cases, cases in which a person’s cognitive states influence the sub-personal level? Some cases in this category might be cases of what Siegel calls a-rational causation. To take a fanciful case, suppose my thinking about the number 354.333 causes certain changes in sub-personal processing prior to my visual experience. This ought not affect the capacity of the experience to provide me justification; the experience is evidence for me, despite its a-rational and epistemically unpropitious etiology. The harder question is what to think about cases in which the cognitive penetration operates in such a way that the resultant seeming appears to be action-like, i.e., appears to be related to beliefs/desires/intentions as actions are to cognitive states in Aristotelian practical inference, in practical inference in which the conclusion is an action. It seems at least conceptually possible that wishful seeming could work this way: it might seem to one that P because one intends it to seem to one that P, and one intends it to seem to one that P because one wants to believe P and knows the way to get oneself to believe that P is to make it seem to one that P. (The story about how such Aristotelian practical inference produces the seeming might refer to effects on sub-personal inferences or on the inputs to such inferences, or both.) Suppose, fancifully again, that I could, just by willing, affect sub-personal goings on in such a way as to produce in myself gold-seemings (or other seemings that would be appropriately connected to gold-seemings). Suppose I could do it in such a way as to rid myself of the memory that I did it and to ensure that I lacked other defeaters. I do it. Am I reasonable to believe the thing I see is a gold nugget? On the receptivity proposal as it stands, the answer would be yes. It would be yes because the seeming is not quasi-inferred. We could modify the receptivity proposal so that it gave an answer of no. We could say that when a seeming is the result of Aristotelian practical inference—one whose inputs are beliefs, intentions, desires and whose output is a seeming as “action”—this too is a kind of quasi-inference. It would be a kind of quasi-inference that, because of its practical nature, cannot transmit (epistemic) justification from any beliefs involved.33 We could make such amendments. They would answer to the feeling that there is something considerably epistemically worse when one affects one’s seemings in this way. There is also some independent theoretical plausibility to the amendments. After all, if a

33

Thanks to Susanna Siegel for discussion here.

244

Dealing with Cognitive Penetration

seeming is the conclusion of a piece of practical reasoning, then it is “made” by the person and not part of what the person is “handed.”34 One worry about making this modification is that we might get certain cases wrong. Compare the standard BIV case with the case of a person who makes himself a BIV by, say, taking a pill, or even by willing himself to be one, if one can imagine this. In the standard BIV case, someone else acts intentionally to produce the subject’s experiences (acts by affecting inputs to sub-personal processes). Intuitively, in the standard BIV case, the intentions and plans of the deceiver do not prevent the subject’s experiences from justifying. Those experiences count as evidence, albeit epistemically unfortunate evidence. In the self-made BIV case, much is the same except for the fact that one has brought the situation on oneself, either with the help of a pill or through sheer willing. It seems to me, however, that one is no less reasonable in one’s beliefs in the self-made BIV case—assuming all defeaters are eliminated, of course—than in the standard case. In the self-made case, as in the standard case, the subject has epistemically unfortunate evidence, and not merely unfortunate in externalistic respects but in ways that indicate epistemic vice or improper cognitive functioning. Still, it is evidence all the same. Presumably, in the self-made BIV case the subject’s previous desires/beliefs are related to the resulting seemings by Aristotelian practical inference (if not, why not?). So, while we could alter our account of quasi-inference to allow for quasi-inference via Aristotelian practical inference, there are reasons not to do so. I leave this matter open.

6. Conclusion So, what is the best option for phenomenal conservatives in the face of the “bad basis” examples? Appealing to defeaters and to the difference between the favored notion of justification (what it is reasonable to believe) and other epistemic statuses (e.g., blameless belief, virtuous belief, good cognitive functioning) can take the conservative some of the distance. Yet these approaches do not save her from counterexamples involving free enrichment. If Markie’s gold-digger seems to see gold because his wish to find gold leads him to “jump” to this seeming-state from his seeming to see something yellowish, then his resulting belief is not reasonable, contrary to PC. PC is refuted by such cases. However, we may revise PC to state that one is prima facie justified in believing P when one has a receptive seeming that P. In free enrichment cases, the output seeming isn’t receptive; it isn’t part of one’s basic evidence

34 The sense of “made” intended in our explanation of the intuitive picture in our thinking about reasonable belief was intellectual: what was “made” was concluded from, and accountable to, what was “handed” to one.

The “Bad Basis” Counterexamples

245

or the grounds one fundamentally is “handed.” It therefore cannot provide foundational justification. Nor can it provide non-foundational justification, because it depends on something very much like bad reasoning, viz., a bad “quasi-inference.”

Appendix: Receptivity and Doxastic Conservatism Doxastic conservatism (DC) holds that one is prima facie justified in retaining the beliefs one already has. DC is subject to “bad basis” counterexamples in droves, since belief is obviously affected by all sorts of psychological states of the person. Markie’s gold-digger surely isn’t reasonable to believe that the pebble is a gold nugget just because he does (wishfully!) believe it, even if he lacks defeaters. Is there any receptivity cure for DC’s ailments as we have argued there is, or plausibly is, for PC?35 What is needed is an account of receptivity for beliefs. The following seems to me promising: a belief is receptive, roughly, if it isn’t, at the time in question, something that is the “making” of the person, but rather is simply “handed” to her. A belief is not, at the time, an instance of what one makes of one’s situation if it is not held at that time on the basis of evidence or reasons (broadly construed so as to include non-doxastic states such as seemings). Thus, we might propose that a belief is receptive so long as it is not currently held on the basis of evidence or reasons. This would have the implication that perceptual beliefs, intuitive beliefs, introspective beliefs, testimonial beliefs, along with beliefs held on the basis of an inference all count as non-receptive. Receptive beliefs would be chiefly limited to beliefs retained from memory. I say “chiefly” because a belief that isn’t based on evidence or reasons that simply pop into one’s head would count as receptive by the account under consideration. Some will find this objectionable. However, here, as in our defense of the receptivity proposal for PC, there is a defense: this belief is part of what is “handed” to one at the time, not of what one makes of what is handed to one, and what is handed to one can provide foundational justification. These ideas are only a first sketch. Problems arise with DC that do not arise with PC. For instance, when they are not based in a quasi-inferential way on other seemings, seemings make a good candidate for being basic evidence. However, it is harder to see how a state of belief could be evidence for its content. How could one’s take on how things are be evidence for oneself at that very time that things are that way? So, to the extent that the intuitive idea of “what is handed to one” should be understood in terms of one’s basic evidence,

35 My earlier defense of doxastic conservatism (McGrath 2007) appeals to an expanded concept of defeat to handle such cases. I now think this answer is not sufficient.

246

Dealing with Cognitive Penetration

it would be harder to think of any beliefs as handed to one. I leave the sortingout of such issues for a later occasion.

Acknowledgments An earlier version of this chapter was presented at the 2011 University of Arkansas Epistemology Workshop. I thank audience members for helpful discussion. I am grateful to Alex Jackson, Peter Markie, Ted Poston, Tom Senor for helpful comments. I want to express particular gratitude to Chris Tucker and especially Susanna Siegel for their trenchant criticisms of the main ideas of the chapter.

References Audi, Robert. 1993. The Structure of Justification. Cambridge: Cambridge University Press. Braisby, Nick and Angus Gellatly, eds. 2005. Cognitive Psychology. Oxford:  Oxford University Press. Broome, John. 2002. “Practical Reasoning.” In José Bermùdez and Alan Millar, eds., Reason and Nature: Essays in the Theory of Rationality, 85–111. Oxford: Oxford University Press. _____. 1999. “Normative Requirements.” Ratio 12: 398–419. Dancy, Jonathan. 2000. Practical Reality. Oxford: Oxford University Press. Delk, John L. and Samuel Fillenbaum. 1965. “Differences in Perceived Color as a Function of Characteristic Color.” The American Journal of Psychology 78: 290–3.   Feldman, Richard. 2003. Epistemology. Upper Saddle River, NJ: Prentice-Hall. _____. 1988. “Having Evidence.” In David Austin, ed., Philosophical Analysis, 83–104. Dordrecht: Kluwer Academic. Goldman, Alvin 1988. Epistemology and Cognition. Cambridge, MA:  Harvard University Press. Goldstone, Robert L. 2003. “Learning to Perceive while Perceiving to Learn.” In Ruth Kimchi, Marlene Behrmann, and Carl R Olson, eds., Perceptual Organization in Vision: Behavioral and Neural Perspectives, 233–78. Mahwah, NJ: Lawrence Erlbaum. Goldstone, Robert L. and Andrew T. Hendrickson. 2010. “Categorical Perception.” Interdisciplinary Reviews: Cognitive Science 1: 65–78. Hansen, Thorsten, Maria Olkkonen, Sebastian Walter, and Karl R. Gegenfurtner. 2006. “Memory Modulates Color Experience.” Nature Neuroscience 9(11): 1367–68. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Jackson, Alexander. 2011. “Appearances, Rationality and Justified Belief.” Philosophy and Phenomenological Research 82: 564–93. Kolodny, Niko. 2005. “Why Be Rational?” Mind 114: 509–63. Kunda, Ziva. 1999. Social Cognition: Making Sense of People. Cambridge, MA: MIT Press. _____. 1990. “The Case for Motivated Reasoning.” Psychological Bulletin 108(3): 480–98. Lyons, Jack. 2011. “Circularity, Reliability and the Cognitive Penetrability of Perception.” Philosophical Issues 21: 289–311.

The “Bad Basis” Counterexamples

247

Markie, Peter. 2006. “Epistemically Appropriate Perceptual Belief.” Nous 40: 118–42. McGrath, Matthew. Forthcoming. “Dogmatism, Underminers and Skepticism.” Philosophy and Phenomenological Research. _____. 2007. “Memory and Epistemic Conservatism.” Synthese 157, 1–24. Pollock, John. 1975. Knowledge and Justification. Princeton, NJ: Princeton University Press. Pryor, James. 2004. “What’s Wrong with Moore’s Argument.” Philosophical Issues 14(1): 349–78. _____. 2000. “The Skeptic and the Dogmatist.” Nous 34: 517–49. Senor, Thomas. 2005. “The Epistemological Problems of Memory.” Stanford Encyclopedia of Philosophy. http://plato.stanford.edu/archives/fall2009/entries/memory-episprob/. Siegel, Susanna.2013. “The Epistemic Impact of the Etiology of Experience.” Philosophical Studies 162: 697–722. _____. 2012. “Cognitive Penetrability and Perceptual Justification.” Nous 46: 201–22. Silins, Nicholas. 2008. “Basic Justification and the Moorean Response to the Skeptic.” In Tamar Szabó Gendler, John Hawthorne, eds., Oxford Studies in Epistemology, Vol. 2, 108–41. Oxford: Oxford University Press. Sosa, Ernest. 2007. A Virtue Epistemology. New York: Oxford University Press. Tolhurst, William. 1998. “Seemings.” American Philosophical Quarterly 35: 293–302. Tucker, Chris. 2010. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24(1): 529–45. Turri, John. 2010. “On the Relationship between Propositional and Doxastic Justification.” Philosophy and Phenomenological Research 80(1): 312–26. White, Roger. 2006. “Problems for Dogmatism.” Philosophical Studies 31: 525–57. Wright, Crispin. 2004. “Warrant for Nothing (and Foundations for Free)?” Aristotelian Society Supplement 78(1): 167–212.

{ 11 }

Searching for True Dogmatism Peter J. Markie I’m a Mentalist, not the entertaining kind who can tell what you’re thinking right now, but the philosophical kind who believes that the following, or something very like it, is true. M If any two possible individuals are exactly alike mentally, then the same beliefs are (doxastically) justified for them to the same extent (see Conee and Feldman 2004; cf. Bergmann 2006).

I’m also a Foundationalist, the kind who believes: F We have some immediately justified (basic) perceptual beliefs about the external world, and we have some immediately justified (basic) beliefs in necessary truths through rational intuition.

As a Mentalist, I  wonder which mental states support justified beliefs. As a Foundationalist, I wonder which mental states support immediately justified beliefs, and I worry, in particular, about how to avoid what Sosa (2003) terms “the dreaded Sellarsian dilemma.” [E]ither the foundational conscious states have propositional content, in which case they would seem to require justification in turn, and could not after all function as a foundation; or else they have no propositional content of their own, in which case it is hard to see how they could possibly provide epistemic justification for any belief founded upon them. (212)

Dogmatism promises to address my wonder and ease my worry. It is the claim: D

Our seeming experiences (or at least some of our seeming experiences under some conditions) immediately support beliefs in their propositional contents.

Our seeming experiences epistemically support our basic beliefs; those experiences have propositional contents but require no epistemic justification. They are neither epistemically justified nor unjustified.

Searching for True Dogmatism

249

Here is where my search begins. There are two general forms of Dogmatism:  Unqualified Dogmatism and Qualified Dogmatism. Michael Huemer (2001) has shown how to develop Unqualified Dogmatism to support Mentalism and Foundationalism, but Unqualified Dogmatism is false. Michael Bergmann (forthcoming) has shown how to develop Qualified Dogmatism to support Foundationalism at the cost of rejecting Mentalism, but, as I said, I am a Mentalist. I seek a form of Qualified Dogmatism that supports both Mentalism and Foundationalism. That view, if it exists, is my kind of Dogmatism. It just might be Dogmatism that is true. Yet, that view is also hard to find. I have organized my search as follows. I start with a brief examination of Dogmatism in both its unqualified and qualified forms. Next, I  show that two major arguments for Unqualified Dogmatism are unsuccessful and that the view is open to compelling counterexamples. With Qualified Dogmatism as the remaining option, I  explore how best to develop it in the context of Mentalism and Foundationalism, focusing on one approach in particular.

1. Dogmatism: Unqualified and Qualified Unqualified General Dogmatism (UGD) is the following claim: UGD Necessarily, if it seems to S that P, then P is prima facie justified for S. (see Huemer 2007; Tucker 2010)

The precise range of Unqualified General Dogmatism is unclear. There are seemings, and then again there are seemings. Perceptual seemings, intellectual seemings, introspective seemings, and memorial seemings are all plausibly within the range of UGD. What, though, about hunches and gut intuitions? How about what Bealer (2000) and others call “physical intuitions?” What of seeming states that result from our inferring one proposition from others we believe? There’s no need to settle boundary issues now. I take perceptual and intellectual seemings to be paradigms of the sort of seemings covered by Unqualified General Dogmatism. I focus my attention on those cases and so on Unqualified Perceptual Dogmatism (UPD): UPD Necessarily, if it perceptually seems to S that P, then P is prima facie justified for S (see Huemer 2001; 2007; Pryor 2000; Tucker 2010).

and Unqualified Intellectual Dogmatism (UID): UID

Necessarily, if it intellectually seems to S that P, then P is prima facie justified for S (see Huemer 2005).

Our perceptual and intellectual seeming experiences give us prima facie justification for their contents.

250

Dealing with Cognitive Penetration

There are various accounts of seemings (see, for example, Huemer 2001, 2005, and 2007; Bergmann forthcoming; Sosa 2007; Tucker 2010), but some points are clear. First, seemings are assertive; in a seeming experience, the propositional content of the experience is presented to us as being the case. Seemings also have various sources. Perceptual seemings stem from, we might with equal vagueness say that they are “grounded in,” sensations. We have certain sensations and, as a result, it visually seems to us that there is a tree before us. Intellectual seemings stem from our understanding of a proposition. We understand the proposition that the shortest distance between two points is a straight line, and, as a result, it intellectually seems to us that the proposition is true. There may be spontaneous seemings, ones that just occur with no obvious source in other experiences or acts of understanding. Second, the principles of Unqualified Dogmatism concern what is necessarily the case. The relations they claim to hold between seeming experiences and prima facie epistemic justification stem, not from contingent facts about our nature, but from the nature of seeming experiences and epistemic justification themselves. Third, these principles are concerned, at least in the first instance, with propositional, rather than doxastic, justification. Doxastic justification is a property of beliefs; our doxastically justified beliefs are those we form on a basis that makes them epistemically appropriate or reasonable for us. Propositional justification is a relation between a person and a proposition. A proposition is prima facie propositionally justified for us just in case we have a basis for believing it. If we were to believe it on that basis, in the absence of rebutting or undercutting defeaters, our belief in it would be doxastically justified for us. According to the principles of Unqualified Dogmatism, seeming experiences—in general or in a particular form (e.g., perception)—are a source of prima facie propositional justification, and a potential source of doxastic justification. Fourth, principles of Unqualified Dogmatism assert that seeming experiences are a source of prima facie justification, independent of any background qualifications. The relevant seemings need not be reliable, need not result from our exercise of a cognitive virtue, or be an instance of cognitive function specified by our design plan, for example. The person who has the seeming experience need not have any reason to think it is of a truth-reliable sort. Qualified Dogmatism is the alternative view that seemings are a source of prima facie justification only when appropriate background conditions are met. Qualified General Dogmatism (QGD) is represented by the following schema, where R is a place holder for the appropriate background condition(s). QGD Necessarily, if R and it seems to S that P, then P is prima facie justified for S.

Qualified Perceptual and Intellectual Dogmatism can be stated in similar principles, containing background conditions on perceptual and intellectual

Searching for True Dogmatism

251

seemings, respectively. What background conditions might we impose on seeming experiences? It depends in part on our other philosophical commitments. Mentalism and Foundationalism are currently on the table. Mentalism implies that any background conditions are reflected in our mental states, which rules out various “Externalist” options (e.g., our seeming states must be truth-reliable or an instance of a properly functioning cognitive faculty). Foundationalism implies that the background conditions do not include any justified beliefs that keep the seeming experience from supporting a properly basic belief. Given this understanding of the territory, let’s start looking for Dogmatism that is true.

2. Unqualified Dogmatism: Failed Arguments We can set Unqualified Dogmatism aside. The main arguments for it are unsuccessful, and it is open to counterexamples that are especially compelling in the context of Mentalism. Let’s start with the failed arguments. According to Michael Huemer’s (2007) “self-defeat argument,” we ought to adopt Unqualified General Dogmatism because any denial of it is self-defeating. These statements by Huemer give the argument’s basic steps: My first premise is an empirical one, to the effect that, when we form beliefs, with a few exceptions not relevant here, our beliefs are based on the way things seem to us. (39) The second premise of the self-defeat argument is, roughly, that if one’s belief that p is based on something that does not constitute a source of justification for believing that p, then one’s belief that p is unjustified. (40) Given these premises, it follows that no belief is justified, unless one may have justification for believing that p in virtue of its appearing to one that p. If, that is, appearances do not confer at least some defeasible justification on propositions that are their contents, then since our beliefs are generally based on what seems to us to be the case (the reason we believe what we do is that it appears true to us; our method of forming beliefs is to believe what seems true to us), our beliefs are generally unjustified. (41)

As I understand it, the argument, as a defense of Unqualified Dogmatism, is the following. (1) Almost all our beliefs are based on the way things seem to us; that is, on some seeming experiences. (2) If our belief is based on something that does not, independently of meeting any background conditions, constitute a source of

252

Dealing with Cognitive Penetration

propositional justification for the content of the belief, then our belief is unjustified. (3) Therefore, if our seeming experiences don’t confer at least defeasible justification on propositions that are their contents, independent of meeting any background conditions, then our beliefs are generally unjustified. (4) If Unqualified General Dogmatism is false, then our seemings don’t confer at least defeasible justification on propositions that are their contents, independent of meeting any background conditions. Therefore, (C) If Unqualified General Dogmatism is false, then our beliefs are generally unjustified, including any belief we have that Unqualified General Dogmatism is false.

We thus deny Unqualified General Dogmatism at our peril. Note that we can extend the argument to the conclusion that Unqualified General Dogmatism is true, by adding the step that our beliefs are not generally unjustified. The argument’s first premise makes a claim about how we, presumably all human cognizers, form beliefs. We base almost all our beliefs on seeming experiences. As Huemer notes, this is an empirical claim, and lacking any other evidence, say, in the form of psychological studies (of which Huemer provides none), we are left to our own reflection to evaluate it. Mine contradicts it. I base some beliefs, indeed a good many beliefs, on other beliefs, rather than on seeming experiences. I form lots of beliefs inferentially, and the beliefs represented in the premises, not some accompanying seeming experience, are the basis for my belief in the conclusion. Perhaps I’m mistaken, however. Perhaps what I take to be my inferring one belief from another is really just my basing a belief on a seeming experience brought about by an act of inference. Mental processes are hard to penetrate. Let’s put the first premise aside and consider the rest of the argument. Something is still very wrong here, and one way to see it is to consider a possible group of inquirers, the “Inferentials.” They base some of their beliefs on other beliefs through inference, even if we don’t. One day, a bold epistemologist among the Inferentials has an insight: Naïve Inferential Justification Necessarily, if S believes that P and S inferentially bases her belief that Q on her belief that P, then Q is prima facie justified for S.

The insight is false. Simply inferring one belief from another is not sufficient to give the conclusion prima facie justification. A background condition must be met: the evidence belief must have some positive epistemic status. An inference cannot transfer epistemic status that is not there in the first place. Nonetheless,

Searching for True Dogmatism

253

our bold epistemologist defends the Principle of Naïve Inferential Justification by a self-defeat argument. (1) When we form inferential beliefs, our beliefs are based on other beliefs. (2) If our belief is based on something that does not, independent of meeting any background conditions, constitute a source of propositional justification for the content of the belief, then our belief is unjustified. (3) Therefore, if beliefs don’t confer at least defeasible propositional justification on the contents of the beliefs we base on them, independent of meeting any background conditions, then our inferential beliefs are generally unjustified. (4) If Naïve Inferential Justification is false, then beliefs don’t confer at least defeasible propositional justification on the contents of the beliefs that are based on them, independent of meeting any background conditions. Therefore, (C) If Naïve Inferential Justification is false, then our inferential beliefs are generally unjustified, including any inferential belief we have that Naïve Inferential Justification is false.

The Inferentials thus see that they deny Naïve Inferential Justification at their peril, and, if they add the premise that they are generally justified in their inferential beliefs, the argument supports the conclusion that the Principle of Naïve Inferential Justification is true. What has gone wrong? The second premise is false. The mere fact that an inferential belief is based on something (in this case, another “premise” belief) that is not a source of justification unless a particular background condition is met (the content of the premise belief must itself be at least prima facie justified) does not imply that the inferred belief is unjustified. The background condition may be met, after all. Huemer’s argument contains the same mistaken second premise. In this case, the mere fact that a belief is based on a seeming experience that is not a source of justification unless a particular background condition is met does not imply that the belief is unjustified. Once again, the background condition may be met. Anticipating this objection, Huemer (2007) places a causal constraint on the sort of background conditions that can apply to seeming states that provide epistemic support. Using the phrase, “Restricted Phenomenal Conservatism,” to designate Qualified Dogmatism, he tells us: One who would avoid the self-defeat argument by proposing a restricted phenomenal conservatism, then, must argue not only that there is some

254

Dealing with Cognitive Penetration

epistemically relevant difference between some appearances and others, but also that this difference makes a difference, causally, to what we believe. It is unlikely that this constraint can be satisfied. (43)

If seeming experiences must meet some background condition to provide epistemic support for beliefs based on them, then the satisfaction of that background condition must play a causal role in the formation of the beliefs. This constraint is indeed unlikely to be met. It is in the nature of background conditions to be normative and in the nature of normative states to be causally ineffective. Imagine though that our Inferential epistemologist makes the same move to defend his argument for the Principle of Naïve Inferential Justification. He offers a similar causal constraint on background conditions: if the contents of beliefs must meet the background condition of having prima facie justification for the beliefs to provide epistemic support for other beliefs, then that background condition must play a causal role in the formation of the beliefs, which of course, it doesn’t. His audience will surely catch the mistake. The causal constraint is false. That beliefs must meet some background condition to support other beliefs that are based on them does not imply that the background condition plays a causal role in belief formation. The same goes for seeming experiences. That seeming experiences, like beliefs, must meet a background condition to support beliefs based on them does not imply that the background condition plays a causal role in belief formation. These problems eliminate the self-defeat argument as an argument for Unqualified General Dogmatism. We can avoid them if we revise the argument as a defense of Dogmatism in its unqualified or qualified forms. Consider: (1) Almost all our beliefs are based on the way things seem to us; that is, on some seeming experiences. (2) If our belief is based on something that does not, under any conditions, constitute a source of propositional justification for the content of the belief, then our belief is unjustified. (3) Therefore, if seemings don’t, under any conditions, confer at least defeasible justification on propositions that are their contents, then, since our beliefs are generally based on what seems to us to be the case, our beliefs are generally unjustified. (4) If no form of either Unqualified or Qualified Dogmatism is true, then seemings don’t, under any conditions, confer at least defeasible justification on propositions that are their contents. Therefore, 5) If no form of Unqualified or Qualified Dogmatism is true, then our beliefs are generally unjustified, including any belief we have that no form of Unqualified or Qualified Dogmatism is true.

Searching for True Dogmatism

255

If we gain good reason to adopt the argument’s first, empirical premise that all our beliefs are based on seeming experiences, the argument gives us a reason to adopt Dogmatism, though not Unqualified Dogmatism. Chris Tucker (2010) argues for Unqualified General Dogmatism on the basis of its explanatory power relative to four epistemological issues. The first issue is the Speckled Hen. When we have a visual image of a speckled hen with just three speckles, we are directly aware of the image and its speckles and the proposition that the image has three speckles is prima facie justified for us. When we have the visual image of a speckled hen with forty-eight speckles, we are again directly aware of the image and its speckles, but the proposition that the image has forty-eight speckles is not prima facie justified for us. Unqualified General Dogmatism explains the difference, according to Tucker. In normal humans, a visual image of a three-speckled hen is accompanied by a seeming that the hen has three speckles, and the visual image of the forty-eight-speckled hen is not accompanied by a seeming that the hen has forty-eight speckles. This difference in the way things seem is what explains the difference in justification. (12)

Tucker’s second issue concerns the relative support relation between sensations and propositions. The olfactory sensation that gives us evidence that a flower is nearby might give another sort of creature evidence that she has a small, hard object in her hand. How is it that the sensation that leads us to have evidence for one proposition leads another to have evidence for a different one? Unqualified General Dogmatism provides an answer. The dogmatist, then, can explain these species-wide differences in justification by appealing to species-wide differences in the way things seem. (16)

The same sensation causes different seeming states in different species, and the different seeming states epistemically support different propositions. It seems to us that a flower is nearby; it seems to the other creature that there’s a small, hard object in her hand. The third issue concerns the difference between experts and non-experts. Although the landlubber and seafarer both see a dolphin, the proposition that there is a dolphin is only prima facie justified for the latter, who has learned what dolphins look like. Unqualified General Dogmatism explains the difference despite the similarity in their visual experience. When seafaring Captain Jack sees a dolphin, it seems to him that the creature is a dolphin, and when Hillbilly considers the image, it does not seem to him that the creature is a dolphin. This difference in the way things seem explains why Jack, but not Hillbilly, has justification that the creature is a dolphin. (17)

256

Dealing with Cognitive Penetration

The fourth issue concerns the epistemic similarity between ourselves and our counterparts in a demon world. Why are the same propositions justified for us and our counterparts and why, despite this fact, do we have knowledge when our counterparts do not? The dogmatist has an intuitively appealing answer at her disposal: the demon victim’s beliefs are justified because they are appropriate responses to her seemings, but they are not knowledge because her seemings, unbeknownst to her, were caused in an inappropriate way. (20)

Even though our seeming experiences are caused by objects of the sort they represent and our counterparts’ are caused by a deceiving demon, we all have the same seeming experiences and that is enough to make the same propositions prima facie justified for us; the difference in the cause of our seeming experiences explains why we know and our counterparts do not. Unqualified General Dogmatism certainly explains each issue, but Qualified Dogmatism explains them just as well. Consider the Speckled Hen. When we see the three-speckled hen, it seems three-speckled to us and the requisite background conditions are met; when we see the forty-eight speckled hen, it does not seem forty-eight-speckled to us or, if it does (perhaps due to the influence of a brain tumor, ray gun, witch’s spell, or whatever), the requisite background qualifications are not met. Consider the issue of the relative support relation between sensations and propositions. When we have a flowerysmell sensation, it seems to us that a flower is present and, in response to the same sensation, it seems to some other creature that she has a small, hard object in her hand. Insofar as different propositions are justified for us and the other creature, it is because we have different seeming states in the presence of required background conditions. Consider the difference between the seafarer and the landlubber. The proposition that there is a dolphin is prima facie justified only for the seafarer. This is because only the seafarer has the seeming experience that there’s a dolphin in the context of the necessary background qualifications. The landlubber either doesn’t seem to see a dolphin or, if he does (brain tumors, ray guns, witches, and such again), he doesn’t meet the appropriate background qualifications. With regard to the fourth case, we can explain the similarity in justification between our demon world counterparts and ourselves by the fact that we all have the same seeming experiences under the same background conditions. We can explain the difference in knowledge by appeal to the fact that we meet, but they don’t, some external condition for knowledge, e.g., our beliefs are true and theirs are false. In short, Tucker’s inference to the best explanation, like Huemer’s self-defeat argument, gives us some reason to accept Dogmatism, but it does not differentiate between the unqualified and qualified forms. Let’s now consider some good reasons to reject Unqualified Dogmatism.

Searching for True Dogmatism

257

3. Unqualified Dogmatism Here is a counterexample to Unqualified Perceptual Dogmatism.1 Prospectors Gus and Virgil are gold prospectors. Gus is an expert at identifying gold. He has learned to do so through long experience. He began with a list of identification rules and consciously applied them. He then reached the point where he could “just see” that a nugget is gold. Virgil is a novice. He has a general sense of what gold looks like, but he is not very good at its visual identification. Virgil, though, is consumed by a lust for gold. He wants very, very badly to make a discovery. When Gus looks at a nugget in his pan, his developed gold-identification abilities come into play and it just seems to him that it’s gold. He believes accordingly. When Virgil looks at his nugget, his strong desire that it be gold comes into play and as a result, it just seems to him that it’s gold. He too believes accordingly. Each believes that his nugget is gold on the basis of his visual seeming experience. Yet, where Gus sees what he has learned to see, Virgil sees what he wants to see. This makes a difference epistemically. Gus’s belief is epistemically appropriate in a way in which Virgil’s is not.

The perceptions of Gus and Virgil are examples of the penetration of a visual experience by other mental states so as to cause things to seem a certain way. Gus’s learning process causes his nugget to seem to be gold to him. Virgil’s lust for gold makes his nugget seem to be gold to him. Prospectors poses a serious challenge to Unqualified Perceptual Dogmatism. It is clear that Gus’s belief is epistemically appropriate in a way in which Virgil’s is not, but what is the difference? A plausible answer is that Gus is prima facie justified, and in the absence of any defeaters just plain justified, in his belief, while Virgil is not even prima facie justified in his belief. Fans of Unqualified Perceptual Dogmatism may not say this, however. According to their view, Gus and Virgil are both prima facie justified in their beliefs. The content propositions of their beliefs are prima facie justified by their seeming experiences, and they base their beliefs in those propositions on those seeming experiences. Unqualified Dogmatists need another explanation of the epistemic difference between the beliefs of Gus and Virgil, but the available alternatives won’t do. Take the defeasibility approach: Gus and Virgil are both prima facie justified in their beliefs, but only Gus is actually justified; the epistemic support provided by his seeming experience is undefeated, while that provided by Virgil’s seeming experience is defeated. There are three significant problems

1 A reminder: since Unqualified Dogmatism is offered as a necessary truth, successful counterexamples to it need only be logically possible.

258

Dealing with Cognitive Penetration

here. First, if Mentalism is true, there’s no guarantee that Virgil has a defeater for his seeming experience. There are propositions, including true ones, that rebut or uncut his seeming experience, e.g., that he strongly wants the nugget to be gold and his strong desires have inaccurately “colored” his perceptions in the past. Yet, given Mentalism, these propositions are not defeaters for Virgil unless they are somehow reflected in his mental states, most likely by his believing them, and they may not be. It is quite possible that Virgil does not believe, or even consider, these potential defeaters. Second, the defeasibility approach locates Virgil’s error in the wrong place. It is not that Virgil shouldn’t form his belief on the basis of his seeming experience, because its epistemic support is defeated. It is that he should not have his seeming experience in the first place. To put the point another way, it is not that he has an epistemically appropriate seeming experience but ignores counter or undercutting evidence to form an unjustified, and in that sense inappropriate, belief. It is that his seeming experience is itself epistemically inappropriate. His nugget shouldn’t seem to be gold to him in the way that Gus’s nugget should seem to be gold to Gus.2 Third, the defeasibility approach has an implausible implication. Consider how Siegel (2012) makes the point. [S]uppose Jill believes that Jack is angry at her, and this makes her experience his face as expressing anger. Now suppose she takes her cognitively penetrated experience at face value, as additional support for her belief that Jack is angry at her (just look at his face!). She seems to have moved in a circle, starting out with the penetrating belief, and ending up with the same belief, via having an experience. From Jill’s point of view, she seems to be gaining additional evidence from this experience for her belief that Jack is angry at her, elevating the epistemic status of that belief. This situation seems epistemically pernicious. (2)

Jill’s completely unsupported belief that Jack is angry penetrates her perceptions, causing Jack to visibly seem angry to her. According to the defeasibility approach, her seeming experience provides epistemic support for her belief. Yet, it is hardly plausible that Jill can gain epistemic support for an otherwise unsupported belief by having that very belief penetrate her perceptions. Another way to deal with Prospectors within Unqualified Perceptual Dogmatism is the knowledge approach: Gus and Virgil are both prima facie justified in their beliefs, and even justified in the absence of defeaters; the epistemic difference between them is that Gus knows that his nugget is gold and

2 This is not to say that Virgil’s seeming experience is epistemically unjustified. Seeming experiences are neither justified nor unjustified. They can, though, be epistemically appropriate or inappropriate in some other way. One task facing Qualified Dogmatists is to explain the nature of this difference.

Searching for True Dogmatism

259

Virgil does not. Virgil’s belief is epistemically inappropriate in that it is not a case of knowledge. The knowledge approach retains the third defect in the defeasibility approach:  it implies that we can gain epistemic support for an otherwise unsupported belief by having that very belief penetrate our perceptions. Another problem with the knowledge approach is that there’s no reason to assume that Gus knows, while Virgil does not. We can locate Prospectors in a demon world in which there is no gold and Gus and Virgil have always been fed their experiences by a demon. Neither has knowledge, but Gus’s belief is still epistemically appropriate in a way that Virgil’s is not. His belief is based on a seeming that results from a learning process involving various reinforcing justified beliefs and experiences, while Virgil’s is just the result of his desire. Tucker (2010) proposes a third way to deal with Prospectors within Unqualified Perceptual Dogmatism, the blameworthiness approach. Noting that our expert, Gus, and novice, Virgil, might be located in a demon world, he suggests that we can explain the epistemic difference between them, by pointing out that Novice is blameworthy for his false belief and his inappropriately-caused seeming and Expert isn’t. The suggestion, then, is this: both Novice and Expert have justification that the nugget is gold, but Novice is nonetheless blameworthy in a way that Expert isn’t. (22)

This strategy again retains the third defect of the defeasibility approach, implying that we can gain epistemic support for an otherwise unsupported belief by having that very belief penetrate our perceptions. Another problem emerges when we ask, What is the negative outcome for which Virgil is epistemically blameworthy? Tucker blames him for two things: having the false belief that his nugget is gold and having an inappropriately caused seeming (22). Yet, if Prospectors takes place in a demon world, Virgil is hardly to blame for having a false belief. The demon ensures that all his external world beliefs are false no matter what he might do. And, if we locate Prospectors in the real world, Virgil may well have a true belief. It is equally hard to see how Virgil can be blameworthy for having an inappropriately caused seeming. What can be epistemically inappropriate about how Virgil’s seeming experience is caused, if the manner of its causation leaves it quite capable of justifying propositions for him? Perhaps, the idea is that his seeming experience is inappropriately caused in that it is caused in such a way that it does not meet some external condition for knowledge, rather than justification, e.g., it does not result from the proper functioning of his faculties or from a reliable belief-forming process, and Virgil is responsible for that fact. He should have ensured that his desire did not penetrate his perception, and, if he had, he would not have formed his belief in a way that was incapable of supporting knowledge. Yet, this again misses what is wrong with Virgil’s belief. We can modify the case so that Virgil meets the relevant external condition, keeping his internal mental state the same. Perhaps, his design plan calls for his desires to penetrate his perceptions.

260

Dealing with Cognitive Penetration

Perhaps, such perceptions are reliable in his environment. He still forms his belief on the basis of a seeming experience resulting from the penetration of his visual perception by his desire. That alone makes his belief epistemically inappropriate in a way in which Gus’s belief is not.3 Prospectors is a counterexample to Unqualified Perceptual Dogmatism. Here is a counterexample to Unqualified Intellectual Dogmatism. Mathematics Machine Elias has a fuller understanding of fractions than Igor and can rationally intuit a truth Igor cannot: (P) 237/148 is greater than 425/266. Nonetheless, they both understand (P), as they both believe it. Elias believes (P) on the basis of his rational intuition. Igor believes it on the basis of Elias’s testimony. (P) just seems true, indeed necessarily true, to Elias. (P) neither seems true nor false to Igor. In an attempt to intuit (P), Igor starts to study fractions, but he is unable to attain Elias’s level of comprehension, and Elias decides to help him. He hooks Igor up to the mathematics machine while he sleeps. He types in (P) and the machine makes an adjustment in Igor’s brain. From then on, when Igor considers (P) and understands it as he did previously, which is just well enough to believe it, it seems to him that it is necessarily true. Elias uses the machine to make similar adjustments for some other mathematical truths Igor understands well enough to believe but does not rationally intuit. When Igor wakes, he knows nothing of what Elias has done. Some days later, when Igor again considers (P), its truth just seems obvious to him. He proclaims, “Now I see it!” and stands taller for his sense of accomplishment, all the better to look down on those who aren’t as smart as he is.

Elias’s belief that (P) is epistemically appropriate in a way in which Igor’s is not. Being hooked up to the mathematics machine is not equivalent to learning mathematics. Elias has a background knowledge of fractions that enables him to “rationally grasp” the truth of (P); the proposition seems true to him as a result of his deep understanding of its content. Given his deep understanding of (P), it should seem true to him. Igor has no such deep understanding of (P). He never “rationally grasps” its truth. The proposition seems true to him because the mathematics machine modifies his brain so that his 3 The knowledge and blameworthiness approaches appear to be based on the idea that Virgil’s case is actually a sort of “Gettier” case in which the subject has a justified true belief but not knowledge. Yet, Virgil’s case is markedly different from the original Gettier cases or their subsequent variations. In those cases, there is nothing inappropriate about the internal belief-forming process. We can remove any problem by just modifying external features (the subject’s belief about who will get the job and has ten coins in his pocket is not false; the valley in which the subject perceives a barn is not salted with fakes). In Virgil’s case, there is something wrong with his internal belief-forming process: his seeming state is penetrated by an inappropriate mental state.

Searching for True Dogmatism

261

meager knowledge of fractions and limited understanding of (P) are sufficient to make it seem to him that (P) is true. Relative to his limited understanding, (P) should not seem true to him.4 Unqualified Intellectual Dogmatism closes the door on an obvious explanation of how Elias’s belief in (P) is epistemically appropriate in a way in which Igor’s is not: (P) is prima facie justified for Elias but not for Igor. The remaining ways to explain the epistemic difference are again inadequate. The defeasibility approach won’t do, as, especially given Mentalism, Igor need not have a defeater for his belief in (P), and the approach again locates his epistemic error in the wrong place. It is not that he has an epistemically appropriate seeming experience but ignores counter or undercutting evidence in forming the belief that (P). It is that his seeming experience itself is epistemically inappropriate. Given his limited knowledge of fractions, (P) should not seem true to him. The knowledge approach is also a dead end. We can modify the example so that neither Elias nor Igor knows, by replacing (P) with a false proposition that, on the one hand, seems true to Elias due to his understanding of fractions and a subtle mistake, and, on the other hand, seems true to Igor due to a connection established by the mathematics machine. The blameworthiness approach fares no better. There is nothing bad for which Igor is blameworthy. Since his belief is true, he can’t be properly blamed for having a false belief, and, according to the blameworthiness approach, his seeming experience supports his belief ’s justification. Perhaps the idea is again that his belief is formed in way that does not meet some external condition for knowledge, but it is hard to see how Igor is blameworthy for that. He doesn’t hook himself up to the mathematics machine. He doesn’t let his desires or other irrelevant mental states “intrude” on his intellectual seemings.5

4 Note the historical precedent for the claim that a rational intuition that (P) requires a degree of understanding beyond that required for mere belief. Descartes writes in The Rules for the Direction of the Mind that intuition is “the indubitable conception of a clear and attentive mind which proceeds solely from the light of reason” (14; my emphasis). Consider too Leibniz’s comment in The Discourse on Metaphysics, 24: “When my mind understands at once and distinctly all the primitive ingredients of a conception, then we have intuitive knowledge” (41-42). On the contemporary scene, Sosa (2007) takes rational intuitions to be intellectual seemings that are explained by our ability to discriminate between true and false modally strong propositions that we “understand well enough” with no further reliance on introspection, perception, memory testimony or inference (p. 61). Bealer (2000, p. 11) requires that in rational intuition we must possess the relevant concepts in a full sense: “A subject possess a concept in the full sense iff (i) the subject at least nominally possesses the concept and (ii) the subject does not do this with misunderstanding or incomplete understanding or just by virtue of satisfying our attribution practices or in any other such manner” (11). 5 Proponents of the knowledge and blameworthiness strategies treat Igor’s case as a sort of “Gettier” case in which all the internal conditions for knowledge are met but the subject fails to meet some external consideration. That’s not Igor’s problem. His defect is internal: his understanding of the proposition is not sufficient to support, in him or any other possible being, an epistemically appropriate intellectual seeming experience of the proposition’s truth. For more discussion of Mathematics Machine and its implications for various accounts of rational intuition, see Markie (2013).

262

Dealing with Cognitive Penetration

Prospectors and Mathematics Machine make the same point. Some seeming experiences are epistemically appropriate and others are not. Gus’s nugget should seem to be gold to him, given his learned identification ability, and (P) should seem to be true to Elias given his sophisticated understanding of fractions. Virgil’s nugget should not seem to be gold to him, given his ignorance of what gold looks like, and (P) should not seem true to Igor, given his limited understanding of fractions. Beliefs based on epistemically appropriate seeming experiences are epistemically superior to beliefs based on epistemically inappropriate ones. Unqualified Dogmatism does not account for this fact. Dogmatism that is true, if there is such a thing, is Qualified Dogmatism.

4. How to Be a Qualified Dogmatist I want to explore one way to develop Qualified Dogmatism, the Knowledge How Proposal. At a high enough level of generality, the same background conditions should explain the difference between epistemically appropriate and epistemically inappropriate seeming experiences, be they perceptual, intellectual, and so on. It is easiest to begin, though, by investigating perceptual seemings and the background conditions applicable to them. Given Mentalism, the background conditions on perceptual seeming experiences are all reflected in internal mental states. Given Foundationalism, those conditions do not require justified beliefs that prevent the associated seeming experiences from supporting basic beliefs. What might they be? Recall Prospectors. It seems to Gus that his nugget is gold, and his epistemically proper seeming experience, rooted in his gold identification ability, evidentially supports his belief. It seems to Virgil that his nugget is gold, and his epistemically improper seeming experience, rooted in his lust for gold, doesn’t evidentially support his belief. Each has a visual experience, each attends to certain features of his experience—perhaps even the same features—and each is caused, by attending to those features, to have the seeming experience that his nugget is gold. The difference between them is that Gus knows how to visually identify gold nuggets, and his seeming state and resulting belief are an instance of his exercise of this know-how; Virgil’s seeming experience is not an instance of any such know-how. Perhaps then an epistemically appropriate perceptual seeming experience to the effect that something is Q is one that is had in the exercise of the subject’s knowledge of how to perceptually identify something as being Q. We need to make two assumptions for this proposal to be plausible. First, “knowledge how” does not require reliable practice. Consider the demon world edition of Prospectors. Demon World Gus has an epistemically appropriate seeming experience and Demon World Virgil does not, just like their actual world counterparts. On the current proposal, Demon World Gus knows

Searching for True Dogmatism

263

how to perceptually identify gold nuggets, even though all of his identifications are mistaken. This is just right. We can know how to do something but fail to do it reliably if our failure is the result of some external intervention. I might learn how to fly a plane in a flight simulator, know how to fly and exercise my knowledge but never actually fly, because my fights are, perhaps even unbeknownst to me, always in the simulator. This is what happens to Gus in the demon world edition of Prospectors. Second, being able to reliably identify things as being Q is not sufficient for knowing how to identify something as being Q. It might be that whenever Virgil looks at a nugget and attends to certain features of his visual experience, he has the desire that the nugget be gold. It might also be that his desire is enough to make it seem to him that the nugget is gold. It may even be that features of his experience that produce his desire are just the ones that are usually produced in visual perception by gold nuggets. Virgil’s gold nugget identifications are reliable. Nonetheless, his seeming experiences are not epistemically appropriate; on the Knowledge How Proposal, he does not know how to visually identify gold nuggets. Reliable practice is not sufficient for knowledge how. In this way, knowing how to identify something is analogous to having a moral virtue. Reliably performing just acts is not, as Aristotle (Nicomachean Ethics, Book II, Chapter 4, 1105a, 17–34) points out, sufficient for having the virtue of justice; the performance must be accompanied by a proper knowledge, intention, and choice. Analogously, reliably identifying things as having a certain property is not sufficient for knowing how to perceptually identify that property; the performance must be accompanied by a proper understanding.6 What then is this knowledge how that Gus has but Virgil, even reliable Virgil, lacks? I  suggest, the following. First, Gus has a disposition to have a gold-seeming experience in response to attending to certain features of his visual perception. Virgil might well have this much. Second, Gus has the information that what looks like that—what presents those features in his visual perception—is gold. Virgil lacks this information. Third, having the disposition and having the associated information are not isolated parts of Gus’s cognitive function; the information “informs” the disposition. He has the disposition, at least in part, by virtue of having that information. This is not to say that his having the information causes him to develop the disposition. It is to say that his having the information helps to determine the character of the disposition—his disposition is a disposition to have that seeming experience in response to attending to those features in part because he has this information—and helps to sustain it.

6 Consider Aristotle on a nonmoral case, being a grammarian: “A man will be a grammarian, then, only when he has both done something grammatical and done it grammatically; and this means doing it in accordance with the grammatical knowledge in himself ” (Nicomachean Ethics, Book II, Chapter 4, 1105a, 24–27).

264

Dealing with Cognitive Penetration

In what way, though, does Gus have the information that what presents certain features in his experience is gold? He may not have any conscious belief to this effect. A plausible option is that he has this information in that he has evidence (experiences, justified beliefs) that prima facie justifies for him the proposition that what looks that way is gold. He may not actually believe that proposition, but he has evidence sufficient to prima facie justify it, and, in that minimal way, he has the information it contains.7 Here then is a suggestion:  knowing how to perceptually identify something as being Q involves, first, a disposition to have a seeming experience that it is Q as a result of attending to certain features of one’s perception, and, second, having that disposition by virtue of having the information that what presents those features—what looks (tastes, smells, feels, sounds) like that—is Q.8 This is all quite sketchy. We need to know more, especially about how the information is had and how it informs the associated disposition. Rather than attempt to make the proposal more precise, however, I want to now consider some of its additional features. This way of developing Qualified Perceptual Dogmatism is consistent with Mentalism, but what of Foundationalism? How can Gus’s gold-nugget belief be basic, if the seeming experience on which it is based derives its own epistemic appropriateness, and so its ability to epistemically support his belief, from his having background information? Consider, first, the form that the background information takes. It is evidence that prima facie justifies for Gus the proposition that what looks a certain way is gold. It does not include that proposition, which he may not even believe. Note too the role that the background information plays. It is not part of Gus’s evidence for his belief that his nugget is gold. He forms that belief just on the basis of his seeming experience. The background information also is not a basis for the seeming experience that makes it justified or unjustified. The seeming experience is neither justified nor unjustified. The background information is not even a basis for the seeming experience that makes it epistemically appropriate rather than inappropriate. The seeming experience is not based on anything. It is epistemically appropriate because 7 Consider a similar suggestion by Steup (1996). Noting that his being appeared to dodecagonally does not evidence for him that there is a dodecagon present, but his being appeared to redly does evidence for him that something is red, he offers the following explanation:

[T]he difference is that I have evidence for taking myself to be reliable in discerning the color red, whereas I do not have evidence for taking myself to be reliable in discerning dodecagons. (105) For criticisms of Steup’s approach, criticisms I now see are mistaken, see Markie (2004). Another option is that Gus has a subconscious belief that what looks a certain way is gold, but on this approach, we still need to require that the proposition be at least prima facie justified for Gus in order to distinguish his position from variations on Virgil’s situation. 8 This understanding of perceptual knowledge how differs significantly from what I  present in Markie (2006) in its inclusion of a requirement of prima facie justification for the proposition that what presents certain features has a certain property.

Searching for True Dogmatism

265

it is an instance of an identification disposition that is determined by—has the character it does in part because of—Gus’s possession of the background information and is sustained by it. If an immediately justified (basic) belief is, as traditionally understood, one that is justified but does not gain its justification from being based on any other justified beliefs (or, more broadly than beliefs, psychological states), then Gus’s nugget belief is immediately justified.9 We can, of course, define forms of immediate justification that are at odds with the Knowledge How Proposal. A belief is immediately justified only if it does not gain its justification from being based on a psychological state that is itself open to some form of epistemic evaluation. Or, a belief is immediately justified only if its justification does not depend on any other justified beliefs of the subject. Yet, such alternatives require defense. We don’t need to adopt either to have a category of immediately justified beliefs capable of ending the chain of justification.10 A second concern about the Knowledge How Proposal is whether it can explain one of the examples considered earlier, Jill’s misleading perception that Jack is angry. Recall that Jill’s unjustified belief that Jack is angry penetrates her perception in such a way that he looks angry to her. Might not Jill be having her inappropriate seeming experience as an instance of her knowledge of how to perceptually identify anger in others though their facial expressions? If so, the Knowledge How Proposal is a nonstarter. We need to consider the details of Jill’s case. She looks at Jack, attends to certain features of her experience, and as a result has the seeming experience that he is angry. How does her prior belief influence this process? One possibility is that it intervenes in the causal relationship between her attending to certain features of her experience and her having the seeming experience. The features to which she attends are not such that her attending them would 9 Sosa (2007) distinguishes between two forms of foundational justification. Basis-dependent foundational justification is foundational justification that derives essentially from the justified belief ’s being based on a given state, a psychological state of the subject’s, one that lies beyond justification and unjustification (50). Virtue foundational justification is foundational justification that derives essentially from the justified propositional attitude’s manifesting an epistemic competence (51).We can say that Gus’s belief that his nugget is gold has basis-dependent foundational justification as it is based on his seeming experience, which itself lies beyond justification and unjustification. Alternatively, we can forgo any reference to Gus’s seeming experience as evidence for his belief and, in a variation on Sosa’s notion of virtue foundational justification, say that Gus’s belief has justification that derives from its being formed through the exercise of his knowledge of how to perceptually identify gold. 10 The last option also threatens to reduce Foundationalism to pure implausibility. As noted by Sosa (2003, 209):

Given that beliefs would not so much as exist without an extensive supporting cast of related beliefs, there is an air of unreality about the foundationalist claim that beliefs might nevertheless be justified independently of other beliefs. . . . [Y]ou could not possibly so much as host the target belief without a lot of the relevant supportive beliefs. Nor does it seem that you could enjoy justification for the target belief in the absence of justification for a good number of those supporting beliefs, absent which you could not hold the target belief at all.

266

Dealing with Cognitive Penetration

normally cause her to have the seeming experience that Jack is angry. Her knowledge of how to visually identify anger includes no disposition to have anger-seeming experiences in response to attending to features of that sort. Somehow, however, her belief that Jack is angry enters into the mix is such a way that, in this case, her attending to those features does cause her to have an anger-seeming experience. If this is how the case goes, it doesn’t pose a problem for the Knowledge How Proposal. Jill does not have her seeming experience in the exercise of her knowledge how to visually identify anger. She is not exercising her knowledge how. She is prevented from doing so by the intervention of her prior belief. Jill’s prior belief may intervene in her perception in a different way. Perhaps, it causes her experience to have features that it would not otherwise have, ones she has learned to take as indications of anger, so that her attending to them does, in the exercise of her knowledge how, cause her to have the seeming experience that Jack is angry. Now her seeming experience counts as appropriate and supports her belief, according to the Knowledge How Proposal. And this is just how things should be. Jill’s believing, with no justification, that Jack is angry results in her gaining epistemic support for that very belief, but there is nothing epistemically pernicious here, so long as the causal relation between Jill’s belief and her seeming experience is mediated by her application of her knowledge of how to visually identify anger. Her belief is functioning as the demon does in the standard demon world case, causing her experience to have features that, with the application of her knowledge of how to perceptually identify various sorts of things, causes her quite properly to have the associated beliefs.11 With worries about Foundationalism and cases like that of Jill addressed, the Knowledge How Proposal is promising. One other point merits attention, though. Can the approach be extended beyond perceptual seeming experiences to support other forms of Qualified Dogmatism? There should, after all, be a unified account of the epistemic appropriateness of all the seeming experiences countenanced by our favored form of Qualified Dogmatism. The results here are mixed. The promise of the Knowledge How Proposal hinges on how extensive a form of Qualified Dogmatism we seek. The proposal can be extended to cover intellectual seemings: an epistemically appropriate intellectual seeming experience to the effect that p is one that is had in the exercise of the subject’s knowledge of how to intellectually identify some

11 For another case, suppose that I believe without justification that the food you just placed before me is poisoned and my belief causes me to examine the food with extra care, enough care that, using my learned abilities to detect certain poisons by smell, I smell the poison that you have put there. I have the seeming experience that the food is poisoned. My seeming experience is appropriate and it supports for me the proposition that the food is poisoned. My seeming experience results from an initially unjustified belief in its propositional content, but it does so through my exercise of my knowledge how, and that makes all the difference.

Searching for True Dogmatism

267

propositions as true. That knowledge how involves the subject’s having a disposition to have intellectual seeming experiences in response to considering particular propositions, by virtue of having a sufficient understanding of the concepts they involve. The subject’s understanding of the concepts “informs” her disposition to have certain intellectual seeming experiences, just as, in the perceptual case, her information about how certain things are presented in her experience informs her disposition to have certain perceptual seeming experiences. In Mathematics Machine, both Elias and Igor can identify (P)  as true through an intellectual seeming experience. The difference between them is that Elias has his identification ability by virtue of his sufficiently full understanding of the relevant concepts in (P), while Igor has his by virtue of the connections established in his brain by the mathematics machine.12 There are, of course, serious questions. What, for instance, makes Elias’s understanding of the relevant concepts “sufficiently full?” What is the nature of his understanding, especially if it does not take the form of a series of beliefs that in turn are part of the evidential basis for his belief that (P)? How does his understanding inform his disposition to have intellectual seeming states? These are all difficult issues, but, nonetheless, intellectual seemings appear amenable to the Knowledge How Proposal (cf. Sosa 2007). The proposal has its limits, however, and memorial seeming experiences appear to be beyond them. If we believe that p based on a memorial seeming experience, we do not appear to apply any form of knowledge how. We do not exercise an ability to identify some proposition as true in response to attending to phenomenological features of our experience or through our understanding of concepts. Is this a serious defect in the Knowledge How Proposal? It is, if memorial justification is gained by seeming experiences. It is not, if Qualified Dogmatism does not extend to memory any more than it extends to hunches and gut intuitions. In that case, the Knowledge How Proposal offers us an account of why true Qualified Dogmatism has the limits it does (see Huemer 1999).

5. Conclusion Is there a form of Dogmatism to join Mentalism and Foundationalism as epistemic truths? If there is, it is a form of Qualified Dogmatism; cases like Prospectors and Mathematics Machine show that much. Whether Qualified 12 What if the mathematics machine gave Igor the same full understanding of the relevant mathematical concepts that is behind Elias’s seeming experience that (P) is true, so that (P) seemed true to Igor because of his full understanding of the relevant concepts? In that case, I believe, Igor too would have an epistemically appropriate seeming experience, one that represented his properly gained ability to identify (P) as true. His ability would be properly gained as it would stem from his sufficiently full understanding of the relevant concepts. The fact that he gained that understanding by being hooked up to the mathematics machine, rather than listening to tapes of mathematics lectures, would not make a difference.

268

Dealing with Cognitive Penetration

Dogmatism can be combined with both Mentalism and Foundationalism remains unclear, however. The Knowledge How Proposal is promising, but it needs to be developed and we need to see if it can cover all the various forms of seeming experience countenanced by a true form of Qualified Dogmatism.

Acknowledgments Matthew McGrath, Andrew Moon, and Chris Tucker provided generous comments on earlier drafts of this chapter, as did participants at the 2011 Midwest Epistemology Workshop.

References Aristotle. Nicomachean Ethics, translated by Richard McKeon. New  York:  Random House, 1941. Bealer, George. 2000. “A Theory of the A priori.” Pacific Philosophical Quarterly 81: 1–30. Bergmann, Michael. 2006. Justification without Awareness. Oxford: Oxford University Press. _____. Forthcoming. “Externalist Justification and the Role of Appearances.” Philosophical Studies. Conee, Earl. 1998. “Seeing the Truth.” Philosophy and Phenomenological Research 58(4): 847–857. Conee, Earl and Richard Feldman. 2004. Evidentialism. Oxford: Oxford University Press. Descartes, Rene. 1641. “The Meditations on First Philosophy.” In Selected Philosophical Writings of Descartes, translated by John Cottingham, Robert Stoothoff, and Dugald Murdoch,73–122. Cambridge: Cambridge University Press, 1985. _____. 1628. “Rules for the Direction of Our Native Intelligence.” In Selected Philosophical Writings of Descartes, 1–19. Huemer, Michael. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74(1): 30–55. _____. 2005. Ethical Intuitionism. Houndmills, UK: Palgrave Macmillan. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. _____. 1999. “The Problem of Memory Knowledge.” Pacific Philosophical Quarterly 80: 346–57. Leibniz, G. W. [1686] 1973. “Discourse on Metaphysics.” In Leibniz, translated by George Montgomery, 1–63. La Salle, IL: Open Court. Markie, Peter. 2013. “Rational Intuition and Understanding.” Philosophical Studies 163: 271–90. _____. 2006. “Epistemically Appropriate Perceptual Belief.” Noûs 40(1): 118–42. _____. 2005. “The Mystery of Direct Perceptual Justification.” Philosophical Studies 126(3): 347–73. _____. 2004. “Nondoxastic Perceptual Evidence.” Philosophy and Phenomenological Research 68(3): 530–53.

Searching for True Dogmatism

269

Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford University Press. Pryor, James. 2000. “The Skeptic and the Dogmatist.” Noûs 34(4): 517–49. Siegel, Susanna. 2012. “Cognitive Penetrability and Perceptual Justification.” Nous 46: 201–22. Sosa, Ernest and Laurence BonJour. 2003. Epistemic Justification: Internalism vs. Externalism. Oxford: Blackwell. Sosa, Ernest. 2007. A Virtue Epistemology. Oxford: Oxford University Press. Steup, Matthias. 1996. An Introduction to Contemporary Epistemology. Upper Saddle River, NJ: Prentice-Hall. Tucker, Chris. 2010. “Why Open Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24(1): 529–45.

{ 12 }

Phenomenal Seemings and Sensible Dogmatism Berit Brogaard 1. Phenomenal Conservatism In Skepticism and the Veil of Perception, Michael Huemer (2001) defends a position he calls “phenomenal conservatism.” He articulates the position as follows: If it seems to S as if p, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p. (Huemer 2007, 30)

William Tolhurst (1998), Jim Pryor (2000, 2005), and Chris Tucker (2010) have defended views akin to phenomenal conservatism. Here I focus on Huemer’s version. Phenomenal conservatism holds that seemings, or appearances, confer prima facie justification on beliefs. Though cognitive factors can affect seemings, the beliefs that are prima facie justified in this way are foundational, that is, they do not depend on any other beliefs for their justification (Huemer 2001, 98). If S possesses a defeater, the defeater prevents the prima facie justification from constituting ultima facie justification. Following John Pollock’s well-known distinction, an undercutting defeater is a reason for supposing that one’s ground for believing p is not sufficiently indicative of the truth of the belief that p (Pollock and Cruz 1986, 38). A rebutting defeater is a reason for holding the negation of p or for holding some proposition, q, incompatible with p. If it seems to me that the table is red, I have prima facie justification for believing that the table is red, according to phenomenal conservatism. But if I find out that I have an eye problem that may or may not cause me to misperceive colors, I have an undercutting defeater of my belief. The red appearance of the table no longer supports my belief that the table is red, so I no longer have justification for believing that the table is red. If it seems to me that you are sad, I have prima facie justification for believing that you are sad, according to phenomenal conservatism. But if you tell

Sensible Dogmatism

271

me that you are not sad, then I  have a rebutting defeater of my belief. The appearance that you are sad is no longer indicative of the truth of the belief that you are sad, so I no longer have justification for believing that you are sad. Likewise, if several individuals in a police lineup seem to an eyewitness as if they committed the crime, each appearance is then a rebutting defeater of the justification for her belief that any one of the individuals in the lineup is the criminal. So the eyewitness does not have justification for believing that any particular individual in the lineup is the criminal, though she may still have justification for believing that at least one of the identified individuals is the criminal. Defeaters should here be understood as reasons to which one has conscious access. Causal influences on beliefs do not count as defeaters.1 For example, suppose a judge makes harsher judgments if he hasn’t eaten than if he is given a sandwich. In this case the sandwich is not a defeater of the judge’s belief. Though Huemer does not provide an analysis of “It seems to S as if p,” he notes that its truth is compatible with S’s not believing that p and with S’s believing that p. In the Müller-Lyer optical illusion, for example, it seems to me that the lines have different lengths, even though I know that they have the same length. So it seems to me that p, although I do not believe that p. The kinds of seemings Huemer has in mind are perceptual seemings, memory-based seemings, introspective seemings, and intellectual seemings. It is hard to say exactly which seemings fall into which categories. But various examples give us a good sense of this distinction. Perceptual seemings include, for instance, the appearances that my cat is black, that the neighbors are having a party, and that you are sad. The appearances that you wore a red shirt last Monday, that Barack Obama won the last election, and that “ranarian” means frog-like are memory-based. The appearances that I  have a headache, that my visual image has no boundaries, and that my thoughts are in English are introspective. Finally, the appearances that 2 + 2 = 4, that the shortest distance between two points in a Euclidean plane is a straight line, and that bachelors are unmarried are intellectual. Huemer uses the principle of phenomenal conservatism to argue against skepticism and to defend a version of ethical intuitionism, the view that some moral beliefs are non-inferentially justified (Huemer 2001, 2005, 2007). On 1

Ambiguous figures may be thought to provide unwanted defeaters. Consider the Necker cube. It seems that a certain corner is closest to me but when my focus changes, it seems that a different corner is closest to me. So it may be thought that I have a defeater of my belief about which corner is closest. However, I don’t think that what things look like in different circumstances can defeat belief. If these kinds of seemings could defeat belief, then my seeming that the grass is green could not confer justification on my belief that it is green, because when it gets dark, the grass does not seem green. Thanks to Matt Weiner here.

272

Dealing with Cognitive Penetration

Huemer’s view, perceptual beliefs are foundationally justified on the basis of the principle of phenomenal conservatism. As this principle does not offer conclusive justification but merely prima facie justification, defeaters can defeat the positive epistemic status of these beliefs. However, Huemer argues, if the notion of prima facie justification is to serve any purpose, skeptical hypotheses cannot defeat the prima facie justification of our perceptual beliefs. So at least some of our perceptual beliefs are foundationally justified to some degree. Huemer’s argument for a version of ethical intuitionism proceeds along similar lines. Intellectual seemings confer prima facie justification on some of our moral beliefs. Though defeaters can defeat the prima facie justification of these beliefs, skeptical hypotheses cannot. So some moral beliefs have some degree of foundational justification. Phenomenal conservatism has received its fair share of criticism. Among other things, it has been argued that too many beliefs that intuitively are not justified come out as having some degree of justification, given phenomenal conservatism. I believe many of these criticisms are justified. Here I want to lay out and defend a close cousin of phenomenal conservatism. I will call it “sensible dogmatism.” Sensible dogmatism is the position that a specific subset of seemings, which I  will call non-epistemic seemings, can confer justification on belief when they are grounded in the content of memory, introspection, or perception. I argue that this position avoids the most devastating problems for phenomenal conservatism while inheriting its virtues. I  begin by considering what I believe are the most damaging problems for phenomenal conservatism. I then offer an analysis of seemings and show that only a subclass of seemings can account for the immediate justification of beliefs. Finally, I set forth the position of sensible dogmatism and outline what I believe are the main virtues of the position compared to alternatives.

2. Challenges for Phenomenal Conservatism: The Experience-Appearance Connection On Huemer’s view, the class of seemings that can confer prima facie justification on beliefs is very broad. It includes perceptual, memory-based, introspective, and intellectual seemings. Huemer makes this assumption because a non-ad hoc version of foundationalism ought to take all foundationally justified beliefs to have the same kind of foundation (Huemer 2001, 102). The problem with this line of argument is that it is doubtful that all seemings are equally epistemically significant. Peter Markie provides two illustrations of this point. The first runs as follows.

Sensible Dogmatism

273

Suppose that we are prospecting for gold. You have learned to identify a gold nugget on sight but I have no such knowledge. As the water washes out of my pan, we both look at a pebble, which is in fact a gold nugget. My desire to discover gold makes it seem to me as if the pebble is gold; your learned identification skills make it seem that way to you. According to (PC), the belief that it is gold has prima facie justification for both of us. Yet, certainly, my wishful thinking should not gain my perceptual belief the same positive epistemic status of defeasible justification as your learned identification skills. (Markie 2005, 356–57)

The problem here is that because it seems to both the expert and the novice that the pebble is gold, Huemer’s view entails that they both have prima facie justification for believing that it is gold. Since the novice possessed no defeater that he has access to, and the novice’s belief is based on his seeming, the expert and the novice are equally justified in believing that the pebble is gold. This is unintuitive, as the novice’s belief is based on wishful thinking. Intuitively, the novice does not have any justification for believing that the pebble is gold.2 So phenomenal conservatism makes the wrong prediction in this case.3 Markie’s second objection runs as follows. Suppose that I perceive the walnut tree in my yard, and, having learned to identify walnut trees visually, it seems to me that it is a walnut tree. The same phenomenological experience that makes it seem to me that the tree is a walnut also makes it seem to me that it was planted on April 24, 1914. Nothing in the phenomenological experience or my identification skills supports things seeming this way to me. There is no date-of-planting sign on the tree, for example. (Markie 2005, 357)

Markie here allows for the possibility that my experience via my seeming can confer prima facie justification on my belief that the tree is a walnut. It could do so because it is plausible that my experience represents the tree as a walnut. But it is not plausible that my experience represents the tree as having been planted April 24, 1914. The property of being planted April 24, 1914, is not a visual property. We cannot detect it through vision alone. But if it really does seem to me that the tree is a walnut and was planted April 24, 1914, then Huemer’s view entails that my belief that the tree is a walnut and my belief that it was planted April 24, 1914, have the same degree of prima facie justification. Supposing I don’t have access to any defeaters, I have justification for my belief

2 If an agent has reasons for p, then p is propositionally justified. If an agent believes p on the basis of her reasons for p, then her belief is doxastically justified. What matters in relation to phenomenal conservatism is that the proposition that the nugget is gold is not propositionally justified. 3 I am somewhat skeptical about this case. I am tempted to think most of us would be sufficiently acquainted with gold to be prima facie justified in thinking that a gold nugget in a pan is gold. But we can change it to something harder for most people to spot, such as valuable antique furniture.

274

Dealing with Cognitive Penetration

that the tree was planted April 24, 1914. But if the only ground for my belief that the tree was planted April 24, 1914, is an appearance that this is so, then clearly I do not have any justification for my belief. The problem with Huemer’s view is that his class of seemings don’t all have a connection to justification. A  proper version of phenomenal conservatism should show how seemings relate to justification and restrict the class of seemings if only some of them are epistemically significant. I propose to do that in the two subsequent sections.

3. Phenomenal and Epistemic Seemings According to Roderick Chisholm (1957), there are three uses of the word “seem”:  epistemic, comparative, and phenomenal. Chisholm doesn’t provide a full analysis of the three uses but illustrates what he has in mind by way of example. Suppose I hear on the radio that a tsunami is going to cause flooding in my area and say to my housemates “It seems to be in our best interest if we evacuate.” In this case, Chisholm would say that “seem” is epistemic, because my claim is not grounded in the phenomenology of my perception. Suppose I say “It seems as if the roads are wet” on the basis of how the roads visually seem to me. Chisholm would say that “seem” here is phenomenal because my claim is grounded in the phenomenology of my perceptual experience of the roads. Comparative “seem” constructions have contents with a comparative structure. For example, going through the same routine as yesterday, I might say “This seems like yesterday.” Chisholm calls this occurrence of “seem” comparative. The epistemic-non-epistemic distinction is orthogonal to the comparative-non-comparative distinction. For example, I might say about War 1 and War 2 on the basis of what I read in a history book that “War 1 looks like War 2.” This claim then is both comparative and epistemic. Though Chisholm is talking about uses of “seem” and other appearance words, I shall here assume that constructions in which “seem” occurs make reference to mental states, which I will call seeming states. Chisholm’s distinction then corresponds to three different kinds of seeming states. How do we distinguish them? I propose that a definitive mark of epistemic seemings is that they go away in the presence of a rebutting defeater if the agent is rational.4 It may seem like a good idea to evacuate because the radio host announced that there will be flooding in 4 If it seems to me that a bimonthly event should occur once a month, and you then explain to me that bimonthly events, by definition, occur twice in a month, then it no longer ought to seem to me that a bimonthly event should occur once a month. So the seeming is epistemic. One could argue that this extends to intellectual seemings. For example, if you convince me that “plus” really means something completely different, one may argue that it no longer will seem to me that 2 + 2 = 4. However, I think there really is a difference between the two cases. I think in the latter case it would continue to seem to me that 2 + 2 = 4 even if I changed my belief about it. One could also argue that there are unwanted

Sensible Dogmatism

275

my area, but if he comes back on the radio and announces that the earlier warning was a hoax, then it will no longer seem like a good idea to evacuate. Likewise, if it seems to me that John is in his office on the basis of seeing his hat hanging in the hall, but you tell me that John took off to Rome this morning, then it will no longer seem to me that John is in his office.5 Non-epistemic seemings persist, at least to some degree, in the presence of a defeater. If the roads look wet, then they will continue to look wet even if you tell me that the city painted them as a part of their “drive safe” campaign. Comparative seemings can be analyzed in terms of non-comparative seemings.6 Consider the following examples of comparative constructions: 1.(a) 1.(b) 1.(c) 1.(d) 1.(e) 1.(f)

John is taller than every girl John is taller than every girl is Ellen is as rich as her father George is richer than his father was and his son will be John dances like Tom Mary eats like a bird

Richard Larson (1988) has argued that 1(a) is best understood along the following lines: (i) the quantified noun phrase (e.g., “every girl” or “one of the girls”) moves to a wide-scope position and (ii) the comparative expression “taller than” combines with two type e expressions (i.e., variables or referring terms). On this view, 1(a) has the underlying form “[Every girl x]taller-than(John, x).” The problem with Larson’s analysis is that it does not extend to the other sentences in (1). 1(b) contains the clause “than every girl is” and “every girl” cannot scope out of that clause. Heim (2006) offers a more plausible account of comparative constructions. On this view, comparative constructions ascribe relations between what she calls “degrees” (i.e., abstract entities like heights, weights, etc.). To account for quantifier scopes, Heim posits semantically vacuous “wh”-items in the sentence structure. 1(b) is to be read as: “John is taller than every girl is wh.” To a first approximation, “every girl is wh” is to be read as: “every girl x: x is this tall.” This item scopes out of the comparative clause, and the “wh”-item raises to a wide-scope position, yielding

defeaters of many perceptual seemings. For example, people who are self-confident are more reliable in recognizing their own voice on tape than people whose self-confidence has been shattered. So a seeming that a taped voice is my own would go away if my self-confidence were shattered. However, shattering my self-confidence is not a defeater. So the seeming is not epistemic. 5 A question arises concerning epistemic seemings of the kind people have with respect to Linda the bank teller case and other similar cases. These seemings do not always go away in the presence of a defeater. I am tempted to say that in these cases people do not exhibit rationality with respect to issues of probability and prediction. 6 Constructions such as “John seems sadder than Mary” are comparative but the verb “seem” is not used comparatively here; “sad” is. Thanks to Chris Tucker here.

276

Dealing with Cognitive Penetration

[wh1[every girl is t1]]2 [John is taller than t2]

Likewise, 1(e) can be read as: “John dances like Tom does wh,” where “Tom does wh” is to be read, roughly, as: “Tom dances this way.” This item scopes out of the comparative clause, yielding “[wh1[Tom dances t1]]2 [John dances t2].” We can assign the following truth-condition to 1(f): for some way w such that w is a way that Tom dances, John dances that way too. As comparative seemings are superficially similar to the comparatives in 1(e) and 1(f), it is very plausible that they have the same underlying structure (Brogaard 2011b). On this hypothesis, “X seems like Y” contains the implicit wh-clause “wh1[Y seems t1].” This item scopes out of the comparative clause, yielding “[wh1[Y seems t1]]2 [X seems t2].” For example, “This seems like yesterday” is to be read as containing the implicit clause wh-clause: “wh1[yesterday seems to be t1].” This item scopes out of the comparative clause, yielding “[wh1[yesterday seems to be t1]]2[today seems to be t2].” The comparative analysis of comparative seemings offers the full answer to the question of how to analyze comparative seemings but it does not address the question of how to assign truth-conditions. The reason is that the analysis makes unreduced appeal to a notion of “seems to be x.” This notion needs further analysis in terms of non-comparative seemings. Chisholm took all phenomenal seemings to be perceptual. I  agree with Huemer, however, that some memory-related, introspective, and intellectual seemings are phenomenal seemings. Phenomenal seemings belong to a class of mental states that is distinct from that of perceptual experiences, memories, and beliefs. Phenomenal seemings are different from belief, as one can believe that p even if it seems to one that not-p. For example, I can believe that the lines in the Müller-Lyer optical illusion have the same length, even though it seems as if they have different lengths. At least some seeming states are also distinct from the experiences that precede them (Brogaard 2011b, Tucker 2010). If I am giving a talk to fifty-four people, my perceptual experience, if veridical, represents fifty-four people in the room but it doesn’t phenomenally seem to me that there are fifty-four people in the room. At best, it seems to me that there are many people or more than twenty people or fewer than one hundred people in the room.7

7 In correspondence Huemer observes that “my experience represents 54 people” probably is best understood as “there are 54 people such that my experience represents each of those people.” But, he argues, then it would also be true that “there are 54 people such that each of those people seems to me to be in the room.” So there would be no difference between what your experience represents and what seems to you to be the case. This certainly is correct. However, as I am currently constituted it could not seem to me on the basis of my perceptual experience that there are fifty-four people in the room. This suggests that perceptual seeming states can be interpretations of perceptual experiences and hence perceptual seemings cannot simply be perceptual experiences.

Sensible Dogmatism

277

Memory-based seemings are distinct from memories for the same reason. If I have an accurate visual image of the room during yesterday’s talk, then my visual image represents fifty-four people in a room. But it doesn’t phenomenally seem to me that there were fifty-four people in the room. Likewise, introspective seemings are distinct from at least some introspective experiences. If you ask me how many people attended the lecture yesterday, I may rely on introspection to determine the answer. In this case, to introspect just is to call up a visual image of the room during yesterday’s talk. In this case, the visual image can represent fifty-four people in a room, even if it doesn’t introspectively seem to me that there were fifty-four people in the room. So states of seeming are distinct from beliefs, perceptual experiences, introspective experiences, and memories. Though seemings are distinct from experiences, they are not completely unrelated. Etymologically, “seem” comes from the old English word “beseon,” which means “be seen as.” Assuming that the meaning of “beseon” is also the meaning of “seem,” saying that I see a as an F is equivalent to saying that a seems like an F. Visual seemings then are interpretations of visual experiences. Presumably, this holds in general. All seemings then are interpretations of experiences. Now, only non-epistemic seemings can account for immediate justification. This is because epistemic seemings, unlike other seemings, are belief states. More precisely, they are beliefs about probability. When the radio host announces that there will be flooding in my area, I form a belief that there will be flooding in my area. From this I may reason that it probably is in my best interest to evacuate the building. We can use “seem” to express this probability belief: “It seems to be in my best interest to evacuate the building.” “Seems” here functions as a modifier much like “probably.” If it epistemically seems to me to be in my best interest to evacuate the building, then it is subjectively probable that it is in my best interest to evacuate the building. Epistemic seemings thus are inferred from beliefs and hence do not confer immediate justification on beliefs.

4. Sensible Dogmatism To distinguish between seemings based on experiences and seemings that are not, let’s introduce a notion of grounding as follows: Content Grounding A seeming of the form [It seems to A as if q] is grounded in a content p of a particular perceptual, introspective, or memory-related experience e had by A iff [Reliably(if p is a content of e, then it seems to A as if q) and Reliably(if it seems to A as if q, then q)].

278

Dealing with Cognitive Penetration

“Reliably” here should be understood as a modal operator similar to the subjunctive conditional. The first conjunct of the right-hand side can be read as:  In the majority of close hypothetical situations in which p is a content of e, it seems to A as if q.8 The second conjunct rules out that seemings can be grounded in experiences as a matter of sheer luck.9 For example, suppose everything seems like a cow to A. Because everything seems like a cow to him, his cow experiences will be accompanied by a cow appearance. But his appearance is not grounded in his experience. The second conjunct does not require that the content q actually be true but only that it is true in most close situations in which it seems to A as if q.10 Given this principle of grounding I  can now state the position I  wish to defend: Sensible Dogmatism If it seems to S as if p and the seeming is grounded in the content of S’s perceptual, introspective, or memory-related experience, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p,  and A seeming to S that p provides prima facie justification only if it is grounded in the content of S’s perceptual, introspective, or memoryrelated experience.

Huemer’s phenomenal conservatism is a version of austere internalism. Whether you have prima facie justification for your beliefs is introspectively accessible to you. Though sensible dogmatism states that an internal and introspectively accessible mental state, viz., a seeming, can confer justification on belief, the proposed view is not a version of austere internalism. The view is a form of weak internalism in the sense that the mental states that confer prima

8 I am presupposing the similarity metric set forth by David Lewis (1979). The measure of similarity is context-sensitive but the following default algorithm applies. 1. Avoid big, widespread, diverse violations of law. 2. Maximize the spatio-temporal region throughout which perfect match of particular fact prevails. 3. Avoid even small, localized, simple violations of law. Consider the Harman case. A physicist sees a certain track in a bubble chamber. Owing to his training, it seems to him that it’s a proton (Harman 1977, 4, 8). The seeming appears to be grounded in his experience, and this is the result we get. The worlds closest to the actual in which the physicist sees the track are worlds in which he has the training he actually has. Thanks to Terence Cuneo here. 9 This principle has the consequence that some lucky seemings that may appear to be grounded are not grounded. Suppose an evil scientist implants a device in Sally’s brain that severs the connection between experiences and seemings. There is a one in a billion chance that the device fails. But one day it happens. The device fails, and Sally has a normal experience and a normal seeming. In this case, her seeming is not grounded in her experience. More on that later. Thanks to Matt Weiner here. 10 Whether this principle entails that seemings cannot be grounded in holistic skeptical scenarios depends on whether our concepts succeed in referring to anything in these kinds of scenarios. They might succeed in referring to whatever causes our experiences (see Chalmers 2005), in which case seemings can be grounded in these scenarios.

Sensible Dogmatism

279

facie justification on belief are states to which you have introspective access but the factors that determine whether these states confer justification is not something to which you have introspective access. One concern about sensible dogmatism immediately arises. At first glance, sensible dogmatism, as stated, does not apply to intellectual seemings. Upon further scrutiny, however, it is apparent that sensible dogmatism does cover intellectual seemings. I believe that there are two kinds of intellectual seemings:  intuitions about how people in my community apply concepts and a priori intuitions. Both kinds of intellectual seemings are ideally grounded in semantic memory.11 If I judge that a subject in a Gettier case lacks knowledge, I make this judgment on the basis of semantic memory of how people in my community apply the notion of knowledge. Likewise, if I judge that p and q entails p, I make the judgment on the basis of semantic memory of the meaning of the logical connectors. So intellectual seemings are a special case of memory-related seemings. A second concern about sensible dogmatism is that some mental states that can give rise to immediately justified beliefs don’t appear to have propositional content. For example, some may hold that my state of having a headache does not have propositional content. But if I have a headache, it seems to me that I have a headache, and this seeming then confers prima facie justification on my belief that I have a headache. In this case, however, I do think that I have a perceptual experience with propositional content. It may be that my state of having a headache doesn’t have propositional content while my experience of myself being in pain does have propositional content. My seeming that I have a headache may be grounded in the content of this experience and hence can confer justification on my belief that I have a headache. A third concern is that sensible dogmatism, as stated, says that seemings must be grounded in experience for them to provide prima facie justification. However, it may be objected that grounding isn’t necessary for a seeming to provide prima facie justification. Consider super blindsighters who get extremely reliable seemings about their environment without getting any visual imagery (or auditory sensations). Their seemings wouldn’t be grounded in any perceptual experience, but it seems plausible that their seemings could provide prima facie justification for their contents.12

11 Semantic memory should be understood as the storage and retrieval of fact-like information and information about concepts and language. A semantic memory experience would be related to retrieval of information. This raises the question of whether we really do have conscious access to the retrieval of information from semantic memory, a question I cannot address here. It also raises the question of how to distinguish between moral intuitions about analytic sentences and moral intuitions about a priori sentences. One could perhaps say that some of these seemings are grounded not just in semantic memory but also in deductive abilities. Thanks to Michael Huemer here. 12 Thanks to Chris Tucker and Michael Tooley for raising these objections.

280

Dealing with Cognitive Penetration

However, as I have argued in previous work (Brogaard 2011b, forthcoming), the seemings in these cases are epistemic. When a super-blindsighter detects the color of a visual stimulus presented to her in her blind field, she has no distinctly visual awareness of the color of the stimulus. So when she reports on the color of a stimulus presented to her in her blind field, she cannot make use of any visual phenomenology associated with the color information. Rather, she must infer from her inclination to guess that the stimulus is red that it is red. Were she to be presented with a defeater, she would—if rational—no longer have the inclination to state that the stimulus looks red. So when she says that the stimulus looks red, her report is evidence-bearing and hence epistemic. It may also be countered that our attention can affect the way things seem. My perceptual experience represents a bird in my peripheral vision. But it doesn’t seem that way until I turn my attention to it. To the extent that attention can affect which seemings we form on the basis of our perceptual experience, our seemings do not appear to be grounded in our perceptual experiences. Yet the fact that these seemings depend on attention should not prevent them from having justificatory power. However, selective attention is a top-down factor that modulates representation (Brogaard 2011a). Attention can affect the representational phenomenal character of a representation. For example, the clock now in the periphery of my visual field is presented to me in a way that leaves out information. I am aware of its round shape and some pattern in the middle. When I attend to it, I am also aware of the numbers and their colors. So, attention can change the representational phenomenology of my experience. When I turn my head, my visual seeming state changes, but so does my perceptual experience. I now argue that sensible dogmatism avoids the problems that plague phenomenal conservatism while inheriting its virtues.

5. Sensible Dogmatism versus Phenomenal Conservatism Sensible dogmatism fares better than phenomenal conservatism in a number of respects. Consider again the gold mining case. Phenomenal conservatism entails that the expert gold miner and the novice both have justification for their beliefs that the pebble in the pan is gold, even though the reason it seemed to the novice that the pebble was gold was that he had an intense desire to find gold. Sensible dogmatism does not have this consequence. As the water washes out of the novice’s pan, his desire to discover gold makes it seem to him as if the pebble is gold. If he had the same perceptual experience in a different context in which he did not desire to discover gold, it wouldn’t seem to him that the pebble was gold. So the grounding principle isn’t satisfied by the novice in this case. The novice’s state of seeming isn’t grounded in the content

Sensible Dogmatism

281

of his perceptual experience and therefore cannot confer justification on his belief that the pebble in the pan is gold. Notice that if the novice normally or always has the relevant desire, then the content of his seemings will be false in scenarios in which the pebble isn’t gold, so the second conjunct on the right-hand side of the grounding principle is not satisfied. We get a similar result in the case of Markie’s walnut tree. Phenomenal conservatism entails that both the belief that the tree is a walnut and the belief that the tree was planted April 24, 1914, are justified on the basis of the seeming. But as the seeming that the tree was planted April 24, 1914, isn’t grounded in the subject’s perceptual experience, his belief that the tree is planted April 24, 1914, is not justified. Sensible dogmatism gives us the right result. In a majority of relevantly close worlds, it will seem to the subject that the tree was planted April 24, 1914, even though it wasn’t actually planted then. So his state of seeming is not grounded in the content of his perceptual experience and hence cannot confer prima facie justification on his belief. Huemer takes it to be a virtue of phenomenal conservatism that it blocks the argument for skepticism and justifies ethical intuitionism. This is not the place to evaluate whether these arguments are sound. Suffice it to say that sensible dogmatism can be used in place of phenomenal conservatism to make equally strong arguments. The argument against skepticism would run as follows. Given sensible dogmatism, some of our perceptual beliefs are prima facie justified on non-inferential grounds. Following Huemer, if the notion of prima facie justification is to serve any meaningful purpose, then skeptical hypotheses cannot serve as defeaters of the prima facie justification of belief. So some of our perceptual beliefs are justified to some degree, contrary to what the skeptic says. How does sensible dogmatism account for moral intuitions? On a popular view of affective states, affective states are perceptual states. For example, anger-at-John might be a perceptual experience of John’s behavior causing certain anger-related bodily changes (Brogaard forthcoming). If moral intuitions are grounded in affective states, then we can explain why we have prima facie justification for certain moral beliefs. We have prima facie justification for the belief that torturing babies is wrong because we have a moral seeming that torturing babies is wrong. The moral seeming, we can imagine, is grounded in the content of the perceptual experience that baby torture causes certain anger and disgust-related bodily changes.13

13 Some moral seemings no doubt are epistemic, for example, the seeming that all wars are morally justified would presumably go away in the presence of a defeater, assuming the agent is rational. The moral seeming that pain is bad, on the other hand, would not go away in the presence of a defeater, given a rational agent.

282

Dealing with Cognitive Penetration

If we take moral intuitions to be grounded in affective states, then many cases in which people have moral intuitions that are very different from most people’s intuitions will have justification for their moral beliefs. For example, if it seems to a serial killer that killing is morally acceptable, then he may have some justification for his belief that killing is morally acceptable.14 This is an unwelcome consequence of taking moral intuitions to be grounded in affective states.15 There is, however, another way to understand moral intuitions. It is plausible that moral intuitions to some extent are grounded in semantic memory of the application conditions for terms such as good and bad. “Bad” may be semantically associated with “harm” and “pain” and “good” may be semantically associated with “benefits” and “pleasure.” If moral intuitions are grounded only in semantic memory, then we can avoid some of the aforementioned unwelcome consequences. Serial killers arguably know how to apply the terms good and bad but it nonetheless seems to them that harming people is morally acceptable to them. The serial killers’ moral intuitions then are not properly grounded in semantic memory. So, their appearances do not confer justification on their beliefs. Another virtue of sensible dogmatism, which it shares with phenomenal conservatism, is that it provides a solution to the old problem of the speckled hen. The problem is that there are aspects of the contents of mental states that cannot serve as grounds for knowledge. For example, if you are looking at a speckled hen, then your visual experience represents a determinate number of speckles but you cannot know which number simply in virtue of having that experience (Ayer 1940; Chisholm 1942; Tucker 2010). But many other aspects of the contents of your mental states can serve as grounds for knowledge. The problem is to specify a principled difference between those aspects that can serve as a ground for knowledge and those that cannot. Sensible dogmatism suggests a way of explaining the difference between those aspects that can serve as grounds for knowledge and those that cannot.

14

This holds only if he lacks defeaters and if his belief is based on his justification. Here is a further problem for this view. As Huemer has pointed out to me, there appear to be ethical intuitions that don’t have any associated emotional or perceptual reactions and it also appears that these are especially reliable intuitions. In his (2008) he provides examples, such as: 15

(a) If x is better than y and y is better than z, then x is better than z. (b) The ethical status of choosing (x and y) over (x and z) is the same as that of choosing y over z, given the knowledge that x exists/occurs (c) If two states of affairs, x and y, are related such that y could be produced by adding something valuable to x, without creating anything bad, lowering the value of anything in x, or removing anything of value from x, then y is better than x. These “formal ethical intuitions” do not by themselves tell you what you should do or even whether a particular thing is good or bad. But as he argues, they can be used to reject certain otherwise tempting and interesting substantive ethical views. So they are important for that reason.

Sensible Dogmatism

283

The problem of the speckled hen can be stated more generally as the problem of explaining why we can be immediately justified in believing truths about the cardinal number of small but not large quantities on the basis of perception. For example, I can be immediately justified in believing that there are three birds on the roof but not in believing that there are seventy-three. As Stanislas Dehaene explains in his book The Number Sense, normal human beings are unable to make immediate judgments about the cardinal number of large quantities, because it was evolutionarily more beneficial to us to make estimations concerning quantities in the environment. If there are seventythree birds on the roof, we can immediately tell than there are fewer than five hundred birds and that there are more than ten birds but we cannot immediately tell that there are seventy-three birds. This is in some sense strange, because we can have perceptual experiences that represent seventy-three birds. However, the reason we cannot tell that there are seventy-three birds, even when our experience represents seventy-three birds, is that we cannot translate the image of seventy-three birds into the number seventy-three. So, it doesn’t seem to us that there are seventy-three birds on the roof. Because it doesn’t seem that way, we cannot be immediately justified in believing that there are seventy-three birds on the roof. So that aspect of the content of our visual experience cannot serve as a ground for knowledge. A further virtue of sensible dogmatism is that it can account for the appearance that an expert and a novice both holding a certain belief do not have the same degree of justification. For example, if an expert and a novice both see an elm tree and both come to believe that it is an elm tree, they do not have the same degree of justification for their beliefs. One way to account for this difference in justification is to say that the perceptual experiences of the expert and the novice have different contents. The expert’s perceptual experience represents an elm tree, whereas the novice’s perceptual experience merely represents a tree. Susanna Siegel (2005, 2011) takes this approach. Her argument runs as follows. Let E1 be a visual experience of someone who has the ability to recognize elm trees (expert) and who is looking at an elm tree, and let E2 be the visual experience of someone who does not have the ability to recognize elm trees (novice) and who is looking at the same tree in the same viewing conditions. The expert finds the tree familiar, the novice does not. So there is a difference in the overall phenomenal character of their experiences. The argument can now articulated as follows: The Argument from Phenomenal Contrast

(1) The overall phenomenology of which the phenomenology of E1 is a part differs from the overall phenomenology of which the phenomenology of E2 is a part (familiarity effects).

284

Dealing with Cognitive Penetration

(2) If the overall phenomenology of which the phenomenology of E1 is a part differs from the overall phenomenology of which the phenomenology of E2 is a part, then there is a phenomenological difference between E1 and E2 (cognitive penetration). (3) If there is a phenomenological difference between E1 and E2, then E1 and E2 differ in content (representationalism). Therefore, (C) If there is a difference in content between E1 and E2, it is a difference with respect to K-properties represented in E1 and E2.

If the perceptual experiences of the expert and the novice have different contents but both hold the belief that there is an elm tree in front of them, then only the expert’s belief is grounded in perceptual experience. The novice’s belief is an instance of mere guessing. So, this explains why we think that the expert’s belief is on firmer grounds than that of the novice. To many people, including me, however, it seems odd that the perceptual experiences of the expert and the novice would represent the tree differently in the two cases (Brogaard 2013). Sensible dogmatism allows us to offer a different explanation. It seems to the expert that the tree is an elm but it doesn’t seem that way to the novice. The overall difference in phenomenology between the expert and the novice stems from a difference in the phenomenology of the two distinct states of seeming. To the expert, the tree seems to be an elm, it seems familiar, and it seems different from the nearly identical neighboring tree which is not an elm. To the novice, the tree just seems to be a tree, it doesn’t seem familiar, and it seems exactly similar to the neighboring tree. So the overall phenomenology of the expert and novice’s experiences are going to be different, even though the content of their perceptual experience is exactly the same. The expert’s seeming can confer justification on his belief that the tree is an elm, because his seeming is grounded in the content of his experience. There is no possibility in which he (as he currently functions cognitively) will have a perceptual experience with the same content but in which it doesn’t seem to him that the tree is an elm. A problem not unrelated to that of high-level properties in perception is the puzzle of occluded objects. It makes sense to say that we cannot see the backside of objects. If, however, objects, such as tomatoes, apples, and oranges literally can be part of the content of our experiences, then the objects as a whole are parts of the contents of our experiences. A fortiori, the backsides of objects are parts of the contents of our experiences. So we do see the backside of objects after all. Personally, I think it is plainly obvious that the backsides of objects are not part of the contents of our visual experiences. At best, the facades of these objects are.

Sensible Dogmatism

285

But why then do we feel at ease talking about seeing people, tomatoes, cars, and so on? I think it is because on the level of seeming, we fill in the missing parts. This then suggests interestingly that the verb “to see” is ambiguous between meaning “to have a verdical visual experience that represents” and “to have a verdical seeming state that represents.” “I don’t see the backside of the tomato” is best interpreted as “I don’t have a verdical visual experience that represents the backside of the tomato.” “I saw a man,” on the other hand, is best interpreted as “I had a visual verdical seeming state that represented a man.” Our visual experiences represent only facades of objects, yet we can have justification and knowledge about objects. What makes this kind of justification and knowledge possible is not our visual experiences but our visual seemings. It is because we have visual seemings of objects as opposed to facades that we have justification for our beliefs that we see objects and not facades. The distinction between visual experience and seemings can also explain the possibility of seeing absences. If I walk into a room expecting you to be there but you are not, I may see that you are not there. But in this case I don’t have a visual experience that represents you as absent. I just have a visual experience of the room and the people in it. This visual experience together with my expectations generates the visual seeming that you are not there. It may be objected that the seeming here is epistemic. Suppose you look in my office and try to find me. It seems to you that I am not there. The person in the office next to mine tells you that I am really in there, you just need to look harder, that I am hiding really well. Assuming you believe the testimony, you will not retain the seeming that I am not in there (especially, if you eventually find me scrunched in one of the cabinets).16 However, in this case, the reason you don’t retain the seeming is that you start to look around. Once you find me scrunched in one of the cabinets, your visual experience changes. When you first enter the office after being told that I am really in there, you may well believe that I am really in there. But it still perceptually seems to you that I am not there.

7. Further Virtues: Elevation Prediction In a recent paper Susanna Siegel (2012) offers two counterexamples to phenomenal conservatism (or close variants). She assumes that beliefs, attitudes, expectations, and so forth, can penetrate and change the content of visual experience. For example, if you expect that someone is sad, you may, partially in virtue of that expectation, have a visual experience that represents that person as sad. This is the sort of assumption I  challenged earlier. But let it be

16

Thanks to Chris Tucker for providing this example.

286

Dealing with Cognitive Penetration

granted for argument’s sake that the assumption is true. Siegel’s first counterexample proceeds as follows. Case 1: Angry-looking Jack: Jill believes, without justification, that Jack is angry at her. The epistemically appropriate attitude for Jill to take toward the proposition that Jack is angry at her is suspension of belief. But when she sees Jack, her belief makes him look angry to her. If she didn’t believe this, her experience wouldn’t represent him as angry.

On Siegel’s view, because Jill believes that Jack is angry at her, Jill has a visual experience that represents Jack as angry. Had she not had the belief that he was angry at her, she wouldn’t have had a visual experience of him being angry at her. According to phenomenal conservatism, if it seems to Jill that Jack is angry at her, then absent defeaters, she has justification for believing that he is angry at her. The problem this case poses for phenomenal conservatism is this. While Jill’s initial belief had no justification, once her belief is firmly grounded in her visual experience, she suddenly does have justification for her belief. This is strange, as it somehow was her unjustified belief that generated the justification for that very belief. The second case is parallel to the first: Case 2: Preformationism: Many of the first users of microscopes favored preformationism about mammalian reproduction. Some of them claimed to see embryos in sperm cells that they examined using a microscope.

Prior to looking at sperm cells under the microscope, the preformationist favored the hypothesis that there were embryos in healthy sperm cells. At the time, no theory of mammalian reproduction was well confirmed, and the epistemically appropriate attitude to take toward preformationism was suspension of belief. But the preformationists did not suspend belief. When they looked under the microscope, they saw embryos in the sperm cells. The preformationists’ visual experiences represented sperm cells as containing embryos. What caused them to see sperm cells in this light, however, was their prior unjustified belief that sperm cells contain embryos. According to phenomenal conservatism, appearances can confer justification on belief in the absence of defeaters. As the preformationists did not have any defeaters of their fundamental beliefs, their visual experience of sperm cells containing embryos conferred justification on their initial beliefs. So, it seems that an unjustified belief can somehow generate justification for itself through the process of cognitive penetration. This just seems plainly wrong. Here it will not help the phenomenal conservative to reject the hypothesis that beliefs can penetrate visual experience. For the counterexamples go through with the weaker assumption that belief can penetrate appearances, an assumption that

Sensible Dogmatism

287

I have embraced throughout the paper. For example, even if Jill’s belief that Jack is angry at her does not cause her to have a visual experience with the content that Jack is angry, it does cause her to have a visual seeming of Jack being angry. But, assuming phenomenal conservatism, this visual seeming can then justify her initial, unjustified belief that Jack is angry at her, which is odd. While Siegel’s two cases present a serious threat to phenomenal conservatism, they do not threaten to defeat sensible dogmatism. This is because the appearances, or seemings, she invokes in her counterexamples are epistemic. Recall that an epistemic seeming is one that disappears in the presence of a defeater in a rational agent, whereas a non-epistemic seeming is one that lingers, at least to some extent, even in the presence of a defeater. Because epistemic seemings disappear in the presence of a defeater when the agent is rational, they are not grounded in the content of perceptual, introspective, or memory-related experiences. Suppose again that I hear on the radio that a tsunami is going to cause flooding in my area and I say to my housemates “It seems like a good idea to evacuate.” If the radio host later announces that the earlier announcement was a hoax, “a tsunami is going to cause flooding in my area” is still the content of what I heard but it no longer seems to me to be in my best interest to evacuate. So my seeming that it’s probably in my best interest to evacuate is not grounded in the content “a tsunami is going to cause flooding in my area.” Of course, it might seem to me to be in my best interest to evacuate on the basis of a coarse-grained memory to the effect that I probably should evacuate. But if all I remember and continue to remember is that I probably should evacuate, then the seeming doesn’t go away as the result of a radio announcement. So in this case the seeming is not epistemic.17 As phenomenal conservatism does not require that seemings that provide prima facie justification are grounded in the content of experiences, it does not have the means to rule out epistemic seemings as a source of prima facie justification. Let us return now to Siegel’s example. If Jack convinces Jill (or gives her some reason to believe) that he is not angry at her (for example, over the phone) before she sees him, she will—if she is rational—no longer believe that Jack is angry at her. So if she is rational, it will no longer seem to her as if he

17 Huemer has suggested to me that it seems that we could devise a hypothetical situation in which epistemic seemings would be grounded in experience. We’d need a case in which

(i) It’s reliable that when you have experience e, then it seems to you that p, (ii) It’s also reliable that when it seems to you that p, p is true, and (iii) if you get evidence against p, the seeming would go away if you were rational. Conditions (i), (ii), and (iii) look mutually consistent, so it looks as though there could be an epistemic seeming that is grounded in experience. If there is a case like this, we might want to say that the epistemic seeming can confer prima facie justification on belief but, as epistemic seemings are belief states, the beliefs on which it would confer justification would not be foundational.

288

Dealing with Cognitive Penetration

is angry at her when she sees him. As the appearance of his being angry at her does not linger in the presence of a defeater if she is rational, the seeming is epistemic and so it does not confer prima facie justification on her belief. The same applies in the same case of preformationism. If someone convinces the preformationist that sperm cells do not contain embryos, or at least gives him some reason to believe that they do not, the preformationist, if rational, will no longer believe that sperm cells contain embryos. So he will no longer have a visual experience of sperm cells containing embryos when he looks in the microscope. Since his seeming does not withstand the presence of a defeater assuming rationality, it is epistemic. So sensible dogmatism does not entail that his visual seeming confers prima facie justification on his belief.

Acknowledgments I am grateful to Sin yee Chan, Terence Cuneo, Louis deRosset, Adam Hosein, Michael Huemer, Don Loeb, Kevin McCain, Mark Moyer, Alastair Norcross, Robert Pasnau, Michael Tooley, Chris Tucker, Adam Wager, Matt Weiner, and audiences at Colorado and Vermont for helpful comments on this chapter.

References Ayer, A. J. 1940. The Foundations of Empirical Knowledge. New York: Macmillan. Brogaard, Berit. Forthcoming. “Perceptual Reports.” In Mohan Matthen, ed., Oxford Handbook of the Philosophy of Perception. _____. 2013. “Do We Perceive Natural Kind Properties?” Philosophical Studies 162: 35–42. _____. 2011a. “Are There Unconscious Perceptual Processes?” Consciousness and Cognition 20: 449-63. _____. 2011b. “Do ‘Looks’ Reports Reflect the Contents of Perception?” Paper presented at the Language of Consciousness workshop, ANU, Australia, July 28–29, 2011. Chalmers, David. 2005. “The Matrix as Metaphysics.” In Christopher Grau, ed., Philosophers Explore the Matrix, 132–76. Oxford: Oxford University Press. Chisholm, Roderick. 1957. Perceiving: A Philosophical Study. Ithaca: Cornell University Press. _____. 1942. “Discussions: The Problem of the Speckled Hen.” Mind 51: 368–73. Dehaene, Stanislas. 1997. The Number Sense. New York: Oxford University Press. Harman, Gilbert. 1977. The Nature of Morality. Oxford: Oxford University Press. Heim, Irene. 2006. “Remarks on Comparative Clauses as Generalized Quantifiers.” Unpublished manuscript. Huemer, Michael. 2008. “Revisionary Intuitionism,” Social Philosophy & Policy 25: 368–92. _____. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Larson, Richard. 1988. “Scope and Comparatives.” Linguistics & Philosophy 11: 1–26.

Sensible Dogmatism

289

Lewis, David. 1979. “Counterfactual Dependence and Time's Arrow.” Noûs 13: 455–76. Markie, Peter. 2005. “The Mystery of Perceptual Justification.” Philosophical Studies 126: 347–73. Pollock, John L., and Joseph Cruz. 1986. Contemporary Theories of Knowledge. Savage, MD: Rowman and Littlefield. Pryor, James. 2005. “There Is Immediate Justification.” In Matthias Steup and Ernest Sosa, eds., Contemporary Debates in Epistemology, 181–202. Malden, MA: Blackwell. _____. 2000. “The Skeptic and the Dogmatist.” Nous 34: 517–49. Siegel, Susanna. 2012. “Cognitive Penetrability and Perceptual Justification.” Nous 46: 201–22. _____. 2011. The Contents of Visual Experience. Oxford: Oxford University Press. _____. 2005. “Which Properties Are Represented in Perception?” In Tamar Szabó Gendler and John Hawthorne, eds., Perceptual Experience, 481–503. Oxford:  Oxford University Press. Tolhurst, William. 1998. “Seemings.” American Philosophical Quarterly 35: 293–302. Tucker, Chris. 2010. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24(1): 529–545.

This page intentionally left blank

{ part vi }

Phenomenal Conservatism

This page intentionally left blank

{ 13 }

Phenomenal Conservatism and the Principle of Credulity William G. Lycan Lycan (1985, 1988)  defended a “Principle of Credulity”:  “Accept at the outset each of those things that seem to be true” (1988, 165). Though that takes the form of a rule rather than a thesis, it does not seem very different from Huemer’s (2001, 2006, 2007) doctrine of phenomenal conservatism (PC): “If it seems to S that p, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p” (2007, 30).1 My Principle was differently motivated and put to uses different from Huemer’s. In this chapter I explore some of the differences.

1. Comparison To make an initial comparison: The Principle was put forward as an epistemic norm, and in particular as an ought rather than merely a permission. It seems fair to suppose that one who obeys an epistemic ought is to some degree justified in holding the resulting belief (by “accept” I meant merely, believe).2 Thus, if we convert the Principle into declarative mood, we may infer “If S believes at the outset a thing [= proposition] that seems (to S) to be true, S’s belief is to some degree justified.” Three remaining differences between that formulation and PC deserve comment. First, “at the outset.” Outset of what? I  proposed my Principle in aid of developing a coherentist epistemology, specifically, an explanationist

1 “Phenomenal” here has only its etymological meaning of seemings or appearances; it does not mean what philosophers of mind do in speaking of “phenomenal states,” “phenomenal character,” and “phenomenal consciousness.” 2 No reference to the more recent distinction in philosophy of science between mere “acceptance” of a scientific theory and actual belief in it.

294

Phenomenal Conservatism

one. My purpose was to solve the problem of what Keith Lehrer (1974) called “explained unexplainers”: Some explananda that justify their explanantia are themselves justified by explaining more primitive data in their turn. But many of the explananda, such as spontaneous perceptual beliefs or memories, cannot be justified in that way, because they themselves do not explain anything else. So now we have an apparently vicious circle: an explanatory hypothesis is supposed to be justified by the data it explains, but it can be so justified only if the data propositions are themselves justified; and the data propositions are justified “by being explained” only if the explanatory hypotheses are themselves justified. Since an explanans presupposes an explanandum, how is the whole explanatory enterprise to get started in the first place? There must be data propositions of some sort that are independently justified before explanatory coherence can come into play. Hence my Principle of Credulity. As written, it applies only to such data propositions, so in that respect it is weaker than PC. Second difference: Huemer adds the qualification “in the absence of defeaters.” I am not sure why he does, for it seems redundant. Possibly, where I am speaking of prima facie justification only, leaving it understood that prima facie justification can be defeated, Huemer means justification overall.3 If that is right, the present difference between the formulations is insignificant. Third, where I  speak of justified belief, PC alleges only “justification for believing that p.” As Huemer (2007) emphasizes, that is an important distinction (going back, in my memory, to Lehrer (1965)). A person may possess any number of justifications or evidential reasons for believing that p, and believe that p, and yet not be justified in her or his belief that p, if that belief is not based on any of those reasons. So in this regard the Principle of Credulity seems to be stronger than PC.

2. Capitulatory Adjudication The first and third differences need adjudicating, because each would be the basis for a Huemer objection to my Principle. Regarding the first, Huemer would complain, as he does (2007, 36–37) against views that privilege some seemings over others as eo ipso conferring justification, that I  have shown no epistemically relevant difference between my “outset” data propositions, the explained unexplainers, and any other seeming. (The point is not that the Principle is untrue, but that it is arbitrarily weak.) He would be right about what I did not show; my Principle as first introduced had only the very limited

3 Huemer (2001) did speak of prima facie justification, and that formulation did not contain the qualification about defeaters.

The Principle of Credulity

295

purpose mentioned earlier and did not need to be stronger. More to the point, I cannot even now think of an epistemically relevant difference, so I concede, and hereby delete “at the outset,” leaving this: “If S believes a proposition that seems (to S) to be true, S’s belief is to some degree justified.” Anent the third difference: at first reading I thought it was illusory. Consider Lehrer’s original example and then an adaptation of it intended as a counterexample to the Principle of Credulity. [A] detective who rejects the truthful testimony of a reliable eye-witness to a crime, but accepts the lying testimony of an ignorant meddler, when both tell him that Brentano committed the crime, would fail to be completely justified in believing this. For his belief is not based on the adequate evidence supplied by the truthful eye-witness but is instead based on the inadequate evidence supplied by an ignorant man. (1965: 169)

The adaptation: It seems to S that p, but S for whatever reason rejects that seeming; yet S does believe that p because S trusts some clearly unreliable extraneous source. S’s belief is not completely justified, indeed not justified at all.

But for what sort of reason does S reject the seeming? Two subcases here. In the first, S has evidence that the seeming is meretricious and not to be trusted. Here certainly S is not justified in believing that p. But that is because S’s phenomenal prima facie justification is defeated by S’s evidence against the trustworthiness of the seeming. This is no counterexample to my Principle. Yet, second subcase: what if there is no such impugning evidence, but for whatever (other) reason S still does not believe on the basis of the seeming? That would be psychologically anomalous; if it seems to S that p, and S has no reason to distrust that seeming, then presumably S does believe at least to some small degree that p. That sounds nearly analytic, but it is not. As has been widely argued,4 seemings are not per se beliefs, but only give rise to beliefs. The situation I have

4 Sosa (1998, 2006), Bealer (1999), Huemer (2001), Cullison (2010). Huemer (2007, 31) points out that a seeming may not even be a disposition to believe. And here for your delectation is a novel argument for the distinction. Lycan (forthcoming) argues that we philosophers do not believe our own views in the same sense in which we believe ordinary things. As Hume and Moore pointed out, there is often a large and striking gap between our doctrines and our behavior. (Similar points have been made about religious belief, notably by van Leeuven (2011).) It is important to see that these are typically not cases of hypocrisy or other insincerity; when we avow a philosophical thesis and defend it, sometimes passionately, we mean what we say. Nonetheless, the thesis does not show in our nonverbal behavior (think of betting), or even in our off-duty verbal behavior. Philosophical “belief ” does not have the functional profile of everyday belief; I think it is a somewhat different propositional attitude, which might be called “taking the position that.” Yet there is no doubt that philosophical propositions often strongly seem true to us, in the everyday sense of “seem”—thus another reason for not understanding seeming as belief.

296

Phenomenal Conservatism

imagined may be psychologically anomalous, but it is not metaphysically or likely even psychologically impossible. So I think we must grant that I am not strictly entitled to say that the belief would be justified. I must fall back to “If S believes a proposition at least in part because it seems (to S) to be true, S has some degree of justification for doing so.” And that last formulation is not wildly different from PC. I seem to have turned into a phenomenal conservative. (And so, I believe that I have done so.) Nonetheless there are still differences in motivation and in uses, and in liability to objections.

3. A Different Defense PC has been defended on each of several grounds: that, versus the justification skeptic, it (obviously) does vindicate gabillions of our commonsense beliefs; that considered as policy it conduces to our pursuit of truth (Huemer 2001); that to deny it is self-defeating (Huemer 2001, 2007); and that it is the best explanation of internalist intuitions (and thereby supports epistemological internalism; Huemer 2006).5 But my argument (1988, chs 7 and 8) for the Principle of Credulity was entirely different, and I believe it extends to PC. It was an argument from engineering design. I used the fanciful example of a benevolent Mother Nature (a slightly Panglossian personification of natural selection), and asked, with what epistemic norms would she endow creatures of our general physical type in an environment like that of ours on earth? I  argued that she would implant the usual principles of explanatory theory preference: given two theories that explain the same data, prefer the simpler; the one that explains more in addition; the more readily testable one; the one that leaves fewer messy unanswered questions behind; the one that squares better with what you already have reason to believe. (I should have added, the more fruitful or fecund one.) There were some much-needed qualifications, most notably that such principles often mutually conflict and have to be weighed against each other. In each case, I offered reasons for why the preference would make for good cognitive design. For example, regarding simplicity (I apologize, but not sincerely, for quoting myself): (1) Simpler hypotheses are more efficient to work with. A simple handbook of rules, such as the Boy Scout Manual, is easier to use than is the 1976 U.S.  tax code. (2)  As Russell (1957) observed in defense of his version of Occam’s Razor, complexities incur greater risk of error. A simpler device has

5

Huemer also deploys apposite examples that go directly against externalism.

The Principle of Credulity

297

less that can go wrong with it (think of a simplified phonograph turntable6 or automobile engine). (3) Simplicity is itself a form of efficiency. The whole point of obtaining simple and unified hypotheses in science is to achieve plenitude of result (in the way of data explained and results predicted) with parsimony of means. If we were not able to mobilize a few simple hypotheses and thereby obtain maximally informative analysis of the news, especially in the way of experimental predictions, we would be far less competent in coping with environmental developments; the world would present us with too many surprises, and they would overwhelm us. (Lycan 1988, 140–41)

Of course the explanatory virtues are open to many skeptical challenges. The Mother Nature argument is not intended as a reply to any such; I had already given entirely independent replies to as many as I could fit in (1988, ch 7). Nor, n.b., did I  claim either that our explanatory virtues’ adaptive utility justifies them in the epistemological sense or that they per se provide any guarantee of true beliefs as output. I take the canons of theory preference to be epistemologically basic and not susceptible of justification by being derived from or being tested against some more fundamental norm(s); as Bentham said, “That which is used to prove everything else cannot itself be proved.” (Formal probability theory, in particular, is not a more fundamental norm [see Lycan 2012].) At best, the canons are justified by reflective equilibrium, which is itself an explanatory coherentist method. The point of the Mother Nature argument was, rather, to offer a convincing reason why it is good and desirable for us to be designed to use the methods we (ideally) do use rather than being designed in some other way. And that reason explains the important sense in which our use of those methods is not arbitrary and is not just a matter of “making our minds feel good” (Hacking 1982).7

4. Conservatism The most contentious of the usual canons of theory preference is that of conservatism. Earlier, I expressed it as the policy of choosing the hypothesis that best squares with what you already have reason to believe, but actually I myself defend an even bolder version:  prefer the hypothesis that best squares with what you already do believe, reasonably or not. That is the really contentious version; critics who do not mind simplicity, testability, fruitfulness, and the rest sometimes balk at conservatism in this bolder sense, because it sounds

6

For those of you old enough to know what a phonograph turntable is. I do suggest that the goodness, in the cost-benefit sense, of Mother Nature’s design for cognizers is the ultimate ground of the value notions of epistemology (“justification,” “warrant,” “good reason,” “legitimacy,” “license”). 7

298

Phenomenal Conservatism

particularly dogmatic, bigoted, pigheaded (Lehrer 1974; Stich 1990. For a more nuanced critique, see Christensen 1994). The Mother Nature argument for it would be that arbitrary and gratuitous changes of belief, like arbitrary and gratuitous changes of institutional policies, come only at a price; they draw on energy and resources (see Harman 1986). (“Arbitrary” and “gratuitous” in the sense of gaining no offsetting advantage; see section 7 of this chapter.) Also, the instability created by a habit of capricious belief change would be inefficient and confusing. As before, these are not deeper justifications of the conservative policy; according to me, there cannot be one. They only explain why it is a good thing that our brains are conservative. Conservatism has a slightly startling consequence (Lycan 1988, 162). Consider my present belief set B, and two theories T1, and T2. T1 is logically stronger than B; T2 is incompatible with B, though T2 and B may overlap. Now, T2 may outweigh T1 in explanatory virtue, in which case we should reject T1 in favor of T2. But suppose it does not. Then according to our strengthened canon of conservatism, we should prefer T1. But T1 entails B, my present belief set. It follows that I am justified in accepting B, merely in virtue of my already holding B. Our rule of conservatism, then, entails the claim that the bare fact of one’s holding a belief renders that belief justified, to some degree however slight; for any belief at all. So be it. Conservatism applies to beliefs and not directly to seemings, but I think a parallel argument supports the Principle of Credulity. Though the appearancereality distinction is (of course) vitally important, separating appearance from a differing reality is costly, again drawing on energy and resources. As before, there is sometimes an advantage to be gained by questioning an appearance rather than just maintaining the corresponding belief, but when none is in prospect, to hold the appearance at arm’s length just runs the battery down. And suppose we are not to take most seemings at face value. Though (according to me) we would still hold many spontaneous beliefs that are not based on seemings, there would still be at least a slight fog of unresolved seemings that probably would slow down action. And these points apply not only to explained unexplainers, but to all appearances or seemings. Thus they recommend, though do not demonstrate, PC.

5. Internalism Regarding the four other arguments for PC mentioned earlier: the first (vindicating commonsense beliefs) should be uncontroversial. I am neutral on the second, about pursuit of truth. I will not address the self-defeat issue, except to note that in the matter of intuitions in particular, no philosopher is in a

The Principle of Credulity

299

position to deny often taking them at face value (nor any logician, nor any linguist). Intuitions are addressed directly in the next section.8 But I  do want to comment on the matter of internalism versus externalism, because therein lies a large difference in motivation between PC and the Principle of Credulity. The Principle has little to do with the internalism/externalism debate. All it does is furnish the explanatory-coherence machine with initial data propositions.9 For that matter, I have never been sure whether the explanatory virtues themselves are internal or external.10 Can I tell from the inside which of two hypotheses is simpler, or which explains more, which is more testable, and so forth? Sometimes I think I can, and it is possible I always can subconsciously. But I could not say with confidence. I myself have no stake in the internalism/externalism debate. For the record, I incline toward the view of Kornblith (1983) and Battaly (2001) that there are two simply different though related notions of “justification,” answering to two separate interests we have in cognition.11 One is that of a person’s usefulness as an informant in one context or another. Reliabilism and hence externalism are appropriate accounts of that. The other is the person’s epistemic responsibleness, whether she or he is behaving as she or he ought. There some however loose version of “ ‘ought’ implies ‘can’ ” applies, and it is reasonable to think that the properties figuring in the relevant norms are internally accessible to the subject. (Granted, that is not the end of the matter. Consider cases of “lost justification” as appealed to by reliabilists. I know that the speed of light in a vacuum is a bit over 186,000 miles per second, but I have no idea when or where I learned that and I could not begin to defend that belief out of my head. I am, I think, a reliable informant on that momentous issue. But does my inability to produce evidence make me epistemically irresponsible? It could be argued that I currently remember that I learned the fact somewhere, and I am justified in trusting that memory. Or it could be argued that my belief is supported by buried memory traces, or some such. I grant that I do not feel irresponsible.)

8 I should add on the opposing side that I do not endorse Huemer’s (2007, 39) premise that “when we form beliefs, with a few exceptions not relevant here, our beliefs are based on the way things seem to us.” The premise may or may not be right if seemings can be unconscious, and “based on” is a merely causal notion, but it is certainly wrong if seemings are necessarily conscious states. In addition, I deny that seemings are always or even very often data for explanation; normally our explained unexplainers are propositions about the external world. 9 However, conservatism more generally has a second role: it breaks ties, when two hypotheses are otherwise equal in explanatory merit (see Lycan 1988, 174–75). 10 Huemer (2006) usefully distinguishes five different versions of the “internal”/“external” distinction. 11 It is remarkable that this view has received so little discussion.

300

Phenomenal Conservatism

6. Intuitions Opinions differ as to what an “intuition” (in contemporary philosophers’ sense of the term)12 is. In this section I argue that intuitions fall into the category of seemings, and so are subject both to PC and to the Principle of Credulity. I would begin by maintaining that intuitions are about cases, actual or hypothetical—Gettier examples, moral situations, particular sentences or argument forms, hypothetical scenarios.13 We might call an intuition a verdictive judgment on a case, not consciously based on inference or on any other particular reason. But that characterization would already be controversial, and I think it is not quite right. Plantinga (1993, 105) speaks of “finding yourself utterly convinced that the proposition in question is true . . .[, and that it] is not only true, but could not have been false.” That is consistent with an intuition’s being a judgment, but as noted in section 2 above (and fn 4), Sosa (1998, 2006), Bealer (1996), Huemer, and also BonJour (1998) question Plantinga’s assumption that intuiting is a type of believing. Rather, Bealer calls it an intellectual seeming, “a sui generis, irreducible . . . propositional attitude that occurs episodically” (28–29, n.  6). BonJour actually assimilates it to perception or apprehending, a matter of “seeing” that a proposition is necessary; the proposition appears to be necessarily true. BonJour (2001, 2005)  moves even further in the perceptual direction, maintaining that intuitions are not propositional at all but are more like perceptual sensations. But Plantinga’s and BonJour’s common focus on the necessity of a proposition as the object of an intuition is far too restrictive: though we do have modal intuitions, and perhaps intuitions of some types do have as their contents the necessity of some proposition, there is no reason to think that a syntactic intuition or a moral one takes that form. (Someone might argue that in the moral case, what is intuited to be necessary is the conditional from the facts of the case to the moral verdict, but (a) that is not the phenomenology, and (b) to say that the conditional was necessary would be at best controversial in the first place.) Also, I  see no motivation for assimilating intuiting to perceiving, once a seeming has been clearly distinguished from a belief. BonJour is right to speak of “appearing,” but the sort of appearing that constitutes an intuition is no more perceptual than cognitive (for a further defense, see Chudnoff 2011a).

12 That sense is technical and differs from those of ordinary English speakers, psychologists, and Kant. 13 What about intuiting the truth of a general principle, say, PC itself? That could be a linguistic or conceptual intuition to the effect that PC is conceptually necessary, or it could be a modal intuition that the scenario—its seeming to S that p and S’s having no defeaters but S’s being to no degree justified in believing that p—is impossible.

The Principle of Credulity

301

Intuitions, then, are intellectual seemings-true.14 And so both the Principle of Credulity and PC apply to them—fortunately for us philosophers. In what remains of this chapter I address one class of objections to PC.

7. The Charge of Liberalism It has been objected to PC that PC is too liberal. The same objection has (understandably) been made to conservatism and to the Principle of Credulity.15 Obviously, if anything that seems true is to be believed, and if any belief whatever is to some degree justified by the mere holding of it, then many completely crazy and paradigmatically irrational beliefs will count as justified. That sounds fatal, but it is not. I begin by clarifying my own position (Lycan 1985, 1988), and then turn to a few more specific versions of the complaint. In my case, it comes down to the degree of justification. In my scheme, the justification furnished by the Principle of Credulity is minute, the faintest edge, infinitesimal if you like. (That is all it needed to be, to feed the explained unexplainers into the coherence apparatus, and to break ties.) And as before, the justification is only prima facie, outweighed by any defeater or explanatory advantage however small (Lycan 1988, 175). So although I  do maintain that “anything that seems true is to be believed” and that “any belief whatever is to some degree justified by the mere holding of it,” those ways of putting it are misleading. They make it sound as though the subject is epistemically fine and can go around freely acting on the beliefs in question without compunction.

14 Williamson (2007) grants this, but offers a deflationary account of it in turn: “For myself, I am aware of no intellectual seeming beyond my conscious inclination to believe . . .” (217). We may or may not grant that every seeming-true is a conscious inclination to believe, but such an inclination is hardly sufficient for being an intuition. There are many things I am consciously inclined to believe: that my daughter was born in Sydney, that there is a moth flitting over to my left, that since the coin has come up heads six times in a row it will almost certainly come up tails on the next toss, and that classes begin on Monday. (Some of these I do also actually believe, some not.) But none would qualify as an intuition in the philosophers’ sense. Not all are even seemings of any kind. Earlenbaugh and Molyneux (2009) too hold an “inclination to believe” view, but dropping the qualification “conscious.” They defend it ingeniously and at considerable length. (Far less convincingly, they also argue that despite advertising and appearances, intuitions are not actually treated as evidence by philosophers.) N.b., like Williamson’s, their view is entirely compatible with that of intuitions as seemings. My main complaint here is the same as against Williamson except more so: if not every conscious inclination to believe is an intuition, then certainly not every inclination to believe is one. Earlenbaugh and Molyneux anticipate this (89, n2): “[W]e think it is possible to informatively classify a species as belonging to a genus without giving a full specification of its nature (consider the case of classifying a platypus as a mammal).” Agreed, but our project in this section of this paper is to say what an “intuition” is, not merely to assign it to a much broader genus. (Mind you, I am not entirely sure that all intellectual seemings are intuitions in the philosophers’ sense.) 15 At the maiden presentation of Lycan (1985), at the University of Pittsburgh, Adolf Grünbaum held forth at great length and at high volume.

302

Phenomenal Conservatism

Of course the latter is not so. I should have emphasized the qualification, “at least to the minutest degree.” There is still a problem, originally put to me by Robert Vishny:  even if the force of conservatism is vanishingly small, why does it go nonexistent in the face of compensating explanatory advantages? What about a situation in which two hypotheses differed so slightly in their other explanatory virtues that conservatism managed to bring the lesser just precisely up to the level of the former in overall goodness? In that case, we would have a tie, and no conservative canon left to break it. That point requires a concession: we cannot regard conservatism as a firstorder rule on a par with simplicity, power, and the rest; it must be a meta-rule, to be invoked only after the first-order rules have been applied and weighed against each other. That is already signaled by the formulation (Lycan 1988, 176) “Do not change your view for no reason,” because that last qualification is recursive; “reason” has to mean “advantage in terms of the other epistemic virtues.” Thus, conservatism occupies a slightly specialized place in our cognitive design. I have never been entirely happy with that feature, but it would help explain why critics are more skeptical about conservativeness than about the other explanatory virtues. I turn now to other charges of liberalism. I think they are bad arguments, and my replies to them will not depend on my particular view that the justification afforded by seemings is minute. Matthias Steup makes the following two objections to PC: Asserting that seemings are a source of justification no matter what, PC would appear to be Ipso Factism about seemings [the view that a sense experience that P is, ipso facto, a source of justification for believing that P]. But in worlds whose inhabitants have evidence of perceptual unreliability, perceptual seemings are not a source of justification. Even in our world, seemings are not always a source of justification. Suppose it seems to S as if P because S has an intense desire that P. S does not thereby have justification for believing that P. So not all kinds of seemings are a source of justification. Some are not because we have evidence for thinking they are unreliable. (Steup 2004, 415)

But each of these ignores the merely prima facie nature of the justification (or in Huemer’s [2007] terms, the absence of defeaters). In a world whose inhabitants have evidence of perceptual unreliability, that evidence defeats the conservative presumption. The case of wishful thinking is slightly more complicated, depending on whether the subject is able to consider that the thinking may be wishful. Suppose (a) the subject does so consider and inclines to admit the fault. Then, obviously, the justification is defeated. If (b) the subject merely concedes that the thinking may be wishful, the justification is at least partially defeated. If

The Principle of Credulity

303

(c) the subject does not consider, despite having reason to (such as knowing she or he is prone to wishful thinking on the topic in question), in that case I  think the justification is defeated. But if (d)  the subject does not consider because she ir he has no reason to, then it seems to me her or his belief is still to some degree justified, so long as there is no other defeater in the form of countervailing evidence or the like. To insist otherwise would be to beg the question against PC. Peter Markie (2005) offers two alleged counterexamples to PC. The first (356–57) is a case in which a prospector who yearns to strike gold finds a pebble and because he wishes it so, it seems to him to be gold. So far, my previous reply would suffice, but Markie adds the wrinkle that to a co-prospector, the same pebble seems to be gold, but in his case because of his “learned identification skills.” “[C]ertainly my wishful thinking should not gain my perceptual belief the same positive epistemic status of defeasible justification as your learned identification skills.” Of course not, but PC does not rule otherwise, even in case (d) cited earlier. Even if the justification provided by seeming is more substantive than my own minute sort, it is hardly as weighty as a seeming based on expertise. The wishful prospector is justified to some degree; the learned one is better justified, certainly on grounds of reliability and probably on grounds of coherence. In the skilled co-prospector’s case, there will be other supporting beliefs about the appearance of the pebble, its disposition in the setting, the locale, the coprospector’s own track record, etc., all of which the wishful prospector lacks. In Markie’s second example (p. 357), I see a walnut tree in my yard, and the tree both seems to me to be a walnut and seems to me to have been planted on April 24, 1914. The first seeming is due to my ability to identify trees; the second is due to sheer cognitive malfunction. Its seeming to me that the tree was planted on that date outstrips the phenomenological character of my experience and my identification skills. My perception cannot directly justify my belief about the planting date. Nonetheless, according to (PC), both my belief that it is a walnut tree and my belief that it was planted on April 24, 1914, are prima facie, and so defeasibly, justified for me.

Markie goes on to insist that no appeal to a defeater will help, for his objection is that the belief about the date should not count as even defeasibly justified. Here there is an even more obvious difference than in the gold nugget case between the two beliefs. But as then, that alone does not show that the belief about the date is not justified to some tiny degree. The point would have to be that the belief could not possibly be justified by the subject’s perceptual experience. This is indisputably true, but not pertinent: in this case, the phenomenal conservative would not cite perceptual experience as justifying, but rather the seeming itself.

304

Phenomenal Conservatism

Markie might still insist that in the absence of empirical evidence, the belief is not even prima facie justified. But that would just beg the question.16 Another putative counterexample is put forward by Michael Bergmann (forthcoming, sec 2.2): it seems to Jack that he has a hard spherical object in his hand, because he has in fact grabbed a billiard ball and has the appropriate tactile sensations; so he believes that he has a hard spherical object in his hand. It seems to Jill that she has a hard spherical object in her hand, and so she believes as well. But Jill’s seeming is based, not on any tactile sensations, but on a smell as of a lilac bush (cognitive/brain malfunction again). Bergmann contends that Jill’s belief is (entirely) unjustified, because “[h]er seeming about the hard spherical object is improperly caused.” So of course it is, and her belief is not nearly as well justified as is Jack’s, who could cite the vivid feel of the actual ball in his hand. But it does not follow that Jill’s belief is not justified to any degree at all. The case is a little hard to imagine—Jill has no appropriate tactile sensations, but it seems to her that she has a hard spherical object in her hand? We would need to hear more (for example, is she aware that she has no such tactile sensations?), and perhaps depending on the details we would agree that her prima facie justification is defeated; but Bergmann has done nothing to show that she does not have even a smidgen of prima facie justification.

8. Conclusion Though the arguments for PC are not crushing, I stand by my own. I also find PC plausible in itself, though not everyone else will. Moreover, I have rebutted the strongest direct objections to PC known to me.

Acknowledgment I am grateful to Kevin McCain and to Chris Tucker for lavish and very helpful comments on my penultimate draft (one of which saved me from an embarrassing error).

References Battaly, Heather. 2001. “Thin Concepts to the Rescue: Thinning the Concepts of Epistemic Justification and Intellectual Virtue.” In Abrol Fairweather and Linda Zagzebski, eds., Virtue Epistemology, 98–116. Oxford: Oxford University Press. Bealer, George. 1996. “On the Possibility of Philosophical Knowledge.” In J. E. Tomberlin, ed., Philosophical Perspectives 10: Metaphysics, 1–34. Malden, MA: Blackwell. 16

For a different but related reply to Markie, see Chudnoff (2011b).

The Principle of Credulity

305

Bergmann, Michael. Forthcoming. “Externalist Justification and the Role of Seemings.” Philosophical Studies. BonJour, Lawrence. 2005. “In Defense of the A Priori.” In Matthias Steup and Ernest Sosa, eds., Contemporary Debates in Epistemology, 98–104. Oxford: Blackwell. _____. 2001. “Précis” and “Replies.” In a Book Symposium on BonJour (1998), Philosophy and Phenomenological Research 63: 625–31, 673–98. _____. 1998. In Defense of Pure Reason. Cambridge: Cambridge University Press. Christensen, D. 1994. “Conservatism in Epistemology.” Noûs 27: 69–89. Chudnoff, Eli. 2011a. “What Intuitions Are Like.” Philosophy and Phenomenological Research 82: 625–54. Chudnoff, Eli. 2011b. “The Nature of Intuitive Justification.” Philosophical Studies 153: 313–33. Cullison, Andrew. 2010. “What Are Seemings?” Ratio 23(3): 260–74. Earlenbaugh, Joshua and Bernard Molyneux. 2009. “Intuitions Are Inclinations to Believe.” Philosophical Studies 145(1): 89–109. Hacking, Ian. 1982. “Experimentation and Scientific Realism.” Philosophical Topics 13: 71–88. Harman, Gilbert. 1986. Change in View. Cambridge, MA: Bradford/MIT Press. Huemer, Michael. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. _____. 2006. “Phenomenal Conservatism and the Internalist Intuition.” American Philosophical Quarterly 43: 147–58. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Kornblith, Hilary. 1983. “Justified Belief and Epistemically Responsible Action.” Philosophical Review 92(1): 33–48. Lehrer, Keith. 1974. Knowledge. Oxford: Oxford University Press. _____. 1965. “Knowledge, Truth and Evidence.” Analysis 25(5): 168–75. Lycan, William G. Forthcoming. “On Two Main Themes in Gutting’s What Philosophers Know.” Southern Journal of Philosophy. _____. 2012. “Explanationist Rebuttals (Coherentism Defended Again).” Southern Journal of Philosophy 50(1): 5–20. _____. 1988. Judgement and Justification. Cambridge: Cambridge University Press. _____. 1985. “Conservatism and the Data Base.” In N. Rescher, ed., Reason and Rationality in Natural Science. University of Pittsburgh Center for the Philosophy of Science Publications. Lanham, MD: University Press of America. Markie, Peter. 2005. “The Mystery of Perceptual Justification.” Philosophical Studies 126: 347–73. Plantinga, Alvin. 1993. Warrant and Proper Function. Oxford: Oxford University Press. Sosa, Ernest. 2006. “A Defense of the Use of Intuitions in Philosophy.” In M. Bishop and D. Murphy, eds., Stich and His Critics. Oxford: Basil Blackwell. _____. 1998. “Minimal Intuition.” In Michael DePaul and William Ramsey, eds., Rethinking Intuition. Lanham, MD: Rowman and Littlefield. Steup, Matthias. 2004. “Internalist Reliabilism.” Philosophical Issues 14: 403–25. Stich, Stephen. 1990. The Fragmentation of Reason. Cambridge, MA:  Bradford Books/ MIT Press. Van Leeuven, Neil. 2011. “Religious Credence and Factual Belief.” MS, Georgia State University. Williamson, Timothy. 2007. The Philosophy of Philosophy. Oxford: Basil Blackwell.

{ 14 }

Michael Huemer and the Principle of Phenomenal Conservatism Michael Tooley Introduction In his book Skepticism and the Veil of Perception (Huemer 2001), and then in his later essay “Compassionate Phenomenal Conservatism” (Huemer 2007), Michael Huemer has argued that it is how things seem to a person that determines whether the person is justified, at least to some extent, in accepting a given proposition. More specifically, Huemer has defended what he refers to as the “Principle of Phenomenal Conservatism,” and he argues that any attempt to reject that principle is self-defeating. I shall argue both that this claim is mistaken, and that the Principle of Phenomenal Conservatism is unsound. My discussion is as follows. In the first section, I  set out and compare Huemer’s two versions of the Principle of Phenomenal Conservatism. What emerges is that Huemer’s revised version of his principle is much more modest than the original version. Next, in discussions of the Principle of Phenomenal Conservatism, it is almost invariably assumed that there is a clear principle that Huemer has advanced, and whose merits are therefore to be assessed. In section 2, however, I argue that Huemer has not provided a satisfactory account of what seemings, in his sense, are, so that he has not in fact succeeded in formulating any clear principle. Then, in section 3, I offer an argument against the existence of seemings by considering mental states that do not involve any non-cognitive qualia. In section 4, I consider whether focusing instead upon the idea of appearances can provide a solution to the problem of explaining what seemings, in Huemer’s sense, are. Next, in section 5, I focus upon the justificatory permissiveness that, I claim, characterizes all versions of a Principle of Phenomenal Conservatism, and I argue that this permissiveness arises out of the failures to distinguish, first,

Huemer’s Phenomenal Conservatism

307

between basic and derived seemings, and, second, between qualia-involving and qualia-free seemings. Then, in section 6, I  argue that Huemer’s contention that denial of the Principle of Phenomenal Conservatism is self-defeating is mistaken, since one can formulate a Restricted Principle of Phenomenal Conservatism, using the idea of basic seemings, that is, in fact, more plausible than Huemer’s unrestricted principle. Finally, in section 7, I conclude by briefly indicating why direct awareness or direct acquaintance approaches to the foundations of justification are superior to approaches based on a Principle of Phenomenal Conservatism, either restricted or unrestricted.

1. Two Versions of the Principle of Phenomenal Conservatism In Skepticism and the Veil of Perception, Huemer formulated what he refers to as the Principle of Phenomenal Conservatism (PC): PC

If it seems to S as if P, then S has at least prima facie justification for believing that P (Huemer 2001, 99).

Then, in “Compassionate Phenomenal Conservatism,” Huemer offered a revised account of the Principle of Phenomenal Conservatism, partly in response to criticisms that I had advanced in conversation: PCR If it seems to S that p, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p (Huemer 2007, 30).

This revised version of the principle involves, as Huemer (2001, 30), points out, two changes from his original formulation:  first, the revised principle is not restricted to non-inferential justification; second, it is concerned, not with whether, in the absence of defeaters, a belief is fully justified, but with whether a person is, in the absence of defeaters, justified at least to some extent in accepting the belief. The idea of being justified at least to some extent in accepting a belief can, however, be interpreted in either a stronger or a weaker sense, depending upon whether belief is understood in the everyday sense, in which to believe that p entails not believing that not p, or, instead, in the technical sense in which belief admits of degrees ranging from complete subjective certainty that p is false through to complete subjective certainty that p is true. Interpreted in the former way, PC could be expressed as Strong

If it seems to S that p, then, in the absence of defeaters, S thereby has some justification for believing that p rather than not p.

308

Phenomenal Conservatism

Interpreted in the second way, and given a framework involving either epistemic or logical probability, PC might be expressed along the lines of Modest 1 If it seems to S that p, then, in the absence of defeaters, that raises the probability of p.

If one wanted to avoid a framework involving probabilities, one alternative formulation would be Modest 2

If it seems to S that p, then, in the absence of defeaters, that justifies S in believing p more strongly than would be the case in the absence of the relevant seeming.

Which type of interpretation did Huemer have in mind? Something like Strong, or something like one of the Modest formulations? The answer is the latter (Huemer, personal communication). What this means is that the modified Principle of Phenomenal Conservatism that Huemer defends in “Compassionate Phenomenal Conservatism” is a much more modest foundational principle than that advanced in Skepticism and the Veil of Perception. This, in turn, has some important consequences. In particular, the defense that Huemer (2001) offers in chapter 5, “A Version of Foundationalism,” in Skepticism and the Veil of Perception, of one of two claims central to direct realism—the claim, namely, that one can be non-inferentially justified in believing in the existence of physical states of affairs—rests upon an appeal to the original, more robust version of the Principle of Phenomenal Conservatism. Whatever the overall merits of that defense may be, once one shifts to the more modest formulation of the Principle of Phenomenal Conservatism, it no longer follows, for example, that if it seems to one that there is something green in front of one, and one has no defeaters, that one is justified in believing that there is something green in front of one in the ordinary sense of “justified in believing,” which entails that that belief is more justified than its denial. In addition, however, one of the two central claims involved in the definition of direct realism may very well need to be modified, so that rather than saying that one can be non-inferentially justified in believing in the existence of physical states of affairs, it says instead that one can be justified (perhaps non-inferentially, perhaps only inferentially) in believing in the existence of physical states of affairs, and in a way not dependent upon justified beliefs about one’s sensory states. Similar problems will arise in other areas. Thus, Huemer appeals to the Principle of Phenomenal Conservatism, and to intellectual seemings, or intuitions, to justify acceptance of various propositions, including those that, if true, express necessary truths. Once one shifts to the modified version of the principle, the question arises whether the level of justification is sufficient to make it more likely that a given proposition is true than false, and the revised principle, in itself, provides no ground for answering that question.

Huemer’s Phenomenal Conservatism

309

2. What Are Seemings? 2.1. Huemer’s Account of the Concept of a Seeming The term seems is crucial to the formulation of the Principle of Phenomenal Conservatism. How, then, is the expression “it seems to S as if p” (or “it seems to S that p”) to be interpreted in Huemer’s statements of the principle? Huemer offers no analysis of what it is for it to seem to S that p. Instead, he attempts to explain his concept of a seeming as follows. First of all, a seeming is “a kind of propositional attitude” (Huemer 2007, 30). Second, its seeming to S that p is not only a distinct state from S’s believing that p, but also a state that is logically compatible with S’s not believing that p (Huemer 2001, 99). Third, there are at least four different types of seemings, namely, perceptual seemings, memory-related seemings, intellectual seemings (or “intuitions”), and introspective seemings. Thus, in “Compassionate Phenomenal Conservatism,” Huemer describes his view of seemings as follows: “I take statements of the form ‘it seems to S that p’ or ‘it appears to S that p’ to describe a kind of propositional attitude, different from belief, of which sensory experience, apparent memory, intuition, and apparent introspective awareness are species. This type of mental state may be termed an ‘appearance’ ” (Huemer 2007, 30). Is there anything more that should be added to this description of Huemer’s concept of seemings? The answer is that there is. In Skepticism and the Veil of Perception, Huemer (2001, 66) says that perceptual experiences typically have three aspects, consisting of sensory qualia, representational content, and forcefulness. He then goes on, in “4.3 The Forcefulness of Perceptual Experience,” to connect up the idea of forcefulness with two other ideas—the idea of representing some propositional content as being actualized, and the concept of seeming: The reason lies in what I  call the “forcefulness” of perceptual experiences: perceptual experiences represent their contents as actualized; states of merely imagining do not. When you have a visual experience of a tomato, it thereby seems to you as if a table is actually present, then and there. When you merely imagine a tomato, it does not thereby seem to you as if a table is actually present. (66)

Moreover, the concept of representing some propositional content as being actualized is also connected, in an earlier discussion, “The Concept of Awareness,” with the idea of “assertive mental representations” (53–54). It is natural to conclude, then, that Huemer’s concept of seemings is such that seemings are assertive mental representations, and this is in fact the case (Huemer, personal communication). Might one hold that, in all of this, it is the concept of a seeming that is basic, and that the concepts of forcefulness, of representing a propositional content as

310

Phenomenal Conservatism

being actualized, and of an assertive mental representation are all to be explained in terms of the concept of seemings? Clearly, this cannot be the case, since there are propositional attitude states that are not seemings but that do involve assertive mental representations, and that represent a propositional content as being actualized—such as beliefs (54), and thoughts that something is the case. The upshot is that, given, on the one hand, that being an assertive mental representation cannot be explained in terms of the concept of a seeming, and, on the other, that it is essential to the concept of a seeming that it be an assertive mental representation, no account of what a seeming is can be satisfactory unless one explains what an assertive mental representation is.

2.2. the basic problem: metaphors that are never cashed out The term assertive, as ordinarily used, applies only to people and is definable in terms of the verb “assert.” The only things that can literally assert anything, then, are things that can have intentions and that can perform actions. Mental states, then, cannot literally assert anything, and if one characterizes a mental state as assertive, that is a metaphor that needs to be cashed out in terms of some literally true characterization. Other philosophers have, in effect, acknowledged the metaphorical nature of such talk. Thus Chris Tucker (2010, 530), who views assertiveness as a crucial characteristic of seemings, immediately after introducing the term assertiveness, says, “The phenomenology of a seeming makes it feel as though the seeming is ‘recommending’ its propositional content as true or ‘assuring’ us of the content’s truth.” Since mental states are not agents, they cannot literally assert anything, or recommend anything, or assure us that anything is the case. So such metaphorical talk needs to be jettisoned and replaced with characterizations that are literally true. Since Huemer and Tucker have not done this, they have not really provided us with a satisfactory account of what they take seemings to be. How might the metaphor of an assertive mental representation be explained in literal terms? If one did not think that there were assertive mental representations that were distinct from beliefs and from related thoughts that something is the case, the task would be a reasonably straightforward one, since one might, for example, explain the idea of an assertive mental representation in terms of conditions that are related to the making of assertions—perhaps along the following lines: Person S’s mental state M with content p is an assertive mental representation =def There is some state T such that if person S were now in state T in addition to being in mental state M, then S would now be sincerely asserting that p, whereas if S were now in state T but not in mental state M, S would not now be sincerely asserting that p.

Huemer’s Phenomenal Conservatism

311

But if one holds that its seeming to S that p is compatible with S’s not believing that p, then there will be seemings that do not satisfy this definition of an assertive mental representation, since there will be no state T such that being in that state, while (1) it seems to one that p, but (2) one does not believe that p, entails that one is sincerely asserting that p. The problem, in short, is to find an account of the concept of assertive mental representations that allows for the possibility of being in a mental state with content p that is an assertive mental representation but that is also compatible with not believing that p.

3. Against the Existence of Seemings 3.1. the case of seemings not involving non-cognitive qualia There is another feature of Huemer’s concept of seemings that I have not yet mentioned but which is very important, since, among other things, this feature provides the basis for a strong argument against the view that there are any mental states answering to Huemer’s characterization of seemings. To explain this feature, one needs, first of all, the idea of qualia, where qualia can be understood in the manner that Huemer suggests in Skepticism and the Veil of Perception: A quale (plural: “qualia”) is a kind of property of an experience. The quale of an experience is what the experience is like, from the subject’s point of view; in other words, what it is like for the person who is having the experience. In the case of emotions and tactile sensations, this can be rephrased as how it feels when you have this experience. (Huemer 2001, 66)

Second, some philosophers hold that there are two fundamentally different types of qualia, which one can refer to as cognitive and non-cognitive. The latter involve, for example, secondary quality properties, such as colors, sounds, smells, and so on. The former are qualia present in certain types of mental states, such as thinking. If there are cognitive qualia—their existence is a highly controversial matter—and if seemings are conscious propositional-attitude states distinct from conscious beliefs and thoughts, then presumably there will be cognitive qualia that characterize seemings. What we are interested in here, however, are seemings that are not accompanied, or that do not involve, non-cognitive qualia. Some mental states are always accompanied by relevant, non-cognitive qualia, and in the case of perceptual seemings, it is rather natural to think that this is always the case—although, as we shall see, Huemer does not think that that is so.

312

Phenomenal Conservatism

In the case of what Huemer refers to as memorial seemings, some seem to be accompanied by relevant, non-verbal images that involve non-cognitive qualia, while others are not. Compare, for example, your memory of what your most recent meal was, with your memory, say, of the fact that Caesar crossed the Rubicon. Similarly, in the case of intellectual seemings, there may very well be a nonverbal image that involves non-cognitive qualia that accompanies its seeming to you that 2 + 2 = 4, but it is rather unlikely that there is such an image in the case of its seeming to you that 12 x 12 = 144. Let us focus, however, upon a case of perception where Huemer (2001, 67)  holds that no (non-cognitive) qualia are present, namely, the case of proprioception: Even in the dark, and when it is not touching another part of your body, you are aware of how your arm is positioned—whether it is pointing up, or straight out, whether it is bent, and so forth. This is a sense over and above the five senses, although most people are unaware of it (they are aware of their body position, but they are not aware that they have a special sense of body position). The reason they are not aware of it is that proprioception has no qualia. It represents one’s body as being positioned in a certain way, but there is no special feeling or other “what-it’s-like” to it—the only noticeable manifestation of proprioceptive “experiences” is that one is inclined to think that one’s body is positioned in a certain way.

Huemer supports this account by appealing to Oliver Sacks’ (1985, 43–54) discussion of the case of a woman who lost her sense of proprioception. But this account of proprioception seems to me clearly mistaken, since when one holds one’s arm in different positions, there are different sensations of muscular tension. Moreover, I do not think that Huemer’s account represents Sacks’ view, since Sacks (1985, 46) says of the woman in question that she had “no muscle or tendon or joint sense whatever,” and since it is surely not true that one has, in proprioception, seemings about the location of tendons, it seems reasonable to view Sacks as referring here to relevant sensations. Huemer also mentions, however, the case of “super blindsight” (Huemer 2001, 89, fn. 30), while Tucker (2010) focuses upon cases of actual blindsight: Subjects who have a damaged visual cortex often emphatically report that they cannot see anything within a certain region of their visual field. Nonetheless, such subjects often show remarkable sensitivity (though less than properly functioning humans) to such things as motion, the orientation of objects, and the wavelength of light within their reported “blind spot.” (530)

Huemer’s Phenomenal Conservatism

313

Tucker takes this as grounds for holding that people can have seemings that are unaccompanied by any visual imagery, by any non-cognitive qualia. His line of thought appears to be that, on the one hand, if one asks such a person about the nature of an object in a “blindspot” part of his or her visual field, that person will say that he or she has no information at all, so that the person has no belief about the object in question. On the other hand, if the person is asked to make a guess concerning the relevant property, he or she will often be right. Tucker proposes, accordingly, the following explanation: The subjects’ “blind spots” are regions in their visual fields that lack visual imagery. Nonetheless, the mechanisms that produce seemings function well enough to provide information about the region of the environment that corresponds to the subjects’ blindspots. (530)

What is one to say about this line of reasoning? The problem is that guessing is an action, and actions are to be explained in terms of beliefs and conative states, such as desires. This type of explanation requires, however, that the idea of belief, rather than being interpreted in the everyday sense, in which believing that p entails not believing that not p, is interpreted in the technical sense in which belief admits of degrees ranging from complete subjective certainty that p is false through to complete subjective certainty that p is true, since one cannot explain the actions of people unless one has a concept of belief that admits of degrees. But given the idea of degrees of assent, ranging from zero to one, the blindsight person who is asked to guess whether there is, say, a round object or a square object in front of him or her, and who guesses that there is a round object, although he or she does not believe, in the everyday sense of “believe,” that there is a round object there, must assent more strongly to the proposition that there is a round object than to the proposition that there is a square object. But, then, given that one must postulate relevant degrees of assent to explain the person’s making one guess rather than another, there is no longer any work to be done by seemings, and so no justification for postulating such mental states. The conclusion here, in short, is that if one considers cases, such as that of blindsight, where a person is not enjoying a relevant mental state that involves non-cognitive qualia, but where the person is nonetheless acquiring information, the fact that the person does not believe the relevant proposition in the everyday sense of “believe” is no ground for postulating the existence of a fundamentally different type of mental state—a seeming. One need merely replace the everyday concept of belief by the broader notion of degrees of belief or degrees of assent, a notion that is perfectly clear, and that one needs if one is to explain any actions.

314

Phenomenal Conservatism

3.2. the argument The upshot is the following argument against the existence of seemings, as conceived of by Huemer: (1) If seemings, as conceived of by Huemer, existed, there should be cases where such seemings exist, but where they are not accompanied by non-cognitive qualia. (2) If there were cases where such seemings, unaccompanied by noncognitive qualia, existed, then there should be cases where one would be justified in believing that such seemings, unaccompanied by non-cognitive qualia, existed. (3) When one considers any candidate for such a case—for example, that of blindsight—it emerges that there is no sound justification for postulating, in such cases, the existence of seemings, as conceived of by Huemer. Therefore (C) Seemings, as conceived of by Huemer, do not exist.

4. Appearances Rather than Seemings? In attempting to get clear about Huemer’s concept of a seeming, it occurred to me that it might very well be helpful to consider what other philosophers in the area of epistemology have had to say about the concept of seeming, to see if there was an interpretation that would enable one to get Huemer’s concept into sharp focus. The outcome of that exercise was, however, rather striking, for an examination of over fifty books in epistemology that I had at hand— going back to some books by Bertrand Russell—resulted in only one in which the index of the book contained the word “seems.” That was G.  E. Moore’s Philosophical Studies (1922), but it did not prove helpful. If Huemer is right in thinking that the basic principle concerning the justification of beliefs involves some concept of seeming, the dearth of references to that concept in other epistemology books is quite remarkable. In the previous section, however, we saw that when one attempts to introduce a concept of seeming according to which seemings involve assertive mental representations that are distinct from beliefs, the natural way of explaining what an assertive mental representation is leaves no room for assertive mental representations distinct from beliefs. We also saw that when one considers cases where perceptual information is acquired in a way not involving non-cognitive qualia, everything can be explained in terms of degrees of assent:  there is no need for the fundamentally different propositional-attitude type of mental state that seemings are supposed to be.

Huemer’s Phenomenal Conservatism

315

4.1. appearances At this point, I want to consider the idea that it may be helpful to shift from talking about seemings to talking about appearances, for while Huemer does not use the terms appears or appearances in stating the Principle of Phenomenal Conservatism, he does use those terms in the subsequent discussions. Thus, for example, in Skepticism and the Veil of Perception, he says, “According to Phenomenal Conservatism, the epistemological default position is to accept things as they appear. The appearances are presumed true, until proven false” (100). Then, in his article “Compassionate Phenomenal Conservatism,” references to appearances and how things appear are much more frequent, where they tend to dominate talk about how things seem. The idea, accordingly, is to consider discussions of the term appears—a term that, in contrast to the term seems, has been carefully discussed by a number of philosophers. In addition, there are terms that are closely related to the term appears, such as looks, that have also been carefully discussed. Among the best-known discussions of the term appears are those by Roderick Chisholm in Perception: A Philosophical Study (1957), and in Theory of Knowledge (1966 and 1977). The view that Chisholm advances is that there are three main uses of the term appears. First, there is what he refers to as the “epistemic use,” and which he suggests is connected with such things as inclinations to believe, and evidence (Chisholm 1957, 43–44). Next, there is what he calls the “comparative use,” where one is describing the qualitative nature of one’s experience, but doing so indirectly, by referring to the type of physical situation that will, under normal conditions, give rise to that type of experience (44–48). Finally, there is what Chisholm refers to as the “noncomparative use” of “appears,” in which one describes an experience not by referring to situations involving external physical objects with certain properties that will give rise to experiences of that sort, but, instead, in what one might describe as a rigid-designator fashion, in which one refers directly, rather than indirectly, to the intrinsic quality of the experience itself (50–52). Or consider Frank Jackson, who, in his book Perception, discusses the term looks in Chapter 2—“Three uses of ‘looks.’ ” Jackson’s discussion there is quite similar to Chisholm’s discussion of “appears,” to which Jackson refers, since Jackson holds, “it is plausible to distinguish three uses of ‘look’ (and ‘appear’) which I will call epistemic, comparative, and phenomenal” (Jackson 1977, 30). Huemer disagrees with the claim that the term appears is ever used in the non-comparative way that Chisholm describes, or in the phenomenal way that Jackson claims that “looks” and “appears” can be used. For, on the one hand, both Chisholm and Jackson hold that when one says that something appears green to one, in the non-comparative or phenomenal sense, one is simply referring directly to a qualitative property. (In Jackson’s case, it turns out that the qualitative property of greenness is a property of sense data, whereas in

316

Phenomenal Conservatism

Chisholm’s case, it is a property that characterizes how one is being appeared to.) Huemer, on the other hand, either does not think that “appears” refers to qualitative properties at all, or else he does not think that “appears” refers only to qualitative properties. For with regard to the use of the word “appears,” Huemer (2001, 90, fn 39)  says, “I think that ‘appear’ words always refer to apprehensions and, in some manner, report their contents”—where an apprehension, for Huemer, is simply “an assertive mental representation” (2001, 53). Consequently, when one says, for example, that a tomato appears green, one is, according to Huemer, in addition to referring to the tomato, referring to an assertive mental representation and reporting its content. Now it may be—I do not know whether Huemer has explicitly said anything about this—that Huemer thinks qualia can be assertive mental representations, and if that is his view, then in referring to assertive mental representations one may in fact sometimes also be referring to qualia. I am not convinced that such an identity claim is plausible if one holds, as Huemer does, that there can be assertive mental representations that are not accompanied by qualia. Moreover, the identity claim does not itself entail, for example, that the proposition that the table appears green involves the concept of qualitative greenness: the term green in “appears green” may be associated with some concept of physical greenness rather than with the concept of qualitative greenness. Fortunately, it does not matter in the present context exactly what Huemer’s view is on this matter. What is crucial is simply that when Huemer talks about appearances, he is either not referring at all to things such as spatial arrangements of qualitative properties that make up visual fields, or else he is not referring simply to such things: he is referring instead, or else as well, to assertive mental representations. This is something that it is very important to keep in mind, and also something that it is very easy not to notice at all.

4.2. an epistemic account of the concept of appearances? We now have four concepts of appearances on the table, namely, Huemer’s, along with, on the one hand, the purely phenomenal concepts, both comparative and non-comparative, mentioned by Chisholm and Jackson, and, on the other hand, some sort of epistemic concept, also mentioned by both Chisholm and Jackson. Will any of these accounts enable us to formulate a clear Principle of Phenomenal Conservatism? Huemer’s own account of the concept of appearances, regardless of exactly how it is to be interpreted, does not make it to the starting blocks, since according to that account, whatever else “appear” words do, they always refer to apprehensions, and given that an apprehension is, for Huemer, simply “an assertive mental representation” (2001, 54), this means that Huemer’s own account immediately leads back to the idea already examined and found wanting.

Huemer’s Phenomenal Conservatism

317

What about the other three types of interpretations? An interpretation according to which appearances are simply a matter of non-cognitive qualia leads one to a radically different justificatory view, for at the heart of Huemer’s approach is the view that justification rests at bottom upon certain propositional-attitude states, and non-cognitive qualia do not, as such, involve propositional attitudes. So the comparative and non-comparative concepts of appearing advanced by Chisholm and Jackson cannot be employed. This leaves us with the idea of what both Chisholm and Jackson refer to as the epistemic use of “appears.” So let us consider how one might attempt to capture that concept, and whether it could be used in formulating a Principle of Phenomenal Conservatism. Chisholm (1957, 43–44), in his discussion of the epistemic sense of appearing, brings in both the idea of inclinations to believe, and the idea of evidence, whereas Jackson (1977, 30) refers only to the idea of evidence. In the present context, the details of the account are not crucial, so I shall go with an account that is closer to Jackson’s, but that diverges from his in incorporating the idea of awareness: It appears epistemically to S that p = def. S is in some state T such that (1)  S is aware of being in state T, and (2)  S believes that state T is evidentially relevant to the proposition that p.

So defined, this state, of its appearing epistemically to S that p, is a propositional-attitude state, and one that is distinct from the state of S’s believing that p. Moreover, it is compatible with S’s not believing that p, since it may be that while state T is evidentially relevant to the proposition that p, it need not make it the case that p is more likely to be true than not p, and so it need not make it reasonable for S to believe that p. But can one set out a non-metaphorical account of the idea of an assertive mental representation such that its appearing epistemically to S that p involves an assertive mental representation of p? It would seem that one can, since one can say that a mental state involves an assertive mental representation of a proposition p if it is a mental state that allows one to make a sincere assertion concerning the probability that p is true. Does this epistemic concept of appearances enable one to formulate a viable Principle of Phenomenal Conservatism? It seems to me doubtful that it does, since given such a formulation, S would be justified in believing that p in cases where, though S believed that the relevant state T was evidentially relevant to the proposition that p, that was not in fact the case, and S was not justified in believing that it was. But surely a belief cannot be justified on the basis of unjustified beliefs about the evidential relevance of some state of affairs to that belief. If this is right, the situation is as follows. On the one hand, Huemer’s account of appearances in terms of assertive mental representations involves, as we

318

Phenomenal Conservatism

have seen, a metaphor that has not been cashed out in literal terms, and where the prospects of doing so in a way that will enable one to justify the claim that there are mental states that involve assertive mental representations other than beliefs—broadly construed so as to include degrees of assent—seems bleak. On the other hand, none of the three ways that other philosophers have proposed for understanding talk about appearances—namely, the comparative, the non-comparative, and the epistemic—can be used, it would seem, to formulate a satisfactory Principle of Phenomenal Conservatism. The conclusion, accordingly, is that we do not presently have any satisfactory account of Huemer’s concept of a seeming.

5. The Justificatory Permissiveness of the Principle of Phenomenal Conservatism, and the Source of that Permissiveness 5.1. an extremely permissive principle of justification Consider, first of all, Huemer’s (2001, 99) original formulation of the Principle of Phenomenal Conservatism in Skepticism and the Veil of Perception: PC

If it seems to S as if P, then S thereby has at least prima facie justification for believing that P.

If this principle is correct, beliefs whose justification is highly controversial will turn out to be easily justified, while beliefs that, according to other approaches to epistemology, have no rational basis at all, will turn out to be fully justified.

5.1.1. Philosophy Made Easy The types of seemings that Huemer lists—namely, perceptual, memory-related, intellectual, and introspective—are related to fundamental areas of epistemology. But this immediately raises the question about seemings in other areas. It seems to most people, for example, that there are other human minds. According to the Principle of Phenomenal Conservatism, then, that belief is prima facie credible, and therefore, if there are no defeaters, such people have a non-inferentially justified belief that there are other human minds. Similarly, it seems to most people that the regularities that have obtained up until now will obtain in the future. So, given the Principle of Phenomenal Conservatism, that belief is prima facie credible, and therefore, if there are no defeaters, one has a non-inferentially justified belief that inductive inference is sound. One consequence of the original Principle of Phenomenal Conservatism seems to be, then, that direct realism conquers all in epistemology. Consider any belief whose justification has proven puzzling but that one would like to show is justified. A solution to this task is readily at hand, since you need

Huemer’s Phenomenal Conservatism

319

merely ask yourself whether it seems to you as if the proposition in question is true. Does it seem to you as if there are external, mind-independent objects? Does it seem to you as if the past exists? Does it seem to you as if there are other human minds? Does it seem to you as if there is mind/body interaction? Does it seem to you as if you have contra-causal freedom? Does it seem to you as if past regularities will obtain in the future? Does it seem to you as if there are objective values? Does it seem to you as if there are other possible worlds—perhaps concrete ones, in the David Lewis style? Then, if the Principle of Phenomenal Conservatism is true, and if it seems to you as if these things are so, all may be well:  provided that there are no defeaters, you are non-inferentially justified in believing that these things are so. So it would seem, then, that David Lewis, for example, was non-inferentially justified in believing that there were concrete possible worlds, and if challenged by a skeptic to justify that belief, he need merely have appealed to the Principle of Phenomenal Conservatism, and have pointed out that he knew of no defeaters for the belief in question.

5.1.2. Dangerous Beliefs that Turn Out to Be Justified Suppose that, while Sue does not claim to have any religious experiences, it seems to Sue as if God exists, and, moreover, that Sue believes that God exists. According to the original Principle of Phenomenal Conservatism, Sue’s belief that God exists is prima facie justified, and if Sue has no defeaters for that belief, then the belief that God exists is non-inferentially justified for Sue. There is no need for Sue to appeal to the purported occurrence of miracles or to struggle with other arguments for the existence of God that may be open to serious objections. As long as it seems to her that God exists, and she has no defeaters for that belief, all is well. Now someone who is attracted to the Principle of Phenomenal Conservatism may say that this situation is not worrying, since there are defeaters for the belief in the existence of God—such as, for example, some version of the argument from evil. So, if some version of the argument from evil is sound, Sue’s belief can be shown to be irrational. Consider then, instead, Tony, a sometime atheist who has recently become a deist. The Tony whom I have in mind has changed his view, not because of consideration of some argument, but simply because it now seems to him that Nod exists—where “Nod” refers to a being who is like God except that, rather than being morally good, he is completely indifferent to the existence of good and evil. The question now is whether there are any defeaters for Tony’s belief that there is an omnipotent and omniscient deity, but one who is indifferent to good and evil. Given Nod’s indifference to good and evil, it will be rather more difficult to find defeaters, though one might try running a probabilistic argument, based on the idea that Nod is one of a very large range of possible omnipotent and

320

Phenomenal Conservatism

omniscient beings, at most one of which can be actual. But even if such an argument is sound, Tony can simply jettison the property of omnipotence, thereby shifting to a belief that would escape a potential defeater of that sort. In any case, the basic point, just as in Sue’s case, is that there is no burden of proof upon Tony to provide evidence in support of his belief. If it seems to Tony that Nod exists, and if he does not have any defeaters for his belief, then his belief is a perfectly rational one. This seems to me unsatisfactory. But perhaps it is not very troubling, since Tony’s Nod sounds harmless enough, as does, perhaps, belief in his existence. But imagine another person, Jim, who believes in Jod, a jealous deity who wants those who believe in him to kill the evildoers who worship false gods, or none at all, and who will reward only those people who do this. Neither Jim nor those who share his belief claim to have had experiences of Jod. But it certainly seems to Jim, and very strongly indeed, and also to those who share his belief, that Jod exists, and they are not aware of any defeaters of this belief, so if the Principle of Phenomenal Conservatism is right, the belief in question is surely justified. Moreover, it may very well also seem to them that they should certainly carry out the will of Jod by killing the evildoers, and the original Principle of Phenomenal Conservatism will then entail that that belief is justified, in the absence of defeaters. Are there any defeaters for Jim’s belief? Here, as in the case of Nod, the prospects do not look that promising. Given that it is not being claimed either that Jod is perfectly good, or that he is omnipotent, neither the argument from evil nor the aforementioned probabilistic argument will supply one with a defeater. In the case of belief in Jod’s existence, as in the case of belief in Nod’s existence, we have a belief that appears to turn out to be justified given Huemer’s original Principle of Phenomenal Conservatism. But in contrast to the case of belief in Nod’s existence, we also have a belief that is highly dangerous, because it makes it reasonable for Jim to believe that it is strongly in his self-interest to attempt to kill unbelievers. The case of belief in Jod, moreover, is not that remote from beliefs that one actually finds in the world, since various religions either now hold, or have held, either that heretics, or apostates, or both, should be put to death. It is true, of course, that religious believers generally want to claim that the deity they worship is good—indeed, perfectly good—and that exposes the believer to potentially powerful defeaters. But some believers are willing either to jettison outright the claim that the deity in question is good or else to maintain that their deity, while not good as judged by merely human standards, is good as judged by correct, non-human standards, and any believer who does either of these things makes it much harder to find defeaters for the belief in question. Abandon the view that beliefs can be justified on the basis of how things seem, however, and adopt an evidential approach, and the epistemic situation is very different. For it will not be the case, for example, that the best

Huemer’s Phenomenal Conservatism

321

explanations of Jim’s seemings, with respect to the existence of Jod, is that things are as they seem to Jim to be.

5.1.3. Disagreement and Defeaters? One way of responding to these permissiveness objections is to appeal to some principle such as the following: If it seems to S that p, then to the extent that it also seems to S that it seems to another person T, and equally strongly, that not p, S has a defeater that undercuts S’s justification for believing that p.

I think that this principle is plausible, and important. Thus, however strongly it seems to some Christians that God is a trinity, it surely seems to some Muslims, equally strongly, that God is a single person, and it may also seem to a given Christian that the latter is the case. Similarly, in the case of some philosophical beliefs, such as David Lewis’s belief in concrete possible worlds, there will be opposed seemings that are at least as strong. But in other cases, such as a belief in other minds, or in induction’s being justified, this may not be so. Would this principle help in the case of Jim? Not in general, I would think. Jim might have no knowledge of how things seem to others, or he might know, but it might seem to him that his own very strong seemings carry more weight than the probably much more tentative seemings of others on this matter. If this is right, the conclusion is that while the claim that opposing seemings provide defeaters is both plausible, and important, this idea can provide at most a partial answer to the permissiveness problem.

5.1.4. The Modified Principle of Phenomenal Conservatism How do things stand given the modified version of the Principle of Phenomenal Conservatism that Huemer sets out and defends in “Compassionate Phenomenal Conservatism”? It will not then follow that a belief that p is justified for S if it seems to S that p, and S has no defeaters. But the modified principle will provide a start toward justification, and in the case of religious beliefs, it seems plausible that coherence with other beliefs that the person in question has, and that are, if the modified Principle of Phenomenal Conservatism is right, also justified to some extent, may very well lift the set of beliefs in question up into the range where they are justified. The situation is less clear in the case of philosophical beliefs, where perhaps there is less scope for coherence to play a role. If so, it may be that, given the modified Principle of Phenomenal Conservatism, the road to justified philosophical beliefs is, near enough, as difficult as it has always seemed to be. But given Huemer’s acceptance of direct realism with regard to the external world, it seems doubtful that that is his view. In any case, the basic point is that once one shifts to Huemer’s more recent and more modest formulation of the Principle of Phenomenal Conservatism, more discussion is

322

Phenomenal Conservatism

needed to determine whether it is or is not relatively easy to justify philosophical claims.

5.2. the sources of the permissiveness The Principle of Phenomenal Conservatism, even in its more modest form, allows far too many beliefs to be justified. What is the source of that permissiveness? One answer, advanced by Peter Markie (2005, 356–57; 2006, 119–20), is that the source lies in the fact that seemings are treated as grounds for noninferential justification regardless of how they are caused. But here I agree with Chris Tucker, who argues that how a seeming is caused is irrelevant to whether the seeming has epistemic weight. Among other things, if one were to modify Huemer’s Principle of Phenomenal Conservatism by restricting seemings to ones caused in an “appropriate” way, one would have a principle that might be relevant with respect to warrant and knowledge, but that could no longer answer the internalist’s question concerning the ground of deontologically justified beliefs. I want to suggest two very different sources of the permissiveness of Huemer’s Principle of Phenomenal Conservatism, both of which could be eliminated without changing the principle to one that is concerned with warrant and knowledge, rather than with justification. The first source is that Huemer’s Principle of Phenomenal Conservatism, in both its original and modified forms, fails to distinguish between what can be referred to as basic seemings and derived seemings, and, thereby, mistakenly assigns justificatory weight to both. The second source is that the Principle of Phenomenal Conservatism, in both its original and modified forms, fails to distinguish between seemings that are associated with non-cognitive qualia, and seemings that are not, and, as a result, mistakenly treats both as having the same epistemological significance.

5.2.1. Basic Seemings versus Derived Seemings The idea here is to draw a distinction between basic seemings and derived seemings: (1) Its seeming to S that p is a derived seeming if and only if there is some proposition q such that it seems to S that q, where that causes it to seem to S that p, by a process internal to S, and where, otherwise, it would not seem to S that p. (2) Its seeming to S that p is a basic seeming if and only if the seeming in question is not a derived seeming. The contention is now that, in failing to draw a distinction between basic seemings and derived seemings, the Principle of Phenomenal Conservatism, in both its original and modified forms, licenses “double counting,” since if

Huemer’s Phenomenal Conservatism

323

its seeming to S that q is a derived seeming based upon its seeming to S that p, then the Principle of Phenomenal Conservatism generates two sources of justification for the belief that q—and sources that are not independent of one another—since, on the one hand, S can appeal to the fact that it seems to him or her that q, thereby supporting the belief that q directly, and, on the other hand, S can also appeal to the fact that it seems to him or her that p, and combine that with its seeming to him or her that p supports q to generate additional support for q. One result of this is that a derived seeming will typically enjoy a higher level of justification than the basic seeming on which it rests. Second, since its seeming to S that q can be a derived seeming that is caused by its seeming to S that p even if S is not justified in believing that p supports q, the Principle of Phenomenal Conservatism, in both its original and modified forms, is granting justificatory weight to seemings that are caused by other, more basic seemings, but that may not be justified in any way by those seemings. Any principle that does this is surely mistaken. The upshot is that no justificatory weight should be assigned to derived seemings and, as a consequence, and at the very least, the Principle of Phenomenal Conservatism should be jettisoned in favor of a Restricted Principle of Phenomenal Conservatism (RPC), such as the following, which assigns justificatory weight only to basic seemings: RPC

If it seems to S that p and, in addition, its seeming to S that p is a basic seeming, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p.

5.2.2. Seemings, Experiences, and Non-Cognitive Qualia A second source of the justificatory permissiveness of the Principle of Phenomenal Conservatism, in both its original and modified forms, is that, as regards experiences that involve seemings, the principle fails to distinguish between experiences that involve non-cognitive qualia, or images, and experiences that do not. But this distinction is crucial, since, as we saw in section 2.3, in the blindsight case, where there are no relevant non-cognitive qualia or images, there is no justification for postulating the existence of a seeming that is anything other than a belief, where belief is understood in the broad sense of some degree of assent ranging from complete rejection of a proposition to complete acceptance. The upshot is this. Suppose it turns out, contrary to what I  have argued, that one can make sense of the idea of assertive mental representations that are distinct from beliefs in the broad sense, and corresponding thoughts. Still, the fact that in the case of experiences that do not involve non-cognitive qualia there is no reason for believing that such assertive mental representations are present means that, just as in the case of the distinction between basic and derived seemings, one should reformulate the Principle of Phenomenal

324

Phenomenal Conservatism

Conservatism slightly, this time by building in a restriction to cases where the experiences in question do involve non-cognitive qualia. For otherwise, in cases where no non-cognitive qualia are present, one will be violating the surely correct constraint that neither a belief that p in the ordinary sense of “belief,” nor a degree of assent to p that is less than what is necessary to have a belief in the ordinary sense of “belief,” should contribute to p’s being justified. Such a change would eliminate certain objections to a Principle of Phenomenal Conservatism. In particular, the cases of God, Nod, and Jod, mentioned earlier, would no longer be cases where the relevant beliefs were justified according to the relevant principle, since the seemings in questions were, by hypothesis, not accompanied by experiences involving non-cognitive qualia.

6. A Quick Corollary: Denial of the Principle of Phenomenal Conservatism Is not Self-Defeating In Skepticism and the Veil of Perception, Huemer (2001, 101)  claims, “[A]ny attempt to deny the Principle of Phenomenal Conservatism will be selfdefeating, for all thought and reasoning presupposes the principle in a certain sense.” He then defends this contention at greater length in “Compassionate Phenomenal Conservatism,” where he both sets out the self-defeat argument in greater detail and then considers two possible responses to it. The first response involves the idea that one can advance a restricted version of the Principle of Phenomenal Conservatism—call it PC*—that will not have the consequence that PC* itself is not justified. The second response holds that the correct foundational principle is one that asserts that what makes a belief that p non-inferentially justified is that the belief that p is appropriately related to some fact that makes p true, and with which one is directly acquainted. Both responses seem to me correct. Here, however, I shall focus only on the first. Consider, then, the Restricted Principle of Phenomenal Conservatism set out at the end of section 5.2.1: RPC

If it seems to S that p and, in addition, its seeming to S that p is a basic seeming, then, in the absence of defeaters, S thereby has at least some degree of justification for believing that p.

My claim is that this principle, in contrast to the alternative principles that Huemer (2007, 42–44) considers in his discussion, not only avoids self-defeat, but also escapes the sort of objection that Huemer directs against the alternative restricted principles that he considers. Notice, first of all, that there is no reason that it cannot seem to one that RPC is true, nor why such a seeming cannot be a basic seeming. Accepting RPC, while denying Huemer’s PC, is not, therefore, self-defeating.

Huemer’s Phenomenal Conservatism

325

Second, unlike one of the alternatives that Huemer (2007, 43)  considers, RPC is not restricted in scope. Any proposition could, given the appropriate seeming, be justified. Third, unlike another alternative that Huemer (2007, 44)  considers, RPC does not appeal to any distinction that is epistemologically irrelevant. In addition, the Restricted Principle of Phenomenal Conservatism is clearly preferable to both versions of Huemer’s Principle of Phenomenal Conservatism in a very important way, namely, in that it does not allow double counting. Suppose that it seems to S that q, and that this seeming is a derived seeming, based on its seeming to S that p. Regardless of whether one accepts PC or RPC, if it seems to S that p provides evidential support for q, then S has some justification for believing that q in virtue of the fact that, in the absence of defeaters, S has some justification for believing that p, and also for believing that p supports q, which together means that S is to some extent justified in believing that q. But if one accepts PC, then its seeming to S that q means that, in the absence of defeaters, S has additional justification for believing that q, even though its seeming to S that q rests entirely on its seeming to S that p, and the support that p provides for q has already been taken into account. By contrast, RPC does not allow such double counting. Finally, according to a foundational approach to justification, there can be beliefs that are justified, but not non-inferentially justified, and the justification of those inferentially justified beliefs must be traceable back to one or more non-inferentially justified beliefs. Given such a structure, there should surely be a principle that deals specifically with the justification of non-inferentially justified beliefs and that does not also cover beliefs that are only inferentially justified. Huemer’s Principle of Phenomenal Conservatism is not such a principle, whereas RPC is. So this is another reason for holding that RPC is superior to Huemer’s PC principle.

7. The Superiority of Direct Awareness and Direct Acquaintance Approaches to Justification One could avoid the problem of justificatory permissiveness by moving on to a Doubly Restricted Principle of Phenomenal Conservatism that, first of all, distinguished both between basic seemings and derived seemings, and also between seemings that are aspects of experiences that involve non-cognitive qualia and those that are not, and, second, that assigned justificatory weight only to seemings that were both basic and aspects of experiences that involve non-cognitive qualia. Should one, then, go that route? The answer is that one should not. The reason is that when one focuses upon seemings that are aspects of experiences that involve non-cognitive qualia, it becomes clear that

326

Phenomenal Conservatism

what really lies at the bottom of the justification of beliefs in such cases is direct awareness of relevant states of affairs, and not seemings. In more detail, the argument is as follows. Huemer’s accounts of seemings and appearances both turn out to involve the idea of assertive mental representations. But Huemer has provided no account of how the “assertive” metaphor is to be cashed out, and we have seen that a natural way of doing so does not leave room for seemings that are distinct from beliefs, understood in the broad sense. Then, when one considers alternative interpretations of talk about appearances, the only possibility according to which appearances involved propositional attitudes was something along the lines of the following, epistemic (or evidential) interpretation: It appears epistemically to S that p = def. S is in some state T such that (1)  S is aware of being in state T, and (2)  S believes that state T is evidentially relevant to the proposition that p.

But given such an epistemic notion of appearing (or seeming), since S’s awareness of being in state T suffices to justify S in believing that he or she is in state T, and since S will also have the trivial belief that being in state T is evidentially relevant to the proposition that he or she is in that state, it follows that it will appear (or seem) to S that he or she is in state T. Let us suppose, moreover, for simplicity, that that is a basic seeming. Then, on the one hand, if p is the proposition that S is in state T, its seeming to S that p will be a basic seeming, but then the belief that p is justified directly via S’s awareness of being in state T. On the other hand, if p is some other proposition, then its seeming to S that p will be a derived seeming, and then, as we have seen, no independent justificatory weight should be assigned to its seeming to S that p. A principle according to which all justification is ultimately based upon direct acquaintance, or direct awareness, is therefore all that is needed. No principle that involves any concept of seeming is required.

8. Summing Up The main conclusions for which I have argued are as follows. First, Huemer’s modified Principle of Phenomenal Conservatism is much more modest than his original principle. Second, Huemer’s account of seemings and appearances involves the crucial idea of assertive mental representations, where the term assertive is a metaphor and where Huemer has offered no account of how the metaphor is to be cashed out. Third, one very natural way of translating the metaphor into literal terms leads to the conclusion that there are no assertive mental representations distinct from beliefs, broadly construed. Fourth, if one focuses on experiences that involve no non-cognitive qualia, there is no

Huemer’s Phenomenal Conservatism

327

justification for believing that such experiences involve any assertive mental representations beyond beliefs or degrees of assent. Fifth, that conclusion is in turn the basis of an argument for the more general conclusion that there are no assertive mental representations that are distinct from beliefs or degrees of assent. Sixth, if one considers alternative accounts of appearances, the only type of account according to which appearances are propositional-attitude states is some sort of epistemic (or evidential) account, and appearances, so understood, will not serve Huemer’s purposes. Seventh, Huemer’s Principle of Phenomenal Conservatism, in both its original and modified form, suffers from justificatory permissiveness. Eighth, the source of that permissiveness is twofold: a failure to distinguish between basic seemings and derived seemings, and a failure to distinguish between seemings that are aspects of experiences involving non-cognitive qualia and those that are not. Ninth, a denial of Huemer’s Principle of Phenomenal Conservatism does not lead to self-defeat, since one can draw an epistemologically relevant, and crucial, distinction between basic seemings and derived seemings, and then adopt a Restricted Principle of Phenomenal Conservatism that assigns justificatory weight only to basic seemings. Finally, reflection upon an epistemic account of seemings, together with the distinction between basic seemings and derived seemings, supports the conclusion that an approach to foundational justification based on the ideas of direct awareness, or direct acquaintance, is superior to one based upon any Principle of Phenomenal Conservatism, be it restricted or unrestricted.

Acknowledgment I am grateful to Chris Tucker for very detailed and helpful comments on an earlier draft.

References Roderick Chisholm. 1977. Theory of Knowledge, 2nd ed. Englewood Cliffs, NJ: Prentice-Hall. _____. 1966. Theory of Knowledge. Englewood Cliffs, NJ: Prentice-Hall. _____. 1957. Perceiving: A Philosophical Study. Ithaca, NY: Cornell University Press. Frank Jackson. 1977. Perception. Cambridge: Cambridge University Press. Michael Huemer. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74(1): 30–55. _____. 2001. Skepticism and the Veil of Perception. Lanham, MD: Rowman and Littlefield. Oliver Sacks. 1985. The Man Who Mistook His Wife for a Hat. New York: Simon and Schuster. Chris Tucker. 2010. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45.

{ 15 }

Phenomenal Conservatism Über Alles Michael Huemer 1. What Are Seemings? 1.1. against analysis Michael Tooley is noticeably frustrated by my failure to analyze the notion of its seeming or appearing to one that P. One reason I have not tried to do so is that as far as I can tell, philosophical analysis has never succeeded. Despite the popularity of the school of linguistic analysis in twentieth-century philosophy, I  cannot name a single analysis of any philosophically interesting term that has not been refuted. There are interesting reasons for this lying in the nature of concepts, which I  will discuss in future work. Some may claim a few alleged successes (philosophers are especially likely to cite their own analyses), yet I think almost any philosopher will agree at least that the overwhelming majority of concepts have never been correctly analyzed, and the overwhelming majority of attempts at analysis have failed. Philosophers should take caution before treading upon a territory so littered with failed theories. But if I  offer no analysis of “seems,” won’t readers be unable to understand what I mean by the term, and hence unable to understand the principle of phenomenal conservatism? Fortunately, there is little cause for fear on this score. I know thousands of words of English. For hardly any of them has anyone ever told me what the word means, and even in the cases where someone explained a word’s meaning to me, almost none of the explanations—perhaps none at all—would stand up to philosophical standards for a correct analysis. Yet I seem to understand all these words well enough to get by. How is this possible? Most words are learned through the language learner’s observation and imitation of their usage. Of course, this presupposes that the learner has some familiarity with the parts or aspects of the world to which the word refers. If, as I claim, there is a class of conscious mental states that may be dubbed “seemings,” including states that occur during the normal operation of

Phenomenal Conservatism Über Alles

329

perception, memory, and intellectual reflection, then these states should be familiar to all normal individuals.1 I therefore need not provide a philosophical analysis of seemings; I need only say enough to draw readers’ attention to these familiar mental states. This can be done (pace Plato) through citing examples, as well as discussing some of the features of these mental states and how they differ from similar mental states (such as beliefs) with which they might be confused. Pace Tooley, metaphors are often a useful tool for directing an audience’s attention to features that are otherwise difficult to describe; I thus see no reason to jettison the metaphor of “assertive” mental representations. Have I  correctly construed the nature of seemings? Perhaps seemings lack some features I have ascribed to them, or possess some features I have denied them. Perhaps “seems” and “appears” do not really refer to a kind of propositional attitude. Or perhaps they refer to a propositional attitude that can be explained in terms of belief and dispositions to believe. These sorts of proposals may dramatically impact the plausibility or even intelligibility of PC. I therefore consider proposals of this sort below.

1.2. the disposition to believe A natural approach to analyzing seeming is to appeal to dispositions to believe. One might hold that its seeming to one that P is simply a matter of one’s being in a state such that one would believe that P, were there no other factors interfering with one’s forming such a belief. It is worth briefly reminding ourselves of one of the main reasons for rejecting that approach. This is the fact that one can be disposed to believe P for different sorts of reasons, other than its seeming to one that P. If I am disposed to believe in the afterlife because I want there to be an afterlife, this is quite different from my being disposed to believe in the afterlife because that seems true. The lesson is that appearances are only one sort of ground for the disposition to believe. This will be important to bear in mind as we consider alternative conceptions of appearances below.

1.3. one sense of “appear” Some authors find three uses of “appears” and “seems”: the phenomenal, comparative, and epistemic uses.2 I have proposed a single sense of these words, in which they refer to a kind of propositional attitude, a sort of mental state 1 I leave aside the interesting question of whether there are unconscious or merely dispositional seemings. 2 Chisholm 1957, ch 4; Jackson 1977, ch 2; Tooley, this volume, section 4; Bergmann, this volume, section 1.1; Brogaard, this volume, section 3 (though Brogaard does not characterize the three uses in the same way as the other authors). I shall simply assume herein that “seems” and “appears” are near enough equivalent that what goes for the one goes for the other.

330

Phenomenal Conservatism

representing the world as being a certain way. This type of mental state provides justification for believing its content and under normal circumstances disposes the subject to believe its content. Can this single underlying sense explain the three uses of “appears”? Begin with the phenomenal use, a paradigm of which is supposed to be “That apple appears red.” Here, “appears red” is said to refer to a certain sort of quale, a qualitative, intrinsic, introspectible property of a mental state, distinct from its intentional content. I disagree with this analysis. I think “That apple appears red” may be understood along the lines of “I have a certain sort of mental state representing that apple as red.” I do not deny the existence of qualia; I simply suggest that the expression “appears F” does not refer to a quale but rather to an intentional content. This helps to explain the vast range of cases in which “appears F” may be used. One may say an apple appears red, where it is plausible that one is experiencing a sensory quale. But one may also say the apple appears round, where it is more controversial that there is a particular quale one might be describing. One may say that a bridge appears unsafe, that a plan appears profitable, that Tooley appears frustrated, or that an argument appears correct. Are there in all of these cases qualia distinctive of just those cases in which an object appears the specified way? Indeed, for virtually any property F, it is intelligible to say “x appears F.” This is easily explained by the propositional-attitude account of “appear” words, for one may take an attitude to the proposition that x is F for virtually any feature F. But on the qualia account, we must either introduce distinct qualia corresponding in some sense to nearly every property that one can name, or else posit an ambiguity in the word “appear,” in cases in which the usages appear parallel—one could claim, e.g., that when I say two word usages “appear parallel,” I am using the epistemic sense of “appear” rather than its phenomenal sense. Of course, many words in English are in fact ambiguous; nevertheless, we should avoid positing ambiguities unnecessarily. In this case, the postulation does not seem necessary. What of the comparative use of “appear”? According to Chisholm (1957, 51), in the comparative sense, “appears red” “may be replaced by an expression of the form “[appears] the way red things [appear] under . . . conditions.”3 I have difficulty seeing the motivation for positing this sense of “appear.” Why can’t all ordinary uses of “x appears F” be interpreted along the lines I have suggested? Chisholm discusses one sort of case that might be thought to motivate postulation of ambiguity (48).4 A  person looks at a round tabletop from an oblique angle. In one sense, we can say that the tabletop appears elliptical. But

3 I have replaced “looks” with “appear” to match the context of my discussion. The ellipsis is in the original. 4 The case moves Chisholm to distinguish the epistemic from the comparative uses of “appear,” but it remains unclear why he distinguishes the comparative from the phenomenal (or “noncomparative,” in his terminology) uses, or why he does not regard the case as illustrating the phenomenal use.

Phenomenal Conservatism Über Alles

331

in another sense, of course, the tabletop appears not elliptical but round; the observer, for example, would not be tempted to say the tabletop was elliptical. Chisholm claims that the sense in which the surface “appears round” is the epistemic sense, whereas the sense in which it “appears elliptical” is the comparative sense—that is, the tabletop appears the way an elliptical object would appear.5 Must we posit an ambiguity in “appear” to explain why “The surface appears round” and “The surface appears elliptical” can be true simultaneously? Not necessarily. As an analogy, a person can desire A and desire B, where A and B are incompatible; rather than postulating an ambiguity in “desire,” we simply say that there can be conflicting desires. Similarly, rather than postulating an ambiguity in “appears,” we may say that a person can have conflicting appearances. Granted, the two appearances in Chisholm’s case are, in a sense, of different sorts. The elliptical appearance is, we might say, a more primitive sort of sensory appearance; the round appearance results from more and more sophisticated information processing in the brain. But we should not allow this to push us into postulating different senses of “appear.” There is no simple dividing line between primitive appearances and sophisticated appearances. Appearances come in many degrees of sophistication, with many different degrees of conscious, unconscious, and semi-conscious processing going into their production, and we do not want to postulate different senses of the word “appear” for all of these cases. Finally, what of the epistemic use of “appear”? In this sense, Chisholm tells us, “x appears to S to be so-and-so” and “x appears so-and-so to S” may be taken to imply that the subject S believes, or is inclined to believe, that x is so-and-so. And I  think that, in this same use, they may be also taken to imply that the subject S has adequate evidence for believing that x is so-and-so. (Chisholm 1957, 44)

If “imply” is read as denoting conversational implicature, then I  agree with Chisholm. Appearances are a source of defeasible epistemic justification. They also normally incline one to believe their content. Thus, when one reports an appearance, unless one indicates otherwise, this will normally be taken to imply that the person having the appearance is inclined to believe the content

5 Since the table appears precisely the way a round table normally appears when viewed from that angle, it remains unclear why we should not say that the table appears round in the comparative sense. Perhaps Chisholm would say it appears the way an elliptical table appears when viewed from above, though it is not entirely clear why “from above” should be considered the privileged viewing condition. It is also not exactly true that the table appears the way an elliptical table would appear from above; if one actually viewed an elliptical tabletop from above, and a round one from the side, one could typically tell the difference. But then, perhaps it is not exactly right in the first place that the table “appears elliptical”; the table only sort of appears elliptical.

332

Phenomenal Conservatism

of that appearance and has justification for so believing. To explain the use of “appear” words to convey these things does not require us to postulate a distinct sense of “appear” words in which they refer to a belief or inclination to believe.

1.4. are epistemic seemings beliefs? Berit Brogaard (this volume, section 3) contends that some seemings, which she terms “epistemic seemings,” are just beliefs. We agree that, if there are cases in which it seems to one that P solely in virtue of one’s believing that P, such seemings are not sources of epistemic justification.6 However, I am not persuaded that such cases exist. Brogaard characterizes epistemic seemings as those that “go away in the presence of a rebutting defeater if the agent is rational.” She then concludes that these seemings are actually beliefs. But why should we draw this conclusion? Why may not a non-belief state also disappear in the presence of a rebutting defeater? Some appearances, particularly low-level appearances such as perceptual experiences, are immune to counterevidence, but there may also be more sophisticated appearances that are capable of responding to evidence. If anything, one might consider such sophisticated appearances even better sources for justified beliefs than primitive appearances that are impervious to counterevidence. Brogaard gives an example in which the radio forecasts impending flooding in her area. As a result, it seems to her that evacuation would be prudent. If she then acquires evidence directly showing that evacuation would not be prudent, it will no longer seem to her that evacuation would be prudent. Brogaard claims that the seeming that evacuation would be prudent is actually just a belief that evacuation is probably prudent. I don’t think this is correct. Suppose that S is a skeptic about probability— according to S, no one knows the probability of anything, and there may not even be any such thing as a “probability.” Still, when S hears about impending flooding on the radio, it seems to her that evacuating would be prudent. A belief about probability is thus not necessary for the epistemic seeming. To see why the probabilistic belief also is not sufficient for an epistemic seeming, first recall the argument given in section 1.2 for distinguishing appearances from beliefs or dispositions to believe—namely, that there are different reasons that a person may be inclined to believe P, only some of which count as its seeming to them that P. If one accepts that argument, it may be 6 On the other hand, if there are cases in which one counts as believing that P (partly) in virtue of its seeming to one that P, these seemings can still provide epistemic justification. There may be cases in which subjects count as dispositionally believing that P in virtue of its seeming to them that P and their being disposed to endorse that appearance.

Phenomenal Conservatism Über Alles

333

applied equally to beliefs about probability. Thus, suppose I believe that there is probably life after death solely because I want this to be true. This would not count as its seeming to me that there is life after death.

1.5. qualia-free seemings Michael Tooley (this volume, section 3) suggests that seemings without noncognitive qualia (hereafter, I  use “qualia” for “non-cognitive qualia”) do not exist, and that this shows that seemings, as I conceive them, do not exist at all. Let us start by briefly reviewing some candidates for qualia-free seemings. I suggested that proprioception, the awareness of one’s body position, lacks sensory qualia (Huemer 2001, 67). Tooley points out that “sensations of muscular tension” often accompany proprioception. I think sensations of muscle or skin tension may be construed as distinct perceptions that sometimes accompany proprioception. There are many ways in which one’s body can be positioned without noticeable sensations of tension but in which one still seems aware of that positioning. Of course, the brain must be processing some information to arrive at one’s sense of how one’s body is positioned, but subjects are not aware of that information as a basis for their sense of body position, so it does not count as part of a proprioceptive experience. A second case is that of blindsight (see Tucker 2010). Tooley suggests that when the blindsight patient guesses what is present in a part of the visual field where he has no qualia, the patient has some (perhaps only slightly) elevated degree of belief that the thing he guesses to be present is present. According to Tooley, we need not posit a distinct appearance state, since we can explain the subject’s guess in terms of this slight degree of belief. But one might still posit appearances to explain why the subject has such an elevated degree of belief. It is natural to think that the subject has a slightly elevated degree of belief that, for example, there is a circle in the blind area because it (slightly) seems that way. A third type of case concerns non-perceptual appearances, including apparent factual memories (such as my seeming to remember that Napoleon was defeated in 1815) and intuitions (such as the intuition that for any property, there is a set of all and only the things possessing that property). It is open to debate whether there are qualia associated with these appearances. I shall not tarry further over the question of qualia-free seemings, however, because I  do not see its relevance to the truth of phenomenal conservatism. Tooley suggests, “If seemings, as conceived of by Huemer, existed, there should be cases where such seemings exist, but where they are not accompanied by noncognitive qualia.” This is the first premise in an argument leading to the conclusion that “Seemings, as conceived of by Huemer, do not exist.” But what is the motivation for this premise? Perhaps seemings are always accompanied by qualia. If so, how would this show that the seemings I purport to refer to don’t exist?

334

Phenomenal Conservatism

We might take a strong reading of “as conceived by Huemer,” such that for seemings as I conceive them to exist, I would have to be correct in every claim I make about seemings. But on this reading, the failure of “seemings as conceived by Huemer” to exist would be of little concern. A more natural reading, and one that renders Tooley’s conclusion more interesting, is that for “seemings as conceived by Huemer” to exist, I would have to be referring to something by my use of “seems” in the statement of phenomenal conservatism. If I  made sufficiently large errors about the nature of seemings, that is, if my understanding of them were sufficiently far from a correct understanding of any existing states, then one would have to say that my use of “seems” did not refer to anything. For example, if when it seemed to one that there was a cat on the balcony, one were not in any relevant mental state, or if the only relevant mental state were one with no intentionality, then it would be plausible to say that my use of “seems” failed to refer. But it is fairly clear that my putative error about whether these mental states are invariably accompanied by qualia does not constitute an error of such magnitude as to prevent me from referring to anything. Another possibility is that Tooley means to suggest that if seemings are always accompanied by qualia, then we need not believe there is any propositional attitude; we can take up Occam’s Razor and say that seemings are just qualia that lack propositional contents but that ground dispositions to adopt beliefs. If this is the idea, I will simply say that I find the claim far from obvious. I would need to hear much more to be convinced that non-representational qualia can do all the epistemological and semantic work that seemings as assertive propositional attitudes can do.

1.6. awareness of apparent evidence Both Michael Tooley and Earl Conee offer accounts of seeming in terms, roughly, of awareness of apparent evidence, thus: It appears epistemically to S that P =df S is in some state T such that (1) S is aware of being in state T, and (2)  S believes that state T is evidentially relevant to the proposition that P. (Tooley, this volume, section 4.2)7

In place of condition (2), Conee (this volume, section 2.1) offers that the subject should be inclined to regard T as displaying the truth of P. These accounts strike me as over-intellectualized, or over-introspectivized, insofar as they make a second-order mental state or disposition a necessary condition on one’s experiencing any appearances.

7

I assume that by “evidentially relevant to,” Tooley means “evidence for.”

Phenomenal Conservatism Über Alles

335

More important, this sort of account is vulnerable to an objection similar to the argument offered in section 1.2 for distinguishing appearances from dispositions to believe. We cannot regard an appearance that P as a mere disposition to believe that P, because, for example, a disposition to believe that P because one wants P to be true is not an appearance that P. When we move to a belief, or disposition to believe, that something counts as evidence for P, matters are no different. A person might be disposed to believe that T is evidence for P solely because he wants T to be evidence for P. This would not count as P’s seeming true to that person. Thus, suppose that because I desperately want to convince myself that theism is rational, I deceive myself into believing that any conscious experience is evidence for the existence of God. It doesn’t seem that way; it’s just what I want to believe. It is not plausible to infer that thereafter, whenever I have a headache, it seems to me that God exists. Conee quotes Tolhurst’s (1998, 298–99) statement to the effect that seemings “have the feel of truth, the feel of a state whose content reveals how things really are” (emphasis added). Conee (this volume, section 2.2) himself goes on to describe “an impression of being presented with something pertaining to” a proposition’s truth (emphasis added). How shall we understand this talk of “feelings” or “impressions”—what sort of mental state are we referring to? The answer fairly thrusts itself upon us: it is an appearance. That is, when it seems to S that P, it is not that S merely believes for some reason or other that his mental state T reveals the truth about P. It is that T seems to S to reveal the truth. But of course, at this point we see that we have failed to analyze seeming. The belief that one’s mental state evidentially supports P also is not necessary for it to seem to S that P. Take the case of an external-world skeptic who denies that we have any evidence for claims about the external world—David Hume, let us say. Hume (1975, section XII) did not believe that any experiences were evidence for any claim about the external world. Yet surely various sorts of external facts still appeared to him. Even the most committed skeptics do not deny that there seem to be external objects around us; they just claim that we are not justified in assuming that things are as they appear. Thus, in the depth of the skepticism of his Second Meditation, Descartes (1984, 19)  remarks, “I am now seeing light, hearing a noise, feeling heat. But I am asleep, so all this is false. Yet I certainly seem to see, to hear, and to be warmed. This cannot be false.”8 Conee asks us to imagine a case in which a subject’s skepticism is so deep that the subject is not even inclined to regard his perceptual experiences as displaying the truth of any contingent, external-world propositions. Somehow, however, the subject is supposed to retain the inclination to form

8 Emphasis in the original. I take it that the factivity of “see” and “hear” is what makes Descartes question that he sees and hears.

336

Phenomenal Conservatism

beliefs based on those experiences. (This is perhaps a slightly more extreme version of Hume’s situation.) In such a case, Conee says, it would not seem to the subject that anything was the case in the external world. But insofar as I can imagine this case, it strikes me, pace Conee, as a clear counterexample to Conee’s theory rather than a confirmation of it.9 Surely, various things would still seem to the subject to be happening in the external world. So the inclination to believe [T is evidence for P] has not helped us analyze seemings.10 Let us return to the original idea in this neighborhood, that when it seems to one that P, there is a disposition to accept P. This seems correct, at least in normal circumstances. But one is not disposed to accept P for just any reason; one is so disposed because of the character of the mental state T itself. A disposition requires a ground. Now, one can try to ground the disposition to accept P in a disposition to accept [T is evidence for P]. But this move does us no good, as we will then just have to inquire into the ground for the disposition to accept [T is evidence for P], and presumably, we will not explain this by introducing yet another disposition. On reflection, both the disposition to accept [P] and the disposition to accept [T is evidence for P] seem to be grounded in an intrinsic feature of T, rather than, say, in some other mental states or in some external states. But once we recognize such a feature of T, there is no need to identify seemings with doxastic dispositions or hold that a state counts as a seeming in virtue of the existence of certain doxastic dispositions. It is more natural simply to identify the seeming with T, and to hold that T counts as a seeming in virtue of having the very feature that also grounds the disposition to form certain beliefs. Moreover, regardless of whether that feature is the referent of “seems” in English, it would be plausible to hold that that same feature is also the source for epistemic justification. That is the view I am defending.

2. The Subject’s Perspective Objection Matthias Steup (Steup, this volume) argues that PC falls to the Subject’s Perspective Objection (SPO). I take the argument to be roughly this: (1) If PC, then whenever it seems to S that P in the absence of defeaters, S has justification for believing P.

9 A case of this sort occurred to me while reading Conee’s chapter, before I reached the point where Conee raises the case himself. Imagine my surprise when I  found Conee describing my counterexample, but citing it as evidence for his theory. 10 I use square brackets to convert sentences into terms denoting propositions; hence, “[φ]” means “the proposition that φ” (read the quotes here as quasi-quotation marks).

Phenomenal Conservatism Über Alles

337

(2) In some cases, it seems to S that P in the absence of defeaters, yet it is at best accidental from S’s perspective that S’s belief that P is true. (3) If it is at best accidental from S’s perspective that S’s belief that P is true, then S lacks justification for believing that P. Therefore, (C) PC is false. (From 1, 2, 3.)

I believe Bergmann (this volume, section 3.3) has shown that Steup’s Internalist Reliabilism is vulnerable to the same style of objection, if the SPO is persuasive to begin with. I find Bergmann’s argument regarding the dilemma for internalism in general well taken; I disagree only with how it is advertised. Bergmann advertises his argument as a problem for “internalism,” whereas I  view the argument simply as a refutation of the SPO. If one has another reason for being an internalist, then one avoids the problem.11 What might be wrong with the SPO? Briefly, advocates of the SPO require that as a precondition on having a justified belief, the subject have positive grounds for believing that his belief-forming method is reliable, that the particular belief in question is likely to be true, or something in this neighborhood. Any such requirement will generate an infinite regress, per Bergmann.12 Advocates of the SPO are right to maintain that a belief is unjustified if, from the subject’s perspective, it is at best accidentally true. However, it is best to interpret this as imposing only a negative condition on justified belief: if S believes or has grounds for believing [P is at best accidentally true], then S is not justified in believing P. This is consistent with PC, which allows the possibility of defeaters for the justification normally conferred by appearances. If one believes or has grounds for believing [P is at best accidentally true], then this will constitute such a defeater.13 Fortunately, this sort of negative condition on justification generates no regress.

11 See Huemer (2011a, 11–12), responding to DePoe (2011), but note that I define “internalism” differently than Bergmann. 12 It is also worth remembering Fumerton’s (1995, 81) rejection of access requirements on justification: “The idea that X constitutes one’s justification for believing P only if one’s awareness of X is added to X is equivalent to holding that X constitutes one’s justification for believing P only if it does not really constitute one’s justification for believing P.” 13 I have elsewhere defended the principle of “metacoherence” (Huemer 2011b), which holds that if one categorically believes that P, then one is committed, on reflection, to the view that one knows that P. I would interpret this as implying that actual doubts as to whether one knows P constitute defeaters for one’s belief that P. Importantly, the meta coherence norm is not meant to impose a precondition on justification for believing P.

338

Phenomenal Conservatism

3. Seemings and Inferential Justification 3.1. teasers about inference and inferential justification A number of objections to my views on phenomenal conservatism are related to inferential justification or indirect appearances (Tooley, this volume, section 5.2.1; McGrath, this volume, section 5; Markie, this volume, section 2). I aim to discuss inferential justification more fully in future work. To avoid spoilers, therefore, I can give only a few laconic hints here, just sufficient to render intelligible my replies to some objections raised by other authors in this volume. There are two kinds of appearances: ordinary (non-inferential) appearances, and inferential appearances. Inferential appearances occur during inference and represent that a conclusion must be true or is likely to be true in light of something else that one believes. An inferential appearance plays an essential role in the process of basing a belief on another belief, because it is what constitutes one’s seeing the premise as an adequate ground for the conclusion. In earlier work on phenomenal conservatism, I proposed simply that appearances confer justification in the absence of defeaters (Huemer 2007, 30).14 But in the case of inferential justification, a subject must normally possess not only an inferential appearance but also some justification for believing the premise of the inference, in order to have justification for believing the conclusion. This observation can be accommodated in either of two ways. One way is to maintain that a subject who lacks justification for a premise of an inference has a defeater for the conclusion (perhaps because whenever one bases a conclusion on an unjustified premise, one has justification for thinking that the premise is unjustified). Another way is to directly build the desired qualification into PC. That is, we can say that, in the case of an ordinary appearance that P, a subject has justification for believing P provided only that the subject lacks defeaters for P, whereas in the case of an inferential appearance that in light of E, P must be the case, the subject has justification for believing P provided that the subject also has justification for E and lacks defeaters for P. Either way, the requirement of justification for E is well motivated and not ad hoc, because the inferential appearance itself represents E as the relevant reason for accepting P.  That is, when one has an inferential appearance, it does not merely (unconditionally) seem to one that P; it seems to one that P given E. It is implicit in such an appearance that the relevant justification for P depends upon E.

14 Though I vaguely intended this statement to include inferential appearances, I did not specifically discuss inferential appearances or inferential justification.

Phenomenal Conservatism Über Alles

339

3.2. justification by fallacious inference My account of inferential justification makes possible cases in which a subject is inferentially justified in believing P on the basis of E, even though the inference from E to P is fallacious. This can happen because a fallacious inference may seem cogent to a subject, while the subject lacks grounds for doubting its cogency. Though some find this problematic (see Tooley, this volume, section 5.2.1), I view it as a correct prediction of the theory. Consider the Case of the Unfortunate Mathematician: S has just gone through what seems to be a genuine mathematical proof of P, starting from E as the sole premise. S is adequately justified in accepting E, he has checked the argument over carefully, and it is, as far as he can ascertain, a sound and rigorous proof of P.  Nevertheless, it happens that there is a subtle error in the argument, as a result of which the argument is neither valid nor cogent. What ought S to think? In the internalist sense—that is, in terms of what makes sense from the subject’s perspective—it seems to me that the unfortunate mathematician should believe P. Indeed, he would be irrational not to believe P at this point. (If he didn’t accept P, what could he say—“I don’t like proofs”?) But his evidence for P does not logically support P. So it is possible to be justified in believing P on the basis of evidence that does not objectively support it. It is an advantage of PC that it enables us to explain this otherwise odd result. The mathematician is justified in believing P because his apparent proof makes it seem that P must be correct, given his premises. I discussed this case in earlier work, where I  incorrectly explained it by claiming that the unfortunate mathematician is justified in believing P, not on the basis of E, but on the basis of the proposition that he has just proved P (Huemer 2002, 335–36). This is a mistake because it requires the mathematician, in order to have doxastic justification, to have inferred P from “I have just proved P” rather than merely from E. While such a case is possible, it is also possible to stipulate that the mathematician inferred P simply from E, and in this case I think we still intuit that the mathematician is doxastically justified in believing P.

3.3. the double counting objection Michael Tooley discusses a case in which a subject correctly infers P from E, thereby acquiring an inferentially justified belief that P. The inference also causes it to seem to the subject that P. Tooley (this volume, section 5.2.1) argues that if PC is true, the subject will then have two justifications for P: the normal inferential justification, plus a (presumably foundational) justification provided by its seeming to the subject that P. PC thus commits us in this case to ascribing too much justification to the belief that P.

340

Phenomenal Conservatism

On my view, an inferential appearance does not provide a subject with a second justification for P because the justification provided by the inferential appearance just is the subject’s inferential justification for P. However, some may doubt my account of inference. And regardless of whether one accepts my account of inference in general, one might imagine a case in which the process of going through an inference causes a subject to have an unconditional appearance that P, rather than merely the inferential appearance that P must be true in light of E. So I will now argue that even if the subject in Tooley’s case has two distinct justifications for P, we should not find this problematic, as the subject will not thereby have a higher degree of justification than that which ought to be provided by the inference from E to P. To see why, it is helpful first to consider a practical analogy. Suppose S has a certain practical end, E, and S knows that M would be an effective means of promoting E. S discovers that action A1 would promote E directly, whereas A2 would promote E indirectly, by promoting M.  A1 and A2 have the same overall probability of producing E, and there are no other relevant available actions or practical reasons. In this case, S has a reason for choosing A1, namely, that A1 promotes E. But S has two reasons for choosing A2, namely, (1) that A2 promotes M, and (2) that A2 promotes E. This must be so, because (1) is clearly a reason to perform A2, (2) is clearly a reason to perform A2, and (1) is clearly not identical to (2). Therefore, since S has only one reason to choose A1 but two reasons to choose A2 (each of which is of equal strength to the reason in favor of A1), it seems that S should definitely prefer A2. What went wrong in this thinking? Although it is technically true that S has two reasons for A2, the inference that S therefore has more reason overall for choosing A2 than for choosing A1 is unwarranted. It is unwarranted because these two reasons are not independent of each other. They are related to each other in such a way that one of them provides no additional support for the action when conjoined with the other; we might say that the one reason is fully redundant with the other for purposes of practical reasoning.15 It is along these lines, then, that we should respond to Tooley’s doublecounting objection if we agree to view the subject as having two justifications for P. If the subject is aware that his appearance that P derives from his belief that E, then from his own point of view, the two justifications are not independent of each other; both are justifications based on E.16 It is very plausible to

15 Chris Tucker (this volume, sec 3.1) makes a similar point using an example in which one witness’s testimony is based entirely on the testimony of a second witness. 16 Matters are otherwise if the subject has an unconditional appearance that P and is entirely unaware of any connection of this appearance with E. In that case, I would say that the appearance strengthens the subject’s justification for P.

Phenomenal Conservatism Über Alles

341

hold that awareness of the way in which they are related to each other prevents the one from providing added reason to believe P when conjoined with the other. Indeed, that is the precise intuition on which Tooley is relying to convince us that the subject in such cases does not have more justification for P than that conferred by a single one of these sources of justification. Just as, in the practical analogy, the two reasons supporting A2 are fully redundant with each other for purposes of practical reasoning, so too, the appearance-based justification for P is fully redundant with the inferential justification for P, for purposes of epistemic evaluation.

4. Self-Defeat and Naïve Inferentialism Peter Markie raises an interesting objection to the self-defeat argument for phenomenal conservatism.17 I argued, in essence, (1) All beliefs (that are plausible candidates for being justified) are based on appearances. (2) A belief is doxastically justified only if what it is based on is a source of propositional justification. Therefore, (C) For any beliefs to be doxastically justified, appearances must be the source of propositional justification.

From there (with some auxiliary assumptions), I  infer that PC is true (see Huemer 2007). Markie asks us to consider the case in which a subject is inferentially justified in believing P on the basis of E. In such a case, the subject’s propositional justification for P derives in part from the subject’s justification for believing E. But the subject’s having justification for E—that is, the normative state of affairs—cannot cause anything, so it cannot be part of the actual basis for the subject’s belief that P. So the belief that P is not based on (all of) what makes it propositionally justified. There are two ways of reading Markie’s example. First, one might take it straightforwardly as a proposed counterexample to my basing requirement (my premise 2): there are inferentially, doxastically justified beliefs; my basing requirement implies that such beliefs would have to be based partly on the justifiedness of the subject’s premises; but no belief is partly based on the justifiedness of one’s premises; hence, my basing requirement is false.

17 See his discussion of Naïve Inferentialism in his section titled “Unqualified Dogmatism: Failed Arguments” in this volume.

342

Phenomenal Conservatism

I think this objection misconstrues the source of propositional justification for the conclusion of an inference. When one correctly infers P from E, it is not the justifiedness of E (the normative state of affairs) that partly confers justification on P. It is the descriptive state of affairs on which E’s justification supervenes that partly confers justification on P. For example, suppose I am justified in believing that the cat is on the printer because I seem to see the cat on the printer. I then justifiedly infer that the cat is not outside. In this case, propositional justification for [the cat is not outside] is conferred by (i) my sensory experience of the cat on the printer, together with (ii) my inferential appearance that in light of the cat’s being on the printer, the cat cannot be outside. (i) and (ii) are both descriptive (nonevaluative) states of affairs, and each is part of the actual basis for my belief that the cat is not outside. Thus, there is no need for any normative state or property to cause my belief, and this case is not a counterexample to my basing requirement. A second way of reading Markie’s example is as a case in which, even though appearances are the (complete) source of propositional justification for P, something more is needed to explain why P is justified.18 In Markie’s view, as I understand it, there are three sorts of factors needed to explain why P is propositionally justified: (i) The source of justification, or that which confers justification, (ii) The absence of defeaters, and (iii) Background conditions, which do not themselves confer justification but merely enable sources to confer justification. Markie appears to view the justifiedness of E as an enabler (a condition of type (iii)) rather than part of what confers justification on P. In contrast, I would account for all epistemic justification using only principles of types (i) and (ii). In the inferential case, the source of justification includes whatever confers justification for E, together with the inferential appearance that in light of E, P must be the case. Given this conjunctive source, no background enabler is needed to explain the case. Markie proposes another background condition needed to enable appearances to confer justification for belief. He holds that a perceptual appearance that X is F will confer justification for [X is F] only if the subject possesses background evidence of the sort that would justify the proposition that things that appear the way X in fact appears are F. It is unclear why this condition, if genuinely required for justification, would count as a mere enabler, rather than a part of one’s source of justification.19 Be that as it may, if Markie’s requirement 18 This reading appears to be what Markie intended. However, the preceding reading strikes me as at least equally interesting. 19 Compare this case: you tell me that there is a calico cat in the parking lot. I say the cat must therefore be female. We would say that I needed a reason to think that calico cats are female. This would be part of my justification for the claim that the cat outside is female; it would not merely be a condition enabling [the cat is calico] by itself to justify [the cat is female].

Phenomenal Conservatism Über Alles

343

applies generally, then we will apparently be plagued with skeptical problems. If, that is, any perceptual appearance can justify a belief based on it only if one has some prior empirical information, then we shall find ourselves unable to acquire any empirical knowledge, for lack of a starting point. Perhaps Markie would draw a distinction between basic and derived appearances, applying his proposed enabling condition only to derived appearances. The motivation for such a double standard is unclear to me; if basic appearances may be presumed innocent until proven guilty, I do not see why derived appearances should be treated with any greater suspicion. But I shall discuss the epistemology of derived appearances further in the next section.

5. The Tainted Source Objection Peter Markie (2005; this volume) and Susanna Siegel (2013) have adduced putative counterexamples to PC in which a person’s appearance that P has unreliable or epistemically blameworthy causes—for example, the appearance that P might be caused by a desire that P, or by an unjustified belief that P. Some argue that the subject’s belief that P in such a case is unjustified, whereas PC implies that it is justified. This class of objection has been addressed elsewhere (Tucker 2010; Skene 2013; Huemer 2013), so here I will be brief. Suppose the appearance that P is caused by mental state M, where M is an unreliable source (such that appearances caused in this manner are not reliable guides to the truth). There are three cases to consider. Case 1 The subject is aware that the appearance is caused by M, and the subject doubts or has justification for doubting that M is a reliable source. In this case, the subject has a defeater for P, and advocates of PC can therefore accommodate the intuition that the subject lacks justification for believing P.

A critic might appeal to an intuition that the subject lacks even prima facie justification for P, rather than that the subject lacks justification tout court. I find this move suspect. I think that to have prima facie justification for P is, roughly, to have something that would justify P if one had no defeaters. So we are asked to have an intuition that, if a certain strange hypothetical scenario obtained,20 it would then be true that if the subject had no defeaters for P, the subject would have had justification for P. I cannot report any such intuition in myself and find it difficult to put much faith in any such intuitions. I suggest

20 Some of the scenarios are quite difficult to relate to, such as the walnut tree that just seems to have been planted on April 24, 1914, or the person who hallucinates a gun because she is afraid of guns.

344

Phenomenal Conservatism

that we instead rely simply on intuitions about whether a belief is justified in a given scenario. Case 2 The subject is aware that the appearance is caused by M, but the subject neither doubts nor has any justification for doubting that M is a reliable source. The difficulty of imagining this case may interfere with forming reliable intuitions about it. If the subject is aware that his appearance that P is caused by his wish that P, or by an unjustified belief, or by an appearance that Q where Q is in fact evidentially irrelevant to P, it would be a very strange world in which the subject would nevertheless have no reason to doubt that the appearance that P was a reliable indicator of the truth. I am not quite sure how to picture such a scenario, but to the extent that I have any intuitive reaction to it, it seems to me that the subject would have some justification for P. Thus, I do not think that this is a counterexample to PC. Case 3 The subject is unaware that the appearance is caused by M.  In this case, I  think we again have no counterexample, because the subject actually would have some justification for believing P. When the subject is unaware of an appearance’s etiology, that etiology is irrelevant to what it is rational for the subject to believe. A few examples may render this idea more plausible.

Suppose I have a completely realistic hallucination of a cat in the parking lot, induced by a hallucinogen that I unknowingly ingested two hours ago. If I have no reason to suspect that there is anything unusual about this experience, I would be justified in thinking there was a cat in the parking lot; any rational person in such circumstances would believe that there was a cat present. Nor does it matter if some other mental state enters the causal chain. Thus, suppose I  have an odd brain malfunction such that (again, entirely unbeknownst to me) having the sensation of a certain shade of violet causes me, five minutes later, to hallucinate a cat. I see the shade in question during my web surfing, then go outside and hallucinate a cat. Once again, I am justified in thinking a cat is present. We can even imagine a case in which I am clearly to blame for my unreliable cat-experience. Suppose that I have a brain-manipulation device lent to me by the local mad brain scientist (mad scientists being, as is well known, the chief source of innovative contraptions). If I push a button on the device, it will cause me to hallucinate a cat outside while simultaneously erasing my memory of my having pushed the button and my knowledge of the nature of the device. I have a desire to see a cat, due to my irrational belief that cats are messengers of God. This desire causes me to activate the device, leading to a cat hallucination, which makes me think a cat is present. In this case, my catexperience is caused by a desire and by an irrational belief, and I am to blame for the unreliable experience. Still, the experience provides justification for me

Phenomenal Conservatism Über Alles

345

to think there is a cat present. From my point of view at the time I have the hallucination, there is nothing about this experience that is different from ordinary perceptual experiences, which are generally agreed to be evidential (cf. Tucker 2010, 539–40). The lesson seems to be that the ability of an appearance to confer justification is independent of whether, unbeknownst to the subject, the appearance is caused by an unreliable process, an unjustified belief, or even an epistemically blameworthy state. Matthew McGrath argues that there is a special class of cases in which, unlike the cases I have just described, an appearance’s tainted source undermines its ability to provide justification (McGrath, this volume, section 4). These are cases of “free enrichment,” in which a subject has an appearance that E, E does not genuinely support P (even given the subject’s background knowledge and evidence), but the subject’s desire (or other unreliable state) causes the subject to base an appearance that P on the appearance that E. As McGrath notes, the original Markie and Siegel cases could be read in this way. I do not think McGrath’s notion of free enrichment alters the situation. We still have three cases: in case 1, the subject is aware that his appearance that P is based on the appearance that E, and the subject doubts or has justification for doubting that E supports P. In this case, the subject has a defeater for P, so PC can accommodate the intuition that P is unjustified. In case 2, the subject is aware that his appearance that P is based on the appearance that E, but the subject somehow neither doubts nor has justification to doubt that E supports P. In this case, I think the subject in fact has some justification for believing P, so this is not a counterexample to PC. In case 3, the subject is unaware that his appearance that P is based on the appearance that E. In this case, I think the subject again has some justification for believing P. In particular, it does not matter if S’s desire that P causes it to seem to S that P by causing it to seem to S that E supports P, rather than by directly causing it to (unconditionally) seem to S that P. In either case, as long as the subject has no reason to think that his desire that P played any such role in generating the appearance that P, the subject is rational in believing P (or at least according an elevated degree of belief to P).

6. PC Über Alles Michael Tooley has suggested that an early statement of phenomenal conservatism held the implication that “direct realism conquers all in epistemology.”21 My more recent statements leave open the possibility that PC enables us to

21 See Tooley, this volume, section 5.1.1, though Tooley’s remarks seem to be intended with less of a tone of triumph than the remarks of this paragraph.

346

Phenomenal Conservatism

surmount all epistemological problems. PC holds out the promise of a simple and unified account of the justification of our beliefs about the external world, about the past, about the future, about values, and so on. It is otherwise very difficult, perhaps impossible, to account for all these justified beliefs, let alone to do so in such a simple manner. It therefore seems to me that this is a powerful argument in favor of some form of phenomenal conservatism. It is of course impossible to review all interesting alternative theories of justification here, nor can I  offer a full treatment even of a few theories. Nevertheless, I would like to briefly gesture at where I find unsatisfying some of the approaches advanced by other authors in this volume.

6.1. evidentialism Earl Conee (this volume, section 7) offers the following “rival evidentialist sufficient condition for doxastic justification”: EJB

If S’s belief in X is based on propositionally justifying evidence S has supporting X, then, in the absence of defeaters, S thereby has at least some degree of justification for believing X.

I have no objection to this statement; I  would merely add the remark that propositionally justifying evidence supporting X always consists of appearances. Is there some alternative way of accounting for propositional justification, without adverting to appearances? To ultimately account for any belief ’s justification, I  think we need to identify the non-evaluative conditions that confer epistemic justification. Evidentialism does not do this, because “evidence” is an epistemically evaluative term. This is not to say that evidentialism is untrue; it is just that, without some account of what makes some particular fact, event, or state count as evidence for some conclusion, evidentialism does not truly (ultimately) explain why any belief is justified. Suppose we say that external world beliefs are typically propositionally justified because subjects possess evidence for those beliefs. What is this evidence? Presumably, it is the subjects’ sensory experiences. But what makes a sensory experience constitute evidence for a particular proposition? The phenomenal conservative has a natural answer:  a sensory experience counts as evidence for [P] when the experience is or contains a certain sort of mental representation (namely, an appearance) with [P] as its content. If we reject this answer, what else might we appeal to? Should we appeal to the experience’s non-cognitive qualia? But it is obscure why some non-representational mental quale should count as evidence for some proposition about external objects. Should we appeal to some non-propositional representation? Or a representation of some proposition other than [P]? Or a representation that has [P] as its content but that does not represent P as actually being the case? All of these

Phenomenal Conservatism Über Alles

347

options seem quite puzzling and unnatural. I doubt that we will have much more luck accounting for the evidential value of memories or intuitions, either.

6.2. acquaintance Thinkers such as Bertrand Russell (1912), Richard Fumerton (1985), and Michael Tooley (this volume) propose to found justified belief on a relation of direct acquaintance. Ironically, one of the main obstacles to such an account is an argument derived from a discussion by Richard Fumerton. Fumerton (1985, 79–80) deploys a version of the argument from hallucination against direct realism. The adaptation to acquaintance theories of non-inferential justification requires the assumption that “hallucinatory acquaintance” is possible— that is, that a subject may be in a state that seems to the subject just like being acquainted with something, yet the subject is not in fact acquainted with what he seems to be acquainted with. If such cases are metaphysically possible, then we may argue as follows: (1) In a case of hallucinatory acquaintance with P, the subject has the same sort of justification for believing that P as in a case of genuine acquaintance. (2) In a case of hallucinatory acquaintance, the subject’s justification for believing that P does not consist in his being directly acquainted with the fact that P. (3) Therefore, in a case of actual acquaintance, the subject’s justification for believing that P does not consist in his being directly acquainted with the fact that P. But is there such a thing as hallucinatory acquaintance? The most convincing examples involve a priori intuitions, states of understanding, or other putative bases for a priori justification. Russell, for example, allows that one may be acquainted with certain universals, and that this acquaintance is the basis for our a priori knowledge (Russell 1912, ch X). But there are cases in which a subject has a false a priori intuition, or a misunderstanding of some abstract object, that is indistinguishable by the subject at the time from a correct intuition, or a correct understanding. If a correct intuition or understanding would count as genuine acquaintance, then these mistaken intuitions or states of misunderstanding constitute hallucinatory acquaintances. Intuitively, the subjects of such hallucinatory acquaintance acquire defeasible justification parallel to that acquired in normal cases. An acquaintance theorist might wish to avoid these cases by rejecting intuition or understanding as potential forms of acquaintance. The problem then will be how one accounts for a priori knowledge. In addition, cases of hallucinatory acquaintance can be constructed for introspection, though they may be slightly less compelling (see Huemer 2007, 34–35; 2006, 152–54).

348

Phenomenal Conservatism

A second major difficulty facing the acquaintance theory is the self-defeat argument. Even if there is such a thing as acquaintance, it can be argued that our non-inferential beliefs are nevertheless based on appearance rather than acquaintance (see Huemer 2007, 44–47; 2011a, 2–8). Therefore, if we accept the basing requirement for doxastic justification, the acquaintance theory cannot afford us doxastic justification.

6.3. internalist reliabilism Matthias Steup’s (this volume) Internalist Reliabilism holds that a subject has justification for P when and only when the subject has an appearance that P, lacks defeaters, and has memory data that support the reliability of the appearance.22 This conflicts with PC, which entails that such memory data are not required. Internalist Reliabilism faces a serious skeptical threat. If appearances on their own fail to provide justification (even in the absence of defeaters) for believing anything, then I do not see how memory data will provide justification for propositions concerning the reliability of appearances. If memory data are themselves a species of appearance, then the same condition should apply to them as applies to other appearances. Thus, suppose S has an appearance, A. For A to provide justification for believing its content, S must have some set of memory data M1, such that M1 provides justification for the proposition that A is reliable. But for M1 to provide justification for anything, S will need a set of memory data M2 that provides justification for the proposition that M1 is reliable. But for M2 to do this, there will have to be some memory data M3 …. On the other hand, if we say that memory data are not appearances (perhaps because merely dispositional memories are not appearances), then Steup’s view as stated implies that memory data cannot provide justification for anything. We could modify the view to make it compatible with the existence of non-appearance sources of justification: we might simply say that an appearance confers justification if and only if it is undefeated and its reliability is attested by memory data—this statement is silent on whether anything other than appearances ever confers justification. One might then propose that memory data are another source of epistemic justification, and memory data do not require evidence supporting their reliability. But this view is not well motivated. If we agree with Steup that appearances unsupported by evidence of reliability are at best accidentally veridical, why wouldn’t we say the same about memories unsupported by evidence of reliability?

22 Because Steup omits any basing requirement in his statement of the view, I  assume that it is intended as an account of propositional, rather than doxastic, justification.

Phenomenal Conservatism Über Alles

349

At some point, we must reject the demand for evidence of the reliability of our cognition. And if we are to reject that demand at some point, we might as well reject it right at the beginning and allow appearances to justify beliefs by default.

6.4. concluding challenge For those who agree with such judgments as that sensory experiences provide justification for external-world beliefs, that intuitions provide justification for a priori beliefs, and that memories provide justification for beliefs about the past, there is an obvious challenge of accounting for the justificatory force of all these experiences. I have not here addressed externalist alternatives, and I will not do so now, as time is short and the motivations for internalism are already well known. I will content myself for now with the observation that within the internalist perspective, there are few options. Not many internal properties of sensory experiences, intuitions, and memories are plausible candidates for sources of justification. Appearances are the most (perhaps the only) simple and natural solution. This suggests that internalists should look to something in the neighborhood of PC.23 It is possible that I have overlooked something. Perhaps there is a way to make non-cognitive qualia, non-propositional representations, or non-assertive representations plausible as a source of justification. Or perhaps there is some other mental state or property that I have overlooked. I cannot now refute these possibilities. I shall simply leave this as a challenge to epistemologists unsympathetic to PC—not the only challenge they face, but among the more salient:  if undefeated appearances are not a source of justified belief, then how is one to avoid skepticism about the external world, the past, values, abstract objects, and so on? Unless this challenge can be met, we would be wise to place our trust in the appearances, and thus to continue to work with something like phenomenal conservatism.

Acknowledgments Besides the appreciation I have already expressed toward the other authors in this volume, I would like to single out Chris Tucker for special praise, in view of his thoughtful and invaluable comments on this chapter, to say nothing of his untiring work on the rest of the volume. 23 Since I also think that everyone should be an internalist, I think everyone should look to something in the neighborhood of PC (see Huemer 2006). I say “something in the neighborhood of ” PC because there is a family of epistemological views that explain justified beliefs in similar ways, by appealing mainly or solely to appearances. Thus, despite their criticisms of PC, Markie, McGrath, and Steup all advance views in the neighborhood of PC.

350

Phenomenal Conservatism

References Chisholm, Roderick. 1957. Perceiving:  A  Philosophical Study. Ithaca, NY:  Cornell University Press. DePoe, J. M. 2011. “Defeating the Self-Defeat Argument for Phenomenal Conservativism.” Philosophical Studies 152: 347–59. Descartes, René. 1984. “Meditations on First Philosophy.” In John Cottingham, Robert Stoothoff, and Dugald Murdoch, eds., The Philosophical Writings of Descartes, Vol. II, 1–61. New York: Cambridge University Press. Fumerton, Richard. 1995. Metaepistemology and Skepticism. Lanham, MD:  Rowman and Littlefield. _____. 1985. Metaphysical and Epistemological Problems of Perception. Lincoln:  University of Nebraska Press. Huemer, Michael. 2013. “Epistemological Asymmetries between Belief and Experience.” Philosophical Studies 162: 741–48. _____. 2011a. “Phenomenal Conservatism and Self-Defeat: A Reply to DePoe.” Philosophical Studies 156: 1–13. _____. 2011b. “The Puzzle of Metacoherence.” Philosophy and Phenomenological Research 82: 1–21. _____. 2009. “When Is Parsimony a Virtue?” Philosophical Quarterly 59: 216–36. _____. 2007. “Compassionate Phenomenal Conservatism.” Philosophy and Phenomenological Research 74: 30–55. _____. 2006. “Phenomenal Conservatism and the Internalist Intuition.” American Philosophical Quarterly 43: 147–58. _____. 2002. “Fumerton’s Principle of Inferential Justification.” Journal of Philosophical Research 27: 329–40. Hume, David. 1748/1975. An Enquiry Concerning Human Understanding in Enquiries Concerning Human Understanding and Concerning the Principles of Morals, edited by L. A. Selby-Bigge. Oxford: Clarendon Press. Jackson, Frank. 1977. Perception:  A  Representative Theory. Cambridge:  Cambridge University Press. Markie, Peter. 2005. “The Mystery of Direct Perceptual Justification.” Philosophical Studies 126: 347–73. Russell, Bertrand. 1912. The Problems of Philosophy. New York: Oxford University Press. Siegel, Susanna. 2013. “The Epistemic Impact of the Etiology of Experience.” Philosophical Studies 162: 697–722. Skene, Matthew. 2013. “Seemings and the Possibility of Epistemic Justification.” Philosophical Studies 163: 539–59. Tolhurst, William. 1998. “Seemings.” American Philosophical Quarterly 35: 293–302. Tucker, Chris. 2010. “Why Open-Minded People Should Endorse Dogmatism.” Philosophical Perspectives 24: 529–45.

{ name index } Achinstein, Peter 104, 128, Alston, William P. 100, 128, 148, 152, 158, 177, 183, 185, 201, Aristotle 263, 268 Audi, Robert 2, 9, 14, 24, 27, 181–83, 185–87, 195–96, 199, 201, 228, 246, Ayer, A. J. 282, 288 Bach, Kent 86–7 Barnett, David 95, 110, 116, 128 Battaly, Heather 299, 304 Bealer, George 5, 9, 27, 156, 177, 249, 261, 268, 295, 300, 304 Bentham, Jeremy 297 Bergmann, Michael 2, 7, 20, 23, 27, 29, 135–48, 151–52, 154–55, 158, 162–63, 168–69, 174, 177–78, 248–50, 268, 304–05, 329, 337 Boghossian, Paul 100, 114–15, 128–29 BonJour, Laurence 23, 27, 115, 129, 136, 139–42, 144–45, 147–48, 152, 168–69, 177, 216, 221, 269, 300, 305 Braisby, Nick 241, 246 Braun, David 45–49 Brewer, Bill 34, 49, 73, 87 Brogaard, Berit 2, 7–8, 11, 20, 26–27, 270, 276, 280–81, 284, 288, 329, 332 Broome, John 229–30, 238, 246 Brown, Jessica 2, 8, 21–22, 47, 50, 71 Buckareff, Andrei 49 Byrne, Alex 36, 49, 116, 129 Cappelen, Herman 85, 87 Casullo, Albert 29, 131, 152 Chalmers, David 41, 50, 278, 288 Chan, Sin yee 288 Chisholm, Roderick M. 2, 115, 129, 140, 152, 155, 177, 181–82, 201, 274, 276, 282, 288, 315–17, 327, 330–31, 350 Christensen, David 100, 103, 107–08, 111, 119, 125, 127–30, 298, 305 Chudnoff, Eli 2, 5, 7–8, 21, 27, 71, 73–75, 78, 87, 96, 129, 300, 304–05 Clark, Michael 82, 87 Cohen, Stewart 89, 116, 129, 136, 152 Coliva, Annalisa 128 Comesaña, Juan 23, 27, 117, 129

Conee, Earl 1, 6–7, 9, 21, 27, 47, 49–50, 52, 149, 152, 154, 156, 177, 248, 268, 334–36, 346 Crane, Tim 34–36, 50, Crisp, Thomas 154, 165, 177 Cruz, Joseph 2, 8, 28, 270, 289, Cullison, Andrew 2, 4–5, 8, 20–21, 27, 33, 50, 295, 305 Cuneo, Terence 278, 288 Dancy, Jonathan 230, 246 Daniels, Norman 75, 87 David, Marian 152 Dehaene, Stanislas 283, 288 Delk, John L. 239, 246 DePaul, Michael 2, 9, 24–25, 28, 75, 87, 202 DePoe, John M. 9, 28, 337, 350 deRosset, Louis 288 Descartes, René 215–16, 261, 268, 335, 350 Dougherty, Trent 47, 50, 152 Dretske, Fred 34, 50, 73, 87, 95, 129, 185 Earlenbaugh, Joshua 301, 305 Ebert, Philip 114, 128 Elga, Adam 100, 128–29 Engel, Mylan 152 Enoch, David 103, 129 Evans, Gareth 34, 50 Fales, Evan 165, 177 Feit, Neil 49 Feldman, Richard 24, 28, 47, 49–50, 76, 87, 100, 107, 129, 135, 149–50, 152, 205, 221, 233, 235, 246, 248, 268 Ferreira, Marinus 27, 128 Field, Hartry 90, 122, 127–29 Fillenbaum, Samuel 239, 246 Flanagan, Matthew 27 Frances, Bryan 215, 221 Frege, Gottlob 44–45, 47 Fumerton, Richard 23, 28, 47, 50, 58–59, 68, 100, 129, 148, 152, 174, 177, 221, 337, 347, 350 Garber, Daniel 127, 129 Gegenfurtner, Karl R. 246 Geirsson, Heimir 46, 50 Gellatly, Angus 241, 246

Name Index

352 Gettier, Edmund 78, 80–82, 88 Ginet, Carl 136, 152 Goldman, Alvin I. 15, 28, 72, 73, 85, 88, 149, 152, 233, 246 Goldstone, Robert L. 242, 246 Greco, John 152 Hacking, Ian 297, 305 Hagaman, Scott 201 Hansen, Thorsten 239, 246 Harman, Gilbert 278, 288, 298, 305 Hasan, Ali 9, 28, 146, 152 Hawthorne, James 122, 127, 130 Hawthorne, John 29, 68, 86, 88, 89, 115, 129–31 Heim, Irene 275, 288 Hendrickson, Andrew T. 242, 246 Hosein, Adam 288 Huemer, Michael 2, 4–7, 9–10, 12–16, 20, 23–28, 53–54, 57, 64, 68, 71, 73–75, 85, 88–89, 96, 100, 130, 140, 152, 154–57, 161, 172, 178, 188, 194, 201, 203–12, 214–15, 218, 221, 225–27, 249–53, 256, 267–68, 270–74, 276, 278–79, 281–82, 287–88, 293–96, 299–300, 302, 305–28, 333–34, 337–39, 341, 343, 347, 349–50 Hume, David 215–16, 295, 335–336, 350

Mackie, J. L. 85, 88 Malmgren, Anna-Sara 79, 88 Markie, Peter J. 2, 9, 11, 15–16, 25–28, 144, 152, 154, 157, 166, 178, 188, 201, 227–28, 233–36, 244–47, 248, 261, 264, 269, 272–73, 281, 289, 303–05, 322, 338, 341–43, 345, 349, 350 Martin, M. G. F. 79, 88 Matheson, Jonathan 4, 23, 29, 139, 152, 154, 156, 165, 168, 172–74, 178 McCain, Kevin 68, 288, 304 McDowell, John 34, 48–50, 131 McGrath, Matthew 2, 9–12, 15–16, 25, 27, 97, 130, 225, 227, 245, 247, 268, 338, 345, 349 Merricks, Trenton 177 Mittag, Dan 49 Molyneux, Bernard 301, 305 Moon, Andrew 268 Moore, G. E. 96, 108–10, 295, 314 Moyer, Mark 288 Nagel, Thomas 86, 88 Nagel, Jennifer 80, 88 Neta, Ram 96–97, 128, 130 Norcross, Alastair 288 Olkkonen, Maria 246 Orlandelli, Eugenio 128

Ichikawa, Jonathan 78–79, 88 Jackson, Alexander 10, 229, 246 Jackson, Frank 4, 315–17, 327, 329, 350 Jarvis, Benjamin 78–79, 88 Jeffrey, Richard 125, 130 Jehle, David 19–20, 28 Kant, Immanuel 117, 300 Kelly, Thomas 100, 102–03, 130 Kershnar, Stephen 49 King, Nathan L. 215, 221 Kolodny, Niko 230, 246 Kornblith, Hilary 299, 305 Kotzen, Matt 95–96, 101, 125, 128, 130 Kunda, Ziva 232, 246 Kung, Peter 19, 28, 104, 115, 130 Lange, Marc 122, 130 Larson, Richard 275, 288 Lasonen-Aarnio, Maria 110, 128 Lehrer, Keith 82–84, 88, 294–95, 298, 305 Leibniz, G. W. 261, 268 Lewis, David 278, 289, 319, 321 Loeb, Don 288 Ludwig, Kirk 74, 88 Lycan, William G. 2–3, 5, 8–9, 15, 26, 28, 293, 295, 297–99, 301–02, 305 Lyons, Jack 4, 7, 13, 15, 28, 242, 246

Pasnau, Robert 288 Paxson, Thomas, Jr. 83–84, 88 Peacocke, Christopher 34, 50, 73, 88 Piller, Christian 128 Plantinga, Alvin 7–8, 28, 76, 88, 106–07, 130, 205, 269, 300, 305 Plato 329 Pollock, John L. 2, 8, 28, 73, 88–89, 95, 228, 247, 270, 289 Poston, Ted 246 Pryor, James 2, 4–5, 8, 10, 16–20, 22, 27–28, 71–73, 75–78, 88–89, 97, 100, 103, 130, 140, 152, 158, 178, 225, 228, 247, 249, 269–70, 289 Putnam, Hilary 41, 50 Pynn, Geoff 103, 130 Rea, Michael 177 Reichenbach, Hans 92, 130 Reid, Thomas 34, 158, 178 Rogers, Jason 4, 23, 29, 139, 152, 154, 156, 165, 168, 172–74, 177 Roush, Sherri 128 Russell, Bertrand 296, 314, 347, 350 Russell, Bruce 152 Rysiew, Patrick 47, 50, 86, 88 Sacks, Oliver 312, 327 Salerno, Joe 152

Name Index Salmon, Nathan 45, 50 Salmon, Wesley 92, 130 Schechter, Joshua 100, 131 Schiffer, Stephen 29, 89, 128, 131 Senor, Thomas 240, 246–47 Sgaravati, Daniele 87 Siegel, Susanna 15, 29, 34–35, 37, 48, 50, 227–28, 233–37, 239, 243, 246–47, 258, 269, 283, 285–87, 289, 343, 345, 350 Silins, Nicholas 18, 94, 96–97, 101–102, 110, 112, 128, 131, 227, 247 Silva, Paul, Jr. 27 Sinnot-Armstrong, Walter 198, 201 Skene, Matthew 2, 5, 9, 16, 29, 343, 350 Sosa, Ernest 4, 7, 24, 29, 74–76, 87, 88, 136, 149, 152, 186, 199, 201, 237, 247–48, 250, 261, 265, 267, 269, 295, 300, 305 Speaks, Jeff 50 Spectre, Levi 128 Spencer, Joshua 49 Steup, Matthias 2, 15, 23–24, 27, 135, 145, 147, 153–54, 157–58, 172, 174–78, 188, 201, 264, 269, 302, 305, 336–37, 345, 348–49 Stich, Stephen 298, 305 Strawson, P. F. 27 Strohminger, Margot 87 Sturgeon, Scott 95, 128, 131 Swank, Casey 152 Swinburne, Richard 3–4, 29, 154, 156, 178 Thompson, Brad J. 43, 51 Tillman, Chris 49 Tolhurst, William 7, 29, 54, 68, 156–57, 159–60, 178, 225–26, 247, 270, 289, 335, 350

353 Tooley, Michael 6, 9, 13–14, 20, 27, 279, 288, 306, 328–30, 333–34, 338–41, 345, 347 Tucker, Chris 1, 5, 8, 29, 33, 49, 51, 53–54, 58–61, 63, 68, 87, 96, 100, 128, 131, 139, 144, 152–54, 177–78, 186, 188, 194, 201, 221, 226, 234–35, 240, 246–47, 249–50, 255–56, 259, 268, 269–70, 275–76, 279, 282, 285, 288–89, 304, 310–13, 322, 327, 333, 340, 343, 345, 349, 350 Tuggy, Dale 49 Turri, John 47, 51, 64, 68, 234, 242, 247 van Fraassen, Bas C. 115, 129 Van Leeuven, Neil 295, 305 Vishny, Robert 302 Vogel, Jonathan 77, 88 Wager, Adam 288 Wagner, Carl 122, 125, 127, 131 Walter, Sebastian 246 Warfield, Ted A. 205, 221 Weatherson, Brian 19–20, 28–29, 86, 88, 116, 131 Wedgwood, Ralph 18, 29, 96, 110, 116, 128, 131 Weiner, Matt 271, 278, 288 Weisberg, Jonathan 19–20, 29, 93, 119, 123–25, 128, 131 White, Roger 9, 17, 29, 89, 103, 106, 108–13, 116, 118, 124, 128, 131, 228, 247 Wierenga, Ed 49 Willenken, Tim 101–02, 112, 128, 131 Williamson, Timothy 3, 5, 29, 73, 78–79, 86, 88–89, 101, 118, 131, 198, 201, 301, 305 Wright, Crispin 29, 89, 113, 128, 131, 228, 247 Yablo, Stephen 118, 131

This page intentionally left blank

{ subject index } Access internalism (see ‘internalism’) Acquaintance 27, 174, 195, 307, 325–27, 347–48 Agent centeredness 17, 25, 204–15, 217–20 Agent neutrality 24–25, 203–11, 215–20 Appearances (see ‘perceptual experience’ and ‘seemings’)

Foundationalism 18, 25, 76, 188, 193–95, 200, 225–28, 237–42, 245, 248–49, 251, 262, 264–68, 270, 272, 287, 308, 324–27, 339 Frege’s puzzle 44–47 Fregeanism about perceptual content (see ‘perceptual experience’)

Bayesian epistemology 2–3, 16–20, 22, 25, 89–128 Jeffrey Conditionalization 19, 98, 106–07, 111, 119–28 Blindsight 184, 279–80, 312–14, 323, 333

Gettier problem 22, 72, 75, 78–87, 260, 279, 300

Cognitive penetration 12–16 (see also ‘objections to dogmatism and/or phenomenal conservatism’) Coherentism 8, 24, 26, 102, 204, 218, 293, 297 Conative conservatism 196–97 Credulism 22, 89, 96–103, 112, 128 Direct acquaintance (see ‘acquaintance’) Disagreement, epistemology of 2, 24–25, 85, 100, 102, 202–220, 321 Dogmatism motivations for (see ‘motivations for dogmatism and/or phenomenal conservatism’) objections to (see ‘objections to dogmatism/ phenomenal conservatism’) perceptual 2, 9, 249, 257–60, 264 Pryor’s definition of 73, 96–97 restricted versions of 25–26, 236–45, 249–51, 262–8, 272, 277–88 Doxastic conservatism (a.k.a. epistemic conservatism) 2, 5, 229, 242, 245–46 Epistemic conservatism (see ‘doxastic conservatism’) Evidentialism 47, 346–47 Expert recognition 8, 12–13, 48, 234–36, 242, 255, 257–60, 273, 280, 283–84, 303 Externalism 2, 100, 102, 122, 135–36, 139, 148, 151, 155, 167, 169, 174, 183–184, 211, 235, 244, 251, 296, 299, 349

Internalism access internalism (AI) 22–24 Bergmann’s Dilemma for 23–24, 135–41, 154–55, 163–72 Replies 141–52, 172–77, 336–37 Strong Internalism 135, 137–39, 147, 151, 163–65 Weak Internalism 135, 137, 139–41, 145–48, 151, 167–72, 278 Internalist Reliabilism 146–52, 174, 337, 348–49 Intuitions epistemic status of 21–22, 24, 71–87, 182, 197–201, 214, 248–49, 260–61, 267, 271–72, 281–82, 298–99, 300–01, 308, 347, 349 vs. seemings 3, 8, 21, 63, 71–74, 197–201, 279, 300–01, 309, 333 Justification a priori 2, 8, 18, 26, 71–87, 96–97, 101, 104–10, 113–17, 124–27, 228, 248–268, 270–88, 293–304, 347 inferential 8–9, 24, 75, 79, 188, 252–55, 338–41 memorial 2, 15, 26, 181–86, 189, 192–93, 196–97, 200, 267, 271–72, 276–79, 282, 287, 294, 299, 312, 333, 347–49 strong modern conception vs. weak contemporary conception 215–220 testimonial 2, 9, 14, 24, 63, 65, 67, 90–91, 95, 157, 172, 175, 183, 186–93, 210, 245, 260–61, 285, 295, 340 Knowledge how 25–26, 262–68 Memory 1–3, 7–9, 11–12, 15, 24, 26, 33, 53, 67, 74, 85, 136, 140–51, 156, 159, 174–77, 181–86, 189, 192–93, 196–97, 200, 202–03, 226, 228,

356 232, 243, 245, 249, 261, 267, 271–72, 276–79, 282, 287, 294, 299, 309, 312, 318, 329, 333, 344, 347–49 Mentalism 162, 249, 251, 258, 261–62, 264, 267–68 Metaphilosophy (philosophical method) 2, 22, 78–87, 297 Moral epistemology 2, 75, 85, 198–99, 206–10, 213, 226, 271–72, 279, 281–82 Motivations for dogmatism and/or phenomenal conservatism internalist intuitions 9, 226, 296 argument from Mother Nature 296–7 self-defeat argument for dogmatism about intuitions 9, 298–9 self-defeat argument for phenomenal conservatism 9, 75, 251–55, 299, 324–5, 341–3 survey of 8–12 Müller-Lyer illusion 271, 276 Objections to dogmatism and/or phenomenal conservatism Bayesian objections 16–18, 89–90, 108–113, 123–27 Replies 18–20, 113–18, 123–28 Tainted Source (a.k.a. Bad Basis) objection 14–15, 25–26, 144–46, 227–29, 232–46, 257–62 Replies 15–16, 301–04, 343–45 Dangerous Belief (a.k.a. permissiveness/ liberality) objection 319–22 Illegitimate Boost/Double-Counting objection 13–14, 190–91, 322–25 Replies 14, 339–41 Perception immediacy about 21–22, 63, 71, 78–85 Perceptual experience content of 1–3, 5, 20–21, 33–49, 158, 225, 248, 279, 309–10, 334, 277–88 Fregean vs. Russellian content 37–49 thick vs. thin (see also ‘qualia’) Phenomenal Conservatism modest vs. strong interpretations of 26–27, 154–55, 306–08, 321–22, 326 motivations for (see ‘motivations for dogmatism and/or phenomenal conservatism’) objections to (see ‘objections to dogmatism/ phenomenal conservatism’) relation to dogmatism 2, 225–26 restricted versions of 199–200, 322–3 unrestricted phenomenal conservatism 188–94, 200

Subject Index Phenomenal Semi-Conservatism (or semi-dogmatism) 11–12 Principle of Credulity 26, 97, 293–301 Problem of Known Illusions 4 Qualia 27, 40–41, 184, 306–07, 309, 311–14, 316–17, 322–27, 330, 333–34, 346, 349 Rational commitment 10–11 Risk 18–19 Russellianism about perceptual content (see ‘perceptual experience’) Seemings Kinds of intellectual (see also ‘intuitions’) 1, 3, 33, 118, 136, 156, 159, 184, 193, 198, 201, 248–49, 260–61, 267, 279, 308–09, 333, 339, 347 introspective 8, 26, 62, 108, 112, 136, 156, 159, 189, 193, 226, 249, 271–72, 276–79, 287, 309 memorial 1–3, 7–8, 11–12, 24, 26, 33, 53, 67, 136, 140–51, 156, 159, 174–77, 184, 186, 189, 193, 196–97, 200, 202–03, 214, 226, 228, 249, 267, 271–72, 276–79, 282, 287, 309, 312, 318, 329, 333, 348–49 moral 1, 8, 33, 75, 85, 156–60, 199, 206–10, 271–72, 279, 281–82, 300 non-epistemic 272–77, 287 perceptual 1–2, 7, 15, 21, 26, 33–49, 61–3, 73, 137, 143, 145, 149, 156–9, 196, 202, 225–8, 236–7, 249–50, 262, 266–7, 271–2, 275–8, 285, 302, 309, 311, 318, 328–9, 342–3 Content of (see also ‘perceptual experience, content of ’) conceptual vs. non-conceptual content 34–37, 48–49, 242 propositional vs. non-propositional content 5, 20, 34–37, 40, 58, 248–250, 279, 309–10, 334 Misc. vs appearances 314–8 as products 190–1 Chisholm on 2, 140, 155–56, 181, 274–76, 282, 315–17, 330–31 quasi-inferential 25, 228, 236–46 receptive 25, 228, 239–45 Ontology of belief view 3–4, 156–7 evidence-taking view 6, 21, 27, 55–6, 334–6 experience view 3–6, 159–60, 276, 329–334 inclination view 3–4, 156–7, 329 vs. sensations 48, 157–9, 279

Subject Index Phenomenal Character of conscious vs. unconscious 160–62, 172, 299, 329 qualia-free 307, 333–34 Seeming truth 52–60, 66–67 Semantics and content externalism 41, 47–48, Semi-conservatism (see ‘Phenomenal Semi-Conservatism’) Sensations (see ‘perceptual experience’ and ‘seemings, ontology of, vs. sensations’) Skepticism 8, 16–19, 21–22, 41, 66–67, 71, 76–78, 96, 104, 108–10, 136, 161, 199, 209–10, 216,

357 225–27, 271–73, 278, 281, 296–97, 335, 343, 348–49 Speckled hen problem 8, 186, 225, 255–56, 282–83 Subject’s Perspective Objection 23–4, 139, 141–44, 167–72, 336–7 Norman the clairvoyant 23, 136, 139, 168 Testimony (see ‘justification, testimonial’) Uncertainty 19–20 Warrant 16, 85, 169, 205, 228, 297, 322, 340 Ways-Millianism 44–48